yC-NRLF 00258 AERONAUTICAL RESEARCH COMMITTEE. Report on Materials of Construction USED IN Aircraft and Aircraft Engines BY Lt.-Col. C. F. JENKIN, C.B.E., R.A.F., M.Inst.C.E., Professor of Engineering Science in the University of Oxford, Late Director of the Materials Section of the Technical Department OF THE AIRCRAFT PRODUCTION DEPARTMENT OF THE MINISTRY OF MUNITIONS. LONDON : PRINTED AND PUBLISHED 1!Y HIS MAJESTY'S STATIONERY OFFICE. To be purchased through any Bookseller or directly from H.M. STATIONERY OFFICE at the following addresses: [MPEHTAL HOUSE, KINGSWAY, LONDON, "VV.C.2, and 28, ABINGDON STREET, LONDON, S.W.I: 37, PETER STREET, MANCHESTER; 1, ST. ANDREW'S CRESCENT, CARDIFF; 23, FORTH STREET, EDINBURGH; or from E. PONSONBY, LTD., 116, GRAFTON STREET, DUBLIN. 1920. Price 21s. Od. Net. CONTENTS. Chapter. Page. Preface ... ... ... ... ... ... ... ... ... ... ... ... ' ... 4 Mechanical Tests . _ I. 5 Local Concentration of Stress ... ... ... ... ... ... ... ... II. 12 Steel III. 16 Steel Tubes ... IV. 47 Stream-line Wires V. 63 Wire Ropes VI. 66 Aluminium Alloys VII. 67 Copper and Tin Alloys VIII. 88 Corrosion IX. 93 Timber X. 95 Glue XI. 131 Rubber Shock Absorber Cord XII. 141 Fabric ... ... XIII. 147 440158 (27364) Wt. 29165-944 2000 12/20 H. St. G.2 PREFACE. This report is written with the object of placing on record for the use of aeronautical engineers the results and conclusions derived from the research work which has been carried out during the war on materials of construction for aircraft. The paramount importance of weight in aircraft construction has led to the develop- ment of various new materials, and has made it worth while to investigate many problems in the theory of the strength of materials which had not previously been of importance. No attempt has been made to compile a complete record of all the investigations which have been made; the object aimed at has been to state the con- clusions arrived at and to describe the more important researches on which the conclusions are based. Most of the investigations were made with the object of solving urgent practical problems, and obtaining data on which design might be based, and the results are given as far as possible in a form in which they may be made immediate use of by engineers. Many interesting points of less urgency have arisen which have had to be left unsettled, and many of the investigations are not yet completed; suggestion for further research are made in most sections of the report. A continuous series of researches were carried out during the whole of the war on timber and the aluminium alloys, but the other subjects were dealt with as occasion arose, and as a consequence the report has a disconnected and scrappy form, which is accentuated by the differences of style between the different sections, many of which are verbatim reprints of the original con- fidential memoranda issued during the war. It has not been considered necessary to fill in the blanks with matter which is well known, or to re-write the whole as a connected text book on materials; the report is intended rather to serve as an addendum to the existing text books. Several sections of the report deal with breakages and failures of parts. Much has been learnt from the study of these, and many of the investigations were originated in order to find the explanation of obscure failures. The importance of tracing breakages to their true cause is much greater in aircraft than in other branches of engineering, where it is often sufficient to strengthen the part till it stands; this method of dealing with failures in aircraft would lead to an inadmissible increase in weight. Many investigations into breakages were also carried out at the N.P.L. and E.A.E., and by Dr. Hatfield and others, but it has not been possible to include their work in this report. The writer has been dealing with this subject all through the war, first in the Admiralty Air Department, then under the Air Board, and finally as Director of the Materials Section in the Department of Aircraft Pro- duction. A considerable part of the work has been carried out by the staff of that Sect inn; researches on timber and strength of materials by Major A. Robertson and his assistants, at first in the Manchester University Engineering Laboratory, and later at the R.A.E.; re- searches on steel by Captain L. Aitchison, of Sheffield University and his assistants at the Metallurgical Laboratory of the Imperial College of Science and Tech- nology lent to the Technical Department by the courtesy of Professor Carpenter; researches on steel tubes, aluminium alloys, and metal spars by Captain Lea (Professor of Engineering), and his assistants at the Engineering Laboratory of the University and at the Technical School, Birmingham; researches on glue by Lieutenant Kernot in the Botanical Laboratory at the Imperial College of Science and Technology. Fabric was not dealt with by the Materials Section till near the end of the war; the chapter on it is written by Professor A. J. Turner, who was in charge of the Fabrics Laboratory at the Royal Aircraft Establish- ment during the war. The chapter on it is written by- Professor A. J. Turner, who at the beginning of the war was an Assistant in the Fabrics Division at the National Physical Laboratory and later was in charge of the Fabrics Laboratory at the Royal Aircraft Establishment. Many investigations have been made for the Materials Section at the National Physical Laboratory by the staff of the Engineering Department under the direction of Dr. T. E. Stanton, C.B.E., F.R.S., and by the staff of the Metallurgy Department under the direction of Dr. W. Rosenhain, F.R.S., who have also carried out much of the work for the Light Alloys Sub-Committee, quoted m Chapter VII. Much work was also done for the Materials Section at the Royal Aircraft Establish- ment, in addition to that referred to above, chiefly by Mr. Dyson in the Metallurgy Department, and a great deal of analysis work has been done by the Government Laboratory. Full use has been made of many investi- gations carried out by the A.I.D. Special investigations occupying a long time have been made for the Section by Professor Lang, F.R.S., and Mr. Robinson in the Botanical Laboratory of the Manchester Uuivers.it \ , who have devoted a large part of their time to the work ; by Professor Dalby, F.R.S., at the Imperial College of Science and Technology; by Professor Edwards at the Manchester University; by Professor Shakespeare, at the Birmingham University; by Drs. Maigh and Br\an at the Royal Naval College, Greenwich; by Professor Coker, F.R.S., at University College; by Dr. Seoble at University College; bv Mr. H. Brearley, Mr. J. II Dickenson, Dr. W. H. Hatfield, Mr. J. W. Fawcett and Mr. D. Fletcher in Sheffield; by Mr. A. A. Remington, Sir Gerrald Munt/. Dr. Hudson and Dr. P.eiigough in Birmingham, by Mr, J. Brunton in Edinburgh. Great assistance lias been gfven lo Hie Section by many manufacturers, who have supplied materials and samples for investigation, and have placed Hie results of their own researches at the disposal of the Section. Chapter). F.o.l. ANNEALED M^n STEEL 'PIECE a STEEU BALL WITH PARALLEL FLATS AXIAL LOADING SHACKLES MACHINE JOU^T 8OU3ER TOGETHER HALVES FINISH HftU^ R?AHER AND OUTSTOE AT SCTT1NC. CHAPTER I. .MECHANICAL TESTS. 1. Tensile Tests. 2. Ult. Strength. 3 Elongation (English, French, American). 4. Reduction of Area. 5. Elastic Limit. 6. Yield Point. 7. Proof Stress. 8. Fatigue Limits. CONTENTS. 9. Notched Bar Test. 10. Bend Test. 11. Reverse Bend Test. 12. Torsion Test. 13. Brittleness. 14. Upton Lewis Test. 15. Abrasion Test. 16. Hardness Test. In no other branch of engineering is an accurate knowledge of the strength of materials so important as it is in aeronautics where the weight of every part has to be reduced to the minimum compatible with safety. Not only is it necessary to know the strengths of the materials of which every part of aeroplane and engine is constructed, but the calculation of the stresses in the different parts has to be made with the greatest accuracy, using every refinement of theory available. Mechanical tests have to be resorted to at every stage, first to determine the qualities of the materials, then to verify the strength of each part, and finally to check the strength of the complete machine. It will, there- fore, be convenient to deal first with the subject of mechanical tests, and then with some stress problems before considering in detail the different materials of construction. Tensile tests. To obtain accurate results several pre- cautions are necessary which have not hitherto received sufficient attention. () The load must be applied truly axially, i.e., the applied forces must act along the centre line of the test piece. Experience has shown that the ordinary spherical seatin< r s used on testing machines are quite insufficient to ensure that the load shall be axial. A satisfactory form of axial loading shackle due to Major A. Robertson" is shown in Fig. 1 ; if these are not used carefully, with turned test pieces, the load cannot deviate from the a-.\is of the test piece more than a very minute distance. If the load is eccentric when ductile materials are tested, the Elastic Limit appears to be lower than its true value, the Yield Point in such materials as wrought iron or mild steel is rounded off and lowered, but the Ultimate Strength is not much affected. If the load is eccentric when brittle materials are tested, the test pieces break considerably below their true ultimate strength. In testing cast aluminium alloys, for example, without axial loading shackles results are often obtained 3 UJ 1 lif J_ M S JJ UI ! i N i^ 1 91 UJ s "v* 4 } O J mi ? S s Q 's 2 UJ I X x UJ o cc N S V. ft 5 N^ o: C fl J i II j ! D Q U. 5 'O 15 20 25 SO 55 -40 PERCENTAGE ELONGATION ON ENGLISH SPECIMEN. ~^ \ 1 . \ _ u PIECE X \ o tf) TEST 5 1 \ CATIOh X in ' Total. Uniform. Test Test Piece. Piece. K. 1 68 20 7 18-7 13-3 ,, 61 22-5 8 21 15-2 46-1 28-4 13 26-8 20-6 ,, ... ... r>9-4 23 T) 8 21-9 15-6 M "* ... 79-8 16-7 6 15-59 11-2 )i .. 65-7 21-1 4-5 19-4 12-5 38-2 29 11 27-13 19-8 S. 6 37-9 31-1 11-1 29-02 20-9 S. 13 40 25 12 23 -65 18-1 S. 14 37-8 29 11 27-13 19-9 S. 15 66-2 19 8-5 17-91 13-7 S. Ifi 84-4 12-8 3-i 11-83 8-0 S. 17 63-2 17-0 3-5 15-6 10-0 "9% carbon ... 82-0 14-8 5 13-78 9-8 S. 27 46'8 28-3 11 26-5 19-5 Mild steel 28 38 22 36-2 29-8 S. 18 air hard 110-6 11-8 o 10-78 6-8 tempered 102-8 12-8 ' 3 11-78 7-8 n p 94 14 -r, 3-5 13-36 8-8 j 'i * 87 14 -r> 4 13-41 9-1 M ) ... 68-6 18-2 5 16-83 11-5 Elongations measured on strips cut from sheet metal are not very satisfactory. Soft metals, such as copper and soft brass, give large elongations even in thin sheets, hat thin sheet steel, even when annealed, gives hardly any elongation, though the material it is made from may give high figures when tested in the standard size. The explanation of this fact has not been found. The effect is too large to explain by the change in the reduction of area due to the difference in form. The facts were fully proved by a long series of tests made for the Air Board by Messrs. Thos. Firth and Son. Elongations are therefore not specified for sheet steels in the Air Board Specifications. Elongations on wires are difficult to measure and do not furnish satisfactory results. Reverse bend tests sire a fairly good substitute for elongations for sheets, wires and stream-line wires. Very small elongations, such as those given by most Cast Aluminium alloys, cannot be accurately measured. Reduction of Area. This is commonly specified, but little attention is paid to the results. In steels the reduction of area is markedly less when the samples are cut " across the grain " than when cut " along the grain. 1 ' By the direction of the grain is meant the direction in the metal along which the maximum elongation has occurred in forging; it is clearly shown by means of etching by copper chloride attack. The reduction of area may possibly be an important figure for copper alloys which differ widely in the way they stretch, some contracting to a narrow neck and some stretching uniformly and breaking with- out any neck. Elastic Limit. The term Elastic Limit has been defined in various ways. It is generally used in the sense defined by the B.E.S.A. to denote the stress at which the stress/strain diagram ceases to be a straight line: it cannot be determined by the ordinary commercial tests but reqiiires more or less delicate laboratory instruments. Comparatively large errors in determin- ing the elastic limit are introduced if the load is not exactly axial, as has already been stated. A very large number of laboratory tests have been made during the war on metals of all sorts, and the meaning of the elastic limit has been widely discussed. The results show that the elastic limit varies widely in the same material and cannot therefore be considered a reliable indication of its quality. The conclusion arrived at is that the elastic limit is very largely dependent on the state of internal stress of the material, and that in interpreting its meaning these stresses must be con- stantly borne in mind. This interpretation of the results explains the fact that the elastic limit is low in quenched steels, in cold worked metals and in castings, and that it is raised by tempering, by blueing, and by alternating stress. It is obvious that all steel sample's which have been quenched in oil or water (and to a less extent those that have been cooled in air) must be in a state of internal stress owing to the different rates at which the surface and interior have cooled. The stress/strain curves for such samples have hardly any straight portion; the elastic limit for 100-ton steel may be as low as 13 tons. (See Fig. 6.) Tempering, par- ticularly if done at a fairly high temperature, relieves the internal stresses to a great extent and tempered samples give stress/strain curves of a very different type. (See Fig. 7.) The effect of tempering in raising the elastic limit is also shown in Fig. ?A, which gives the results of tempering three samples of air-hardening steel at various temperatures. Similarly castings are left in a state of internal stress ; this probably accounts for the very low elastic limits of the aluminium alloys. (See Chapter VII. > Cold worked metal is also obviously left in a state of internal stress. This stress may be relieved by blueing (that is, heating to a temperature usually between 300 and 500 0.), which consequently raises the elastic limit to a marked extent. The effect of blueing on cold-drawn steel tubes is shown in Figs. 8 and 9 Chapter IV. The effect on steam-line wires i* shown in Figs 1 to 4 Chapter V. The effect of blueing on brass rods which have been severely cold drawn is remarkable. In the cold-drawn con- dition the internal stresses are often sufficient to cause the bar to crack spontaneously after a time (" season cracking "). After blueing the material is in a stable condition. Whether raising the elastic limit by blueing is advantageous or not depends on how the material is going to be used. It is useful for drawn -steel tubes which are used in compression, but is considered useless for steam-line wires which are used in tension. The internal stresses due to quenching, and probably those due to cold working, may be relieved by sub jecting the material to alternating stresses for a short time. (See Fig. 9.) It is always important to consider how far a test piece accurately represents the material to be tested. Thus if the test piece is separately heat treated the internal stresses may differ from those in the part it represents and its elastic limit will be different. Again, it is probable that the surface stresses produced by the turning of the test piece may be sufficient to produce a perceptible deviation of the stress/strain curve and thus to give a false elastic limit. The elastic limit may be raised by very slightly stretching the material, as is w^ll known. This fact may he used in various ways. For example, the elastic limit, of a cast aluminium alloy cylinder may be raised when the steel liner is shrunk in. The elastic limit lias often been used as the point on which to base strength calculations. The above facts, which show how uncertain its meaning is, are sufficient to show that calculations should only be based on it after fully considering its meaning in each particular case. Yield Point. When wrought iron or mild steel are tested in tension a stress is reached slightly above the elastic limit when the test piece suddenly stretches a considerable amount without increase of load. When accurate methods of testing are employed it is found that in these materials the stress actually falls 20 to 80 per cent, during the stretching.* The stress at which this sudden stretching commences is called the Yield Point. Most other materials do not exhibit this pheno- menon, and the stress /strain diagram, instead of having a sudden horizontal kink (or actual drop), bends away gradually at an increasing rate. In some cases there is a fairly sharply marked corner in the diagram which then rises again nearly in a straight line; in most cases the curve is more or less parabolic. (Cf. Fig. 6.) There is, properly speaking, no yield point for such materials, but. unfortunately, the term has been gener- ally used to denote the entirely vague point correspond- ing with the corner of the stress/strain curve. Where this corner is located depends on the fancy of the tester. The B.E.S.A. have published two definitions of the yield point, as follows : (i) " The Yield Point is the point where the extension of the bar increases without increase of load." (ii) " Practical Definition of Yield Point. The Yield Point is the load per square inch at which a distinctly visible increase occurs in the distance between gauge points on the test piece, observed by using dividers ; or at which when the load is increased at a mode- rately fast rate there is a distinct drop of the testing-machine lever, or in hydraulic machines, of the gauge finger." The first of these definitions is accurate, and accord- nig to it most metals have no yield point. The second definition, which is called the " practical definition," is thoroughly unsatisfactory, as indeed any definition is bound to be which attempts to define a property for materials which do not possess it, but unfortunately this second definition is still in common use. Steel when stressed to about 67 tons per square inch will give an elastic expansion of V J- - inch in a 2-inch gauge length ; such an extension is easily observed with dividers, and consequently, according to the second definition, the yield point can never be above this stress. When carefully tested materials which have no true yield point never show any " drop of the beam." In practice the tester considers the yield point to be reached when he notes a considerable change in the relative rates of turning the handles which apply the load and take up the extension respectively; the record therefore depends on the idiosvncraey of the tester. The sooner the Yield Point is eliminated from all specifications of materials which have none the better. * Vide Proe. R. S. A. Vol. 88. 1913. It has already been removed from Specifications R IP and S 2H, and from the drafts of the future s|iecilira t i, nis I'm- I lie case-hardening steels; it is also bein:', replaced in the copper-alloy specification by the Prooi Stress. Proof Stress. The Proof Stress has been introduced into some of the copper alloy specifications as a. sulisti tute for the Yield Point. It is a stress which is to he applied for 15 seconds, and must not produce a perma- nent set after the stress is removed. In applying the load the beam of the testing machine is to be kept floating. No change in the distance between the gauge points of less than | per cent, is to be considered as proof of permanent set. A very interesting use of the proof load occurs in the Specification T 14 for tempered carbon steel axle tubes. A small but limited permanent set is allowed under the proof load. The relation between the proof stress and the stre strain diagram of the material is shown in Fig. 8. A straight line is drawn parallel to the straight part of the diagram at a distance from it equal to an elongation of \ per cent. This line will cut the stress/strain starve at some point P. The specification requires that tin- point P shall be above the stated proof stress, or, in other words, that at the stated proof stress the stress,' strain curve shall not yet have crossed the sloping line. The specification does not require the stress at P to be determined, the test can therefore be made quickly, and it is perfectly definite in its meaning. Sufficient experience has hardly been gained yet to settle whether a Proof Stress will prove a satisfactory substitute for the Yield Point. Fatigue Range. It can no longer be doubted that the life of all parts which are subject to alternating or fluctuating stresses (such as most, engine parts and those aeroplane parts which are subject to vibration) depends simply on the fatigue range of the material they are made of. Owing to the lengthy and laborious nature of the usual tests required to determine fatigue ranges, the subject has not yet been fully investigated; it is one of the most important subjects requiring investigation at the present time. When the load on a part is not steady 1 but varies cyclically, there are limits which the stress must not exceed or fracture will occur after a smaller or greater number of alternations. The Fatigue Range is the range of stress within which an indefinitely large number of alternations will not cause fracture. Some parts may be stressed equally in compression and ten- sion others between no stress and a tensile stress, or in any number of different ways; there is therefore a whole series of fatigue ranges. If the inferior limit of stress is given, then the range determines the upper limit. The fatigue range for torsional (shear) stresses is as important as that for tension and compression. For the present it will be conveni3tit to confine, our attention to the fatigue range for equal tensile and compressivo stresses or equal + and torques ; with this limitation the, fatigue range is double the fatigue Until. Various methods have been used to determine the Fatigue Range. The commonest is the Woehler test. Eden. Rose and Cunningham* have used a modi- fication of the Woehler test, in which the test piece is subject to a uniform bending moment * Min. Proc. I.M.E. Oct. 1911. Chapfcer. I Fie. 8. 6000 COPPER ALLOY. 945. 4O SO 6O I OtVI&ON - 33OO FiG.9. REDUCTION OF HYSTERESIS RAISING OF L. LmtT BY ALTERNATING STRESS More recently methods have been devised by Kapp, Hopkinson* and Haighf using alternating current electromagnets to apply direct tension and com- pression instead of bending. In all these methods the alternations are continued till the sample breaks or till some determined large number of alternations (usually millions) has been applied. Stromeyerf has devised a very mucli quicker test depending on the detection of the first, rise of temperature of the sample, which takes place as soon as it ceases to be perfectly elastic. For equal plus and minus stresses the fatigue limit generally coincides with the stress at which hysteresis commences and heat begins to be developed, but this is not true for fluctuating stresses in which the plus and minus stresses are unequal: Bairstow has shown that under such fluctuating stresses there is at first an elongation which gradually ceases as the metal reaches a fresh stable condition; during tin's elongation heat is developed. Even under equal plus and minus stresses some materials at first show considerable hysteresis which afterwards disappears; this occurs in samples which have considerable internal stresses, such for example as those made of hardened (untempered) steel. Stromeyer's method, therefore, though it promises to be very useful, must not be relied on without discrimination. The following tests illustrate this point. A bar of steel was air-hardened and its fatigue range measured in Haigh's machine and found to be + 25 tons per square inch. A fresh sample was then stressed statically to + 25 tons per sq. in. and the hysteresis loop determined see the first curve in Fig. 0. Tliis sample was then subjected to 1,000 reversals of stress of + 25 tons/sq. in. and then retested statically. The hysteresis loop is shown in the second curve in Fig. 9, which will be seen to be much smaller than the first. This procedure was re- peated and after a little over a million reversals of stress the hysteresis loop had practically vanished, as is shown in the figure. The elastic limit rises as the hysteresis loop diminishes. This test was made on a sample of steel which did not harden intensively; the results would probnblv have been more marked in a 100-ton steel. There are still many matters in connection with these tests to clear up. It is not yet certain whether there is any finite fatigue range which will insure an infinite life; there is some evidence to show that the limiting stress may be zero for an infinite number of repetitions This is not a matter of essential importance if we can determine the range for a number of alternations cor- responding with the desired life of the part. For example, a crankshaft of an engine running at 1,000 r.p.m. in 100 hours will have made 6 million revolutions. It is customary at present to quote fatigue ranges for f> million alternations. They are often quoted for 1 million alternations, but this is certainly too few. Torsional Fatigue Eanges have been determined by Stsnton and Batsonl| and Stromeyer (loc. ci'i.j, but only for a few materials. They are specially impor- tant for springs and shafts. * Proc. R.S.A. Vol. 86. p. 131. t Journal Inst. of Metal?. 1917. J Manchester Steam Users Association. Memorandum by the Chief Engineer. 1913, and Sheffield Soc. Enjr. Proc. 1914, Part II. The Elastic Limits of Iron and Steel under cyclical variations of stress. Phil. Trans. Roy. Soc. A. Vol. 210, p. 35. II Brit. Assoc. Section G. Newcastle, 191fi. Stress Distribution in Engineering Materials. Interim Report. Appendix III 27264 The Fatigue Ranges of brasses and bronzes require full investigation. Some of the alloys having the highest ultimate strengths have very poor fatigue ranges, which is probably the explanation of some of the failures which have occurred with high tensile bronzes. The fatiguo ranges of the Cast Aluminium Alloys are considerably better than might be anticipated from their very low elastic limits. Fatigue Eanges will have to be deter- mined at high temperatures for many parts of engines. Katigue ranges for timber also require investigation. Fatigue fractures exhibit certain well-known features. There is no elongation* or reduction of area and the surface of the fracture shows a series of roughly con- centric ridges like contour lines on a map round the point where the fracture originated. These ridges are often likened to the marks on shells or to ripples or wave fronts. When seeking for a cause for such a fracture, it is necessary to remember that the fatigue range need only have been exceeded over a very small part of the section at the point where the fracture started, the mean stress may have been much less; the origin may, therefore, have been a sharp corner or a very small flaw, crack or scratch (see Chapter II.). When the fracture is due to a sharp corner it is fre- quently found that failure after failure will occur in engines of the same type in approximately the same number of running hours. Some data on the Fatigue Eanges of the following materials are included in different sections of this report : Steels (of many types and in different conditions) Chapter III. ' Stream-line wires (unsatisfactory), Chapter V. Wire ropes (bending round pulleys), Chapter VI Brasses, Chapter VIII. Copper pipes (petrol and oil), Chapter VIII. Aluminium (a few cast allovs only), Chapter VII. Eubber, Chapter XII. Fatigue under steady load. Timber and probably other organic substances fail at loads considerably under their ultimate strengths if kept loaded for long periods. As has already been stated, the results of ten- sile tests depend considerably on the rate of loading. The creep, which is observed after the elastic limit is passed, apparently goes on at a diminishing rate for very long periods and may cause final failure after many months under loads 20 or 30 per cent, less than the ultimate load borne in the ordinary tests. Gradual changes occur in metals with the lapse of time, possibly analogous to gradual tempering at ordinary temperatures. These changes have to be borne in mind when testing parts which have failed. A material which is found to be below the specification may have passed the tests when first put into use. The parts most liable to these slow changes in aeroplanes are probably those made of cold-worked metals, such as wire ropes and Stream-line wires. Brinell Test. This test is chiefly used as a method of estimating the ultimate strengths of parts or of com- paring the condition (due to heat treatment) of parts with that of test pieces; for which purposes it is very useful. It has also been used to measure the hardness of steels, aluminium alloys and copper alloys at high * This statement requires some qualification. Vide Bairstow's results for unequal fluctuating stresses (Joe. cit.). 10 temperatures by means of Prof. Edwards' dynamic modi- fication of the test. Notched Bar Test (Izod) The meaning and value of the notched bar test have been widely discussed during the war, and a great deal of research work has been done on it.* The principal outcome of the researches has been to show : (1) The result of a test depends largely on the form of the notch, the sharper the notch the more sensitive the test. (2) The result does not depend much on the type of the machine. (3) The result is nearly independent of the velocity of the hammer. (4) The result does not indicate the resistance of the material to " impact " or " shock" It should not be called an impact test. (5) The result is mainly valuable as an indication whether the heat treatment has been carried out. (6) The results indicate differences of condition which cannot be demonstrated in any other way so far as is at present known. (See Krupp Krankheit, page 22.) (7) It is generally believed that the condition of a steel giving a high notched bar test is better than that of the same steel giving a low test; but no conclusive evidence of this fact has been obtained yet. Experience with broken engine parts tends to confirm it. The notched' bar test is usually the most convenient one for showing the effect of the " direction of the grain " in a forging. This subject is fullv discussed in the section on macro-structure, Chapter III, and appears to be of greater importance than has hitherto been recognised. The following information (published in C.I.M. 738) explains why the notched bar test has been included in most Air Board Steel Specifications: All the principal Air Board Specifications for steel now contain a requirement that the steel shall give at least a certain specified Izod value. It is considered desirable to indicate briefly the reasons for the inclusion of this test in the specifications and also its usefulness to the builder of aero engines. (1) The notched bar test is put into specifications principally because it is the test which gives most in- formation as to whether the heat treatment which has been given to the steel is satisfactory or not. If the steel under consideration has been heat treated to the best advantage then the Izod value will be relatively high. If, on the other hand, the heat treatment given has been inferior, then the Tzod value will be relatively low. (2) After the best heat treatment the actual Izod values obtained from different steels will depend verv largely upon the nature of the steel. The values wbich will be obtained from three steels (e.v., a plain carbon steel, a nickel chrome steel, and a high speed steel), all of which have been heat treated in the best way to give the same tensile strength will be nuite different. The values, however, in all cases would be good for the particular class of steel represented. It is therefore impossible to compare two steels upon the basis of the absolute Izod value obtained from them. If such com- parisons are made, then the resulting conclusions would be entirely wrong in many cases. This shows that two steels cannot be compared by their absolute Izod values in the same way that they can by their tensile strength. It shows also that the actual Izod value obtained from any steel must not be judged by itself, but must be evaluated after taking into account the type of steel upon which the result was obtained. (3) The actual Izod value which is obtained from any particular steel after proper heat treatment depends upon the ultimate tensile strength of the sample. Various samples of the same steel may be heat treated to give varying ultimate strengths, and the satisfactory Izod value which should be obtained from the steel will be different for each ultimate strength. This is dealt with in the specifications by the provision of sliding scales making the Izod value vary with the ultimate tensile strengths (e.g., Specifications K.I or 3.S.17). The proper Izod value for different steels of the same type wbich are specified to give different ultimate strengths will be different. The steels in Specifications 2.S.18, K.3, S.83 and 2.S.12 are all of the same type (nickel chromium steels), but the required ultimate strengths are quite different ranging from 100 to 55 tons per square inch. The satisfactory Izod values are consequently different, and range from 10 foot Ib. with 100 tons to 40 foot Ib. with an ultimate strength of 55 tons per square inch". These examples show that the Izod value obtained from a steel must never be judged apart from the ultimate tensile strength with which it is associated. (4) The use of the Izod test as an indication of the heat treatment has been mentioned above. If the heat treatment is good, then the Izod value will be high, and if the treatment is bad, then it will be low. Con- sequently, it may be assumed that all steels which have given an Izod value, above the specification re- quirements have been properly heat treated. It is therefore not correct to say that a steel which gives fin Izod value of 60 foot Ib. is any better from the point of view of usefulness or suitability in an engine than one which gives a value of 50. if the Specification value be 40 foot Ib. Both the 60 and the 50 foot Ib. indicate a proper heat treatment. If the steel has not been heat treated well and properly, then the Izod value will be quite low. Izod values are really one of two things, either good or bad. If they are equal to or above the specified value, then they are good, whereas if they are far below the specified value, then they are bad. * \\ilt. Capt. Phil)>ot'8 paper on "Some experiments on Notched Bars." Proc. Inst. Automobile Enjrs. Vol. 12. (5) It is to be borne in mind when considering the Tzod value for engine parts, that the value obtained in a machined forging, may not be exactly the same as the value obtained upon the test sample upon which the part was inspected. The difference is most pro- bably due to the inevitable difference in the mass of the two parts, which affects the amount of work done upon the steel and also the heat treatment. Conse quently, a sample ctit out of a part which has failed may give an Tzod value a little below the specification value, whereas the test sample upon which this part wns passed, being a smaller mass, gave a higher value. Tlio specification values are framed with a knowledge of this probable difference and it is unsafe to assume 11 that because the Izod value on a part is a little below the value given in the specification, that the material of the part is therefore in an unsatisfactory condition. If the material were really unsatisfactory, the Izod value would be quite low and would not be anywhere near to the specification value. (6) Whilst the above statements show how the Izod value given by a steel is to be regarded as a criterion of the heat treatment which a steel has received, it is to be borne in mind that the Izod value also shows whether or not a steel is really in the most satisfactory condition to be used in an engine part. The study of a large number of failures of engine parts goes to show that steel with a bad Izod value fails more often than steel with a good Izod value. The existence of a good Izod value is no safeguard against bad design, but it certainly is of use in an engine part. As explained above, the criterion for the suitable Izod value is not an absolute value. The way of judging the suitability is rattier to find out whether the Izod value is high or low, i.e., whether the heat treatment has been properly carried out or not. If the value is low, then the parl is not likely to serve so well as if it is high, but .the " high " and " low " value must be judged as indicated in paragraphs 2, 3 and 4 above. A large number of notched bar tests have been made on aluminium alloys, but no interpretation of the mean- ing of the results has been suggested. A notched bar test has recently been introduced for measuring the brittleiiess of timber. Bend Tests Bend tests are used in addition to or as substitutes for elongation measurements in some specifications. They are specially suitable for sheets and wires. Bend tests are not easy to make; it is very difficult to insure that the radius of the bend shall be exactly the radius specified. It is important that there should be no sharp edges on the test pieces. Strips of sheet metal must have their edges smoothed off and rounded; sheared edges are damaged by the shears and are left in a ragged condition which will initiate cracks which spread across the bend. Keverse bend tests. These tests have been used for a long time for wires. They have been introduced during the war for stream-line wires and swaged rods (Speci- fication W.3 and W.8). They have the advantage over simple bend tests that they give numerical results which can be compared, instead of being merely a test which the sample passes or fails to pass. It is much easier to insure that the radius of the bend is correct in a reverse bend test than in an ordinary bend test, because the angle of bend is only 90 instead of 180. Torsion tests. These tests have been used for a long time for testing wires. They have been found useful in detecting brittle wires among the very hard drawn strands of wire rope. They are also used as the stan- dard method of measuring shear stengths. Brittleness test for wire. A wrapping test is being introduced into some of. the wire specifications as a test for brittleness. The wire is to be wrapped 8 times round its own diameter and then unwound again (except the last turn) and must show no fracture. The Upton-Lewis test This test somewhat resem- bles the reverse bend test, but the angle the sample is bent through is very small and the number of bends required to produce fracture is large. The results pro- bably depend mainly on the ductility of the material'. The sample is given a permanent set at each bend, the results therefore are not comparable with true fatigue tests. The meaning if the results is doubtful and the test has not been adopted in the Air Board Specifica- tions. The same remarks apply to the Arnold test. Abrasion test A satisfactory method of measuring the wearing quality of metals when subject to abrasion would be very useful ; the problem is complex and- has not so far been satisfactorily solved. The rate of wear probably depends on the nature of both the metals in contact, and certainly depends on the lubrication. It is probably affected by the speed of motion and by the manner in which the pressure between the surfaces fluctuates. Trials were made with an abrasion machine of the type designed by Stanton* with the intention of comparing different types of crankshaft steel, but the results were disappointing as far as the tests were carried. Tests are required suitable for comparing bearing brasses, bronzes and white metals. Hardness tests. A satisfactory method of measuring the hardness of ease-hardened parts is badly wanted. The Shore scleroscope is capable of giving valuable results but is not generally applicable; in skilled it hands it gives good results but in unskilled hands is liable to serious errors. It cannot be applied to curved surfaces, small parts, &c., the results appear to depend on the mass and method of clamping of the part being tested, and on the foundations on which the instrument rests. The rough and ready method of " scratching with a new file " is still the only test available. Timber tests. Owing to the structure of the material which is anisotropic, the testing of timber is difficult and complex; the subject is dealt with in Chapter X. Glue tests are dealt with in Chapter XI. * Report of the Hardness Test Research Committee. Inst. Mech. Engrs. Proo. 1U18, page 677. 27264 B 2 12 CHAPTER II. LOCAL CONCENTRATION OF STRESS AT CHANGES OF SECTION, SHARt> CORNERS, SCREW THREADS, CASTELLATIONS, KEY WAYS, SCRATCHES, &c. It is well known that the theory of the distribution of stress in an elastic body indicates that the distribu- tion of stress will not be uniform at any cross section where the form or size is changing, and that where the change is rapid the concentration of stress will be large. For example, if there is a shoulder in the middle of a straight round rod (see Fig 1) so that the two ends are different sizes, the stress when the rod is put under tension will not be uniformly distributed across the section in the neighbourhood of the shoulder, but there will be a concentration of stress at A. The extent to which the stress will increase in the surface layers just under the shoulder will depend on the form of the shoulder; if the shoulder has the form shown in the figure, it will depend on the radius of the fillet at A. A convenient measure of the increase of stress is the ratio of the maximum stress at A to the mean stress at B. If the corner at A could be made perfectly square (the radius of the fillet being nil) the stress at A would theoretically be infinitely greater than at B. The theory is only applicable to materials while they remain elastic, in this example therefore the theory means that even the smallest stress in B will cause the surface metal at the sharp corner A to begin to stretch. A very minute yield at A will be sufficient to relieve the stress in the immediate neighbourhood of the corner, and the stress will then redistribute itself, the maxi- mum at A never much exceeding the elastic limit of the metal. Similar concentrations of stress occur at all changes of shape or section. When the stress is tensile or compressive the only irregularities of section which matter are those shown in a longitudinal section, but when the stress is torsional, irregularities in both longi- tudinal and cross sections produce concentrations of stress. For example, if a shaft have a screw thread turned on it and also a keyway cut in it, then under tensile stress, which may of course be due to bending moments, there will be increased stresses at the bottom of the threads, but not at the bottom of the keyway (except at the ends if it does not run the whole length of the shaft) ; on the other hand under torsional stress there will be increased stresses both at the corners of the keyway and under the thread (and also at the ends of the keyway). In estimating the importance of such concentrations of stress as have just been described, it is necessary to draw a sharp distinction between parts which are loaded with a constant stress and those which are loaded with an pliernating or fluctuating stress. Under a constant load all metals appear to be sufficiently ductile to give the minute yield required to permit the stress to redistribute itself in a safe way. Tests have been made to check this point on the brittlest cast iron obtainable; the test pieces were standard test pieces with a sharp groove turned round the middle. The breaking stresses (reckoned on the area at the bottom of the groove) were the same as those of uugrooved test pieces, but the material was not sufficiently uniform to enable very accurate comparisons to be made. Test pieces of brittle materials which have too rapid ;i change of section at the shoulders usually break at the shoulder instead of in the middle owing to the local concentra- tion of stress, but the breaking loads are not appreciably different when they break in the middle; the effect is only sufficient to determine where the specimen will break. Marked effects are only produced when the dis- tribution of stress is very unequal; for example, immediately under the head of the test piece; if the corner at the point B is left sharp in test pieces of 100-ton steel, they frequently break there, though the cross section may be 25 per cent, bigger than that in the middle of the test piece. When the load is alternating the results are entirely different. If any portion of the metal, however small, where the concentration of stress occurs is stressed beyond its fatigue limit it will fracture after a time under the alternating stress and thus cause a minute crack. The concentration of stress at the bottom of this crack will be, if anything, greater than before, and the process will repeat itself, and in this way the crack will spread and ultimately cause a fatigue fracture of the whole. Thus the yielding of the overstressed part instead of saving the part will inevitably initiate a fracture. When the load is fluctuating, that is alternating be- tween unequal tension and compression loads or between two different tension loads, the results are more com plex. Bairstow* has shown that a condition of equili- brium may be re-established tufter a certain amount of yield, if the range of stress is not beyond the fatigue range corresponding to the particular range of fluc- tuation, it is therefore possible that the yield in such cases may be sufficient to reduce the stress at the point of concentration within the safe fatigue limits. It seems not unlikely that this is the explanation why sharp corners and scratches do not always produce such serious effects as might be expected from the calculated values of the concentrated stresses. This subject re- quires full investigation before any definite conclusions can be arrived at, but in the meantime it is necessary for safety to avoid all conditions which will allow the concentration of stress to exceed the ordinary fatigue range for alternating stresses. * Phil. Trans. A. Vol. 210, pp. 35-55, To farf page \ 3. ] [CHAPTER II. '- .'.":.- - .- - ?'"f,f-' ' :,- r -i-"'/ S^A'iJ/l FIG. 3. Fatigue crack, starting in fillet of crankshaft. X 5. FIG. 4. Branching crack. X 470. 13 Numerous examples have occurred during the war of fatigue fractures due to concentrated stresses ; it is hardly possible to exaggerate the importance of this source of failure. As examples of the positions where failures have been frequent the following may be given : Crankshaft fillets; maneton shoulders; key ways; rotary cylinder shoulders ; gear-wheel teeth ; screw threads; fcoolmarks inside connecting rods; toolmarks. grooves and sharp corners in springs; toolmarks inside gudgeon pins; file mark* on copper pipes (where brazed). The first step in dealing with the subject is to find methods of estimating the magnitude of the concen- trated stresses under the various conditions which occur in engine design. In a few simple instances the in- creases of stress due to change of section can be calcu- lated, for example the stress at the sides of a round hole in a broad plate under tension has been shown to be three times the mean stress in the plate; this ratio has been confirmed by Professor Coker* by means of his elegant optical stress metliod. This result is & special case of the wider theorem that the stress at the sides of an elliptical hole is greater than the mean stress in the ratio 1 + 2 - :1 P \vhere c( is the semi-axis of the ellipse perpendicular to the direction of stress, and p is the radius of curvature of the ellipse at the sides, i.e., at the end of a in Fig. IA. The above ratio is also almost exactly true for a semi-elliptical notch in the edge of a wide plate and also for a semi-elliptical groove in a round shaft. Radius a FIG. IA. The corresponding ratio for the increase of torsional stress for an elliptical groove in a round shaft is : 1 + A' V p Ratio of increase of torsional stress at the root of scratches of various shapes. The two curves representing these forinulse for various values of - arc shown in Fig. 2. It is noteworthy P that the ratio is much larger for tensile than for tor- sional stresses. Scratches. These results may be applied at once to the estimation of the effect of elliptical scratches and toolmarks on round shafts. In order to extend the theory to scratches of slightly different shapes an in- vestigation was made by Mr. Griffith t by his elegant soap-bubble method. The results are shown in the following table and in Fig. 2. The scratch or groove in the shaft was assumed to be V-shaped with a rounded bottom. The results are given for various ratios of (where a is the depth and p the radius of curvature of the root of the groove, as before) and for various angles of V. * Min. Proc. I.M.E. Feb. 10, 1913, p. 93. t Ad. Comtee. Report. B. 29. genl. 20. Raiio " : P i t 1 3 5 9 Angle of V. 2-01 2-66 3-J3 t-54 60 1-84 2-00 2-54 3-06 3-99 90 1-81 1-95 2-40 2-64 3-12 120 1-66 1-75 1-95 2-00 2-13 These results apply to torsional stresses only, but Mr. Griffith has shown that the corresponding figures for tension may be obtained by increasing the torsional figures in the ratio 1 + jj /,!.+ / ~ V p V p just as iii the case of elliptical grooves. He has also proved that the direction of the groove is immaterial in torsion, the concentration of stress will be the same whether the groove is round the shaft, along the shaft, or at any intermediate angle. The direction of the groove, however, is very material in tension. The results so far given, apply to grooves perpendicular to the direction of stress, i.e., round the shaft; if the groove makes an angle with the circumference of the shaft (-- e ) with the length of the shaft), then in tension the stress is in- creased iii the ratio: so that there is no increase when = 9t)- The proof of this is given in Appendix I. The theory does not suggest any limit to the small ness of the groove its effect depends simply on its shape. To estimate the importance of the theory it is necessary to measure the shape of such scratches as occur in practice ; this is being done by making gelatine casts of surfaces prepared in various ways filed, polished with emery, ground, &c. and micro photo- graphing sections of the gelatine casts. To complete this investigation it will be necessary to make fatigue tests on samples which have scratches of accurately determined shapes and to compare the results with the theory. There are various records in existence of the effect of scratches on samples used for fatigue tests*; the reductions in strength have not usually exceeded 20 or 30 per cent., but as no record exists of the shape of the scratches these results cannot be used to check the theory. Experience during the war shows that frac- tures in a large proportion of cases start at sharp corners or machine scratches but such examples give no indication of the extent to which the part has been weakened, nor have records been kept of the shape of the scratches or of the radii of the corners. The value of for an actual crack is very large, so that if fatigue p at a scratch is capable of starting a crack, fracture occurs very rapidly. The extreme sharpness of the bottom of a crack is shown in Fig. 3, which is magnified * Cf. Eden, Rose and Cunningham's paper. M. in Proc. I.M.E., 20 Oct., 1911. 14 about five diameters, aud in Fig. 4, which shows one end of a branching crack under a magnification of 470 diameters. Crankshaft fillets. No accurate calculation is possi- ble on the concentration of stress in a crankshaft at the juncture between the webs and the shaft or crank- pins. Nor is the optical method yet able to deal with models in three dimensions, but it appeared probable that a fairly accurate estimate of the stresses could be obtained by that method, using a flat model represent ing a median section of the shaft. At the request of the Materials Section, Professor Coker made such a model and measured the stress round the fillet between the crankpin and the crank. Professor Coker's report is given in Appendix II. ; the result of the measurement was that the stresses round the fillet were 60 per cent, higher than the calculated stresses due to the bending moment at that point. This result is of considerable importance, as it shows how large an effect change of form may have, even when a fairly large fillet connects the parts. It is not often possible to find fatigue cracks in a crankshaft in process of formation, but they have been found in a few cases. Photo Fig. 3 shows a fatigue crack starting from one of the fillets in a crankshaft which broke at a neighbouring fillet. Keyways and Oastellations The concentration of torsional stress at the bottom corners of keyways and under castellations has been determined by the soap- bubble method; (Ad. Committee Eeport B2g General 26. T1076) the results show the importance of rounding off the corner with a sufficiently large fillet. In the propeller shaft of the design investigated it was found that if the radius of the fillet was ,' th the depth of the keyway the stress in it was 2^ times the stress in a shaft having the radius equal to the radius under the keyway. If the radius of the fillet was Ith of the depth of the keyway, the increase of stress was reduced from 2j to 1J. As a result of this investigation, instructions were issued requiring the radius of the fillet at the bot- tom of the keyways in engine construction to be at least Ith of the depth of the keyway. If the stress at the fillet is compared with the maximum stress in the shaft of the full size, instead of with the stress in a shaft of the size under the keyway, the increase is much more striking. For the shaft with the fillet Jth the depth of the keyway investigated, the ratio was 3'37 instead of 1. As the strength of the shaft for alternating stresses is inversely proportional to the maximum stress at any point in it, it is legitimate to say that the keyway, even when it has the required fillet, reduces the strength of the shaft to 0'297 times its strength without a keyway. These figures assume that the shaft is subject to an alternating load; if the load is fluc- tuating between different torques of the same sign, the strength will probably not be reduced so much. The concentration of stress at the bottom of castella- tions was investigated in the same way. The results showed that with serrations of common proportions the stress was more than double that in a shaft of the same radius as the bottom of the serrations, or, in other words, adding plus serrations to a shaft reduces its strenght to one-half a remarkable result. These figures are subject to the same assumption as to the alternations mentioned above. APPENDIX I. STBESSES IN SCRATCHES. Consider the surface of a piece of material subjected to a tension THbs./sq.in. in a direction parallel to OY (Fig. 4a). SCRATCH = FIG. 4A. The tension in a direction making an angle with OY is T cos' 9, and, at an angle with OX, Tain* 9. The shear stress in the direction 9 is T sin 8 cos 9. Now put a scratch on the surface in the direction 9, as shown in the sketch. Assume that the tension T cos 2 9 is thereby increased to ......... (A) and the shear T sin 9 cos 9 to T(l + re) sin 9 cos 9 ... (B), the tension T sin' 9 parallel to the scratch remaining unaltered. Then the tension parallel to Y becomes T {(1 + m) cos 1 9 + sin* + 2(1 + ) sin' 9 cos 2 ft} = T{\ + cos 2 (m cos' + 2rtsin 2 0} ... (1) and the tension parallel to OX T(l + m) sin- 9 cos- 9 + Tain- 9 cos- 9 - 2T ( 1 + n) siu j 9 cw = T(m - 2n) sin- 9 cos- 9 ... (2) while the shear parallel to OX or OY becomes T(\ + ) sin cos 9 (cos 2 - sin 2 0) - T{(\ + m) cos 2 - sin 2 (>} sin 9 cos 9 = Tsiu 9 cos 9 {(-m) cos' 9 - n sin 2 9} ... (B). In a similar way the stresses due to making the scratch on a body subjected to a compression T parallel to OX are given by Tension parallel to OX = - 2'{l + 8in 2 0(TO8in>6 + 2cos 2 0)} (4), OY= - T(_m - 2ra) sin* cos' ... (5), Shear parallel to OX (or OY) = - Tain 9 cos 9 {( - TO) sin' 9 - n cos' 0} (6) . Now combine these two stress systems. The resultant stresses are Tension parallel to OF= r(l+mcos'0) ... (7), 0jr = -r(i + 8in') ... (8), Shear parallel to OX (or OY) = T(2n - m) (cos 2 - sin' 0) sin cos (9). But the resultant of the stresses T and T on the unscratched surface is a shear Tin directions at 45 with the axes. Herce the stresses given by (7), (8) and (9) are those due to placing a scratch on such a surface, at an angle of 45 9 with the shear. Chapfcer-H. F.e-2. INCREASE OF STRESS DUE TO SCRATCHES a/p O. * DEPTH OF SCRATCH. p - RADIUS AT BOTTOM OF SCRATCH. P - ANGLE. OF V. Chapfcer.il. I U sr i'SI6S/a.iOOO.C* B.1T" 27 12/18. 15 The shear stress in the 45 direction, due to the stresses (7), (8) and (9), is Tain 45 cos 45 (2 + m) + Tain cos 9 (2n - HI) (cos 2 9 - sin 2 9) (cos 5 45 - sin 2 45) which is independent of 9. Hence, by assumption (B), we must have Hence the shear stress (9) = 0. Tensions parallel and perpendicular to 45 direction are, by (7) and (8), i 7'i (cos 2 9 - sin 2 9) ......... (12). Summary of Remits. I. Maximum tension parallel to OY, due to placing a scratch at an angle on a surface subjected to a tension T parallel to Y, is T (1 + m cos 2 0) from (1) and (11), Tension parallel to OX = (2) and (11), Shear in directions OX and OY = %mT sin cos from (3) and (11). II. Maximum shear in directions OX and O Y, due to placing a scratch at an angle with OX, on a surface subjected to a shear 7 in the directions OX and ft Y = *(' + ?) {do)}- Tensions in directions OA'and OF are = i7>sin20 {(12)} (since = 45 - 0). APPENDIX II. THE STRESS I\ T THE FILLETS OF A CRANK SHAFT. BY PROFESSOR E. G. COKER. The occurrence of fractures in certain types of crank shafts has recently caused much difficulty, and at the suggestion of Lieut-Commander Jenkin, I have examined in a preliminary fashion the stress produced in a model of a crank shaft of the 4B Type when subjected to an overhanging load. Before com- mencing this work a few experiments were made upon the effect of curvature in bodies of crank shaped form, and some of the results obtained formed the subject matter of an earlier report. The present note relates to observations on a model of the above crank shaft shaped from a sheet of xylonite having a mean thickness of 0-2 ins. and of the form shown in the accompanying Fig. 5. It will be observed that the dimensions of the model correspond to those of the actual shaft to a scale of 1 inch = 100 units of the drawing supplied, but in one or two instances there are slight errors in the dimensions of the model, which, however, proved to be of no importance as regards the problem in hand. The drawing of the crank shaft shows a cylindrical hole throughout its length. while the crank pins are also hollow, and in addition there, are oilways in various positions so that the shaping of an exact model would be somewhat difficult and is probably unnecessary, since for the case taken the extreme stresses are found to occur at the junctures of the crank pin and main bearing with the crank web. At this section there is a cylindrical hole running through the centre of the web, but as will be shown this probably has little effect on the stress distribution at this place. A more important difference is that the model is a flat sheet of material with appropriate contours and therefore does not truly represent the stress condition of the crank shaft as actually made. It seems possible that the effect of the fillets at the junctions of the cylindrical parts with the web may cause more intense stresses than those found below. At the commencement of the experiments the model was shaped s > that the fillets at A and B, Fig. 1, were of greater radius than those shown on the drawing. The radius at A was at first 0-05 ins. instead of 0-04 ins. while the radius at B and at three corresponding places was 0-2 ins. thereby producing a sharp change of direction as shown in the accompanying sketch. After examination under load the fillets at A and B were reduced to their priper curvature of 0'03 ins. and (1-04 ins. respectively. The method of clamping the specimen and loading are indicated in Fig. 5. An examination in circularly polarised light showed at once that with tha loading chosen there is an extreme concentration of stress around the points A and B and nowhere else. The stress distribution is determined by the use of a standard calibration member under tension. At any place of the contour, where there is no local application of load the principal stresses must be along and perpendicular to the contour, and moreover, the latter must be zero, so that the value of the stress can be determined when the calibration value of the tension specimen is known. This is subject to a slight error when determining the magnitude of compression stresses owing to the local increase in thickness, but this latter effect is small, and is negligible in a rapid survey of the stresses where extreme accuracy is not attempted. In all cases the stresses found are plotted perpendicularly to the contour in Ibs., the reading of the spring balance employed, and subsequently the stress scale was found to be 736 inches ;= 1000 Ibs per sq. ins. The stress distribution at the fillet A for a radius of - 05 ins. ia shown on the accompanying Fig. 6 for loads of 1 Ib. and 2 Ibs. respectively : the dotted lines indicate that the values obtained are somewhat doubtful. This is owing to the small scale of the model, it should have been much larger to get accurate values of the stress, in fact the radius|here of 1/20 inches proved too small for the purpose of experiment. It is, however, probable that the results approximate to the true values and a test was applied to ascertain this. The horizontal section through the point C the junction of the fillet with the vertical part of the web is evidently under nearly pure bending and direct stress, and for a load of 2 Ibs. the bending moment is 3-6 inch Ibs. and the cross section is 0-225 x 0-198 inches with an I = -000188. The bending moment stress is therefore 3-6 000188 while the direct stress is 225 = 2120 Ibs. per sq. in. 045 = 45 Ibs. per sq. in. so that the actual stress pi-p s = 2075 Ibs. per square inch, while the measured value is found to be about 2230 Ibs. per square inch. Having regard to the possible error in the assumption made this may be regarded as a fairly satisfactory result. The maximum stress is about 3600 Ibs. per square inch, but as indicated by the dotted curve this is somewhat problematical. At the other fillet, Fig. 7, there is a very sharp change of direction and the local increase of stress at this point is very pronounced and could not be determined with accuracy at the apex of the angle. On the completion of this set of observations the fillets were reduced to the dimensions shown on the drawing, viz., -03 ins. at A and -04 ins. at B, and the stresses determined as before. As might be expected the reduction of the fillet at A causes a corresponding increase of stress ; thus at the point corresponding to C a stress was found of 2530 Ibs. per sq. ins. with a maximum value of about 4000 Ibs. per square inch, Fig. 8, while at the opposite fillet, Fig. 9, the maximum stress is about 3700 Ibs. per sq. in., but again the difficulty of locating the exact position of maximum stress was such that these figures can only be regarded as first approximations. I am satisfied, however, that they can be determined with considerable accuracy on a larger scale model. 16 CHAPTER III. STEEL. (This Chapter is nearly all based on reports drawn up by Captain Leslie Aitchison, R.A.F., of the Materials Section.) CONTENTS. Introduction. 1. Breakages of Steel Engine Parts. 2. Fatigue Ranges. 3. Hair Cracks. 4. Krupp Krankheit. 5. Macrostructure. 6. Forging Gear Blanks. 7. Casebardening. 8. Welding. 9. Valves. 10. Springs. 11. Sheet Steel. 12. Cold Rolled or Drawn Steel. 13. Non-Magnetic Steel. H. Cast Iron. Appendix I. Casehardening Spec. 2 H. 5. Research work on steel has been carried on very actively during the war, and many Committees have been formed to investigate various problems connected with the standardisation, specification, supply, manu- facture, and defects in steel. No attempt is made to deal with all these activities, but this chapter is confined to problems which have arisen in aircraft manufacture and mainly to the investigations carried out for the Materials Section, but it includes particulars of some important researches placed at the disposal of the Sec tion by Mr. H. Brearley and others. Great assistance has been given by steel makers all over the country and particularly by the Aircraft Steel-makers Associa- tion in Sheffield, and the Association of Drop Forgers and Stampers, and the committees from which this association developed. The conclusions arrived at on the various subjects dealt with are partly based on advice and assistance received from many sources and on experience and investigations generously placed at the disposal of the Section by many firms. Several of the subjects dealt with are still matters of controversy, and require further investigation before they can be finally settled. The problems which have arisen are of different sorts : (i) The causes of failure of steel parts which have broken in service or during tests see the section on Breakages. (ii) Questions concerning the effective strength of the metal; see the sections on Fatigue Limits, Case-hardening Steels, and Welding. (iii) Problems in manufacture and difficulties in meeting the specifications see the sections on Hair Cracks, Krupp Krankheit, Macro- structure, Forging Gear Blanks. Cold- Rolled Steels, and Sheet Steels. (iv) Difficulties with specific parts see the sec- tions on Valves, Valve Springs, and Cast Iron. fv) Problems on the efficiency of non-magnetic steels see the section on those steels. The testing of steel and the effects of local concen- trations of stress have already been dealt with in earlier chapters. Steel tubes, stream-line wires, and wire ropes are dealt with in later chapters. 1. BREAKAGES OF STEEL ENGINE PARTS. When an engine part breaks the engine is often exten- sively wrecked, and the first question is to determine which part failed first. Assuming that this has been settled the cause of the primary failure has to be found. It is convenient to distinguish between the following causes of failure: (i) Local flaws in the material, (ii) Material not up to the specification, either throughout or locally. (iii) Improper methods of forging resulting in weakness in certain directions. (iv) Local concentration of stress due to sharp corners or faulty machining, resulting in the stress at some point being above what the material will stand. (v) Overloading of the material, i.e , stressing it beyond what it will stand. This may be due to: (a) A defect of design, so that the dimensions are too small or the strength specified is too low. (b) Overloading the engine. (e) Stresses due to synchronous vibra- tions. (d) Stresses due to want of rigidity of the crank case. The primary failure almost always results in a " Fatigue Fracture," and it is generally possible to deter- mine the exact point at which the fatigue fracture commenced. These facts are very useful, first CHAPTER III. [To face page 16. FIG. !. Spiral fracture in crankshaft commencing at the bottom corner of the keyway. FIG. IA. The shaft shown in Fig. 1, cut open. Flu. 2. Fracture in crankshaft eommencing'in oil-hole. FIG. 3. Fractured propeller shaft. To face page 17.] [CHAPTER III. FIG. 5. Gear wheel with teeth crushed back. FIG. 6. Gear wheel tooth cracked under the case. FIG. 7. Gear wheel teeth cracked at sharp corners. FIG. 8. Sun-wheel carrier fractured " along the grain.' 17 enabling the primary failure to be distinguished from consequent fractures which take place suddenly when the break-down occurs; and secondly, in making it easy to settle whether the failure was due to a local fault, to local concentration of stress, or to simple over- stressing. A failure in sound material always commences at the weakest spcrt for example, at a shoulder or keyway and iihis fact makes it almost impossible to distinguish with certainty between failures due to concentration of stress and failures due to overloading. Eemoving the cause of local weakness will certainly strengthen the part, but will not stop failure if there is serious over- loading, due, for instance, to synchronous vibrations. Failures often occur repeatedly in the same position in engines of one type. When this occurs it is safe to assume that they are not due to local defects though they may be due to unsuitable methods of forging but are due to defects of design. Repeated failures in one type of engine at different points, for example first in the propeller shaft then in the gear then in the crankshaft, point clearly to over- loading, and probably to synchronous vibrations or to want of rigidity in the engine frame. Crank Shafts. The number of failures in crank shafts is not high, but when they happen they generally occur in batches, that is to say, that if in any engine a crank- shaft breaks, then many more shafts will break in similar engines in the same manner. In consequence of the very complete inspection which is made upon the crankshafts, it is very rarely that a shaft is put into use which does not come up to the specification as regards tensile strength, though a large number of shafts have undoubtedly come into use which have a low Izod value. Many broken shafts have been tested and found to be made of steel with a low Izod value, and it is probable that this has had something to do with failure, though the reason is not understood. There is no evidence to show that the type of fracture in shafts which have a low Izod value differs from that in crankshafts which have a high value, the growing crack type of failure occurs in both alike. A large number of crankshafts in one type of engine failed in consequence of a sharp corner at the junction of a journal and the web, and this trouble was definitely overcome by increasing that radius. Other failures have commenced in keyways which have not a sufficient radius at the bottom (see Figs. 1 and la) ; others at the oil holes in the journals or in the webs, in which screw threads have acted as sharp corners (see Fig. 2). In most eases where this matter has received careful attention and the radius in these positions made suffi- ciently large, the trouble has been overcome. These fractures (and similar ones in connecting rods) are frequently spiral in form, as shown in Figs. 1 and la, which represent the same shaft before and after it was cut open. Fig. la shows that the spiral crack began along the bottom corner of the keywav. A full explana- tion of these spiral fatigue cracks has not yet been found, i.e., the reason why they grow spirally. Occasionally the failure of a crankshaft can be definitely traced to overstressing due to synchronous vibration. A connection has been found occasionally between the method of forcing, as revealed by the macrostruc- tnre of a crankshaft and its failure ?72fi4 Propeller Shafts. Propeller shafts have not failed in many engines, and where they have it has generally been possible to account for the failure quite definitely. By far the greatest number of failures have occurred in one type of engine, and these have been traced to over- stressing due to synchronous vibrations. In the majority of cases the actual break has started at the keyway in the shaft. (See Fig. 3.) Connecting Rods. Connecting rods have failed from several causes, the different types of rods failing in different ways. The connecting rods in aero engines vary considerably in design, the large master rod in the rotary engines being quite different from the ordinary rods of a stationary engine, and also different from the auxiliary rods in the rotary or stationary engines. The majority of the rotary master rods have one end which is at least 8 inches in diameter, and this is joined to a tubular stem connected to a small end of ordinary size. They are always made as drop forgings, and many diffi- culties have arisen in their heat treatment in conse- quence of the very sharp change of section which occurs where the tubular rod joins the massive big- end. Many failures have occurred owing to cracks arising at this point during heat treatment. To over- come this difficulty it is necessary that a very large radius should be used at this point in the drop forging, so that the junction of the big-end with the stem may be as gradual as possible. It is desirable also to remove as much metal as possible from the big-end before carrying out the heat treatment. Other failures are due to distortion in hardening; when distorted rods are straightened cold they usually give trouble at a later stage. Finally, failures have occurred owing to the big-end being below the specified strength ; unless a suitable steel is used the big-end is liable to be oversoftened though the tubular rod may have the specified strength. A second frequent cause of failure is rough machining inside the hollow stem; it is by no means uncommon to find deep tool marks and scratches in the inside of the stem in rods which have broken. An isolated but interesting failure occurred in a rotary engine; the rod, after running for some time, split quite evenly along its axis, the two sides of the biff-end falling away. When this rod was examined it was found that the rod really consisted of two pieces of steel joined together by a layer of slag. In the other kinds of connecting rod which are much more uniform in section over their length, troubles due to unequal heat treatment do not often occur. Several failures have occurred at a keyway near the small end of one design of rod, which has caused them to break off quite short. Gear Wheels. The gear wheels in aero engines are made of various types of steel. Speaking generally, the only gears which have given trouble are those which are subjected to high duty, and which are consequently made of fairly powerful steel. These gears are usually made in one of three classes of material. (1) Case-hardening steel, usually containing about 5 per cent, of nickel. (2) 100 ton air-hardening steel. (3) 60 ton oil-hardened and tempered. made in the latter material have given no C 18 trouble except, in one or two instances, in which the wear has been rather excessive. This has generally been quickly overcome, and appeared to be accidental and to be due to lubrication troubles. With the 100 ton steel most of the failures were due to wear. It appears to be the fact that air hardening 100 ton steel is liable to wear excessively during the early stages of running. Many gears have been examined which have worn appreciably on the pitch line after running for a very short time, and it was expected that they would be quite useless, and would have to be removed from the engine. Some of them, however, were run for a further period to see what would happen, and it was found in practically every case that no further wear took place. Apparently the steel wears to a small extent very rapidly at first, and then settles down into a condition in which it will resist wearing for a much greater length of time, and in fact reaches a stable state. It is no uncommon thing for a gear to run well over 100 hours after having been apparently seriously pitted during the first two or three hours running. Trouble has been experienced occasionally with steel which has been hardened to give 100 tons per square inch tensile strength, but was unsuitable for the purpose, and only gave this strength after fairly drastic heat treatment. A large number of gears in one engine were made of steel which only gave 100 tons after quenching out in oil, with no tempering, and it was found that the ma- jority of these gears failed in consequence of the uneven- ness of their machanical properties, due to the variation in mass in different parts of the gear. There was no serious wear on the teeth, but the body of the gear behind the teeth was not equal in strength to the teeth, and failed by crushing back. (See Fig. 5.) It is essential that large gears should be made of air-harden- ing steel, so as to ensure that the more massive parts attain the required strength ; small gears may be made in oil-hardening steel. The case-hardened gears have given a considerable amount of trouble of various kinds. In many instances the failure of the gear is due simply to the imperfection of the case-hardening. In quite a number of gears the thickness of case on the teeth has been found to be only about l-200th of an inch, a result usually due simply to bad practice in the case-hardening depart- ment of the manufacturer's works. This defect is not always, however, the fault of the case-hardener, for manv instances have been found in which the original depth of case has been quite satisfactory, but the grind- ins; allowance left on the gear has been too large, and a considerable proportion of the case has been removed by grinding. The distortion which occurs during the hardening of the carburised gear is a trouble which ; s not so easily overcome. In some cases the distortion is so serious that the subsequent grinding operations remove almost the whole of the case on one side of the teeth. In all these instances failure occurs bv excessive wear on the pitch line of the teeth, though this fre- quently results in the breaking off of the worn teeth. Other failures have apparently been due to the pro duction of too much case. If a depth of case of l-10th inch is put on a gear tooth it is probable that the sub- sequent operations of hardening will produce very fine cracks under the case, which consequently chips off, and ruins tlie gear. (Ker Fig. 6.) Manv s.roars fail bv the ((racking of the teeth at the root. Quite a number of examples have been examined in which the radius at the root of the teeth is exceed- ingly small, and it is by no means uncommon to find that the tooth is actually undercut. The sharpness of the corner is made still more dangerous by the greater depth of carburisation which takes place there. The concentration of stress at the sharp and brittle corner inevitably starts a crack, which spreads inwards, till the tooth breaks off. Several examples have been found in which cracks, evidently due to concentration of stress, have started on both sides of a tooth, and have grown inwards till they meet. (See Fig. 7.) Several instances have occurred in which the material in the gear wheels has been much below the specified strength. The steel generally used for making case- hardened gears is the 5% to 6% nickel case-hardened steel, which has a very low carbon content, usually .in the region of 0-1 to 0-12%. If a gear is made from steel of this kind it requires a fairly drastic heat treatment to give the required strength in the more massive parts of the gear, which may be at least l\ inches thick. It is not surprising if the strength of the material in this part of the gear is very much below what the designer expects it to be. Instead of 60 tons, which the steel would give when in the form of 1| inch round bar, only 40 tons is found in the actual part. It is a question for consideration whether the use of this very excellent case-hardening steel should not be confined to gears in which the sections are not too large, and whether it would not be desirable to employ a steel containing some chromium, as well as the nickel, for thicker gears, in order to ensure a satisfactory core strength. A factor which has contributed to the failure of gears in several instances is the method by which they have been manufactured, as is explained in the section on Forging Gear Blanks. Valve Springs. When spring breakages were first looked into it was found very difficult to assign any causes for the failures; it was clear that most failures were not due to faulty material. Further investigation, however, showed that the stresses at which the steel worked were so high that the difficulty became rather to explain how many of the springs could stand at all, rather than why they failed. A few instances have however occurred in which the steel wire had incipient cracks which were sufficient to start fractures. Many broken springs have been found made of wire which shows roughness or notches on the surface. A very slight scratch may easily initiate a crack in such stressed parts. In volute springs a similar defect is not uncommon, namely a roughness in the edge of the strip from which the spring is made. The majority of springs appear to be made by cutting off a length of strip and then pro- ducing the necessary taper by grinding it quite roughly to shape. The grinding generally leaves a very sharp corner and often a small burr, and whereas the original strip had nicely rounded edges the. spring has along the ground edge of the coil a verv sharp corner. It is not unusual in consequence to find that tiny cracks have been started in the ground edge which have con- tributed to the broakace in the spring. Cams Cams have in general not been a prolific source of trouble. One or two interesting failures have occurred however in these parts. Speaking first of caR] shafts which are case-hardened in the earn block and uot on the shaft, it lias frequently been found that the case-hardened surface comes away easily by chipping. This has happened most frequently when the cam shaft has been made from plain carbon steel containing about 10 to "20% carbon, and the most common reason has been found to be the presence of well-marked slag lines in the steel. In some cases the slag line appeared on the surface and was sufficiently large to cause the rejection of the piece during machining, but in the majority of cases the slag line lay not far below the surface. It was small enough not to be detected by the naked eye but sufficiently large to cause the thin material immedi- ately above it to fail under load. This has been overcome by the use of high grade case-hardening steel, or better by the use of separate cam blocks pinned on to the shaft. By this arrangement the cam block can be made and hardened very much more satisfactorily than when it is part of the shaft. Another interesting way in which some cams have failed is worthy of mention, a,s it indicates possibilities in connection with case- hardening steels which are not generally accepted. Several instances have occurred in which the cam has failed not by the cracking of the case but by the distortion of the case. The core appears gradually to have crushed and the case has been able to follow it without cracking. This is rather unusual for case- hardening material, but quite a number of instances have occurred on cams of different engines in which a plain carbon case-hardening steel was used, and in general the trouble has been overcome by using a steel which gave a stronger core, such as a 3% nickel C.H. steel. Rocking Levers. Rocking levers have been a very prolific source of trouble, but the breaks which have occurred can be summed up very shortly. (p.) Rocking levers have been case-hardened all over, tin the jwins as well as on the hammer and on the ball end. As a result it is not surprising that they regularly broke in the arms, particularly if the lever was highly stressed. (6) Levers made of other steels have been hardened and tempered to give a strength of 90 to 100 tons/sq. in. As a result they were exceedingly brittle. (c) Levers have been made of mild steel, unheat- treated. As a result they distorted and quickly became useless. The most satisfactory rocking levers have certainly been those made of a 55-ton hardened and tempered alloy steel with a case-hardened roller and a suitable inserted bearing or bush. Gudgeon Pins. Gudgeon pins which fail usually do so in consequence of unsatisfactory manufacture. The causes which have led to their failure are briefly as follows : (fl) Exceedingly deep case-hardening, which has often been carried out on tubular gudgeon pins on the inner as well as the outer surface. (!>) Rough machining inside the pin leaving too] marks from which cracks start. (c) Insufficient case, due to the same causes which have been described in connection with Gears. 27264 (d) Grinding cracks in air-hardening steel gudgeon pins. Such cracks have frequently caused breakages. Sunwheel Carriers. A very interesting batch of failures occurred on the sunwheel carriers on one of the best known engines. They were made from 5% nickel case-hardening steel, lightly case-hardened on the jaws only. (See -Fig. 8.) The failures originated quite definitely in cracks which started at sharp corners and which followed the macroscopic grain of the steel most faithfully from start to finish. There is no doubt that the failures were partly due to the weakness of the steel and partly to the fact that the sunwheel carrier was made by a process similar to the first method described for the manufacture of gear blanks from a flat bar, instead of by the " up-ending " method Pistons. The only pistons considered here are those made of cast iron, which are now only found in a few engines. Several cases of failure of cast iron pistons have occurred, notably during the earlier part of the war. It has not always been possible to state the reason for these failures, and in many cases the failure has been put down to such things as unsatisfactory distri- bution of graphite or excessive phosphorus (when there was a content of only 1%), or too much or too little combined carbon. In all probability these were nothing more than contributory causes of failure. One piston failure was however of interest; the piston failed by cracking in the gudgeon pin socket. The iron was cer- tainly not very strong (about 11 tons per square inch), but the interesting point was that the composition of the skirt and of the head were quite different in regard to the percentage of combined carbon It may reason- ably be assumed that the difference was not due entirely to a difference in chilling during casting, and also that the composition in the skirt represents fairly the com- position of the material in the piston as originally cast. In this part there was about 0~6% of combined carbon but in the head there was less than 0~1%. The piston had run for nearly 100 hours before cracking and the cracks appeared at the junction of the head and the skirt. It appears likely that the difference in composi- tion in the two parts was brought about by the heat while running and that failure was due to the conse- quent change of strength or deterioration of the iron. Piston Rings 1'iston rings have been a very fruitful source of trouble, and still are. What the properties of the ideal ring are, and what material to specify to pro duce those ideal properties, is not known. It is hoped that further information will be available before long. At the present time it appears to have been established fairly definitely that many cast-iron piston rings fail simply by a process of deterioration. Many piston rings have been examined which have failed in practice sim- ply because they became soft. There is no apparent change in the chemical composition of the material, and there is no apparent change in its macroscopic con dition, and yet there has been a vast change in the mechanical properties of the ring. These changes in mechanical properties can be expressed by means of the ratio of the permanent extension after stressing the ring to a definite load to the temporary extension which is produced whilst the load is actually on the ring. In new good piston piston rings this ratio is quite small, but in the rings which have become soft the ratio is very high. There is no question as to how the piston U 2 20 ring has failed, it is undoubtedly due to the deterioration of the iron, but it is by no means clear what this deterioration is, and there is no information as to how it can be avoided. Valves The failure of valves is dealt with in the section on valves, Section 9. Scries 1. Case-hardening steels in accordance with specifications S. 13, S. 14, S. 15, S. 16, and 8. 17 (two samples). 2. FATIGUE RANGES. The importance of the fatigue ranges of all materials used under alternating or fluctuating stresses became more and more apparent during the war, and fatigue tests were consequently made on steels complying with 13 different Air Board specifications, of different chemi- cal compositions, and in many different conditions of heat treatment. Fatigue tests are laborious, each fatigue range deter- mination represents the result of a number of separate tests, usually five or six. The following tables include particulars of 25 fatigue range determinations for equal plus and minus stresses, with the chemical composi- tions, heat treatments, and tensile tests of each steel tested. Most of the results were made by Woehler's method at the N.P.L., but a few were made by Dr. Haigh on his electrical alternating tensile and com- pression testing machine. This machine usually gives slightly lower results than the Woehler. The results show that the fatigue limit (half the fatigue range) for nearly all the different steels lies between '45 and '51 times the ultimate strength. The constancy of this fraction is remarkable. The only steels which have a fatigue limit markedly outside the above figures are those complying with specifications S. 6 and S. 26, which are plain carbon steels, not cold worked. There is no uniform relation between the fatigue limits and the elastic limits or yield points. This fact confirms the opinion expressed in the chapter on Mechanical Tests, where the unreliability and compara- tive unimportance of the elastic limit and yield point are pointed out. List of Fatigue Range Determinations. Series (1.) Five case-hardening steels in accordance with specifications S. 13, S. 14, S. 15, S. 16, and S. 17 (two samples). Series (2) Steel in accordance with specification S.,28 in the fully-hardened condition, and also after tem- pering at temperatures from 200 to 600 C. Series (3) Steels in accordance with specifications S. 11 and K. 1. Series (4) Steels in accordance with specification S. 2 both nickel-chromium steel complying with speci- fication 2 S. 11, and also a chrome-vanadium steel). Series (5) Steels in accordance with specification S. (5 and S. 26. Series (6) Steel in accordance with specification S. 1, part 1, both as drawn, and after blueing. Specification. SI ' S.14. S.15. S.I 6 S.17. S.17. Chemical composi- tion : Carbon, per cent. 0-17 0-19 0-19 0-20 0-16 0-11 Silicon , 0-12 0-05 0-32 0-28 0-11 Manganese 0-46 0-65 0-45 0-39 0-43 0-20 Sulphur , 0-032 0-049 0-028 0-033 0-044 Phosphorus , 0-023 0-025 0-021 0-022 0-033 Nickel Nil Nil 3-00 3-15 4-65 5-76 Chromium , Nil Nil 0-48 0-58 0-29 Heat treatment : Refining temper- ature C 900 880 850 840 820 820 Hardening tem- perature C 770 770 770 770 760 760 Tensile, test : Elastic limit, tons/sq. in. 17-6 15-1 16-0 14-5 14-0 20-0 Yield point 24-7 22-1 38-2 38-0 44-1 30-0 Ult. strength 39-5 38-2 65-5 77-5* 60-0 57-8 Elongation, per cent. 27-0 30-5 23-0 16-5 20-1 21-7 Red. of area 63-5 64-8 44-5 37-5 47-0 37-2 Fatigue tests : Fatigue range (Woehler) 18-4 17-3 +31-0 29-5 30-0 26-5 Ratio Fat. Lim. to Ult 0-47 0-46 0-47 0-38* 0-50 0-51 * NOTE. This appears to be a freak result. Series II. A steel in accordance with specification S. 28, fully hardened and tempered at different temperatures (See Fig. 9.) Chemical composition: Carbon, per cent. ... ... ... ... 0-30 Silicon Manganese Sulphur Phosphorus Nickel Chromium 0-22 0-56 0-041 0-015 4-30 1-40 Heat Treatment. All the specimens were air hardened from 800 C., and tempered at different temperatures. Tempering Temperature. 200 C. 400 C. 500 C. 600 C. Tensile testx : Elastic limit, tons/eq. in. 20-0 36-2 53-3 SI -7 40-9 Yield point ,, 78-8 77-5 80-5 71-3 63-3 Ult. strength 109-1 101-2 97-9 82-4 70-1 Elongation, per cent. ... 10-8 12-5 10-0 15-0 17-5 Red. of area 36-6 41-6 36-2 46-4 55-1 Fatigue tettt : Fatigue range (Woehler) +45-5 -f-51'5 +47-5 -1-41 T, -1-35 -5 Ratio Fat. Lim. to Ult.... 0-42 0-51 0-49 0-60 0-50 FATIGUE LIMITS FOR STEEL TO SPEC* $28 AlR -HARDENED FROM 600* C AND TEMPERED AT VARIOUS TEMPERATURES MATE STRENGTH 00" 200 500" 400 500 TEMPERING TEMPERATURE J c 9/65/ * 2 000. '. 4 n f ' 273 6OO Scries III. was quenched in oil from 850 C. and tempered at 6000 C. Chemical Composition : '}'.) S. 6. 8.26. Silicon ,, 0-23 Tenxile test : Elastic limit, tons/sq. in. Yield point Ult. strength Elongation, per cent.... Red. of area, per cent. 23-1 24-9 37-9 31-1 57-9 -+ 1 7 (Woehler) 15 (Haigh) 0-45 (Woehler) 0-40 (Haigh) 28-4 31-5 46-8 28-3 62-3 -4- 19 (Woehler) -+- 18 (Haigh) 0-41 (Woehler) 0-39 (Haigh) Manganese Nickel ,, ... 0-40 3-48 Chromium ,. 0-78 Sample. A B. Fatigue range ... Ratio Fat. Lim. to Ult. Tensile tc.t's : Fatigue range -4-33 (Woehler) -t-27'9 (Haieh) Ratio Fat. Lim. to Ult. ... 0-50 0'49 Series IV. Steels in accordance with specification S. 2. A. U. Chemical eumpntition : Carbon, per cent. 0-30 0-44 Silicon , 0-31 0-25 Manganese , 0-49 Sulphur , 0-022 0-029 Phosphorus , 0-0'4 0-038 Nickel 3-30 Nil Chromium . 0-60 0-95 Vanadium , 0-10 0-20 Tensile test : Elastic limit, tons/sq.in. 44-8 Yield point 53-6 54-7 Ult. strength ,, 60-5 62-9 Elongation, per cent. 22-0 20-0 Red. of area 64-0 54-6 Fatigue test : Fatigue range / +34 (Woehler) L -1-27 -2 (Haigh) j32-8 (Woehler) Ratio Fat. Lim. to Ult. ( 0-56 (Woehler) \ 0-46 (Haigh) 0' 52 (Woehler) Series VI. Steels to specification S. 1, part 1 both as drawn and after blueing at different temperatures. Chemicil Composition. A. B. Carbon, per cent. 0-26 0-25 Silicon 0-05 0-08 Manganese, per cent. 0-54 0-68 Sulphur 0-031 0-051 Phosphoru" 0-027 0-043 STEEL A. Tensile Tests. Blueing Temperature : Centigrade. Not blued. 250 400 550 Yield point ... 36-8 36-6 28-4 26-4 Ult. strength ... 40-8 40-1 39-9 36-6 Elongation 13-3 15-0 16-6 23-3 Reduction of area 52-6 62-6 so-o 52-8 Fatigue Tests. Fatigue range Ratio Fat. Lim. to Ult. 19-1 0-47 18-4 0-46 19-0 0-48 18-0 0-49 Series V Haiti carbon steel in accordance with specifications S. 6 and S. 26. Chemical composition : Carbon, per cent. ... ... ... ... 0'36 Silicon ,, Manganese ,, Sulphur ,, Phosphorus ,. 0-25 0-66 0-041 0-020 Heat Treatment. Steel to specification S. 6 wat normalised at 850 C. Steel to specification S. 26 was STEEL B. Tensile Tests. Yield Point 26-1 31-2 30-5 27-2 Ultimate Strength ... 36-2 38-4 37-9 35-2 Elongation 20-8 22-5 20-0 28-3 Reduction of Area ... 54-1 50-9 50-1 52-6 Fatigue Tests. Fatigue Range Ratio Fat. Lim. to Ult. 17-0 0-47 17-2 0-45 17-0 0-45 16-7 0-45 i'he results are summarised in the following table : Summary. Specification. Treatment. Ult. Strength. Fatigue Range. S. 13 Refined and Hardened 39- + 18- 4 W. 14 i > 38- 2 17' 3 15 i 65- f. 31 16 77- r> 29- 5 17 T> 60- 30' 17 57- 8 26- 5 S. 28 , 109' 1 -4- 4'- 200 C. lot- 2 ~ r.i 5 400 97- 9 47- 5 500 82- 4 41' 5 600 70- 1 35- 5 S. 11 orK . 1 ' A. To specification 65' 7 -I-33- ** it i, 57- 5 27- 9 H. S. 2 A. 60- .-, + 34W. -4- 27-2 H. B. 62- 9 32- 8 W. 86 Normalised 850 37- 9 + 17 W. + 15 H. S 26 Q 850 temp. 600 46- 8 19 ,, 18 Blued at 40- 8 -1- 19- 1 W. S. l,part A. 1 250 C. 400 550 40- 39- 36- 1 9 6 18- 19- 18- "* t) o o f 36- 2 17' S. 1, part B. 1 1 250 400 I. 550 38- 37- 35' 4 9 2 17- 17- 15- 2 o 7 W. Woehler Test. H. Haigh Test. 3. HAIR CRACKS. A great deal of trouble has been caused during the war by small Haws found in many forgings, particularly in- crankshafts. Hundreds of crankshafts at a time have been rejected because of them. These flaws are of various kinds; some of them are of quite well-known types, and are fairly easily visible, for example, laps, heat treatment cracks, large slag inclusions, and elongated blowholes. Besides these, however, there are much smaller cracks, which are very much harder to find by inspection, though they are usually much more numerous. These have been named " Hair Cracks." (See Fig. 10.) They are usually about half to three- quarters of an inch long, and vary in depth from about one to three thousandths of an inch. In width they very somewhat, but are generally so fine as to appear no more than a line when viewed under an ordinary hand magnifying glass. Hair cracks are rarely found singly; if one is discovered usually one or two more will be found in the immediate neighbourhood. If a hair crack be discovered on a journal and a thin cut of one or two thousandths be taken off, more cracks will generally be found on the new surface. This process may be repeated until the whole of the journal has been turned off, hair cracks being found all through it. Hair cracks are very troublesome to the engineer, as they can be discovered only after the machining .and polish- ing have been completed It is quite certain that the true hair crack originates 111 the ingot and is not produced during the subsequent operations of forcing and heat treatment, it is probable, in fact, that the operation of forging diminishes the hair cracks rather than increases them, and in no case can these operations start a crack. Hair cracks are now believed to be the relic of very small contraction cavities or discontinuities in the original ingot. When the ingot has completely solidified and is still at a high temperature, e.g., 1350 C. the final outside shape of it is fixed, not only on the part which is bounded by the mould, but also on the upper crust. As the metal cools it contracts, aud as the inner part is hottest it contracts more than the cooler shell. The inner part being hotter is also weaker, and the difference in con- traction results in the formation of very small discon- tinuities between the macro-crystals in the interior of the ingot. (See Fig. 11.) As stated in the section on macros tructure, the ingot is built up of an agglomeration of crystals bound to- gether chiefly by their interlocking; when therefore the contraction stresses arise in the middle of the ingot, the crystals are pulled apart to a greater or lesser de- gree, and intercrystalline spaces are formed. The part of the ingot where this will occur most easily is that part which was fluid last, as this will be the most loosely knit. By forging the ingot these spaces may be eliminated for all practical purposes; forging distorts the crystals, thus causing proper interlocking, and it also to some extent welds them up. If forging is not completely or sufficiently carried out then the spaces will persist, and, although distorted, will not be removed. It is these persisting spaces which constitute the hair cracks in the finished forging. The portion of the ingot in which the original inter- crystalline spaces occur will not always be the same. In some ingots, particularly those poured from the top, the middle of the ingot will be the last to solidify, and will therefore contain the defective steel. In other ingots the part which free/es last is an annular ring lying about l/5th of the diameter from the outside of the ingot, and in these ingots the cracks will be found chiefly in this space. It is desirable to arrange that the part of the ingot which is richest in the defects shall be discarded during the early stages of manufacture and before the steel for the forgings is cut off. If the centre is likely to be the part affected, then it should be cut out and not used in the forgings at all. This has been done successfully by one firm, who roll the ingot into slabs and then reject the middle portion of the slab. With ingots in which the trouble does not lie in the middle the process of removing the defective part is not so simple, but it can be effected. In order to avoid hair cracks the utmost possible care- should be paid to the production of good sound ingots. The most satisfactory method depends to some extent upon the plant available, but it is certain that the whole trouble can be avoided or reduced to a minimum by careful attention to details in the manufacture of the ingots. It is a must important question whether or not these tiny flaws are harmful to the crankshaft in which they occur. Do they cause failure or do they not ? It is impossible at present to give a definite reply. As n ii suit of the examination of many shafts which have broken and of all the records available, it is fairly cer- CHAITKI: IV. [ To face page 22. FIG. 10. Crankshaft journal showing hair cracks. They appear as two fine white lines just above the middle of the smoothed portion in the middle of the photograph. _ 'W?-'*^& -- ^t?3r--. "-^- .. ^ ' >'*V >. ' '.' : ....-- - ' '. V-.v.-, .. Xi^;-^.- >i - - ^V >Sp ; FIG. 11. Discontinuity in ingot. Photograph of a sulphur print, the location of the cavities being indicated by large dark areas. The cavities are not so large as the dark areas, which appear large owing to the segregation of sulphur round them. Chapter. III. 65O 550 250 450 350 Fio.12. IZOD FIGURE OF TOUGH SAMPLES CURVE A WHEN COOLED SLOWLY FROM 65O* TO DIFFERENT TEMPERATURES. THEN QUENCHED CURVE B WHEN QUENCHED IN LEAD BATH AND COOLED SLOWLY FROM DIFFERENT TEMPERATURES. 150'C 11/20. Chapter ill 120 z 5 60 40 * B CARBON NiCKEL. CHROMIUM 29 SS 3. SO 1-46 too 2OO 300 400 Fis. 13. 500 600 yoo'c RELATION OF Izoo FIG. TO THE TEMPERATURE AT WHICH THE STEEL is TEMPERED. SHEWING EFFECT OF KRUPP KRANKHEIT BETWEEN 2OO&550. 1200 40 10 050 SSO'C 550 A50 FIG.I+. IZOD FIGURE OF TOUGH SAMPLES CURVE C WHEN RE-HEATED FOR I HOUR AT TEMPERATURES. THEN QUENCHED. CUP.VE O WHtM Rt-HEATtO FOR. I HOUR AT DIFFEKEKfT TEMPtRATURtS AXO COOLED SLOWLY. *50 450 350* C FlO. 15. RECOVERY OF BRITTLE SAMPLES. CURVE L. WHEN REPEATED TO DIFFERENT TEMPERATURES AND QUENCHED. 23 tain that these flaws are practically harmless unless they occur in very vital parts. A large hair crack or a collection of small ones in a fillet of a shaft may be dangerous, but it is quite certain, owing to the diffi- culty of detecting the cracks, that a large number of cranks have been used in which they they have existed and that no trouble has resulted. When definite ex- periments have been made to determine whether or not observed flaws would cause failure, the results have been quite inconclusive. In one or two cases the shaft has broken through the flaw; in several other cases the shaft has broken at a point away from the flaw. The majority of the evidence is to the effect that the flaws are generally not harmful. Hair cracks are not a new type of defect. They have probably always existed in steel ingots, and the reason why they have not caused trouble before is not because they were not there, but because they were not dis- covered, the inspection having been less rigorous than that carried out on aeronautical parts. Bearing this in mind, it is reasonable to conclude, as the parts with the undiscovered flaws have done satisfactory service, that hair cracks are not very serious defects. 4. KRDPP KRANKHEIT. Krupp Krankheit is the name given to a defect, shown by the low value of the notched bar test, pro- duced in some Nickel-Chrome steels by cooling them slowlv from or through a certain range of temperatures. Mr. H. Brearley was one of the first to discover this phenomenon and how to prevent it, and he has been investigating it for some years. The following informa- tion was supplied bv him to the Air Board for publica- tion and was issued in C.T.M. 42. Krupp Krankheit does not occur in any steels except Nickel-Chrome steels unless they are cooled extremely slowly. The figures given below have been obtained on acid open hearth steels containing from 1 to 1'5% Cr. and 3'5 to 4-5% Ni. ; the same phenomenon occurs in all Nickel-Chrome steels, but to varying extents. It has been observed in steels containing as little as '45 Chromium, with Nickel varying between 1'25% and 5%, and in steels containing as much as 2% Chromium, with Nickel varvinsr from 1% to 4%, and in steels of inter- mediate composition. It is less marked in electric furnace steels, and is often practically absent in cru- cible steels. The exact cause of the illness is not yet understood, but the way to produce it and the way to cure it are known. It may be produced and cured repeatedly in the same piece of steel. The phenomenon is shown most strikingly by the following series of tests. A dozen pieces of Nickel-Chrome steel were oil hardened from 820 C., and tempered by heating to 050 and quenching them. So treated, they had an ultimate strength of 60 tons/sq. inch, and gave an Izod figure of 65 ft. Ibs. Five of them were then re-heated to 650 and allowed to cool slowly in the furnace (4 C. per minute). As they coolfd. thp sampler wore drawn 'Hit of the furnace one at a time and quenched, They gave the following results, which are plotted in Curve A, Fig. 12: Cooled from 650 C. slowly to, and quenched quickly from 650 C 550 C 450 C !350o C 250 C. Izod figure. . 65.66.70 . 76.65.75 . 28.34.34 . 23.22.30 22.19.23 The curve shows that slow cooling to 550 did no harm, but slow cooling to lower temperatures rapidly reduced the Izod figures. The remaining seven test pieces were put in a furnace at 650 and kept at that temperature and were with- drawn, one at a time, and quenched in a lead bath which was slowly cooling from 650 (2 C. per minute). They gave the following results, which are plotted in Curve B, Fig. 12: Quenched from 650 C. quickly to and cooled slowly from 6500 C 600C 500 C 4000 C 3000 C 2500 C 15 C. Izod figure. 7&14 6& 6 ... 33.87.39 ... 71.69.68 ... 70.74.75 ... 63.65.66 ... 66.63.65 The curve shows that slow cooling (in the lead bath) from all temperatures above 400 C. lowers the Izod figure. Considering both curves together it appears that slow cooling does no harm from 650 to 550, or from 400 to zero; but slow cooling through the intermediate range, 550 to 400, seriously reduces the Izod figure. This series of tests is particularly striking because the tensile properties of the steel are the same in all the samples. They were originally tempered at 650, and the subsequent treatments below that temperature do not alter the ordinary tests (Ult. Strength, Elastic Limit. Elongation, and Reduction of Area). These tests, however, do not indicate what will happen if the same steel is tempered at different temperatures. Under these conditions all the properties of the steel will differ with each temper, and it is not possible to demonstrate so clearly the effect of the Krupp Krankheit; but the results are more important as they represent the con- ditions which ordinarily occur in practice. The results of a series of tests on samples tempered at temperatures between 100 and 650 are plotted in Fig. 13. The tests show a serious drop in the Izod figures for a range of temperatures between 200 and 500. or a little higher. In steels which are not affected by Krupp Krankheit the Tzod figure rises more or less regularly with the rising temperature. Translated into a general rule for the heat-treatment of forgings made from Nickel-Chrome steels, this means that the re-heating temperatures should be fixed as high as possible, consistently with tlie specified ulti- mate stress, in order to obtain good impact figures. 24 To emphasize the serious effect of re-heating to tem- peratures below 550, a series of tests were made on samples which originally gave Izod figures between 60 and 65. The samples were heated for an hour at various temperatures and then quenched. After this treatment they gave the following results, which are plotted in Curve C, Fig. 14 : Water quenched after re-heating at Izod figure. 400 C. ... ... 62 ft. Ibs. 425 C ... 56 450C ? 4750C ... 52 5000 C 24 _ t> 550 C 27 6000 C 52 ,, 6500 C 63 A similar set of samples were treated in the same way, only allowed to cool slowly, instead of being quenched. They gave the following results, which are plotted in Curve I), Fig. 14: Cooled slowly after re-heating at Izod figure. 450 C 21 ft. Ibs. 5000C 2 5500 C 2 6000 C. ... l n n To show how the steel recovers if given proper treat- ment a further series of samples, which had been made brittle by slow cooling from 650, were heated to various temperatures and quenched. They gave the following results, which are plotted in Curve E, Fig. 15 : Water quenched after re-heating at Izod figure. Not treated 21 ft. Ibs. 450 C 25 5000 C 26 5500 C 25 6000 C 68 ti M The following practical rules may be deduced from these results : 1. Always quench after tempering; never allow the steel to cool in the furnace. 2. "Use a steel which will give the required ultimate strength when tempered at 550, or, better, at 6000 c. As has been already stated, different nickel-chrome steels suffer from this illness to different extents. A suitable measure of the extent to which they suffer is the ratio between the Izod figures for two samples, one tempered and quenched at 650, and the other tempered and slowly cooled from 650 (cooling, say, in 10 hours). This ratio varies from 30 to 2. What the physical meaning of Krupp Krankheit may be has not yet been ascertained. The appearance to the naked eye of the fractures of the, two samples 37 and 38 are very different, and suggest that the steel has failed in different ways; but hardly any difference can be seen between the microstructure of the two samples. As a result of the information supplied by Mr. Brearley, clauses have been introduced into some of the steel specifications requiring the steel to be quenched after tempering. (See Specification S, 33, Clause 5b.) The following additional information was prepared by Captain L. Aitchison and issued in C.I.M. 717 and 735. The above information supplied by Mr. Brearley does not give any indication as to what would be the effect on the notched bar test of re-heating a correctly tempered steel to the temperature at which Krupp Krankheit is produced. That this information is necessary will be real- ised if certain practices in connection with crankshafts are considered. There are certain processes used in the manu- facture of crankshafts which involve the heating of the steel, after hardening and tempering, to a temperature well within the dangerous range. In some cases quenched and tempered crankshafts are heated to an undetermined temperature in order that they may be straightened. In other cases the crankshafts are heated to a temperature of approximately 500 C. and allowed to cool slowly, in order to remove as far as possible the strains* set up in the metal during the tempering and quenching processes which might produce distortion during the machining operations. In order to determine the effect of this second treat- ment upon the steel, two series of tests have been car- ried out. The first series was made upon steel which was not liable to Krupp Krankheit, whilst the second series of tests were made upon a steel which was very liable to it. In each series the steel was quenched and tempered, and cooled after tempering by quenching in water. Test pieces were cut from the steel and then heated to various temperatures lower than the one at which they had been tempered and the Izod figure of each determined. The results are shown below. FIRST SAMPLES. A steel which is not liable to Krupp Krankheit. (a) Chemical Composition. Carbon Silicon Manganese Nickel Chromium . 0'44 per cent. 0-14 0-49 3-22 0-56 (b) Heat Treatment, Heated to 820 C, and quenched in oil. Reheated to 650 C, and cooled in water. (c) Mechanical Tents after the above treatment. Yield point ... ... 48'0 tons/sq. inch. Maximum stress ... 58-0 ,, Elongation per cent. ... 26'0 Reduction of area per cent. 62-0 Izod figure (i)66, 56, 68 foot Ih. (ii) 58, 71, 59 * It is doubtful whether strains are actually produced. 25 (d) Izoil Tr.sfs on samples which had been hardened and tempered as in (b) and then reheated to various temperatures : (i.) Reheated to 400 C., cooled in air ... 58, 59 foot 11 J ' ') J oil ... 63, 60 (ii.) , 450 C., , air ... 58, 63 i) oil ... 70, 55, 65 (iii.) 550 C., air ... 53, 56 J1 oil . 56, 61 (iv.) 000 C., air . 58, 52 M oil . 54, 60 (v.) 050 C., air . 70, 55, 71 J oil . 61, 64 (vi.) Reheated to 600 C,, and allowed to cool slowly in furnace during several hours ... 35 foot Ib. These results clearly show that the notched bar tests are not affected by these various treatments and that reheating within the range of temperature which in many steels produces Krupp Krankheit lias no dele- terious effect upon this steel. SECOND SAMPLES. A steel which is liable to Krupp Krankheit. (a) Chemical Composition. Carbon ... ... ... 0'35 per cent. Manganese ... ... - 41 ,, ,, Nickel 3-42 Chromium ... ... 0'81 ,, ,, (b) Heat Treatment. Heated to 850 C. and quenched in oil. Reheated to 050 C. and cooled in water. (c) Mechanical Tests after the above treatment. Yield point Maximum stress Elongation per cent. ... Reduction of area per cent. ... Izod impact 49'0 tons/sq. inch. 57-5 25-0 59-0 47. 48, 50 foot Ib. Izod tests were also taken on samples quenched as above, but cooled from 650 C. in other ways than in water. These show the effects of Krupp Krankheit. (i.) Cooled in oil (ii.) air (iii) furnace... ... 42, 44, 47 foot 11). ... 15. 16. 19 ... 7, 7, 6 ) Izod Tests on samples which had been hardened and tempered as in (b) and then reheated to various temperatures : (i.) Reheated to 450 C., cooled in water oil 550 C., 650 C.. 20, 25, 26 foot 11. 31,32,35 air ... 28, 32, 33 furnace... 19, 20, 18 water ... 11, 6, 13 oil ... 11, 10, 11 air ... 7, 7, 9 furnace... 5, 6, 5 ., water ... 52, 52, 56 ., oil ... 56, 55, 53 air ... 16, 15, 1-3 .. furnace... 6, 5, 2 , temperature within the danger zone however sub- sequently cooled. Test (d iii.) shows that reheating such steels to tem- peratures above the danger zone must be followed by quick cooling to prevent Krupp Krankheit, but this pro- cedure would be useless for removing quenching stresses. Reheating any steel to temperatures below the danger /one produces no bad effect. This has been proved by a thorough investigation made by Mr. J. H. S. Dickenson. As it is impossible to distinguish between crankshafts which are liable and those which are not liable to Krupp Krankheit it is clear that reheating, if used, must be confined to temperatures below 400 C. To illustrate the obscure nature of the illness that attacks the steel, and to show how uncertain its inci- dence is, some interesting figures, compiled from data supplied by Mr. H. Brearley, are given in the table below. Each line in the table represents a separate cast of steel. The Izod figure in column 5 was obtained from a specimen which was hardened in oil, then reheated to 650 C. and cooled in water. The Izod figure in column 6 was obtained from a specimen hardened as before, reheated to 650 C. and cooled at a very slow fixed rate, so that the steel took eight or nine hours to cool from 650 to 400 C. The chemical composition of the steel in each case is given, and shows that all the steels are practically identical so far as composition is concerned. The final column gives the ratio of the Izod figures for furnace-cooled and water-cooled samples; this figure is a convenient measure of the susceptibility of the steel to Krupp Krankheit. The tests are arranged in order of susceptibility and show the wide variation which exists. In only the last three cases is the Izod figure higher after slow cooling than after cooling in water. SUSCEPTIBILITY OF DIFFERENT CASTS OF STEEL TO KRUPP KRANKHEIT. Tests (d i.) and (d ii.) slio\v the bad results which follow heating steels subject to Krupp Krankheit to a Carbon. Man- ganese. Nickel. ' Chromium. Izod Figures. Suscepti- bility. (1) Water Cooled. (2) Furnace Cooled. 1 :2. 33 55 8-71 91 46, 48, 49 6. 4, 3 12-5 31 44 3-74 86 33, 40 6 6-3 29 60 3-81 90 50, 51, 50 8 6-3 so 45 3-44 1-27 48 14, 12 3-7 85 59 3-61 1-00 37, 38 11 3-5 32 67 3-75 94 21, 24 7 3-3 36 57 3'62 9(i 31, 40 11 3-2 31 59 3-61 94 52 10, 23 3-2 30 58 3-61 1-03 47 20, 12 3-0 36 55 3-79 93 40 17, 9 3-0 35 48 3-54 94 55, 53, 58 20, 18, 16 3-0 29 60 3-81 1-51 35, 32 15 2-3 30 53 3-71 94 53 20, 26 2'3 34 63 3'77 95 43 20 2-2 81 54 3-57 94 53 25, 25 2-1 33 55 3-63 1-01 30, 30 16 1-9 38 51 3-44 1-73 37, 41 23 1-7 33 51 3-67 85 35, 33 20 1-7 34 54 3-63 92 34, 32 21 1-7 35 38 3-68 91 42 33, 31 1-3 32 47 3-43 88 60, 56, 54 72, 72 0-7 30 60 3-64 1-15 40 53, 51, 53 0-77 38 62 3-56 I'll 32, 31 43 0-75 2726 Conclusions. 1. The effect on the notched bar test of reheating a quenched and tempered steel depends on the tempera- ture to which it is reheated, and on whether the steel is liable to Krupp Krankheit. 2. E cheating to any temperature below 400 C. does no harm in any steel. 3. Reheating to any temperature between 400 and 600 C. will reduce the notched bar test to an extent depending on how far the steel is liable to Krupp Krank- heit. This result cannot be modified after reheating by quenching the steel. 4. Reheating a steel liable to Krupp Krankheit to a temperature, above 600 C. will lower the notched bar test unless the reheating is followed by quenching, which makes the process useless for removing quenching, strains. 5. Most nickel chrome crankshaft steels are liable to Krupp Krankheit, but to very varying degrees. 6. For safety it must be assumed that every crank shaft is liable to Krupp Krankheit, and therefore no shaft, after it has been hardened and tempered, should be heated to a temperature above 400 C. The following is an extract from C.I.M. 735 : THE METHOD OF COOLING STEELS AFTER TEMPERING. Attention is drawn to the following points: 1. That there is a fair proportion of nickel-chrome steels which give perfectly satisfactory Izod values if cooled in air after tempering. For these steels there is no point at all in cooling them in water after tempering. 2. That other steels give quite good Izod values when cooled in oil after tempering. This treatment is per- fectly satisfactory for such steels. 3. That the critical factor in the case of the steels which give poor Izod values when cooled in air after tempering is the rate of cooling. The rate of cooling of a part depends upon its mass, consequently a part of large mass may require to be dipped in water after tem- pering in order to cool it at a sufficiently high rate, whereas a thin part, or one of small mass, made in the same steel will cool in air at a sufficiently high rate to give proper Izod values. 4. That as there are certain objections to cooling forgings and drop forgings in water after tempering, this process should only be applied to those steels and those large parts which need to be dipped in water in order to make them cool at a sufficiently quick rate to give good Izod figures. It is not usually necessary to experi- ment in order to ascertain the required treatment for the steel, as this is given in the A.I D. release note, which shows in what way the steel was treated at the time of acceptance at the steelmaker's works. 5. That the process of cooling steel in water after tempering has no hardening effect upon the steel, and that this process in no way affects the tensile properties of the steel. After the initial hardening the tensile pro- perties of the steel depend simply upon the temperature at which it is tempered. 6. That it is extremely improbable that any part will be cracked or appreciably distorted by cooling it in water after tempering. The operation of cooling in water after tempering is sometimes regarded as similar to that of quenching the steel in water in order to harden it. This latter operation at times has the effect of cracking or distorting the part. The former operation is not likely to do this, and the effect of the two opera- tions must not be confused. 5. MACROSTRUCTt'RE. If a section of steel is polished and is then attacked by suitable agents, such as Heyn's reagent, consisting of a mixture of hydrochloric and sulphuric acids, or by a solution of copper chloride containing a small quantity of free hydrochloric acid, it is found that the steel surface does not dissolve uniformly, but that certain parts are attacked more rapidly than others, which therefore remain in relief. The resulting pattern may either be photographed or transferred to paper by using the st'eel section as a type and taking an impression with printers' ink. The pattern discloses what is called the macrostructure of the steel. It gives a great deal of information about the steel, as certain types of macro- structure are associated with steel which has been cast properly, others are associated with steel which has been cast improperly, whilst others again are associated with forged steel. From an examination of the macro- structure of a steel part it is generally possible to state exactly how the part has been made, and it is frequently also possible to determine the quality of material, and to state in what way defects in it may have arisen. It has long been known that cast steel has a structure which is built up of individual crystals formed from the molten steel at the time of solidification. These crystals are true crystals in so far as they represent the original form in which the liquid steel solidified. (See Fig. 16.) The actual form in which the crystals appear depends considerably upon the conditions under which the steel was cast, but in general they may be described as long and thin. They resemble the elevation of a distant pine tree (Fig. 17) and may be considered to be some- what similar in structure to that tree, for they consist of a core from which large branches have sprung off. and from these large branches smaller branches have separated, the process going on repeatedly. In a steel ingot these crystals form as regularly as they may, starting from the side of the mould or any similar solid foundation, and interlocking with one another until they occupy the whole space. These crystals are revealed as the macrostructure of the cast metal. This form of macrostructure is more or less common to all cast steels, whether the ingot is large or small. After forging there is, however, a distinct and complete change in the macrostructure. It may be assumed for purposes of description that the steel consists essentially of two constituents, which differ in some way at present not known definitely, though often attributed to phosphorus segregation, but which is sufficient to cause the two parts to re-act differently under etching attack. It may be assumed also that each constituent retains its characteristics during the process of forging, so that the constituent which before forging was attacked by the etching agent is acted on similarly after forging. The. result is that the part of the steel which stood up CHAPTER III.] [ Tu face page 26. FIG. Ifi. Macro-structure of large ingot. FIG. 17. Steel crystal. FIG. 18. Macro-structure of forged steel. 2o face page 27. | [CHAPTER III.. ::i ~-^:-'- JT. '=3S~^..:-.... FIG. 19. Snaky-grained crankshaft. Flo. 20. Snaky. grained crankshaft. FIG. 22. Straight-grained crankshaft. FIG. 21. Straight-grained crankshaft. 27 in relief when the material was in the cast state also stands up in relief after forging. It is found that the effect of forging upou the macrostruoture depends entirely upon the extent to which it has been carried out. In large slabs, which have been rolled from ingots, and in which the amount of deformation is relatively small, it is not uncommon to find the macrostructure of the inside of the slab almost identical with that of the ingot. On the other hand, when the ingot is forged, and then rolled down into a small bar, all traces of the original structure are obliterated completely, and in place off the collection of pine tree crystals the steel bar apparently consists of elongated strings of drawn-out material, the appearance being not unlike that of an untwisted wire rope. (Sec Fig. 18.) This difference in macrostructure has a considerable influence upon the properties of the metal, i'or instance, in the interior of a cast ingot (in which presumably the arrangements of the crystals is the same in all direc- tions) it is found that the mechanical properties of the steel are not affected by the relation of the axis of the test piece to the original lump of metal. That is to say, that if test pieces are cut parallel to all three axes of an ingot, or iu any other direction the ultimate strength, elongation, reduction of area and notched bar test, are practically identical for all the test pieces. Quite a different result is obtained, however, if the test pieces are cut from steel after forging. For purposes of expla- nation it is convenient to speak of two directions, namely, along the grain and perpendicular to the grain of a forged piece, the word grain being used in the same way as it- would be if a plank of wood were being dis- cussed instead of a piece of forged steel. It is found in forged steel that the ultimate strength is the same in whichever direction the test piece is cut, but that the elongation, the reduction of area, and the notched bar test, are very much lower in the sample cut across the grain than they are in the sample cut along the grain. The following are typical figures: Along the Grain. Across the Grain. Ultimate Strength Elongation ... Reduction of Area Notched bar (Izod) 60 tons/sq. in. 18 per cent. 45 per cent. 40 ft. IDS. 60 tons/sq. in. 3 per cent. 8 per cent. 5 ft. Ibe. Assuming then that the ductility of a steel, as shown by its elongation, reduction of area, and notched bar test, have some value, it is of importance in an engine part that the grain should be arranged as far as possible so that the main stresses in the part are applied in a direction parallel to the grain. The possibility of doing this may be shown very clearly in connection with forg- ing* such as crankshafts, gear blanks, and valves. Two methods of manufacture may be adopted for crank- shafts. In the first the grain is arranged so that it follows the contour of the machined shaft from end to end, so that it lies parallel to the axis of the shaft in the journals, and in the pins, but perpendicular to the axis of the shaft in the webs, the grain lying along a snaky line. (See Figs. 19 and 20.) In the other method the grain lies parallel to the axis of the shaft in all members, without variation of direction in either the pins, journals or webs. (See Figs. 21 and 22.) In the 27264 first method of manufacture an arrangement of the grain of the shaft is produced such that the stresses in the webs are applied principally in a direction parallel to the grain, whilst in the second method the stresses in the webs are applied in a direction perpendicular to the grain. It is reasonable to suppose that the first arrangement is better for the shaft than the second, when the difference between the tests along and across the grain is remembered. An interesting light is thrown on the effect of the macroscopic structures of shafts made by the two methods by the different ways in which they break in service. An examination has been made of a large number of crankshafts which have broken, and it has been found that in general the shafts with a snaky grain break iu quite a different way from those with a straight grain. Most of the shafts which have been examined have broken through a web, and in the majority of cases the break has commenced in the fillet at the junction of the web and the journal. In the shafts which have a snaky grain the break has pro- ceeded in a direction which bisects the angle between the web and the journal, thus running in a direction normal to the fibres of the grain at this point. In the shafts with a straight grain the crack runs parallel to the grain in the shaft thus running along the fibres of the grain, and not across them. It is probable that the stresses in the shaft which produce the crack would tend always to make the crack extend in the direction which it follows in the shafts having the snaky grain, and that this direction is only changed if the shaft possesses a path of least resistance in some other direc- tion. The fact that in the straight -grained shafts the crack follows a path along the grain suggests that in this direction there is less resistance to the passage of the crack than across the grain. Assuming this to be cor- rect, it is important that the path of least resistance should be arranged so that the stresses imposed are not at right angles to it. The macrostructure of an engine part also reveals the form of the ingot from which it was made, and the extent of the forging which has been done upon the material. When the shafts with the snaky grain are made the material is taken from an ingot which has been slabbed out and then cut into large bars, which are then hammered into the rough shape of the shaft and finished off in dies. The amount of work done is large, and the resulting grain is very regular. The usual method of manufacture of the straight-grained shaft is to roll out a large ingot into a slab, and then to cut off a smaller slab as large as the biggest overall dimensions of the finished forging. The result of this process is that the amount of work put upon the different parts of the finished shaft may be quite unequal. The journals, which are taken from the middle part of the slab, have generally been worked much less than the crankpins, which come from the outer parts of the slab. The struc- ture in the journals of some shafts is practically the same as that in the ingot from which the slabs have been cut, so that there is a great difference between the two parts of the shaft, and consequently there is likely to be a corresponding difference in the mechanical pro- perties. In addition to this the surviving ingot struc- ture is generally that of the interior part of the ingot, which is the portion most likely to be unsound. This does not imply that there are large pipes and blow holes, but that there is a lack of continuity between the D 2 28 crystals in consequence of the contraction of the ingot. These discontinuities are discussed in the section deal- ing with Hair Cracks. They are largely eliminated by the forging operations, in which the crystals arc forced up against each other and distorted, or made to interlock. In making large slabs the amount of work done is not sufficient to distort the crystals and to make them interlock, and hence the spaces remain. They may not always give trouble, but sometimes they do, for example, in the majority of six-throw shafts. These shafts are made by twisting the journals of the roughly- shaped shaft. (See Fig. 23.) The operation of twisting is probably quite harmless if done properly, but when the steel contains a space or a discontinuity between two crystals the operation of twisting opens out of the space. (Sec Fig. 24.) No space can be created by twisting but a minute space may be enlarged. Forgings have been examined in which the twisting has opened out a discontinuity which probably in the original ingot was only microscopic, into a hole about 4 m.m. diameter. A shaft containing such a hole in the journal is, of course, rejected if discovered, and is dangerous if not discovered. In valve forgings the importance of the macrostruc- ture is considerable. Valves have been made very largely in the past by turning them out of a bar of steel; the resulting arrangement of the grain is shown in Fig. 25. When the valve is made properly the arrange- ment of the grain is as shown in Fig. 26. The importance of the macrostructure of gear wheels is referred to in the next section. To sum up, it may be said that evidence is accumu- lating pointing to the importance of the direction of the grain in forgings. The direction of the grain is fully re- vealed by the macros tructure. The extent to which the direction of the grain affects the properties of the steel may be shown by the falling off of the elongation, re- duction of area, and notched bar tests in directions inclined to the grain. The notched bar test is usually the most convenient for this purpose. 6. FORGING GEAR BLANKS. There are two well-recognised methods of manufac- turing drop-forgings for gear blanks. By the first method, or " flat-bar method," a number of blanks are manufactured from one length of bar, the bar being cut to a multiple length, and not to the length required for a single blank. The bar is placed between the dies with its axis running in a direction perpendicular to the di- rection of the blow of the hammer. It is consequently splayed out in one direction, perpendicular to the axis of the bar, but in the other dimension it is relatively unaffected. The form of the steel before and after drop- forging is shown in Fig. 27. By the second method, or " upending method," each blank is made from a piece of the original bar cut to dead length. This length is put in the dies with its axis parallel to the direction of the blow of the hammer, and the steel is forced out- wards equally in all radial directions, so that all similar parts of the gear blank are subjected to approximately the same amount of work. The form of the bar before and after drop-forging is shown ir. Fig. 28. FIG. 28. Up-ending method of forging gear blanks. The gear blanks made by the two methods possess very different structures so far as the direction of the grain of the steel in the blanks is concerned. The macrostructure of blanks made by the two methods is shown in Figs. 29 to 32. Figs. 29 and 30 represent sections of a wheel blank made by the first method, Fig. 29 being a section in the plane of the wheel, and Fig. 30 a section perpen- dicular to it along the original axis of the bar, i.e., along the diameter A.B. Figs. 31 and 32 are corresponding sections prepared from a blank made by the second method. The structure shown in Fig. 32 would be shown in a section cut along any diameter, but the structure shown in Fig. 30 is only shown in sections cut approxi- mately along A.B. Comparing these results it will be seen at once that the direction of the grain in the teeth in the resulting gears cut from the two blanks will be very different. In the gears made by the second method the arrangement of the grain in each tooth, from what- ever position in the gear it is taken, will be the same, and will run in a curved direction approximately as Fia. 27. Flat bar method of forgiug gear blanks. Si' At FIG. 33. Grain in teeth of wheel made by the " up-ending " method. shown in Fig. 33. In the gears made from blanks manu- factured by the first method the arrangement of the grain in the teeth will be different according to the position of the tooth in relation to the original axis of the bar. A tooth cut from position C. or D (see Fig. 29) will have the grain running perpendicular to the pitch line of the tooth, as shown diagrammaticaJly in Fig. 34. On the other hand, a tooth cut from position A. or B. will have the grain running in a direction parallel to the pitch line of the tooth, as shown in Fig. 35. The difference in the physical properties of the teeth CHAPTER III.] [To face page 28. >T*^S" -**-i4~rj-*-?* >S|g^jpS f'^'^.'i^f ;*-> ";- FIG. 25. Macro-structure of valve turned from bar. 3Sr~ _ FIG. 23. Macro-structure of twisted crankshaft journal. J FIG. 24. Twisted journal showing discontinuity. FIG. 26. Macro-structure of forged valve. To face page 29.] [CHAPTER_III. FlG. 30. Diametral section (along diameter A B, Fig. 29) of a gear blank made by the flat-bar method. FIG. 29. Section in the plane of a gear blank made by the "Flat bar method.' Diameter A B is along the length of the original bar. FIG. 32. Diametral section throngh gear blank made by the " up-ending " method. Fro. 81. Section in the plane of a g *- ,.^' ' K) a: I/} -J >o LU S x: h- UJ 4 LU J de pe ^ ^ LU LU \ ^9 > V \j h- r CO j * -.%! 15 u u h 2 y ID s "10 ? <0 u. 29 of the two gears (or of the teeth in different positions in the blank illustrated in Fig. 29) may be shown by the difference of the elongation, reduction of area or Izod values of the steel in the teeth. When a gear is at work the direction of the stress applied to a tooth is perpen- dicular to the radius of the wheel which passes through the tooth. The root of the tooth on the bearing face may be considered to be the part at which the stress is concentrated, and this line in one case cuts across all the fibres of the grain and in the other runs parallel to them. The effect of the direction of the fibre on the Izod tests is shown bv the following tests made on teeth shows the minimum notched bar test possible for forged steel as is shown by some of the teeth in gears made by the first method. 7. CASKHAKDKNKU I'AUTS. Information on casehafdening and casehardeniug mixtures was issued as Airboard Specification H.5. The last edition of this specification is printed in Appendix I. A steel part which has been easel lardened is no longer homogeneous. The outside layer or case consists of a very different material from the inside core, and as a FIG. 34. Grain in reeth of wheel at the points C and D. Fijf. 29, made by the " flat-bar " n ethod. Ill III 1 / III 1 Ijll Hlliii'iU. i' \ FIG. 35. Grain in teeth of wheel at the points A and B, Fig. 29. made by the " flat-bar " method. in different positions in a wheel made from a blank forged by the first method : Direction of Grain. I nod Results. Perpendicular to radius of wheel (Fig 8) 15 foot BJ. ,, lt>'5 Parallel to radius of wheel (Fig. 9) ... 52 ... 54 In gears made by the second method, and having the arrangement of the grain as shown in Fig. 80, the resis- tance to impact should not be affected by the position on the wheel from which the tooth to be tested is cut. That this view is correct is shown by the following results obtained on similar samples cut from different positions in a gear made by the second method : Izod Result*. 25 "foot Ib. 1 27 28 29 33.5 ,. In gears made by the second method there are no teeth in which the arrangement of the grain in the teeth is in a direction perpendicular to the radius of the gear, ;vnd consequently there is no position in which a tooth consequence the strength of the part cannot be calcu- lated by the ordinary theorems, which are only true for homogeneous materials. Nor is it possible to modify the theory in any simple manner, because of the effect that the hard case will have in preventing the plastic yield of the core, which introduces a factor of unknown magni- tude. The support given to the core will depend on the shape of the part, and on whether it is casehardened all round. For example, the conditions in a thin disc-cam hardened on the edge, but not on the faces, is very dif- ferent from that in a round rod hardened all over, and experience shows that the core of such a cam distorts freely under pressure en the casehardened rim the soft core bulging out on either side, whereas in the rod the core cannot yield till the case cracks. Many parts are only casehardened locally, and the question of their strength is of little importance, but for such parts as gudgeon pins the strength is of great im- portance. The question of strength arises in two forms: How should the strength of a casehardened part be cal- culated ? and what type of casehardened steel will have the greatest strength ? The properties of the case do not differ very much for different steels, but the core, which is little affected by the casehardening process, has very different pro- 30 perties, depending on the type of steel used. Owing to lack of accurate information many engineers prefer the soft ductile core given by a niild carbon casehardeued steel, while others claim that there is a great gain in strength when nickel or nickel-chrome casehardeued steels are used, owing to the greater strength of the core. The series of tests described below were made to in- vestigate these points. Owing to several unforeseen difficulties the results were not as satisfactory as they might have been, and further investigations are needed, but some valuable results were obtained: (i) The beam tests, which may be taken to represent the conditions in a gudgeon pin, show that the stronger (alloy) steels stand about 50% higher load before the case cracks than the carbon steel ; the first failure in all casehardened beams is the cracking of the case. (ii) The fatigue tests (Woehler) of the cores show that the fatigue ranges of the stronger (alloy) steels were a little more than 50% higher than those of the carbon steel. (iii) The strengths of the cases were roughly equal to the strength of a 0'9% carbon steel. This does not imply that the hardnesses of the different cases were equal. Fatigue tests are in hand on casehardened bars, and till the results are available a complete comparison can- not be made, but the above results are sufficient to show the considerable advantages possessed by the alloyed casehardening steels. Tests on Casehardening Steels The following steels were tested : (a) Steel to Specification S. 13 (b) S. 14 (c) S. 15 (d) S. 16 (e) , S. 17 (f) S. 17 (g) 0'9% plain carbon steel. Two samples of S. 17 steel were used in order to test both types of steel in general use complying with that specification, one being higher in nickel and lower in carbon than the other. The sample (g) was tested in order to show approximately the properties which might be expected from the case of the steels. The composition of the different steels were as follows : Mark. sg 11 Carbon. Silicon. > o> n C Sulphur. Phos- phorus. Nickel. Chro- mium. ,.v S. 13 0-18 0-12 0-46 0-032 u-023 (&) 8.14 0-19 0-05 0-65 0-049 0-025 00 S. 15 0-19 0-32 0-45 0-028 0-021 3-00 0-48 If) S. 16 0-20 0-28 0-39 0-033 0-022 3-15 0-58 (<0 S. 17 0-16 0-11 0-43 0'044 0-033 4-65 0-29 GO S. 17 0-11 0-20 5-76 o/> ~"~ 0-91 0-17 0-27 0-027 0-030 ~ The following tests were made : (i) Tension tests upon the core of all the steels (a) to (f). (ii) Tension test upon (g). (iii) Tension tests upon the case of all the steels (.) to (f). (iv) Fatigue tests (Woehler) upon the core of all the steels (a) to (f). (v) Beam tests upon a casehardened specimen of each steel. The specimens were all chosen so that the part ac- tually under test should be heat treated when 1| inch diameter. For the tension tests i, ii and iii bars of the dimensions shown in Fig. 36 were used. For tests iii the cores of the specimens were bored out after heat treatment. For test iv the bars were plain 1^ inch diameter rods inches long. For test v the specimen shown in Fig. 37 was used. The specimens which were to be casehardened were carburised by Mr. David Flather, after machining to approximately the dimensions given above. Those which were not to be carburised were heated in a sarbuming box, but without carburising mixture, for the same length of time as those which were to be carburised. They were then all refined and hardened by quenching in water first from the upper, and then from the lower temperature given in the following table : Before each quenching the specimens were kept at the specified temperatures for 15 minutes. Steel. Refining Temp. C. Hardening Temp. C. 00 900 770 () 890 770 00 860 770 C*) 840 770 (O 830 760 (/) 820 760 GO "" 760 During the carburisiug operation a specimen of the steel was heated along with the test pieces, and the resulting carbon penetration was determined on these pieces by removing the case layer by layer and deter- mining the carbon content in it. The carbon content is plotted against depth of layer in Fig. 38. TENSION TESTS ON CORES OF (a) TO (/) AND UPON ( '" 'S o O (g S3 Mark. 'I- 2 & "8 aa o III HHJ m a l|f as a n O 3) Si 00 S. 13 17-25 24-1 40-3 25-1 58-4 S. 14 15-3 21-4 38-0 30-2 ii.V9 00 S. 15 18-0 40-0 06-2 19-0 3S-1 S. 16 16-0 84-4 12-8 80-1 00 S. 17 16-0 50-0 63-2 17-0 46-0 S. 17 20-0 29-2 51-7 23-9 lo-ii (i) 33-6 60-2 83-4 14-2 41-1 The low values of the elastic limit are good examples of the results commonly obtained with quenched steels which have not been tempered. TENSION TESTS ON C4SES OF STEEL () TO (c). m _g a J,-* e8 ^ e """ Ultimate Mark. B g.a J w 's ? 3l Stress. g-s- ** ~OQ~ H^ 43 !*<&< g 3 tous/cq. in. 00 0-013 0-0528 27-1 IS- 2 (Soft < note below) m 0-0185 0-0639 36-o tin MI (p. ) (<0 0-0161 0-0668 34-2 53-0 80-8 n-Olli d-0558 37-6 1)4-8 Apparently the presence of the alloying element, CHAPTER III.] [ Tufiuv page 30. M 553 Fia. 40. Casehardened beam tests. To face page 31.] [CHAPTKH III. t,. > , f l ' V ' ; , : "'' V '^iiiv FIG. 42. Cracks at weld. (The weld is horizontal.) FIG. 41. Cracks at weld. (The weld is horizontal.) FIG. 46. Cracked valve. FIG. 43. Cracks at weld. (The weld in horizontal.) 31 i.e., the nickel, has very little effect upon the strengtli of the case, all the results being very much alike. BEAM TESTS ON CASEHARDE \ T ED SAMPLES. (Se Fig. 40.) The specimens were loaded so that in the gauge length there was a uniform bending moment. 3 n f$ Max. Stress at Mark. Specifi- cation. Sd'l la-? ii j. J a *o *-* "* jit 5S-1 t CD a Deflec. in inches on 3" which Case cracked in S "MS 3 length. tension. 00 S. 13 16-5 81-50) 11,400 0-24 Not cracked. (.V S. H 20-1 71-8 12,100 0-20 64-4 00 00 S. 15 S. 16 39-2 34-5 94-3 95-5 12,700 13,000 0-06 0-028 92-8 95-5 CO S. 17 39-0 95-6 12,900 0-18 95-6 (') Not broken. Steel a. In this specimen it is probable that the case was not hardened properly. It did not crack until the modulus of rupture was reached. The speciman, as a whole, was quite ductile. The elastic limit of the beam is practically identical wibh that of the core. Steel b. In this specimen the case cracked in ten- sion before the core gave, and when once started the latter flowed freely. It was observed that the elastic limit was reached first on the compression side. Steel c. The core required increased load to produce fractures after the case had cracked in tension, and apparently failed in compression. The elastic limit was reached on the compression side before the tension side of the beam. Steel a and e. The elastic limit was reached on the compression side before the tension side. The beams had little ductility and immediately the case cracked the cores flowed freely without any increase of load. FATIGUE TESTS ON THE CORES. The determinations wore made by the Woehler method and gave the following results: Mark. Specifica- tion. Fatigue range, tons/sq. in. Fatigue Limit (') Fatigue Limit ('). Fatigue Limit (') Klaatic Limit. Yield Point. Ultimate Stress. 00 S. 13 -1- 18-4 1-06 0-77 0-47 C) S. 14 -t- 17-3 1-13 0-81 0-46 (e) S. 15 + 31-0 1-73 0-77 0-47 00 S. 16 -1- 29-r> 1-84 0-35 00 S. 17 -+- 30-0 1-88 0-60 48 to S. 17 -1- 26-5 1-33 0-91 0-52 Note ('). The Fatigue Limit is half the Fatigue Range. This table shows that there is no constant relation- ship between the fatigue range and the elastic limit or the yield point, but that with one exception (this appears to be a real exception in so far as the ultimate strength of the sample was concerned), there is a fairly constant relationship between the fatigue range and the ultimate strength of the steels. 8. WELDING. The material which is usually welded in aircraft is ordinary mild steel, generally to Specification S 3, which is employed for minor structural parts of aeroplanes, and is also used to some extent for cowls and exhaust manifolds in engines. Welding on this material does not present any particular features of interest; it is usually done with an oxy-acetylene flame, and suitable filling material is employed. The parts should be nor- malised after the welding has been carried out, and by this means the ill-effects of the welding, if any, are easily removed. Recently, a considerable amount of welding has been used on other parts of the aeroplane, namely in the manufacture of steel spars and struts, which are usually made of steel of much greater strength than is called for by Specification S 3. For these parts the steel is re- quired to have a strength, varying according to the design, from about 65 to 70 tons per square inch up to 85 tons per square inch. Obviously it is not possi- ble to emproy plain low carbon steel, and the types of material which have been used have been, firstly, an alloy steel containing about O20% of carbon, about 3-5% of nickel, and about 1'3% of chromium; and secondly, a plain carbon steel containing about 0'45 to 0-55% of carbon which is cold rolled and blued to give the necessary strength. The first of these materials is an air hardening alloy steel, and it is used after suitable heat treatment. The welding of these steels is usually done not by means of an oxy-acetylene flame, but electrically, either by spot welding, or by roller welding, the heat for the welding being obtained by the resistance of the metal to the passage of an electric current between two electrodes pressing upon opposite sides of the two pieces of steel which are to be welded. It has been shown that moderately good results can be obtained by this process from the point of view of the strength of the weld, but it is found that the operation of welding lias a distinctly deleterious effect upon the metal. The effect of welding the alloy steel which is used in a heat treated condition may be considered first. Material of this kind does not weld very easily, and an examination of a large number of welds shows that in very few cases has a really good weld been produced. This defect might be overcome with further experience, but a more serious defect, due to the welding operation, is the production in the steel of a very large number 32 of cracks. During the welding the material is heated to a high temperature, and the cooling which follows is necessarily rapid since the volume of metal which is heated is only small, and it is in perfect metallic contact with a considerable volume of cold steel which rapidly conducts away the heat. The resulting structure in the material is similar to that of steel which has been quenched out in oil from a very high temperature. The microstructure is very coarse and triangular and similar to that associated with cast metal. The steel at the weld is hardened considerably, and the original heat treatment is entirely superseded. During cooling crack- ing necessarily ensues in a steel which hardens so intensively as this. An examination of a large number of welds has shown in practically every case the exist- ence of considerable cracks running into the body of the steel at right angles to the surfaces which have been welded. These cracks are not large in material which has not been stressed, but when any appreciable stress is applied they develop considerably. The presence of these cracks in stressed members will be dangerous, particularly in any parts which are subject to alternating stresses. Typical examples of these cracks are shown in Figs. 41, 42, and 43. The line of weld is horizontal in all the figures. Up to the present it has not been found possible to avoid these cracks. It appears probable that the only way in which they could be avoided would be by arrang- ing to cool the metal after welding much more slowly. The cracking is certainly due to the combined effect of the high temperature attained by the metal during welding and the rapid cooling which follows. It is pos- sible that the ill-effect of the welding upon the mechanical properties might be overcome to some extent by re-tempering the steel after the welding. This, how- ever, would only lessen the trouble, and would not over- come it altogether. The second material, namely the cold rolled and blued steel, does not suffer so seriously from the effect of the welding operation as the alloy steel. In this material, as in the other, the high temperature of welding produces a casting structure which is undesirable, but in general the welded material is free from cracks. The principal effect of the welding is to remove the effect of the cold working in the neighbourhood of the weld, because the temperature reached is well above the normalising tem- perature. The cooling is rapid, and the material is left more or less in a quenched and tempered condition after the operation is completed. To some extent this neutralises the loss of strength resulting from the destruction of the cold work effect, but it produces an unfortunate structure in the material. In a welded specimen there are three separate structures : (1) That of the original material. (2) That of the part completely affected by the welding, and which retains the chief high temperature effect. (3) The intermediate /.one which has not been at the full welding temperature, and therefore possesses an indeterminate structure. It is usually in this zone, at the junction of the first and second zones, where fractures occur when the welded steel is stressed. It is probable that electrical spot or roller welding can only be recommended under two sets of conditions. It may be satisfactory for parts which are not sub- jected to any appreciable stress in the welds, for example, in a stream line interplane strut in which the welding runs longitudinally, and where the welding only fixes the material in position. Secondly, it might be ap- proved when the whole part after welding can be normalised. It is only by this operation that a uniform structure can he restored throughout the material and the dangerous effects of high temperature completely eliminated. 9. VALVES. Valve Failures. In selecting steel to be used for valves in aeroplane engines it is important to take into consideration the nature of the work which the valves have to do, and the conditions under which they do it, Possibly the most instructive way of determining what steel is to be employed is to enumerate the ways in which valves fail, and from this endeavour to choose a steel which will overcome the various causes of failure. Aero engine valves may fail in any of the following six ways : (1) By distortion of the head of the valve, in consequence of which the valve does not close on to its seat properly, and therefore fails to perform its proper functions. (2) By elongation of the stem of the valve. (3) By wearing of the stem of the valve. This causes the valve to work loose in the guides, and allows too much play, with the result that it does not seat properly. (4) By the burning out of the head of the valve. (5) By the breaking of the head or the neck of the valve. (0) By excessive wearing at the end of the stem which comes in contact with the tappet or tappet roller. The actual conditions under which the valves work are not the same for all types of engine, those in a stationary engine are very different from those in a rotary, and many of the various probable causes of failure which occur in one engine do not occur in others; also there is a marked difference between the conditions under which an inlet and an exhaust valve work. As the cylinder is fixed in a stationary engine, the track of the valve head is simply backwards and for- wards along a straight line, and the, only sideways force on the stem is that produced by the cam and the tappet action. The exhaust in a stationary engine, however, is often fairly confined, and, speaking generally, the valves in a stationary engine have to work at a higher temperature than those in any other type. The radial engine has very much the same conditions as regards the sideways thrust on the valves, but in almost all cases the temperature conditions of the valves are less severe, since better cooling of the valves is generally obtained. The rotary engines, speaking generally, have very much more satisfactory cooling conditions, and, as 33 a result, the valves work usually at a lower temperature. Tlie exhaust from the rotary engines usually goes directly into the air, and hence is not concentrated by an exhaust pocket upon the stern or any other part of the valve. In these engines, however, there is a great deal of sideways wear upon the stem of the valves. It is important to know how a valve gets rid of its heat; there are three distinct ways by which it does so. The heat may be conducted from the head down the stem of the valve and be dissipated by that channel; it may be lost by direct radiation from the upper side of the head or from the stem; and finally, it may be con- ducted away through the seating. It is found that most of the heat is removed by the third method, that is, by direct conduction from the valve to the seating and thence through the cylinder head, a relatively small amount being carried down the stem. In some cases the position of the valve in the exhaust pocket is such that the waste gases come closely into contact with the valve stem, and actually heat it as hot or possibly hotter than the valve head itself; under these conditions the amount of heat carried down the stem from the valve head will be nil, and, in fact, heat may be given to the head rather than taken away from it. The various ways in which the valve fails may now he considered more fully. (1) The distortion of the head of the valve may be produced by several causes, and whatever the initial cause the effect is apt to be cumu- lative as the distortion, by allowing leakage, causes local heating. If the valve has not been heat treated in such a, way as to remove as far as possible all the stresses which result from forging or quenching, then the heating during running may give rise to considerable distortion. A valve forging should be normalised most carefully after forging, and if further heat treatment is adopted the method employed should be such as will leave the minimum possible residual stress in the metal. If the valve is so placed in the engine that the head is unequally heated, it may distort; this may sometimes be prevented by using a valve which rotates when it is working. Finally, if the steel softens during running, and the spring action is heavy, distortion of the head may be produced in a manner which resembles the closing of an umbrella. (2) The elongation of the stem of the valve takes place by the softening of the steel at high temperature; examples have been found in which the elongation has amounted to one- eighth inch. It is sometimes possible to get over the difficulty by using weaker springs and a steel which is stronger at high temperatures, but the proper cure is a modification of the design of the cylinder head so as to keep the valve stem cooler. The fault in the design may be in the shape of the exhaust pockets or in the design of the valve guides which do not conduct away the the heat passing down the stem efficiently, or, finally, in (lie dimension of the valve stem which is insufficient to bear its load. ?72fi4 The valve stem sometimes is reduced in size by excessive scaling till it stretches under its load. (3) The valves which fail in consequence of exces- sive wearing of the stem are usually those in rotary engines. Wear may be provided against by a suitable selection of steel. In general the stem of the valve does not become heated to a temperature higher than, say, 400 0., and there should be no serious weakening or softening in a suitable valve steel at that temperature. (4) The burning out of valves is generally due to defects in the running of the engine; probably the most common defect is pre- ignition. If ignition takes places before the valve is properly seated, the escape of gas, then at its maximum temperature, is very destructive Imperfect seating of the valve has much the same effect; it also interferes with the conduction of the heat from the valve head to the seating, and thus causes the valve to run too hot. Oxidation and scaling of the steel sometimes causes burning of the valves; the scale breaks off unevenly and causes defective seating of the valve with the results already mentioned. Scaling may be due to the quality of the steel. Finally, a crack in the valve head, even if very minute, is sufficient to initiate its destruc- tion; the crack allows a leak to start, and by admitting the hot gases into the metal appears to start a general destructive action. (See Fig. 46.) (5) The breaking of the head of the valve is almost invariably a steel trouble. The valve may have been left brittle owing to improper heat treatment before being put into the engine, or it may have been made from a steel which has hardened in cooling after running at a temperature higher than its critical range. This hardening induces a certain amount of brittleness, and when the valve is restarted cold or cool, fracture occurs. A fracture of the head may also occur in consequence of lack of strength at high temperatures, though this is unusual, as the elongations possessed by the steels at these temperatures are so high that they will draw out rather than break. (6) Excessive wearing at the top of the stem is usually a consequence of using too soft a steel. In many cases the valve stem itself is exposed to the blows of a tappet or to the continual wearing action of a roller. Unless the steel is sufficiently hard the end of the stem very quickly becomes hollowed out, and wears away rapidly. Steels for Valves At the beginning of the war there was hardly any information on the proper types of steel for valves. Each engine maker chose a named brand of steel almost at haphazard, and used it till trouble arose, when he tried another brand. As all brands of steel are subject to gradual changes, and sometimes to sudden considerable changes, great confusion and un- certainty arose. When the subject was first investi- gated by the Material Section it was found that for exhaust valves alone there were 23 different brands in use, the composition of very few of which were known to the users. The valve steels then in use may be roughly classified as follows: (1) High tungsten steel containing approximately: Tungsten 17%, and carbon - G5%. (2) Lower tungsten steel containing approximately: Tungsten 12%, and carbon 0-6%. (3) Low carbon tungsten steel containing carbon approximately 0'25 to - 3%, and tungsten about 12%. (4) Stainless steel. (5) 25% nickel steel. (6) 3% nickel steel, with carbon about 0-3%. (7) 3% nickel steel, with carbon about 0-6%. (8) 5% nickel case-hardening steel. (9) Nickel chromium steel of the type used for crankshafts. (10) Nickel chromium steel as used for air-harden ing gears. (11) Nickel chromium steel of a more mild air- hardening type. It was evident that in the above list of steels there must be several which had no properties to mark them as superior tc the others, and also that many olf the classes of steel could be combined in one without caus- ing any read hardship or drop of efficiency. It was also evidently desirable that the more suitable classes of steel should be picked out and made the basis for specifications, and that the others should be suppressed. The first thing done was to weed out those classes of steels which were only used in a very small number of engines, or in the unimportant engines, and which intrinsically had very little to recommend them. 25% nickel steel was not in very wide use, although it has been recommended on account of its supposed great resistance to corrosion. This property was found not to be Aery reliable, and as its other properties are poor this sleel was removed from the list. Two 3% nickel steels could not be considered to be necessary, and from the point of view of ease of forging the lower carbon one was distinctly preferable, whilst the mechanical properties obtained from it after satis- factory heat treatment were quite as good as those obtained from the higher carbon steel. It was stated that the amount of pitting which took place in the valves was much, less with 0-6% carbon than with O3% carbon, but th>'s statement did not bear a complete investigation, and the 0-6% carbon nickel steel was there- fore deleted. In the same class came the 5% nickel case-hardening steel There was, no reason why this steel should not make satisfactory valves, but no evidence was brought forward which suggested that it would make a better valve than a 3% nickel steel with 0'3% of carbon. This latter is cheaper and easier to make than a good 5% nickel casehardening steel, and there- fore the latter steel was also removed from the list. Of the various nickel chrome steels which were in use it was decided to retain two classes, both of which were air-hardening steels. The 100 ton air-hardening steel was retained, as it had apparently given consistently good results, and was in wide use. A mild air-hardening nickel chrome steel was also retained, as it also was found to be in very wide use, and, generally speaking, had given satisfaction. The stainless steel was very widely used, and. although it was rather soft, and therefore frequently wore badly in the stems, and also appeared to be rather weak at high temperatures, yet the great advantages of the steel made it evident that it should be retained. Of the tungsten steels, the third type mentioned was used to only a small extent, and was only manufactured in this country in very small quantities; it was there- fore deleted from the list. Of the other two the deci- sion lay between one containing 17% of tungsten and one containing 12% of tungsten. Of these the 17% tungsten had the better history, and it was decided that this was the one which should be standardised. In consequence of this weeding out it was possible to prepare suitable specifications for quite a reasonable number of valve steels, as follows : (a) Specification S. 19 for the stainless steel. (b) K. 8 for the 17% tungsten with 0-65% of carbon. (c) Specification K. 9 for the milder air-hardening nickel chrome steel. (ii) Specification K. 10 for the 3% nickel steel with 0.3% of carbon. (e] Specification K. 14 for the 100 ton air-harden- ing steel. In practice these steels were apportioned roughly as follows : Steels to specification K. 10 were used as inlet valves in all cases. Trials were made after a time with a view to the extension of the use of this steel for the exhaust valves of some of the cooler running engines, such as rotaries. The trials have been successful, and the steel is now in use in some engines. . Steel to specification K. 9 was used for the inlet valves of some of the stationary engines, and also for the exhaust valves of several rotary engines. This steel has given satisfaction in both these positions, but it has been found that it is a difficult steel to produce free from cracks, and it is probable that this type of steel will be discontinued in future standardisation. Steel to specification K. 14 was used principally for the exhaust valves on the rotary engines. It has been used also in some of the stationary engines for the exhaust valves with complete success, and it is probable that if this steel were tried out it would give quite satisfactory results for this purpose. The stainless steel is used for inlet and exhaust valves in many engines, and has given satisfactory results, except in those engines which run at exceed- ingly high temperatures. The tungsten steel is reserved for use in those engines in which the exhaust valve reaches the highest tem- peratures; it is never employed for inlet valves. 35 In order t<> determine whether or not this method of standardisation was satisfactory an investigation was made into the properties of the various valve steels at the temperatures at whidi they might be expected to run in the engines. The complete results of the tests made by the N.T.L., Professor Edwards and Mr. Brearly are given below. The results of the tensile tests at high temperatures are summarised in tho following table, and some of them shown in Fig. 47. ULT TENSILE STRENGTH TONIS/D" 25 20 IS 10 5 TENSILE TESTS OF VALVE STEELS AT DIFFERENT TEMPERATURES. 1 \ N \ I ^ N s. X x^ * i ""^ ^ ^ JlK^ *; - ^ ^^-^ *"-N/*-* 65O TOO 75O 800 85O 900 9SO IOOO TEMPERATURE. *C. lia. 47. ULTIMATE STRENGTHS OF VALVE STEELS AT HIGH TEMPERATURES. 650 700 750 800 85(l c 900 950" Temp. Centigrade. 15 . Ultimate Strength. Tonspersq.in A. 18% Tungsten 58 25 _ 13 10 11 9 7 B. 12% 50 20 10 7 10 8 C. Stainless 43 14 8 6 y 7-5 5 D. High Carbon Chromium 67 21 10 7-5 U-6 7-5 6-5 E. High Silicon 1% \ Chromium Steel/ 67 21 7-3 F. Cobaltchrome 54 21 10 11 9 6-5 6. 3 __ 4-2 J. 100 ton Nickel Chrome 98 13 8 5-4 4-6 3-4 K. Ordinary Nickel"! Chromium / 55 16 8 7 5'4 4-6 As a result of these tests it became clear that the standardisation which bad been effected was right in the main ; the following conclusions may be drawn from the experiments : (1) The 17% tungsten steel is stronger at high temperatures than any of the other steels. (2) If the carbon or the tungsten in the tungsten steels are lowered, then there is a distinct falling off in strength. The carbon percentage is, however, considered to be of much 27264 greater importance than the tungsten, and the reduction of 0'2% of carbon produces a greater loss of strength than does the reduc- tion of 5% in the tungsten content. (3) The stainless steel is much weaker than the tungsten steel, but it is possible to increase the strength of this steel at high tempera- tures considerably by increasing its carbon content. This affects its stainless properties and in fact the steel ceases to be stainless and scales much more at high temperatures than the true stainless steel. It is probable, however, that a steel which has the same percentage of chromium as the stainless steel, but a higher percentage of carbon, could be usefully standardised and used as an alternative to the stainless steel. (4) All the nickel chromium steels which have been tested have exactly the same strength at high temperatures. It appears likely, there- fore, that there is no gain in standardising more than one nickel chromium steel, and it is probable that for most conditions the best would be one which does not air harden inten- sively, but is of the oil hardening and tem- pering type. Such a steel would have advantages in manufacture and in forging, whilst there would be little likelihood of its hardening up in the engine and becoming brittle. (5) Plain nickel steel is of approximately the same strength at high temperatures as the nickel- ehromium steels, and it is probable that such steel might conveniently be used instead of the nickel chromium steel in all cases. So far as the general trend of the research work which has been done can be summed up, it indicates that the only steels which need to be specified in future are: (1) Higli tungsten steel for use in the hottest engines. (2) High chromium steel containing either a " stainless " proportion of carbon or a much high proportion. This steel could be used in the medium hot engines as its machining properties are very much superior to the high tungsten steel. (3) Plain nickel steels which will serve generally for the inlet valves of all engines and pro- bably for the exhaust valves of rotary engines. Tests of Valve Steels at High Temperatures. The steels tested fall naturally into three groups (1) The tungsten steel group corresponding roughly to the class of steel called for by specification K.8. These include not only those which fall definitely within the limits of composition called for by specification, but som } which do not, the object being 'to determine the influ- ence of variations in carbon and in tungsten content-- upon the properties of the material, and also the effect of vanadium. E 2 (2) The chromium steel group oa.scd originally upon steel to specification S. 19, i.e., stainless steel. In this group tests have been made not only upon steel falling within specification S. 19 but upon steels containing high percentages of carbon, much lower percentages of L-hromium, and varying percentages of silicon and nickel, the object being to determine what limits of composi- tion could usefully be placed in the specification, and to determine the effect of the variations of the different constituents upo i the mechanical properties of the ma- terial at high temperatures. (3) The ordinary engine steels of the 3% nickel class, with varying percentages of chromium. In this group tests have been made to show the influence of heat treatment upon the results. In addition to ordinary tensile tests at high tempera- tures made by the N.P.L., a certain number of other tests have been carried out; these include measure- ments of the Brinell hardness of the steel at high temperatures made by Prof. Edwards, using his dynamic method of testing,* and a series of notched bar tests at high temperatures were made by Mr. H. Brearley. using a Charpz machine and test piece but an Izod notch. (Test piece 10 mm. x 10 mm. Notch 2 mm. deep, 45. Radius 0-25 mm.) TUNGSTEN STEKL GROUP. Chemical Composition. Tensile Tests at High Temperature* (N.P.L.) P Steel. ' | d I si 1 ' % Jl s is 5 AS frj il O 33 s & 1 *' i w a H \ 0-71 0-11 0-05 0-079 0-01 3 3-86 17-30 0-75 2 0-43 0-34 0-37 0-055 3-19 12-10 3 0-60 0-10 0-08 0-073 Trace 3-64 17-44 1-00 4 0-60 0-095i 0-09 0-041 , 3-?6 16-37 0-06 3 0-67 0-10 0-09 0-050 , 3-70 13-56 o-io 6 0-45 0-094 0-07 0-037 , 3-62 13-04 Nil. 7 0-45 0-095 0-08 0-037 3-75 16-73 Nil. 8 0-47 0-14 0-07 0-064 ' , 3-62 12-68 0-80 Heat Treatment. All the steels were refined at 950 C., and tempered at 800 C. Tensile Tests a; ordinary Temperature Steel. Yield Point, Ult. Strength, Elongation, Red. of Area, tons/sq. in. tons/sq. in. per cent. per cent. 1 3,9-4 58-4 . 19-0 19-0 2 ' 19-0 52-7 19-0 19-0 3 40-2 48-6 18-0 30-8 4 32-1 52-5 18-0 24'6 5 44-6 50-9 24-0 38-6 6 39-9 47-8 25'0 45-3 7 39-0 44-6 30-0 59-3 8 39-6 1 4.V5 29-0 49-4 Ultimate strength at various temperatures (tons per square inch). Steel. 600 700 750 si 10' 860 900 950 I 24-7 13-2 10-:! 11-1 8-8 (i-6 2 20-1 12-6 9-0 lO-'J y-i 7-4 3 17-7 8-3 4 17MI . 8-95 5 15-9 7-65 6 14 -7r, . 7-45 7 ],V5 6-3 * 16-85 ~ 7-25 ; These tests were made by the dynamic method de- scribed by Professor Edwards in his paper published in the Proceedings of the Institute of Mechanical En- gineers in 1918. The Brinell Hardness numbers are obtained from the diameter of the impression produced by a blow of 63 inch Ibs. by the formula H = - ^ where H = Brinell Hardness number, and d = diameter of im- pression in mm., vide page 362 loc. cit. It will be noted that these Brinell tests at high temperatures are not proportional to the tensile tests at high tempera- tures. At high temperatures the rate of applying the load affects both tests. Brinell Tests at High Temperatures (by Professor Edwards). Fig. 48. Brinell Hardness Number at various Temperatures. nteel. 200 300 400 600" 700 750 800 8. r -0 3 193 170 II '.I 131 121 1111 loi 4 238 217 K06 193 101 184 111 5 219 lt>6 162 158 143 118 1011 6 807 1 111 129 123 li:, 10!) 11.-. 7 200. 148 121 110 107 118 112 8 209 156 128 H8 111 in? 118 Notched Bar Tests at High Temperatures (Mr. H. Brearley). The tests were made in ;i Oliarpy machine, using a test piece of the Charpy size, but having an Izod notch. (Fig. 49.) Pro-. lost. Meoh. Eng., 1918, pp. 335-369, and Journal lust, of Metals, No. 2, 1918, VoL XX. Temp. Cent. 18 100 :2oo 300 400 500 600 700 800 900 1,000 Steel No. Kilogram-metres. 84 1 95 3-6 3-0 3-2 3-d 8-Ofl 2-67 :;<; 3-9 5 1 / -88 \ 1-27 1-4 2-1 3-2 2-9 3 ' 11 3-1 3-3 2-9 r.MI 7-8 . / 1-05 'Ml 2-9 2-8 3-2 2-9 3-4 3-3 4-2 7'5 8-8 ,\l 3-23 B * 3-05 7-4 8-1 9-1 8-6 8-2 8 9 8-8 16-0 15-8 IB ' 9 1 1 .18 1-27 2-58 4-3 SMI 9-0 9-3 8-0 7-5 8-0 9vl 9-5 8-1 8 2-50 .VII 5-5 5-8 7-1 5-2 fi'3 i ' .) 8-6| 9-0 9 :. Chapter III. 5 f\ y w .J CL il LJ LJ fc ERATU CL VALVE LJ i Z i LJ fc * O ^_ Z i/) Id H h I c: O 03 z D u I o h Z ~ $-. S) 3 K -S6 Sf;ifTfs" cu IS'-s u 6 o 6 =6 h ~ c" fi > O ^ i* 10 ~ - 6 6 no i * ID * 5!S - fi c N S ~ 01 - -66 9 o A 00-0 I T 9S? - - o o 66 2o !c *0 K >O (j A 10 ~ 6 O b 6b5o g z SJ oz " fl O J oil U (C O o <3< o Z7S.n/io 500 6OO 7OO C 800 9OO IOO( TEMPERATURES DEGREES cEHTiCF?*Dei. Chapter. III. -* I U OL z o CU 10 LL U O I a: < a a u o o 37 CHROMIUM STKKI.S (luoujf. Chemical Composition. Tcimilc Tents at LL'ujli Temperatures (N.l'.Jj.) Fig. 50. Steel. g 2 a 65u 700 750 800 850 '.'00 920 950 1,000 C Steel. 3 p (3 L* 3 5 3 O Si j Q. -5 1 A ^ P. 1 1 o 9 18-6 7-0 6-5 8*6 6-5 6-7 5-2 o CG a 1 "a CO -^ Jj o 1U 21-5 10-0 7-5 9-6 7-5 6-5 11 21-0 10-5 7-0 9-6 8-8 6-5 12 14-0 7-35 _ ., _ 4- 1 B 0-37 0-25 0-31 0-067 0-030 0-15 12-37 13 15-15 8 ' 55 7-4 1-05 1 11 13-2 7-95 7-5 3 ' 85 HI I -04 0-15 0-28 10-42 15 13-35 _ 7-9 7-65 8-98 16 12-08 6*64 4-8 4-0* 11 1*11 0-85 0-21 0-23 ll-o:!l 0-50 11-35 3-110 17 17-0 7-05 IS _^ 14-9 , 7-25 12 0-96 0-17 0-33 0-104 0-034 o-l 5 13-1 19 17-75 7-6 20 18-45 _ , . 7*5 13 1-08 0-17 0-34 0-106 0-030 0-50 13-1 21 21-0 7-35 _ 2-i 18-18 5-6 14 1-18 0-1(1 0-29 0-106 0-024 0-45 13-1 18 1-42 ' 35 0-35 0-101 0-030 0-44 13-1 * Temperature of specimen 970 C. 16 36 0-lti 0-25 0-060 0-025 0-23 11-2 17 0-54 0-14 (1-15 n-071 0-027 0-43 6-3 _ Complete Tests of Stainless Steel No. 1(3 is ' 55 0-75 0-13 O-'OW 0-025 0-50 7>1 (see Fig. 51), Mr. H. Brearley. 111 0-56 0-17 0'15 0-074 (1-023 3*08 Temperature. Y.P. Ult. E. <) (, It. A. % 20 l-oii 0-19 0-18 0-065 0-020 0-42 6-3 21 1-08 0-56 0-23 0-068 0-025 0-51 6-7 18 38-8 48*48 27 59-3 100 37-0 48-52 21 54 (> 22 1-11 0-19 0-17 0-074 0-1123 2 -SIC, 5-S 200 37-0 44-04 21 55-8 30(1 33-44 39 56 18 5 57-0 400 83-6 39-87 17 5 88-6 450 31-32 34-92 18 5 63-7 Heal Treatment. 500 550 28-68 22-92 29*72 ' 23-6 22 25 71-7 76-8 9 ... Air-hardened from / ii .O 880 C. tempered at 700 0. 700 a 20-8 10-84 21-24 12-08 25 40 75-9 90-9 10 ... n 880 C. 700 C. 800 5-08 6*64 40 5 91-9 11 ... .. 11 } ,000 C. 850 " 8( C> 900 4-68 3-8 6*64 4-8 40 41 5 92-4 65-8 12 to 16 .. 11 ,. 900 C. 700 C. J7(J ] 96 4-0 31 5 52-2 17 to 22 .. Oil-hardened from 820 C. .. 700 C. Notched bar tetts at high temperatures (Mr. II Brearley). The tests were made in a Charpy machine using a test-piece of the Charpy size, but having an Izod notch. The results are given in kilogram-metres. Fig. 52. TEMPERATURES. Steel. 18. 100. 150. 200. 300. 350. 400. 450. 500. 550. 600. 700. 750. 776. 800. 825. 850. 900. 950. 1001/ 12 1-27 2-41 b-52 3-23 3-87 3-42 3-52 3-62 4-02 4-07 4-02 4-02 5-35 5-93 7-34 8-86 13 1-12 1-43 2-76 2-45 2-95 3-04 2-90 2-81 2-95 3-33 2-85 3-19 3-46 4-02 5-25 6 7'J 14 0-98 1 73 2-23 2-45 2-49 2-02 2-27 2-06 2-19 2-14 2-49 2-27 3-47 3-37 6-03 6-Ofl 15 0-71 0-87 1-34 1-50 1-73 1-50 1-77 1-86 1-73 1-58 1-86 2-07 2-62 3-09 4-17 5-83 16 ti'SO 6-92 7-28 6-92 7-&0 8-60 8-43 7-40 7-40 7-40 6-92 6-56 7-28 8-50 8-68 S-45 9-11 14-00 14-82 1 4 23 3% NICKEL STEEL GROUP. Chemical Composition. d a h i ^ a ~ ~ Steel. 2 o 27,600 50,300 :, 2-097 1 732 2 ' 992 14,600 63,400 48,100 68,000 (1 1-555 1-102 3-766 78,500 KKS.OOO 80,600 110,1)00 7 2-2.-> 1-875 i 25 5S 4IK) 69,600 55,500 86,800 It will be seen from this table that the actual stress upon the spring when the valve is open varies, in ths specimens selected, from about 20 tons to about 50 tons per square inch, and when the valve is closed from about 10 tons to about 35 tons, and the usual range of variation of stress is about 15 tons per square inch. The springs therefore are exposed to a fluctuating torsional stress varying over a range of about 15 tons per square inch, the number of alterations varying from 050 to 1,200 per minute according to the type of engine. A certain number of tension and torsion tests have been made to determine the actual condition of the steel in springs which have been put into engines. Owing to the minute size of the specimen which could be cut out of the springs, the only information which it ha^ been possible to obtain lias been the ultimate strengths in tension and in torsion. For the torsion test two values are given, one corresponding to a linear distribu- tion of stress and the other to a uniform distribution across the section (between these two there is a constant ratio of 4 :8). TP:STS ON SAMPLES CUT OUT OF SPRINGS. Ultimate Strength Torsion Spring. Ult. Strength Tension. Tons / sq. in. Tons per sq. in. 1 2 8 95-5 91 68-2 93-2 97-8 73-3 84 63 9 104-7 108 81 j 104-3 94 -ft 95 -B 70-8 71-7 10 81-3 79-8 59-8 82-2 69-8 52-3 74-5 55-8 11 109-8 89-0 6C.-7 107-3 83-2 62-4 91-6 68-7 12 117 88-6 66-5 116-8 86-7 66-5 81-8 61-3 js 88-5 86-2 64-7 '.17-0 89-5 67-2 These tests show that the materials actually used in springs have ultimate tensile strengths of round about 95 tons per square inch. In order to determine the effect of different heat treatments upon the strength of spring steels, a con- siderable number of experiments have been carried out. The steels examined were as follows : (14) Plain carbon steel about 0'9% carbon. (15) A plain carbon steel about 1'2% carbon. (16) Chrome vanadium steel as described above. (17) Plain carbon steel about 0'5% carbon. No. 17 (-5 carbon steel) was included because the springs found in the engine of a captured Gotha aero- plane were of this composition. All the specimens were in the form of spring wire be- tween 14 and 10 gauge and wore heat-treated in that size. They were quenched, half of them at 800 C., and half of them at 850 C., and then tempered at 400 C., 500 C'., 550 C., and 600 C., and tested in tension and torsion at atmospheric temperature. There is very little difference in the high carbon steels between the speci- mens quenched at 800 C. and at 850 C., but in the low carbon steels (i.e., 16 and 17), there is a distinct difference, quenching at 800 C. not appearing to harden these steels completely. Selecting the most suitable quenching temperature for the steels, the following table shows the effect upon the tension and the torsion tests of variations in tempering temperature. It is evident that within the range of temperature covered there may be a large variation in the physical properties of the spring steels. Steel. Tempering Temp. Centigrade. Ult. Tensile Strength. Tons per sq. in. Ult. Torsional Strength.* Tons per sq. in. 14 400 107-0 92.8 ,, 500 95.6 75-6 it 15 600 400 72-5 112-5 62-3 95-8 R 500 Hi in W-8 60-7 74-0 66-0 If) 400 100-0 74-4 It H 17 500 600 400 83-6 68-0 102-4 63-9 59-4 86-2 500 81-0 68-6 it 550 66-6 61-6 * Assuming linear distribution of stresses. Springs on aero engines have to work at temperatures distinctly higher than the temperature of the air. The actual temperature attained by the springs varies con- siderably in different engines and depends upon the design of the cylinder head. Observations were made upon several different types of engines to determine the temperature attained by the springs; the maximum tem- perature observed of any of the springs was about 440 C. In view of the possibility of springs being worked at tem- peratures of this order it was thought desirable that tests should be made on the actual properties of spring steels at temperatures up to 450 C. Tests were accordingly made on the steels numbers 14 to 17 described above, and in addition on two other steels numbered 18 and 19 of the following compositions : (18) Nickel chrome steel conforming in composition to specification 2S11, and heat treated to give the physical properties required by that specification. (19) Chrome vanadium steel of the composition given above, but heat-treated to have the same approximate physical properties as those called for by specification 2S11. Numbers 14 to 17 were tested in the form of wires as before, numbers 18 and 19 in half-inch rods. The results of the tests are set out in the following two tables, the first of which gives the results of a restricted series of tests upon steels 14, 15, 17, 18, and 19. whilst the second gives the results of a more complete series upon steel number 17. (See Fig. 58.) 40 Strengths of Spring Stcfh tested at High Temperatures. "S 43 as Temperature of Specimen . 1 3 o during Test. Centigrade. 1 11! 1 *> C-g 15 200 3(M 400 450 H H Tons per square inch. 14 800C sooC Elastic limit .. I V 47 67 Ult. Strength 99 1 8B-5 73-2 14 850 :>IKI Elastic limit 88 4(> BIS 18 Ult. Strength '.>'2 fi 85-(; 8-7 50-8 16 800 500 Elastic limit 7(1 37 20 10 Ult. Strength !)l-(i 78-8 59 -4 49-6 15 850 50(1 Elastic limit M 88 24 17 Ult. Strength 98 89 62 -5 4S 17 850 500 Elastic limit 47-1 39-7 22-0 8-2 Ult. Strength 81-9 76-2 R2-G 39-0 18 830 620 Elastic limit 51 8j6 34 15 Ult. Strength 69 S3 -4 88 41-7 19 830 f,30 Elastic limit 44 3D 20 23 68-8 S3 SI -9 49-3 STEEL No. 17. B jj m Temperature of Specimen during Test. 5 O Centigrade. o S g &* 1 100 200 230 260 300 380 460 ^_ 3 S o -2 r Tons per square inch. 850C 400C Elastic limit 51-0 40-8 38-0 35-3 28-2 14-0 7-4 Ult. Strength 98-7 98- fi 100-9 99-8 89-0 59-7 42-9 850 500 Elastic limit 49-8 47-1 43-4 42-7 39-7 22-0 8-2 Ult. Strength 80-6 81-9 84-7 79-0 76-2 62-6 39-0 850 550 Elastic limit 45-0 38-3 37'0 37'fi 36-4 22-3 6-1 Ult. Strength 66-1 68-5 70-8 74-3 72-8 55-0 31-7 The following conclusions may be drawn from these experiments : (1) That there is a marked falling off, both in the elastic limit and in the ultimate strength of the steels, as the temperature rises. The falling off in the elastic limit is the greater. (2) That the falling off of the elastic limit and ulti- mate strength in alloy steels is distinctly less than in the carbon steels; the ratio of the elastic limit or ulti- mate strength at a higli temperature to that at normal temperature is nearly twice as great in the alloy steels as in the carbon steels. This may possibly be due to the fact that the alloy steels wer'e tempered at a rather higher temperature than the carbon steels before being put under the test. No fatigue test at high temperature have yet to be made. Till data are available it may perhaps be as- sumed that the fatigue limit (\ fatigue range) is between 45 and '50 of the ultimate strength. (See Section '2, p. 19.) 11. SHEET STEEL. There are three Air Board specifications for sheet steel, S. 20 for tinned sheets, S. 3 for mild steel sheets for welding, and S. 4 for higher tensile steel sheets which are to be heat treated. The tinned sheets specified in S.20 require no description ; the material called for in specification S. 3 is of greater importance and calls for some comment, whilst the material to the third speci- fications (S. 4) is relatively new, and has been the subject of a considerable amount of experimental work. Specification S. 3. Mild steel sheet for welding in accordance with this specification is used for a very large number of aeroplane parts, the greater proportion of which are made as cold pressings. The specification calls for an ultimate strenght of not less than 26 tons per square inch, and a yield point of not less than 18 tons per square inch, and states that these tests shall be obtained from material which has been normalised. Ordinary dead mild steel will not give these tests in the normalised condition; the material must contain about 0'20% of carbon. The steel has also to be capable of welding satisfactorily, which imposes an upper limit upon the amount of carbon; probably it would be quite safe to allow up to 0-30% of carbon, but this figure should not be exceeded. During the war a very large quantity of sheet steel has been supplied to this specification containing approximately 3% of nickel with about 0-12% of carbon. As it is difficult to avoid the mixing of these two types of sheet in the shops it is important to ascertain whether or not the same heat treatment (i.e., normalising), if given to the two different types of sheet, would give equally satisfactory results in the final pressing. A considerable number of experiments on this point were carried out in conjunction with the A.I.D. and the N.P.L. for the British Engineering Standards Association, and it was shown that the same heat treatment would give satisfactory results for both steels. The tests in- cluded tensile strengths, but dealt more particularly with the bend tests which the material would endure. The following table gives a summary orf the results : g 3 ( a u 5 SS.2 S .d.s its' 5 23 &! Direction of test piece. Reverse bends. Close bends on strip. I! a - 0. _. 4J 13 d |* S* 10 17-7 29-2 Longitudinal H O.K. 0-19 Nil. 19-0 30-1 Transverse... B S.C. 0-19 Nil. 10 20-5 31-4 Longitudinal 8 O.K. 0-22 Nil. 20-0 32-2 Transverse... 8 O.K. 0-22 Nil. 10 22-2 31-7 Longitudinal -li S.C. 0-12 3-33 21-4 81-0 Transverse... 7 O.K. 0-12 3-33 22 16-1 30-6 Longitudinal BJ O.K. 0-24 Nil. 17-6 34-8 Transverse... 7 O.K. 0-24 Nil. 22 18-6 29-0 Longitudinal 13 S.C. 0-17 Nil. 19-7 2fi-<> Transverse... 11 O.K. 0-17 Nil. 22 25-8 34-2 Longitudinal 9 O.K. 0-16 3-51 L 2 ^ 35-3 Transverse... 9 O.K. 0-16 3-51 S.C. = Sl-.'ht crack. O.K. = Passed the test. All the specimens had been normalised afresh after delivery from the makers. They were heated to a tem- perature of 845 C. in a salt bath furnace, and then cooled in air. The inclusion of normalising in the specification lias been much criticised, both by makers who wished to supply sheets in a different condition, and also by users who objected to the black surface produced by (lie normalising. It is possible to supply material which is not normalised, but will satisfy the other clauses in the specification. I'Y>r example, cold-rolled and close- annealed sheet containing about 0-32% of carbon, or indeed ordinary dead-mild cold-rolled sheet which has no heat treatment at all. Both these materials 41 have the advantage of a surface free from scale, and are therefore easy to press and do not wear out the dies. The objection to the use of these two materials is that the pressings made from them, after being normalised, are no longer up to the specified strengtk The normalising of the fittings is necessary, because they are seriously injured at the points at which they have been bent in the cold press, having been hard- ened by the cold work, and are left in a condition of considerable stress. By the operation of normalising the effect of the cold work is removed, and the stress is taken out of the material, and the weakness at the distorted part is removed. It is obvious that this treat- ment must be carried out, and that the material which is employed must give the strength specified after it has been normalised Ordinary cold-rolled mild steel sheet cannot therefore be used, because the normalised fittings will be below the specified strength. The same objec- tion does not hold in regard to the cold-rolled and close- annealed sheets containing rather more than - 3% of carbon, as this material would be quite up to strength after normalising, but this material is difficult to weld, and it certainly is desirable to avoid even the risk of making bad welds. It is generally supposed that nickel steel will not weld. It is true that nickel steel is more difficult to weld than ordinary carbon steel, but as the result of a large number of experiments it has been shown that it is quite practicable to weld 3% nickel steel containing only a low proportion of carbon, provided that suitable care is taken in the operation It has been found that the most satisfactory way of welding it is to employ as filling material strips of the same metal rather than the dead mild steel or iron wire ordinarily used. Specification S.4. This Specification calls for material which after suitable heat treatment will give an ultimate strength of 48 tons per square inch, and a yield point of 34 tons per square inch. To meet these tests an alloy steel must be employed, and it is not easy to produce an alloy steel which will at the same time give suitable bend tests. The only type of steel which has been found to be quite satisfactory contains approximately 5% of nickel with 0'20% of carbon. It is practically impossible to obtain material which will give quite such good bend tests as the normalised plain carbon sheets, but it is possible to approach closely to them. A large number of experiments have been carried out upon sheets of this kind varying in composition as regards both nickel and carbon. The heat treatment employed was varied and included considerable varia- tions in the quenching temperature and large variations in the tempering temperature. As a result it was found that the chemical composition must be kept within fairly narrow limits, namely, carbon O18 to O24%, and nickel 4-5 to 5'0%. With material within these limits of com- position it is possible to get quite satisfactory results if the steel is quenched at a temperature of not less than 830 C., and there is no gain in quenching from a tem- perature much higher than this unless the gauge of the steel is very heavy. The tempering temperature should vary somewhat with the gauge of the sheet, but in general a temperature of between 580 C. and 620 C. will give satisfactory results. In the following table are given a selection of the results obtained in tests showing the effects of different percentages of nickel and carbon inside and outside the selected range, and also showing the effect of different heat treatments. Gauge. 1! H isc , c Sd % Oi o fa* 8 S a tiS H H ^g 2-S o- (jj _H x C O tn ft | Ultimate strength tons/sq. in. t 'O M 18 0-23 2-80 830 i 400 30-2 38-9 180U. 0-23 2-80 900 300 82-7 93-8 75B. 0-23 2-80 900 400 72-9 83-0 90B. 10 0-27 3-28 830 500 62-1 6o-3 75B. 0-27 3-28 830 530 62-8 65-3 90B. 0-27 3-28 830 650 48-3 54-0 1800. 16 0-34 3-09 830 500 67-5 69-3 180C. 0-34 3-09 830 530 60-3 63-7 180C. 0-34 3-09 830 650 37 9 51-6 180C. 10 0-10 4-91 830 400 28-3 36-6 180U. 16 0-15 4-97 830 500 25-6 35-1 180U. 0-15 4-97 830 530 27-4 32-9 I80U. 20 0-21 4-61 830 500 43-4 69-8 180U. 0-21 4-61 830 530 58-2 68-2 180U. 20 0-20 4-08 830 500 52-0 61-2 180C. 0-20 4-08 830 600 41-6 49-1 180U. U. = Unbroken ; B. = Broken ; C. = Cracked. Material to this specification is not suitable for weld- ing, and it is intended that the heat treatment should be carried out after the parts have been formed. In the non-heat treated condition, as supplied, the material is quite easily worked, and there is no difficulty in heat treating the parts made up from it. 12. COLD DKAWN OB ROLLED STEEL. The most important Air Board specification for bright rolled steel is S.I, part 1, which calls for a material which after bright drawing gives an ultimate strength between 35 and 42 tons per square inch, with an elonga- tion not less than 15. This specification applies to bars of all sizes, and does not define the chemical composi- tion. It has recently been revised so as to ensure that the most satisfactory type of steel shall be used for making each different size of bar. The bright drawn bars are made by drawing down hot rolled bars, in which condition the drawer receives the steel. The results obtained by different drawers using the same hot rolled bars differ somewhat accord- ing to the details of the process employed. For example, the same reduction in size can be obtained by simple cold drawing or by alternately cold drawing and annealing -but the tests given by the finished products will be very different. The final result depends on the amount of reduction of area from the hot rolled to the bright drawn size, but the most important factor is the reduction of area after the final annealing, normalising, or " patenting." It is common practice in bright drawing to reduce the diameter by approximately the same amount in the drawing dies, whether the finished bright bar is J inch diameter or J inch diameter. The amount of work which is put into the steel therefore depends upon the size of the bar, and may be taken roughly as varying inversely as the square of the diameter of the finished bar. The specification on the other hand calls for a uniform value for the ultimate strength of the material, namely, between 35 and 42 tons per square inch, con- sequently the ultimate strengths of the black bars from which the different sizes are produced, or rather the ultimate strengths after the final heat treatment, must be different. For small bars the ultimate strength need 27264 42 not be nearly so high as for large bars ; with small bars it is not uncommon to add as much as ten tons per square inch in the process of bright drawing, whereas with the large bars it is not uncommon to add only two or three tons. It was therefore decided in the revised specific a- tion to grade the black bars according to the size of bright bar to be drawn from them, and the compositions have been varied in such a way that the black bar will have an ultimate tensile strength varying from about 26 ton per square inch for small bars to about 34 tons per square inch for large bars. To be successful this arrangement requires that the reduction of area during drawing shall be more or less constant for the different sizes of bars, which in point of fact is generally true. The variation in ultimate strength has been secured by varying the carbon content in the black bars, the final agreed compositions for the different sizes of black bars being as follows : CHEMICAL COMPOSITION OF BLACK BARS FOR BRIGHT DRAWING. Diameter or Width across Flats. T y and under. Between |J" and 1J. Between l^" and 2 T y. 2 T V" and over. Carbon between 0-15 and f 0-20 and 0-25 and 0-80 and 0-25 96. \ 0-30% 0-35% Silicon not more than 0-30 96 0-30% 0-80% 0-45% 0-30 o/ Manganese between 0-5 and ^"0-5 and 0-5 and 0-5 and 0-9%. I 0-996 0-9% 0-9 o/ Sulphur not more than 06 % 0-06 % 0-06 % 0-06 % Phosphorus not more than 0-0696 0-06 o/o 0-06 % 0-06%. The table of compositions which has been adopted assumes that black bars made by different steel makers and rolled by different firms will have the same tensile strength which in practice is found to be the case within reasonable limits. In the following table are given the results obtained from tests on black bars obtained from a number of different manufacturers, all of which were drawn to the same size of bright bar. The bars were tested exactly as received from the hot rolling firms, and in addition they were tested after they had been normalised at a temperature of 850 C. Carbon, per cent. Heat treatment. Yield point, lons/sq. in. Ult. strength, tons/sq. in. a ^ '3 c a 5)0 n b, I& Reduction of area, per cent. As rolled ... 17 28 41 62 As normalised 22 29 37 63 A. 01 J As rolled ... 17 29 41 63 As normalised 19 29 41 63 0-91 ) As rolled ... 21 32 38 63 As normalised 21 81 40 65 0-21 [ As rolled ... As normalised 25 24 32 32 43 41 65 67 It will be seen that there is relatively little difference between the tests obtained from the different samples certainly not sufficient to throw material outside the 7 tons range allowed by the specification. It will also be seen that the operation of normalising has had singularly little effect upon the test results. There is no doubt that much more regular results would be obtained if all black bars were normalised before they were bright drawn, but the shortage of heat-treating plant has hitherto made it impossible to specify this treatment. By normalising the irregularities due to the cooling of black bars under different conditions would be removed. The bars sometimes cool singly, sometimes in a heap, sometimes in a draught, and sometimes in a puddle of water in the rolling mill. Each of these con- ditions of cooling produces its effect upon the tensile strength, and it is only by properly normalising the bars that uniform material will be produced. A more difficult problem, and one which has quite as great effect upon the finished product, is the exact size of the black bar. The black bars are supposed to be rolled to size with a definite tolerance, and the drawer requires that they shall be within this tolerance. In actual practice it is found that the tolerance is frequently exceeded, with the result that the amount of reduction during bright drawing is not what was intended; much of the irregularity in bright-drawn steel can be referred to this cause. Another difficulty is that the elongation and reduction of area tend to vary con- siderably according to the size of the bar. With the smallest size of bar it is specially difficult, for the reasons already stated, to ensure that over-drawing will not occasionally occur, with a resulting fall in the elonga- tion; it is therefore necessary in the specification to graduate the elongation according to the size of the bar. Experiments have been made to determine the effect of blueing upon bright-drawn bars to specification S.I part 1. The elastic limit (limit of proportionality) is low in cold-worked steel, and the effect of blueing is to raise it, but this effect is not so clearly marked as it is in streamline wires (See Chapter V.) It is interesting to determine the fatigue limit of material of this kind in view of the low values of the limit of proportionality. Such experiments have been carried out upon bright-drawn bars blued at different temperatures, and the results are shown in the following table : WOEHLEH TESTS ON BRIGHT-DRAWN BOUND BAR f" DIAMETER. D A a Ratio of Treatment. 1|jr to C 8 f If 5! Fatigue Limit * is g .cafe ^ S a to Tilt. 3 CC B, C-. 3 Strength. As drawn ... 36-8 40-8 13-3 52-6 19-1 0-468 Blued 250 C. 36-6 40-1 15-11 52-6 18-4 0-46 ,. 400 C. 28-4 3!) 11 16-6 50-0 19-0 0-477 55(1 C. 26-4 36-1: 23-3 52-8 18-0 0-492 WOEHLER TESTS ON BRIGHT-DRAWN HOUND BAR DIAMETER. s Jl a a Ratio of s - . SPw tJ a ? B -S 1 Fatigue Treatment. T>.S_S^ s-? d ^ ^.3 j*. ! Limit **l H S Cd ' % X " 4 i >> K "s | s V a ^s \ 5! 9 A S| 1 1 tt ^> ^ \ k s 5! \ ^ S \ % s S 9 I I) 1 1 i t i 1 i s b3SsBestc< in o ul STRENGTH OF CAST IRON AT DIFFERENT TEMPERATURES . FIG. 56. 100 90 SO O 70 60 50 Ul K 30 20 K> WO* *00* 300* 400" 500* 6OO" TOO' "TEMPERATURE *C 43 It will be seen that the ratio of the fatigue limit to the ultimate strength is remarkably constant, the value being almost exactly the same for all the bars, despite the fact that the ultimate strength varies a good deal. The fatigue limit apparently possesses no relationship to any property except the ultimate strength, which is in accordance with the results obtained on other types of steel. The effect of blueing upon the steel is very slight so far as the fatigue limit is concerned up to a tem- perature of 550 C., and the subsequent alteration in the fatigue limit is found to agree with the change in the ultimate strength and not with the change in any other property. The value for the fatigue limit of the bright-drawn material is therefore singularly little dif- ferent from that obtained from ordinary normalised steel of the same ultimate strength, despite the high yield point which the bright drawn bar possesses. 13. NON-MAGNETIC STEEL. Non-magnetic steels are valuable for use in positions in an aeroplane where magnetic materials would affect the compass. The non-magnetic bullet-proof steel to Specification 2S. 7 has been used for petrol tanks and armour plate. (It is the material used for making hel- mets.) It has been proposed for the hull of a flying boat and the tests described below show that a compass would be unaffected inside such a hull. Non-magnetic steels are also used for magnetic spindles. There are two types of completely non-magnetic steels, one con- taining 12 to 13% manganese and the other 5% man- ganese, with about 16% Nickel ; a third steel containing 25% Nickel is often called non-magnetic, but it is far more magnetic than the others. 12% Manganese Steel This is the steel specified in 2S. 7. Its magnetic properties in low fields were care- fully measured by Dr. Bryan at the Royal Naval College on two samples, one made by Messrs. Firth and one by Messrs. Armstrong Whitworth. The B/H curves are given in Figs. 54 and 55. The effect of various heat treatments on this steel were determined by Sir E. A. Hadfield and Professor Hopkinson (vide Journal of the Iron and Steel Institute, No. 1, 1914). The following reports by -the writer and Commander Finch Dawson, of the Admiralty Compass Department, show that this steel is sufficiently non-magnetic not to affect a compass within a hull built of it. The Effect of the Steel Hvll of a Flying Boat on a Compute inside the Boat. In order to ascertain whether Messrs. Firth's non-mag- netic bullet-proof steel sheets had a sufficiently low permeability to allow of the compass being placed inside a hull of a flying boat made of these sheets, arrange- ments were made with Messrs. Firth and the Compass Department for a test on a rough model hull. A copy of Commander Finch Dawson 's report on this test is attached. This report shows that with sheets of 12 gauge and a model hull less than half full size, the screening effect was almost neglible. In order to deduce the effect on a full-sized boat it is necessary to know the law according to which the screen- ing alters with the size of the hull. Professor Ernest Wilson, of King's College, who is engaged on experiments on magnetic screening, states that for the same thickness 272(14 of sheet, the screening effects diminish rapidly with the size of the hull. The screening effect also diminishes with thinner sheets. It follows that the screening effect of the hull of the proposed flying boat, which will be made of 12 gauge and thinner sheets, will be negligible. tii'port by Commander Finch Dawson, R.N. On 7th instant I visited Messrs. Firth's Tinsley Works, at Sheffield, to test the non-magnetic character of their bullet-proof steel. A very suitable site for the experi- ments was found inside the works a clear open space with no iron within at least 75 feet. An ordinary Thomson's compass with usual ship's card was erected on a stand, and the four cardinal points marked out. The card was deflected 75 and the time taken to come to rest observed; '6 minutes 20 seconds. The compass was now removed and a vibrating needle substituted this when steady was deflected 90, coming to rest in 4 minutes 30 seconds. Bullet-proof plating of 12 gauge was now built up round the compass stand to form an imaginary boat bull, the bows being 5 feet before the stand and the hull extending 20 feet abaft; sheets being placed on the ground, vertically for the sides, and roofed in overhead. First with fore and aft line north and south, with bows heading north. Compass showed no difference in bearing of north and south points. Card deflected 75, came to rest in 3 minutes 40 seconds. Vibrating needle substituted, deflected 90, came to rest in 4 minutes 48 seconds. The imaginary boat's hull was now shifted round on to the east and west line, bows heading east. In this position the compass showed no difference in bearing of the east and west points. Card deflected 75, came to rest in 3 minutes 38 seconds. Witli vibrating needle substituted, deflected 90, came to rest in 4 minutes 45 seconds. From the above experiments it is plain that the bullet- proof steel tested was non-magnetic, although there is apparently a very slight decrease of directive force on any magnetic needle surrounded by this material. Manganese-Nickel Steel. The following data were issued as C.I.M. 725. Steel complying with A.A.D. Specification 2.S.7, which contains a high percentage of manganese, is non-mag- netic, but it is unmachinable, and as a consequence its use is practically restricted to armour plates and tanks. Samples of another alloy, made by Messrs. Samuel Osborn and Co., of Sheffield, have recently been tested and found to be non-magnetic and fairly easily machin- able. It has good strength and remarkable ductility. It appears to be very suitable for use where strength and non-magnetic properties are required. Messrs. Osborn and Co.'s alloy contains about 6% of manganese, and 16% nickel. Heat treatment has very little effect on it. It may be heated to 1,000C., and quenched in water or cooled in air without seriously modifiing the tests. Subsequent tempering also has little effect. After any of these treatments it gives: Ultimate strength ... ... 44 tons per sq. in. Elongation 70 to 80% Reduction of area ... ... 75 to 80% In non-magnetic properties it is if anything a trifle better than the Manganese steel. The permeability F 2 44 has been measured and found to be about 1'13 for values of H. from 2 to 22. This steel may be used safely close to a compass. 14. CAST IRON. Cast iron was used in the earlier aero engines for cylinders and pistons. It has also been used both in the earlier and the later engines for exhaust domes and simi- lar parts, and is widely used for piston rings. Most of these parts particularly the cylinders, pistons and the piston rings have to work at temperatures which may be quite high, in some of the earlier engines the cylinder was often red hot. In order to ascertain to what extent cast iron retains its strength at high temperatures, a series of tests was therefore made on various samples of iron prepared under different conditions and tested at temperatures ranging from room temperature to about 700 C. Owing to the incompleteness of metallurgical knowledge, it was not easy to prepare satisfactory sam- ples, and the difficulty in controlling the composition of the samples within any narrow limits has made it im- possible to deduce from these experiments the effect of any particular element on the strength of the iron. The tests were made on different classes of iron including those used for cylinders, and for piston rings, and also on a special cast iron or semi-steel, in which the percen- tage of phosphorus had been brought down to a very low figure, and the carbon content also considerably reduced. This semi-steel has been proposed for cylinders, but owing to the difficulty in machining it there is little likelihood of its being adopted, nor do these experiments show that it has any marked superiority at high tem- peratures over the other irons tested The samples tested differed not only in composition but in the size of the castings from which they were cut, and the cast- ings may 'have differed in the extent to which they were chilled so that the results cannot be considered as strict- ly comparable or representative of the chemical composi- tions, but they serve to show the general effect of high temperatures on strength. In order to make the effect of high temperatures of test as clear as possible the ratio of the strength of the iron at each temperature to its strength when cold is shown in the tables. Series 1. The specimens in this series comprised of four different chemical compositions, each of which was cast in 'three different forms a test piece (1 inch round bar), a piston ring pot, and an aero cylinder. The analyses of each sample is indicated in the tables by the numbers 1 to 4 and the form of the casting by the letters T (for the one inch round test piece), P (for the piston ring pot), and C (for the cylinder). Portions cut from the 1 inch round bars analysed with the following results : were Analysis number. 1. 2. 3. 4. Total carbon per cent. 3-11 3-35 2-93 2-96 Combined carbon per cent. 3-01 2-86 2-54 2-60 Graphite 0-10 0'49 0-39 0-36 Silicon 3-05 1-26 1-41 1-97 Sulphur 0-055 0-071 0-079 0-060 Phosphorus 1-16 0-47 0-95 0-43 Manganese 0-44 0-62 0-39 0-65 TESTS ON 1 INCII ROUND BAR TEST PIECES. ULTIMATE STRENGTH IN TONS/SQ. IN. AT VARIOUS TEMPERATURES. Mark. ir,c 200C 300C 400C 500C 600C 650C 700C 1 T 18-0 13-6 12-8 14-8 9-5 6-6 5-0 3-6 2T 14-6 16-0 15-3 14-4 12-5 8-8 5-9 3-6 3T 18-6 16-9 18-5 19-1 16-4 12-8 8-5 5-5 4 T 18-6 16-4 17-9 19-1 13-9 8-7 6-5 5-6 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 1 T 2 T 3 T 4T 1-04 1-09 85 825 986 1-05 99 96 1-14 99 1-03 1-03 731 855 825 750 51 60 69 468 i!S4 405 455 350 277 245 295 301 TESTS ON SPECIMENS CUT FROM PISTON RING POTS. ULTIMATE STRENGTH IN TONS/SQ. IN. AT VARIOUS TEMPERATURES. i P 2P 3P 4P 9-0 11-2 16-8 14-9 7'7 11-9 16-5 15-0 7'8 9-8 15* 5 13-9 9-2 11-5 16-6 14-0 8-1 10-4 15'5 13-4 5-4 6-6 10-3 8'9 4-3 5-1 9-3 7-0 3-6 4-0 6-2 5-4 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 1 P 2 P 3 P 4 P 852 1-06 &85 1-01 865 885 925 935 1-02 1-03 99 94 90 93 925 90 60 59 615 60 477 455 55 470 325 355 370 362 TESTS ON SPECIMENS CUT FROM CYLINDER CASTINGS. ULTIMATE STRENGTH IN TONS/SQ. INCH AT VARIOUS TEMPERATURES. 150 9-0 8-8 8-2 8-7 7-7 4-8 3-1 2-7 2 C 13-6 12-0 12-2 12-4 10-6 7'9 6-0 4-9 3C 16'8 15-5 16-7 16-4 15-2 10-5 8-5 6-5 4C 12-6 12-0 11-2 13-0 11-2 7-1 5'8 4-2 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 1 C 1 98 91 97 85 533 345 300 2C 1 882 895 913 98 580 440 361 3C 1 923 995 975 905 625 505 372 4C 1 952 8<>0 1-03 890 565 460 333 Se-ries 2. The specimens in this series comprised irons used for making piston rings of three different compositions, each of which was cast in three different forms piston ring pots, 1^-in. diameter test pieces, and |-in. diameter test pieces. The analyses of each sample is indicated in the tables by the numbers 5 to 7, and the form of the casting by the letters P. (for the piston ring pot), L.T. (for the large test piece), and S.T. (for the small test piece). 45 Portions cut from the large test pieces were analysed The chemical analyses of the two samples were as with the following results : follows : Analysis Number. 5. 6. 7. ?otal carbon, per cent. 3-31 3-19 3-48 Combined carbon 0-71 1-09 0-84 Jraphite 2 -GO 2-10 2-64 Silicon 1 ' ")5 0-96 2-16 5ulphur 0-082 0-205 0-079 "hosphorus 0-79 0-59 0-099 Manganese 0-92 0-47 0-70 TESTS ON SPECIMENS CUT FROM PISTON RING POTS. ULTIMATE STRENGTH IN TONS/SQ. INCH AT VARIOUS TEMPERATURES. Mark. 15C 200C 300C 400C 5 P 12-7 11-7 11-3 13-4 11-5 9-3 6 P 16-0 17'4 18-0 15-3 19-1 14-4 14-1 16-8 7 P 22-0 19-8 19-6 21-5 20-2 19-4 17-9 21-5 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 5 P 6 P 7 P TESTS ON 14 INCH DIAMETER ROUND BAR TEST PIECES. 1 918 888 1 862 695 1 1-08 1-13 96 1 756 740 882 1 900 892 980 1 962 892 1-07 ULTIMATE STRENGTH IN TONS/SQ. INCH AT VARIOUS TEMPERATURES. 5 LT 6 LT 17-7 17-4 19-3 22-1 17-4 16-7 22-8 21-2 17-2 17-8 21-0 21-9 17-8 17-3 23-2 22-9 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 5 LT 6 LT 983 905 1-18 '.Kill 1175 1-02 1-09 990 1-06 994 1-20 1'04 TESTS ON INCH DIAMETER ROUND BAR TEST PIECES. ULTIMATE STRENGTH IN TONS/SQ. JNCH AT VARIOUS TEMPERATTIRES. 6 sr 7 ST 21-3 22-3 12-6 19-3 20-5 10-3 11-3 19-5 20-2 10-3 10'4 21-5 21-3 10-9 11-7 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 150 C. 6 ST 1 905 915 1-01 1 920 902 955 7 ST I 890 890 940 Series 3. The specimens in thip series comprised two different special low phosphorus irons (or semi- steels), each of which was cast in the form of a 1 inch round bar. Analysis number. 8 9 Total carbon, per cent. 1- 72 1 44 Combined carbon, per cent. o- 59 53 Graphite, per cent. 1- 13 91 Silicon, o- 75 75 Sulphur, o- 137 0-137 Phosphorus, o- 071 072 Manganese, o- 19 17 ULTIMATE STRENGTH IN TONS/SQ. INCH AT VARIOUS TEMPERATURES. Mark. 15"C. 200C. 300 'C 400C. 8 28-4 27-3 28-1 26-9 25-3 24-3 25-2 23-4 9 26-8 28-2 32-0 33-1 32-7 27-0 33-8 36-1 RATIO OF STRENGTH AT EACH TEMPERATURE TO THAT AT 15 C. 1 965 990 948 1 960 995 920 1 1-05 1-19 1-23 1 825 1-03 1-11 Conclusions. The tables show that temperatures up to 400 C. have only a very small effect upon the strength of the iron. The fact that the results (as shown by the ratios quoted), are very irregular and are sometimes higher at high temperatures than at room temperatures shows how great the difficulty is of providing perfectly comparable specimens of cast iron. It is, however, safe to conclude that the iron retains at least 90% of its strength up to a temperature of 400 C. At 500 C there is a distinct reduction in tensile strength, say, to 85%, and after that the falling off in strength is marked. The following table gives a fair representation of the strength of the irons at the different temperatures. The results are plotted in Fig. 56. Percentage of strength at 15 C. 100 98 95 90 60 Temperature 200 G. 3000 C. 400 C. 500 C. 6000 c. 6500 c. 7000 C. 45 33 These results show a great similarity to those obtained in the tests upon hardened and tempered alloy steel (see Section 9), though the final strength is a higher proportion of the original than in steel. The results do not suggest any marked superiority of one kind of iron over another, nor is it possible to pick out any particular element which has a special effect upon the strength of the iron at high temperatures. 46 APPENDIX I. Air Board Specification 2 H. 5. December, 1918. CASE HARDENING. CASE-HARDENING MIXTURE. 1. COMPOSITION. (a) The essential constituent of a case-hardening mixture is carbon, which may be present in the form of wood charcoal, charred leather or charred bone, or mixtures of these materials. Charcoal alone, however, is not very effective at satis- factory temperatures, but can be much improved by the addition of 30 to 40 per cent, of barium carbonate, or 5 to 10 per cent, of soda ash. A simple mixture of this kind is quite as satisfactory as more expensive and complex mixtures which frequently contain undesirable additions. (i) Thorough mixing is necessary for uniform results, and the material must be stored in a dry place. No advantage is obtained by adding water or heavy oils, and if coarsely granular charcoal be used, troubles with dust are obviated. The mixture may be used repeatedly by the addition of about 20 per cent, fresh mixture each time. (e) Objectional shrinkage will not occur with the mixture recom- mended. 2. TESTING OP CASE-HARDENING MIXTURES. (a) It is desirable that purchasers of case-hardening mixtures should test the con- signments of material supplied to them, in order to determine the suitability of the mixture for its work. (*) The most satisfactory method of determining the value of a case-hardening mixture is to estimate the volume and composition of the gases evolved on heating the mixtures to various tempera- tures. This test is rather lengthy, and requires fairly elaborate apparatus with efficient handling. (c) The most satisfactory method amongst those which are applied easily is that of carbonising a sample with the proposed mixture. In this case it is essential that the method and conditions of testing be standard, and applied in all cases. The following method has been found to be quite satisfactory : (i) Prepare a stock of discs J-inch thick and f-inch diameter from a standard sample of steel to specification S. 13 or S. 14. The furface of the discs must be quite clean and free from rust. (ii) The case hardening box should b3 one which can be pro- cured easily and replaced exactly. A very useful type is a gas-fitter's blank end (4-inch size is quite suitable) with a t-crew-in cap. (iii) Fill the box loosely with the mixture, placing one of the discs about half-way down, and screw home the cap. (iv) Heat the furnace to 9 C. and keep it at that tempera- ture during ihe operations. (v) Place the box in the furnace and allow it to remain for three hours after it has attained 9CO. Allow the box and its contents to cool in the furnace. (vi) When cold saw the sample through a diameter and polish the re\ealed surface. Etch the surface w'th dilute nitric acid and so observe the depth of carbur.sation. A good mixture will give a depth of not less than 1 mm. (vii) A second test on the same sample of mixture with a second disc is also useful, as a good mixture will give practically the same penetration on a second heating under the same conditions. CASE-HARDENING PROCESS. 3. PREPARATION OP THE ARTICLES. (a") The articles to be case- hardened should first be thoroughly cleaned. (i) The most satisfactory method of dealing with surfaces which are required to be left solt is to remove the cise by machining after carburising and before hardening (see Clause 6). When this cannot be done, the surfaces may be protected from carburisation by one of the following methods : (e) Painting the surface with a suitable enamel and covering with a paste of fireclay, which should be allowed to dry before it is put in the carburising box. (d) Covering with a mixture of water-glass and fine sand, which should be allowed to dry before the articles are put in the car- burising box. (e) Electro-plating with copper or nickel ; a thick, dense deposit is required. (/) The last method (e~) is recommended as the best 4. PACKING IN CAUBUEISING Box. When ready the articles are to be packed in an iron box with the carburising mixture. The articles should be separated from the sides and bottom of the box by at least 2 inches of the mixture, and from the lid by more than 2 inches. When packed full the lid of the box ia to be luted down. An approved rotating furnace also gives satisfactory results. 5. CARBCRISING (CEMENTING). (a) The box when charged is to be put in the furnace and heated to the carburising temperature (see Table, col. 3), and kept at that temperature for sufficient time to give the required thickness of case. As a guide to penetration, it may be taken that carburising for 3 hours at 900 C. gives a case approximately 1 mm. thick. (i) After carburising, the box is to be removed from the furnace and allowed to cool. Articles which are liable to warp should be left to cool in the box. When cold the articles are to be cleaned from ash, &c. 6. MACHINING. After cooling, the articles may be machined if desired. The case may be removed from auy part where it is not needed 7. PREPARATION FOR HARDENING. Parts of the article which are required to be left soft may either be protected from the quenching medium (oil or water) by coating them with a paste of fireclay before they are heated, or the temperature of those parti may be lowered before quenching by aj plying a pad of wet asbestos to them. 8. REFINING. (a) The articles are to be heated to the refining temperatu e (see, col. 4 in the Table). Small articles of simple design may be put into a hot furnace, but large articles of complex shape which are liable to warp or crack, particularly if made of steel to Specification S. 15, 16 or 17, should be put into a cold furcaea so as to be heated slowly. (J) To avoid decarburisation the best method of heiting is in a salt bath. An alternative method is to heat the articles in boxes, packing them in the box with old carbur'sing mixture. (c) After reaching the refining temperature, the articles are to be allowed to soak at that temperature long enough to ensure that the temperature has become uniform throughout their mass. When they have attained a uniform temperature throughout, the articles are preferably to be quenched in water or oil. Man . articles cannot be quenched with safety from the refining temperature in either water or oil. Such parts may be allowed to cool from the refining temperature in still air. 9. HARDENING. After refining, the articles are to be reheated with the precautions described in Clause 8 to the hardening tem- perature (see col. 5 in the Table), and then quenched in water or oil. 10. ALTERNATIVE METHODS. (a) Articles made of steel to Specification S. 17 are very little improved by the refining process, and may be satisfactorily treated by giving them a single quench in oil from the temperature given in the Table (line 6). (i) The case produced by carburising steel to Specification S. 17 will harden if cooled in air ; articles made of S. 17 may there- fore be air hardened from the temperature given in the Table (line 7). If this method is used, the case will be hard, but the core will be left in the normalised condition and will give rests only about equal to those named in Specification S. 14, which are far below those called for in Specification S. 17. Air hardening is, however, of value, as it provides a method of producing a hard case on articles which, owing to their complex shape, could not be quenched without risk of fracture or distortion. TABLE. oi a 13 Steel Specifica- tion. Carburising Temperature. Refining Temperature. Hardening Temperature. i 2 3 4 5 6 S. 13 S. 14 S. 15 S. 16 S. 17 S. 17* 880 to 900 C. 880 to 900 C. 880 to 900 C. 880 to 90u C. 880 to 90 . C. 880 to 900 C. 900 to 920 C. 900 to 920 C. 85a]hy). then plotted against Lie U'lnperatures at which the show that all the constants ot the steel vary more or less samples had been blued. The resulting curves (Fig. 8) in the same way, increasing to maxima for blueing SO 45 30 G 20 o Coi ipression on Stn Le C/asfic L 12 9 Ions* TO/ npffu Fir,. 6. Stres/Strain diagrams for Brighf-Drawn Tube in Tension and Cotnpression(Dalby). Oi 'C L G C L f.L C.L. C L C_L_ C L C L E L i. Sfra/n FIG. 7. C L. L Compression Stress/Strain Diagrams for steel tubes blned at, different temperatures (Lea). E L = Elastic Limit. C.L. = Collapsing Load. Tubes. 1 J inches diameter. 20 gauge thick. 0-2 per cent, carbon, represent the observed curves. The corresponding collapsing loads have been marked and connected to the curves by dotb convenience of reference. 27264 G 2 52 t,emj;ei\.U,ies between 300 and 450 C., and then rapidly falling off as the material reaches the abnor- mally weak condition already refei'red to, produced by temperatures of about 600 C., and finally rising again when the recalescence point is passed and the steel reaches the annealed condition. The form of the curve as it passes the recalescence point has not been deter- mined; it is probably discontinuous through the critical range. It is shown dotted in the figures. These curves represent average values for a con- siderable number of tests. The effect of blueing de- pends to some extent on the time the steel is subjected to the blueing temperature, a longer time at a lower temperature producing results similar to a shorter time at a higher temperature. Different samples also give somewhat different results, depending on their previous treatment particularly on the amount of work which has been done on the steel in the last pass p.nd on the ratio of reduction of diameter to reduction of thickness of wall and on the heat treatment in previous softenings. Such variations are characteristic of tubes which have not been annealed before the last pass. The effects of blueing are more marked in steel con taining a higher percentage of carbon. Fig. 9 shows the curves obtained from a - 5 per cent, carbon steel tube. This tube had been annealed before the last pass. It will be noticed that the abnormally weak range is at a rather higher temperature for the O5 C. steel than for the 0'2 C. steel. This confirms the in- dications given by the micro-photos referred to on page 49. The shape of the stress/strain curve for the bright- drawn tube Fig. 7, is particularly interesting. The -Elastic Limit is very low, orly 9'2 tons per square inch, after which the curve bends away rapidly and then shows a series of steps each of which may be described as a partial Yield Point. The lowest of these partial Yield Points has been taken as the Yield Point in plotting the results in Fig. 8. The contrast between this curve and the curves which represent properly blued tubes is very striking. The Elastic Limit in the blued tube is almost as high as the Yield Point and the collapsing load. The importance of a high Elastic Limit will be real- ised if the two curves in Fig. 10 are compared. These curves show the loads which struts of any length will bear. They are; drawn for two tubes of the same dimen- sions, one made of steel with an Elastic Limit of 28 tons per sq. inch and the other with an Elastic Limit of 14 tons per sq. inch. It is only for long struts that the Elastic Limit becomes of relatively small im- portance. L too soa FIG. 8. The effect of blueing at different temperatures on the Collapsing Load, Yield Point, Elastic Limit, and Young's Modulus of mild steel tubes. Diameter of tubes, Ij inches. Gauge, 20. Carbon, 0-2 per cent. I . 260 C.L Y.P. \ >AOOO '3000 200" 7CO soo'c <100' 300" 600" FIG. 9. The effect of blueing at different temperatures on the Collapsing Load, Yield Point, Elastic Limit, and Young'* Modulus of medium steel tubes. Diameter of tubes, 1 inches. Gauge, 20. Carbon, 0-5 per cent. To face page 53.] [CHAPTER IV. Bright Drawn. 330 C. :,30 C c. 630 C. 660 C. 740 C. Annealed 850 C. FIG. 11. Compression test pieces which have been blued at different temperature?. (Lea.) Diameter of tube Ij, inch. Gauge 20. Carbon 0-296. The importance of a high Elastic Limit for tubes loaded as beams is even greater than for struts, since their strength is directly proportional to the Elastic Limit. The variation of Young's Modulus is also interesting ; the experimental results of all the observers agree in showing that the value of E is not so constant as is generally assumed, but there is not sufficient evidence to settle entirely on what it depends.* Values have been observed from a minimum of about 12,200 to a maximum of about 14,400 tons per square inch, a variation of about 16 per cent, on the mean value. The range of values for blued tubes is only about half as wide; the values vary- between 13,200 and 14,400. It appears to be safe to take a value 13,600 for normalised tubes, the lowest observed value falling less than 3 per cent, below this figure and most of the values falling above it. . The curves in Figs. 8 and 9 show the variation of E. as determined by Professor Lea, but the observations on different samples do not fall very closely on the line, and show that the value of E depends a good deal on the previous treatment of the steel. Fig. 11 is a photograph of a selection of the samples tested, and illustrates the different ways in which such tubes collapse, according to the temperatures at which they have been.biued or annealed. Blueing after the last pass is specified in all the Air Department Specifications. The Effects of Annealing Bright-drawn Tubes. The effects on the Yield Point, Elastic Limit, and Young's Modulus, of annealing, i.e., heating above the critical " The curious irregularities found for E suggest that the method of measuring it may be subject to some unknown source of error. range, are shown by the heights to which Professor Lea's curves, -tigs. H and 9, rise on the right hand of the diagrams. The values of all these quantities recover from the abnormally low values corresponding to blue- ing at 600, but do not rise so high as the values for tubes blued at temperatures between 300 and 450 G. Professor Dalby carried out a series of tests for the Admiralty on struts made of annealed tubes for com- parison with the series on bright-drawn and blued tubes already referred to. The following is an extract from his second report, which emphasises the sudden way in which annealed tubes collapse after reaching the Elastic Limit : In the case of thin tubes which are either in the bright-drawn or blued conditions, after the Elastic Limit is reached, the tube when tested as a strut will continue to support a practically constant load under a continu- ally increasing lateral deflection, and if the lateral deflection is sufficiently increased, the load which the strut will sustain begins to gradually diminish. (See load curve, Fig. 4.) As the strut is unloaded it straightens itself and, al- though there is a distinct permanent set in the strut when the load is entirely removed, this permanent set need not necessarily be very great. It is therefore clear that during the application of the load, and whilst the specimen is being deflected laterally, a considerable amount of work is being stored up in the strut, which work is largely recoverable when the load is released. For the purpose of resisting shock or of sustaining large stresses which may be accidentally applied for a short while, this property of the strut is decidedly satisfactory. The case of thin annealed tubes is, however, very different. Almost as soon as the Elastic Limit of the sqooo 1 .* 1 44000 I 5 M % 34000 ; 20,000 /oooc Y/e/JPoir, 'ietdPo/nr /4 Tons per ly "^"-4 Eu/crs C 30 40 60 SO tOO J.enqtf> //? inches. /6O FIG. 10. for i Comparative strength of struts made of steel with Yield Points of 28 and U tons per square inch respectively. The curves are drawn r tubes of 24 inches diameter and 0-058 inch thick, with S = 0'2 inch (constant for all lengths). E = 13,400 tons per square inch. material is reached, the tube buckles under the load and fails almost entirely, the load dropping immediately to a very small value. Consequently comparatively little work is necessary to produce failure, and practi- cally none of this work is recoverable when the load is entirely released. It therefore follows that tubes of this kind are totally unable to resist anything in the nature of shock or accidental high loading, inasmuch as, if the Elastic Limit of the material is passed by even a small amount, the strut immediately buckles, and the strut, and structure of whch it forms a part, is permanently and badly deformed. Effects of Welding, Brazing, and Soldering. When two tubes are welded together, the weakest part of the joint is usually a fraction of an inch away from the weld. It is practically certain that the weakness at this point is due to the fact that the temperature there has reached between 600 and 800 C. if the tubes have been softened during manufacture at 600 C., and not annealed before the last pass, then since the steel is in the condition shown in Fig. 2, the tubes will be reduced in strength to the abnormally weak condition indicated in Figs. 8 and 9 for temperatures between 600 C. and 800 U. ; but if the steel has been annealed before the last pass, there will hardly be time during the welding for the balling up of the pearlite to occur to any serious extent, and the strength will only be reduced to that of the blued steel. Brazing will produce the same effect. Soldering does no harm if care is taken not to heat the tubes hotter than is necessary. Solder melts at about 200 C. Testing and Inspection. Compression Tests. Pro- fessor Dalby's tests, which show how far the Elastic Limit in compression may fall short of that in tension, indicate that tubes should be tested in compression if they are to be used for struts or beams. Compression tests are specified in the Air Department Specifications. Straightness. The importance of having the tubes straight is shown by the effect of g in the formulae for calculating their strength. It is important that the tubes should be straightened before blueing, for any straightening carried out subsequently necessarily leaves part of the steel in an overstrained condition and liable to return to its crooked state under relatively small forces. Clauses providing for the inspection of the tubes for straightness have been introduced into the Air Depart- ment Specifications. Elastic Limit and Yield Point. In commercial testing the Elastic Limit is much more difficult to determine than the Yield Point, and in blued tubes the two points are not far apart; it therefore appears to be best to use the Yield Point rather than the Elastic Limit as the criterion of the quality of the tubes, and if this is done it is also most convenient to use the Yield Point in the formulae for the limiting load. The Yield Point in Compression is specified in the Air Department Speci- fications, and has been used in constructing the curves representing Southwell's formuse given in the Appendix. 2. BRITANNIA WELDED STEEL TUBES. Early in the War there was a shortage of solid- drawn steel tubes for aircraft construction and the suita- bility of welded tubes was investigated. The most suitable welded tubes appeared to be those made by the Britannia Tube Co., of thin sheet steel rolled into tubular form and welded along the butt joint by the oxy-acety- lone fiame; but there was considerable doubt as to the reliability of the weld and the effect it might have on the strength of struts and also whether it would stand bending. The tubes were liable to crack at the weld when struck by a hammer or flattened in a vice. A microscopic examination of the welds was made for the Materials Section by the N.P.L. which showed that the welds were sound but that the steel in the neighbourhood of the weld was considerably altered iu structure by the process. Tests on struts were made for the Materials Section by Professor Dalby, F.E.S., and gave remarkably good results showing that the tubes were quite as strong as solid drawn tubes for struts. Tensile tests of the weld were made for the Materials Section by Mr. Percy Kirk, at the Hackney Institute. Circumferential strips were cut from the tubes and straightened out and tested in tension; the results showed that the strengtli of the weld in tension was between 50% and oO% of the strength of the steel strip. This weakness of the weld appeared to be of little impor- tance as the tubes were not required to stand internal pressure. These reports were considered sufficiently satisfactory and the welded tubes were brought into use; at the same time experiments were made by the makers, in conjunction with Lieut. Currall, B.N.V.E., and Prof. Lea at the Birmingham University on modifications of the process of manufacture in order to improve the weld and restore the steel to a thoroughly satisfactory condition. The result of these experiments has been the introduction of several modifications in the process of manufacture which have greatly improved the weld and strengthened the tubes. The uniformity and excellence of these tubes as now made is shown by the following series of 15 strut tests made by Lieut. Currall and Prof. Lea. The tubes used in these experiments were taken from the Britannia Company's stock, and were not specially prepared or selected in any way. The results show that their strength is remarkably uniform and that their combined crookedness and eccentricity of bore is much less than is allowed under Air Department Specification for solid drawn tubes. This result is to be expected since they are formed from sheet steel, and consequently the bore is less eccentric than is common in solid drawn tubes. The results of the tests agree very closely indeed with the theoretical figures^ as is seen by the plotted curves. Such close agreement has rarely been found before. The tubes were 2 inch diameter by 17 gauge made from 0'35 per cent, carbon steel; they were annealed after welding and previous to the final pass, at 830 C., and subsequently heat-treated at 450 C. in accordance with Air Department Specifications. The yield point in compression was 35 tons per square inch, and the ultimate strength was 42 tons per square inch. The value of E, deduced from bending tests, was 13,600 tons per square inch. The struts had values of L/K varying from 10 to 125, and were all tested between parallel knife edge supports. Fifteen tests in all were made, and the results are plotted in Fig. 13. At failure the plane of bending did not appear to be connected in any way with the weld, and in no case did the weld, after failure, show any sign of a crack. Chapter. IV. i. 15. fe *3 Ji 11 90 n 17 W "* t 2 " 4 "> -5 " w *> a a n 10 a a 4 z i WELDED TUBE STRUT TESTS WITH KNIFE EDGE ENDS I r 4 ^ 3d E- 13.600 TONS D.' YIELD -35 TONS D'. ~* v ^ \ A is Soulr* B C O is E tiler h 'city = Standard too + W ^n s k s i ^ ^ ^ V c S A- 0\\ s B V \ ^ \ \ \ * 1 A s v \ \ s \ A \\ f : K ' i. . \ \ \ 1 - free Length- L -> well Cur re with equivalent eccentr Curno * \ \ - V i s \ t \ ' \ \ V \ . \ ^L \ V \ s N k . s \ ^ V < ^ \i i = Half Standard = Quarter Standard - Zero V V V s ' ,- v s - \ V \J *v > \V '*#* tons per sf in './in inches k^ !^ ^ *v ^s, 'v X ^ J 2 V s p- co /lapsing stress in ^ v^- ^> k. ft- field 'point stress c ^ > ^ 6- equivalent eccentnci -^, h - outside radius, of tube - ^ "*-. -. i - length ctec of knife edfs inches. for Z"+IJo.tube. ~ z K- radius of ^/ration in .- 13.600 tons per sq in +- experimental points WO 170 ISO Values of L /K CHAPTER IV.] [To face page 54. ".- . < a . \ : r ..-.. x ' v '" -^ .-.' . - - . ' - " ''' l \ >', '- >. ;,; \ -->. .*- - FIG. 14. 0-35C. At the weld, x 100. FIG. 15. 0-35 C. At the weld, x 1,000. FIG. 16. 0-35 C. Opposite the weld. X 100. 'V -Vs . .* FIG. 17. 0-35 C. Opposite the weld, x 1,000. I face pai/r .".">.] [CHAri'ER IV \ FIG. 18. 0-35 C. Welded and Annealed 650 C. to 700 C. At the weld, x 100. FIG. 19. 0-35 C. Welded and Annealed 650 to 700 C. At the weld, x 1,000. vliMl^fi^ *-" : te:%: _>- - ;**'~> I >*, 1 N ,: .-r > >-,-;; -.^gtoV FIG. 20. 0-35 C. Welded and Annealed fir>0' to 700 C. Opposite the weld. XilOO. FIG. 21. ll-SS C. Welded and Annealed 650 to 700 C. Opposite the weld, x 1,000.| 55 In the figure, the curve D is Euler's curve for E = 13,600 tons per square inch. The curves A, B, and C are curves representing the formula P = V "2 as given above. The curve A corresponds to the value of the equivalent eccentricity allowed in the Air Depart- ment Specifications. The curve B corresponds to a valve of S one half as large, and to a value of c one quarter as large. The points representing the results of the tests follow the curve C, corresponding to an equivalent eccentricity equal to one quarter the specified maximum. If ball-ends instead of knife edges had been used, the apparent equivalent eccentricity would have been larger. 7T The following account of the experiments made on modifications of the process of manufacture is abstracted from Lieut. Ourrall's report. The original process of manufacture consisted of the following operations : (1) Rolling up steep strip into a close jointed tube, at red heat. (2) Welding this tube by oxy-acetylene torch along the joint. (3) Annealing the tube at 670 to 700 C. (4) Pickling in sulphuric acid to remove the scale. (5) Washing in water to remove the acid. (6) Heating the tube to about 100 C. to drive off the moisture. (7) Cold drawing the tube to harden it and make it the required size. (8) Blueing the tube at about 400 C. The experiments were made on two samples of tube. Sample A : 0-35% Carbon Steel, 0-070 inches thick, rolled to a close jointed tube 1'39 inches outside diameter and welded and subse- quently drawn down by one pass to 1'313 inches outside diameter and - 062 inches thick; a reduction of 5-6% in diameter and 11'4% in thickness. Sample B : 0'15% Carbon Steel, 0-042 inches thick, rolled to a close jointed tube 1'250 inches outside diameter and welded and subse- quently drawn down by one pass to 1'188 inches outside diameter and 0-036 inches thick; a reduction of 5'8% in diameter and 14'3% in thickness. Microphotos numbers 14 to 21 show the structure of the steel at the weld and away from the weld at different stages in the manufacture of Sample A tube. Photos Nos. 14 and 15, at the weld, show that the structure is close-grained and angular, indicating that overheating has occurred close to the weld. Photos 16 and 17, away from the weld, show the nor- mal structure of the rolled steel. Photos Nos. 18 and 19. at the weld, after annealing at 650 to 7000 C. show that although a considerable amount of refining has taken place the structure is not normal, the temperature of annealing not having been high enough. Photos 20 and 21, away from the weld, after annealing as above. Photographs of the tube after drawing are practically identical with Nos. 18 to 21, and are not reproduced. The process of blueing at 400 C. makes no visible change in the micro-structure. These photographs confirm the results obtained at the N.P.L., and indicate, as the N.P.L. Report stated, that the tubes should be annealed at a higher temperature after welding to bring the steel into normal condition. While arrangements were being made for higher tem- perature annealing, a preliminary set of experiments was made to ascertain whether blueing at higher tem- peratures would make the weld less brittle. Bright drawn tubes of Sample A were blued at various tem- peratures at the Birmingham University Laboratory and tested in compression with the following results: Blueing Temperature GO As drawn 400 4700 Yield Stress tons/sq. in. ... 37-4 ... ... 40-0 ... ... 37-5 ... Ultimate Compression Strength tons/sq. in. ... 42-8 ... 43-6 40-8 500 570 ... 35-7 ... ... 34-8 ... ... 39-5 38-6 The test pieces blued at 400 and 470 split down the weld before one complete fold had formed. Those blued at 500 and 570 formed one complete fold, and as the fold completed, opened slightly at the weld, but the split did not extend down the tube showing that over-blueing did soften the weld. In order to ascertain how far complete annealing would soften the weld a sample of the tube was an- nealed for half an hour at 900 C. and tested in com- pression. It gave a yield Stress of 23-6 tons/sq. in. and an ultimate strength of 30-5 tons/sq. in. and crushed from 4 inches to 3 inches without sign of crack or opening at the weld. The micro-structure of the steel is shown in photos 22-25. The structure is norma! both at the weld and away from it. This test confirms the suggestion that the weld can be restored to a good condition by thorough annealing. The second series of tests on tubes annealed at higher temperatures were made on Sample B tubes ('15% C.). At the time there were difficulties in getting adequate supplies of -35% C. strip and it seemed likely that the lower carbon steel might have to be used. The anneal- ing temperature was not raised above 830 to 850 0. as excessive scaling e:ave trouble when higher tempera- tures were used. The micro-structure of the tubes is shown in micro-photos numbers 26 to 29 which show that the annealing at 830 is not quite sufficient to re- store this lower carbon steel to its normal condition. The mechanical tests showed that the tubes usually formed two or three complete folds without cracking at the weld, and test pieces would frequently crush completely down without cracking. The yield stress was 32 tons ner sq. in. and the ultimate strength 35 tons per sq. in. As these tests showed that sufficiently high tempera- tures to fully anneal the -15/ C. steel could not be usd without excessive, scaling it was necessary to revert to the use of the '35% C. steel. Arrangements were successfully made for a supply of the necessary strip, and '35% C. steel has since then been used for all welded tubes for aircraft. The micro-structure of tubes as now made of '35% C. steel, annealed at 830 to 850 C. after welding, and given one cold pass and then blued at 400 C. is shown in micro-photos Figs. Nos. 30-33 which show that the structure of the steel at the weld is normal. The average tests on this tube are: Ultimate Tensile strength, 45 tons/sq. in. Y.P. in Tension -10 Y.P. in Compression ... 40 ,, The tubes easily stand crushing to form one fold without cracking, as required in the specification. 3. AEROPLANE AXLES. Aeroplane Axles are now practically all made of solid drawn chrome nickel steel tubing, complying with Airboard Specification T.2 The tubes are hardened by cooling in air from about 850 C. and are afterwards tempered at 250 C. which greatly improves the elastic properties and only slightly reduces the tensile strength. Every axle is proved by having a load applied to ; t which produces a bending stress of 78 tons per sq. inch and most axles pass the test with a hardly perceptible set. Bending Loads on Aeroplane Axles. It frequently happens that an axle is bent in making a heavy landing. Many are bent at one end only when the machine has landed on one wheel, but the majority are uniformly curved almost from end to end, indicating that the load has been fairly shared by the two wheels. When an axle is thus bent the wheels must splay, as shown in Fig. 34, and this splay greatly increases the leverage, of the forces bending the axle. If an undercarriage ic some extent with very little restraint. This point is now being investigated at the R.A.E. but in the mean- time it is safe practice to neglect any possible restraint No load. Unstrained axle. loaded while stationary, splaying of the wheels is re- strained by friction between the tyres and the ground, and the bending moment in the central portion of the axle between the shock absorbers may be practically zero. The wheels are, however, not stationary but are running when making a landing and it seems probable that in running forward they may also run outward to \-:> tons load. Actual deflections. FIG. 34. Average; Air-hardened Nickel-Chrome Axle, untempered. and to assume that the maximum bending moment in an axle is the load on one wheel multiplied by the dis- tance h + c shown in Fig. 34, where 3 is due to the splaying. An approximate formula for for axles of T.2. quality, and a fibre stress of 90 tons per sq. in., is : 230 x Axle diameter' Where T = track in inches, D = wheel diameter in inches, and d tyre diameter in inches. Fig. 35 shows the extent to which the load-carrying capacity of an axle is affected by the splaying of the wheels. Curve 1 shows the relation between the curva- ture and the load carried, assuming the lever arm to remain constant and the steel to be perfectly elastic. Curve 2 shows the same relation when the lever arm is increased by the amount g due to splaying. A tempered chrome nickel axle tube complying with specification T.2. is nearly enough elastic up to a fibre stress of 80 tons per sq. in. to be represented by Curves 1 and 2 up to that fibre stress. Curves 3 and 4 show the same relation for an air-hardened but untempered chrome nickel axle tube with low elastic limit such as was used prior to the issue of specification T.2. Curves 5 and 6 show the same relation for a carbon steel axle tube complying with Air Board speci- fication T.5 and made, approximately 30% thicker than the chrome nickel axle. Curves 4 and 6 are super- posed in Fig. 35, and it will be seen that this extra thickness makes the T.5. tube fully as strong as the nickel chrome one. Broken Axles. A considerable number of axles have been examined that have broken in service. In every case fracture has been due to holes drilled in the axle at or near to the shock absorber where the maxi- mum bending stress occurs. Laboratory tests have been made on the axles that have fractured in service by bending them until they failed. In some of these tests the load was applied slowly, in others it applied in about 1/lOth of a second, which is of CHAPTER IV. 1 [To fin-? page, 56. */ v*^_ m- * Fie. 22. 0-35 C. Welded and annealed 900 C. At the weld. X 100. FIG. 23. 0-35 C. Welded and annealed 900 C. At the. weld. X 1,000. ^^ . * - -wf ' "'<% ^*tSft"*I ' "ik ^^Vi^^S^-^ ^*^S.| FIG. 2-t. 35 C. Welded and annealed 900 C. Opposite the weld. X 100. FIG. 25. 0-35 C. Wolded and annealed 1)00 C. Opposite the weld. X 1 ,OOU. Fie. 2R. 0'15 C. At the weld. X 100, I - V V* x r \ " - A A FIG. 27. 0-15 C. At the weld. X 1,000. liofuce page 57.] [CHAPTER IV. , ?i$*? *" i - t t ' ' "* ' A v H ." _. * ** -, _ ' y ~* t T A* \ * ' * > 'ua'. ' '>' *(" ' ' < ,' A ' ..*>* * ' i. . * N .. .' v. ,r V "*' *. V, ' ^ . ; . .- is r FIG. 28. 0-15 C. Opposite the weld, x 100. s X i- I 'I* I ' I mw > f?-.$iii'*i'v-> JM~F &&t\ - .-if ' * FIG. 30. On the weld, x 100. . 29. O'loC. Opposite the weld. X 1,000. FIG. 31. On the weld. X 1,000. FIG. 32. OfiE the weld, x 100. FIG. S3. Off_the weld, x 1,000. Chapbe r. IV. (> CO GO u -J x: o UJ Y a % s t s 1 2 ? 3 HONI Oft U3d 6K40JL 6S3UJ.g o I S 5 ui bi I O u KJ CVJ Ul i CD Ul Ul CO (0 >U 19 W fc g 10 CO O 1L s P s a ? a 2 57 order of the time taken to stress an axle to failure in making a landing. The axles that had holes drilled in them always failed by breaking in tension through the holes when the load was rapidly applie 1, and some- times broke through the holes even when the load was slowly applied, whereas no axles broke at an undrilled part when loaded either suddenly or slowly. These tests showed that the breakages were due to the holes and not to brittleness of the steel. The results were particularly striking as some of the drilled axles that had broken in the service were of T.5 quality and relatively ductile, and one of the axles which was tested under the shock load when undrilled was an under-tempered T.14 axle which was much more brittle than any axle known to have broken in service, yet it did not break. Various designs have been successfully used which avoid the necessity for holes in stressed parts of the axles. Chrome Nickel Steel Axles. 90 ton steel axles were being made commercially early in 1914 by Messrs. Accles and Pollock who obtained hot rolled chrome nickel steel tubes from Sweden and reduced them to the necessary size and thickness by cold drawing and finally hardened them by cooling from about 850 C. in air. The process is described in Mr. Hackett's paper to the Eoyal Aero- nautical Society (Aeronautical reprint, No. 7). The poor elastic qualities of hardened axles as then used and the improvement effected by tempering is clearly shown in a series of compression tests made by Major Robert- son, the results of which are given in Fig. 36. Tension tests made at Messrs. Accles and Pollock's works gave results in agreement with Major Robertson's. These tests showed that tempering at 250 C. reduced the tensile strength by about 5 tons per sq. in., and temper- ing at 400 C. reduced the strength by about 15 tons/sq. in. Other tests made by Accles and Pollock showed that varying the temperature of hardening from about 820 C. to 900 C. resulted in very little change in the physical properties; the higher tempera- tures appeared to favour higher tensile strength and greater elongation and lower elastic limit. Tempera tures sensibly above 900C. caused deterioration in all respects. Before tempering was introduced axles were fre- quently put out of service by reason of being bowed by the shock of landing though the set was frequently little mors than the maximum permitted before an axle was replaced, viz., f in. in 7 ft. length. It was therefore agreed that axles should be tempered to im- prove their elastic properties. It was also decided that the tensile tests used for passing axles were unsatis- factory, and should be replaced by bend tests arranged to show whether the axles had the necessary elastic properties as well as strength. Mr. Hackett suggested that every axle should be proved, and made a machine to carry this into effect. This proposal was embodied in the specification and has not been found difficult to carry out. It was decided that axles should be tempered >at 250 o. rather than at 400 C. although the latter tem- perature raised the elastic limit to the maximum. The reasons for this choice were that axles tempered at 250 C. develop a considerably higher moment of resistance, before failure than axles tempered at 400 C., and thus provide a greater margin of safety against 27261 complete failure. The lower temperature, moreover, improved the elastic properties sufficiently to enable the axle to develop a stress of 80 tons per sq. in. without being BEND TESTS ON TEMPERED CR: Ni STEEL AXLES. CURVE I:- TEMPERED AT 230* c CURVE 2- .. - 450- c 120 110 100 90 80 70 CO 60 50 40 3C 20 10 X 'FAILED BY BUCKLING. ZO 40 60 80 100 120 MO 160 180 200 220 2-M 260 STRAIN. FIG. 38. permanently bent to an extent which would involve replacement. In case of a heavy landing on one wheel the axle tempered at 250 C. would bend without buckling until the wheel fouled the undercarriage, whereas the axle tempered at 400 C., if it were thin, would probably buckle and fail completely before the wheel came up against the undercarriage. This latter point is illustrated by curve 1 and 2 of Fig. 38, which shows the behaviour of two parts of the same axle tempered respectively at 230 C. and 450 C. and tested by single point loading. It will be seen that the strain at failure was only about 150 for the first, as compared with 210 for the second. Since 1917 chrome nickel axles have been mostly made in Sheffield by Messrs. Vickers. Testing. T.2 specification requires each axle to be held at one end in half collars about 10 ins. apart and to be loaded as a cantilever near the other end till the maximum stress is 78 xZ tons per sq. in., where Z is the modulus of the section. After the specified load has been sustained and removed the remaining deflec- tion must not exceed 5% of the elastic deflection if measured at the point of application of the load, or must not exceed ^ in. if measured at a distance of 10 ins. from the support. The set of ^ in. at 10 ins. from the support very nearly corresponds to the same strain (vim., 5% of the elastic strain) in axles between 1J in. and 2J ins. dia., and as checking the set at this point with a fixed gauge is much the more rapid method of working, it is always used except with axles of large size. The advantages of the cantilever method of bending are that it is easy to carry out and that it H 58 enablas the axle to be highly stressed without being excessively bent. Most axles pass through the test with a set of less than '01 in at 10 ins. from the sup- port, and do not require to be straightened before being passed into service. It was proposed at one time to bend one or two in every 100 axles to a radius of curvature of about 29 diameters (see Fig. 39) to prove that if severely bent in FIG. 39. Extreme Curvature provided for in Specification T. 14. service the wheels would foul the undercarriage before the axle broke. This was found to be a difficult test to carry out and, since it was shown that even some what brittle axles would fail not in tension but by buckling in the compression wall, this test was replaced by a test for brittleness which requires the axle to stand being pressed oval to a specified extent without cracking. This test is very simply made but it is not altogether satisfactory as it is relatively severe on axles having a small ratio of diameter to thickness. Specification T.5. Axles to this specification are made from '4% to '55% carbon steel rendered hard and elastic by cold drawing and blueing. They usually have a yield point of 45 tons per sq. in., and an ultimate strength of 50 to 54 tons per sq. in. Although such tubes have only one half the tensile strength of hardened chrome nickel steel tubes they require to be only about 30% thicker to possess equal strength as axles. (See page 57, Fig. 35.) T.5 tubes were very extensively used first by the Naval Air Service, and later by the Royal Air Force before supplies of T.2 axles became sufficient. Though the tubes are usually made of the strength stated above, considerably greater strength can be obtained as is shown in the following report. Test Report on 300 T.5 Axles. 2-165 ins. o.d. x 0'102 ins. thick x 7 ft. long. Made from "45% carbon steel. Hard drawn and blued for 1 hour at 400 C. No. Yield. Crush. Tons per Tons per sq. in. sq. in. 1 58-3 ... 63-3-| 2 ... 56-6 ... 62-5 3 ... 56 ... 61 I 4 ... 59-7 ... 64 5 ... r>7-3 62 J 6 7 8 9 10 56-7 55-4 56-3 54 61 62 61 61 62 Test pieces cut from ends of tubes nearer back end of muffle. Test pieces cut from ends 2 \ of tubes near door or cooler I end of muffle. J The tests were made by crushing samples 3 ins. long with accurately turned ends between compression heads. All the specimens formed a partial fold without cracking, but afterwards cracked when shortened by '36 ins. The results are exceptionally good. The results are exceptionally good. The following table gives the results of bend tests made by the manufacturers on some experimental axles. These results are also exceptionally good. If inch x 12 gauge T.5 Axles. Proof Load, 856 Ibs. Leverage, 28'1 inches. Stress, 50 tons per sq. in. Tube. Heat Treatment. Permanent Set. Tube. A J hr. 350 C. 040 040 Mean thickness 101. 400 C. 030 030 .. diameter 1 748. 450 C. 020 020 Tulifli i hr. 350 C. 030 030 Mean thickness 100. 400 C. 020 030 .. di. meter 1 747. 450 C. 030 030 Tube C ... J hr. 350 C. 030 030 Mean thickness 102. 400 C. 030 030 .. diameter 1 747. 450 C. 020 030 The axles were bent as cantilevers. The permanent set was measured 20 ins. from the end. It will be seen that the permanent set was less than 2J% of the elastic deflection. The result of a set of tests on a more average quality of T.5 axle tubes is given in Kig. 40. The object .if these tests was to determine the best blueing treat- ment. To make the test of the treatments as severe as possible the tube was given a set before each treat ment and then bent in the reverse direction. The im- provement effected by blueing from the state shown by curve 2 to that shown by curve 4 is noteworthy. Hard drawn tubes of the quality required by T.5 specification do not appear to be made abroad. In France tubes made from steel of the same composition have been used as axles, but they were left in a half hard state having an effective strength as axles only about 60% as great as T.5 tubes. The original French axle for the Nieuport machine weighed 24 Ibs. This was replaced by a T.5 axle weighing 12'2 Ibs. and having greater strength. If a T.2 axle of equal strength had been used the weight might have been reduced to about 9J Ibs. The largest T.5 axles made so far are 4 ins. o.d. x J in. thick. These were satisfactory, but are now re- placed by T.2 axles. T.14 Specification was issued at a time when supplies of chrome nickel steel tubes were insufficient. As will be seen from curve T.G.3, in Fig. 14, great strength can be obtained by properly heat-treating plain carbon steel. The T.C.3 axle was of English plain carbon steel containing just under "4% of carbon, oil hardened and slightly tempered. Such steel had much less tough- ness than the chrome nickel steel that had been used, and the question of the degree of toughness necessary was closely looked into. It was concluded at the time Chapter. IV. TESTS ON BLUEING T5 AXLE TUBES. Fio 4O. I -I * S J :$ CURVE I - BRIGHT DRAWN - MT TO orr.Mr SHOWN BV DOT-TED LINKS CURVE 2 UNT IN MVCMC OIKCCTION exrewr >^s w i CURVE 3 - BLUED I HOUR _ N ~ '" *tvtxst DiKtcriON >T 350- c * TO iAMC EXTENT AS N- 2 CURVE. -4 - BLUED I HOUR _ " '* VMC DIICTK>N >3"O.DX-09*AXL5 MADE XT <4OO*C. ATD4AMC CXTCNTAS fVCLOLCSS CO. 5TggL (CARBON ABOUT ~X, Af ABOUT ttO (40 WO STRAIN too tao MO MO ceo 500 59 that if an axle would bend round a radius equal to 29 times its diameter without fracture it was sufficiently tough. This degree of curvature is illustrated bv Fig, 14. Supplies of T.2 axles became ample shortly after the issue of the specification, and as special heat treatment plant is necessary for quenching the tubes they were very little made. Duralumin Axles. Solid drawn Duralumin tubes were at one time much used for aeroplane axles, and when suitably made they compared favourably with steel axle tubes for light machines. Duralumin tubes are given a cold pass through dies to straighten theiii after being heat-treated. This pass can be made suf- ficiently heavy to raise the yield from about 16 to 28 tons per square inch, and to improve the elastic pro- perties without hardening the tubes too much to be satisfactory as axles. A If ' in. x '128 in. thick dura- lumin tube of this quality has the same strength as a 1J in. x '048 in. thick steel tube complying with Specification T.2 and weighs 10% less. It is also some- what more springy, which is an advantage when the axle is short. The demand for duralumin axles fell off after the issue of Specification T.2; no specification for duralumin axles has therefore been issued. APPENDIX I. STRUT FORMULAE. Two formulas give results in close agreement with the tests. In the first (a modified Perry formula) the effects of crookedness, eccen- tricity of bore, eccentricity of loading and deflections due to lateral forces are all allowed for by assuming the tube to have an equivalent crookedness defined by the distance c between the centre of the tube and the line along which the load is applied. In the second (a modified Smith formula) the results of the above defects are all allowed for by assuming the load to be applied eccen- trically by an equivalent eccentricity f. The formula; are : Modified Perry formula*: *y- v/T - p ^' where Modified Smith formula : , L 1 2 ' In bi/th formula; the symbols have the following meanings : p = Limiting average stress. /;= Yield stress of the material. E = Young's Modulus. h = Greatest distance of extreme fibre from C. 6. of cross section. (. 30) the maximum load per square inch that can be sustained is t at given by the yield stress. For materials that have no drop of stress at the yield point and which have stress strain diagrams deviating from the linear relationship, the only theoretical treatment is that of Southwell (Engineering, 1912), which deals with perfectly straight struts. As yet there is no theoretical method of allowing for the inevitable defects of struts made of such materials. 27264 The equivalent eccentricity of a strut, $ used in the Smith formula is the sum of four terms : 1. The crookedness of the tube, i.e., the maximum deflection at any point on the centre line of the tube from the line through the centres of the pin joints (due to crookedness of the tube). In commercial tubes the crookedness need not exceed 0'02 inch per foot run of tube ; with careful manufacture it may bs made much less. 2. The eccentricity of the bore. The eccentricity of the bore may be taken as 0'025 times the bore of the tube, if the maximum eccentricity is due to a variation of thickness of the wall of the tube of + 10 per cent. (Total difference between thickest and thinnest side = 20 per cent of the mean thickness). This is the maximum allowed in the Air Board Specifications. 3. The radius of the friction circle of the pin joint (<'.., the radius of the pin X coefficient of friction). This term may be omitted if the tube is pinned to attachments which do not rotate so as to assist the bending of the strut. 4. The deflection produced by lateral forces acting alone. The Equivalent crookedness of a strut, c, used in Perry's formula, is made up of the same four terms, but the two terms Xos. 2 and 3 must ach be multiplied by 6/5 before they are added together. The Perry formula is explicit, and the value of p can be calculated directly, but the Smith formula has p on both sides so that its value cannot be obtained directly ; Southwell has shown how to construct curves to enable this formula to be usod. To construct a Curve representing Smith's formula. The co-ordinates chosen are the Limiting Average Stres, p, and i' the ratio of the length to the radius of gyration of the cross-section of the strut. First draw Euler's Curve (SP, Fig. 12). given by the equation also the horizontal line p = Elastic Limit or Yield Point (LS, Fig. 12) . Next, choose any value of X with its corresponding value of-p, from I the following table, and mark off M so that LM _ hS MO ~ A"' ! and on the horizontal MP mark off Q, so that MQ_ I MP I 1 ' Then Q is a point on the required curve. Any number of points may be found by selecting a series of values of X and p, from the taMe. H X 60 L 1 ^f/ojf/c \s Lt/ntf 'f <* \, \ k \ Eater's cu \. r-ve. ^ ~ k FIG. 12. Construction of Southwell's curve. TABLE 1. Values of X corresponding to different Ratios r L | I I I I V X I' X J> X V X o-o 1-000 0-25 1-082 0-50 1-414 0-75 2-613 0-05 1-003 0-30 1-122 0-55 1-540 0-80 3-236 0-!0 1-012 0-35 1-173 0-60 1-701 0-85 4-284 0-15 1-028 0-40 1-236 0-65 1-914 0-90 6-329 0-20 1-051 0-45 1-315 0-70 2-20:i 0-U5 12-745 It will be noticed that the curve constructed in this way depends on the values chosen for Young's Modulus B, the Elastic Limit or Yield Point, and the Equivalent Eccentricity S. For a given quality of steel (or for tubes made in accordance with one specification) Young's Modulus and the Elastic Limit or Yield Point may be assumed to be constant, but the value of S will generally depend on the length of the strut. The curve, as drawn above, is only correct for one value of S. If a series of curves be drawn for a series of values of S, then the limiting stress may be read off one or other curve for any strut made of the steel (determined by E and the Elastic Limit or Yield Point) for which the curves are drawn. If the value of i' be assumed to be a definite function of the dimensions of the tube say for commercial tubes : S = then a new curve may be drawn for a reasonably straight tube of any selected size, giving the Limiting Stress for all lengths. Th3 curves in the Air Hoard Specification have beea drawn in this way and give the Limiting Loads for struts of any length made of tubes of standard dimension and the quality of steel specified in the Specification. The formula and curves may be used for struts of all lengths, and all thicknesses of wall greater than - A strut whose length is very great compared with its diameter may give a large deflection under the limiting load ; it it therefore desirable to calculate the deflection of such a strut. This may be done as follows : Deflection of a Strut. The Deflection of a strut can be readily calculated using either of the two formulae. By Smith's formula : The total deflection of the centre of the strut from the line of action of the load is Deflection = h where /,, is the stress AC, /is the stress BC and f y J is the stress AB in Fig. 10, page 53. Thus Deflection = x - h BC By Perry's formula : Deflection = I APPENDIX II. EXTRACT FROM AIR BOARD SPECIFICATION T.I. FOR 35-TON CARBON STEEL TUBES. HARD DRAWN AND BLUED. CARBON ABOUT 0'3 PElt CBXT. (For Softened Tubes, see Specification r.21.) If blued tubes are annealed, braced, or welded, their strength will I/a reduced at the parts where they are so treated to the values given in Clause 8, Sojtening Test. 1. MAIEMAL. The steel is to be obtained from approved makers. A copy of all orders for steel billets or blooms intended for manu- facture under this specification, stating the quality, brand, and analysis is to be shown to the Inspector on demand. The steel in the fully annealed condition must have an ultimate strength of not less than 28 tons per square inch, and a yield point not lets than 18 tons per square inch. The Contractor is to supply the analysis of the steel when required to do so by the Inspector. The Inspector may also select samples and hare them analysed at the Ministry of Munitions expense. 3. MANUFACTURE. The tubes are to be seamless and cold drawn. The tubes are, if possible, to be annealed at a tiemperature between 830 C. and 870 0. before the final pass, and to prevent brittleness are to be reduced in diameter as little as possible in that pass. 4. GENERAL CONDITION. The tubes are to te smooth, true to section, of uniform sectional thickness, and of equal diameter throughout, free from sojile, dirt, specks, longitudinal seaming, lamination, grooving or blistering, both internally and externally. 5. ACCURACY OP FORM, SIZE, AND STHAIGHTNESS. () Round tubes are to be accurately circular to the satisfaction of the Inspector. (/>) The mean outside diameter (i.e., the mean between the maxi- mum and minimum diameter), at any point, is not to differ from the size ordered by more than + O'OOi in. (or for tubes over 2 ins. diameter by more than diameter/500). The mean insid-j diameter is not to be less than the correct outside diameter miniit twice the maximum permissible thickness, nor greater than same minus twice the minimum permissible thickness. (c) Oval tubes are to be of the correct form and dimensions within the tolerances specified in Schedule T.ll. (rf) No tube is to have a mean thickness less than the specified gauge, or exceeding it by more than -004 in. except tubes thicker than -060 in., for which the tolerance is to be 1\% of their thickness. (Tubes ordered to be 22-gauge thick are to be -025 in. thick with a tolerance of + -004 in., to ugree with the dimensions set out in Schedule T.10.) () At no point in a tube is the thickness to fall short of the normal thickness by more than 10% or exceed it by more than 15%. (/) The tubes are to be as straight as possible, and in no part of the length is the departure from straightness to exceed one six- hundredth of the length of that part. li. HEAT TREATMENT. (a) All tubes are to be straightened and then blued. (*) Blueing. The tubes are to be heated to a temperature between 380 and 450 C. in a uniformly heated muffle, and after they have attained a uniform temperature throughout they are to be allowed to soak at that temperature for 20 minutes and then allowed to cool freely in the air, sheltered from draughts 7. OILING. The tubes are to be coated inside and outside with a suitable non-acid oil to preserve them from rust. 8, TESTS. (a) The tubes are to comply with the following 61 mechanical tests, which are to be carried out by the Contractor in the presence of the Inspector and to his satisfaction. (J) Tension and Comprenion Tests. Test pieces consisting of short lengths cut off the tube selected as specified in Clause 9 must give the following results, without further heat treatment or other manipulation : Ultimate Strength in tension not less than 35 tons per square inch. Yield Point in tension not less than HO tons per square inch. Yield Point in compression not less than 30 tons per square inch. (c~) Softening Test. This test is intended to prove that the tubes will not be unduly soft after they have been annealed, welded, brazed, or otherwise heated. (<7) A test piece consisting of a suitable length cut off the tube selected as specified in Clause 9 is to be heated to a full red heat at one end for at least five minutes while the other end remains cold, and is then to be allowed to cool freely in the air. The sample so treated must give the following results when tested in tension : Ultimate Strength in tension not less than 28 tons per square inch. Yield Point in tension not less than 18 tons per square inch. () Flattening Test. The tube is to be flattened at the end or at any point where defective material is suspected by a few blows (not more than six) till the sides are not more than three times the thickness of the metal apart. The tubes must stand this treatment without cracking. (/) Crushing Test. Samples of the tube selected as specified in Clause 9 are to be crushed endwise until the outside diameter is increased in one zone by 25%, or until one complete fold is formed. The Sample must stand this treatment without cracking. 9. SELECTION FOR TESTING. (.a) Five tubes in every 400 feet of ordinary sine or two tubes in every 400 feet of tubes larger than two inches in diameter or J inch in thickness are to be flattened as specified in Clause 8. (?/) Sufficient samples, selected by the Inspector, are to be cut from every 400 feet of tubes which have passed the flattening tests for the following tests : (c) One tension test and one crushing test (or two compression tests) and, when required by the Inspector, one softening test. (rf) The samples may be selected from any part of the length of the tube if inequality is suspected. The samples are to be marked as directed by the Inspector before they are cut off and are not to be annealed, hammered, or otherwise treated before they are tested, except as specified in Clause 8 for the softening test. EXTRACT FROM AIR BOARD SPECIFICATION T.2. FOR NICKEL-CHROME STEEL AXLE TUBES. NOTE. Axle tubes are seriously weakened where drilled. 1. MATERIAL. The steel is to be obtained from approved makers. A copy of all orders for steel billets or bloom? intended for manu- facture under this specification, stating the quality, brand, and limits of analysis of the steel ordered, is to be given to the Inspector on demand. The Contractor is to supply the analysis of the steel when required to do so by the Inspector. The Inspector may also select samples and have them analysed at the expense of the Government. 3. MANUFACTURE. The tubes are to be seamless and cold drawn. They are then to be air -hardened and tempered. 4. GENERAL CONDITION. The tubes are to be smooth, true to section, of uniform sectional thickness, and of equal diameter throughout, free from dirt, specks, longitudinal seaming, lamination, or grooving, both internally and externally. 5. ACCURACY OF FORM, SIZE, AND STRAIGHTNESS. () The mean outside diameter (t ., the mean between the maximum and minimum diameter) at any section is not to differ from the size shown in Col. 2 of the Schedule by more than 0-005 inch. (e) The ends of the tubes for a distance of 1 4 ins. from each end are to be rounded by pressure, so that no diameter exceeds the nominal diameter (Col. 1). ((/) No axle is to exceed the maximum specified weight (calculated by multiplying its length by the weight per foot given in Col. 5 of the Schedule). (e) At no point in a tube is the thickness to fall short of the nominal thickness by more than 10% or exceed it by more than 2(1%. (/) The tubes are to be as straight as possible, and in no part of the length is the departure from straightness to exceed one three- hundredth of the length of that part. 6. STRAIGHTENING. The tubes are to be straightened before heat treatment. After they are hardened and tempered the tubes are not to be straightened again till they have been tested under proof load, but are to be straightened, if necessary, after the proof load. 7. HEAT TREATMENT. (a) Harilenmr/. The tubes are to be heated uniformly to a temperature above (but not more than 50 above) the upper critical point and cooled at a uniform rate in air. (J) Tempering. The tubes are to be uniformly heated to a temperature between 200 and 250 C. and allowed to soak at that temperature for a short time, and then quenched. (Tempering should preferably be done in a bath.) 8. OILING. The tubes are to be coated inside and outside with a suitable oil to preserve them from rust. 9. MECHANICAL TESTS. (a) The tubes are to comply with the following tests, which are to be carried out in the presence of the Inspector and to his satisfaction. 'J>) Proof Load. Every axle is to be tested by having a proof bending moment applied to it near one end, and at least one axle in ten is to be tested in this way at both ends. .The proof load is given in column 7 and the leverage L is given in column 8 of the Schedule at the end of this Specification. Contact between the tube and the support at A is not to extend for more than f in. along the tube. (c) A suitable apparatus for applying the bending moment is shown diagrammatically in Fig. 41. The load is to be applied in two parts, the first part being any convenient amount between \ and J of the proof load. (d) After the first part of the load has been applied a gauge is to adjusted to a standard distance from the tube at B (Fig. 1) ; the second part of the load is then to be added and a'lowed to remain on for at leist 5 seconds and then removed. If any set is produced the tube will not return to the standard distance from the gauge at B. The set must not be more than T V in. Any tube giving a greater set will be rejected, but may be re-heat-treated. Accepted tubes are to be re-straightened if necessary. (See Clauses 5 (/) and 6.) (e) Tensile Test. One test piece to represent every 100 axles is to be cut from one of the tubes from which the axles have been cut ; it is to be heat treated with the axles it represents and is then to be tested in tension ; it must give the following results : Ultimate Tensile Strength not less than 85 tons per square inch. Elongation on 2 ins. not less than 5%. Elongation on 4 ins. not less than 3%. SCHEDULE. 3S7 4 +-005 339 1-24 138 96 i 25-0 1-50 1-49.5 14 . --000 I80 +-006 357 1-30 120 966 21-7 1-50 1-493 16 . n( .,--ooo U64 +-005 289 1-05 099 797 21-7 1-50 1-493 18 ., -000 I48 +-004 219 80 077 620 21-7 nfwi 1-10 1-095 14 .AQ(l~ UUU '+-006 256 93 061 807 13-2 62 EXTRACT FROM ADMIRALTY AIR DEPARTMENT SPECIFICATION T.5. FOR 50-TON CARBON STEEL TUBES. 1. MATERIAL. The steel is to be obtained from makers approved by the Admiralty. A copy of all orders for steel billets or blooms intended for manufacture under this specification, stating the quality, brand, and analysis, or, alternatively, the tests the material is to comply with, is to be shown to the Inspector on demand. Check analyses and tests are to be made on billets or blooms or on finished tubes when required by the Inspector, or, alternatively, the material for such analyses and tests is to be supplied. 3. MANUFACTURE. The tubes are to be seamless and cold drawn so as to remove all traces of the hot process and leave smooth surfaces inside and outside. The tubes are to be annealed at a temperature between 780 C. and 820 C. before the final pass, and to prevent brittleness are to be reduced in diameter as little as possible in that pass. 4. GENERAL CONDITION. The tubes are to be straight, smooth, true to section, of uniform sectional thickness, and of equal diameter throughout, free from scale, dirt, specks, longitudinal seaming, lamination, grooving or blistering, both internally and externally. 5. ACCURACY OF DIMENSIONS. The mean diameter of any tube is not to differ from the size specified by more than + '002 in. The mean thickness of any tube is not to be less than the specified gauge, and is not to exceed it by more than '004 in. in tubes thinner than '08 in., or by more than 5% in thicker tubes. Any variation of thickness due to eccentricity of the bore is not to exceed W/ of the specified thickness. 6. HEAT TREATMENT. All tubes are to be normalised 1 by heating to a temperature between 300 C. and 450 C. After the whole of the metal has reached a uniform temperature they are to be left for at least 10 minutes at that temperature before cooling. 7. OILING. The tubes are to be coated inside and outside with a suitable non-acid oil to preserve them from rust. 8. TESTS. The tubes are to comply with the following mechanical tests, which are to be carried out by the Contractor at his works in the presence of the Inspector and to his satisfaction. Tension and Compression Tests. A test piece consisting of a short length cut off the tube must give the following results, without further heat treatment or other manipulation. Normalised' Tubes : Ultimate Stress in tension not less than 50 tons per square inch. Yield Point in tension not less than 45 tons per square inch. Yield Point in compression not less than 45 tons per square inch. Softening Test. Additional tensile tests are to be made when required by the Inspector, to prove that the tubes will not soften unduly when brazed or otherwise heated. For this purpose the test piece, before it is tested, is to be heated to a full red heat at one end, while the other end remains cold, and is then to be allowed to cool. When tested the Ultimate Strength is not to be less that) 30 tons per square inch, and the Yield Point is not to be less than 18 tons per square inch. Flattening Test. A sample of the tube of length equal to its diameter is to be flattened till the sides are not more than eight times the thickness of the metal apart. The samples must stand this treat- ment without cracking. EXTRACT FROM AIR BOARD SPECIFICATION, T.14, FOR TEMPERED CARBON STIiEL AXLE TUBES. NOTE. Axle tubes art seriously weakened where drilled. 1. MATERIAL. The steel is to be obtained from approved makers. A copy of all orders lor steel billets or blooma intended for manu- facture under this specification, stating the quality, brand and limits of analysis of the steel ordered, is to be given to the Inspector on demand. The Contractor is to supply the analysis of the steel when .Vote. (>). .Vote. ( z ). " Blueing " is meant. " Blued " is meant. required to do so by the Inspector. The Inspector may also select samples and have them analysed at the expense of the Government. 3. MANUFACTURE. The tubes are to be seamless and cold drawn. They are then to be hardened and tempered. 4. GENERAL CONDITION. The tubes are to be smooth, true to section, of uniform sectional thickness, and of equal diameter throughout, free from dirt, specks, longitudinal seaming, lamination, or grooving, both internally and externally. 5. ACCURACY OF FORM, SIZE, AND STRAIGHTNESS. (a) The tubes are to be accurately circular to the satisfaction of the Inspector. (J) The mean outside diameter (i.e., the mean between the maximum and minimum diameter) at any section is not to differ from the size shown in Col. 2 of the Schedule by not more than 0-005 inch. (c) The ends of the tubes for a distance of 14 ins. from each end are to be rounded by pressure, so that no diameter exceeds the nominal diameter (Col. 1). (d) No axle is to exceed the maximum specified weight (calculated by multiplying its length by the weight per foot given in column 5 of the Schedule). (e) At no point in a tube is the thickness to fall short of the nominal thickness by more than 10% or exceed it by more than 2096. (/) The tubes are to be as straight as possible, and in no part of the length is the departure from straightness to exceed one three- hundredth of the length of that part. 6. STRAIGHTENING. The tubes are to be straightened before heat treatment. After they are hardened and tempered the tubes are not to be straightened again till they have been tested under proof load, but are to be straightened, if necessary, after the proof load. 7. HEAT TREATMENT. () Quenching. The tubes are to be heated uniformly to a temperature above the upper critical point, and are to be quenched in an approved manner. (i) Tempering. The tubes are to be uniformly heated to a suitable temperature, and allowed to soak at that temperature for a short time, and then either quenched or allowed to cool in still air. 8. OILING. The tubes are to be coated inside and outside with a suitable non-acid oil to preserve them from rust. 9. MECHANICAL TESTS. (a) The tubes are to comply with the following tests, which are to be carried out in the presence of the Inspector and to his satisfaction. (>) Proof Load. Every axle is to be tested by having a proof bending moment applied to it near one end, and at least one axle in ten is to be tested in this way at both ends. The proof load is given in column 7 and the leverage L is given in column 8 of the Schedule at the end of this specification. Contact between the tube and the support at A is not to extend for more than f in. along the tube. (e) A suitable apparatus for applying the bending moment is shown diagrammatica'lly in Fig. 41. The load is to be applied in two parts, the first part being any convenient amount between \ and { of the proof load. (rf) After the first part of the load has been applied a gauge is to be adjusted to a standard distance from the tube at B (Fig. 1) ; the second part of the load is then to be added and allowed to remain on for at least 5 seconds and then removed. If any set is produced the tube will not return to the standard distance from the gauge at B. The set must not be more than in. Any tube giving a greater set will be rejected, but may be re-heat-treated. Accepted tubes are to be re-straightened if necessary. (See Clauses 5(/) and 6.) () Flattening Test. After heat treatment a short piece, one or two inches long, is to be cut from one axle in ten and bent into an oval section by pressure, so that its diameter is reduced when under load to 0'85 times the original diameter. The sample must stand this treatment without cracking. The sample may be nearly severed from the tube before heat treatment while the tube is soft. (/) Destruction Test. At least one sample tube from each batch heat treated together is to be bent to destruction. The sample must be of sufficient length to test, but need not be the full length of an axle. (y) The sample is first to be tested under the proof load to verify that it will pass that test ; if it fail, another sample must be selected which will pass it. The sample is then to be further bent in the same apparatus by increasing the lever arm till the gap at B reaches the value given in column ! of the Schedule. It must stand this without cracking. If it fails, the batch it represents will be rejected, but may be re-heat-treated. (A) Tensile Text. Occasional tensile testa are to be made when called for by the Inspector. The tensile test pieces are to be cut from the undamaged ends of the samples which have undergone the destruction test. The Ultimate Strength, Yield Point, and elongation on 2 ins. and 4 ins. are to be recorded. SCHEDULE. 5 7 4-t ""aj Nominal Dia lal 11 Thickness. Min. Area o Section. '** s . fc. M i a ft Modulus of Section Z. Proof Load. Proof Load Leverage L Destruction Ti Min. Deflectii without Crac 1. 2. 3. 4. 5. 6. 7. 8. 9. ins. ins. ins. ins. 8 Ibs. ins. 3 Ibs. ins. ins. 2-375 2-362 120~' 0-850 3-12 0-460 2,080 34-65 1-5 2-165 2-152 .,o n --ooo 30 +-oio 0-831 3-05 0-399 2,080 30-0 1-5 2-165 2-152 ' 090 +-007 0-587 2-15 0-292 1,527 30-0 1-5 1-75 1-743 120~' -"+010 0-615 2-25 0-235 1,475 25-0 2-0 ,/ij in 1-75 1-743 .AQft UUU 90 +-007 0-4C9 1-75 0-185 1,160 25-0 2-0 1-75 1-743 072-' 000 +-OOfi 0-380 1-40 0-153 960 25-0 2-0 1-75 1-743 . O r )6 -'000 "' )b +-004 0-298 1-08 0-122 766 25-0 2-0 iAAA 1-50 1-493 90 +-007 0-399 1-46 0-133 960 21-7 2-n 1-50 1.493 072 ~' 000 0-323 1-18 0-110 796 21-7 2-5 1-50 1-493 56 +-004 0-254 0-923 0-0884 638 21-7 2-5 I'M 1-093 *'+:SS? 0-286 0-43 0-067 795 13-2 3-0 tiffing /?od in Guide* FIG. 41. DlAOBAM OP AXLE-TBSTINQ APPAKATUB. CHAPTER V. STREAM-LINE WIRES. No part of an aeroplane has caused greater anxiety or been more discussed tban the stream-line wires, with their fork-end fittings. Numerous tests have been made en them, but no satisfactory method has yet been revised for testing them under conditions similar to those ex- isting in an aeroplane, nor have any tests been devised by which reliable comparisons can be made between wires made by different processes. When fractures occur in rlight it is usually extremely difficult to determine their cause with certainty. All these circumstances combine to make the responsibility for the specification and use of stream-line wires onerous. Manufacture. The steel (for analyses see specifica- tion W. 3) is cold-drawn down to the size of the screwed ends after a special process of normalising known as patenting." Tin's process brings the carbon into the sorbitic condition, and makes the steel specially tough, so that it will stand a large amount of cold work; It is the same process which is applied to all rods used for drawing into wire for wire ropes. The amount of reduc- tion in section by drawing is arranged so that the round rod shall have the strength called for in specification. The rod is then cut into lengths and the middle part of each length reduced to the oval section by cold rolling. Considerable experience is required to determine the best diameter for the rolls, and the best series of grooves in the rolls to produce a finished oval of the correct form and strength; considerable skill is also required to roll the oval part to the specified length between the shoulders. The wire, instead of being drawn to the exact size for the screwed ends, is sometimes drawn to a rather larger size, and the ends are turned down before screwing. Also the blanks are sometimes heat-treated, by blueing at a temperature between 450 and 500 C. From the above description it will be seen that the manufacture of stream-line wires is difficult and slow, numerous suggestions have therefore been made for manufacturing the oval section in long lengths which could be cut up and have end fittings attached, but so far no method of attaching the fittings has been devised which is as strong as the oval section. The oval section owes a large part of its strength to the cold work done on it, and any method involving forging, welding, or bru/.ing reduces its strength greatly. The type of end fitting most commonly proposed is attached by bending back the oval section round a pin and drawing the 64 looped end into an oval socket. When tested this fitting always allows the wire to draw through at a load 20 to 30% lower than the ultimate strength of the wire. Stream-line Wire, Specification W. 3 The specifica- tion for stream-line wires has been repeatedly modified in details, but the following objects have been constantly aimed at. 1. The weakest part of the stream-line wire should be the oval, not the screwed end. The object of this is to provide the maximum " stretching length " in case of accidental overloading. It is generally agreed that fractures may often be avoided in cases of momentary overloading or shock if the parts are able to absorb the energy of the shock by stretching. 2. The oval part of the wire must not be rolled so hard as to become brittle. It is generally recognised that wires used for making wire ropes can be spoilt by over- drawing; this defect is revealed in them by the reverse bending or torsion tests. The reverse bending test is specified for stream-line wires with the same object. The amount of cold work allowed and the number of bends specified is a matter of judgment; no tests have been made to investigate the matter experimentally. The dimensions of the screwed ends being fixed by the dimensions of the standard threads, the ultimate strengths of the wires are determined by the ultimate strength of the round ends. The strength to which the round rod should be drawn and the proportion between the section of the oval and the section of the round end must be determined so that the above two require- ments may be complied with; this is by no means easy. The ultimate stress which the oval will bear increases as the section is reduced, so that the oval is apt to- finish stronger than the screwed end, notwithstanding its reduced size. Success in rolling the wires depends on using rolls of a suitable diameter, and reducing the section in a suitable number of passes. Heat-treating the blanks may in some cases facilitate manufacture, because it reduces the strength of the oval more' than that of the round end. Using rounds of a larger section than the screw thread makes the difficulty greater. The special methods of testing the wires is described in the specification. Heat Treatment. The blueing of cold-worked metal is for many purposes very advantageous and has been strongly advocated for stream-line wires. Considerable numbers of stream-line wires in service have been so treated, and many fests have been undertaken to prove their superiority over untreated wires, but none of these tests have shown that heat treatment is beneficial. The question cannot, however, be finally settled till a com- pletely satisfactory method of testing is devised. The latest specification forbids heat treatment unless special permission is granted to use it. A number of investigations have been made to deter- mine the effect of heat treatment on the properties of the wires as shown by the ordinary mechanical tests. Figs. 1 and 2 show the results of a series of tests made at the R.A.E. by Professor Edwards, Fig. 1, giving the results on the round rod, and Fig. 2 on the oval sections. Fig. 3 shows the results of a large number of tests made by Professor Goodman at Messrs. Brunton's Research Laboratory: these tests were made on the oval section only of wires of 4 B.A., 2 B.A., 1 jnch, 9.32 inch, 5-16 inch, g inch, 7-16 inch, and \ inch sizes; the results were similar for all sizes, and the curves give the mean results. Similar results were obtained by Professor Arnold at the Sheffield University. The curves show the well-known effects of heat treatment on cold-worked steel. The elastic limit (limit of proportionality) is con- siderably raised by blueing at temperatures between 350 and 500 C. The ultimate strength is not much affected till temperatures of 400 or 500 C. are reached, and then begins to fall off rapidly, at which temperature the elon- gation begins to increase. The effect of heat treatment on the number of reverse bends the wires will stand is shown in the following test made in Sheffield; somewhat similar results were ob- tained at Brunton's Research Laboratory. Reverse bend test on 1 * r incli stream-line wire. Gold rolled ... 13 reverse bends before fracture. Blued at 350 G. 12 ., ., 400 C. 8 4500 C. 8 n 500 G. 10 5500 C. 12 Tests made at Brunton's Research Laboratory show that ageing has an effect on the reverse bend test; a reduction of 25% in the number of reverse bends which the wires would stand occurred after a period of four months (June to October). This point should be further investigated. The raising of the Elastic Limit produced by blueing at temperatures about 400 G. might at first sight appear to be an advantage, but, as is explained on pp. 7 and 8, the elastic limit has really very little meaning, and is quite unimportant for wires used in tension. The effect of the blueing is merely to release the internal stresses in the wire, and there is no reason to suppose that these stresses do any harm. There is, on the contrary, much evidence to show that cold-worked steel is a reliable material; it is used for the wires in all wire ropes, for bicycle and motor car wheel spokes, and for many other purposes. i Three other tests have been made with the object of showing the superiority of heat-treated wires, but all have given negative results. The first was a series of tests on the Upton Lewis machine; they were made on the assumption that they represented the effect of vibra- tion on the wire, but this assumption is certainly not warranted. In the Upton Lewis test the material is repeatedly bent backwards and forwards sufficiently to give it a small permanent set each time; if vibrations ever produced sufficient stress to do this the life of the wire would be extremely short. Vibrations must be treated as a fatigue problem, and the power to withstand their effect depends on the fatigue range of the material ; the Upton Lewis test does not measure fatigue ranges, as is explained on p. 11. The results of the Upton l.c\vis test on the stream-line wires is shown in Fig. 4, from which it will be seen that blueing reduces the re- sults more and more as the temperature is raised. The second test was made in an Izod machine, but the samples had no notches cut in them, and the quan- tity measured was the energy absorbed to bend the sam- ple through an angle of 140. Two sizes of wire were .H SONOOd Ml QV Z o tt CO OJ u Q. O O k. o o Z 1,500,000 1.000000 500000 100 200 300 400 500 600 700 800 900 50 150 ?50 350 450 550 650 750 250 350 J UNTREATED BLUED ANNEALED ,- ,' . ,,"V4-). ilOQO C* B Cf> 27B. /a/19 tested, and the results are shown in the following table from which it will be noted that the heat-treated wires show no superiority. Impact bend tests on heat-treated and untreated stream- line wires. The figures given are the number of foot Ibs. absorbed in an Izod machine when the wires were bent to an angle of 140. The wires were not notched. Smaller wire. Larger wire. Un- treated. Treated. Untreated. Treated. Oval section 8 ft. Ibs. 8J ft. Ibs. Oval section 15 ft. Ibs. 11 ft. Ibs. ' i 8 16 144 , 8 , 7i , I5J 14 . i 8J , 6J , 13J 15 , 8 6 18 , 15J , 8 7* , 14 ., 16 , Round , >* 45 , Round section 88* { 111 98 The third test was Arnold's alternating stress test in which the result is expressed as the number of back- wards and forwards bends the sample will stand before failure; it does not differ much from the Upton Lewis test. A typical set of results were: test on J B.S F. stream-line Wire (oval 213 171 186 170 165 lgfl Arnold section). Cold rolled ...... Mean of 4 tests Blued at 350 C. ... ., ,. 4000 C. ... n 4500 C. ... 5000 C. ... 5500 C. ... ~ These tests also show no gain from blueing. These are distinct objections to the use of heat treat ment which makes it undesirable unless it can be shown to improve the wires; it introduces an additional pro- cess into the manufacture and it is not easy to make certain that the treatment of such long thin parts is uniform and constant. Vibration Tests. Many vibration tests have been made on stream-line wires, but in none of them liave the conditions been varied in the manner necessary to give truly comparable results. It is very desirable that the nature and conditions of the failure of stream-line wires under vibration should be fully investigated, but for this purpose the tests must be made on the lines of a fatigue test and the limiting load determined which the wire will bear when vibrating with a given amplitude. It may not be necessary to keep the speed constant the speed has not a great effect on the fatigue limit, and in actual use the speed will be the natural one for the load on the wire, but the test must be run to determine the limiting load which the wire will bear. Comparisons between the number of vibrations required to cause frac- ture under equal loads are entirely misleading such ratios have no relation to the ratios between the limiting loads the wires will carry. In making vibration tests the 27264 inertia of the attachments must be allowed for and accu- rate notes be made on the control of the ends of the wire exerted by the fork-end for it has not yet been ascertained with certainity how far swivel fork-ends allow the ends of the wiro free play. If plain fork-ends are used the amount of play they allow must be carefully measured. There is reason to believe that the breakages of the wires under moderate vibration may be entirely due to want of freedom at the ends, for lubrication of the swivel fork ends lias a very marked effect. A special end could probably be devised with a hardened ball to take the load, like Robertson's axial loading shackle, which would give perfect freedom and settle the influ- ence of other types of fork end. The shape assumed by the wire when at its maximum amplitude should be determined and the stresses due to its bent form calcu- lated. Direct tension and compression fatigue range tests should also be made on the wires for comparison. If tests are made on these lines it is probable that the nature of the failure of the wires under vibration will be elucidated and direct comparisons will be possible be- tween wires made by different processes and of different materials. Causes of fracture of Stream-line Wires in Service A very careful investigation made in April, 1917, into the fractures of wires in the air showed that they were due to a number of different causes : 1. Excessive vibration. 2. " Flapping " of loose wires. 3. Misuse of lock nuts. 4. Imperfect tuning of duplicate wires. 5. Overloads due to landing shocks. 6. Overloading in stunt flying. 7. Defects in the wires. 1. There is good evidence from pilots that vibrations of amplitudes of 4 to 6 inches are not uncommon in the air. Laboratory tests have shown that such large vibra- tions result in fractures after a comparatively short time; what the safe limits for vibrations are has not yet been determined. 2. There is ample evidence that stream-line wires break quickly if they are slack in the air. The type of vibration of a slack wire is quite different from that of a tight wire, and for distinction may be called " flapp- ing." No experiments have so far been made on flapping. All stream-line wires which are liable to be slack and to flap should be replaced by wire ropes. 3. Many fractured stream-line wires are found to have their threads stripped. This is often due to over-tighten- ing of lock nuts. A better method of locking stream-line wires is very desirable. 4. Duplicate wires are decidedly dangerous owing to the difficulty of insuring an equal distribution of load between them. 5. Anti-lift wires frequently break in landing. Exam- ples have occurred of these wires being partially broken on landing. If such partial fractures are not discovered, complete fracture may occur in the air on the next flight. It is not easy to distinguish with certainty whether a failure of an anti-lift wire has been due to landing shock or to slackness which lias led to flapping. 66 6. At the time when the report was drawn up it was not generally admitted that stunt flying could overload the wires sufficiently to break them. This fact is now universally recognized. 7. Defects in wires have been very rare, but several fractures have taken place in stream-line wires the cause of which may definitely be considered to be defec- tive material. Cracks may occasionally be found in stream-line wires sometimes running parallel and at other times perpendicular to the axis of the wire. The first class are generally found in the oval and the most usual size is half to one inch in length In some cases, however, the crack is in effect much longer as several of the above kind are only separated by short lengths of sound material. In general, it may be assumed that these cracks are not produced during the cold working, although, of course, they may be intensified by it. They are most likely the relics of some small surface crack in the hot rolled bar. and may be a rolling lap or may be due to defects in the original ingot or billet. These defects are generally not very deep although cases have been found in which they penetrated into the wire for quite one-third of its thickness. This type of flaw is not detected by the bend test, nor has it any influence upon the breaking strength of the wire. In actual use, how- ever, several instances have been found in which the \vires have broken through such cracks. The other type of crack runs perpendicular to the axis of the wire and is generally found along the edge of the oval. They are generally fairly shallow but quite sharp, and usually have been produced during the cold working of the stream-line section, the steel having burst at these points. These cracks are probably more dangerous than the longitudinal ones, but can usually be observed in the course of inspection. Instances have occurred in which wires have broken in flight through one of these cracks. Wires having this kind of crack very rarely pass the specification tests, but fail on the reverse bend test, fracture taking place through one of the cracks. CHAPTER VI. WIEE ROPES. Wire ropes are used extensively in aeroplanes, airships, and balloons, the very highest quality possible being required for towing kite balloons. (See specification W. 6.) This report only deals with the ropes used in aeroplanes; of these there are two classes, known as " straining cords," used for standing rigging, and " extra flexible," used for control wires. The less flexible strain- ing cords were introduced early in the war to facilitate manufacture; the extra flexible ropes are suitable for standing rigging, but cost more and take longer to manu- facture. The smaller sizes of extra flexible rope are made of very small wires, which require great skill to manufac- ture ; to facilitate manufacture the four-strand type was introduced to replace the seven-strand type in the smaller sizes, thus avoiding the use of the smallest wires, but fatigue tests on ropes running round pulleys show that the four-strand ropes wear out considerably sooner than the seven-strand ropes, as was to be expected. It is specially necessary to use large pulleys for the four- strand ropes. On the other hand, the four-strand ropes have one definite advantage over the seven-strand. The seven-strand ropes are formed of a single strand in the centre surrounded by six strands; if a piece of seven- strand rope is accidentally given a few twists, so as t < twist the outer strands more closely, they shorten and leave the core slack; the core then frequently pushes out between the outer strands and forms a dangerous kink, which may jam in a guide or pulley. Accident have occurred from this cause, and kinks have given considerable trouble, for the kink frequently does not appear at once, but only after the rope has been used for some time. Testing Wire Hopes and Single Wires. Special wires, a special clump is used, ns shown in Fig. 1. the testing machine. The following methods have been found the most satisfactory by the A.I.D. : FIG. 1. 1. J''or small wires, i.e., up to 1-32 inch diameter for high tensile wires, or up to 1-10 inch for low tensile wires, a special clump is used, as shown in Fig. 1. Chapber.VI. V \ COMPLETE BENDS TO FffoCTURE & TENSION ON CABLE. - 500,000 NUMBER Or COMPL.E.TE BEt>K36. 29KS/944- 2000.C & R.ir-27 12/19. fIGURES IH INCHES DENOTE 0/AMETERS OF PULLEYS. oo< HJOISN3J. u Chapfcer.VI CO d LU -J CQ UJ Q: u. o a z UJ 00 Q (U h- Q. UJ a: 319 VO NO NOISN30. -OVQ1 3JVS KIO NOISN3J. Chapfcer.VI. FiG.7 TENSION ON CABLE TO BREAK AFTER 1.OOO. OOO CYCLES. i o 8 o 8 o a REPEATED BENDNG OF WIRE 1 CABLE OVER PULLEYS. DIAMETER Of PULLEY & / / LOAD TO PERMIT 1,000,000 COMPLETE BENDS. 20a.25cwr. 7 x 19 FLEXIBLE CABLE!. > ,/ */ 1 5 / / f 8 3 4 DIAMETER OF PULLEY. 5 C INCHES 67 The plate A. is fixed in the ordinary jaws of the testing machine; the wire is fixed in the clip C., and given two turns round the fixed groove wheel E. 2. For larger wires the ordinary wedge grips are satis- factory, but the serrations on the grips should not be too prominent. The ends of the wire should project beyond the grips, or if this cannot be arranged the end of the wire should be tapered off on a grindstone. 3. For small cables (up to 20 cwt. ultimate strength) the ends may be whipped with wire and soldered, and held in the ordinary wedge grips, or they may be splayed out, tinned and cast into conical sockets with a low melting-point alloy. 4. For larger cables the ends should be cast into sockets, as above; for very large cables the ends of the wires should be bent inwards to form hooks before they are cast into the socket. Fatigue of Rope running round Pulleys. Figs. 2-7 show the results of a series of tests made by Capt. Scoble at University College. They show what a small load ropes will carry when running round pulleys even of comparatively large size, and the importance of using large pulleys. Fig. 2 shows the results of tests on four-strand ropes of 5 cwt., 10 cwt., and 15 cwt. sizes tested round 2 inch, 3 inch, 4 inch, and 6 inch pulleys. Fig. 3 shows similar results for seven-strand ropes on 1 inch, 2 inch and 3 inch pulleys. Fig. 4 shows similar results for seven-strand ropes of 20 cwt. and 25 cwt. sizes. Fig. 5 shows the results given in Fig. 2 plotted to show the effect of the size of pulley (100,000 leversals). Fig. 6 the same (100,000,000 reversals). Fig. 7 shows the results given in Fig. 4 plotted as in Fig. 6. Elasticity of Ropes. Loaded ropes stretch elas- tically rather more than a solid steel wire of the same strength. The elasticity of ropes is of importance from several points of view. Comparing ropes with stream- line wires it is important to know their relative resilience, i.e., the amount of energy they will store, the larger the resilience the greater the safety when the machine is subjected to sudden shocks. Again, the elas- ticity of the rope is of importance in tuning up the aeroplane; it also partly determines the range of adjustment needed in turn-buckles, etc. The elastic stretch is also important in control wires. For practical purposes the most useful comparison to make is between the elastic strains per pound load of ropes and stream- line wires of equal strengths. Measurements have been made on a few sizes with the following results : Strain per Ib. load of 1/19 wire rope .. , ... Ditto of S.L. wire of same nominal strength ~~ Ditto for 10 cwt. flexible 7)7 wire rope = 1-9. Ditto for 20 cwt. flexible 7/19 wire rope = 2'2. In each case the comparison was made between ropes and stream-line wires of the same nominal strengths, i.e., supplied to specification calling for the same ulti- mate strengths. It is probable that the wire rope has a larger margin of safety above the specification than the stream-line wire and it is estimated that the above ratios would be increased about 15% if the ropes and wires had actually the same ultimate strengths. The resiliences of wire ropes and stream-line wires are in the same ratios. CHAPTER VII. ALUMINIUM. CONTENTS. 1. Wrought Alloys. 2. Duralumin Struts. 3. Cast Alloyp. 4. Testing (including high temp. Brinell tests). 5. Fatigue Ranges. 6. Preparation of Alloys. 7. Founding. Appendix. Specifications. 8. Poron* Castings. 9. Burning of Pistons. 10. Steel liners in Aluminium Cylinders. 11. Aluminium Crank cases. 12. Bearing metals. 13. Welding. 14. German Cast Alloys. At the beginning of the war there were a large number of proprietary alloys of aluminium on the market, but there was very little reliable information about the pn - perties of any of them except duralumin. As the use of light alloys increased the difficulties of obtaining suffi- cient supplies of pure aluminium became acute, and it became necessary to study the effects of the impurities in the different alloys. The results of tests on the cast alloys were very erratic, and manufacturers were un- willing to have material rejected on the basis of tests which could not be relied on. A committee was formed to investigate these problems and was subsequently taken over by the Advisory Committee for Aeronautics as their Light Alloys Sub-Committee; this sub-com- mittee has carried out a large amount of research work, 27264 mainly on cast alloys, and has published a series of short reports summarising the more important results arrived at. Their work was not confined to laboratory investigations, but included a large amount of work in foundries and tests on engines, both at the R.A.E. and at engine builders' works. Reports Nos. 2, 3, 4, 7 and 9 are reproduced on pp. 72 to 76. For the remainder see Advisory Committee Reports. The Air Board Specifi- cation for Aluminium Alloys (except those for dura- lumin) are based on the results of the sub-committee's work. The aluminium alloys may conveniently be divided into two classes: the wrought alloys (including rolled extruded, drawn and forged alloys) and the cast alloys. I 2 68 1. WROUGHT ALUMINIUM ALLOYS. By i'ar the most important wrought aluminium alloy during the war has been duralumin. It has been very extensively used in the manufacture of the framework of rigid airships in England and in Germany. Next in importance is pure aluminium, which has been used extensively in sheets for cowlings, seats, panels, etc., in aeroplanes; it has not been used for parts requiring strength. Three other alloys have been worked out at the N.P.L., and suggested for practical use, viz., the -3/20" alloy (3 Cu, 20 Zn), the N.P.L. High-Tension Alloy (2$ Cu, 20 Zn, Mg, $ Mu), and the Copper- Nickel-Magnesium Alloy (4 Cu, 2 Ni, 1| Mg), but manu- facturing restrictions have delayed their production on a commercial scale during the war ; they are now being developed. A fairly satisfactory alloy called Aeromin was being developed before the war, but has not been used much during the war. Duralumin. Duralumin is an aluminium-copper- magnesium alloy, usually containing about 4% copper and \% magnesium. This alloy responds to heat treat- ment, probably owing to the presence of magnesium; the usual heat treatment is to quench the metal from a temperature of about 480 C. in water, and then to allow about a week for the " ageing " to take place; the effect of this treatment is to increase the ultimate strength about 50%. The specified minimum ultimate strength after heat treatment is 25 tons per square inch, with an elongation of 15%, but these figures are usually con- siderably exceeded. By heating it to 380 C. Duralumin is softened, and can then be easily 'vorked cold. The American Bureau of Standards has carried out elaborate tests to determine the effects of various heat treat- ments; the conclusions indicate that for high tensile strength the best temperature to quench from is between 510 and 515 C., and that the quenching bath should be between 125 and 135 C., and that ageing should take place at the same temperature. The heat treatment causes serious warping, which introduces considerable manufacturing difficulties. It is just as important in the workshop to avoid heating duralumin parts which nave been heat-treated as it is to avoid heating heat- treated steel parts. Duralumin corrodes rather badly and its use in aeroplanes has bc x 5 x 27 mm. 90 Notch. 3 mm. radius, 0-t mm. depth. TABLE 2. N.P.L. HIGH TBNSILE ALLOY (Density 3-1). Tensile Test. Fatigue Range. Notched Bar Test. Condition. Ult. Strength, i Yield Point. Elongn. At 20 C. At 150 C. At 20 C. At 130C. A l 240 C. Quenched from 350 C. tns./sq. in. j tns./sq. ir. 37-8 32-9 % 11-6 tns./sq. in. 9-7 tns./sq. in. 5-1 kg.m. o-aj kg.m. 0-44 kg.m. Tensile Tests an Sheet 0-041" thick. Condition. Ult. Strength. Yield Point. Elongn. As rolled Quenched from 400 Annealed at 250 tns./sq. in. 33 39-7 25-7 tns./sq. in. 26 30-9 14-7 % 18 11 20 Young's Modulus 9-8 X 10 s . 32/20 ALLOY (Density 3-1). Tensile Tests. Fatigue Range. Notched Bar Test. Ult. Strength. Yield Point. Elongn. At 20 C. At 150 C. At20C. Atl50C. tns./eq. in. 26-6 tns./sq. in. 16-8 % 22 tns./sq. in. 8-7 tns./sq. in. 4-5 kg. m. kg. m. 0-53 0-66 Young's Modulus 9'5 X 10 6 Ibs./sq. in. Modulus of Rigidity 3 '75 X 10 6 Ibs./sq. in. TABLE 3. COPPEK NICKEL MAGNESIUM ALLOY. Tensile Tests mide at various Temperatures on Heat-treated Samples Cold. 100 C. 150 C. 200" C. Yield Point. Ult. Strength. Elongn. Ult. Strength. Elongn. Ult. Strength. Elongn. Ult. Strength. Elongn. tns./sq. in. 14-27 tns./cq. ins. 24-27 % 24-0 tns./sq. in. 22-4 % 21-5 tns./sq. in. 21-4 % 23-0 tns./sq. in. 19-48 % 2J-0 Fatigue Range. Notched Bar Test. At 20 C. At 150 C. At 20 C. At 150 C. At 240 C. tns./sq. in. 10-2 tns./sq. ip. 8-4 kg.m. 0-26 kg. m. 0-24 kg.m. 0-15 AEROMIN. Tensilt Test on, Sheet. Condition. Ult. Strength. Yield Point. Elongn. % 4-5 2-5 Hard Rolled ... Annealed tns.'sq. in. 29 23 tns./?q. in. 22-6 Fatigue Range at 20 C. + 6-7 tns./sq. in. The sample tested for Fatigue Range contained 6-15% Mg. and 0-88% Fe. 70 German Wrought Alloys used in Zeppelins. A large number of samples have been secured and tested. The bulk of the framework in most Zeppelins is made of Duralumin, but occasionally a " Zinc Duralumin " is used, with a composition: Zinc, 9%; copper, 0-6%; magnesium, 0'7%. In the thin sheets used for the framework this gives : Ultimate strength ... ... 26'5 tons sq. in. Elongation 17'0% The rivets are usually pure aluminium, but some have been found containing 1'3% of copper. A single sample " plug " was found, which is made of a magnesium-base alloy, with a composition : Mag- nesium, 91-2%, aluminium, 2-4%, zinc, G'2%. The density was 1'84. Tested by a modification of the Brinell method this sample appeared to be " weaker than Duralumin." Uses of Wrought Aluminium Alloys Some early accidents due to the corrosion of the Duralumin used in aerofoil construction have greatly retarded the employment of wrought aluminium alloys but metal spars are now being built experimentally of Duralumin, and are lighter and stronger than timber spars. It appears to be worth while to experiment with wrought alloys for connecting rods, bolts, nuts, bearing caps, pistons, piston rings, and many small parts of engines. The comparatively high fatigue ranges of some of the alloys suggest that they may be found satisfactory. Suggestions have been made for using extremely thin high-tensile alloy sheets a few thousandths of an inch thick instead of linen for covering wings, the sheet being stretched tight and not depending on its own stiffness. Corrugated aluminium alloy covering for wings has been used in Germany, but there appears to be no prospect of success in that direction. 2. DURALUMIN STRUTS. The following report by Professor Lea, on Duralumin struts, was published in C.I.M. 720 It includes valuable information on the secondary failure of angles, channels, etc., and gives data necessary for the design of spars : The memorandum summarises the results of com- pression tests on elements of Duralumin. They are shown plotted in the figures, the failing stress per inch being plotted in all cases as ordinates and the ratio of the length to the least radius of gyration, that is, l/k, as abscissas. When l/k is large, say more than 120, the points lie very near the well-known Euler curve, which is shown plotted in all cases for the equation P = rr 2 El. AZ 2 . the value of E for Duralumin being taken as 4,800 tons per square inch. Extensometer and bending tests on suitable specimens show that the value of E varies some- what : the figure stated is probably a little higher than the mean. Cross sections of the elements are given on the figures ; and details of thicknesses, least radius of gyration, etc., are shown in the tables. The load was applied in some cases through steel balls and in other cases through knife edges, the centre of the steel balls and the knife edges being in all oases placed as nearly as possible in the axis of the specimen and in the plane of the axis respectively. it is interesting to note that the specimens failed in two ways, which are referred to as primary and secondary. For example, if the angle H, Fig. 1, is considered, it will be seen that the thickness is 0-118" and the width of the flanges nearly 1". The ratio of the width of the angle face to thickness is there- fore less than 9. In this case all angles tested failed by primary failure, that is, they bent into a continuous curve from one end to the other about the axis of minimum radius of gyration. Similarly the angles P, Fig. 1, failed by primary bending. On the other hand, the angles L and K failed primarily when the value of l/k were greater than 90 and 70 respectively. For values of l/k less than these the angles not by bending as a whole, but by the lips buckling, and the strut apparently twisting. The section N, Fig. 7, showed lip buckling when l/k was 41, and section E, Fig. 6, when l/k was 68. By comparing H and P with L and K (Figs. 1 and 2), it will be seen that the point of secondary failure depends upon the ratio of the width of the angle face to the thickness. As long as this is less than 10, except in very short specimens, secondary failure is not likely to occur. The same phenomenon can be illustrated by bending tests of the lip channels, the loading being applied as in Fig. 10s. Stress-deflection curves for such tests are shown in Fig. 10. When the channel is so loaded that the lips are in compression, the apparent elastic limit is about 14 - 5 tons per square inch, and the lips buckled when the apparent stress, as calculated by the formula p = My/I, was 17 tons per square inch. On loading so that the lip was in tension the beam was loaded until this apparent stress was more than 30 tons per square inch. A study of the curves shows at once that in using such section for building up the flanges of struts or beams, economies can be effected by carefully choosing the sections, and in all cases where elements of struts or girders are in compression, by a careful consideration of the unsupported length. For example, suppose it is required to build up a strut having angles at four corners connected by suitable bracing. Suppose the value of l/k for the strut as a whole is 50; then, if angles similar to H, Fig. 2, are used, the failing stress, if the strut is carefully loaded, and the angles are braced together at such distances apart that they are formed into bays of about 9" to 10" long (that is, 50xl'94" = 9-2), will be from 10 to 12 tons per square inch. On the other hand, as seen from the curves, it would be practi- cally impossible, however close the bracings are put together, to obtain such a stress from a strut made of the angles L or N, Figs. 1 and 2, in which the ratio of thickness to width of angle face is 17'5 and 20 respec- tively. For values of l/k greater than, say, 150, it would appear unimportant which section is used. In Fig. 9 are shown plotted the results of tests on Duralumin tubes. For values of l/k less than 100 a, higher stress is necessary to cause failure in a tube than in any other section. As a strut the tube form may, therefore, be very economical: large values of k for a given amount of material cannot, however, be obtained. On reference to Figs. 3 and 4 it will be seen that " lipping " the edges of the angle is not very effective if the thickness of the angle is less than one-twentieth of the width of the face. Chapter. VM. ienffh of 'Strut divided by Least Radius of Gyration. O I O 10 2O 30 O SO Ml Tn Kn \ 1 \ Vark "oint or 'jirri 9imenj/onJ in Inches tit ti xt DeYtlc/fd Width d tv" Sect! 4rfi w rad ffy-ation Inches \ . 'n 1C s; \ K* N P H L Fie . I. 0*7; 8S7J 9*7^1 203S 197 1372 09S7 o 9*-9* 'OS 1-0 -10, -OS 1-3* r3*-iZS S '-S*-Of /7.S- I-9S 2-475 55 i. i \ \ y ^ o 1 IS t \l \ V 1 \ f s y 1 i\ \ 1 \' \ \ c i\ 1 1 1 \J \* i * An rlr 75 \ \ \ ,\ \ \ o k \ 1 \ v 4 I t \ ^ r $ y V t X \ s. s M s ^ yj>< ;s s *"' -* -N. ^ - "> M X \\ "** -. ^ M V \ h N * J i \ \\ "r \ Kj 1 Jv FIG a. N X X X x FIG. 2. X ^> . i-- >* -, s ^ r^l -.. ^ t I "-"*: - --i. 1 < Length of Strut divided by Least Radius of Gyration. Length of 13 II \ \ =i \ I )j I-1S I-7S ns OSIS M7S \ f \ \ D > I -s \P \ \ v V '* \ \' 1 i \ v V CtK 3 it -\ /\ /\ & /\ ^*'- ^^f ^L|5 -'Xj*' y' x^/ ^*^ / Sections C.O&K. 9 8 7 5 4. a L 1 \ S A "iA \ \ ^ \ S S ^ /i N V \ E \ s ^ s x N ^, |8\ \ - ^ ^ ^ '--- ^ \ V 4_5 X N N X x \ X X ^ Sections A& B " - ' - 1-v ~ <^. ^^ ^ //ff 4, -. -^ ~-^ * ** *** n v- * KJ 10 30 40 SO 60 7O 80 90 XX) MO l?0 I JO 140 ISO 160 HO 180 ISO J-ength of Sfrut divided by Least- Radius of Gyr . C* K Chapter. VM. IO 20 4O SO 100 120 140 i 1 1 1 r i [ ( i 1 1 1 1 i r - Teste of Duralumin Channel Sections AS Free-end Struts ^ loaded through Centroid bv means of 3 /b diam Balls J. i i I t\ 1- i i t MMt Pwrtt Curve T QtmtnsionJ tnetiel Otrrkftd d w ' Sro! Stln. Hast m g il 4J SectiondE C. I 6 x > ]*. \ \ V- I V \ v^ \ ^ V n V t, V ^^ * ^ X _ F IG: . |[. X v^ s ^^ | j '-N k. i i 10 20 30 <4C SO 60 70 80 30 KM 110 120 130 Length of Strut divided by Least Radius of Gyration, in ZQ 30 40 50 60 70 M SO MO 110 130 UQ ISO 160 170 180 ___L Tests of Duralumin Channel Sections as Free-end 3trufa loaded through Centroid by means of Knife - ec/fe nd Pieces. i I 1 ': \ \ J ^ u * \ \ \ .\ It \ \ Hbrk faint mSOiaviSm en f Angle Curre\ e* "' Diptti rh,clr- ness f" fneltfttl Width Area Lent Kad. inches . It V V T e F G H *- Q X 36-S 3S 40 38 ass l-s.i 1-291 668 rs4 662 oses 012 ~ 3-22 Z;38 Z-35 Ot3t 1454 1081 ZI6 243 23* 21 * .\ 1 | S r \t 1, * S fl i \ k X ^ [ W f Section . * J_ . .\ * \ * . \ V > t> \ ^ I> ^ ^ S v^< t i FIG: 6. o ^ s. ^ ^f* ^^^. f^-%l^^K Ma- 1 i , 1 I * 1 10 30 40 M 60 70 *0 90 tOO 110 120 13O I4Q ISO I6O I7O :80 Length of Strut divided by Lrast Radius of Gyration. Chapfcer.VII. Length of Strut divided by Least Radius of Gyration. O 10 tO 30 50 60 "70 80 90 MO IZO 130 140 150 160 170 180 190 JO 20 30 40 SO 60 TO 80 90 IOO IK} divided by Least Radius of Gyration . 2000. c& R.I QhapberVll 71 In Figs. 6, 7 and 8 other channels are shown. For economy of material the ratio of d to the thickness should evidently not be more than 10. In the case of the channel I, Fig. 7, it is about 17, whereas for M the ratio is less than 10. The curve for the channels J, K and L, Fig. 8, lies a little below that for M, the mean ratio of d/t for the former being slightly greater than for the latter. In designing the elements of the compression flanges of spars, three considerations will have to be taken into account : (1) The value of the minimum radius of gyration if the element is free to bend about that axis. (2) The thickness of the element. (3) If the spar as a whole is in compression the radius of gyration of the spar considered as a whole. Thin flat surfaces, or a thin free lip of an element are to be avoided. Similarly, in designing the compressive elements of braced spars, thin flat bars are quite unsuitable. The minimum value of the radius of gyration must be con- sidered and the thickness of the material at the free edges and on the flat surface is important. In Fig. 10A, which is the section of the bracing element that lias been largely used for members in certain light structural work, it is desirable that the free edges A and B shall be very nearly on the neutral axis of the section. The stress on the free edge due to bending is then a mini- mum, and secondary failure is not so likely to occur. For struts the load may be applied non-axially and in such a way that the lip is in tension. The section will then be much more stable than when the lip is in compression. The tensile strength of Duralumin is given in Air Board Specification L.3 as 25 tons per square inch. This figure can be relied upon and higher breaking strengths are to be expected. Duralumin spars have been designed, in which care was taken that secondary failure should not take place, which did not fail on test under direct thrust and bend- ing when the stress in the compression flange was 20 tons per square inch. 3. CAST ALUMINIUM ALLOYS. The most important parts made of cast aluminium alloys are : Crank cases, cylinders, pistons, pump bodies and carburetters, but many other small parts are also made of them. Enormous development has taken place in aluminium-alloy casting during the war, and most of the work of the L.A. Sub-Committee has been devoted to the cast alloys. When the Sub-Committee began their work it was anticipated that new alloys would probably be found superior in their properties to any of the numer- ous alloys in existence, but subsequent experience showed that little improvement could be made on some of the recognised alloys. A certain number of these were there- fore selected and the best proportions for the constituents determined and the numerous inferior variations rejected- The mechanical and thermal properties of these alloys has been fully investigated over a wide range of tempera- tures. Accurate methods of testing have been deter- mined and numerous foundry difficulties have been investigated. Experience has shown that, speaking generally, more depends on skill in the foundry and the use of suitable methods than on the exact alloy used. A grent deal of research has been done on the influence of impurities, particularly iron,* the proportion of which in aluminium ingots steadily rose during the war. Most of the purest ingots were required for making Duralumin, so that the lower grade material had to be used for the cast alloys. The results has been to show that the pre sence of iron, though it adds to the difficulties of making sound castings does not very seriously affect the mechani- cal properties of the cast alloys. Suggestions for the use of many different alloying metals have been investi- gated, but none have been found useful except zinc, copper, manganese, magnesium, nickel, and possibly tin. Vanadium, silver, mercury, cobalt, antimony and others have been found useless. It would be an endless problem to test every combination of metals, but it is believed that a sufficient number have been tested to make it improbable that any new alloying metal will be found which will give better results than those already in use. Light alloys, consisting mainly of magnesium, have been investigated, but are disappointing. Aluminium bronze (copper + 10% Al.) is discussed under the heading of Copper Alloys. The principal aluminium alloys suitable for casting are: Specification L.5 Alloy. Zinc between 12'5 and 14-5% Copper 2-5 3-0% This alloy has proved thoroughly reliable for general castings. It is the most suitable alloy for crank cases and carburetters. It has been largely used for pump bodies, but a less corrodible alloy is wanted for them, and trials are being made with the copper-nickel-magne- sium alloy. A record of 80 consecutive casts in the chill of L.5 alloy gave a mean ultimate tensile strength of 13'6 ton per square inch. (Only 11 tons is at present specified as the minimum ultimate strength.) This alloy is not very suitable for die castings. Specification L. 8 Alloy. (Copper between 11 and 13%. The rest aluminium.) This alloy has proved thoroughly reliable for pistons. It is more suitable for die casting than L. 5. Specification L. 10 Alloy. Copper between 9 and 11%. Tin between 0'5 and 1-5%. Aluminium the rest. This alloy is slightly more ductile than L. 8., and has proved thoroughly reliable for cylinders. Specification 2. L. 11 Alloy. Copper between 6 and 8~%. Tin between 0-5 and 2'0%. Aluminium the rest. This alloy originally contained 1% of zinc in addition, and was largely used in the R.A.F. It is a good deal softer and more ductile than L. 8, but also a good deal weaker, particularly at high temperature, neverthe- less it has been successfully used for pistons and cylinders. Copper-Nickel-Magnesium-Alloy. Copper between 3'5 and 4'5%. Nickel between 1'75 and 2-25%. Magnesium between TO and 1'5%. Aluminium the rest. * See Report No. 8. 72 This alloy is being tested in practice, but has not yet come into general use. The specified minimum ultimate strength is 11 tons. Laboratory tests show that it resists corrosion better than any other aluminium alloy, and trials are being made on pump bodies made of it. It is more difficult to make in the foundry than L. 5, L. 8 or L. 10. Magnalite. This name is usually given to an alloy containing : Copper, 2 to 2$%. Nickel, 1^ to lf%. Magnesium, 1 to 2%. It is similar, but inferior to the copper-nickel-magne- sium alloy already described. Lynite The samples of this alloy which have been tested have differed widely in composition. The original alloy, said to give good results for pistons in America, contained 2J to 3% of copper, 1J% Magnesium, and 1J% of iron. It is therefore similar to magnalite, with iron instead of nickel. One of the latest samples tested was practically the same as specification L. 8, i.e., a 12% copper alloy, and another contained 8% copper and 1% iron. The mechanical and thermal properties of these alloys are given in the L.A. Reports Nos. 2, 3, 4, 7 and 9, which are reproduced below. (For Reports Nos. 1, 5 and 6 see Advisory Committee Reports.) Report Number 2 deals with the strength of different alloys at high temperatures. The effect of low tempera- tures (down to that of liquid air) has been investigated, and it was found that no harmful effects were produced. (see Report No. 6. ) Report Number 3 deals with the contraction during solidification in the mould. The different alloys differ very little in this respect, and difficulties which are experienced in the foundry cannot be attributed to dif- ferences in the " contraction." The same patterns may be used for all the alloys. Report Number 4 deals with the thermal expansion of the different alloys, their growth and distortion. The growth of the castings is an important point when ex- treme accuracy of dimensions or good fits are necessary. Report Number 1 deals with the thermal conduc- tivity of the alloys. Most of the alloys have about the same conductivity, but the manganese alloys are deci- dedly inferior. The high thermal conductivity of aluminium alloys is a very important property, and is the main cause of success of aluminium pistons. Report Number 9 deals with the eopper-nickel-magne- sium alloy. L.A. REPORT. No. 2. TENSILE STRENGTH OF ALUMINIUM ALLOYS AT HIGH TEMPERATURES. For aluminium cylinders and pistons it is desirable to use an alloy which retains its strength at high tempera- tures. The actual maximum temperatures reached in cylinders and pistons are not yet known with certainty. The barrel of the cylinder in a water-cooled engine pro- bably reaches about 100 C., and in an air-cooled cylinder has been found to reach from 180 to 220 C. The maxi- mum temperature in the head or near the exhaust valve is between 250 and 300 C. Aluminium pistons occa- sionally reach the eutectic inciting point, 530 C., but such temperatures are certainly abnormal*. .Measurements of the strength of different alloys at high temperatures have been made at the National Physical Laboratory and the Royal Aircraft Factory, and by Prof. Lea. The results differ among themselves con- siderably, depending partly on the exact method of cast- ing the samples and possibly on the impurities present. but the characteristic way in which each alloy behaves as the temperature rises is fairly definite. The accom- panying eight figures show average results for sand cast and chill cast samples. In Figs. 13, 14, 15, 16, 19 and 20 the ultimate strength of the sample is plotted against the temperature at which it is tested. In Figs. 17 and 18 the effect of the per- centage of manganese in the Cu.-Mn. alloy is shown at various temperatures. These manganese alloys have the remarkable property of rising in strength with the tem- perature to about 250 C.f Fig. 13 shows the variation of ultimate strength with temperature for the simple Cu.-Al. alloys. The 12% Cu. alloy is the alloy specified in Air Board Specification L. 8. The samples were cast in chills. Fig. 14 shows corresponding curves for the Cu.-Mn. alloys, both sand and chill cast; the 12 per cent. Cu.-Al. alloy curve is reproduced from Fig. 1 in this and in other figures for comparison. The 14 per cent. Cu. 1 per cent. Mn. alloy is the alloy specified in Air Board Specifica- tion L. 7. The remarkable rise in strength referred to above is clearly shown. These curves may be compared with curve 2 in Fig. 14, obtained by Professor Lea. Fig. 15 shows the corresponding curves for: (1 and 2) Cu. 7, Tin 1, Zinc 1. (Known as the R.A.F. alloy.) (3 and 4) Cu. 8, Zinc 15 (approximately the same as Air Board Specification L. 5). The weakness at high temperatures of these alloys is marked. Fig. 16 shows similar curves obtained by Professor Lea. The addition of small quantities of zinc and tin to the manganese alloy lowers its strength, but not excessively. Figs. 17 and 18 show that at all temperatures the strongest copper-manganese alloy is that containing 1 per cent. Mn. Figs. 19 and 20 show the variation of strength with temperature for various alloys containing copper with other elements, viz., nickel, chromium, cohalt, vanadium id iron. Several of these are promising. L.A. KKl'OHT No. 8. LINEAR CONTRACTION OF ALUMINIUM ALLOYS DURING SOLIDIFICATION' IN THE MOULD. Accurate measurements have been made at the National Physical Laboratory on the linear contraction of aluminium alloys in the mould. Two methods were employed which gave results in close agreement. The most satisfactory method con- sists in casting a bar 10"xl"xJ" in a sand mould between the faces of a steel template, which form the * Subsequent investigation has shown that this statement is doubtful. t The thermal conductivity of these Cu.-Mn. alloys is unfortu- nately considerably lower than that of some of the other alloys, which may more than counterbalance the advantage of this rise in strength. Chapter.WI. Flo. 13.14,16,16,17; 16.19. ana SO. i3. i*. HIGH TEMPERATURE TENSILE TESTS, is. 16. i x n n \ n Ml L l N L \ 1 Nl L to fm i ;*'! E, 4>p mn I 1 I; s 4 2 1 n tt 9 1 1 t> A ) * | Si Ws \ , t . -- s K .1 t 1 t n s \ \ t \ . , B-N - s N V i i z 1 II K) ^ _^ Kl ~- 4*-. (^ *x 5 J * 3- \ *-- 5 S * I~ ^ ^ Ce \ s ^ feii > N Sj ^ ..- , / ^ A s "--> -- iV \ \ \ \ s ^ \ *\ \ i -s, * t T* x ^ C \ Ri -1. 'Offer H, MMsntKl. t t~ * H. /. . -J- * . /. 4. A /. 1 i 1 wu W Qm &CM ^ l.'dfpar?. til. 2>K/. Cm -z- I */. i. m 3- " 3. VnclS. Cm -* .. CM II II r^ i a~ t* J- j (4 Uirllitt H. > /. CyfwW. , !>/. 71 1 1 l r//T rrz inr N of of of taf or aaf str o CfMl'IGRAOl 17 jo' IDC' 06' of ad 1 #d" 350' MfflMfW 18. ^0" JOtf JStf" HJtf MO' JOC" J50" CtTI6**i>C 19. 50" lOfl" Off K1C" WO 1 300" 350" CtHTHiMOC 2O. 1 Kl L K N i L H ^ Kft> Nf >L NF L H K ^ "" MMV 'ftp fJV fwcu \ ^ ^ * 9 R ^ ^ =# -^ s. fl r ^ ^ ^ s* 2 \ \ ^ \\ 'MM// 'I it >V ,^-- / 2 \ ^, zs k ^ 1; I 8 i: a i* *^~ \ V "" ^**i ^z. i,./4 ^ fe JS V- ^ " s* X rV \, 5 /- \ ^ s *3 4 7 "v 1 & CHIU CAST BAA 14% Copper f*ct of Waryinf Per of Msnvtnemo f. Cent I * fl ?> /.-CtpperHMii -t- n W. i i i 1 kel -i- * A- - .;. -* - . If Lf L \ I -f-' MX Copper rtct of Varying PtrCtn of Maaganetf. i ill I. t IfU /Ok tvn* ./ ft 1 ' Vffffiai Iron 1. <;; IJTi ul 1 1 1 1 1 .. J O 0-S Ml I'S /*/ crr. o t i ' w ' H o w wo- wx* ^ "r MANGAIV CCPtTISPAOC ft* CtKT. eetmcMOf C* A. 1.^0.27*. Chapter. VI I. \ in Sr, ! \ - '&AIQ queue 73 ends of the mould and measuring the difference in length between the cast bar and the template when they are cold. A number of different alloys have been cast in this way, and give the following results, which are in each case the means of several tests : Percentage Composition. fl CO Alloy. 'a a III o ""' C a | d a H 'a a ,3 a 8 2 * o 3 |P ho i"" __ ^_ _ 15 _ 85 1-42 1/69 2 10 88 1-40 1/70 " L.5 " 2-75 13-5 83-75 1-27 1/78 3 15 82 1-25 1/79 " 8/1 " 1 8 91 1-31 1/75 " 14/1 " " L.7 " } 1 14 85 1-21 1/82 " 7/1/1 " " B.4 " }- 7 1 1 91 1-19 1/83 10 90 1-25 1/79 " L.8 " 12 88 1-25 1 /79 10 1-25 88 '75 1-22 1/81 12 1-5 86-5 1-25 1/79 See also Report No. 9. L.A. KEPOETNo. 4. THERMAL EXPANSION GROWTH AND DISTORTION OF ALUMINIUM PISTONS. Aluminium pistons have been observed to increase permanently in size during use. This permanent in- crease is called growth. The cause of the growth has been investigated by the National Physical Laboratory, the Royal Aircraft Factory, Professor Lea, and Dr. Shakespear. When an aluminium piston is heated it first expands, as most metals do, but the coefficient of expansion is very large. After reaching a temperature of about 250 C. the increase in size becomes more rapid, and continues for a considerable period though the tempera^ ture remains constant. This increase is the growth, and is more or less permanent. At the same time the piston usually alters in shape, probably due to the release of casting strains. Such alteration in shape is called distortion. It is also usually more or less permanent. If the piston is cooled down again it contracts regu- larly, and at the initial temperature the piston has regained its initial size except for the permanent growth and distortion. Subsequent heating and cooling usually proceed in a normal manner, but the growth and distor- tion may be slightly modified. Special attention is directed to the large thermal ex- pansion, which probably accounts for some of the seizures which have occurred with aluminium pistons. "When estimating the clearance required between a pis- ton and cylinder, allowance must be made for the changes due to all the three causes mentioned above, but the thermal expansion is in most cases much the largest. An example will make this clear. It is not certain what temperature aero-engine pistons reach under normal working conditions, but it is probably fairly high.* Temperatures of 250 C. have been mea- * Nate. See results of actual measurement* quoted on page 86. 27264 sured on the cylinder wall of an air-cooled engine, and the piston is probably at least 100 hotter, and may be considerably more. If we assume these temperatures and take the coefficients of expansion for the aluminium piston and steel or cast iron liner as 26xlO~ 6 and 10 x 10~ 6 respectively, the difference in expansion, for a 4-inch diameter piston, will be 26 thousandths of an inch. The growth of the piston will probably be less than 4 thousandths, and the distortion probably less than 4 thousandths. These figures may be compared with the difference of thermal expansion between a cast iron piston and the cylinder under the same conditions, which will only be 4 thousandths of an inch. These figures show that very much larger clearances are needed with aluminium than with cast iron pistons. Professor Lea's experiments were made on pistons of various alloys. The maximum distortion observed was + '016 inch on one diameter, accompanied by '009 inch on another; it is usually much smaller. The growth never exceeded - 0032 inch; all these measure- ments being made on a 4-inch piston. Dr. Shakspear's experiments were made on bars the length of which was recorded continuously. The bars were first heated slowly and the increase of length due to expansion was measured. A typical curve is shown in the lower diagram, Fig. 21. It will be noted that this curve is not straight and that the rate of increase in length rises with the temperature. This is due to the growth. After reaching 260 C. the temperature was kept constant and the growth observed and plotted against time. The resulting curve is shown in the upper diagram, Fig. 21. After 320 minutes the growth had practically ceased and the specimen was cooled again ; the contraction is shown in the lower diagram. On sub- sequently reheating the sample no further growth occurred and the expansion curve coincided with the contraction curve both being practically straight. Dr. Shakespear's curves show that the coefficient of expansion for 12 per cent. Cu, Alloy (A.B. L. 7) is 26-4 xlO-", and for 14 Cu. 1 Mn. alloy (A.B. L. 7) is 25'9 x 10" 6 . The maximum growth observed was '0006 (i.e., '0024 inch on a 4-inch piston). At the National Physical Laboratory one sand cast and three chill cast pistons were tested in an alloy of nominal composition aluminium 85, copper 14, man- ganese 1.* An old experimental sand cast piston of composition approximating to the Air Board L. 5 alloy (3 per cent, copper, 15 per cent, zinc) was also tested. The heat treatment involved exposure to a temperature of 350 to 370 C. for two periods of three hours with intermediate cooling, with a final exposure of 42 hours; alternate heating to 350 C. and cooling as rapidly as possible a number of times; exposure to 450 for three hours and afterwards for twelve hours. Measurements were made after each period, being taken across four diameters at 45. The change in mean diameter did not in any case exceed 2-3 thousandths of an inch. In every instance an initial growth was followed by a con- traction, the final mean growths ranging from O'l to 0-6 thousandths of an inch. The E.A.F. have made two tests on pistons running in an engine. In the first test, using a piston of 71 : 1 : 1 * On analysis the sand cast piston was found to contain only 0-26 % manganese. 74 the growth amounted to '002 inch after 40 hours. In the second test, using a 14 :1 alloy, the growth amounted to 0044 inch after 55 hours. Professor Lea has made a series of tests to ascertain whether distortion and growth can be eliminated by a preliminary annealing of the pistons. His results show that if the pistons are rough turned all over and annealed at 340-360 C. for five hours, subsequent heating pro- duces changes which do not exceed - 0015 inch on a 4-inch piston. It is, however, doubtful whether anneal- ing is worth while, as the changes which occur in practice are not usually enough to cause trouble. It will be seen from Dr. Shakespear's curves that the permanent growth is considerably less than the ordinary thermal expansion at a temperature of 260. L.A. REPORT No. 7. THERMAL CONDUCTIVITY OF ALUMINIUM ALLOYS. Probably the most important advantage of aluminium alloys over cast iron and steel for cylinders and pistons is their high thermal conductivity. This advantage be- comes more marked at high temperatures because the thermal conductivity of aluminium alloys increases with temperature, whereas that of cast iron, steel and other metals decreases. Measurements of the thermal conductivity of alloys have been made at the National Physical Laboratory and at the Royal Aircraft Factory. At the National Physical Laboratory the thermal conductivity has been determined up to 300 400 C. for aluminium and its alloys and up to 700 C. for cast iron (Table 8), by an absolute method in which the heat transmitted is measured by a flow calorimeter. Tem- peratures along the rods are obtained by a series of iron-constantan thermo-j unctions. The rods are enclosed in magnesia-asbestos lagging to diminish the heat flow through the sides, and appropriate corrections are applied for the heat transmitted through this insulating material. One end of each bar is wound with a heating coil of Nichrome ribbon. The electrical input is measured and serves as a check on the method. Observations have been made for various temperatures of the hot end, the cold end being maintained at about 40 C. through- out. The values are given to two significant figures only, since different specimens of the same nominal composi- tion do not give results concordant with a higher degree of accuracy. Over 100 conductivity measurements made by this method are given in the Tables below. The alloys have in all cases been prepared and analysed in the Laboratory under carefully controlled conditions. The composition of the aluminium used in making the alloy is given in Table 6 In the chemical analyses of the alloys the added metals only have been determined; the aluminium is obtained by difference. Small quan- tities of iron and silicon, in proportion to the amounts found in the aluminium (Table 6), are present. At the Royal Aircraft Factory the methol used was the usual one of heating a uniform rod at the ends and measuring the temperature at three equidistant points after the rod is in temperature equilibrium with its sur- roundings. Rods of various alloys were compared with a rod of some metal such as zinc or aluminium taken as standard, whose heat conductivity was assumed to be given with sufficient accuracy in the books of reference. The results in general are in agreement with those obtained at the National Physical Laboratory, with the exception of that for pure aluminium; this discrepancy is being further investigated. The present report contains the results of the work at the National Physical Laboratory only. An inspection of the data given in the Tables shows that : (1) All the aluminium alloys have a conductivity less than that of pure aluminium. (2) Their conductivities increase with the tempera- ture, while that of pure aluminium is prac- tically constant. (The conductivities of cast iron and zinc decrease with temperature.) (3) There is very little difference in the conduc- tivities of the chill cast and sand cast specimens, the latter possessing a very slightly lower conductivity. (4) Annealing up to a temperature of 450 C. has a beneficial effect on the conductivity in practically all cases. This is particularly marked in the alloys of aluminium with copper and manganese. In comparing the results given in the tables for the " as cast " and annealed states, it must be remembered that the heating up of one end of the bar causes some annealing of the metal during the course of the tests. The conductivities of aluminium and zinc (Tables 6 and 7) were determined for the purpose of checking the accuracy of the methed. In Table fi are given the results of some experiments on cast iron. In Table 9 some of the published data on the thermal conductivities of aluminium, cast iron, steel and zinc are given for comparison. Tables of Thermal Conductivity. Thermal conductivity in the following Tables is given in terms of gramme calories conducted per square centi- metre per second across a slab 1 centimetre thick, having a temperature difference of 1 C. per centimetre. To obtain conductivity in terms of pound calories conducted per inch per second across a slab 1 inch thick having a temperature difference of 1 C. per inch the values must be multiplied by 0'0056. TABLE 1. MAGNESIUM-NlCKEL-CoPPER- ALUMINIUM. Composition by Analysis. As Cast. Annealed (450C.). Per cent. Sand. Chill. Sand. Chill. Mag- nesium. Nickel. Copper. | r~ o O g CO 1 o cs o CO 8 O 8 O 8 o 0^ ?! O O 1-06 2-33 8-04 34 39 39 40 1 *2 9-08 - 33 35 37 - 38 39 39 1-45 1-91 7-96 3J 36 37 37 38 1-6 *2 *8 34 35 36 39 39 40 * Nominal content. Increase on annealing approximately 5 to 10 per cent. 75 TABLE 2. MANGANESE-COPPER-ALUMINIUM. Composition by Analysis. As Cast. Annealed (450C.). Per cent. Sand. Chill. Ohill. Man- ganese. Copper. 100 200 300 100 200 300 100 200 300 0-98 13-9 26 27 29 27 28 SO 37 37 38 0-98 8-0 25 27 29 25 27 30 36 38 39 Increase on annealing approximately 26 to 30 per cent. TABLE 3. NlCKEL-CoPPEK-ALUMINIUM. Composition by Analysis. As Cast. Annealed (450C.). Per cent. Chill. Chill. Nickel. Copper. 100 200 300 100 200 300 1-12 7-93 38 39 40 41 41 42 2-16 9-04 36 37 39 40 40 41 2-86 7-90 38 39 40 42 42 43 1-96 11-87 35 38 39 39 2-87 12-06 34 35 37 38 39 40 Increase on annealing approximately 8 per cent. TABLE 4. iRON-CorPER-ALUMINIUM Composition by Analysis. As Cast. Annealed (450C.). Per cent. Chill. Chill. Iron. Copper. 100 200 300 100 200 300 0-82 7-91 35 : -37 38 40 41 42 2-09 8-18 34 -35 36 38 38 39 Increase on annealing approximately 10 per cent. TABLE 5. MISCELLANEOUS ALUMINIUM ALLOYS. Composition by Analysis. Per cent. As Cast. Annealed (450C.). Sand. Chill. Sand. Chill. oo v> t) "> Oj * ^. \ \ Z z ,0/fNOJL 't*)US c\i O LL > o 1 K Q j -J u ^c (A O p 1 Q *^^ -J Ul 1 i J Ul K ,D/SNOH SS3VJS IQ <*> -*>* Chapter. VII- o LL c OB N U it % K * o o \ *0 a h ^: IO o j. CO N 5 ^ \ 8 d 8 & >- C 0. (0 w u K J 4 i j (V d (O Z c Q. W f|- c\l 6 cp o U (L CO 6 C'v NX SNOJ. Nl SS3M1S Q H3d SNOJ. Nt S93U19 i: 1 UJ in i CO C) u a. CO CHILL CAST TESTED COLD ALLOY L8 CLASTIC LIMIT 1-8TO o> CM N 0) en U C o> ! X in O Z O -1 u YOUNG'S MOO. 5.000 DOTTED LINE SHOWS ^ 3 O fe t- B: u TO ENLARGED SCALE Ml I -i 4 1 u o . \ \ \\ AI X. .a Mid SNOi. Nl SI1U1I cy 6 i t Ul . > ul ^ fS K V. J o 00 o 5 ZJ in ^ o O U CO r " en U. in CO d U 0. CO CHILL CAST TESTED AT ALLOY L 8 ELASTIC LIMIT OLT: STRESS* V- Z 3 w DOTTED LINE ! LOWER PART C TO ENLARGED a M3d SNOi. Nl SS3M1S 77 advisable to limit the nickel content to 20 per cent : with this composition the alloy has a melting point of approximately 780 C. The temperature required in the final stages of preparing this alloy, just before casting it into ingots, is about 830 C. By using a 10% nickel alloy hardener these tempera- tures can be considerably reduced. The melting point in that case is about 680 C. and the temperature used need not exceed 730 C. In the preparation of the T,iloy itself the aluminium is first melted and then carefully skimmed. The nickel and copper hardeners are then added together, or tho copper hardener may be added first. (The copper hardener is the usual 50% copper alloy). The magnesium should be added last, just before pour- ing. It is used in the form of pure metal, preferably in sticks about 1 inch diameter which may be conveniently ?ui into short pieces. Before adding the magnesium the metal must firNt be skimmed and the magnesium slumid then be stirred into it under the surface. A convenient appliance for doing this consists of a small inverted cup, with perforated sides, which is attached to the end of a long iron rod, the other end of which may be provided with a wooden handle. Such a cup can be conveniently irjade of a small salamander crucible. If the magnesium is merely held below the surface of the alloy in tom,s, pieces are very apt to break off, float to the surface and there burn away. 4. TESTING CASTINGS. 'I he following instructions have been issued in T.D.1.M.-3: The strength of castings depends considerably on tbe rate of cooling of the metal in the mould; the faster it cools the stronger it becomes, so that chill castings are considerably stronger than sand castings, and th'.n parts stronger than thick parts. In sand a inch plate is three or four tons per square inch stronger than a 2 inch plate. As the strength of the casting depends on the rate of cooling, one test piece cannot represent all parts of the casting, and it has been decided to cast samples to represent the alloy used, rather than the casting, and it has been agreed to use as a standard sample a rod 1 inch diameter, cast in an iron chill. The chill is to be between 7 and 9 inches long, and the lower end is pre- ferably to be closed with a sand or clay plug not with a metal end. The chill is to be well warmed before the metal is poured into it. It is to be held at about 45 with the vertical while being filled, and tipped upright when full. 'I he test piece is to be turned from the middle of the sample to the form shown in specification L. 5. The shape of the shoulder in common use is too abrupt for such a brittle material. The ordinary methods of gripping the test pieces be tween wedges does not insure that the load will be Hpplifd axially, so it is necessary to have the ends of the test pieces turned to a shoulder or screwed. These precautions are generally sufficient to insure accurate results with full-sized test pieces of alloys having 3 or 4% elongation (such as L. 5), but are net sufficient for smaller test pieces, or for alloys such as are used for pistons, etc., which have only about 1% elongation. For these it is necessary to use axial load- ing shackles, such as those designed by Lieutenant liobertson. (See Fig. 1, Chapter I.) With these special shackles accurate results can be obtained, even with very minute test pieces of brittle material. Elastic Limit of Oast Alloys. A very large number of stress/ strain diagrams have been drawn for the different alloys : typical examples are reproduced in Figs. 22A 22s and 22D. In all cases the primitive elastic limit is very low, usually between 1J and 2J tons/square inch. It can be raised by stretching the test piece, as is shown by the loop in Fig. 23, which shows that the elastic limit has been more than doubled by a very small strain. This method of raising the elastic limit is probably made use of in cylinders with shrunk-in liners. The manner in which the stress/ strain curve changes as the temperature rises is shown for L. 8 alloy in Figs. 22c, 22o, 23, 24, 25 and 26. Similar curves have been plotted for many other aluminium alloys at high temperatures. The changes which occur are of the same nature in all, but the rate at which the alloys lose strength with the rise in temperature varies considerably (see L.A. Eeport No. 2 above). Notched Bar Tests. A great many notched bar tests have been made on the cast alloys at various tempera- tures, but no meaning has been suggested for the figures obtained, and it is not considered necessary to include the results in this report. Brinell Hardness Tests Professor Lea has carried out a full investigation into the hardness of a number of alloys at various temperatures. One of his reports to the L.A. Sub-committee is reproduced here: ALUMINIUM ALLOYS, BRINELL HARDNESS NUMBERS AT HIGH TEMPERATURES. These experiments have been carried out in the hope that they would indicate a rapid and reliable method of judging the relative suitability of the various aluminium alloys for pistons and other engine parts subjected to high working temperatures. The following alloys have been tested up to the pre- sent, both chill and sand cast.* Specification. L5 L8 L10 LI I Cu. 2 Su. 1 Zn. Nominal Composition. (per cent.) 13-5 Zn. 2-75 Cu. 12-0 Cu. 10-0 Cu. 1-25 Sn. 7-0 Cu. 1-0 Sn. 1-0 Zn. 9-0 Cu. 2-0 Sn. 1-5 Zn. The tests have been carried out at 15, 150, 200, 300 and 400 degrees Centigrade. Owing to the. large variations which occur in these alloys the following precautions had to be taken : (a) All tests were made on specimens cut from 1 inch diameter test bars in order that the results for the different alloys might be rea- sonably compared. (b) In investigating the hardness of each particular alloy all the specimens tested were cut from the same bar. This eliminated, as far as possible, variations due to differences in com- position and manufacture. * AUo \% Cu., 2 Ni , I ( Mg. Set end of table, p. 79. 78 (c) In each alloy, wherever possible, two specimens were tested at each temperature, and the average of the figures thus obtained was taken as the hardness number. By this means it was possible to detect and discard low or erratic figures due to mechanical faults, such as small blow holes. Where these extra specimens were not available, however, the second impression had to be made on the reverse side of the specimen. Whenever this was done the fact has been noted in the accompanying table of results. It will be noticed that in most cases the duplicate figures thus obtained agree very well together. The specimens which as stated above were all cut from test bars were about 1 inch diameter and |-inch to f-inch deep. The flat surfaces were polished to eliminate tool marks and scratches and thus facilitate the accurate measurement of the impressions. The tests were carried out in a small electric furnace which had been specially designed for use with one of the Briuell testing machines made by the firm of Aktiebol, Alpha, of Stockholm. The general arrange- ment of this furnace is shown in Fig. 27. The details are not however drawn accurately to scale. The specimens to be tested were placed on the steel base of the electric furnace, which was then heated to the required temperature. This temperature, with a varia tion of not more than two degrees Centigrade in either direction, was maintained m every case for eight to twelve minutes before the test was made in order to make sure as far as possible that the specimen was always uniformly heated throughout to the required tem- perature. As it usually took about five minutes to make a test, change the specimen, and bring the furnace back again to a steady condition, it was possible to test four specimens per hour when the furnace had once been heated up. The temperatures (as indicated in Fig. 27), were mea- sured by an iron-constantan thermo-couple, the bare end of which was allowed to touch the specimen. The couple was connected to a direct reading instrument on the scale of which every ten degrees was marked tod each degree could easily be estimated. While the specimens were being heated the furnace was allowed to stand on a sheet of asbestos on the bench. As soon as they were ready, the furnace together with the asbestos sheet was transferred to the table of the machine and the test made by bringing the steel ball in the machine into the depression on the top of the plunger in the furnace and loading this by means of the oil pump in the ordinary way. The pressure was thus transferred to the second steel ball attached to the bottom of the plunger, as chown in Fig. 27. The ball which made the impression was therefore always in the furnace and at the same temperature as the specimen at the time of testing. In every case a 10 mm. diameter hardened steel ball was used with a load of 500 kilos, and the Brinell hard- ness number was calculated from the diameter of the resulting impression. It has already been shown in a previous report that reliable results are obtained in tests at 15 C., when the load is applied for 10 seconds, and th? 1 if this period is increased there is no corresponding uniform increase in the diameter of the impression. As a result however of a lew preliminary tests at 300 C., it seemed probable that the period of application of the load would bt: of considerable importance at higher temperatures. To investigate this therefore a few tests were first made in which the loads werj applied for periods varying from 10 to 70 seconds. The figures obtained are given in Table 1, and are shown plotted in Fig. 28. It will be seen that the total variation in the hardness number is fairly small, while the variation due to an error of 2 or 3 seconds in the loading period is quite negligible. In all subsequent tests therefore the load was applied for 10 seconds. The complete results are shown in Table 2, and are plotted in Figs. 29, 30 and 31. In order to show what variation may reasonably be expected in the same alloy, two bars of L. 10, chill cast, have been tested. The two resulting curves, which lie quite close together, have been plotted in Fig. 29. For convenience in comparing the various alloys the curve for U.S. P. 35 (L. 10, chill cast) has been plotted on each figure. TERMINALS FIG. 27. It will be seen that on the whole, the higher the per- centage of copper the harder is the alloy at all tem- peratures irrespective of the other metals present in the alloys in small quantities. The four alloys which are highest in copper show a considerable variation in hard- ness at 15 C., but these variations rapidly decrease as the temperature rises, until at about 400 C. there is very little diference between them. L. 5. how- ever, the high zinc alloy, though one of the hardest alloys tested at 15 C., loses its hardness with very great rapidity as the temperature increases. (Fig. 29.) It will be noticed, 'moreover, that L. 5 seems to lose its hardness most rapidly in the first 150 when compared with the alloys containing a higher percentage of copper. Most of the sand-cast bars tested, as noted in Table 2, showed the characteristic small pin holes or air holes that are often to be seen on the machined surfaces ot cylinder-blocks and other castings in these alloys. It is Chapter. VH. O L- O CO CO U Z CO U ft CC U 0, z U O FIG. 28. SPEC. L10 CHILL CAST TESTED AT 300*C. 10 20 70 40 DURATION OF APPLICATION OF LOAD SO SPEC L11 CHILL CAST TESTED AT 500*0 NUMBER 30 20 c * . " MMMi 10 tO 30 40 50 U RATION OF APPLICATION OF LOAD 80 70 & ?9I6S/*.2000.CA. H. CHAPTER VH. i b. o co cn u M3QWnN S33NQWVH CO U. O CO CO u a: z LJ U 2J3BNON SS3NQMVH 79 noticeable that in two cases (Fig. 29) the sand-cast bars give an indication of the resistance of any new alloy to were harder than the chill-cast bars in the same alloy. high temperatures. One interesting phenomenon was noticed. When test- ing the alloys containing both zinc and tin small globules of metal were found on the polished surfaces of the impressions made at 300 and 400 C. These were pro- bably minute globules of the zinc-tin eutectic squeezed out of the alloy. Conclusion. The results on the whole seem instruc- tive. They indicate the general similarity between the alloys which are at the present time in general use, and clearly point out the superiority of the high copper alloys over the high zinc alloy (L. 5) for all castings H.S.P. 25 which are exposed to high temperatures. The results moreover show sufficient uniformity to justify further \y M c A similar tests on other alloys at present in use, and to warrant the assumption that such tests would rapidly TABLE 1. BHINELL HARDNESS NUMBERS. The Effect of Duration of Load at 300 C. Mark. Specifica- tion. 10 sec-'. 30 sees. 60 gees. 70 sees. I.8.P. 25 L10 Chill. 27-1 I 27^, 27-8 f 27 ' 27-6/ 28 ' 2 2I ' 5 \24-S 23-8 / 24 S iV.M.C.A. Lll Chill. 28-2 1 ,. 27-0 C" 26 ' 5 i2fi-0 25-4 f 26 !} TABLE 2. BRINELL HARDNESS NUMBERS AT HIGH TEMPERATURES. Mark. Specifi- Actual cation. Analysis. Load 600 Kilogrammes. Brinell Hardness Numbers at 16 C. 150 C. 200 C. 300 C. 400 C. H.S.P. 25 Chill. L. 10 9-72 Cu. 1-25 Sn. 61 Fe. 62-8 1 V62-2 61-8 } 48-9 ^i U8-9 63-7 J 27-1 1 \27-f, 27-8 J 11-8 1 J-ll-5 11-1 J H.S.P. 35 Chill. L. 10 10-03 Cu. 1-24 Sn. 61 Fe. *fi2-2 -v \ 63-0 S63-7 j 59-5 1 \ 59-5 59'5 J 50-3 ) U9'9 49- J 26 1 V2ti-4 26-7 J 10-8 -i U04 10-0 J H.S.P. 8 Sand. L. 10 10-41 Cu. 65-3 -I U3-8 62-2 ) 61-8 -I Ul-7 61-5 J 56-0 ~| US' 7 51-3 J 27-6 1 >27'7 27-7 J 11-3 1 I- 11-2 11-0 J W.M.C.A. Chill. L. 11 7-61 Cu. 1-50 Sn. 1-46 Zn. 60-2 -j U9-8 59-5 ) 58-7 -v U8-7 58-7 J 47-6 -v V46-2 44-8 J 28-2 1 ^27-6 27-0 J 12-7 1 M2-2 11-6 J 74 Fe. W.M.S.A. Sand (c). L. 11 6-95 Cu. 1-45 Sn. 1-66 Zn. 56-0 1 V54-6 53-2 } 54-9 -I V53-5 52-0 J 41-5 "i U3-3 45-0 J 24-8 -\ 1 86-0 25-1 J 11-1 -v UO-8 10-4 J 58 Fe. M. 6 Chill L. 5 3-01 Cu. 13-01 Zn. 61 Fe. 59-5 -i J-61-6 68-7 J 43-0 -1 Ul-9 40-7 J 32-1 -i Uo-1 3S-0 J 165-1 M57 14-i. J " L. 5 Sand O). L. 5 3-26 Cu. 13-35 Zn. 70 Fe. 80-4 -I 179-2 78-0 } 44-3 ^ . H3-8 48-3 J 35-2 1 \-S5-9 3-5 J 14-7 1 U4-8 14-9 J K. 6 Chill. L. 8 12-17 Cu. 70 Fe. *804 -v Ul-7 *8S-0 } J70-8 -i V71-7 J72-5 J J63-7 -i U>3'4 *63'0 J J33-8 1 U4-7 J35-6 J J13-5 -i [13-0 J12-5 J J. 5 Sand O 1 )- L. 8 12-13 Cu. 69 Fe. J68-9 1 J-OT-0 &G5-0 J 69-0 1 V69-0 69-0 J 59-0 1 V57-9 50-7 J J32-3 1 U3-2 &34 J J15-6 1 [14-4 *13-1 J H. 7 Chill. 9 Co. 2 Sn. IJZn. 9-32 Cu. 2-15 Sn. 1-42 Zn. 61-8 1 V62-2 62-6 J 61'0 1 V57-8 54-5 J 47-1 ) Us-i 49-1 J 27-4 1 V26-7 2'i-O J 9-7 -I } 9-8 10 J 50 Fe. G. 7 Sand 00- 9Cu. 2 Sn. 1J Zn. 9-01 Cu. 2-14 Sn. 1-53 Zn. 60-2 1 V60-4 60-6 J 56-6 -> V55-1 53-6 J 49-7 -> V48-3 46-8 J 26-8 -i V26-3 25-7 J 10 5 1 VlO-4 10-2 J 62 Fe. Notes. (a) Figure discarded as obviously unreliabls. (J) Both impressions at the same temperature made on the same specimen, one on each side, (c) Whole bar full of very minute blow or pin holes. ADDENDUM JULY 1918. Sand 4Cu. 4-00 Cu. 71-3 61 45-7 17-2 Chill 2Ni. 2-12 Ni. 76-2 73-2 67-6 52-7 liMg. 1-56 Mg. 80 5. FATIGUE RANGES. Investigations on the fatigue ranges of aluminium alloy castings have only recently been begun. The following reports by Professor Lea show that the fatigue ranges are well above the primitive elastic limits, just as they are in hardened (untempered) steel; the explanation is pro- bably the same, namely, that the low elastic limits are due to internal stresses arising during the cooling of the metal. It should be noted that the tests were made at the unusually high rate of 4,000 r.p.m., and were interrupted by periods of rest, but it is not known whether the results are affected thereby. Another point of interest in these reports is the evidence they contain that test pieces are actually strengthened by being sub- jected to alternating stresses below their fatigue range. This fact is very suggestive, and is well worth further investigation. If confirmed it will have an important bearing on the methods employed in making fatigue range tests, as it indicates that different results may be obtained if the limiting stress is approached from above or from below. REPORT ON WOEHLER FATIGUE TESTS ON ALUMINIUM ALLOY L. 8. In order to ensure as far as possible a uniform com- position throughout 'the whole series of tests here recorded, the bars for both the Tensile and the Woehler tests were cast at the same time from the same pot of molten metal. Pieces of sand cast and chill cast specimens in this series were analysed with the following results : Chill Cast. Cu 12-00% Sn Nil Fe -69% 05% 38% Zn Si Sand Cast. 12-55% Nil 82% Trace 37% The stress strain curves obtained in the tensile tests are shown in T'ig. 32, and the results are also recorded in tabular form in Table I. The strains were measured by a 2 inch Ewing extensometer. It will be noticed that the small pin-holes present in test bar D.l.C.2. do not seem to have lowered the elastic limit though they have appreciably affected the ultimate stress and elongation. The results of the Woehler tests are recorded in Table II, and are shown plotted in Fig. 33. The tests were carried out in a specially designed machine in which the test specimen is a rotating cantilever carrying a known weight at the end. The shape of the specimen (see Fig. 34) is so designed that it should break at a particu- lar place, and where the specimen completed a test without breaking the stresses were calculated at the point where fracture should have occurred. Owing however to imperfect machining and lack of uniformity, some test bars did not break at the desired place, and in that case the stresses were calculated at the actual point of fracture. All tests were carried out at an average speed of 4,000 revolutions per minute and vibration during the tests was hardly noticeable. During the tests it was not found possible to run the specimen continuously until fracture occurred. Each test was therefore divided approximately into continuous running periods of about nine to ten hours followed by periods of rest of about fourteen hours. During these periods when the machine was not running the weights were allowed to remain hanging on the specimen. It will be seen from Fig. 33 that specimens which ran for one or more tests without breaking tended to give points slightly above the line when they were finally broken. This is shown by specimens C.I.C.2., C.I.S.I., and C.l.S.4. It should further be noticed that specimen C.l.S.4. was allowed to run for over 15 millions at a stress not quite high enough to produce fracture, and was then allowed to rest for seven complete days before it was again tested at a higher stress, as can be seen from Fig. 33, this point is well above the asymptote. This rather suggests that a large number of reversals at a low stress followed by a rest considerably improves the strength of this alloy, though this cannot be stated definitely without further exhaustive experiments. Fig. 33 shows that the strength of the sand-cast ma- terial is practically the same as that of the chill cast material when they are subjected to alternating stresses varying from a maximum compressive to an equal maxi- mum tensile stress, thongh the elastic limits and ultimate strenghts of the chill cast material are con- siderably higher than those of the sand cast material. Thus the elastic limits of the specimens tested are approximately : 3'3 tons/sq. in. 2-25 , Chill cast Sand cast while the fatigue range is about + 3 tons per sq. in. for both the sand and the chill cast specimens. TABLE I. RESULTS OF TENSILE TESTS. Chill Cast. Specimen. Elastic 1, mil. Ultimate Stress. Elongation on 2". Dl C2 Dl C3 ... 3-4 tons/sq. in. 3-2 *7'4 tons/sq. in. 13-08 *i% H% * These low results are due to the fact that the specimen wa< full of small pin-holes. Sand Cast. Both specimens full of small pin-holes. Specimen. Elastic Limit. Ultimate Stress. Elongation on 2". Dl S2 Dl S3 ... 2-3 tons/sq. in. 0. f) * * t ! 7-725 tons/sq. in. 7-12 \Wo \% Chapter. VII. r-: N IS" Mid CNOi HI seaui V 5 K\ \ s V \ \ N O k N !L \\ v_ S\ V \ \ \ \ \ \ X \ \ \ \ V \ \ ,MI/ JNOJ. SJ3KJ.S MM 113d OL -U. X < 5 3 -w lucjei MM Chapber.VII. ALUMINIUM ALLOY WOEHLER TEST PIECE F.c.34-. N3 KORSE TAPER SIZE^TO GAUGE ^- A I t ^* F" -l-f, t f *. k *- i I I 1 ! ~ CLEARANCE AT SH( HJLDER \ RAP - X V p v NO EMERY OR GLASS PAPER TO RF 1 REAMEREO ii T 5 *- * LS FOR CURVES TO BE GROUND ACCURATELY TO DIN B PLATE WITH 1"D\A HOLE TO BED AGAINST FACE 81 TABLE II. RESULTS OF WOEHLER TESTS. Chill Gnat. All specimens were sound and free from blow-Roles and pin-holes. Specimen. Ran 3-839 Broken. Sand Gftst. Specimens free from pin-holes, but one or two specimens had small blow-holes. No fracture, however, occurred at any blow-hole. Specimen. Rai'ge of Stress. Reversals in Millions. Remarks. Cl S2 +3 82 tons/sq. in. 0-643 Broken. Cl S3 3-33 , 3-214 01 Cl 2-4 6 , 7-038 Unbroken. 3-04 , 7-098 s-no . 7-430 3-15 , ., 5-642 Broken. 01 84 2-92 , ,. 15-730 Unbroken. (Allowed to rent for 7 days before next test.) 3-71 tons/sq. in. 12-256 Broken. REPORT ON WOEHLER TESTS ON AN ALUMIMIUM ALLOY, CONTAINING 8% Cu, 6% ZN. The details and conditions of these tests are exactly similar to those described in the previous report on the Woehler tests on Aluminium Alloy L. 8. Pieces of sand cast and chill cast specimens were analysed and gave the following results : The results of the tensile tests are tabulated in Table I., and the stress strain curves are shown plotted in Fig. 35. The results of the Woehler tests are re- corded in Table II., and are plotted in Fig. 36. It will be noted that the results of the tensile tests are fairly uniform with the exception of the elastic limits of the sand cast specimens, which appear to be very irregular. It will further be noted that the superiority of the chill cast specimens over the sand cast specimens is con- siderably more marked under repeated stresses than under direct tensile stress. There is apparently no direct connection between the fatigue ranges and elastic 27264 limits and ultimate strengths in the corresponding ten- sile tests, as shown below. Woehler Test Fatigue Range. Average Elastic imit. Average Ult. Stress. Chill Oast ... Sand Cast ... -4-5-25tons/8q. in. 2 '30 3-7 tons/sq. in. 2.Q 11 I) 12-28 tons/sq. in. 8-87 ., TABLE I. TENSILE TESTS ON ALUMINIUM ALLOY (8% Cu, 6% ZN). Chill Cast Specimen. Elastic Limit. Tons per sq. in. Ultimate Stress. Tons per sq. in. Elonsation % on 2" Rem rka. D2 01 ... D2 02 ... I >2 03 ... Average... 5-9 3-5 3-8 12-28 11-88 12-68 \% Wo Wo 8-7 12-28 O-lo/o Sand Cast. No. of Specimen. Elastic Limit. Tons per sq. in. Ultimate Stress. Tons per sq. in. Elongation o/o on 2" Remarks. D2S1 D2S2 ... D2S3 ... Average... 3-1 2-4 1-5 9-08 8-52 9-00 Wo Wo Wo Small pin holes in metals. Do. do. Do. do. 2-3 8-87 Wo . TABLE II. WOEHLER TESTS ON ALUMINIUM ALLOY (8%Cu, 6%ZN). Chill Gnat. __ Chill Cast (C. 2 C.I). Spee Sand Cast C. 2 S. 1). met Cu. Sn. Fe Zn. *i 8-7% Nil. 0-ii796 5-26% 0-33% 9-54% Nil. 0-75% s-8396 0-35% ' 02 Cl 0202 02 C3 no n Speci- men. Range of Stress. Tons per sq. in. Reversals in millions. Remarks. C2C1 ... + 4-0 8-380 Unbroken. 4-52 5-727 Do. 5-25 5-131 Do. 5-89 1-626 Broken. 02 02 ... 5-4S 075 Broke through small blow 02 03 ... C2C4 ... 4-56 5-28 1-334 4-668 Do. do. Broken. [hole Sand Cflst. Speci- Range of Stress. Reversals in Ilrumi k. men. Tons per sq. in. millions. C2S1 ... 4- 5-31 012 Broken . Pin holes in metal. C2S2 ... 3-82 142 Do. do. do. C2S3 .. 2-32 6-263 Do do. do. C2S4 ... 2-66 2-0113 Broken . Metal free from pin holes on surface. 82 REPORT ON WOKHLER FATIGUE TESTS ON ALUMINIUM ALLOY, ZINC 12%, COPPER 2%. The details and conditions of these tests were exactly similar to those described in the previous report on aluminium alloy L. 8. Pieces of sand-cast and chill-cast specimens in this series were analysed, nnd gave the following result: Chill-cast (C. 3 C. 3). ' Sand-cast (C. 3 S. 4). On 2 07% 2- 17% Su tm 05% o- 06% Fe Zn 11 62% 40% o- 12- 7*% 13% Si ... 27% o- 30% The results of the tensile tests are tabulated in Table I, and the stress/strain curves are shown plotted in Fig. 37. The results of the Woehler tests are tabulated in Table II, and are plotted in Fig. 38. Considering the stress/strain curves it will be seen that in all four cases the curve up to about 4 tons per square inch has been shown as two consecutive straight lines. It is possible that the true stress/strain curve for this range may be a slight curve throughout its length, or the first portion may be a straight line followed by a curve. The total deviation from the straight line is however small, and within the reading of the extenso- meter (2" Ewing) the curve appears to be best repre- sented by two straight lines. The end of the lower straight line is quoted as the " primary elastic limit," and the end of the upper straight line as the " higher elastic limit." TABLE I. Specimen. Primary Elastic Limit. Tons per sq. in. Higher Elastic Limit. Tons per sq. in. Ultimate Stress. Tons per >q. in. Elongation on 2'. Per cent. Chill. D.3. 0.1. D.3. C.H. 2-6 2-2 4-0 3-8 12-04 13-76 1'5 2-5 Sand. D.3. 5.2. D.3. 5.1. 2-0 2-4 3-8 4-0 12-12 12-12 2-0 1-5 From the above results it will be seen that the sand-cast and the chill-cast specimens give very similar results, the chill cast having possibly a higher primary elastic limit. The results obtained from the Woehler tests were not on the whole so satisfactory as have been obtained from the alloys previously tested, as a number of the specimens broke owing possibly to defective material or machining at sections where the stress was not a maxi- mum. In such cases the maximum stress and the stress at the section at which failure occurred are both re- corded. One specimen C.S.C.l. was run for eight million complete reversals without fracture; the stress range was then increased, and fracture occurred after about three-quarters of a million reversals. Two curves have been drawn for each hatch of material; the upper one in each case (shown dotted) possibly represents the con- ditions for uniformly satisfactory specimens, and seems to indicate a fatigue range for the chill-cast of + 3-25 tons per sq. in., and for the sand-cast of +2'5 tons per square inch. The lower curves (in full) show the rela- tionship between number of cycles and stress at frac- ture of the specimen, and these appear to indicate rather lower fatigue ranges, of the order of +2'95 tons per sq. in. for the chill cast and of +2'3 tons/square inch for the sand-cast. There does not appear to be any very definite connec- tion between the fatigue ranges so found and the elastic limits of the alloys, and although there is little differ- ence between the properties of sand and chill cast- ings as shown by the static tests, there appears to be a marked difference between their fatigue ranges. TABLE II. Specimen. Rauge of Stress at point of maximum stress in the specimen. Ran<,'e of Stress at point in specimen where failure occurred. Reversals. Tons per sq. in. Tons per ej. in. Chill. C.3. C.I -4- 3-08 8.H73,36U C.3. C.I 3-78 -+- 3-87 774.2 16 C.3. C.2 3-5 3-35 3. 989.232 C.3. C.A 3-4 3-18 4,825,512 C.3. C.3 3-4 3-02 6,112,708 Sand. C.3.5.1 3-31 2-88 1,544,544 C.3.5.2 3-14 2-96 1,586,304 C.3.5.3 2-65 2-65 4,833,792 C.3.5.4 2-51 2-32 8,218,440 6. PREPARATION OF ALLOYS. The following instructions have been issued in T.D.I.-M.5. : Stock Alloy No. 1. 50% Cu + 50% Al. Melt the copper first completely, then add the alu- minium, solid, in small quantities at a time. After about one third of the aluminium has been added allow the temperature of the mass to fall whilst the remaining alu- minium is being added. Stir the alloy well, and pour it into ingot moulds. The melting point of this alloy is 580 C. Specification L.8. 12% Cu + 88% AI. Melt the aluminium first, then add the necessary quantity of stock alloy No. 1 in a solid state. Stir well. Overheating is to be carefully avoided. Specification L.5 13'5 Zn + 2-75 Cu + Al. Melt the aluminium first, then add the necessary quantity of stock alloy No. 1 in a solid state, and theii the necessary quantity of solid zinc. Stir well and pour. Specification 2 L. 11 7% Cu + 1% Tin + Al. Melt the aluminium first, then add the necessary quantity of stock alloy No. 1 in a solid state, and (lieii add the necessary quantity of tin. Stir well and pour Copper-Nickel-Magnesium Alloy. The preparation oJ this alloy is described in L.A. Report No. 9, p. 7(1. 7. FOUNDING OF ALUMIXIUM. The following report on foundry practice has heen drawn up by Professor Lea. There are certain difficulties that have to be faced and overcome in the founding of aluminium. For good and reliable work, one of the first essentials as indeed, it is for all good founding is cleanliness in melting, in the pots, and in the moulds. Hardly too much emphasis can be laid upon this. The melting pots, whether of the 83 ordinary plumbago crucible type or iron pots covered with some coating, should be kept clean, and in the latter case the coating should be applied frequently and regularly. The second essential is careful regulation of the tem- peratures. It is not always possible to lay down abso- lutely definitely the best temperature to use in pooling any casting, as of necessity the fluidity of the metal must depend upon the thinness of the casting to be produced, and also upon the intricateness of it; generally speaking however the lower the temperature the better. What is important is that when the tem- perature has been decided upon and found to be correct for the effective running of any casting it should be carefully controlled, preferably by means of pyrometers, so that the pouring temperature is very nearly the same at each cast; and further, the temperature the metal is allowed to reach in the melting pot should also be the same within reasonable, which means narrow, limits. At present there is hardly sufficient information avail able to say definitely what is the effect of rapidity of melting, or of allowing metal to " soak " for some time before being poured. If metal is run down rapidly it must of necessity mean a very hot furnace, and there is a danger of the bottom of the pot getting very hot, and of some of the metal reaching a high temperature; with reasonable care however even with a very hot furnace this danger can be avoided. During the war output has been the all important consideration in foundry practice, and risks have been run that in more leisurely times will hardly be found necessary, and it may be therefore that aluminium foundry managers will find it will pay in the long run to sacrifice speed to quality and reliability. With such an expensive material as aluminium, and considering the care that has to be exercised in the preparation of moulds and cores, from a commercial standpoint it will not pay to run any risk in the melting of the metal which may in any way endanger the finished casting. The cost of actually melting the metal must always be relatively small, and the saving due to increased speed of melting is still smaller as compared with the cost of the raw material and of the preparation of the mould and cores; if by paying increased attention to the melting conditions the scrap can be diminished, it is clear that economy will be effected; there is need for careful investigation in this connection. If the temperature of all the metal can be kept during the whole of the melting below a cer- tain temperature there is not much danger, but if in pro- longed heating it exceeds that temperature the rapidity of oxidization is greatly increased, and the risk of absorp- tion of silicon from the melting pot or iron from the stirrers is facilitated. To what extent the oxidised aluminium enters into the aluminium or comes to the surface as a scum is by no means easy to determine; Mr. John Rhodin has done much valuable work in this connection and is still pursuing his investigations. There seems no doubt that very badly " burned " alu- minium contains a considerable percentage of aluminium oxide, but other unexplained defects which occasionally appear such for instance as the brittleness of some ingots of virgin aluminium of normal analysis cannot be traced to the presence of oxide Another and probably ever more elusive problem is that of the absorption of gases by aluminium. Before cooling the surfaces of masses of aluminium very fre- quently look as if occluded gases were breaking through It is 'not always clear that the correct explanation 27264 of this appearance is that gases are actually escap- ing. Cooling of the surface of the metal may not take place perfectly uniformly, and solidification may begin suddenly at needle points, which give the appear- ance of small bubbles of escaping gas. Whether high temperature melting encourages the absorption of gases such as nitrogen, and the products of combustion from a coke or gas or oil fire, is a matter for investigation. Whether it is better to prevent the air coming in contact with the hot metal by passing the products of combustion of the furnace over the pot is doubtful. It is not possible to stop all oxidation in this way for the products of combustion always contain free oxygen, and they may do harm by facilitating the absorption of carbon dioxide and even of nitrogen. The possibility of melting under a vacuum upon a large scale is a very doubtful commercial proposition, but it is probable that melting in special pots, which can be easily covered and thus protected from the hot gases, would reduce any risk of gas absorption, and also probably of oxidation, to a minimum. Certainly, the metal could more easily be kept free from dirt and dust. The use of Sciap. Difficulties have frequently occurred in the foundries due to inaccuracies of mixture. Unless special care is taken mistakes are likely to occur in those foundries in which a number of alloys are being made up and used in different types of castings ; in such cases it is very desirable that furnaces for melting each kind of alloy should be kept quite separate. When cast- ing from virgin metal reasonable care in the weighing out from the stores will prevent mistakes. The mistakes generally arise from the use of scrap, headers, risers, etc., owing to the possibility of getting scrap from castings made of different alloys mixed. As a rule a furnace-man knows the types of castings made from the alloy he melts, so that he can recognise scrap made of another alloy, but headers, risers and swarf cannot so readily be checked. For first-class work the use of swarf should be avoided, and never used unless quite clean. In order to prevent the possibility of errors due to wrong mixtures it is also desirable to keep the alloys as simple as possi- ble in constitution. During the war production on a very large scale has been very essential, and large num- bers of workmen have had to be engaged, with com- paratively little supervision. Simple alloys that can be easily made, and introducing not very serious melting or mixing difficulties, have been adhered to, and the results have justified this policy. Special alloys containing particular metals that were supposed to give satisfactory qualities to the alloys have frequently been suggested, but unless the advantages to be derived from their use could be shown to outweigh the dis- advantages arising from the increased difficulties in making them up, and the possibilities of mistakes in composition, their use was riot encouraged. In the more leisurely times of peace, during which supervision can be better exercised, it may possibly be found that alloys containing such metals as nickel, magnesium, man- ganese, and even rather more iron than is generally thought good, will give valuable properties outweighing the disadvantages of the extra care necessary in the foundry to prevent mistakes, and the greater skill re- quired to prepare the alloys. It will probably be found best to make the "hardners" for the complex alloys in the laboratory and not in the foundry. The Foundry and Design. In the design of crank cases, engine cylinders, and other large and complicated castings it is important that before the final design is 84 determined iu the drawing office serious consideration should be given to the difficulties of casting. Sudden changes of thickness should be avoided as much as possible, and for production on a large scale it is not advisable in large castings to attempt to make them too thin. An extra one or two millimetres thickness on a casting may make a very large difference in the percentage of scrap produced. Again, careful considera- tion should be given to the " venting " of the mould. It should not be overlooked that the aluminium founder in order to overcome the effects of drawing and to diminish the risk of cracking finds it desirable to run the metal as cool as possible consistently with the intri- cacy of the casting. If possible the temperature of pouring for most of the ordinary alloys should be less than 700 C. If however the castings are very thin, or the metal has of necessity to flow a considerable distance in order to fill some part of the casting, higher tempera- tures may be necessary. Low temperature pouring has however one very distinct advantage. Unless great care is exercised in the path the fluid metal is made to take so as always to force air before it, there is considerable danger of cooling taking place before the air can escape and of " shuts " or blow holes being formed. In arrang- ing for the metal to push the air before it it is fre- quently necessary to feed the casting from the bottom of the mould, and therefore to suspend the cores. In the casting of crank cases very considerable difficulty has been experienced with air which has been trapped under cores ; in one case the bearings were carried on a double rib which with the side of the crank case and the bearing made a " box," which had to be formed by a core. To get this core out it was necessary to leave holes in the ribs and through these it was necessary that the ah 1 should escape from under the core. Con- siderable difficulty was also experienced with blown bosses at the bottom of a crank case, due to the diffi- culty of getting the air away from similar box-shaped cores. A somewhat similar difficulty was experienced in the combustion heads of the Arab engines, owing to the difficulty of getting the air away from the cores. In some cases, in order to be sure of getting air away, it has been found necessary to make venting holes in the cores themselves. Cores. Both from the point of view of venting, and also to prevent the cracking of the castings on contrac- tion, cores have to be made very fragile. In iron found- ing the temperature of pouring is sufficiently high to cause destruction of the binding materials such as oils, molasses, or other proprietary articles which are used as binders. In aluminium founding the tempera- tures used are not sufficiently high to dissociate these materials, and they are not therefore often used. To make the cores sufficiently fragile they are made of a sand that is not too binding and frequently a little saw- dust is mixed with the sand; it is nearly always neces- sary to stove the cores. For cylinder work care is neces- sary to see that the water jacket and valve port cores are made in boxes that retain their correct shape and they should be dried on flat plates. It is desirable that the cores should be tested by jigs before being placed in the mould. Gates and Risers As the metal is poured at low temperatures wide " gates " must be used, and if the casting is large a number of gates are necessary which, generally speaking, should be as large as possible, having regard to the thickness of the part of the casting to which they are attached. Wherever there is a consider- able change of thickness or where there is a considerable mass of metal as for instance hear the bearings of a crank case, it is generally necessary to have a good riser in order that this part of the casting may be well fed. A large head to the gate will generally help to pre- vent drawing, and the risers should be sufficiently high. Risers assist in the effective ventilation of the mould, and thus small risers are sometimes found necessary even in the cores. Designers will always do well to consult the aluminium founders about new types of castings as slight modifications in the design may fre- quently give very considerable assistance to the moulder. Chills. The use of chills in sand moulds is to be avoided as much as possible, but in many cases they are necessary; they may be made of brass or cast iron It is difficult to lay down general principles but, as in other castings, the general object of chills is to induce reasonable uniform cooling in the casting. Where bosses project from a thin surface or at fillets where a thicker mass connects to a thin wall chills are frequently found necessary. Care must be taken to see that chills are not too massive; cracks are occasionally developed due to over-chilling. Chills in cores are sometimes found necessary, but care must be taken to ensure that when cooling takes place the chills will allow for sufficient contraction. Die or Chill Castings. The art of die casting alu- minium alloys has been considerably developed during this last five years. No mechanical or rluid pressure is however used in ordinary aluminium die casting. It is perhaps better therefore to speak of the method as chill casting rather than die casting. Aluminium alloy pistons, parts of crank cases, wheels and other details have been produced in dies. Die casting from the point of view of production has considerable advantages over sand casting. Girls- and youths can be fairly quickly taught to carry out the work and with reasonable care defects in castings due to dirt, specks of sand, etc., such as are frequent in sand castings are entirely avoided. In addition the castings have a much finer grain and are stronger. It is not easy to lay down any general principles that have to be observed iu designing dies for aluminium. The best methods of pouring, the arrangement of die gates and risers, and the methods of ventilating the moulds are all of very considerable importance and sometimes require a good deal of experimenting before successful dies are made. The making of the dies requires very skilled workmanship and can hardly pay unless a very large number of castings are to be produced. It is generally desirable that the die shall not be too cold when the metal is poured into it or in other words that the chilling effect should not be too sudden. It can generally be taken that the opening out of the mould should be made as soon as possible after solidification, provided that the metal has cooled sufficiently to resist the slight tensile stresses that may be set up in opening the mould. A fine-grained structure of the metal will be produced if the cooling down to 520 ( '. is fairly rapid a cold chill is not needed to insure this. Under skilled supervision the workers very quickly develop a sense of the amount of time necessary before the mould should be opened out. For small die castings it is undesirable to take molten metal from the actual melting furnace; it is better to have pots kept at a constant temperature and supplied with metal from time to time from the melting furnaces. 85 The design of parts which are to be cast in dies requires careful consideration, but very slight modifi- cations, as a rule, are sufficient to allow any casting to be made in dies. Semi-Die Casting. Semi-die casting has been adopted in a. number of cases with great success, that is the outside of the casting is a metal chill while the inside is a sand core. This method can be adopted when it is impossible so to design a casting as to allow the cores to be withdrawn while hot. If the sand core is made sufficiently fragile in semi-die casting it will not be necessary to remove the casting from the chill very quickly, especially if the gates and risers are so arranged that all the contraction takes place on to the core. 8. POROUS CASTINGS. It is difficult to avoid porosity in aluminium alloy castings. Experience has shown that porous castings can be made perfectly water and petrol-tight by treating them with water-glass, and that the results of the treat- ment, if properly carried out, are lasting. The following instructions for treating porous castings have been issued as T.D.I. M.8 : Tlie Treatment u[ Porous Aluminium Castings it'itk Water Glaus. 1. Suitability of the Process. This treatment is suit- ably for aluminium castings that sweat under water pressure or under spirit test. It is not to be used when such tests disclose areas through which the liquid spurts freely owing to the presence of blow holes or " draws " in the casting. 2. Preparation of Solution Prepare a solution of sodium silicate by dissolving one part of commercial water glass (sodium silicate) in from three to five parts of hot water. The solution may be used repeatedly. 3. Preparation of the Casting Plug up the casting in such a way that it can be rilled with the solution under a known pressure, making certain, as far as possible, that the solution reaches one side only of the porous sections. Small porous castings may be treated before or after machining, as is most convenient in manufacture. Where there is a large mass of metal to machine off, or where the porous sections, after machining, are very thin, the casting should be treated after machining. 4. Applying the Solution -When the casting is ready, pour in the hot solution and apply a pressure (about 70 Ibs. the square inch has been found suitable if the casting is strong enough) by means of a water pump. It is not necessary to pump in sodium silicate. Maintain this pressure until the sweating ceases, or. if the solution does not sweat through, for about 10 to 20 minutes. The temperature at which the solution should be put in depends on the relative mass of the casting and the solution. The right temperature to use for any particular casting can best be found by experi- ment. It should be kept as low as possible consistent with stopping the porosity say between 50 and 70 C. in order to leave the least possible deposit on the surface. This is particularly important when the sur- face to be treated is already machined to its finished size. 5. Washing and Casting. Eelease the pressure, thoroughly wash out the casting with hot water, brush off any white powder which has been deposited, and then allow the casting to stand until thoroughly dry. 6. Alternative method. Where the castings are completely machined before treatment it is sufficient in some cases simply to immerse the casting in a hot bath of the sodium silicate solution. When this is done the temperature of the solution may generally be lower than when the first method is used. This method should only be used when it lias been proved that it is satisfactory for the particular casting in question. 7. Testing. The casting is to be tested hydraulically after treatment. 9. BURNING OF ALUMINIUM PISTONS. Ever since the introduction of aluminium alloy pistons engines have occasionally failed owing to the " burning " of the pistons, and many theories have been advanced to explain how it is caused. In the initial stages of burning the surface of the metal becomes roughened, and minute beads, apparently extruded from the mass of the metal., appear on the surface; the defect is usually confined to one side of the crown of the piston, and extends part of the way down the side. As the burning proceeds the surface becomes pitted, and pre- sents a porous appearance; the porosity gets deeper and deeper, and may extend completely through the metal, but usually cracks develop, or small portions of the burnt metal crumble off, and finally the piston seizes in the cylinder. This type of failure must not be con- founded with another which bears some resemblance to it, but originates in a casting crack. Some designs of piston are specially liable to crack during casting be- tween the gudgeon pin hole and the crown ; when work- ing these casting cracks gradually extend across the crown and the leakage of high temperature gases through them increases their size till a hole is formed through the piston crown, which may ultimately be as large as | inch in diameter. A complete explanation of the burning of pistons has not yet been found, but investigations on many different lines have brought several important facts to light. Burning is undoubtedly primarily due to the overheating of the piston. This overheating may be due to various causes: (1) The commonest cause is defective lubrication between the piston and the cylinder liner. (2) In some engines part of the heat is removed by the oil splashing against the inside of the piston; shortage of this oil will allow the piston to get too hot. (3) Overheating of the liner; this inevitably causes overheating of the piston. Defective con- tact between the liner and the aluminium shell is not uncommon, and causes the liner to overheat. (4) Local concentrations of heat occur in some cylinders, possibly due to the relative posi- tions of the sparking plug, the inlet and the outlet valves, or to local defects of contact between liner and shell. The actual temperature necessary to produce burning is much more difficult to determine. The appearance of the extruded beads of metal on the burnt piston's sug- gested that the temperature must have reached the eutectic melting point, 530 C., and as slight overloads in some engines almost always caused the pistons to burn it appeared as if the ordinary running temperature could not be far below 530 C. If this were true the limiting size of all pistons appeared to have already been reached. It therefore became of great importance to measure the actual piston temperature when working; 86 the practical difficulties were great, but were finally overcome by Professor Gibson at the It.A.E., and the results shewed that the pistons never reached any tem- perature approaching the eutectic melting point, but usually worked between 200 and 240 G. An attempt was made to measure the temperature of a piston while it was actually burning; for this purpose a water-cooled engine was selected which had previously burnt out pis- tons in one of its cylinders. The measurements showed that the liner in that particular cylinder was about GOC. hotter than in the other cylinders, and the piston tem- perature rose to 275, compared with 240 C. in the other cylinders, but unfortunately, though the engine was loaded as heavily as possible, actual burning did not take place. These results, though not conclusive, indicate that burning probably takes place at a temperature only <* little over 275, and much below 530, the eutectic point. Further investigations are in hand, particularly to test whether some constituent of the alloy may not soften at temperatures of perhaps 300 sufficiently to allow of its being squeezed out to form the beads, and so initiate failure. Professor Lea has found that beads of metal are squeezed out of the metal at relatively- low temperatures in the Brinell test. The investigation of the micro-structure of burnt pis- tons has led to many theories being suggested, but none of these have been generally accepted. 10. STEEL LINERS IN ALUMINIUM CYLINDERS. Calculation shows that unless special precautions ai'e taken it is not possible to fit a steel liner in an alu- minium cylinder so that it will remain a tight fit when ihe engine is working. The reasons are that the coeffi- cient of expansion of the aluminium alloys is 2'6 times that of the steel liner, and that the elastic limit of the aluminium is very low; the result is that if the liner is tight when hot it will stretch under the stresses due to contraction when cold and the next time the cylinder is heated the liner will be slack. When the liner is slack it has to rise to a much higher temperature than the cylinder. The space between liner and cylinder is a very bad conductor of heat and the liner rises in tem- perature till its expansion brings it almost in contact with the aluminium cylinder again. An actual calcu- lation based on the thermal conductivity of the gases between the cylinder and liner shows that the gap must be reduced to about O0003 inches before the jacket heat can pass from liner to cylinder. Slack liners have undoubtedly often been the cause of overheating and failure of engines, so that it appears to be desirable to modify the present designs in some way to get over this defect; promising modifications have already been tested. 11. ALUMINIUM CRANK-CASKS. Many failures in engines have been traced to want of rigidity <>f the crank-cases. The low elastic moduli of the aluminium alloys is unfortunate in this connection, but the rigidity depends mainly on the design, and a. great deal can be done to improve designs if the impor- tance of rigidity is borne in mind. Crank-cases distort in two ways, by bending and by torsion; both actions must be remembered in the design. A full investigation of the rigidity of a Sunbeam engine has been made by Professor Lea, and of a Hispano Suiza engine by Major Robertson. The additional stiffness due to the addition of the oilwell and cylinder blocks were measured, and very interesting results were obtained which explained many of the fractures which had occurred. The stiffen- ing effects produced by adding extra bracing were also measured. This subject, however, hardly falls within the scope of this report. 12. ALUMINIUM ALLOYS AS BEARING METALS An investigation of the maximum loads which alu minium alloy bearings would carry, and on their coeffi- cients of friction was undertaken, but the method of testing was not considered satisfactory and the meaning of the results was doubtful. Engine tests with gudgeon pin bushes of Duralumin and of L 5 alloy gave excellent ivsults, the wear after 140 hours being quite small. This subject is worth following up ; considerable saving in weight and complication could be effected if bmn/c bushes could be eliminated. It is not known whether white metal can be used in aluminium shells. 13. WELDING AND SOLDERING. Welding Aluminium Alloys 1'ure aluminium and all the aluminium alloys may be welded. The process has been used for making tanks, and for patching and repairing at the front. No investigations have been made on it in the Material Section. The best flux is given in Air Board's Specification, L. 13. 14. GERMAN ALLOYS. German cast alloys, used in engines. Maybach Crank-case Zn 13'0 Ten. test 10 tons/sq. in Cu 1-0 Mercedes Zn il'6 8' 7 Cu 4-2 Benz Zn 9'6 8'7 Cu 6-3 Benz Piston Zn 12 -0 Cu 6-0 Fe 1-4 None of these alloys, so far as is known, show my superiority over the standard Air Board specifications. APPENDIX. The following are the principal clauses from the Air Hoard Speci- fications L'), LS, LIU and Lll. SPECIFICATION Lr> FOU ALUMINIUM ALLOY CASTINGS. SUITABLE COR CRANK CASKS AND GENERAL USE. '/'/in ipfcijie graeifi/ of thif alloy in 3'0. 1. QUALITY OF MATERIAL. The aluminium used for making this alloy is to assay not less than 98 per cent. The copper need for making this alloy is- to assay not, less than 99 3 per cent. The zinc used for making this alloy is to be of the best quality. The alloy is to consist of Zinc, not less than 12 <> per cent, or more than 14 '5 per cent. Copper 2-5 8-0 Aluminium, remainder. Impurities Lead, not more than 0' 1 per cent. Silicon 1-0 , Iron ,,1-0 :>. MECHANICAL TESTS. The metal is to comply with the follow 87 ing tests, which are to be carried out by the Contractor at his works in the presence of the Inspector and to his satisfaction. Tensile Test. Test; pieces turned to the dimensions of British Standard test piece C, with ends shaped as shown in Fig. 2, p. 5, Chap. I, from sample pieces cast as specified in clause 6, must give the following results : Ultimate strength not less than ... 11 tons per square inch. Elongation ,. ,, ... 4 per cent. The teat pieces are not to be annealed, hammered, or otherwise treated before they are tested. The ultimate strength of this alloy should be over 12 tons per square inch, and the alloy will not be considered first-rate till this strength is attained. 6. PROVISION OF TEST SAMPLES. At least one sample is to be cast to represent each Crank case or other large casting. The number of samples for small castings is to be settled by the Inspector, so that all the alloy used is tested. The samples are to be caiit from the same ladle as the castings and are to be poured first. The samples are to be 1 inch in diameter and from 7 to 9 inches long. They are to be cast in iron chills which have been heated before they are filled. The bottom of the chill is to be closed with a clay or sand plug, not with a metal end. SPECIFICATION L8 FOB 12/ COPPER-ALUMINIUM ALLOY CASTINGS. The speeijic gravity of this alloy it between 2'83 and 2-91. 1. QUALITY OP MATERIAL. () The aluminium used for making this alloy is to assay not less than 98 per cent. (4) The copper used for making this alloy is to assay not less than 99 3 per cent. (c) The alloy is to consist of Copper, not less than 11 per cent, or more than 13 per cent. Aluminium, remainder. () The samples are to be 1 inch diam. and between 7 and It inches long ; they are to be cast in iron chills which have been heated before they are filled. The bottom of the chill is to be closed with a clay or sand plug, not with a metal end. The samples are to be cast from the same ladle as the castings, and are to be poured first. SPECIFICATION Liu FOK ALUMINIUM ALLOY CASTINGS. The specific gratify oj this alloy in 2'95. 1. QUALITY Ob 1 MATERIAL. () The aluminium used for making this alloy is to assay not less than 98 per cc-nt. (//) The copper used for making this alloy is to assay not less than 99'ii per cent. () The alloy is to consist of Copper, not less than 9 per cent, or more than 11 per cent. Tin, 0-5 1'5 Aluminium, remainder. Imp irities Lead, n it more than O'l per cent. Silicon, ,, I'O Iron, ,, I'O 5. MECHANICAL TESTS. (a) The metal ie to comply with the following tests, which are to be carried out in the presence of tho Inspector and to his satisfaction. (V) Tensile Tests. Test pieces turned to the dimensions of British Standard test piece C, with ends shaped as shown in Fig. 2, p. 5, Chap. I, from sample pieces cast as specified in clause fi, must give the following results : Ultimate strength not less than ... 9 tons per square inch. (c) The teet pieces are not to be annealed, hammered, or otherwise treated before they are tested. 6. PROVISION OF SAMPLES FOR THE TENSILE TEST. () At least one sample is to be cast to represent each cylinder block or other large casting. The number of samples for small castings is to be settled by the Inspector, so that all the alloy use I is tested. The samples are to be cast from the same ladle as the castings and are to be poured first. (J) The samples are to be 1 inch in diameter and 7 to 9 inches long. They are to be cast in iron chills which have been heated before they are filled. The bottom of the chill may with advantage be closed with a clay or sand plug instead of with a metal end. 7. HYDRAULIC TEST. (a) All cylinder castings and others which are required to be tight are to be tested for porosity by means of water, methylated spirits or petrol. The test is to be carried out in an approved manner to the satisfaction of the Inspector. (J) Castings which have been tested are to be marked to show the result of this test as the Inspector may direct. All castings which in the opinion of the Inspector are too porous for use will be rejected. Castings which are only slightly porous will be accepted subject to their proving tight after subsequent doping. 8. DOPING. (a ) Doping may be carried out by the Contractor or at the machine shop as may be arranged. If the Contractor dopes the castings he ia to do so in an approved manner to the satis- faction of the Inspector. SPECIFICATION 2L11 FOR 7:1: ALUMINIUM ALLOY CASTINGS. The tpeeific grarittj oft/tie alloy if between 2'87 and 2'93. I . QUALITY OP MATKRIAL. () The aluminium used for making this alloy is to assay not less than 98 per cent. (A) The copper used for making this alloy is to assay not leas than 99 3 per cent. (0) The alloy is to consist of Copper, not less than 6 per cent, or more than 8 per cent. Tin 0-5 2-0 Aluminium, remainder. ( ^ fi kt * n H i flC k! Q_ o o: Ui -Jo 89 brazed into the unions (unless designs are altered so as to replace screwed unions by rubber tube connections). The following report on fatigue tests made at the N.P.L. with plain copper tube, arsenical copper tube, and nickel-copper tube indicate that the arsenical copper is decidedly the best. The present specification calls for this type of metal. 2. FATIGUE TESTS ON COPPER TUBES BY THE ENGINEERING DEPARTMENT OF THE NATIONAL PHYSICAL LABORATORY. The tubes used in these tests were taken from the batches supplied by three different makers, who are described in this report as A, B, and C respectively. The samples are distinguished as follows : Maker. Grade. Marks. A Soft copper KZ 1 B Deadsoft copper . . ... ... KZ 4 B Deadsoft copper-nickel ... ... KZ 7 C Soft arsenical copper KZ 10 The method of testing is shown in Fig. 1. The tubes, which were f^-inch outside diameter and about 0'040 inch thick (I.S.W.G.19), were cut into lengths of about 4i inches, which were then fitted with a plug at each end; these plugs were well rounded on the inside, as shown, so as not to damage the inside of the tube during the test ; the tube was then firmly fitted to the spindle of the testing machine and a bail bearing fitted to the overhanging end, and the load was suspended from trunnions attached to this ball bearing, so that the tube formed a cantilever with a length of about 1'4 inch. The spindle was rotated at a speed of 2,000 revolutions per minute. If under a given load the specimen did not break after several millions of revolutions (as shown by the table below), the load was increased and the test continued until fracture occurred. Several pieces of each kind were broken, and the results are given in the table below, and also in Figs. 2 to 5. The stress in the material was determined on the assumption that the material was not strained, either in tension or compression, beyond the elastic limit during the test. From the appearance of some of the broken test pieces, however, this was not the case as several of them had decidedly bulged out near the point of frac- ture, indicating that the compressive stress was beyond the elastic limit of compression. This was particularly the case with the sample KZ 10. With three of the materials, the graphs of " range of stress " / " number of reversals " form fairly definite curves from which the limiting range of stress can be determined, but with the sample KZ 7 this was not the case as the tests did not appear to give concordant results; this may be due to lack of uniformity of the material. The tests definitely establish the superiority of the sample KZ 10 for withstanding alternating bend- ing stresses, although two samples from this tube, namely. KZ lOFj. and KZ 10f!. broke with a very low number of reversals. The large circles in the graphs represent the results of the fractured specimens, and the small dots repre- sent tests which were stopped before the point of the fracture was reached, these dots should therefore always lie under the curve, 87264 RESULTS OF TESTS Mark. Maker. No. of Reversal* Range of Stres ton/sq. inch. Remarks. KZ 1A .. A 1,123,900 + 2-93 Not broken. ,. A 250, These results are plotted Figs. 6 and 7. In the case of alloys with a Brinell number less than 50 it was difficult to maintain the load, and the figures obtained at 450 C. for such materials should there- fore be taken with some reserve, as it is doubtful in these cases whether the flow had ceased at the end of 10 seconds, and the numbers quoted for these tests are perhaps higher than they should be. The hardness of the 60/40 type alloys drops rapidly after a temperature of 250 C. is reached. Of the alloys of this type which were tested Delta IVE is the hardest at all temperatures above normal, the presence of nickel probably accounting for this. The cast manganese bronze appears to be harder at all temperatures than the drawn manganese bronze. Phosphor bronze maintains its hardness up to 350 C., and then commences to soften rapidly. The sample of sand-cast phosphor bronze P is somewhat harder at the lower temperatures than the other two samples. The gunmetal, as might be expected, gives results very similar to those obtained from phosphor bronze. It is interesting to note that at 350 C. all the brasses and bronzes have a hardness number of about 60 to 70. Of the other materials tested nicro-copper is the softest when cold, but maintains its properties so well that at 450 C. it is harder than any of the brasses or bronzes. Monel metal and the 3' 7% nickel steel (included for comparative purposes) are harder at all temperatures than the ordinary copper alloys, and botli are but slightly softer at 450 C. than they are cold. Two samples, 73 and 74, of iron-bearing aluminium bronze (ferra-cupralium) were tested. The individual results seemed to be a little erratic, which may be due to poor mixing. Both samples are very hard when cold, and well maintain their hardness up to 350 C., but above that temperature they appear to soften very rapidly. 4. TENSILE TESTS OF SOME COPPER ALLOYS AT VARIOUS TEMPERATURES. In order to obtain further information on the strength of the different alloys at high temperatures, Professor Lea carried out p series of tensile tests. Tho results nvf Chapfcer.yill. 00 6 LL VJ 7; NOIJ.VSW13 5/64-*. EOOO C ft F L. T1 91 given in tlic following report which again shows the merits of Monel metal :- These tests were carried out with the object of deter- mining some of the mechanical properties of a few copper alloys. Altogether eight alloys, which were obtained through the ordinary commercial channels in the form of 1 inch diameter bars, were investigated. Tests were carried out at 20, 250, 350, and 450 C. on i sq. in. standard specimens. The specimens were tested in a horizontal machine and were heated during the hot tests bv means of an electric furnace. The various temperatures were reached in from 20 to 25 minutes, and then the specimens were allowed to soak for 2 to 2| hours at that temperature, in order to acquire a steady condition, not only through the specimen, but also along the bars leading out of the furnace, as well as along the extensometer. The temperatures were measured by means of a thermo-couple attached to the specimens. The stress-strain diagrams were drawn for all the tests, and the results are summarised in the following table and graphically in Fig. 8. TENSILE TESTS OF COPPER ALLOYS AT HIGH TEMPERATURES. Metal. Composition. 0C. . a? $ 250 C. 350 C. 450 C. 3 O a N & g i CM i OU (U ! fe * |j w3 535 e* a ^ S3 - || & bi fa S3 A* |1 005 a ta SE 33 A. Young's Mod. M a 2 M Muntz Metal.. ! I .CO 1 - - i - - - | 6-4 21-3 6800 31-49 26 1-8 5-8 4890 23-02 28 Below 0-4 - - 12-8 42 Below 0-2 - - 3-92 !11-0 Cast M anga- J neee Bronze. ) i OS s o - I - i - * r-t | 6-3 10-4 6810 32-5 20-5 4-8 10-3 6140 123-15 44-5 Below 0-4 - - 12-6 84 Below 0-4 - - 4-08 . Drawn Manga- ( nese Bronze. | i in 1 IO 1 a S - o - 1 OO O | 8-0 11-4 C660 34-4 24-5 2-0 10-0 6130 22-84 54 0-8? 0-95 5260 11-74 82 Below 0-2 *- - 2-72 67 Delta Metal .. j I 1 g S - 2 - 13 | 4-6 ' 24-6 6890 32-65 36-9 2-8 '8'8 6290 28-0 30-5 0-8 2-3 3870 18'6 13 Below 0-2 - - 7-32 35 Gun Metal .. j 1 en 1 - - - S - - | 3-2 6-5 i 5780 15-82 14-5 2-8 5-8 5800 16-06 16 2-2 6-9 5000 10-27 4 1-6 4-50 3880 5-2 2 Phosphor Bronze, j 1 s s w o - g I - i j 4-4 8-43 6i.lO 13-49 1-5 3-4 7-73 /5800 13-46 3-5 2-6 6-95 4040 972 1-5 0-5 2-16 3510 9-40 2 Nicro-Copper j 1 - - p - - - - 20 4-0 11-75 8160 116-69 41 2-4 10-4 7700 12-98 30 3-4 8-55 6930 11-4 19 2-0 7-0 5150 11-33 7-6 Monel Metal .. J 1 - - i CO - - ^> C-I ~* | 10-4 14-8 12700 38-77 49 10-4 11-64 12100 35-76 43 10-0 11-8 11530 36-0 45-5 6-0 11-25 11000 31-0 44-5 A. Stresal which ^ ic deformation = 1 Elastic deformation 4 All stresses are in tons per B([. in. Elongations given as percentages on 2 ins. Tensile Strength In Fig. 8 it will be noticed that all the 60/40 copper-ainc alloys, i.e., Muntz metal, drawn and cast manganese bronze, and Delta metal have very similar values for ultimate stress, and that the addition of 2% of nickel in Delta metal adds considerably to the strength at high temperatures. The curves of the bronzes, i.e., gunmetal and phosphor bronze, are of quite a different character. The diminu- tion in the value of the ultimate stress does not occur until a temperature of 250 C. is reached, and then it is of a much smaller order than in the case of the 60/40 ' alloys. Monel metal is superior at all temperatures, and nicro-copper retains its strength very well with increase of temperature. It is interesting to note that at 350 all these alloys, with the exception of Delta metal and Monel metal, have approximately the same ultimate strength and also the same Brinell hardness number. Ductility. Keferring to the elongations (taken on 2 inches) of the 60/40 alloys (Fig. 8), it will be seen that Muntz metal, and more particularly drawn and cast manganese bronze, possess a greatly increased duc- tility at 350 C., which falls off very considerably at 450. Delta metal appears to be hot short at 350 C., 27264 but recovers its ductility at 450 C. The bronzes are much less ductile than the 60/40 alloys, especially at 350 and 450 C. Monel metal possesses a very con- stant elongation at all temperatures up to 450 C., while the elongation of nicro-copper falls off, until at 450 C. the metal is comparatively brittle. Elastic Limit. It will be seen from Fig. 8 that only the two bronzes, nicro-copper and Monel metal, have any elasticity at 450 C., and that the curves of the first three of these do not fall off very rapidly. At 350 C. cast manganese bronze and Muntz metal are the only alloys lacking in elasticity, but at 250 C. cast man- ganese bronze has quite a high value for its elastic limit. Young's Modulus. Keferring to the table, it will be seen that nicro-copper and Monel metal have much higher values for Young's Modulus than have the brasses and bronzes. In conclusion it may be stated that the properties of the bronzes are much less affected by heat than we those of the 60/40 alloys. The ductility of nicro-copper diminishes as the temperature rises. The ductility of Monel metal remains constant. In alJ properties it is superior to the other alloys. M 2 5. WHITE METALS FOR BEARINGS. There are many proprietary brands for which great claims are made, but there is no real evidence that there is much difference between the results obtained with the different brands. As has already been pointed out in connection with prosphor bronze, there is great need for a reliable method of testing bearing metals. The different alloys were very thoroughly discussed by the sub-committee of the E.S.C. and the alloy specified in 2. L9 was agreed on, but it must not be forgotten that there is hardly any evidence that one alloy is better than another. A well-known engineer who has made extensive tests on white metals has stated that the bearing which gave the highest results in his testing machine was made of lead obtained by melting down an old lead pipe. L 9 alloy is quite suitable both for lining bearings and for making die castings; it contains : Copper, between 3 and 4%. Antimony, between 4 and 5%. Copper plus antimony, not more than 8%. Tin the remainder. 6. SOFT SOLDER. Three grades of soft solder are recommended, viz. : Grade A. Tin, between 63 to 68%. Lead remainder. Total impurities not more than 1'5%, of which anti- mony shall not exceed 1'0%; zinc, 0'05%; arsenic, 0-05%; iron, 0-2%. Grade B. Tin, between 49 and 52%. Antimony, not more than 3%. Lead the remainder. Total impurities not more than 0'5%, of which zinc shall not exceed 0-05%; arsenic, 0'05%; iron, 0'2%. Grade G. Tin, between 28 and 32%. Lead the remainder. Total impurities not more than 1'5%, of which anti- mony shall not exceed 1'0%; zinc, 0'05%; arsenic, 0-05%; iron, '0-2%. The following instructions for using solders have been issued : Grade A. Solder This solder melts at about 170 C. This grade must be used when making sweated socket joints in steel tubing. Grade B. Solder.- -This solder is not completely melted until about 205 C. This grade must be used for building up radiator casings and tanks, petrol tanks, and for tin and coppersmiths' work generally. Grade C. Solder. This solder' is not completely melted until about 225 C. This grade must only be used as a " dipping " metal or for making " wiped " joints. It must not be used with a soldering iron, or gas, or blow-pipe flame. General Instructions. Surfaces to be soldered must bo thoroughly cleaned and tinned (that is, coated with a film of the same solder which is to be used on the job), and then coated with flux before being brought together. They are to be sweated or soldered together whilst the tinning is fresh. Holes for pins through soldered joints should be drilled, and the pins inserted before the joint is sweated together. Oil must not be used in the drilling. Cleaning the Surfaces. The surfaces must be cleaned mechanically if possible. Pickling must only be used if other methods of cleaning are not practicable. If pick- ling has to be resorted to, the residue of acid which is always left must be carefully removed, preferably by heating- the work to a temperature of 150 C., or by thorough washing and drying. Sweating In sweating, care must be taken that the joint is completely filled with solder, and that the solder is not run out by over application of heat. A soft, non- luminous gas-flame is best for sweating. Final Cleaning. AH traces of the flux used on any soldered job whatever must be completely removed either by boiling or washing the ivorlt immediately after soldering with clean hot-water and carefully drying. Strength of Soldered Joints. If the film of solder is too thin the joint is too brittle, if too thick, the shear strength is low. The best thickness of film is from 0-003 inch to O010 inch. The shear strength of a con- tinuous solder film is about three tons per square inch, but in practice no soldered joint is found to be continu- ous. In good work, at least two-thirds of the overlap- ping surfaces are united, so that a shear strength of at least two tons per square inch of the whole joint should always be developed, but more than this cannot be relied upon. As it is almost impossible to inspect sweated joints, occasional tests must be made by pulling the surfaces apart and examining them to ascertain whether a suffi- cient proportion of the surface (at least two-thirds) has been united. Fluxes for soft soldering. Samples of most of the leading proprietary brands of flux have been purchased and analysed with a view to finding, if possible, one which should not exert any corroding action when it can- not be cleaned off. The results are given in the following table, from which it will be seen that there is little difference between them, and that all contain tho objec- tionable zinc chloride. Miker. Minerul Jelly. Zinc Chloride. Ammonium Chloride. Water. Flowing Temperature Centigrade. JELLY FLUXES. A B C D 76-1 64-2 66-1 61-5 11-8 16-1 15-0 17-2 0-5 0-6 1-6 0-4 12-1 19-1 17-3 20-9 45 '5 e 40-8 48-5 46-2 LIQUID FLUXES. E ... F ... 39-0 H6-5 1-9 3-6 59-1 59-9 93 'The Engineer's Department of the Post Office use two fluxes : Flux for copper wires in cables : Paraffin wax ... ... .. 5 parts Tallow 3 Resin 2 The resin is added in powdered form to the wax and tallow when melted at 100 C., and well stirred in. For other purposes : Zinc chloride ... ... ... 56 Ib. Ammonium chloride ... ... 10 Ib. Water 8 gallons The object of the ammonium chloride is to keep the zinc chloride in solution. CHAPTER IX. CORROSION AND PROTECTION AGAINST CORROSION. The work which has been completed so far by the Corrosion Committee has dealt with: (1) The protection of steel rigging parts, such as wiring plates, fork ends, turnbuckles, and streamline wires. (2) The protection of high tensile steel wire and cables. (3) The corrosion of aluminium and its alloys. (4) The corrosion of two metals in contact. 1. The protection of steel rigging parts. The right method of protecting steel rigging parts has been con- sidered very fully, and a large number of experiments have been carried out. The methods which have been investigated include almost all those possible. The work has been divided into two parts, namely, protection by paints and varnishes, which will be referred to as organic protectives; and secondly, protection by metallic or partly metallic coatings, such as zinc plating. In so far as the organic protectives are concerned the work is not yet completed, but it has been found that there are relatively few paints which give satisfactory results under the conditions of experiment. The protec- tion afforded by the organic protectives has been tested on metals exposed to corrosion by the atmosphere, by tap water, and by sea water, both by continuous immersion, and also by alternately wetting and drying. The pro- tective coatings have also been exposed to the attack of lubricating oil and of petrol, and tests have been made of their ductility. This latter test was made by bending the coated specimen and then examining the coating to determine whether it had cracked. Speaking generally, the most satisfactory protectives are stove enamels and some ot the high-grade varnishes. This work is proceeding. Investigation has been made on the following methods of protecting parts by metallic coatings : (a) Ordinary galvanising, that is, the hot-dip process. (6) Electro galvanising, that is, plating with zinc by electro deposition. (c) Sherardising and the D.N. process. (d) Coslettising and Jonetising. (e) Aluminium painting. (/) Copper plating, that is, plating with copper by electro deposition. (g) Lead plating, that is, coating with lead by electro deposition (h) Tin plating by the hot dip process. The method of testing which has been employed in all these cases has been similar to that used in the testing of organic protectives. It was found that several of the processes could be ruled out very quickly as giving quite unsatisfactory results. The process of galvanising by hot dipping failed both in the atmosphere and under the attack of salt water. The process of tin plating was found to be even less satisfactory; it gives a thin coating which is very rarely continuous, and the coated samples are very rapidly attacked, even in the atmosphere. The process of Jonetising, which appears to be an inferior method of Coslettising, did not give completely satis- factory results in the atmosphere, and in sea water the specimen was completely destroyed. Coslettising was found to be quite a satisfactory process for protecting steel against the action of the atmosphere, but it was not good for resisting the action of sea water under alter- nate conditions of wetting and drying. It is a process which would be quite satisfactory for aeroplanes, but not for seaplanes. Copper-plating gave very satisfactory results in so far as actual resistance to corrosion was concerned, but where the copper-plating was scratched or where it came into contact with any naked steel the corrosion of the steel was so violent that it was con- sidered undesirable to recommend the process. Alu- minium painting did fairly well in the atmosphere, but badly in sea water. It is probable that the process might yield satisfactory results if special care were taken in the application of the aluminium paint, for it appears that the efficiency of this process depends entirely upon the skill of the painter. Experiments upon this coating are being continued in connection with the organic pro- tectives. Lead plating gave moderately satisfactory results. It was found, however, to be distinctly inferior to zinc under all conditions, and as it is more expensive to carry out than zinc plating, and as the weight of the deposit is much greater, the method was abandoned. Turning to the zinc coatings, experiments were tried upon material which had been zinc-coated electrolyti- cally by different firms in various ways, and also upon Sherardised material. It was found that Shenzdising, if properly carried out, affords a moderately satisfactory protection against atmospheric corrosion, but not at all 94 a good protection against corrosion by sea water. The JJ.JN. process, which, is a modification of the Sherard- ising process, was found to be superior, and gave satis- factory results, particularly against atmospheric corro- sion. The electrolytic zinc coatings were found to be exceedingly good, and in all cases resisted corrosion both in the atmosphere and also in sea water. Without question electro-plating with zinc is the most satisfac- tory of all the methods tried, and can be highly recommended. The practical result of the investigation is that for aeroplanes the following processes, which are given in order of merit, can be utilised satisfactorily: (1) Electro-galvanising, i.e., the electrolytic deposi- tion of zinc. (2) The D.N. process. (3) Coslettising. For seaplanes the only process which can be recom- mended confidently is electro-galvanising. 2. The protection of wires and cables Various pro- cesses have been used to protect high-tensile wires and cables, and all have been tested. The process most widely employed up to the present is tin-plating, whilst the next most widely employed is ordinary galvanising. A process which was brought to the notice of the Material Section by Mr. Brunton was found to be a great improvement upon both the others; it is known as " Bruntonising," or " hard-drawn galvanising," and con- sists of carrying out an ordinary hot-dip process upon wire which has been properly cleaned, and then draw- ing down the coated wire in dies. By this process it is found that a satisfactory tensile strength can be obtained in the finished wire (within certain limits) whilst the operation of drawing has the effect of consolidating the zinc coating, and of removing the flaws which are so fatal to the ordinary hot galvanised wire. In test it was found that Bruntonising gives a much better pro- tection against corrosion than any other kind of coating which has been tried; samples of it have been exposed to alternate wetting and drying for as much as 12 months without any apparent effect upon the wire. Both tinning and the ordinary hot dip galvanising are very bad. The principal drawback to Bruntonising is that there is a limit to the size and to the ultimate strength of the wire which can be protected in this manner. For ordinary high-tensile steel wire (to specification W.I) no difficulty arises, but if it is attempted to draw the wire down to the exceedingly small sizes required for the small flexible steel wire ropes (to specification W.2) it is not found possible at present to carry out the pro- cess satisfactorily. It is possible that this difficulty may be overcome, and work is proceeding with this end in view. 3. The Corrosion of Aluminium The results which have been obtained so far cannot be regarded as final, and the work is proceeding in order to confirm the con- clusions arrived at in this preliminary report. The experiments' have included tests on the following materials: (a) Duralumin, in accordance with the Air Board specifications. (6) Hard-drawn and soft-drawn aluminium of com- mercial purity. (c) Specially pure aluminium. (d)' Cast alloy to specification L. 5 (sand and chili- cast). (e) Cast alloy to specification L. 8 (sand and chill- cast). (/) Cast alloy to specification L. 11 (sand and chill- cast). (g) Cast alloy containing about 4% copper, 2% nickel, and 1|% magnesium. (h) Rolled rods of an alloy containing about 18% of zinc. These have been tested under atmospheric condi- tions, also by immersion in distilled water, in tap water, in sea water, and have also been sprayed with solutions of various salts, including ordinary sea water. It is found that the results of the cor- rosion under the different conditions vary very little, and that material which will resist atmosphere cor- rosion will resist most of the other forms of attack. The general conclusions which have been reached so far show that the alloys containing a considerable propor- tion of copper corrode very violently, whilst those con- taining a high proportion of idnc are almost equally bad. Tin- alloy containing 18% of zinc corrodes probably more violently than any of the other alloys. The resistance of duralumin to corrosion was very good on the samples which were tested, and particularly upon those samples which had no cut surfaces. If the surface were left as rolled or drawn, then the corrosion was particularly slight, but when the surface was cut corrosion in some cases rapidly followed. Pure alu- minium corroded more than did the duralumin, and it was found that there was practically no difference between the corrosion of the commercial and the pure metal. It was found also that there was no observable difference in corrosion between the hard, half-hard, and soft-rolled aluminium. Of the cast alloys, that containing 4% copper, 2% nickel, 1J% magnesium, gave by far the most satisfac- tory results. Experiments were also carried out to determine the most suitable organic protectives for aluminium. It was found that it was much easier to protect aluminium than steel with an organic coating, and that an organic pro- tection which was satisfactory for steel would also probably be quite satisfactory for the aluminium alloys. 4. Contact of two Metals There is a general impres- sion that corrosion is bound to be accelerated when two dissimilar metals are in contact with each other in the presence of a corroding medium; this opinion is based upon the belief in the galvanic theory of corrosion. If this belief is sound it follows necessarily that brass and steel, or aluminium and steel, or aluminium and brass, when in contact with each other will corrode at a much greater rate than if the two are separate. It also follows that in the course of the corrosion the metal which has the highest positive electro potential will corrode the most rapidly. It would be expected, therefore, that when a galvanised steel fitting was in contact with an uncoated fitting, or when a galvanised fitting has been scratched, the zinc would corrode and so protect the iron. As a matter of fact, in actual practice (and usually in labora- tory experiments) no such result is obtained, and zinc does not act in this way as a protective at all. As this is directly_ contrary to what should happen according to the above theory, it was decided to try an elaborate 95 series of contact experiments. Two forms of test piece were used. The first consisted of two discs of dissimilar metals fastened together, after polishing the two plane surfaces of contact, by means of a thread. The second consisted of a nut and bolt arrangement in which the nut and the bolt were made of the two metals to be tested. The metals which were employed were mild steel, pure zinc, pure copper, pure aluminium, alloys to specifications L. 8, L 5, L. 11, and brass. Each of these metals was tried in conjunction with every one of the others, and in addition was also tried by itself, that is to say, in contact with another piece of the same metal. Control experiments were carried out by attacking pieces of each of the metals not in contact with any other metal. It was found that the action of the corroding media, which were the same as employed in the other experiments, namely, the atmosphere, tap water, and sea water under alternating wet and dry conditions, was apparently no greater in the case of the metals in con- tact than it was in the case of the non-connected metals. These results, which have been repeated several times in the laboratory, are surprising if viewed from the standpoint of the galvanic theory, but they have received ample confirmation in actual practice. It has been quite general to use, even on seaplanes, fittings in which brass and steel are continually in contact, and no harm ful results appear to have come from this arrangement. The definite conclusion derived from these contact experiments is that it is just as safe in practice to have contact between two dissimilar metals as between two pieces of the same metal. It appears to be likely that the laboratory results can be explained fairly readily by the fact that in all cases the galvanic action is not between the two metals except at the commencement of the operation. A very brief exposure to corrosive conditions gives rise to pro- ducts of corrosion which tend to stabilise and to retard, if not to prevent altogether, the ordinary galvanic action which otherwise would go on. CHAPTER X. TIMBER. CONTENTS. Introduction. 1. Spruce Strengths. 2. Micro-structure and Strength. 3. Teitinp. 4. Strength in directions inclined to the Grain 5. Elastic Constants. 6. Deviation of Grain. 7. Brittleness. 8. Fatigue. 9. Chemical Treatment. 10. Built up Spars. 11. Strength of built-up Spars. 12. Strength of Struts. 13. Stiffening Timber. 14. Use of Heavy Timber. 15. Moisture. 16. Varnish. 17. Plywood. 18. Propeller Bosses (Coefficients of Friction). 19. Reinforced Timber. 20. Substitutes. Appendix I. Specification 2 V. 1. II. Inspection and Testing. HI. Elementary Theory of Stresses. IV. Calibration of Impact Testing Machine. V. Summary of Materials Section Tests on various Timbers. VI. American Tests on various Timbers. This chapter is mainly based on the researches made by Major A. Robertson, of the Materials Section, and compiled from his reports. Timber has held its position up to the end of the war as the best material for spars, struts, longerons ani many other parts of aeroplanes, owing to its remarkable strength and lightness; it is, however, probable that steel will ultimately replace wood for spars and struts in all large machines; steel tubes are already used extensively for struts. The conditions in which timber is used in aeroplanes are very different from those in any other engineering structure. Timber is rarely used in machines and experience of its properties is mainly confined to its use in buildings, ships, docks, scaffolding, etc., where it is used in large sizes and where its exact strength is not important. Its use in aeroplanes is more analogous to its use in furniture, but there again no calculations of strength nro needed. In aeroplane construction it is used in relatively small sections in the form of columns and beams, the strength of which must be subject to accurate calculation. Most of the data on the strength of timber before it was required for aeroplanes have related to its strength and other properties in largo baulks, often in the green, unseasoned condition, and are based on practical experience so different from aeroplane construction that they are quite useless for this purpose. What is wanted is accurate information on the strength and elasticity in all directions of the grain, on which complex calculations, such as those on the strength and stiffness of continuous beams or on propellers, mav safely be based. Extensive investigations have been made on these lines into the strength of many different timbers in laboratories all over the country during the war. The problem is complex because wood is not an isotropic material, its properties differ widely in different directions of the grain, and the difficulties of the in- vestigation are considerable because wood is not a uni- form material, but differs from point to point in th\> 96 same tree, and each tree differs from its neighbour. Never- theless, great advances have been made in our know- ledge of the mechanical and physical properties of various timbers, but chiefly of Silver Spruce (Ficea sitchensis, Carr.) the most important timber for spar and strut construction. Tests on the strength of Spruce at all angles with the grain were begun early in the war, and despite many difficulties have been brought to a successful conclusion, and similar tests have been made on ash, walnut and mahogany. From the re- sults obtained it has been possible to calculate the nine elastic constants for each of these timbers. The results are given in Figs. 22 to 52; they apply, of course, only to the single specimen of each timber which was tested. Much work remains to be done in this direction. The first report on Silver Spruce (Admiralty C.I.M., No. 2) issued in 1916 is reproduced below. It includes tables giving the tension, compression, and shear strengths, and elastic moduli of a good sample 'of the timber in all three principal directions of the grain. The shear tests were not very accurate, but they served to show the order of magnitude of the various strengths and moduli. Fuller data are given in later sections of this report. 1. PRELIMINARY REPORT ON THE STRENGTH OF SPRUCE. Silver Spruce is neither homogenous nor isotropic. but its elastic properties are so good that the ordinary engineering theory of bending as used for long beams and struts is applicable. Silver spruce obeys Hooke's Law, i.e., the stress strain diagram is a straight line up to the elastic limit. The elastic limit in compression is about 80% of the ultimate strength. The values of the shear modulus for beams cut along the grain and bent in a longitudinal plane are very low roughly 3% of the value of Young's modulus along the grain so that in short beams the deflection due to shear is not negligible. The effect of the low shear modulus on the theory of struts has been in- vestigated; it is less than 1%. The elastic limit and ultimate strength of silver spruce along the grain in compression are very much lower than in tension so all beams and struts fail by compression first. Failure by shearing usually only occurs in beams with very thin webs. It may occur in very short beams but these usually fail by crushing. The deflection-load curve for a beam is straight up to a point (often called the elastic limit) and then bends off up to the point of rupture. The end of the straight part of the curve corresponds to the commence ment of actual crushing on the compression side, an.1 the calculated stress in the extreme fibres at this point agrees closely with that which causes failure in direct compression tests. As soon as this point is passed the distribution of stress across the section changes, so that the stress at the point of rupture cannot be calculated. The modulus of rupture does not corre- spond to any fundamental property of the wood. The shear strength of silver spruce is much greater than the values given in the text books and published papers. In the methods usually recommended for testing for shear strength the timber, in addition to shear stress, is subjected to a tensile stress across the grain, and it is this latter stress which causes the failure, and not the shear stress. In the table of strengths given below the average values of the shear stress and the maximum values (assuming parabolic distri- bution) are both given. The ultimate tensile strength across the grain (in any direction) is very small. Timber frequently fails under small secondary stresses across the grain which cause it to split before it fails under the principal stress; such stresses must be carefully guarded against. There is nothing in timber stressed across the grain corresponding to the plastic flow in steel which allows a localised stress to distribute itself, so that a very small force may be sufficient to start and extend a split. . The remarkable differences between the strengths and elastic constants of a sample of silver spruce when tested in different directions is shown in the following table. The figures in the table were measured on .a single very good sample and must not be taken as average values. The direction of the stresses are defined by means of three axes fixed relatively to the trunk of the tree from which the timber was cut, as shown in Fig. 1. FIG. 1. O L is along the grain. E across the grain radially T across the grain tangentially. A shearing stress is defined by the plane in which it acts and its direction in that plane. These may be indicated by suffixes, the first defining the normal to the plane and the second the direction. Thus Prl re- presents a shear stress in the plane LOT (perpen- dicular to OR), in the direction O L. Diagrams are given in the table to make the direc- tion clearer. TENSION AND COMPRESSION STRESSES. Silver Spruce. Tension, Lbs. per sq. in. Compression, Lbs.persq.in. Axis... Ult. El. Limit. Modulus. Ult. El. Limit. Modulus. X K> X 10 L .. 18,000 Over 10,000 1-8 5,000 4,000 2-0 OR ... SOOto 1,000 280 05? 600 270 0-09 to -12 OT ... 400 to 800 200 or>? 700 180 0-06 to -07 CHAPTER X.] [Ti> face page 96. FIG. IA. Spruce. Cross Section. R.O.T. ' ,. ! M FIG. 2. Spruce. Radial Section. L.O.H. FIG. 3. Spruce. Tangential Section L.O.T. L FIG. 4. Crushed Spruce. Radial Section. L.O.R. 97 SHEAR STRESSES ix RECTANGULAR BEAMS OF SILVER SPRUCE. g .n 03 I Elastic Limit. Lbs. per sq. in. Aver- age. Par- abolic Ultimate Stress. Lbs. per so. in. Aver- age. Par- ibolic. Modu- lus. Lbs. per sq. in. LOR Prl Plr LOT 550 820 870 1,310 60,000 LOT Ptl Pll LOR 250 910 1,370 63,000 TOE Prt* Plr* Very small. 200 Approxi- mate!' . 2,300 Approxi mately. " These shear stresses may conveniently be termed " Rolling Shear Stresses," as they tend to roll the fibres of the wood rounJ. 2. EELATION BETWEEN STRENGTH AND MICRO-STRUCTURE OF TIMBER. The tensile strength of spruce is about three times its compression strength. The contrast between timber and metal in this respect is striking. An investigation into the cause of the comparative weakness of spruce in compression was made at the request of the Materials Section by Professor Lang and Mr. Robinson, and an apparently satisfactory explanation was found in the microscopic structure of the wood. The structure of spruce is shown in the micro-photographs, Figs, la to 5. Fig. IA is a cross section, cut in the plane ROT (i-ic. Oregon Pine. FIG. 5u. Mahogany. FIG. ~>E. Louisiana Cypress. 99 In almost all the air-dry timbers tested there has been found a considerable drop of stress immediately after the ultimate stress has been passed. This is well AUTOGRAPHIC LOAD- DEFLECTION DIAGRAMS COMPRESSION. OREGON PINC. POON t'lG. 9. shown in the autographic records. Fig. 9. In spruce this drop amounts to about 18% of the failure stress. This drop of stress depends to some extent on the rate of loading which also affects the ultimate stress. In green timbers there is scarcely any drop but the material flows steadily under the load. Rate of Strtiinintj. The effect of rate of straining has been investigated for compression and bending tests by Tiemann (Proc. Am. Soc. for Testing Materials. Vol. Vin.. Philadelphia, 1908). He found within the range of rates that he used that for compression the ultimate strength does depend on the rate, but the elastic limit and modulus of elasticity are practi- cally unaffected, and, for bending the modulus }f rupture in affected and not the modulus of elasticity. A summary of Tiemann's results is given in the following table. It is noticeable that the wet timber is much more affected than the drv. Rate of Strain.* Per min. 0002 0006 0018 0054 0162 0486 Compressed Tests : Rel. strength, Dry 100 100-8 102-7 1 1 ):> 5 iox-3 llll Wet ... 100 103-4, 107-5 113-9 121-3 128-8 Bending Tests : Rel. strength, Dry '... 1(1(1 102-1 ior,-8 108-6 KJ'.I-I; 110-0 Wet ... 100 105-1 111-3 117-9 123-7 126-3 * In a compression test a "rate of strain of '0002 per minute" means that the test piece is reduced in length by -0002 times its length every minute. In a bending test the same expression means that the extreme fibres are extended on the bottom and compressed on the top by -0002 times their length every minute. 27264 The standard American rates of straining are 'OOlo per minute for compression and '001 per minute for bending. This is approximately the same rate as that specified in Air Board Specification V.I, which was arrived at after a short series of trials. Comparatively large changes in the rate round about these figures do not affect the Ultimate Strength greatly. It was at one time suggested that the value of FJ found by the vibration method (in which the rate oC straining was very high) would be different from that found by the static method. Several comparative tests have been made but in all cases the agreement was very close indeed. The value of E therefore is practi cally unaffected by the rate of straining. Shear Tests These are the most difficult of all timber tests, and many of the methods used are entirely erroneous. Modulus of Rigidity. The shear moduli are not readily determined satisfactorily. Torsion tests which are usually the most satisfactory for shear problems always involve two of the moduli. In a circular speci- men the mean modulus of rigidity for torsion about the axis of L is , ^ 2 where ui and u.-, are the two pi + pa moduli involved. Fairly good values can be obtained by using a very deep beam and measuring the de- flection of a point on the neutral surface relative to a cross section about 1 inch away. This method is open to objection on account of the uncertainty of the stress distribution. Using the type of specimen shown in Fig. 13, which has been found useful in obtaining the ultimate strength in shear, with specimens 5 inches deep it is probable that the error is comparatively small. In these specimens the stress is assumed to be uniformly distributed and the strain is probably uniform in the region tested. Typical results are as follows : Moduli of Rigidity. Calculated /'? t'l f'3 Mean Mo- dulus for LR. LT. RT. Circular - Specimen. Ibs.per sq. in Ibs.per sq. in. Ibs.per i-q. in. Ibs.per sq. in. Spruce 10-4 X. lO'j 7-2 X 10 4 0-46 x HI 4 8 5. X 10 4 Ash 12-3 X 10* 8-1) X 10' 3-6 X 10< 11-3 X 10' Walnut 13-9 X 10 4 10-1 X 10* 3-4 X 10 4 11-7 X 10< Mahogany.. 8-75 X. 10 4 6-8 X 10 4 2-22 X 10 4 7-65 X 10 4 For circular specimens, tested in torsion, cut from the same pieces of timber the observed mean moduli of rigidity were: Spruce 8-2 x 10' Ash 11-3 x 10 4 Walnut 14-0 x 10* Mahogany ... ... ... ... 8'1 x 10 4 Ultimate Strength. The usual methods of deter- mining shear strengths are all faulty and do not give the real sheer strength of the timber tested. Thus the method adopted by Johnson, see Fig 10, and also by Julius gives values which are frequently only one half those obtained by Warrsn's method (Fig. 11), or N2 100 by that adopted by the U.S.A. Forestry Service (sec Fig. 12). In all these tests the surface which it is intended to shear is also subjected by the conditions of the test to a bending moment and therefore to ten sjon. This tension is at right angles to the grain, in which direction the timber is very weak, so that the failure is initiated by tension. Thus spruce has by the Johnson type of test a shear strength of about 300 Ibs. per square inch, and by the Warren type of test about 400 Ibs. per square inch, whereas the shear tests about to be described and the compression and tension tests at an angle with the grain (see Section 4) point to a figure of about 1,100 Ibs. as being the real shear strength of spruce. Some tests have been carried out in an apparatus shown in Fig. 13. This apparatus is a modification oi that used by Professor Coker in his optical tests on the distribution of shear stress across a thin plate (Proc. E.S.A., Vol. 86, 1912). The test piece may be regarded as a cantilever loaded at the end by a force and a couple. The resulting bending moment diagram is shown in the figure ; the bending moment is nil at the middle section, while the shear stress is approxi- mately equal' to the force divided by the area of the cross section, i.e., the distribution of shear stress is rectangular, not parabolic (vide loc. cit.). The results obtained were : Ultimate Shear Strengths. Prl, Plr. Ptl, Pit. Prt, Ptr. Lbs. per square inch. Lbs. per square inch. Lbs. per square inch. Spruce Ash Walnut... Mahogany 1,200 2,000 2,000 1,540 1,100 2,000 2,000 1,050 Approximately 300. ., 5UO. 500. 500. The value of the shear Prt is very difficult to deter mine as, even when using this type of specimen, the timber (which is cut across the grain) generally fails by tension. The approximate values given have beei 1 deduced from the results of tension and compression tests on specimens with the grain at 45 to the axe<3 R and T. Whatever method is adopted of applying a shear stress there will always be two equal shear stresses i>i planes at right angles to each other, and the timber will fail in the plane in which it is weakest. For example, if a thin radial or tangential plank is tested in this apparatus it is immaterial whether the grain is placed vertically or horizontally, failure will take place along the grain in either position. It is impossible to break the plank by shear across the grain and it is therefore impossible to determine by direct test three of the six shear strengths. The plane in which the failure takes place is stated in the third column of the table (page 99). The higher strengths across the grain are of some value in Plywood. The results of torsion tests on spruce, mahogany, ash and walnut are shown in Fig. 14. The values of the moduli given are the mean values. The values of the elastic limits are calculated on the reasonable assump- tion that up to tlie elastic limit the stress varies from the centre to the outside fibre in linear relation with the radius. The values of the ultimate strengths are indefinite, as the stress distribution is no longer known after the elastic limit is passed. The values tabulated are calculated on two assumptions, the first that the stress varies in proportion to the radius, and the second that the stress is uniform from centre to outside. Rate of Strain intj. The effects of time on the torsion moduli have been investigated to some extent by MessrSj Griffiths and Wigley (Advisory Committee for Aero- nautics, Report T. 1077). They found that in any torsion test there is a creep with even the smallest loads. This creep was greatly affected by temperature, so much so, that at 90 C. the twist after the same time may be several times its value at ordinary tempera- ture. They have also made vibration tests to determine the mean value of N, and find that the value may be 20% and in some cases 30% and 40% above the value obtained under the normal rate of loading. Some few tests have been done at high temperatures by the vibration method and these indicate that at 95 C. the value may be 25% less than that at ordinary tempera- tures. Their values of the moduli of rigidity deter- mined both statically and dynamically are given in the following table : Material. a -^ Modulus Modulus 6-2 '3 a of Rigidity by Vibration of Rigidity by Static OQ^ Method. Method. American walnut (ordinary temp.) 4 Ibs. per sq. in. 150 x 10 6 Ibs. per sq. in. 133 X 10" )5 at 72 C. 4 129 X 10 6 ., ., at 95 C. 4 112 X 10 6 )) ,, 2nd sample (or- (dinary temp.) 1 ' 148 X 10 6 12(1 X 10" J) 3rd sample (or- dinary temp.) }. 135 X I0 fi 132 X 10 6 Mahogany IB 111 X 10' J03 x lye 4 103 X 10 6 090 X 10 6 1 '090 X 1(J 6 086 X 10 6 Spruce 2 134 X 10 6 "080 X 1 6 4 '0*18 X 10 6 "073 X 10 6 " '124 X 10 6 '0 ( )0 x 10 fi 6 '096 X 10*" Light PadouK 4* 215 X 10" 220 X 10 6 Dark u 270 X 10 6 American Birch 1 158 X 10 6 159 X 10 s M ,. 7 179 X 10 148 X 111* 8 174 X 10 6 166 X 10' fib 209 X I" 6 165 X 10 The rate of loading employed in the tests recorded in Fig 14 was such that full load was reached in about 15 minutes. Failure under Prolonged Loading The load that a piece of timber will carry for an indefinite time is very considerably less than the values of the ultimate strength found in any of the ordinary tests. Dr. It. H. Thurston, in America, found that 00% of the ordinary ultimate stress would break yellow pine beams if left for some nine months; Tiemann states that dry long- leaf pine beams may be safely loaded permanently to within at least 75% of their immediate elastic limit, and the deflection will ultimately cease under that load. Fortunately in aeroplane work these prolonged applications of load are never experienced as the length of time during which the load in a member approximates to the maximum load is extremely short. It may, Chapfcer.X. OJ cr ai UJ V / 2 O H to 250 30O 35O 400 DEFLECTION* IN I/IOOO INS. FlG. 17. MODULUS TOR RUPTURE or I BCAMS(5*1i) 2-0 y CORRESPONDS TO FRACTURE BY TENSION AT 18.000 -4 -6 -8 10 VZ t 4- 16 DEPTH OF COMPRESSION FAILURE 1-8 Chapber.X. FlG. 20. LIMIT5 OF COMP&E55IVE 5TDEN6TH 50 40 5O 60 ANGLE Of GBAJN 6 FIG. 2 I. 20000 IQOOO 18000 r o z S Q. o go in U.I a vn LIMITS OF TENSILE STRENGTH 30 40 5O 60 ANGLE Of 6CAIN ff 70 80= Q0 M/ZO. Chapber.X. SPRUCE (AVERAGE MOISTURE 12-2% ) RADIAL PLANK (PLANE L.R) VARIATION or ULT COMPRESSION STRENGTH WITH ANGLE OF GRAIN. TOOOl 10' 20* 30* 40* 50*60* 70* 80*3O. ANGLE OF GRAIN CHAPTER.X. SPRUCE (AVERAGE Fio.23. TANGENTIAL PLANK (PLANE LT,) VARIATION OF LJLT: COMPRESSION STRENGTH' WTTH ANCLE OF GRAIN. O* to' 30" 40* 50* 60 10' ANCLE OF GRAIN 90 Fio.24-. S PRUC E /AVERAGE MOISTURE >a-2*/. } CROSS SECTION ( PLANE R.T.) VARIATJON or ULT. COMPRESSION STRENGTH WITH ANGLE OF GRAIN. 10* 20' W ) S0 60^ TO- 60' 90 ANCLE OF GRAIN tO.000 l-t- F 10 .25. SPRUCE PLANK W.2. VARIATION OF TENSILE STRENGTH. Fio.26. sy SHEAR STRESS of tsoo ja/a" w 30' 4* so" -W W *o- to' *> scT ANCLE or GRAIN f ASH.(AvERACE MOISTURE 1*9^ PLANK ( VARIATION OF Uu-p COMPRESSION STRENGTH WITH ANCLE OF* GRAIN. AWCLE. or GRAIN. Fto.29. ASK (AVERAGE MOISTURE 13-9 */.) CROSS SECTION (PLA.NE T.&) VARIATION OF ULT COMPRESSION STRENGTH WITH ANCLE OF GRAIN 7.000 rrso ISOO IS.5O 1000 03 750 500 250 IO* 2O* 3O* 40* SO* O T 70* 80* ANGLE OF GRAIN. FiG.3O. ASH. (AVERAGE MOISTURE CROSS SECTION (PLANE TR) VARIATION OF ULT TENSILE STRENGTH WITH ANGLE OF GRAIN. MOO 1-500 O* IO* 2O* 3O* 4O* 50* 60* 70 80 T ANGLE OF GRAIN. R ULTIMATE TENSILE STRENGTH, AtONC THESRAJNH CHAPTER.X. Fic.31, WALNUT (AVERAGE MOISTURE 11%) RADIAL PLANK (PLANE LR)~ VARIATION OF ULT: COMPRESSION STRENGTH WITH ANCLE OF GRAIN, 650t r F|C.3Z. WALNUT /AVERAGE MQ.STUPP ..*) TANGENTIAL PLAKIK ( PLANE LTj VARIATION OF ULT COMPRESSION WITH ANCLE Or GRAIN. 10* tar *> 40* so* eo' 70' eo* o* ANSLE OFORAIM. WOO 10* * 30' 40* 60' SO* TO" 0* ANCLE OF GRAIN. WALNUT. (AVERAGE MOISTURE it' CROSS SECTION ( PLM>IE RTJ~ \^RIATON OF ULT: COMPRESSION STRENGTH WFTH ANCLE OF GRAIN. mD SUOO 2000 1300 I" 00 5>o / / 1 f I / i / \ /' [ / I / \ / * ^^ j* r M ^' 1300 ( 1 > W ao" 30* 40* 50' 60' -70* t?* ANCLE or GRAIN . F.o.34, WALNUT (. AVERAGE MOISTURE \\%) CROSS 5ECTION (PLANE R~f]~ VARIATION OF ULT TENSILE STRENGTH WITH ANCLE OF GRAIN . TENSILE STRESS L&S/SQ-.1N-. WOO 200O 1800 .1800 1700 MM y / / y /^ 1500 y " r 1300 1200 1100 1000 / ( / ^X ^ taMi , -* 10* HO* 30' 40* SO* 60* ' O* 90 T ANGLE OF GRAIN. CHAPTER.X. Fic.35. MAHOGANY (AVERAGE MQ.STURE RADIAL PLANK (PLANE LR) VMHATION OF ULT COMPRESSION STRENGTH WITH ANCLE OF GRAIN no. 3 6 (MAHOGANY AVERAGE MO.STURE g+ % ) TANGENTIAL. PLANK (PLANE LT). VARIATION OF ULT COMPRESSION STRENGTH WITH ANCLE OF GRAIN. 10* ao* 30' 40* so* eor 10' to' so* ANGLE OF GRAIN. c&'oa 6OOO 5500 5000 4300 4000 3500 3000 ; t i ^ \\ s \ 1 , esoo aooo 1500 1000 >y \ 3 5 r^* 500 ANCLE. OF GRAJN. Fic. MA HOGAN Y AVERAGE MOISTURE n-6*/ ) CROSS SECTION (PLANE RT) VARIAT>ONOF ULT : COMPRESSION STRENGTH WITH ANtCLE OF GRAIN MAHOGANY (AVERAGE MOISTURE n CROSS SECTION ( PLANE TR) VARIATION OF ULT TENSILE STRENGTH O- K>* 20* ANCLE OF GRAIN woo WITH ANCLE OF GRAIN.. / / I8CO 1100 IOOC 900 800 / ' ^ 1 , X -1 r ' ^x X* 600 500 ?* 0' 10* 80* ISO' 40* SO' 60' 70' 80' *>' ANCLE OF GRAJN. CHAPTER. X. F.o. 39. OPRUCE AVERAGE MOISTURE 12 -g% RADIAL PLANK (PLANE L.R) VARIATION OF E WITH ANCLE OF GRAIN . Fio.4-1. SPRUCEfAvERACE MOISTURE 122 V.) CROSS SECTION ( PLANE Rjl~ VARIATION OF E WITH ANCLE OF GRAIN 10' 20* 30* 40* 50' 60' 70' BO' 90* ANCLE OF GRAIN. F.G.4O. o PRU C E (AVERAGE MOISTURE 12 g*/. ) TANGENTIAL PLANK (PLANE LJ.) VARIATION or E WITH ANCLE OF GRAJNJ Z ) TANGENTIAL PLANK (PLANE LT.) VARIATION OF E WITH ANGLE OF GRAIN . Fio.45. WALNUT /AVERAGE MOISTURE RADIAL PLANK (PLANE LR) VARIATION OF E "WITH ANCLE OF GRAIN o* to* eo* 30* 40* o* TO* ao* so* ANCLE OF GRAIN. FIG. ASH. (XvERACE MOISTURE 133 tt CROSS SECTION ( PLANE T.Rj VARIATION OF" E WITH ANCLE OF GRA\N 10* tO* 50* 40* SO" 60 70* 60* SO' ANGLE OF GRAIN Fic.46. WALNUT (AVERAGE MOISTURE ii'/J TANGENTIAL PLANK /PLANE L.T j VARIATION OF E'wrm ANCLE OF GR/MN ^ z ^ / J N. 1C x^ k== 1 . ^^iH^ iF" r" 1 UJ O* 10* 20* ISO* 40* SO' >0 ISO* 40* 50-60* 70* ANCLE OF GRAIN . 80" 90* z I ' 2 jj ro 1*' 14 l-e JO 8 6 4 a < \ \ V \ V \ \ \i >t-v. O* eo* 30 - 4O' SO' 60' TO" 8O* 9O' ANGLE OF GRAIN CHAPTER. X. F.c. 47. WALNUT (AVERAGE MOIST, mr n't) CROSS SECTION ( PLANE RT) VARIATION OF~ E "WITH ANCLE OF GRAIN. FiG.49. MA HOC ANY (AVERAGE MOISTURE -ray/.) IANCELNTIAL PLANK (PLANE LT) VARIATION orE WITH ANGLE or GRAIN f.(J [IP 1 18 ^* M X 1? / 10 .*x LX i 08 >- r - 06 < 1" 1C > a a* 3 0" 4 y s 0* G 0* 7 0' ft 0* 90' ANGLE OF GRAIN. Fio.48. MAHOGANY ^AVERAGE MOISTURE: RADIAL PLANK (PLANE LR) VARIATION or"E WITH ANGLE OF GRAIN. 0' 10' to' 30* 40* 50" 60' 70' 80* 9O* ANGLE OF GRAIN. FIG. SO. MAHOGANY /XvERACE MOISTURE -ii-e'/.) CROSS SECTION ( PLANE RT) VARIATION OF E WITH ANCLE OF GRAIN 20* 30- -4o- SO" SO* 70" SO' 30' ANCLE OF GRAIN o X Z d n J Ld 16 14 12 10 M ./ ^" / / r~~ S 06 CM oe 0* 10' 20- 30- 40' SO* 60' 7O* 0* 90* ANCLE OF GRAIN CHAPTER. X. FIG. 51. PLANK Z UNIFORM BENDING VARIATION OF MOD: OF RUPTURE WITH ANGLE OF GRAIN. Fio.52. SPRUCE PLANK UNIFORM BENDING. VARIATION OF"F* WITH ANGLE OF GRAIN. % 10* 80* 3O* 40* SO* 60* 70* CO* SO* L E OF GRAIN . 0* 10* t R and q taken from the table I. It is clear that as the load p is increased the test piece will fail when it reaches the lowest of the three values given by the equations. The maximum ultimate strength of the timber is therefore given by the broken line A B C D. The same equations hold for the tensile stress if the constants p L and p R are changed in equations (iv.) and (v.) to the tensile strengths along and across the grain. Equation (vi.) remains the same. The corresponding curves are plotted in Fig. 21. If the timber fails in ways other than those assumed the strength may be lower than that given by these considerations. Tests on various timbers show that the curves for shear and compression at right angles to the grain are closely followed, but that for small deviation of the grain the values are lower than those indicated. This is due to the timber failing in a plane at right angles to the stress and not at right angles to the grain as is assumed in the above theory. Spruce The results of a series of tests on a sample of spruce are given in Figs. 22 to 26. Fig. 22 shows the variation of Compression Strength in plane LOR. Fig. 23 shows the variation of Compression Strength in plane LOT. Fig. 24 shows the variation of Compression Strength in plane TOR. Fig. 25 shows the variation of Tension Strength in plane LOR. Fig. 26 shows the variation of Tension Strength in plane TOR. It will be seen that in the planes LOR and LOT the points lie reasonably near the theoretical curves for failure by shear and compression at right angles to the grain, but that for small deviations the strength decreases rapidly with increase of the angle. In this region the plane of failure runs square across the speci- mens and not at right angles to the grain as it would if failure were conditioned by the compression strength along the grain. The tension tests in plane LOR are also in good agreement with the theoretical curves for shear and tension at right angles to the grain. In tension a very small deviation, not more than 2| is sufficient to cause failure by shearing. It is not surprising, there- fore, that it is difficult to obtain real tension failure in timber since with quite small deviation the shear stress is sufficient to produce failure. The tension stress is of the order of 18,000 Ibs., or more than three times the compression stress. The results for the strengths in plane T R do not follow any of the curves closely except those in com- pression for deviations between 20 and 60. The shear strengths deduced from these tests are in fair agreement with those found by the direct shear tests. The shear strength Prt appears to be of the order of 200 to 250 Ibs. per square inch. Ash, Walnut and Mahogany. Similar sets of curves for specimens of ash, walunt and mahagony are given in Figs. 27 to 38. Ash Fig. 27. Compression in plane LOR. 28. LOT. 29. TOR, ,i 30. Tension TOR. Walnut. Fig. 31. Compression in plane LOR. 32. LOT. :, 33. TOR, 34. Tension TOR. Mahogany. Fig. 35. Compression in plane LOR 36. LOT. ;, 37. ',, TOR, mrflJ 38. Tension TOR. Young's Modulus of Timber in Directions inclined to the grain. The calculation of the value of E for any specimen whose axis is inclined to the grain is somewhat complex. It depends on all the shear moduli and principal Poisson's ratios. The equation connecting E and Q for inclinations in a principal plans can, however, be obtained if the value of E for 6 = 45 is obtained in addition to the values for 6 = and 6 = 900. The equation then becomes : T^ _ 1 } E 0=0 = 45 The results of a number of tests on spruce, ash walnut and mahogany are given in Figs. 39 to 50. Spruce. Fig. 39. E/e, Curve in plane LOR. 40. , ,. LOT. 41. ,. TOR. Ash. Fig. 42.- -E/0, Curve in plane LOR. ,, 43. LOT. 44. TOR. Walnut. Fig. 45. D/fl, Curve in plane LOR. ,. 46. LOT. 47. ., TOR. Mahogany. Fig. 48. E/e, Curve in plane LOR. 49. LOT. 50. , TOR. The points plot fairly well on the theoretical curve. There is however a tendency for some of the tests at 104 small angles of deviation to give rather low values. This may be due in part to the end effects on the short specimens used for these compression tests. Bending Tests on Beams of Spruce cut at various angles with the grain. The results of a series of tests on beams of spruce Z cut in the plane LOR (i.e.. deviation in radial direction) are given in Figs. 51 and 52. Young's Modulus, E. The points plot fairly well on the theoretical curve. Modulus of Rupture. As explained above the modu- lus of rupture is not connected with the fundamental properties in any simple way. The results (spruce only) have, therefore, merely been plotted in Pig. 52, and a smooth curve drawn through them. The results of these bending tests are in the main in good agree- ment with compression tests on the same spruce. All these tests bring out very forcibly the importance of straightness of grain of timber that is to be used for construction work either as spars or as struts. 5. ELASTIC CONSTANTS OF TIMBER. Since timber is not a homogeneous material but a complex cellular structure the theory of elasticity does not strictly apply to it, but as experience shows that even the simplest approximations commonly used for ordinary strength calculations give results of fair accu- racy, it is reasonable to attempt to carry the elastic theory a step further so as to enable more complex calculations to be attempted. For this purpose the timber, still assumed to be homogeneous, is treated as anisotropic with three axes of symmetry, namely, the longitudinal, radial and tangential directions. On these assumptions the elastic constants may be calculated and the accuracy of the approximation may be checked by comparing results of direct tests with figures calculated by means of the elastic theory from tests of a different type. The elastic constants have been obtained for selected specimens of spruce, ash, walnutt and ma- hogany, and as a check, the values of the shear moduli have been calculated and compared with those obtained by direct shear tests. The agreement is good for walnut and mahogany but not for spruce and ash. This result is probably connected with the nature of the cell structures of the timbers; in walnut and mahogany the structure is much more homogeneous than in spruce and ash. St. Venant's theory shows that for an anisotropic material with three axes of symmetry there are nine fundamental elastic constants. Adopting the symbols used in Todhunter and Pearson's History of Elasticity, we have : The axis X is radial ; Y is tangential ; Z is longi- tudinal. XX, YY and ZZ represent the Stresses along the axes. Sj, S,, and S, represent the Strains along the axes yz, zx and xy ., ,. Shear Stresses in the principal planes. pi, /<} and pz ,, ., Shear Moduli in the principal planes. if represents 6 Poisson's Ratios ; Strain along OY thus ;,,, = =- i-f^f Strain along OX when stress is in direction OX . a y ., (T,. e and ,ts as cut by a plane surface, and it is at once seen that the resin ducts can only appear in Figs. 53 to 56 as small dots, which indeed may be seen by a close inspection. It is clear then that the deviation that requires most attention is that in the tangential direction and that this is most readily de tected on a true tangential surface. Detection of Spiral Grain Spiral grain differs from diagonal grain in that the deviation varies with the dis- tance from the centre of the tree. For its complete determination therefore it is necessary to observe the deviation on each face of a plank or part. In general, however, since the deviation is greatest at the largest radius, it will be sufficient to determine the deviation at the face where the curvature of the annual rings is least. 7. BRITTLENESS. Brittleness is a property of timber, the effects of which it is very difficult to assess properly. It some- times happens that timber with a good compression value and a good E breaks off quite short. In the beam test this is shown by a small deviation of the stress strain curve from the straight line after the elastic limit has been passed. Some timbers, andaman padouk for example, appear to be elastic practically up to the ultimate stress. It is not easy to decide whether this property spoils the timber for any portions of an aero- plane, but it is safer to reject it. The simplest test to detect brittleness is the notched bar test specified in several of the Air Board Specifica- tion, e.g., V.4. The standard test piece adopted is shown in Figs. 61 and 62. If the radius at the bottom of the notch is kept the same, viz., -^ inch, the notch may be cut open to 90 or left as a saw cut, as shown in the figures, without producing an appreciable differ- ence in the results. A simple pendulum apparatus for breaking the test pieces is illustrated in the specifica- tion and a metjiod of calibration is given in Appendix TV. It is important that the blow should be struck in the tangential and not in the radial direc- tion, as with many timbers there is a decided difference between the strengths in the two directions. A good specimen always shows on the tension side a large number of fibres cleanly broken, and on the compression side a fibreless structure, i.e., the typical fracture on the compression side of a beam, while a very bad sample is marked by a complete absence of fibres, the whole section having practically the same appearance. A photograph of a good sample of English poplar and a bad sample of spruce are shown in Fig. 63. The exact cause of brittleness and its relation to other properties is not fully determined. It does not appear to be produced by treatment in a drying kiln under normal conditions, though brittleness is a characteristic of timber which has been seriously over- dried. As a general rule it would appear that it is more nearly connected with the tension value than with any other property. Certainly brittle specimens fail quite " short " in tension at values which are decidedly lower than those of good samples. 8. FATIGUE. The fatigue of timber may be an important matter in connection with the strength of parts subject to vibration, for example, in engine bearers and all parts of aerofoils and fuselage structure which vibrate. Almost nothing is known on the subject. A few tests have been made but they are not sufficient to base any con- clusions on. It seems probable that the difficulties which are met with in static testing will be more serioua in fatigue testing, hut the subject is well worth in- vestigation. So far as is known there is no evidence of fatigue failures occurring in use. , 9. STRENGTHENING TIMBER BY CHEMICAL TREATMENT. A lengthy investigation was made by Professor Lang and Mr. Eobinson to ascertain whether spruce could be strengthened by impregnating the wood with any substance which would stiffen or harden the cells. A large number of substances were tried, but it was found that most substances could not be made to enter the wood to any but a trifling depth below the surface, and that the few substances which would enter it had little effect on its strength. In one or two cases in- crease of strength was obtained, but the increase of weight mere than counterbalanced the gain of strength. Claims are made for some methods of seasoning timber that they increase, its strength, but no basis for these claims has been found for those processes which have been rigorously investigated. Apparent increases of strength are easy to produce by over-drying the wood, but the increase disappears when the normal moisture is re- framed. The strength of timber increases considerably as the moisture content is reduced (see Section 15. p. 113), and the possibility of making practical use of this extra strength was fully investigated, but no means could be found of completely sealing the timber against the slow entry of moisture (see Section 16, p. 117) The following report by Mr. Eobinson gives particulars of some of the trials. TTTE EFFECT OF IMPREGNATION WITH SODIUM SIT.ICATK ON THE STRENGTH OF SII.VEK SPRUCE. Some preliminary experiments on the injection of spruce wood with a solution of sodium silicate have CO LjJ O 5 a: k 55 b a: LU H o o u Q: D h O I E (t UJ CQ X o s Z 3 a z 1- z F cc o V) Ul a g HRECTIC J < P Z Ul NUAL R u. > u O Z i S 5 < Z Tg ^ ..ur. "I ^ V o 3 o 5 a' 2 ? > CQ g o 8 a a o "I > ^ u O Q O O O o t o lO Ul a u I x I 1 le.> *w- >i< i *K) J *p N * 7y w 53 il ^ 8 a: a. cO u. o o o ui CC a. Q U O s JE i 5 U cc hi 4 o flf a u o i age 106. o. 02 - bo o a S To face page 107.] [CHAPTER X. FlG. OS. Tough English Poplar. Brittle Spruce. FIG. 66. Warping. FlO. 67.;; Twisted Box Spar. 107 resulted in an increase iu the strength of the wood as tested by end-wise compression. A large number of cylindrical test specimens (.1 inch long x J inch dia- meter) were impregnated with a 35% solution of sodium silicate. The air was first exhausted from the speci- mens, and the solution then forced into the wood under atmospheric pressure. The treated specimens were thoroughly dried at the temperature of the labora- tory, and were then brought to different moisture con- tents and tested in endwise compression. In this wa.v the strength-moisture curve (Fig. 64) was obtained. For comparison the strength-moisture curve for the same wood untreated is plotted alongside. The in- crease in strength holds for all degrees of moisture, and the strength is at no point less than 30% higher than that of the control wood. Table 1, however, shows that the treatment increases the weight on the average by about 28% the increase in strength would therefore appear to correspond closely to the in- crease in weight. TABLE I. Initial dry weight. Dry weight after treatment. % increase in weight. I 2 3 4 1 102 grains. 1-151 1-105 1-039 1 3St> grams. 1-623 1-4045 1-335 25-79$. 31-1% 27-1% 28-3% It must be pointed out however that if treated speci- mens are soaked for two days in water the silicate washes out, and that on drying again the strength of the wood is na greater than that of the control wood. Since the above tests were carried out on small specimens it was desirable to know whether it would be practicable to impregnate large beams of spruce with sodium silicate solution. Trials were therefore carried out in which beams measuring 6 ft. by 2 ins. by 2 ins. were treated with a solution of the same strength as that used in the former experi- ments. The treatment was carried out in a special iron cylinder with air-pump and force-pump attached. The cylinder containing the wood was first exhausted for some hours by the air-pump and then the solution -was drawn in and pressure gradually applied to 100 Ibs., which pressure was maintained for 12 hours. At the end of the treatment one of the beams was cut across at selected distances from the end. The change in colour of the wood due to the penetration of the silicate afforded a sufficient indication of the degree f penetration. Fig. 65 shows that the silicate is con- fined to the immediate vicinity of the ends of the beam, and to an outer shell about -^ inch deep along the length of the beam. A control beam 6 feet long, treated in the cylinder in exactly the same way, but using water instead of sodium silicate solution, was com- pletely penetrated by the water throughout the length of the beam. Other experiments, using shorter beams and warm solutions of sodium silicate, gave equally un satisfactory results as regards penetration. These trials indicate the practical impossibility of impreg- nating large beams of spruce with sodium silicate. It- has, on the other hand, been found possible to thoroughly penetrate thin veneers of wood with the same solution. 27264 A more serious objection to the process is that wood treated with sodium silicate is more hygroscopic than untreated wood. For example, it has been found that while pieces of control wood attained equilibrium is regards moisture content in the laboratory at 12- 1%, the treated pieces contained 15% moisture under the same conditions. In an atmosphere saturated with water- vapour the figures after 42 hours were 22'1% moisture for the control, and 38 4 8% for the pieces treated with sodium silicate. Unless, therefore, the entry of mois- ture could be completely prevented the increased strength would be neutralised by the increase in the moisture content of the wood. In addition to sodium silioate a number of other sub- stances have been tried; it has been found that sodium sulphate and alum (ammonium alum) increase the strength of spruce wood, but to a less marked degree than sodium silicate. On the other hand, sodium chloride and glucose have been shown to have no such effect. With sodium sulphate and alum, as with sodium silicate, the increase in strength was roughly propor- tional to the increase in weight. 10. BUILT-UP SPABS AND STRUTS. All built-up parts depend on the glue and no addi- tional strength is given them by adding bolts or lashings ot canvas or cord, so long as the glue holds. Numerous tests, which are confirmed by experience in use, show that good glue is stronger than wood in all directions, except in tension along the grain, so that if plain built butt joints in parts under tension are avoided, built-up parts are as strong as those made of solid timber. All wooden parts fail first by compression, and the portions under tension have an excess of strength, so that buct joints may be used even on the tension side if they only extend through a part of the section. The glued joints have no effect on the elastic properties of the timber, so the built-up parts may be treated as if they were solid when calculating their strength, stiffness, etc., and no modifications in dimensions need be made when changing from solid to built-up parts. As stated above bolts and lashings add nothing to the strength of a joint as long as the glue holds, they may, however, be used as a " second string to the bow " in case the glue should fail; they are used, for example, in the Canadian Splice described below, which retains 60% of its strength after the glue has entirely failed Lashings may also be valuable in preventing an in- cipient crack in a glued joint from extending all along it; they are therefore recommended wherever a glued joint " runs out," e.g., at the two ends of a scarfed joint or at the ends of the side laminations of a strut. The best methods of building up spars, making splices, etc., are given in the following instructions, originally issued as T.D.I., Nos. 502a, 515a, and 517. The most serious trouble met with in built-up parts is warping. This can only be avoided by using laminae, etc., having the same moisture content and which will expand and contract at the same rate in each portion of the sec- tion. Photo. 66 shows the bad effect of glueing a radial and tangential plank together. A box spar may he built of different timbers in the flanges and webs but the flanges must all be of timber expanding alike and the two webs must behave alike or distortion will occur. Photo. 67 shows an end perspective view of a box spar made with webs having the grain sloping one way on 2 108 one side of the spar and the opposite on the other; in a dry atmosphere the spar is quite straight and free from twist, but in a damp atmosphere it twists seriously, as shown in the photograph. T.D.I. 5176, SPLICES IN WOODEN SPARS. Splices in solid or laminated spars should be made iu accordance with one or other of the accompanying drawings : Plain Sczrfed Splice. This splice is shown in Fig. 68. Properly-made splices of this type are as strong as the- full-sized section of the wood, and retain about 40% of their strength even if the. glue entirely fails. There is no objection to the addition of a few bolts, if the bolt holes are kept within the middle third of the depth of the spar. If bolts are used, three of the ash pegs may be omitted. Vertical dowels and wedges, such as are shown in earlier designs, have been proved to reduce the strength of the splice, and are omitted in the new design; the blunt end to the scarf also reduces its strength; the ends are pointed in the new design. The fabric lashing adds greatly to the strength of the splice if the glue fails; the lashing has been increased in the new design. Canadian Serrated Splice. This splice is shown in Fig. 69. The faces of the scarf are serrated as shown in the section. The serration must be accurately formed and may be cut by spindling. The thin end of the scarf may be supported while it is being cut, by glueing on a backing piece which is afterwards removed. Large- washers or plates must be put under the bolt heads and nuts. The drawing shows both alternatives. Properly-made splices of this type are as strong as the full-sized section of the wood, and retain about 60% oi their strength even if the glue entirely fails. They are therefore stronger than the plain scarfed joint if the glue fails. Position of Splices. Splices are allowed in spars at any unloaded point where the bending moment does not exceed one-half of the maximum ; they are not allowed at loaded points, that is, under the ends of the inter plane struts. T.D.I. 515a,, Box SPABS. 1. Box spars made in accordance with the following in- structions may be substituted for solid or laminated spars : 2. Top. and Bottom Flanges. The top and bottom flanges may be made of a single thickness of wood, or may be built up of laminations. All the laminae of both flanges are to be made of the same timber. 3. Webs. The webs may be made of a single thickness of wood, or may be built up of laminations, or in some designs (see Clause 8) may be made of a high quality plywood. 4. Design. The webs are to be glued to the edges of the flanges, as shown in Fig. 70. Kebating the webs into the flanges as shown in Fig. 71 may be allowed if the flanges are thick, but there is no advantage in this construction over the simpler method shown in Fig. 70 and the inspection of the joint is much more difficult. The depth of glued surface should be at least twice the thickness of the webs. Fillets should be glued in the corners as shown in Fig. 70. Tacking pieces are to be fitted wherever the load is applied, also at all joints in the webs, and also at all scarfed joints through the flanges (see Clauses 5 (a) and 6). The packing pieces are to be tapered at the ends as shown in the figures. Intermediate diaphragms may be used between the packing pieces. 5. Joints in Flwges. (a) A plain scarfed joint, sloped 1 in 9, as shown in Fig. 72, may be used in the flanges, whether made of a single thickness or laminated. A scarfed joint is always to be supported by a packing piece which is to be solid in small spars, but may be made hollow in large spars, as shown in Fig. 73. The pieces which are to form the flange are to be scarfed and glued together and finished to size after glueing. It is not advisable to use screws, but small brads may be used if adequate cramping facilities are not available. The brads should be staggered so that not more than one hole in the wood occurs on any one cross section. (6) In laminated flanges it is rather better to scarf each lamina separately and to stagger the joints. This method is recommended when the laminae are sufficiently thick, say ^ inch or more. In laminated flanges made up of four or more thin laminae, simple butt-joints may be used, staggered so that no two joints come close together, as shown in Fig. 74. (c) Joints made in either of the ways described in the last paragraph need not be supported by packing pieces, but butt-joints in the innermost lamina are to be covered with a butt strap of a total length about 20 times the thickness of the lamina, and butt-joints in the outermost lamina should be treated in the same way or covered by a fabric lashing. BUTT JOINTS.- FLANGE MADE OF THIN LAMINA WITH BUTT JOINTS. FIG. 74 PACKING PIECE. v 1 ,i\ -l-\ 1_,^_. li. v JOINT IN WEB. WRAPPED WITH FABRIC. JOINTS IN WEBS, SUPPORTED BY BUCKING PIECE. FIG. 75. 0. Joints in Webs. A plain scarfed joint sloped Iin9, as shown in Fig. 75, may be used in the webs, whether made of a single piece or laminated or made of ply- wood. The joints are to be made in the same way as those described in Clause 5 (a) for the flanges, and are to be supported by packing pieces. 7. Position of Joints. Joints in flanges are not to be made at any point where the bending moment ex- ceeds one half the maximum bending moment in the spar. This applies to joints in individual laminae, as well as to joints in the whole flange. CHAPTER X. FIG 68. PLAIN SPLICE S/x equal spaces Y \ Lashings 3 Layers of Fabric \ 1. V 5 /l6 cfia Ash afprox. Straight Scarf / in 9. /VOTE . W Bolts are used omit Pegs marked thus X. Bo/ts must be provided with ^arge Washer under B * S E R R AT E D SPLICES Sketch of 1 in 9 Splice showing Two Bolts each end. FIG. 69. r dlr 'Il6 Pfafe or /fr" Washers 3/16 Bofts and Nuts B Enlarged V/ Serration. .Lashing 3 /syers of Fabric - CHAPTER X. Fie, 70 71 7 a & 73. = FtQ-70. FILLET. CROSS SECTIONS OF Box SPARS. JOINT IN FLANGE. I IN 9. JOINT WRAPPED WITH FABRIC. -PACKING PIECE.- -X JOINTS IN FLANGES, SUPPORTED BY BVCKIIMG RECE. ALTERMATIVES FOR LARGE SECTIONS. \ 1 fiq73 PACKING PIECES. 109 Joints in the webs may be made at any point in the spar, but preferably should not be placed at points of maximum shear. 8. Use of Plywood in Webs. The modulus of elas- ticity of plywood is only about 0'9 x 10 6 Ibs. per square inch, whereas the modulus for spruce is TO x 10 6 Iba per square inch. Plywood, therefore, is not well adapted for resisting bending moments. For this reason it is not economical to use plywood webs with light flanges, but when the flanges are heavy, plywood may be used with little loss. The ultimate shear stress for webs made of plywood should not be estimated at more than 800 Ibs. per square inch. The joints in plywood webs are to be made exactly in the same way as those for solid or laminated webs. 9. Crftinping Glued Joints.- The strength of the glued joints depends largely in the efficiency of the cramping while the glue sets. It is important that ample cramp- ing facilities should be available so that the joints may be left to set while subject to uniform pressure during the whole period. If properly clamped it will not be necessary to use screws or brads; if screws are used to help to secure the webs to the flanges they should be staggered so that not more than one screw comes on any cross section. 10. Fabric Lashing. All joints appearing on the out- side (except the joints in laminated flanges which are covered with butt straps) are to be covered with three layers of fabric lashing. No tongues or dowels should be used. A shallow rebate may be used to keep the plies in position during glueing, as shown in Fig. 5, but this construction adds nothing to the strength and is not necessarv. 5. Design of Spar. (Vertical Laminations) (a) The simplest design is made of two plies. (Figs. 76 and 77.) The spars may be spindled out into an I section in the usual way. (b) An alternative construction is to spindle the plies before glueing up and to reverse them and so make a hollow spar (Fig. 78). Tin.-: design has rather better torsional stiffness; special care must be taken to ensure per- fect glueing along the flanges. (c) Three plies may be used (Figs. 79 and 81). It is best to keep the middle ply thin, so that the spindling does not go quite through the outer plies but it is also quite safe to allow the spindling to cut clean through the outer plies. If this is done an ample fillet is to be left all round the spindled-out portion. (See " A," Fig. 79.) (d) Joints in the Plies. Any ply may be jointed by a plain 9:1 scarfed joint. Outer plies must not be spindled out where they are jointed. No joint must be made where ,the bending moment is more than half its maxi- mum value. Joints in the middle ply of a 3-ply spar must also not be made close to FIJ. 76. FIG. 77. FIG. 78. FIG. 79. FIG. 80. FIG. 81. FIG. 82. 11. Varnish. All glued joints are to be thoroughly protected from moisture by waterproof varnish. T.D.I. 502(1, LAMINATED SPARS. 1. Properly made laminated spars are as strong as solid spars. To obtain the full strength the following instructions should be followed. 2. Quality of Timber. The wood for making the spars must be of Grade A as specified in Air Board Specification V.I. 3. Warping. To avoid warping or twisting after they are made, the following precautions must be taken: (a) The plies of which each spar is to be built must all be equally dry. To ensure this they should all be chosen from the same batch of timber and should all have been kept together for the same time in the workshop after cutting to size for glueing. (6) The plies for each spar should have about the same number of annual rings per inch. (c) The plies for each spar should be cut at about the same angle with the annual rings. (See Figs. 1, 2 and 3.) 4. Jointing and Glueing. Special care is to be taken to secure a sound glued joint between the plies. Hot " hide " glue or cold casein glue may be used. the points of maximum shear. Canvas lashings should be put on the spar over all joints in the outer laminations. 6. Design of Spar. (Horizontal Laminations) (a) A satisfactory spar may be built up of numerous thiu horizontal laminae* (see Fig. 82). If built in this way the thickness of the laminations should be arranged so that the joint between the outer laminations divides the depth of the flange into two equal parts; these laminee may be butt-jointed at any point where the bending moment is less than half the maximum bending moment in the spar. The web laminae may be butt-jointed at any point where the shear is less than three-quarters of the maximum in the spar. (b) The ends of the laminations at the butt-joints should be planed accurately square with the faces of the laminations and be sized with weak glue before glueing up. (c) Butt straps of thin wood should be glued over the joints in the flange laminations. 7. Splicing. Laminated spars may be spliced as if they were made from the solid, by a simple 9 :1 scarfed joint (See T.D.I. 517), but it is generally better to joint the separate laminations as described" in Clauses f. (d) and 6. " There should be sufficient laminae to make at least one joint in each flange and at least four joints in the web (see Fig. 7). iio 11. STRENGTH OK SPAUS. The results of a very large number of tests on full sized spars show a close agreement between the test results and calculated figures for the failing load. (Sec Beam Tests, page 17.) The spars are always tested under combined bending and end compression loads similar to those commonly met with in an aerofoil. They fail fry crushing under the compressive stress m the outer fibres on the compression flange. The first sign of failure is the appearance of a fine compression line across the flange at the point of maximum stress. The calculated stress at which this line appears is about 10% (for the usual I and box section) above the ultimate compressive strength of the wood. The reason that the spar carries a slightly higher load than the calculated load is that the elastic limit of the timber is exceeded at about 80% of the full load, and after that the distri- bution of stress is no longer in linear relation to the distance from the neutral axis (cf. page 101, and Fig. 15). Spars rarely fail by shearing stress; it is hardly practicable to make the webs thin enough to do so. 12. STRENGTH OF STRUTS. A very large number of wooden struts of many different sections have been tested; the test results for the strength and elastic behaviour agree very closely with the calculated results. This applies to solid timber struts, laminated struts, tubular struts and hollow stream-line section struts with or without internal bracing. The strength is calculated by the modified Perry formula which takes into account the initial cur- vature of the strut and the eccentricity of the load. (See page 48 and Appendix I, Chapter IV.) LOAD Let c be the initial curvature (Fig. 83). e ,. eccentricity of the load. Then the curvature which is equivalent to these is e = c a + . The formula is p = |Q - Q = p v + (I + ^\ p K where and .p = ultimate load per square inch, p v = ultimate compression strength of timber, ' ~p p E = Eulerian value of p ; K = radius of gyration of cross section of strut (about major axis), c = equivalent curvature = c,, + f,e, as above ex- plained, d = thickness of the strut (along minor axis), E = Young's modulus of the timber. If c,, and e are not known, the value of ^^ may be taken as O'OOl ~. for an average value, or 0'003 == for K J^- tue minimum strength ; these two approximations are found to be in good agreement with the results of all carefully conducted strut tests. The result may, however, be actually above the Euler value. This will occur if the axis of the applied load deviates to the same side as the curvature of the strut, as shown in Fig. 84. This combination of curvature LOAD Fia. 84. and non-axial loading reduces the effective length of the strut as shown in the figure, and the test load may be higher than for a straight axially-loaded strut of the full length. When testing struts the load should be applied through hard steel balls so as to leave the strut free to bend in any direction. The equivalent length of the strut is the distance between the centres of the balls. Fig. 85 shows the Euler curve and the two curves for ~ equal to -001^ and -003 ' for spruce of B quality. Also a series of test results on a sample which happened to be almost exactly of this quality. i.e., having E = T2 million and p y = 4,000 Ibs. per square inch. The struts were square in section. Fig. 86 shows the results of a series of tests on four different types of round struts, viz.. Solid, Tubular McGruer struts, Tubular Kyan struts, and Tubular (with internal cross) Lawrence struts, shown in Fig. 87. The samples of spruce of which they were made were not identical, but the results have been reduced in the manner described below to correspond with a standard timber of Grade A (E = 1'6 million p y ,= 5,000 Ibs. per square inch) and plotted with the three reference curves for these constants. The figure shows how closely the tests of all the different sections fall on the theoretical curves; this hpwever does not mean that they are all of equal merit, as the weights were very different. CHAPTER. X. .0 1 8 O LA LU i u 2 U O a: 0, < in i Q. 2 O U Q Z I to <3 *9 "" h- x S to H UJ T 2 h- Q it Q^ h- UJ U iu fc5 ID :> H a: Q ^ h- CA UJ cc O N 8 D or u o o o u o D a: a. (0 cd LJ 70 } V Jv ' sufficiently accurate to multiply the observed strengtl E by the reduction factor -. EO where E A is the value of E for Grade A viz., 1-6 million; and Eq is the value of E for the timber the strut is made of. For short struts (rr< 70) the reduction is rather more troublesome. The test result will fall somewhere between the Euler curve and the lower curve / i \ I *^ K ) ^ rawn ^ or * ne constants of the wood of which the strut is made. The corresponding point on the standard drawing (for Grade A timber) should be placed so as to divide the ordinate between the Euler and lower curve in the same ratio. The deflection of a timber strut may be calculated by the formula given for steel struts (Appendix I, Chapter IV). Many types of strut have been designed to be built up of veneers or plywood. Such designs can only be suc- cessful if the grain of all the veneers or layers of wood runs parallel to the length of the strut. The values of E and p,, for the veneers with inclined grain rapidly fall with the inclination of the grain (see Section 4), so ihat the load which the strut made with inclined veneers will carry is greatly reduced. This fact has been repeatedly confirmed by tests on such built-up struts. 13. A METHOD OF STIFFENING TIMBER TO RESIST BENDING ROUND A LONGITUDINAT, Axis. (C.I.M. 742.) The extent to which timber differs in strength in different directions of the grain has already been dis- cussed. For example, the strengths and elastic moduli of spruce in the three principal directions of the grain given in Admiralty C.I.M. No. 2, are as follows: Axis. Ult. Strength, Tension. Modulus. O.L. along the tree. 18,000 Ibs./sq. in. 1.800,000 Ibe./sq. in. O.R. radial. 500 to 1000 Ihs./sq. in. 50,000 Ibs./sq. in O.T. tangential. 400 to 800 Ibc./cq. in. 50,000 Ibs./gq. in. The differences between the moduli are even more striking than the differences between the ultimate ptrf'net.hs, The stiffness of timber when bent round axes radially, tangentially, or on the quarter, will only be about 1/30 as stiff when bent round an axis along the plank as when bent round an axis across the plank. As an example the websi of a box spar are very weak to resist bulging inwards or outwards. Tests have been made at the R.A.E. to ascertain the extent of the stiffening produced by means of thin layers of veneer on the sides of a plank. The results are very striking. The addition of a veneer 1/100 inch thick on each side of a plank 1 inch thick increases its 'stiffness (to resist bending round axis O.L.) about '4$ times, and increases its modulus of rupture to about twice its former value. Veneers 1/50 inch thick similarly increase the stiffness about six times, and the modulus of rupture 2-6 times. The experimental values for the stiffnesses agree fairlv well with values calculated on the assumption of plane sections remaining plane during the bending (i.e., the usual assumption in rein- forced concrete problems). Details of the tests are given in Fig. 88. The attention of designers is drawn to this method of stiffening timber in directions in which it is normally very weak. It may be applied with advantage in many parts of aeroplane structure. The following are typical examples : 1. Webs of Box Spars and Rib. If not reinforced the webs of box spars and ribs need to be proportioned from consideration of stiffness, so as to prevent undue bulging and consequent splitting. If reinforced in the manner described above, the total thickness of web could be reduced considerably, for the inner ply could be proportioned to take the shear and the outer plies would provide the stiffness necessary to prevent bulgin? laterally. Such a web is virtually a special plywood with the plies proportioned to the particular work re- quired of them, instead of being all the same thickness. As only a small proportion of the web has the grain at right angles to the direction of the stress, the web takes nearly as much of the bending stresses as if it were solid. ^ A variation of this method of stiffening is used in the German Pfalz Scout. In this case very thin three-ply is used on each side of the centre ply instead of a single veneer. Probably this three-ply is used for reasons connected with production. 2. Strengthening Timbers which are Liable to Split. Rome timbers are very liable to split, particularly if 112 the section does not closely approximate to a radial section. This can be largely prevented by the use of thin veneers with the grain at right angles to the grain of the piece it is desired to strengthen. An example of this application is to be found in the box spars of the German Albatross machine. In this machine the flanges are rather thin and wide, and are made of a pine which splits readily. To avoid this, thin three- ply of birch is glued to both sides of the flange. As with the case of the Pfalz Scout, the three-ply has been adopted probably for manufacturing reasons. The same effect would be produced by a single veneer on each side. 3. Plywood for Ribs. Where plywood is used for the webs of ribs a stiffer web is produced if made on the lines indicated above, whether the web is designed to take bending or only shear. 4. Covering Outer Portion of Propeller. Propellers frequently fail owing to lack of strength in bending round an axis through the blade. The outer portions where the section is thin are the weakest for this type of failure. If these portions were covered with a veneer having the grain at right angles to that of the blade, a considerable increase in stiffness would be obtained. This method has frequently been used by the Germans on their large propellers. 14. THE USB OP HEAVY TIMBER IN SPARS AND STRUTS. Owing to the lack of sufficient light timber of good quality during the war the possibility of using heavier timber was considered. The conclusion arrived at was that heavier timbers might be used without increasing the weight in the flanges of box spars, particularly in the larger sizes, and that the most promising timbers were those imported under the name of Pitch Pine, but that heavier timbers could not be used for struts. In order to estimate the effect of using heavier timber it is necessary to distinguish between the uses to which the timber will be applied. The principal uses are: (a) For very short compression members. (b) For spars. (c) For long compression members, e.g., interplane struts. (a) Very Short Compression Members. In these parts there is no bending, and the strength is practic- ally determined by the compression strength. For equality of strength, then, the compression strength should increase in direct proportion to the density. There are very few of such parts in practice. (b) Spars. The criterion of failure in a spar is the attaining at the extreme fibre of a stress equal to the compression stress of the material. If the spar is fairly deep and the, flanges are relatively thin, the condition of equality of strength is the same as for short com- pression members, i.e., compression strength in direct oroportion to density. When the section is stoekv (as in the Avro rear spar) the comparison is not so simple ns a large part of the cross section is not fully stressed. Each design of spar requires separate consideration. In Fig. 89 the results of the U.S.A. compression tests on timbers containing 15% moisture, given in Appendix VI, are plotted on a density base. The lines A and B are drawn through the origin, and the points corresponding to Grade A and Grade B spruce respectively. In Fig. 90 similar data are given for timbers containing 12% moisture. The figures show that almost any timber except the Oaks are just about equal to Grade B Spruce. Many of these timbers, however, will not be suitable for spar construction owing to the difficulty of getting supplies of reasonably long and straight- grained wood. Probably the best timbers to use will be some of the heavier conifers. From the results of a large number of tests by Janka on Pines, Larches and Firs, the following equation connecting f e , the com- pression strength and p, the density in Ibs. per cubic foot has been obtained: / = 234,0 - 850 (1). Bauschinger from a number of tests on similar German woods deduced the equation: f c = 221 P - 900 (2). These equations are practically identical and show that the strength increases more rapidly than the density. Equation (1) is drawn in both figures. It will be seen that the line agrees with the results of the Long-leaf (Pinus palustris, Mill.), Short-leaf (Pinus echinata, Mill.), Loblolly (Pinus taeda, Linn), and Cuban Pine (Pinus heterophylla, Sudworth.). It would, therefore, appear that these Pines, which are sent to this country as "Pitch Pine," may be suitable timbers for making spars. If this timber were used, the flanges could be reduced to give spars of the same weight as those made of lighter wood. The difficulty would probably be mainly in the shakes to which it :s liable, but these may be minimised if the pieces are sawn as nearly radial as possible, since the shakes tend to be radial. Difficulties may be experienced in glueing these heavy timbers which are resinous. The Western Hemlock and Larch which give good values probably come to this country among the Silver Spruce. The results of a large number of tests by Julius on the West Australian timbers are also given in Fig. 90 ; also a few tests on some Victorian timber by Mr. G. Hankins, of the N.P.L. (Vide Ad. Committee for Aero- nautics Eeports and Memoranda No. 395.) Of those tested by Julius, almost all are well below grade B. figures. Of those tested by Hankins, several are above the grade A. line, and the Victorian Blackwood in parti- cular gives wonderfully high results. If this material can be obtained it would certainly be worth further investigation. The other Australian timbers are probably far too heavy for small spars even where they are above the Grade A line, as it would be impracticable to make the flanges thin enough. It is also probable that many of these timbers, owing to the peculiar way in which the annual layers are formed (with the grain of each at a different inclination to the vertical) may not be suitable for the small scantlings required for aeroplane work. _^ (c) Long Struts In order to compare different timber when used as long struts, the following approxi- mate method of calculation will give results which are sufficiently accurate. Interplane struts are practically Euler struts; the failing load is therefore, p = ! r 2 EI The problem is therefore to ascertain the variation of E. with density in order that struts of given length and CHAPTER. X. FiG.90 4000 40 .50 60 DENSITY L.B3. PER. CUBIC FOOT. 00 Fio.9!. 59tG5/94>V 200O. C I R H<* S79 12/20 CHAPTER X. ill a: LJ 03 H u. o g UJ cc u a. a. O H Ld 1- o o UJ ft: ZD in o 2 I h UJ k! u o: i o ,,a/sQ-| OVCn LJ U ^ a: a. CO a z UJ or te CO LJ or Q_ O o z o H z LU h- o O LJ a: o 23(65 /!XW ^000 C S R. tf." 273 II/2O Chjpber.X. Fia. SPRUCE VARIATION OF ULT STRENGTH IN COMPRESSION WITH MOISTURE^ V "K s N f 0,0 00 \ 9,000 X N v 6.OOO \ V 3.0OO 4 .s V i i 2.000 PER CCNT OF ASH. Fie. VARIATION Or UL.T. STRENGTH IN COMPRESSION. WITH MOISTURE. 9,904 \ tC^M ,o - ^ ^ < I^Btt X *L It AM ^ k Vs s^ S e.oo* 4> 6* W MOISTURC PER CCffT W O*tY WCIOMT. FIG. 95. WALNUT VARIATION OF UL.T STRENGTH IN COMPRESSION WITH MOISTURE. f..-J s: N MOISTURE PER CENT or DRY WEIGHT. MAHOGANY Fie. 9 VARIATION Or ULT. STRENGTH IN COMPRESSION, WITH MOISTURE:. 15, oCK^"" ( II 000 1)000 9.000 t>,aao \ \ 1 >V A ^ . (.. OOO b.OCIO i nAA 1 s v ^> CKapber.X. Fie. 99. WALNUT SPRUCE VARIATION Or"E V> WITH MOISTURE M 2-4 VARIATION OF E WITH MOISTURE 3 < t O I 1 "1 2 ?-C 18 '6 I 4 2 O 1 ft_ " t s^< tt 8 X V - 2 > S-' "V, ( s < i is \ V ^^ h i * << ^ 'v I" X Ik,. - *V V_ a * 4 MOISTURE F*CR C-ltl^T* or" ncV M/K-I^.U-V 10 a r 1 * X ASH. F.e.98. VARIATION OF"E" WITH UfNnT|FFTF *>*% rvtOiaTURK PBR CENT Or DRY WWOHT. <: f( MAHOGANY VARIATION OF "^"TH MOIST FiG.IOO. V i s URE1 \ < o X ? c 1 Z y tP v^ x z a < I 1 B-0 18 16 V s X 6 ^1 ?!*, , ^ ; - ~ U r-4 ^ s i e w 4 *74 > 6> 8> I0> |2Ji I4X I6^( |^x J(J J2jt ZVf te .( Styyp. MOI6TUPIC pen CENT OF DRY WBVGMT 29165/344 2000 C J R. L'. 7S. n/SO. CHAPTER X. (0 o >0 O oo o O I 1 I I I O F> LL I CO O cn 113 strength may have the same weight. The cross section will be assumed to be the same shape for all struts. Let I be the moment of inertia of the cross section. A , area of the cross section. I , length. W , weight. P ., , failing load. t , thickness of the strut. p , density of the wood. Then I = U* where 6 is a constant, A = at 2 a W = P _ E where ;j is a constant. In order that the strengths of two struts of the same shape and weight, but of different material, may be equal the Es must vary as the squares of the densities, i.e., P 1 -r 2 The value of E. for various timbers is plotted in Fig. 91, and it will be seen that E. is approximately propor- tional to p. The two curves on the drawing show the way in which the E. would have to vary in order that the timber might be equal to grade A. or grade B. Spruce. Very few of the timbers are even equal to grade B. ; white pine appears to be very good, but this is not in agree- ment with tests made in this country on samples of this timber. The heavy timbers, therefore, will not be suit- able for long struts. In this connection is may be noted that the com- parison used by Hnnkins in the Advisory Committee Report (/of. eft.) is not a legitimate one. Using the load ratio W(jjo rj- t as the figure of merit he shows that Vic- toria Blackwood is as good as (Trade B Silver Spruce. It is evident, however, that this is not a legitimate criterion since the aim is not to put in a stronger strut but one of the same, strength as Spruce and of no greater weight. Comparing Blackwood and Spruce by the method recommended above it will be found that a Blackwood strut of the same weight as the Spruce strut would only sustain about half the load. Again it is hardly fair to compare a picked sample of Australian wood with the lowest grade Spruce. Com- paring a good sample of Spruce with Blackwood, with the same moisture contents the following figures are obtained which aro not nearly so striking: Compression. Strength. Density. Compression. Density. Moisture=9% Victoria Blackwoo'l, (light) 93i HI N 26fi 8800 32-6 270 Spruce 7i:.o 27 265 Moisture 9-9% Victoria Blackwood. (dark) 13,700 44-7 306 12,000 42-6 282 Spruce 7000 2fi 259 Moisture 12% Woolybutt 11.700 57-3 204 Ironbark i 14.9110 64-6 230 Blackbutt 13,200 r,9-8 221 Spruce ... 6400 27 237 15. THE EFFECTS OF MOISTURE ON TIMBER. Any change in the moisture content of a wooden part has important effects, since it alters both its strength and its dimensions. Though the alteration in dimen- sions is comparatively small it may loosen the metal fittings and cause general slackening and distortion of the structure. If the change of moisture content takes place at all rapidly serious warping may occur, and if the change takes place gradually distortion of built-up parts may take place owing to the different rates of shrinkage of different timbers and of the same timber in different directions of the grain. All wooden parts of aeroplanes are protected by varnish, but experience confirmed by accurate tests shows that no varnish is perfectly moisture proof. Professor Lang and Mr. Robinson, at the request of the Materials Section, have made a very thorough in- vestigation into the relations between the moisture con- tent of a number of different timbers and the humidity of the surrounding atmosphere; they also devised methods of testing the rate of the passage of moisture through varnishes, and measured the rates for many different varnishes. Their reports are reproduced below. A note on the moisture contents of some timbers in Egypt by Major Atkin is also given. Variation of Compression Strength with Moisture Content. The results of a series of tests on a sampl.j of Spruce are given in Fig. 92. The relation of ultimate strength in compression to the moisture content is practically linear over the usual range of working. In the same figure is given the curve obtained by using the correction factors given by Tiemann (A. B. Tiemann. Bulletin 70, U.S.A. Forest Service) for Red Spruce, and in Fig. 93 are given the curves for Western Hemlock and Western Larch. For all these timbers the average rate of increase of strength with decrease of moisture at about 15% moisture is practically the same, i.e.. 230 Ibs. per square inch for 1% moisture. The Ameri- can tests give curves of slightly different shape to the English ones. This is possibly due to the different size of test pieces employed, i.e., 2 ins. by 2 ins. as compared with jj in. diameter; with the large test piece it is difficult to ensure uniform moisture condition. There would appear to be considerable variation among dif- ferent samples of timber. Thus another sample of Spruce gave the tests plotted on Fig. 93A. In this sample the effect of a change of moisture is about double that of the sample given in Fig. 92. Similar sets of results for Ash. Walnut, and Mahogany are given in Figs. 94, 95, and 96. _ It is noticeable that Mahogany shows the least variation. Variation of E with Moisture In Figs. 97, 98, 99 and 100 are given the results of a series of tests on Spruce, Ash, Walnut and Mahogany. The tests are not entirely satisfactory, and some points require ampli- fication. It may be noted that another sample of Spruce ' showed considerably loss variation than the one given here. They serve, however, to show that the propor- tional change in E for a given moisture change is not so great a* the proportional change in the compression strength. As in the compression test Mahogany shows decidedly less variation with moisture than the other timber* In Fig. 93 are given typical sets of result* for Western Hemlock (U.S.A. Forest Rorvice, Bulletin 272H 114 115) and Western Larch (U.S.A. Forest Service, Bul- letin 122). In these timbers the variation of E is smaller than in those just described. Variation ol Elastic Limit and Modulus of Rupture with Moisture Content. The variation of these quanti- ties are of the same order as those of the compression strength. The American curves for Western Hemlock and Western Larch are given in Fig. 93. Variation of Tensile Strength with Moisture Content. From the few tests available it would appear that the tensile strength is much less affected than any of the other quantities. It is probable that when the moisture content is decreased to very small amounts some change occurs and the tension value is seriously diminished. Certainly over-dried timber beams frequently fail en- tirely by tension, and not as is usual in normal timber by compression. Such timber does not completely re- cover when it regains moisture. Variation of Shear Strength with Moisture Content. As generally determined the value of the shear strength also increases with decrease of moisture approximately at the same rate as the compression strength. For the fullest account of the effect of moisture on strength reference should be made to A. B. Tiemann's paper, Bulletin 70 of U.S.A. Forest Service. Expansion of Timber with Moisture. In Figs. 101, 102, 103 and 104 are given the results of a series of observations on cubes of Spruce, Ash, Walnut and Mahogany placed in a saturated atmosphere and weighed and measured at intervals of time. The varia- tion in the longitudinal direction is very slight in most cases, but that in the radial direction is very appreciable, and in the tangential direction still more so. There is generally no change in the dimensions with increasing moisture content after the fibre saturation point has been passed. Between 12% and 28% moisture the sample of Spruce expanded 2'7% in the radial direction and 4-6% in the tangential direction. The curves plotted on a time base, Figs. 105, 106, 107 and 108 show how very slowly absorption takes place even in a sample so small as a 1 inch cube. It has already been noted that there is little plastic flow in timber when under tension stresses. If, there- fore, timber is placed so that the natural change of size under different hygrometric conditions is not allowed to take place, stresses will be set up which may easily split it. Assuming E to be 60,000 and the ulti- mate tensile stress in the tangential direction to be 600 Ibs. per square inch, the strain at rupture of a perfectly elastic material would be 1%. If, therefore, the sides of a tangential plank of Spruce were rigidly held and the moisture content diminished by 3'4%, the specimen would be close to the point of rupture. This is virtually what occurs when a log dries, the outer layers dry quicker than those inside and are prevented from contracting, with the result that radial shakes are produced. The same phenomenon is observable in planks that are not uniformly dried, either owing to irregular ventilation or exposure to sunlight. This variation of size with moisture content is very troublesome in the shops, where it constantly produces shakes and warping. In timber in which there is little curvature in the annual rings a rectangular plank cut with its sides parallel to the radius and the tangent respectively, will retain its rectangular shape as it dries. If, however, there is a decided curvature in the rings Flfi. 10!). then the shape is not retained; for example, AB (Fig. 109) being nearly tangential will contract more than C D which is partly radial, so that when dry A B will be concave upwards. There will be less distortion in a radial plank than in any other. With straight-grained timber which is fairly uniform across the section, the main change will be in the shape of the cross-section, but with cross-grained timber, and even more with spiral-grained timber, the diagonals of the different faces of the plank will in general be at different angles to the grain and will contract differently, so that the plank will bend and twist. These effects of varying moisture content are of con- siderable importance in all built-up structures. Not only have the distortions of the individual parts to be considered, but also the effects of the several parts on each other. The following are some of the more im- portant points to which attention should be directed : 1. All pieces entering into a built-up structure should have the same moisture content. 2. As expansions vary somewhat with density, etc., all the pieces should be of approximately the same density, number of rings per inch, etc. 3. Only straight-grained timber should be used and in particular spiral grain should be excluded; to secure good results it may be necessary to " rive " the pieces. 4. Tn built-up parts there will be a minimum of distortion if the annual rings of the pieces on opposite sides of a joint have the same inclination to the joint but slope in opposite directions. Tf the two portions to be jointed are from the same piece of timber sawn down the middle (as is usual in making aileron spars that have to take torsion), then it is probably 115 slightly better to reverse tho grain and put the sawn surfaces on the outside. What must never be done is to glue a radial plank to a tangential one. 5. With timbers which are specially liable to split as, for example, Oregon Pine all the planks should be cut as nearly as possibly radially. THE EFFECT OF THE RELATIVE HUMIDITY* AND THE TEMFEKATUKK OF THE ATMOSPHEKE ON THE MOISTUKI: CONTENT OF WOOD. BY W. ROBINSON. In an atmosphere containing a given percentage of moisture, wood behaves as a hygroscopic substance and attains 'i state of equilibrium at a definite moisture-con- tent. The moisture-content at this equilibrium may differ for different woods, but is approximately the same for many woods. Table I shows the moisture-content attained by small pieces of a number of different woods under three different sets of conditions, via,, those (a) of an ordinary laboratory; (b) of the open air, but under cover; and (c) of a moist chamber with a saturated at- mosphere. In all of these the average temperature was roughly the same, but the relative humidity of the atmosphere differed in each case. The relative humidity were re- spectively 55%, 82%, and 100%, and the results given in the table make it clear that as the relative humidity increases, the moisture-content of the wood at equilibrium also increases. In this property wood behaves like most other hygroscopic substances, and the property can be regarded as a special case of adsorption. This assumption is justifiable since, in coming to equilibrium with an atmosphere of a given moisture-content, wood obeys the laws which govern other cases of adsorption. Adsorption is the process whereby a substance in contact with the surface of another substance is con- centrated at that surface if by so doing the free energy at the surface is diminished. The process is character- istic of all surfaces, but is most striking in substances possessing surfaces which are large in proportion to the mass of the substance. Charcoal for example is such a substance, and the effect of charcoal in clarifying liquids is an adsorption effect, as is also the occlusion of gases by charcoal. Wood, like charcoal, is a substance pos- sessing a large internal surface and on this account water vapour is adsorbed by it. The degree of adsorption of substances from a solution in a liquid depends on the concentration of the solution and on the temperature. The relation between adsorp- tion and concentration follows a simple parabolic law * The relative humidity may be defined as the ratio of the actual vapour pressure at a given temperature to the maximum possible vapour pressure at the same temperature. 27264 as expressed in the ordinary adsorption formula x 1 _. = aC, where x is the amount absorbed by a surface m n in from a solution whose final concentration is C, a and being constants for the particular surface and solution.* The relation to temperature is different, for while the rate of adsorption is increased by rise in tem- perature the amount adsorbed at equilibrium is dimin- ished. This also holds for the adsorption of gases, of which that of water vapour from air is here regarded as a special case. Baylisst has explained the decrease in adsorption with rise in temperature by the fact that surface energy (on which adsorption depends) is itself anomalous in having a negative temperature coefficient. Water, for example, which, at 17 C. has a surface tension of 73'8 dynes per sq. cm. at 60 C. has only a surface tension of 65 dynes per sq. cm. From this it appears that, as the temperature rises, the surface energy between the particles of a hygroscopic substance and water will diminish and the particles will therefore be able to hold less adsorbed water than before, even though the actual amount of water in the atmosphere is greater. The work of previous investigators on various hygro scopic substances has shown that the moisture-content at equilibrium depends on the relative humidity and to a lesser extent on the temperature. It appears from Eegnault 's studies on hygrometry, that the weight of water absorbed by sulphuric acid at temperatures ranging from 5 C. to 35 C. depends on the relative humidity (i.e., on the ratio of the actual vapour pres- sure to the maximum possible) even though the actual amount of moisture present in the atmosphere for the ratio is very different at different temperatures. For the range of temperatures used by Eegnault, mis., 5 0. to 35 C. the effect of temperature on the behaviour of sulphuric acid was so slight that little or no differ- ence can be detected in the curves obtained by plotting moisture-content against relative humidity at 5 C. and 35 C. respectively. Since sulphuric acid, itself, has practically no vapour pressure, Eegnault made use of different concentrations of it to obtain atmospheres of desired degrees of humidity in order to study the hygroscopic properties of hair used for hygrometers. Van Bemmelen more recently has made use of the same method in investigating the properties of silicic acid gels, and SchloesingH has employed the same method in the study of textile fibres. Other methods of experiment and expression of results have been used by Masson and Eichardsf, Trouton**. and also by Traverstt in studying the hygroscopic properties of substances like cotton, flannel "and glass wool. The method introduced by Eegnault has been adopted in the present research on wood, and the results for this See Freundlich, " Kapillarchemie " (1909), pp. 170-186. t Baylies. "Principles of General Physiology" (1917), p. 61. I Regnault, "Etudes sur 1'hygrometrie," Ann. de Chem et de Phys. (1845), 3 Her., Tome 5, pp. 129-236. Bemmelen Van, "Die absorption" (1910), Dresden. || Schloesing, "Comptes Rendus" (1893), Vol. 116, pp. 808-12. ^f Masson and Richards, "Proc. Roy. Soc.," 78 (1907), pp. 412-429 * Trouton, "Proc. Roy. Soc.," A. 77 (1906), pp. 292-314. ft Travers, " Proc. Roy. Soc.," A. 78 (1906-7), pp. 9-22 ; A. 79 (1907), p. 204. P 2 116 substance are expressed below in the form used by liegnault, and after him by Schloesing and Van Bemmelen. The exact procedure carried out for specimens of wood may be briefly described. The chambers used in the experiments were wide-mouthed bottles of 600 cc. capacity, with well-fitting ground glass stoppers. About 100 cc. of sulphuric acid of approximately known strength was placed at the bottom of each bottle and the specimens of -wood were suspended in the air above the surface of the acid. The pieces of wood were all cut in the same direction, across the grain, and measured about " x j" x ". Specimens of silver spruce, English ash, mahogany, swamp cypress and of a dark and light sample of padouk were tested. In addition a small roll of cotton wool was suspended in each chamber in order to ascertain whether the method was giving results similar to those previously obtained by other investigators for cotton. Ten cham- bers were prepared using for eacli a different strength of sulphuric acid ranging from H 2 S0 4 + 1H 3 O to H 2 S0 4 + 17H 2 0, and calculated to give relative humid- ities ranging from 8% to 83%. The bottles were placed in a cupboard in a position where temperature variations were known to be slight; with them was placed a recording thermometer which registered the temperature continuously. During the experiment the temperature gradually rose from 16 C. to 18 C. and remained at 18 C. for the last four days of the experiment. The pieces were allowed 21 days to attain equilibrium. At the end of this period the moisture-content of each specimen was determined, and expressed as a percent- age of the dry weight. The exact composition of the sulphuric acid in each chamber was obtained by titra- tion against standard solutions of sodium carbonate, and the strength of the sulphuric acid was expressed as HjSO,! + xH 2 0, the value of x being in each case calculated from the results of the titration. The curve representing the relation, at the tempera ture of the experiment, between the strengths of the solution of sulphuric acid and the relative humidity of the air in equilibrium with these solutions was drawn. The figures forming the basis of this curve are given in Eegnault's table for temperatures ranging from 5 C to 35 C. The strength of the sulphuric acid in the different bottles being known, the relative humidity of the atmosphere in equilibrium with the given strength of acid can be read off from this Regnault curve. These data having been obtained, the curve showing the relation between the moisture-content and the relative humidity was plotted for each of the different woods and also for cotton (Fig. 110). Owing to the fact that Eegnault's table only gives figures extending to a humidity of about 85%, the highest point actually measured for the curves in Fig. 110 was at about relative humidity 84%. Some of the earlier tests, however, on the moisture-content attained by woods in a saturated atmosphere showed that the percentages were: spruce about 24%, mahogany about 20%, ash about 25'5%, and Andaman padouk about 15%, These figures, broadly speaking, give points which would fall on the respective curves in Fig 110. This question of the moisture-content attained by woods in a completely saturated atmosphere has, however, not beon fully investigated. liegnuult's. table of the comparative tensions of aqueous vapour given by pure water and by different proportions of sulphuric acid and water at different tem- peratures shows that the effect of temperature is com- paratively slight. The results of Schloesing* for textile fibres for temperatures ranging from 12 0. to 35 G. agree with this. In order to ascertain to what degree temperature affects the moisture-humidity curve for wood, an experi- ment was carried out on specimens of spruce. The small pieces of wood were suspended in a series of sU chambers similar to those described and kept at 55 G. for 14 days while other similar series were kept at 30 C. and at 16 C. The curves showing the results of these experiments are plotted in Fig. 111. From these it will be seen that the difference produced by the range of temperatures used is relatively slight at the lower tem- peratures, but at 55 C. the spruce wood attains equi- librium with a distinctly smaller moisture-content at each degree of relative humidity than at the lower temperatures. Discussion of Results. The chief factor determining the moisture-content of a given piece of wood is the relative humidity of the atmosphere with which it is in equilibrium. The curves in Fig. 110 show that the relation between the moisture-content at equilibrium, and the relative humidity of the atmosphere is very similar for most of the woods tested. These results are in agree- ment with those given in Table I. The curve for cotton in Fig. 110 is very similar to that given by Schloesing for this substance. The results obtained in the present tests on wood may therefore be regarded as comparable with those obtained by other investigators on hygroscopic substances. It is obvious from Fig. 110 that most woods are more hygroscopic than cotton. The specimen of mahogany differed partially from the other woods in that it attained a slightly less moisture-content at the higher degree* of humidity. Of the present series, however, Andaman padouk is the only wood showing marked deviations. The curve (Fig. 110) indicates that it is much less hygroscopic than the other woods. No explanation for this striking difference of padouk from the other woods can at present be offered, but it is perhaps of interest in the connection that a much less dense and lighter-coloured sample of padouk gave results intermediate between those for other woods and for the dark-coloured sample. Within the range of temperatures used in these experi- ments the effeet of the temperature on the moisture- content at spruce wood at equilibrium is relatively slight at lower temperatures, but is more marked ut higher temperatures. In the curves (Fig. Ill) for spruce at 16 G., 30 C. and 55 G. respectively the general trend of the curves is similar, but with rise in temperature the curve moves distinctly to the left, less moisture being held by the wood at the higher tempera- tures even though the actual amount of moisture present in the air is greater. This temperature effect is exactly similar to that obtained in cases of adsorption. The assumption made earlier in this report that the relation of wood to water vapour is a special case, of adsorption is thus justified from the results for spruce at different temperatures. * Schloesing, "Comptes llcmlus" (1SU3), Vol. 116, pp. 808-1?, O 00 oo - > M '* 45 n * +J O 4 Under cover in open for 4 days. Under cover in open for 31 days. In saturated atmosphere at lab. temp, for 8 days. Sat;u Walnut Pitch pine % % 17-6 16-8 17'2 o/c 17-1 16-0 17- 1 % 24 -9 23-4 24-6 Aspen ... 17-0 16-7 24-2 Aspen ... 16-9 16'1 24 '2 English poplar Birch 13-8 17-0 16'7 16-4 16-7 23-3 25-2 Birch 16-9 17-5 25-7 Bass (5 ply) White beech (Aus.) ... Ash (Eng.) White Wood (Can.) Oak (Eng.) Mahogany Mahogany (Africa) Mahogany (Lagos) ... Mahogany (Aus.) Mahogany Mahogany (Benin) 13-3 10-0 12-4 15-8 14-8 17-0 17'5 16-0 12-8 17-3 16-4 18-6 15-2 17-2 16-1 15-U 16-5 17-0 16-1 14-8 16-8 17-1 18-0 15-6 16-8 16-9 23-3 19-8 25-0 23 '3 H>B 17-7 23-8 21-2 25-2 18-9 20-4 24-0 Spruce ... ... .. ... 17-0 23-8 Boxwood Walnut 11-6 16-9 15-6 24-5 23-0 Walnut (Black) Sycamore Oregon pine ... Indian rosewood i Andaman padouk 9-8 14-8 17-0 16-9 13-0 23-0 25-4 23 -3 17 -5 14- 9 * The laboratory temperature range J from 1.1 to 16 C. at 9 a.m Note to Table 1. The specimens used in the tests in Table I were small pieees cut across the grain, measur- ing about 1 in. x 1 in. x J in., and the per cent, moisture is calculated on the dry weight of the wood. It is clear that the majority of the woods show a very close similarity in the amounts of moisture taken up, under tho respective atmospheric conditions to which they were exposed. Two of tho varieties of mahogany tested and the Indian rosewood, Andaman padouk, and to a less degree Australian white beech under all the different conditions, have taken up less moisture than the majority of tho woods. No explanation of these differ- ences can at present be offered. been over two years in the laboratory or in Egypt else- where. It may be added that the coastal zone, which includes this station, is much damper than the country inland, at Heliopolis, near Cairo, for instance. Far more trouble has been experienced there than here, owing to the loosening of internal cross-bracing wires due to shrinkage, aggravated in some cases by insufficient thrashing of the wires. In the following table are given the minimum moisture-content values found in Egypt for various timbers, together with the specified limits, if specifica- tions for the material are available here. The figures are percentages on the dry weight. Timber. Egyptian minimum. Specification limits. Specification. Mahogany 8-3 j 16-1 2 summer 14-10 winter | V. 7 Walnut s-ii 18-12 summer 16-10 winter } V.5 Ash, English ... 8-6 20-15 summer 18-14 winter ! v - * Spruce 8-6 17-14 2 VI. Oregon pine 8-3 17-14 2 VI. Three-ply 7-8 Under 15 V. 3 Five-ply 8-4 _ Swedish pine 7-7 _ Padonk 7-6 -.. Bombay black wood 7-6 _ Lein 9-3 Cypress 9-8 __ Kail (Bhotau pine) 11-8 i ~~" It is probable that neither the Leiri, Cypress nor Kail had been sufficiently long in the country to reach a condition of equilibrium. The above figures were obtained by drying about two grams of the shavings, taken from the centre of the block, for three hours at 60 C. in order to dry with as little oxidation as possible. A further period of one hour at 100 C. was usually sufficient to dry completely though a second hour was sometimes necessary. Further heating leads to an increase in weight due to oxidation it is believed. This method was checked by a comparison with the usual commercial method, namely, the heating till constant, or for a given period, of a one-inch slab cut across the plank. The results were found to agree- to within 01 per cent. when the heating of the slab was prolonged to about nine hours at 100 C., after tliree at 60 C. Provided due care is taken in selecting tho sample the use of the shavings seems preferable on account of its rapidity. MoiSTI/UK-CoNTENTS OK DIFFERENT TlMBKRH IN THE COSTAL ZONE OF EGYPT. }'>\ MA.IOR W. K. G. ATKINS. A number of moisture estimations, calculated on tlu' dry weight, have been made here, during the last year to see whether imported timber was sufficiently seasoned to use without risk of shrinkage. Determina- tions were also made upon certain samples which had 16. VARNISH. A METHOD OF TESTING AND COMPARING THK " WATER- PROOF " PROPERTIES OF VARNISH AND OTHER PROTECTIVE COATINGS OR TREATMENTS OF WOOD. By W. H. LANG. Statements are commonly made as to the effect ot various coating and treatments in rendering wood " waterproof." Little appears to have been done towards obtaining such a measure of this property a3 will enable the efficiency of different treatments to be estimated and compared. 118 The following method has been found to give useful information in studying the effect of coatings or treat- ments of wood used in aeroplane construction. It can also be used to test the " waterproof " properties of any coating that for the purpose of testing can be applied to wood. 1! octangular test specimens of spruce measuring about 1O5 x 5 x 4'5 cms. are carefully cut and smoothed. For purposes of comparison it is convenient to keep the same wood and size of test specimen, but this is not absolutely essential. A test specimen is weighed in the air-dry condition and then varnished or otherwise treated. It is then placed in a moist chamber, in which the air is practically saturated, along with a weighed, untreated, control test-specimen of the same wood. The absorption of moisture is ascertained by weighing the specimens at intervals. In the full tests the test-specimens were kept in the moist chamber for 14 days. They were then allowed to dry in the laboratory for 14 days. The control and treated specimens were then submerged in water for 14 days and finally allowed to dry in the laboratory for 14 days. The gain or loss of moisture can be expressed as a percentage of the original air-dry weight of the wood and the changes in" moisture-content during the experiment be plotted as a curve. Alternatively the rate of absorp- tion of moisture during any period can be expressed in grammes per sq. cm. per day. A record of the temperature during the course of the experiment is kept. In this way the rate of entry of water vapour and of liquid water into a test-specimen of wood which has been varnished or otherwise treated is compared with that into a similar specimen of untreated wood. In doing this, the general relation of wood to an atmos- phere saturated with vapour and to liquid water must be borne in mind. In a saturated atmosphere the hygroscopic substance of the wood can absorb about 29% of its dry weight of moisture. This moisture absorbed as water-vapour is not present as liquid water in the cavities of the cellular structure of the wood. In atmospheres of lower re- lative humidity less moisture is held by the wood. The wood in our laboratory, which is used for the test speci- mens, may be taken as having about 12% moisture. When wood is submerged it absorbs a much larger amount of water, which accumulates in the capillary cavities of the wood. This absorption of water is rapid at first, then becomes slower, but (in a piece of wood of the size of the test specimens) continues for a long period at a much more rapid rate that the absorption of water vapour. With regard to the effectiveness of any coating or treatment, it can be ascertained by. this method: (1) Whether it prevents the entrance of water vapour or not. (2) Whether is prevents the entrance of liquid water or not. (3) When it does not prevent the entrance of vapour or liquid water, how it affects the rate of entrance. It will bo evident that this draws a distinction be- tween the vapour-waterproof properties and the liquid waterproof properties of a coating or treatment. It will be shown in the case of varnish that this is a distinc- tion of practical importance. If a coating is liquid- waterproof but not vapour-waterproof the gain in mois- ture when the specimen is submerged is of the sairi'j order as when it is placed in a saturated atmosphere. The procedure described above, while useful in study- ing the effect of any particular coating, is inconveniently long for a practical test. It can be shortened by making parallel tests for water-vapour and liquid water. An opinion can be formed as to the waterproof proper- ties of a coating in one or two days, though a longer test is desirable to ascertain if the protection is maintained. The problem of obtaining the most rapid reliable test is under consideration. The expression of the rate of entry of moisture in grammes per sq. cm. per day is only satisfactory when a coating has been found to be liquid-waterproof. It cannot be applied to the control wood since the entry of liquid water is greater at the ends than the sides of the block. Similarly, when a coating shows local break- downs in water this way of expressing the result is misleading. The thickness of the coating must be taken into con- sideration. This can be directly measured by making sections at the end of the experiment. The rate of entry of water-vapour depends on the temperature, and this must be stated. While a general correspondence between the rate of absorption of mois- ture and temperature is evident, the details of this relation are complicated. This and some other factors require further investigation. SlIOIlTKNlil) FOHM OF THE MV.TI1O1) OF TESTING THE " WATEK-l'KOOF " i'KOPKHTIES OF A COATING OF VAKNISH ON WOOD. BY W. H. LANG. Further experience has shown that the test can be completed and reliable results obtained in a few days. Longer tests are only required for special purposes. The test specimens, prepared as described in the pre- ceding reports, are tested in the moist chamber and by submergence in water. The same specimen may be first tested in the moist chamber and then submerged, or two specimens may be tested simultaneously under these respective conditions. The temperature should be about 60 F., and should be stated. In each case the increase of weight due to absorption of moisture is ascertained at the end of 24 hours, and expressed as grammes, per sq. cm., per 24 hours. The result is checked with that for the first 48 or 72 hours. If the rates agree (taking any change in temperature into consideration) the test may be stopped. The method in this shortened from may be illustrated by the results for four linseed oil varnishes supplied by A.I.D. as given in the following table. Although the mode of expression of the results is not strictly applic- able to the control specimen (owing to the difference in capillary absorption by ends and sides) the corre- sponding figures for this are given to indicate the degree of protection conferred by the coatings of varnish. CHAPTER X. I NJ h J ^ \ o I ', 9*S*UP 6s * t O v) S 5 uj r < >n nc 5 o ec <-> 95 9C 80 75 CURVE 2JH LINSEED on PROPELLER VAAHISH ON WALNUT TEST SCECIMf.NS OF ISO SQ. CM. 5:UCTt FlUlRtl.Z, .*, e. 16, COATS OF VARNISH. temp" is'f 53' Si' CHAPTER X. FIG. 117 Cim XVII ZOOO. CR ! 8. I?/IS. 119 TESTS ON PERMEABILITY TO MOISTURE OF FOUR LINSEED OIL VARNISHES (Fig. 112). (Results expressed in grammes per sq. cm. per 24 hours.] Temperature 63 F. Thickness of Coatings of Varnish, 70 80/i. Moist Chamber Days. Submerged Days. Description of Specimens. 1 2 3 1 2 3 Spruce + Silicate Filler + 4 Coats Varnish, No. 1. 00048 00046 00068 00066 00062 Spruce + Filler + 4 Coats Varnish, No, 2 ... 00040 00036 00052 00054 00052 Spruce + Filler -4- 4 Coats Varnish, Xo. 3 ... 00068 00060 00088 00082 00081 Spruce -4- Filler + 4 Coats Varnish, No. 4 ... 00036 00034 00048 00050 00049 Spruce Untreated Control Specimen 00544 00450 00394 06392 04036 02676 ON THE PERMEABILITY OF CERTAIN VARNISHES TO MOISTURE. BY W. H. LANG. Rate of Entry of Moisture into Air-dry Wood By means of a method which has been described in the previous reports, a number of varnishes have been tested and compared as regards the protection they afford against the entry of moisture into wood. While the theoretical explanation of the results requires further investigation, the tests summarised in this report will, it is hoped, be sufficiently consistent and reliable to be of use. They will also serve to show the effectiveness of the method and encourage those concerned in testing varnishes to apply it for themselves. In this method of testing, the rate of entry of moisture into a test specimen of air-dry wood is determined per square centimetre (or square metre) per 24 hours, at a particular temperature. The air-dry wood in our labora- tory has been found to contain 12-13% of its dry weight of moisture and is in equilibrium with an atmosphere of a relative humidity of 55 to 65. When placed in the saturated atmosphere of a properly constructed moist chamber it tends to absorb moisture up to about 28% of the dry weight of the wood. When submerged in liquid water the wood absorbs a much greater amount, which is mainly held by capillarity in the cavities ol the woody tissue. Waterproof Properties of Varnish. A complete coat- ing of varnish has a remarkable effect on the relations of the wood to moisture. Such a complete coating can usually be obtained by carefully smoothing the test specimen, treating it with silicate wood filler and giving four coats of the varnish. The coating of varnish is not impermeable to moisture. It allows water-vapour to pass from a saturated atmosphere into the wood at a rate some ten times slower than it enters an un- varnished control specimen. When the varnished speci- men is submerged, no water enters by capillarity, but moisture continues to pass in at the same rate as from a saturated atmosphere. Varnishes are Liquid-waterproof, but not Vapour- waterproof. Put shortly, an efficient coating of varnish renders the wood liquid-waterproof, but not vapour- waterproof. though it reduces the rate of passage of water-vapour. Method of Testing. Tests have usually been carried out by placing the test specimen for three days in the moist chamber and then for three days submerged in water. The area of the surface of the specimen being known, and the gain in weight on various days during the test, the gain in weight per unit area is easily ascertained. When this is plotted against the time, the rate of entry of moisture is expressed by the slope of a line which is the same whether the specimen was in a saturated atmosphere or submerged. (Of. Curves I. XI.) This rate depends on the temperature, which must always be noted. The tests summarised here were all carried out in the neighbourhood of 60 F. (55 to 65 F.), and are thus reasonably comparable. For a strict comparison of varnishes, the thickness of the coatings should be measured in each case, but the thickness of the, coatings was sufficiently similar for this to be disregarded here, and four coats of each varnish have been taken for the routine coating. NOTES ON THE TESTS. The various tests are sufficiently described on the accompanying Figs. 113, 114, 115, 116, and 117. It is only necessary to add a few notes regarding the varnishes tested. Oil Varnishes. A number have been tested, some being linseed oil varnishes, others China-wood oil var nishes, while others had a mixture of drying oils. Curves I.-II. A Linseed Oil Propeller Varnish ob- tained from Messrs. A. V. Roe. Numerous tests gave similar results. One specimen was re-tested after ex- posure on a roof through August and September, and showed that the waterproof properties were unimpaired Curve III. An Oak Varnish, used for dipping the fuselage and framework, obtained from Messrs. A. V. Roe. Curve IV. Chins-wood Oil Varnish obtained from A.I.D. A similar result was obtained with other samples. A number of tests made on the above three varnishes confirm their relative positions as regards permeability to moisture. One of these tests (Test M) was described in a previous report. Others have been continued for periods of over 30 weeks' submergence. At the end of this time the rate was slowing off, but the wood still contained less moisture than it could have absorbed from a saturated atmosphere. This fibre saturation point would probably not be passed and no fluid water accumulate in the cavities of the varnished specimen though submerged. 120 The somewhat higher rate of passage of moisture shown by the oak varnish and China-wood oil varnish as com- pared with the propeller varnish may in part be ac- counted for by the two former giving slightly thinner coatings. Curve V. Result of three tests of a good oil varnish supplied by Dr. Morell. It may be mentioned that another varnish sent by Dr. Morell gave an even lower rate, viz., -00024 gr. moisture per sq. cm. per 24 hours at 56 to 58 F. The rates for eight tests of this varnish with three different wood-fillers only ranged between -00021 and -00024 gr. per sq. cm. per day. Curve VI. Eesult for three linseed oil propeller varnishes from A.I.D. ; one markedly inferior to the others as regards permeability to moisture. These tests indicate the range in degree of protection against moisture given by oil varnishes. Since the coating is fairly extensible and not brittle, the protection continues as the wood is expanding. Some Other Varnishes. Curve VII. Shellac in Methylated Spirit, and Brown French Polish from A.I.D. The protection given is similar to that of an oil varnish. The 'coating is, however, brittle, and when- ever the test by submergence was continued a break- down ultimately occurred, cracking of the film taking place and liquid water entering by a capillarity. Curves VIII.-IX. Aerovarn obtained from E.A.E. This varnish when applied as four coats gives efficient protection against moisture, but, probably owing to its brittleness, is very liable to flaws. This brittleness is a serious disadvantage as compared with the properties of oil varnishes. Curve X Black Varnish supplied by Naval au- thorities. This gave a very brittle coating, readily in- jured and detached from the wood. Though it gave a high degree of protection against water-vapour, repeated trials were necessary to get a coating to hold up even for a few days in water as in the test shown. In this the low rate of entry of moisture in vapour was maintained for three days when the specimen was submerged and then the usual breakdown occurred, the coating cracking and liquid water entering. Though it is clear that such a brittle coating is practically useless for wood, the result is instructive from the high degree of resistance to the passaee of moisture afforded by this coating be- fore the breakdown occurs. Curve XI.Bitumastic Solution supplied by Naval authorities. This coating, though with care it gives good protection, is defective in an opposite way to the brittle black varnish just described. It is very slow in drying and never forms anything but a viscous coating, which, owing to its softness, is readily injured, without, how- ever, breaking down. Comparison of Results. The degrees of protection afforded by the various varnishes enumerated above, when applied as four coats over wood filler, can be compared with one another on the diagram in Curve XVII. On this diagram is also plotted the rate of absorption of water-vapour from a saturated atmosphere by untreated control specimens of Oregon pine. It will he seen that the rate of absorption of moisture through the varnish coatings is from about one-tenth of the rate of absorption of water-vapour by unprotected wood down- wards. Three other points may be briefly dealt with. These are the effects of the number of coats of a varnish, the effect of the addition of aluminium leaf or aluminium powder to a varnish coating, and the effect of the fabric- enamel-varnish finishing of propellers. Effect of Number of Coats of a Varnish By a method in which the entry of liquid water at any point could be recognised by the effect of a dye dissolved in it, it has been possible to determine the defective points left after one, two or more coats of a varnish have been applied. It can be said generally that defective places are always present after one coat and that they have usually disappeared with three coats. This accords with the experience that four coats give a safe coating to prevent the entrance of liquid water. Speaking generally, it may also be said that the degree of protection against the entrance of water- vapour is about in proportion to the number of coats. While the coating has defective spots, this does not, of course, hold for submergence in water. A similar rela- tion may hold for submergence tests when more than four coats are applied, though there are indications that the increase in protection falls off as the thickness is increased. These conclusions are rendered probable from the test with one, two and four coats of Aerovarn shown in Curve IX. and the tests with four and eight coats of Aerovarn in Curve VIII. A fuller test done in duplicate on pieces of Oregon pine and of walnut, with one, two, four, eight and sixteen coats of the same good propeller varnish gives more conclusive evidence. The results are plotted in Curves XV. and XVI. A practically important aspect of this question is the degree of protection afforded by a single " dipping " in a varnish, such as is often given to the fuselage and frame- work ">f the planes. The oak varnish, four coats of which gave the protection shown in Curve III., was used for this purpose 1 , and Messrs. Roe kindly had a number of test specimens " dipped " in the usual wn} for me. The result for the " dipped " specimens is shown in Curve XII. The protection against a saturated atmosphere is about one-fourth of that afforded by four coats of the same varnish (cf. Curve TTI.). The entry of liquid water is far more rapid, however, owing to the imperfection of the coating obtained by a single dipping, especially at the ends of the test specimen. Effect of Aluminium Powder and Aluminium Leal in a Varnish Coating. Attention was directed to the effect of aluminium leaf by statements in an American report eommuTiicated to me by Lieut. -Colonel Jenkin, R.A.F. who also suggested the trial of aluminium powder and supplied me with a sample of this. The two tests shown in Curves XTTI. and XIV. illus- trate the effect of combining aluminium powder and aluminium leaf with a China-wood oil varnish. It will be evident that the presence of the powder reduces the rate of entry of moisture by more than 50%, and the presence of a laver of aluminium leaf by about 80%. The difference was maintained in tests continued for over six weeks' submergence. If further tests confirm these results, the method promises to be of practical value. It may be mentioned that preliminary tests indicate that powdered graphite may have a similar effect to aluminium powder. Propellers Covered with Fabric, Filling Enamel anrt Varnish. Under certain schemes for finishing propellers 00 I o U. X 111 U 121 the wood is covered with liiicn fabric glued on, then given several coats of a filling enamel flatted down, and finally two coats of varnish. Messrs. A. V. Hoe kindly provided me with a number of pieces of propeller blades completely covered with this complex coating. Tested by submergence for some weeks, it was found that the rate of entry of moisture through the coating was about the same as that through four coats of a good oil varnish (via., '00030 grammes per sq. cm. per day). The rate fell off rather more rapidly as the test was continued, 'so that at the end of the fourth week the average rate was about '00020 grammes moisture per sq. cm. per day. Conclusions. It seems safe to draw the following conclusions from the experience gained in making these tests : (1) The method affords a satisfactory \vay of in- vestigating, testing and comparing the " water-proof " properties of various coatings that can be applied to wood, if due care is taken in carrying it out. The further use of the method depends on its application in the regular testing laboratories. (2) The coatings investigated (Oil Varnishes, Shellac, Aerovarn, Black Varnish, Bitumastic Solution), when properly applied, all render the wood liquid-waterproof, but differed in the degree of protection afforded against the passage of water-vapour. (3) The method, when all the conditions of the test are kept as similar as possible, allows of even closely similar oil varnishes being compared as regards their resistance to the passage of moisture, and readily separates more dis- similar coatings. Further work on these lines can only be profitably carried on if the exact composition of the varnishes, etc., is known. (4) Various coatings in use afford, if properly ap- plied, such high degrees of protection against moisture that the choice of a particular method may depend upon other properties of the coating. (5) Brittleness of the coating is always to be re- garded as a serious disadvantage. The flaws readily produced in a brittle coating let in liquid water, and lead to the progressive breakdown of the coating. (6) When protection against atmospheric moisture is specially important, it is essential not only that a coating of good moisture-resisting properties should be selected, but that a com- plete coating preventing any entrance of liquid water should be obtained and pre- served. A sufficient number of coats to ensure this must therefore be applied. (7) No coating which completely prevents entrance of water-vapour has been met with among those investigated. AN EXPERIMENT ON THE BELATIVE ABSORPTION OF FLUID WATER BY THE ENDS AND SIDES OF SPRUCE. It was found necessary to ascertain the exact degree to which the ends and sides of a piece of spruce differed in their power of absorbing fluid water. Three small cubical blocks were cut as closely as possible to the same 27264 size and weighed and measured. The pieces were then carefully embedded in paraffin wax so that in one the end only was exposed, while in the other two the radial and tangential faces respectively were left exposed. The embedded blocks were weighed and then floated with the exposed surfaces downwards on water in a dish and weighed at intervals for 28 days. The results are shown in Fig. 118, in which the percentage increase in weight is plotted against time. It is evident that the rate of absorption in the first 48 hours is very much greater by the end grain than by either of the sides. It is equally clear, however, that after this time the rate of absorption, as indicated by the slope of the curves, 18 approximately the same for the end and the two sides. The greater weight of water at the close of the experi- ment in the piece with the end exposed is wholly due to the rapid initial rate of absorption, caused by the capillarity of the open tracheides of the wood. 17. PLYWOOD. The name plywood is used to describe the material of any wooden part built up of several thicknesses with the grain running in different directions in the different plies. It is generally used to mean boards built up of three or more plies such as are sold ready for cutting up for use, but it is also used to describe thin shells built up in place of several layers of veneer, for example, in the fuselages built on the monocoque system, and also the relatively thick blocks built up of many thick- nesses used for cross frames or gun-mountings. The veneers of which the plywood is built up are generally sliced off logs of timber revolving in a lathe, they may therefore be described as thin circumferential planks, but as the stem of the tree generally tapers somewhat and is not accurately round, the veneers are not cut truly parallel to the grain or exactly in a tangential plane. This latter point is of no importance, but the want of parallelism between the length of the grain and the length of the veneer seriously reduces its strength wherever it occurs. As a consequence plywood is very variable in strength and cannot be relied on to have more than half the strength it would have if the veneers were truly parallel to the grain. Experiments have been made in riving off veneers so that they shall be exactly parallel to the grain, and a successful method of doing this has recently been devised. There is good prospect of a considerable advance in this direction. The thickness of the veneers used varies greatly; the Germans have made admirable use of extremely thin sheets l-100th inch thick or even less. In three-ply boards the thickness of the outer and middle plies are usually different. There is con- siderable scope for plywoods of different proportions to suit special purposes. The angle between the grains of the plie-3 is usually a right angle, but for special pur- poses, such as the skin of a fuselage, other angles have been used. What the best timbers are for making plywoods is still a matter of doubt. The middle ply is usually made of a different timber from the outers. In choosing the timbers there are many points to consider, and there is room for valuable investigation in this direction. The chief points to be considered are: strength, flexibility, freedom from warping and good glueing qualities. The plies are glued or cemented together, and the value of the plywood depends largely on the success of this operation. Most plywoods in England are made Q 122 with Blood glue, used hot, or with Casein Cement, used cold, but neither of these gives completely satisfactory results; researches are in progress on new glues, and there is good reason to expect that one at least of the new materials will prove to be very much better than those at present in use. The ideal glue should be strong, adhere to all sorts of timber, resist moisture (hot or cold), not perish in dry heat, resist fungus and bacteria, be unaffected by fireproofing chemicals, and be easy to apply. It is possible to treat plywood with chemicals so as to make it fireproof; such wood chars slowly and does not flame. This process has not hitherto been successful, because the chemicals inter- fered with the glue, but it is probable that in the future it will be extensively used. Theoretical Strength of Plywood The accurate theoretical treatment of the stress-strain relations in a piece of plywood is very involved, since it is apparent that the one set of the plies must affect the lateral contraction or extension of the other set. As a first approximation the values of the modulus of elasticity and the ultimate strength can be calculated on the assumption that the cement has no effect beyond pre- venting lateral buckling of the plies, the individual plies taking loads each in proportion to its modulus of elasticity. Compression along and across the Grain. Let the suffix I define the direction along the grain and suffix r the direction at right angles to the grain, and let : A, = area with grain end on. E, = modulus in direction I. f cl = ultimate strength in compression in direction I. f n = ultimate strength in tension in direction I. Similarly A,., E,.,/.,., f tr correspond to direction at right angles to the grain. Consider first compression or tension and assume that the wood is elastic up to failure ; then Load = Stress x Area = E x Strain x Area = (E,A, + E,.A,.) Strain. Stress = ( -M_+ E r A,.) Strain ^ A where A = Whole Area, E r A r I f E,A, x Strain 1 E,A, I 1 A ) The ultimate stress will be attained when the strain reaches the value , the maximum for the end-on plies, therefore Ultimate Strength = I 1 + f^ } (& \ . ( E,A, ) ( A ) The corresponding modulus of elasticity will be . - Stress K,A r / E,A, \ A~ In order to check the accuracy of this theory a series of tests has been made on pieces of plywood built up of veneers cut carefully along the grain from a piece of Silver Spruce of known properties; tests on commercial plywood are too erratic to be of any use for this purpose owing to the varying inclination of the grain in the veneers. A comparison between the test results and the calculated strength is given in Fig. 119, from which it will be seen that the values calculated as above agree very well with the observed values for the follow- ing cases, i.e., 2 plies end on, 1 at right angles, i.e., lengthwise of usual 3-ply board. 1 ply end on, 2 at right angles, i.e., across usual 3-ply board. 1 ply end on, 1 at right angles, i.e., along or across a 2-ply board. The direction ot the grain in the plies is shown in the drawing by means of arrows, the load being assumed to be applied in the direction up and down the page. Compression at 45 to Direction of the Grain. It is evident that with this typo of stress the approximations suggested will not hold good. A specimen of solid timber compressed in the direction of 45 to the grain will fail by compression at right angles to the grain at about 1,300 Ibs. per square inch. (See Section 4.) In a plywood, however, whilst one layer tends to fail in this manner, the other prevents it, because its grain is end-on to this direction. Under these conditions the type of failure to be expected would be by shearing along the grain of each layer. The shear strength of the spruce of which the test pieces were made is about 1,250 Ibs. per square inch, and the end stress required to produce this shear on a plane at 45 is twice the shear stress, i.e., about 2,500 Ibs. per square inch. In the actual tests the average of four results gave 2,706 Ibs. per square inch at the end stress. The value of E is of course also affected considerably, being increased from 21 x 10 6 to '39 x 10 6 . Practically the same values arc obtained with plywood made of alternate layers at +45 to the axis, and with plywood made of two layers at + 45 and one at -45. For combinations built up of 2 plies at +45 and one end-on the agreement between the value as calcu- lated on the lines indicated above is quite good when the single ply is horizontal (test 9) but not quite so good when the single ply is vertical (test 8). The latter is being investigated further. Tension. Comparatively few tests have been done in tension. With the testing gear available it was difficult to get satisfactory tension tests on the 3-ply spruce with two layers end-on to the stress. For 3-ply with only one layer end-on, the actual values 6,000 and 8,280 Ibs. per square inch, are of the same order as the values calculated in the same way as indicated for compression, viz., 7,100 Ibs. per square inch. For plies at 45 the tensile strength is practically the same as the compression strength, i.e., 2,500 Ibs. per square inch, which is twice the real shear strength of the timber. Shear There is no simple way of estimating the strength of plywood in shear. Some tests were made as described on page 100 with the apparatus shown in Fig. 13. The stresses were calculated on the assumption that the distribution was uniform, not parabolic. For plywoods made up of an inner ply which is equal in thickness to the sum of the outer plies, the shear strength when the stress is along the grain is raised considerably above that of the solid timber, i.e., to about 2,000 Ibs. (Tests 15 and 16.) When all the plies are equal the results show a differ- ence between test 19, where the two outer plies are parallel to the direction of the applied shear, and test 18, where these plies are perpendicular 'o the applied shear. CHAPTER X. ll.H o O O Q? o C> o a * i o o =i 0 iO >0c s o ^ "0 K u t u O ^> PC u CM CM *~~ Cvt 2 i f e oo 00 00 o H (j r - i z ) H ^~ Q QCt < 2 U O M Q U ^ ko ~ Oi 9 M ^ y x 0. J Z ~7 C^r (TJ Ov flQ CM 9 C S t u _ O Q o o ifc J t ? Ul ,t ^ H Q Q iO Q 01 H- CT> O) >O <'l 3 <0 IO IO w a J I ' j g Q Si 01 Q Q S Q O 2; O o Q Q (0 o < Q (O O N 01 O CM Q O 8 Q O id )O )O 01 O O o\ CM Q 8 0) 6 coo &S60 X^i'i-.o A CD - Si (A CM 2? j 5 O O O !o O -0 Q Q Q 03 (0 N O o ^f^ o u H 00 00 u j *) Ci ~ CM "* Q ^ A* Om ( I 4) | t i i ( 1 ( t 4 1 < I t ~? i 4 l ( t < t ( i 1, i ' ' ' ( 1 1 1 i 1 t ( f 1 3000 4 , > I t l F 4 i ! ' 4 i 1 < > T __ 1 1 4OOO ^/3000 eooo ^ i t < ! 1 I 1 1 \ , 1 , ' t ( i i I 1 1 1 < i i 1 ^ 1 1 * < 1 ' r U. 1 i ( 1 i 1 , 1 1 lejLiJ V F A teJInJI Aw A - A i *le A t M i A A t> ty 9& l B Hw * p ik! ftl H V> P ? LNUT OP Uuw r 4 ' M H K Ci H ! W i u if. 5 ft rw > r a > 4- -if e 7 a 9 is 19 to ti u. 11 M AS e 27 t 19 io^i 33 34 35 a ST- w 33 to ig || PLYWOOD ULT. SHEAR STRENGTH. FIG. 123. SOW) 2000 1000 1 t i { 1 1 i i l < ( i i 4 > 41 i i 1 4 , i 1 ; 4 I 1 | 4 1 i * 4 t 4 1 1 4 1 1 4> "1 | bQOO 33 2000 2 1000 ~ '1 r 4 i < > i 4j i 1 4 i 4 i 4 < : t 4 . < I 1 i < l 4 i i i ! i I 1 4 i i 4 } i | ( > i < , ( > 4i 1 u 400fl 2000 Nor i 4 4 1 1 1 i i l i ( < > I ( 1 SK.VJ I i < i 1 1 *?"* i < i Of* 1 I,-*.*, A ( > pop A i i ,. t j i 'oc e t*tr A * A pSp*^ fop MAM BV p. w * 1M rests', I 5 6 Y B 9 IS 19 2.0 Zt 22 23 24 25 26 2r 2 I 3 3 3 i J 7 -1 * " 40 CHAPTER.X. FIG. \d.cL. PLYWOOD BEAM TESTS. YOUNG'S MODULUS A MODULUSOFRUPTURE. 5* o a: r i i i__i tu % 1-0 ' \ 1 1 i 1-2. i .9 8 i a 34 567 as laoaoeiz^ OUTER Ate AASH Afcp.EPop AAwAAaiABiROtA^RCH Aftp Aftr APOp. EPWEPop _. iNfCR Aftp. AA Aftp AftfLA*. AteM ABincn MAN. AA4aABwo4. NAM 29165/34* SOOO. CiR.L T = 2T9 IS /1 3. 123 It is probable that this is due to inaccuracy in the shape of the test piece, and it is, therefore, being investigated 1.1 further. When the plies are arranged at 45 to the direction of the shear there is a further considerable increase of shear strength. As a shear is equivalent to a tension and a compression at right angles, there is naturally a difference according to the amount of material available for resisting the compression. Thus, test 17, with half the section to resist compression, gives values in good agreement with the compression in test 16 and higher than test 20, with only $ the area to resist compression. The result of the latter test is considerably higher than the theoretical, but this is probably due to the shape <>f the test piece which was not of suitable proportions to test this point accurately. Summary. The above results may be briefly and not very inaccurately summarised as follows: (a) For all plywoods tested in tension or compres- sion parallel to one set of plies, the strength and elasticity is approximately that which would be obtained by neglecting altogether the plies with grain at right angles to the stress; more accurate figures are obtained by calculations in the manner described above. (b) For all plywoods tested in tension or compres- sion inclined to the grain, the modulus of elasticity is raised above that corresponding to the same angle in solid timber, and failure is generally determined by the shear strength of the timber. (<' For plywoods tested in shear, the strength is raised approximately 60% above the shear strength of the timber, but for particular arrangements of plies the shear strength in certain directions is raised to nearly thrao times that of the solid timber. Strength of Commercial Plywood. A number of tests on various plywoods (3-ply) have been made, and the results are summarised in Figs. 120, 121, 122, and 123. The results are erratic, so much so that occasionally the test with one ply end-on is stronger than the test with two plies end-on. These irregularities are entirely due to the variation in the direction of the grain, which is inevitable with rotary-cut veneers. Perhaps the most consistent test is that with the plies at 45, because in this case the failure is one of shear, and the curve connecting ultimate shear stress with the inclination of the grain is fairly flat in this region, whereas the curvos connecting ultimate tensile and compression stresses with the inclination of the grain near the end-on posi- tions are exceedingly steep, particularly the tension curve. (See Figs. 20 and 21.) It may be noted thatthis 45 test probably represents the kind of stress which will cause failure in large panels such as are used in fuselage construction with the plywood put on in the usual manner with grain horizontal and vertical. If the ply- wood were put on with the grain approximately along the diagonals of the bays, the fuselage would be stronger and stiffer. Plywood for Sides of Box Spars The value of F for ordinary plywood is low on account of the low E 27264 of the ply which has its grain at right angles to the stress. For the sides of a box spar it is desirable to have a high E so that the sides can take a reasonable proportion of the load. This is best secured by making the inner ply as thick as possible with its grain along the spar and the outer jphes as thin as possible. The function of the outer plies is merely to stiffen the inn-r one against buckling and to prevent splitting. 18. PROPELLER BOSSES. Considerable trouble has been caused from time to time by the heating of the bosses of propellers, which has been sufficient in some cases to set the propeller on fire and often to char the wood. The causes which lead to this trouble are complex and are not within the scope of this report, but two factors which enter into tlie problem must be referred to. There is good reason to believe that in many of the burnt bosses the crushing load of the timber had been exceeded and the hub flanges had been squeezed into the wood. The distribu- tion of pressure over the flanges depends on their design and flexibility; the pressure is usually not uniform. The maximum pressure should not exceed the safe load the timber can carry. The propeller is driven by friction between the flanges and the boss, and there is reason to believe that the shearing strength of the timber on the face of the boss has been exceeded. The shearing strength is much greater along the grain than across it, and in several instances it has been observed that the charring was confined to the two quadrants of the face of the boss where the shear was across the grain (Boiling Shear). As the driving is by friction, the pressure necessary between flanges and boss depends on the coefficient of friction between them. Two series of tests have been made to determine the coefficient of friction between the flanges and the boss, the first on square wooden blocks, and the second on actual pro- peller bosses. The results are given in the followin- report by Mr. Skelton. COEFFICIKNTS OF FRICTION BETWEEN STEEL AND TIMBER. Blocks of various woods, having] contact surfaces measuring 1-2 inches square, were tested under loads viirying in steps up to the crushing strength of the timber used. The loads were applied through a carriage fitted at the top with ball bearings to eliminate friction as far as possible except at one contact surface. The carriage was pulled by means of a calibrated device depending on the measurement of the deflection at the centre of two parallel steel bars linked together at the_ends (Fig. 124). The force was applied by a screw acting on the centre of one of the bars and 'from the centre of the other bar to the carriage by a stirrup fitted with a knife edge. Tests were made both along and across the grain of the timber, on radial and tangential surfaces and on three varieties of steel surfaces, vie. (n) Milled surface (along path of milling cutter). (b i Planed surface (along tool marks). (c) Planed surface (across tool marks). Q 2 124 Of these, the second condition most nearly approxi- mates to that obtaining in practice in the case of a propeller boss and the steel disc of the hub. Several consecutive /ests were made on the first block to find whether the results were consistent, but no noticeable differences occurred after the second test. Two tests were ther ?f >re made on each subsequent block. A further test of the same kind was made to deter- mine the coefficient of friction between planed steel and Ferodo fabric.* The results of the tests are shown in Figs. 126 to 130 and a summary is given in the following table. The results vary so much that the tabulated figures can only be considered a very rough approximation. Approximate Summary of the Tests. Coefficients of Friction. Along Grain. Aoros 9 Grain. Static Sliding Static Sliding Milled Steel and : - Spruce 16 13 16 13 Mahogany 15 11 IS 10 Walnut 16 12 16 12 Ash 16 13 16 13 Planed Steel (across Tool Marks) : Spruce 27 19 22 17 Mahogany 20 16 20 17 Walnut 24 18 24 18 Ash 23 16 26 20 Planed Steel (along Tool Marks): Spruce 20 15 14 12 Mahogany ... 16 14 14 12 Walnut 19 15 16 13 Ash 19 14 17 13 Ferodo fabric 25 18 smooth grey surface whilst the steel was turned to approximately the same finish as a normal hub washer. A further test was made with a washer of Ferodo fabric inserted between the boss and the turned stee'l plate. The pressure was transmitted to the boss in a testing machine through a thrust ball bearing and the torque was applied by means of a screw and wheel. The coefficients of friction were calculated on the assumption that the loads were evenly distributed over the areas in contact. This assumption may be far from the truth in the tests made on a flange supported on a ring erf the same diameter as the bolt hole centres; the distribution of pressure in these tests probably corre- sponds with the distribution under the flange when bolted up. The results are given in Fig. 130, and summarised in the following table, which, however, is only a rough approximation. Measurement of Coefficient of Friction between Machined Steel or Cast Iron and Propeller Boss. In these tests a propeller from an S.E.5 was rotated under various pressures by means of a couple applied through a spring balance to two points on the blades (Fig. 125). The maximum load (15 tons = 772 Ibs. per square inch) is that which would be given by the fixing bolts if they were tightened up to their yield poim. This was determined by tests on actual bolts. Tests were made with the ordinary varnish and paint finish of the propeller, as well as with the surface of the boss planed smooth and clean. Measurements were made on a cast iron machined surface, on a turned steel surface corresponding to the actual disc of a hub except for the absence of lightening holes, and on an actual hub supported by a ring at the radius of bolt holes. The cast iron was machined to a * A proprietary woven material used for lining clutches, friction brakes, Sec. Walnut Propeller Boss and : Cast Iron Plate. (Boss varnished and painted as usually fitted. Cast Iron Plate turned to a smooth grey finish). Turned Steel Plate. (Boss varnished and painted as usually fitted Steel Plate turned similar to finish of hub washer). Turned Steel Plate. (Boss surface planed clean and free from varnish, &c.). Hub Washer. (Boss surface planed clean and smooth. Hub washer supported on J" square section ring at radius of bolt holes). Ferodo Washer. (Boss planed smooth and clean, washer J* thick as actually used). Ferodo Coefficients of Friction. Static. Sliding. 41 42 28 27 26 25 35 20 21 18 The figures for the friction on the complete propeller are not in good agreement with the tests on separate pieces of wood (not of course identically the same). When the Ferodo washer is introduced the results agree fairly well It is observed that the Ferodo results are very regular, but the friction is not as great in the propeller tests as with the boss in the condition as normally used. On the small samples, however, the Ferodo re- sults are better than the timber along the tool marks and about the same as the timber across the tool marks. It is noted, too, that the total friction and also the difference between static and sliding tests, is least when the tool marks and grain are parallel. The tests show that the coefficient of friction is not sufficient to cause the Eolling Shear strength to bo exceeded unless the normal load (applied by the bolts) exceeds the crushing strength of the wood. fVJ o LL O ir UJ fc Q cr UJ s: LJ O 1 cc L_ O LU CO Z o o ct: LL. o I- UJ o O O c/3 u 2 C o ETERMIN l LLJ I LJI Chapfcer.X. FIG. 126. COEFFT. OF FRICTION BETWEEN PLANED STEEL & PLANED WOOD. MAHOGANY. TANGENTIAL PLANK MAHOGANY. TANGENTIAL PLANK. M-ONO TOOL. MARKS ON STfEL, .ALONG GRAIN. OO 4OO 500 too LOA6 IBVCi 1 SPRUCE, TANGENTIAL PLANK. ALONG TOOL OH STtCL ALONO GRAIN. o foe 00 D ALONG TOOL MARKS ON 3TEEL ACROSS GRMN. S^ ^ ^-1 H ^.p" 4 me too too 400 ax> too x SPRUCE. TANGENTIAL PLANK. AlONC TOOL M*P*S ON SIU.L. ACROM GRWN HTTttT COEFFT OF FRICTION BETWEEN PLANED STEEL fc PLANED WOOD ASH. TANGENTIAL PLANK. TOOL MARKS ON STEEL ACBO53' OffMN. ^ a s^ A s* ; 'i _- Ml D *OO 4OO A o oo 1000 laoo 1400 WALNUT. TANGENTIAL PLANK ALONG TOOL MAJTW ON STCJL ALONG OHMN foo aoo 400 300 LOAO LHa-'Q' ASH. TANGENTIAL PLANK ALONG TOOL. MARKS OH STEEL ALONC ORAIM. MO MM UM0 M00 LOAO. tnyo" WALNUT. TANGENTIAL PLAMK ALONG TOOL MARKS ON STZEL. Acnoas QHMN. I! soo 4*0 aoo oo roo 800 Chapher.X. QO oj LU Q_ & LJ i! th 5 LJ CO g b cc Liu o o 00 (\j d ^ a_ i ^ S LJ fe! 1 L- te ^ S o o s z. u o u CO b UJ CD, O Chapber. X. CD cvj O NUT I WO * LJ IT) o LJ s LJ LJ CO W E u_ u_ M o o Chepber. X. O O 6 ui u. V) o Is E NOU3UL a ao i of. TV* 8 O rO 6 U 55 E I* J dO .LM JIM J J3- CD JO < 3< 5 g s S UJ 6 o 29I6S/ 94 4- ?OOO C4W- IZ* 125 19. KEINFORCED TIMBER. Many methods of reinforcing wood with steel or other metals have been proposed during the war, but the principle does not seem likely to be really useful. The only parts in which a moderate amount of success has been attained are stream-line struts. The difficulty of attaching the wood to the metal and of connecting fittings to the metal reinforcement are very great, and speaking generally, it appears to be better to change over from wood to metal in construction rather than to attempt to strengthen the wood. Samples of plywood with embedded wire netting have been tested, but as was to be expected offer no advantages over plain ply- wood. Two methods of reinforcing wooden interplane struts have been tested, both of which gave fair results, but the connection between the metal and the wood was unreliable in both. The following report on sample Haskelite struts from America shows that in the larger sizes there is some gain in strength over solid wooden struts. HASKELITE STRUTS. Heskelite Struts. A series of six struts lias been tested with free ends. The construction of the strut Outer ply grain fibres lie circumferentially (in plane of paper) -~ fnner ply grair, fibres lie parallel cfslrvf. /Steel strips. Fid. 131. is shown in Fig. 131. The results of the tests are given in the accompanying table, and also the calculated load for solid spruce struts of the same external di- mensions. The test figures agree fairly well with the calculated values with one exception, No. 36. Taking the ratio of collapsing load to weight of the struts, it is seen that the small size is a little better than Grade A spruce, and the other size decidedly better. This is to be expected, as the strength is mainly due to the steel which is at the outer fibres where it is most effective. There is no doubt that struts of this type can be made to give good values in a laboratory test, but it does not follow that they will be suitable for service conditions. The steel is held to the wood by nails and it is doubt- ful how these will stand up under the vibration and varying stress to which the struts are subjected. At- mospheric changes, too, are a probable source of trouble, in that the timber will expand and contract more than the steel. Strut Number. * 3') 31 32 33 36 37 Minimum dian.eter, too ins. 1-30 1-29 1-47 1*51 i-7<; 1-72 ,. middle 1-34 1-35 1-52 1-51 1-75 1-75 ., ., liottom ,. 1-27 1-25 1-46 1-49 l-72 : 1-72 Maximujn ., top , 4-03 4-03 4-94 4-89 5-8 5-87 ., middle 4-10 4-12 5-01 4-87 5-86 5-80 ,, bottom ., 4-12 4 '03 4-98 4-85 5-S4 5-76 Length, ins. ... 60-81 60-83 66-23 66-24 C.6-90 66-'<7 Weight, Ibs 3-88 3-88 5-34 5-28 6-37 6-50 1 It. compression strength I of timber ... 1 6,300 6,300 6,300 6,300 6,300 6,300 E of timber HI 8 Ibs./sq. in. ... 1-72 1-72 1-72 1-72 1-72 l-',2 E (steel) 10 C Ibs. sq./m. 30 30 30 30' 30 3D Collapsing load, Ibs.... 2,150 2.05" 3,000 3,100 ' 4,200 5,000 Euler value, Ibs. 2,200 2,200 3300 3,400 5,600 5,600 Ra io Colla P sin f? Load 1 555 K90 K9 587 660 770 Weight I Solid Spruce Stru-.s of Grade A quality having the same dimensions as Haskeite Struts. Euler value | 1,904 2,720 4,696 Ratio Collapsing Load . | _ 529 473 fioi; Weight ) Weight, Ibs j 3-6 5-76 7-75 20. SUBSTITUTES FOK TIMBER. Many sorts of manufactured materials have been proposed as substitutes for timber, but none have shown any promise when tested. The material most commonly suggested is some form of vulcanised fibre. The follow- ing report on a sample indicates the sort of properties of such materials. Vulcanised Fibre The results of the tests are given in the following- table. The material is very heavy, having a density of 69 Ibs. per cubic foot for the red fibre, and 79 Ibs. per cubic foot for the grey. The modulus of elasticity is low, varying from '78 to 1-15. The material under compression flows at compara- tively low stresses and continues to do so at an increased rate as load is applied. Thus, in the compression test the elastic limit is from 1,500 to 2,900 Ibs. per square inch, and in the ultimate stress about 6,500 to 7,500 Ibs. per square inch, and in the bend test the material cannot be broken with the usual arrangement for timber owing to the extent of the plastic flow. Beams 1 inch x 1 inch will bend double before breaking. In tension it appears to break off quite short and sharp with but little elongation. The tension values are of the order of 13,000 Ibs. per square inch. Glue Tests. The tests that have been carried out are relatively few, but all point to the fact that the fibre does not glue so readily as timber. Thus, in the Spandau test the values obtained are only about 800 Ibs. per square inch; in the adhesion tension test about 460 Ibs. to 860 Ibs per square inch; in the shear test on surfaces inclined 20 to line of pull the values are 1,400 to 1,700 Ibs per square inch. In the A.I.D. shear test the values are 480 to 640 Ibs. per square inch. All these values are less than are obtained with good timber, such as is employed in aeronautical work. 126 FIBRE TESTS. Murk Density. Bend Test. Ten- sion. Ult. i Compression Tf at. Shear Strength. Glue Tests. Uniform Bend. Central Load. E x 10". x K Elastic Limii. Uli, Strength. Spandau. Adhesion. A.I.I). Shear Test. Glued to - E x 10. Kl. Lt. I 1 No. 1. Grey No. 2. Grey No. 3. Bsd Ibs./ fa in. 79'3 77-6 69'2 Ibs./ sq. in. ris I'M 86 Ibs./ sq. in. 1,910 1,770 2,740 Ibs./ fa. in. I'OO 97 837 Ibs./ sq. in. 12,720 13380 11,790 lb-./ sq. in. MIS 984 778 1's./ sq.in. 2,1''0 2,900 1,500 Its./ sq. in. 7,;>44 7,499 6,fO:, Ibs/ ^q. in. 1,479 l;<88 1,707 Ibs./ -q. Ml. 240 880 FOO Ibs/ sq. in. Si* 516 160 Ibs./ sq. in. 640 480 480 Ibs./ sq. in. m 610 620 Paper gave way. Not j_'lue. APPENDIX I. EXTRACT FROM AIR BOARD SPECIFICATION 2V. 1. FOR SILVER SPRUCE AND APPROVED SUBSTITUTES. 1. QUALITY. (a) The timber is to be the first quality of any of the following woods, viz. : Silver Spruce (Picea siti'hensh, Carr.), Quebec Spruce (Picet alba and Pieea nigra, Link), White Sea White Deal (Picea excelsa, Link), White Sea Rjd Deal (Piuus gylcettrit, L.), West Virginia Spruce (Pirea Rubens, Sargent), and North Carolina Spruce, when this is the same wood as West Virginia Spruce, but grown in North Carolina, Port Orford Cedar (Chamaeeyparii Lauitaniana, Murr.), New Zealand Kauri (Agathii [ Dammara] Auitralis, Salisb.), Canadian White Pine (P, Strobus, L.), Oregon Pine (Pneudiitsiuja Douglassii, Carr.). (ft) It should be butt-lengths, slow grown (not less than six annual rings per inch) and preferably rift-sawn. (c) Grade*. All approved timbers complying with Grade A tests are to be classed as Grade A. All approved timbefs complying with Grade B tests are to be classed as Grade B. 2. FREEDOM FROM DEFECTS. -The timber is to be clean, straight- grained, free from dote, deleterious shakes, knots, and resin pockets. It is to be cut parallel to the grain (as determined by the Splitting Test specified in Clause 5). 3. SEASONING. The timber is to be thoroughly seasoned naturally, if possible ; but, if not, may be conditioned after cutting into over- head sizes. The conditioning is to be carried out in a well-ventilated place at a temperature not exceeding 85 Fahr., under systematic A.I.D. inspection. The moisture at the end of the process is to be between H and 17 per cent., calculated on the weight of the dry wood. An autographic record is to be kept, showing the temperature and humidity during the process. 4. WEIGHT. The weight is not to be less than 25 Ibs. per cubic foot when it contains 15 per cent, of moisture. 5. MECHANICAL TESTS. (a) The timber is to comply with the following tests, which are to be carried out in the presence of the Inspector and to his satisfaction. ($> Suitable methods and apparatus for carrying out the tests are described in V. 1, Schedule 1. These are not ins'sted on, but in case of dispute check tests are to be made in the manner there described, and the results so obtained are to be accepted as deci.- ive. (e) Cumi'retxion Tent. Test pieces turned parallel to the grain and lo the form and dimensions shown in Fig. 1 (or alten atively cut 1 inch square and between 2 and 3 inches long), from samples selected as specified in Clause (>, when tested in compression must give an ultimate strength not less than : Ultimate st 'engtb, Gra'^e A ... 5,000 lb*. per square inch. ., B .. 4,000 The load is to be applied at a rate I.etween 3.000 and 6,000 Ibs. per min. (d) Drynefs Correction. These compression tests are for timber containing 15 per cent, of moisture. If the timber when tested con- tains more moisture the specified strengths are to be reduced by 230 Ibs. per square inch for every 1 per cent, increase of mo'sture above 15 percent., and if the timber contains less moisture the spicified strengths are to be increased at the same rate. The percentage of moisture is calculated on the weight of the dried sample. (e) Bending Test. Test pieces, 40 inches long by 2 inches deep by 1 inch wide, cut from samples selected as specified in Clame b', parallel to the grain, are to be loaded so as to produce iu the middle part a pure bending moment (without shear). The deflection of the part subject to simple bending is to be measured and the value of Young's Modulus calculaled from it. The results must not be less than Young's Modulus Grade A ... 1.6'iO,000 Ibs. per square inch. B ... 1,200,OCO ,. (O Splitting Test. Short samples, s:iy 4 to 6 inches long, are to be split in two planes, one tangentinl and one radial. The split faces will show the true direction of the grain, which must not be inclined to the length of the plank by more than 1 in 20 for Grades A and B. (0) Jirittlenesi Test. A notched test piece of the dimensions shown in Fig. 2 when broken in a notched bar testing machine must not absorb less than : For Grade A timber 8 ft. Ibs. B ... ... ... 4 APPENDIX II. AIR B04RD SPECIFICATION V.I. SCHEDULE I. INSPECTING AND TESTING SILVER SPRUCE AND APPROVED ALTERNATIVES. SPECIFICATION V. 1, CLAUSE 2. () The Umber is tv lie utraight- omitted. The grain should not be obviously wavy. If the grain is wavy the elastic limit and modulus and ultimate strength are con- siderably lowered and a beam is liable to break by a shear fracture on the tension side under a much lower load than it would carry if the grain were straight. (ft) The plank is to be nit parallel ta the grain. It is often not easy to to see from the appearance of a plank whether it is parallel to the grain or not. If the plank has one of its faces in a tangential plane, it is almost impossible to tell if it is incli-ied in that plane or not. The splitting test (Spec. V. 1, Clause 5) should therefore never be omitted. No deviation of more than one in twenty from parallelism should be allowed. (e) Cross grain can easily be detected by the workmen, and they should be directed to call attention to it. 127 SPECIFICATION V. 1, Clause 5. (a) Cvmprettion Tents. The samples should be turned approximately to the sketch in the Specifi- cation. The ends must be turned quite flat ; the small centre pip left when parted off in the lathe may be smoothed off with a chisel. (>) It is usually sufficiently accurate to test the samples between the faces of the testing machine, but if the faces are not quite parallel, or if the testing machine does not load the sample quite evenly, too low results will be obtained. To eliminate errors due to defects in the testing machine and to ensure that the loading is truly axial, the samples may be fitted with end collars and steel balls, such as are shown in Fig. I. (c) Several test pieces should be turned from each sample to be tested and the quality of the timber judged from the average of the results. (W) The rate of applying the load should be within the specified limits. Very fast loading will give too high results ; very slow loading too low results. At the specified rate the test takes about one minute. A close watch must be kept on the beam of the testing machine when approaching the maximum load und the loading stopped the moment the beam begins to drop. (e) Plain rectangular blocks are often ued for compression tests ; suoh blocks usually fail at slightly lower stress than pieces turned to the shape shown in the Specification. They always fail by crushing at the ends. (/) Bending Tet 1 . " Four point loading " (Fig. 2) is to be used, so that there will be no shearing stress in the middle parto the beam where the deflec ion is measured. The load should be applied through saddles to avoid crushing the timber. The points of application of the load should be in the neutral plane of the beam. Endways stresses should be avoided by means of rollers or ball bearings. Suitable arrangements are shown in Fig. 3. The deflection d is to be measure'! on a length I in the middle part of the beam ; a suitable deflection meter is shown in Fig. 4. ( ~ 12 where J> = breadth of beam, A = depth of beam, a and I are shown in Fig. 2. f A) Several samples may be tested and the quality of the timber judged from the average of the results. BKITTLBNESS. A few samples of timber have been found which comply with the Specification, but are brittle (so-called "dead wood"). It is therefore desirable to carry ac least one of the be ding tests on each sample of timber as far as rupture. Good timber fails first by crushing at several points and finally breaks on the tension side with a long splintering fracture. Brittle timber breaks off short and shows very few and shallow crushing lines on the compression side. Such brittle timber should not be accepted. The tests on notched bars give clear indications if the timber is brittle. APPENDIX III. PRINCIPAL STRESSES. P FIG. 132. The following brief summary of the ordinary elementary theory of stresses will make the meaniug of the statement in section 4 about the components of the principal stresses and the presence of shear stresses clear. Simple compression or tension. Fig. 132 represents a body under a simple compres-ion (or tension) stress p. The stress across any plane BB, inclined at an angle 9 with the normal plane AA, may be resolved into two components, a normal ttress n and a shear stress q, whose magnitudes are : n = p cos* 9 q = \p sin 2 9. These magnitudes are plotted in Fig. 133. The maximum shear stress occurs when 8 = 45; q is tl.e ; %p. The stresses and forces on an elementary prism, when 0=45 are shown in Fig. 132. L'he existence of tlie normal force across the plane at 45, as well as the shear, must not be forgotten. Shear. Shear stresses cm only exist in pairs in mutually per- pendicular planes Such a pair of shear stresses are equivalent to two equal simple stresses of opposite signs on planes bisecting the angles between the shear planes. Such a pair of shears in horizontal and vertical planes are shown in Fig. 134. The stresses and forces on two elementary prisms are shown in the same figure. The normal stresses p are equal to the shear stresses q. Ihis condition of stress must not be confounded with that shown in Fig. 132 to which it lias a misleading similarity. Shear stresses exist in conjunction with all ether conditions of stresses (except ''hydraulic " s'ress, with which we are notconceri ed), and are not confined to the few ca=es usually dealt with in text-books, where alone they are of importance in steel structures. Indeed, the common method of treating them is liable to be misleading, for example the parabolic distribution of shear stress in a beam only 128 refers to the horizontal and rrrticid shear stresses and not to the iiiajcimuiu shear stresses. Instead of being nil at the top and bottom of a beam the maximum shear stresses are there equal to half the at once from the ellipse. principal stresses and its plane of action bisects the angle between the principal stresses. The stresses across any plane may be found tension or compression stress, as explained above, and are inclined at 45 with the horizontal. Shear stresses are difficult to estimate because of the uncertainty of their distribution over the planes in which they act. If th<* stress 1-0 9 & 7 6 5 10 ?0 30 40 50" 60 70 60 90 FIG. 133. w. Stress -I Force - -/2 FIG. The most satisfactory method of representing the stresses com- FIG. 135. is distributed uniformly the maximum value is only two-thirds of pletely is to draw the ellipses of stress at a number of points, what it is if they are distributed parabolical ly. The ellipses of TU~ :~ f, n A ,.,;,,,,. nwa fartfraont Hi o , ,,- 1 ,, / - i , 1-1 1 ati>aaac,a onH fho stress are drawn for a few positions in a simple beam in Fig. 135 and The major and minoi axes represent the principal stresses and the maximum shear stress is half the algebraic difference between the the values of the principal and maximum shear stresses are shown. APPENDIX IV. CALIBRATION OF IMPACT TESTING MACHINE. Let a = initial position of pendulum (Fig. 136) /3 = final position of pendulum after breaking specimen. Energy of blow = 2 $ wr (1 cos ) = WR (1 cos a) where WR = moment of pendulum about shaft when the pendulum is horizontal. Energy in pendulum after fracturing specimen = WR(1 cos/3) Energy absorbed = Wit (cos 13 cos a), If a is 90, i.e., initial position horizontal, then Energy absorbed = WR (cos /3), WR is obtained by measuring the force at a convenient distance to maintain pendulum horizontal. It will generally be convenient to work with a = 90. If, however, greater sensitiveness is desired when testing brittle material it will be preferable to work with a considerably less and thus use a more open part of the scale. The scale may be cotstructed to give the energy directly for any given initial setting, but the following will generally be the most convenient way to use the machine. Adjust the moment to 40 ft. Ibs. when a = 90 and mark off on the dial the angle for increments of 1 ft. Ib. : 1 ft. Ib. /3 = cos- 1 fa 2 ft. Ibs. /3 = cos" 1 fa &o. FIG. 136. 129 The scale should be fixed so that the pointer is over the 40 ft. Ib. mark when the pendulum ia hanging free, and the drop should be adjusted eo that at the end of a free swing (with no specimen in) the pointer is pushed over to ft. Ib. For greater sensitiveness the drop should be adjusted to give 20 ft. Ibs. and this amount subtracted from the reading of the scale. For more detailed investigations, particularly on brittle timbers, it will be convenient to have, in addition to this scale, a scale of angles so that any drop may be used and the values of the energy calculated APPENDIX V. SUMMARY OF TESTS MADE BY THE MATERIALS SECTION ON VARIOUS TIMBERS. Density Ibs. / cubic foot. Modulus of Elasticity. Compression. Name. Botanical Name. y' m 1 "3 <4H (J 3 be . . gj f>s 1 1 ** H O> M m o P ac If a a bo"* If ll| H be p> dP ^* il 1 H SF p * a M q -g e5 o "U V 1 1 ,J3 > f3 GQ 3 a| a d I gj > II? S) O "^ o 1 t< fcjD 3J 8-SrP 3> O 'T? o d 0) 8.JU III tfi ad PH^H fi-j? * PH ^ ^ PH" ja * Mahogany V Swietcnia mahagoni (Jacq.) ... and 9 32-9 + 2-1 -;>*{ 3 1-492 3 1-492 + 1-4 + 3-2 - 3-7( - 3-7) 9 6,516 +20-9 18-2 Cuban Mahogany J Kit'icti-nia niacroi>hijllii (Kin^) 1 43-8 1 1-68 1 11,100 Australian Rosew'd | or Pencil Cedar. 1 Dynu'ijl inn /; i-ff;(Benth.) 2 35-8 + o : 1 2 1-655 2 1-665 + 1-2 + 4-2 - 1-8-1 - 4-8/ 2 7,270 + 1-2 - 1-3 Gaboon Mahogany ... 10 27-84 + 9-6 - 9-1 10 1-46 + 21-2 -17-2 10 5,400 +30-3 -39-9 Andaman Padouk ... Ptenicarpvs indicnn (Willd.) ... 4 48-44 + 5-6 -12-4J 4 1-818 3 ! 1-968 + 11-1 + 3-9 -16-1 { - 5-2) 4 11,745 + 12-8 -19-2 Australian Beanwood Dijto-rylum muelleri (Benth ) ... 2 43-75 + 5-9 - 5-9J 2 1-877 2 2-061 + 5-1 + 5-2 - 5-2/ 3 8,537 + 5-9 - 7-7 American Whitewood lAriudfndrimt'uHpifera (\Arrn.)... 1 34-02 + 5-8 - 4-4{ 6 6 1-72 1-67 + 10-5 + 16-7 -16'9\ -25-2J 6 7,208 + 17-6 -11-5 English Willow Snli.r caprea (Linn) 3 30-55 + 1-8 - 1-SJ 3 3 1-157 1-209 + 4-7 + 6-5 - 3-3\ - 4-9/ 3 4,608 + 3-9 - 6 Louisiana Cypress Taxadium distichum (Rich ) ... 6 44-52 + 11-4 - 7-9 6 1-219 +53 -29-8 7 4,887 +27-9 -16-6 (over 25 o/o mois- ture). Cypress (16 % and Tnxodium dittichum (Rich.) ... 25 30-46 + 18-5 -11-4 25 1-382 +49-7 -56-8 42 5,887 +46-08 -35-1 under). Benin Walnut Khuya sp. 6 36-9 +15-1 - 4-9 6 1-245 + 14-8 -11-6 6 10,365 +27-3 -21-1 Black Walnut Jmjlam nigra (Linn.) ... 14 34-67 + 24-0 - 3-6 2 1-45 + 14-7 -14-8 2 7,895 + 10-4 -10-1 Japanese Kuru ri Juglam mandshwrica (Maxim) 3 32-07 + 10-4 -13-0 3 1 -4*5 + 4-1 - 4-9 3 7,040 + 16-6 -10-3 (Walnut). African Wa'nut Khaya sp. 2 31-05 + 3-1 - 3-1 7 1-126 +27-88 -33-4 12 5488 + 8-2 -31-8 Spruce from Scotland Pirea ercflsa (Link) 4 25-92 + 7-2 - 3-5 4 1-34 + 8-2 - 8-9 8 4644 + 12-2 - 8-5 Quebec Spruce Picea alba and P. Nigra, L. 12 19 29-50 30-48 + 12-6 + 8-3 - 6-4 - 9-1 12 19 1-378 1-797 +33-4 +12-4 -19-7 -15-3 12 19 5,222 5,49 1 + 12-7 - 9-7 -10-1 Spruce, Snail-grain ... Picea Sttelteni (Carriere) 5 33-0 +10-9 - 1-5 5 2-26 + 5-3 - 3-5 5 6,708 + 11-2 - 8-4 12 1.41 +21 *9 1 7 . Q . Silver Spruce Pit-fa Sitchentit (Carriere) 1J 26-39 +23-0 -13-4] 12 A T 1 1-38 +27-5 i 1 O 1 -17-4 j 12 4,902 +26-6 -19-1 Afara Triploc/iiton sp. ... 5 33-76 + 7-2 - 9-7 5 1-52 +12-5 -LS-9 5 6,8.13 + 5-8 - 8-1 Arere... Termiitalia sp. ... 5 22-72 + 7-3 - 4-0 5 0-804 + 16-9 -12-9 5 4,19.) + 4-0 - 8-4 Idigbo Terminalia sp. ... 5 37-1 + 4-6 - 4-3 5 1 45 + 2-1 - 4-2 6 6,901 + 5-9 -21-3 Aspen ... Populus tremidtt (Linn.) 2 36-65 + 0-7 - 0-7 2 1-39 + 11-5 -10-8 2 7,040 + 5-8 - 5-9 American fottonwood Pupulut monilifera (Ait.) 10 29-95 + 11-8 - 7-8 10 1-35 +56-3 -37-1 10 6,169 + 9-2 -20-7 Wych Elm Uhnux intnitana (Sin ) ... 19 38-2 + 1-6 6-oJ 6 1-66 6 ll-76 13 1-098 7 IO-917 + 13-85 +10-8 +54-8 + 11-2 -22-9\ -23-3 / -23-91 -17-SJ 6 18 11,768 5,134 +12-8 +21-3 -10-1 - 8-5 Australian Silkwood 3 37-67 + 0-6 - 1-2 3 1-79 + 8-3 - 5-6 3 6,710 + 4-9 - 5-4 Quartered American Quercns Ma (Linn.) 6 46-95 + 9-0 -13-5 6 1-766 +29-3 -31-9 1? 9,204 +24-9 -23-7 White Oak. Japanese Oak Quercus sp. 6 38-16 + 4-8 - 9-1 6 1-475 +17 -24-5 6 5,995 + 12-7 -18-5 Austrim Oak Qurrcvs Hubtir (Linn.) 12 42-77 + 6-9 - 6-0 12 1-61 +31-05 -23 12 6,516 +31-0 -19-7 Canadian Whitepine 1'iiius StrobUB (Linn.) 4 80-62 + 3-9 -5-7 4 1-539 +25-5 -20-9 4 5,802 + 5-8 -13-H Indian Rosewood Dalbergia latifilia (Roxb.) 6 51-2 + 3-3 - 5-7 6 1-88 + 12-7 1- 9-6 6 9.592 + 12-3 -11-4 Australian Rosewood Dytoxyfa in frasrru n u in ^ Berth .) 6 47-87 + 1-9 - 4-3 6 2-252 +27-8 - 9-7 6 8,025 + 3-4 - 6-2 Vontwiifln Rnvwnnrl 6 60-18 + 3-0 4 * S 6 1-96 + 17-8 24-5 6 7,653 L 4-fi 3-7 Ceiba Ceiba petandra (Gaertn.) 8 11-01 +71-0 -U7-2/ 6 2 0-212 0-566 + 18-4 -22-2 + 0-35 0-35 6 4 885 +44-5 -26-9 2.504 + 5-4 -14-2 White Sea White Deal Picea excelta (Link) ... 7 28-86 -j 6*1 - 2-3 7 jl-53 + 9-1 - 9-8 7 5,263 + 6-6 -11-9 White Sea Red Deal finus syh-estris (Linn.)] 7 33-47 + 14-1 -10-1 7 1-75 + 12-5 -10-3 7 6 13(i + 9-6 -19-8 27264 B 130 Appendix V continued. SUMMARY OF TESTS MADE BY THE MATERIALS SECTION ON VARIOUS TIMBERS cortinued. Density Ibs./ cubic foot. Modulus of Elasticity. Compression. Name. Botanical Name. 1 CO a "3 ii C3 $ 'ca |i "8 S & OJ r^H C3 Sr C8 ^ 1 111 ll| S > 4) a cs *- 1e * H , > *M ' fcj0 fjS ||| M > > B 49 fi> g a f | 1 o 2 V CO O J9 E * fe g o} o te O 2 o S !* o > t-l bC Q M | O a> T, d ! . > ^ J2 6 f ||> 00-3 " - c is Pn-J < p . ,] O P-l '" .0 " om- on volume and weight when oven from green to oven-dry conditioni. Static Bending. pression parallel to grain. perpen- Uicular Shear- Hard- Impact Bending. No. COMMON AND BOTANICAL NAMES. dry. Weight at 16 per to grain. ing strength ness per- pendicu- li t> f r. oS a-o aS moisture _. '3 ! It SH o JL M +? _s 3 60J 1.2 *5 ** *3 parallel to grain. lar to grain (a) 11 as. a 1 1* |1 ^1 (/) S3 jjl o '2 tSJ HARDWOODS. Lbs. pel cu. ft. Per cent. Per cent. Lbs. per Lbs. per sq. in. 1000 Ibs. per In.-lbs. per Lbs. pc-r sq. ill. Lbs. per sq. in. Lbs. per sq. in. Poun ds. Lbs. pei sq. in. In.-lbs. per cu. ft. sq. in. cu. in. cu. in. 1 Ash. Commercial white (Frixinus 62 56 40 4-5 7-1 7700 12700 1500 14-2 6000 1300 1750 1150 15220 7-3 auiericana. Fraxinus lanceo- laia, Fraxinus quadrangnlata) 2 Ash, Black ( Fraxinus nigra ) 53 48 35 5-0 7-8 6800 10600 1400 14-1 4900 800 1350 740 10020 4-1 3 Basswood (Tilia americana) 40 36 26 6-6 9-3 4700 7200 1300 6-4 3800 400 880 340 8350 3-1 4 Beech (Kagus atropunicea) 66 60 41 4-8 10-6 7400 12600 1500 13-3 5900 1100 1700 1060 16430 7-4 5 Birch (Betnla lutea, lenta) 67 61 43 7-0 8-5 8400 13500 1800 17-6 lilillll 1000 1C20 1070 16660 7-3 6 Cherry, Black (Prunus serotina) 53 48 35 3-7 7-1 7300 10600 1400 12-0 5800 700 1500 830 12330 4-9 7 Cottonwood (Popnlns deltoides) 43 39 28 3-9 9-2 4500 7000 1200 7-3 3800 400 800 380 7280 2-3 8 Elm, Rock (Ulmus racemosa) ... 66 6(1 44 4-8 8-1 6700 12500 1400 19-3 5800 1200 1650 1200 14260 6-4 9 Clam, Red (Liquidambar styra- 53 48 34 6-2 9-9 6700 10400 1400 11-0 4900 700 1500 650 16090 8-1 cifiua). 10 Hickory (True Hickories) (Hico- 81 73 50 7-3 11-4 8900 16300 1900 28-0 7300 1800 1800 19910 10-1 ria glabra, laciuiosa, alba, ovata). 11 Mahogany (True) (Swietenia 54 50 36 3'5 4-2 7000 10000 1300 9-1 5500 1000 1420 860 mahogani). 12 Mahogany, African (Khaya sene- 50 46 34 4-8 5-6 7100 10400 1400 10-3 5100 900 1270 730 galensis). 13 Maple, Hard (Acer saccharum) 66 60 42 4-8 9-2 8100 12900 1600 12-9 6500 1200 1990 1200 KilOO 7-5 14 Oak. Commercial white (Quercus 72 65 46 5-3 9-2 6700 12000 1400 12-7 5900 1300 1760 1270 14980 6-6 alba, macrocarpa miner, mi- chauxii). 15 Poplar, Yellow (Liriodendron 42 38 28 4-1 6-9 4800 7500 1300 6-2 4100 400 900 370 11290 4'6 tulipifera). 16 Walnut, Black (Juglans nigra)... 56 52 38 5-2 7-1 7900 11900 1500 13-1 6100 looo I3oo 950 14000 6-4 17 Cedar, Incense (Libocedrus de- 36 32 26 3-3 3-7 4900 7100 1000 6-0 4300 COO too 430 8390 3-2 currens). 18 Cedar, Port Orford (Chamaecy- 47 42 31 5-2 8-1 6200 10300 1700 9-7 5300 700 1160 580 13070 4-7 paris lawsoniana). 19 Cedar, Western Bed (Thuja pli- 34 31 23 2-5 5-1 4200 6400 1000 5-5 4000 400 790 300 7960 2-7 ca ta). 20 Cedar, White (northern) (Thuja 32 29 22 2-1 4-9 4200 5800 750 5-1 3400 )150 800 300 1)530 t-6 occidentalis). 21 Cypress, Bald (Taxodium di- 47 42 31 3'8 6-0 6100 8800 1300 0-8 5400 670 940 4(iO 9450 3-6 tichum). 22 Douglas Fir (Pseudotsuga taxi- 52 47 34 6-0 7-9 6800 9746 1780 7'2 6000 750 1020 580 10930 3-7 folia). 23 Pine, Sugar (Pinus Lambcrtiana) 39 36 27 2-9 5-6 5300 7400 1100 5-0 (MO 540 950 410 8940 :i i; 24 Pine, Western White (Pinus 45 40 29 4-1 7-4 5100 7800 1400 6-9 4800 480 670 360 10230 3-6 monticola). 25 Pine, White (Pinus strobus) 39 36 27 2-2 5-9 5100 7400 1200 6-1 4500 530 too 380 78M 2-7 26 Pine, Norway (Pinus resinosa) ... 51 46 33 4-6 7-2 7900 10900 1700 6-1 lilOO 720 1150 540 13290 5-3 27 Spruce, Red, White, Sitka (Picea 41 36 27 3-9 7-5 6100 7900 1300 7-4 4300 500 920 430 9740 3-5 rubens. canadensis, sitchensis). (a) Load required to embed a 0'444 inch steel ball in the wood to a depth of 0-222 inch. CHAPTER XL GLUE. The importance of glue has steadily increased with the introduction of built-up spars, struts and longerons and with the increasing use of plywood. A glue to bo satisfactory must fulfil very exacting requirements. It must be strong and resist moisture and dry heat, also fungus and bacterial attacks, and must not deteriorate with time. Researches on glues of many sorts have been carried on during the whole of the war. 27264 One of the principal difficulties in investigating the pro- perties of glues has been to devise any reliable mechani- cal tests to measure their strengths ; the following report by Major Robertson deals with this subject very fully. The whole subject is at present being investigated by the Adhesives Committee of the Conjoint Board of Scientific Societies. The present state of knowledge is not sufficient to make it possible to make any completely E 2 132 satisfactory comparison between the merits of Bono Glue, Hide Glue, Blood Glue, Casein Cement or the recently invented Bakelite Glue. Lieut. Kernot's report on the Manufacture and Application of Glue (C.I.M. 707) is reproduced here and also his report on Casein Cement. 1. STRENGTH TESTS OF ADHESIVES FOE TIMBER. By Major Robertson. In a joint made by cementing two pieces of wood together, failure under a given set of applied forces may occur in three ways, by the rupture of the adhesive, or by the breakdown of the adhesion of the cement to the wood, or by the rupture of the wood. From the prac- tical point of view, a cement may be considered satisfac- tory which ensures that under the actual conditions of loading the joint is stronger than the wood, but in an in vestigation on the properties of adhesives it is obviously necessary to devise methods of measuring its actual strength and adhesion. Joints in timber work are generally parallel to, cr inclined at an angle of approximately 1 in 9 to the fibres of the wood. End-grain joints are employed, but are not generally regarded as satisfactory for taking stress. In joints of these types the principal stresses to be con- sidered are tension across and shear along the joint. Compression may safely be neglected as the adhesive is always in the form of a very thin layer and cannot therefore produce failure. So far tests have only been made under static conditions, but it is desirable that in- vestigations should be extended to dynamic stresses and alternating stresses. The true strength of an adhesive cannot be measured if it is greater than the strength of the wood the test piece is made of. The following table gives approxi- mately the strengths of a few timbers which are com- monly used for testing adhesives. that cutting the joint down to less than J its original length actually enabled it to support a greater total load. The results of these tests were : Spruce. Walnut. Mahogany. Lbs./sq. in. Lbs./sq. in. Lbs./sq. in. Tension, along the grain ... 18,000 28,000 21,000 , across the grain, 800 1600 1800 radially. across the grain, 600 1100 800 tangentially. Shear, along the grain 1200 2400 2200 Compression, along the 5000 7000 6000 grain. Thus with Walnut test pieces, the tensile strength of joints between tangential faces cannot be measured if it exceeds 1,100 Ibs. per square inch, nor can the shear strength of a joint along the grain if it exceeds 2,400 Ibs. per square inch Tension Tests. Joints made along the Grain. Two types of test pieces were formerly used at the R.A.E., the B.M. type and the E.A.F. type (see Figs. 1 and 2). In both these test pieces the joint is parallel to the fibres, and the direction of pull at right angles to the fibres. Both are unsatisfactory on account of the un- equal distribution of stress produced by the bending of the specimen when loaded. This is particularly marked in the B.M. test piece which is very shallow; in a series of tests using this test piece it was found Ultimate Stress on Joint (Casein Cement). B.M. Pattern. R.A.F. Pattern. Full s : ze test piece Lbs./sq. in. 78, 48, 37, 64 62. Lbs./sq. in. 144, 140, 11*, 156. Glued up full size and then *id.-e cut away to leave a joint J in. long, as shown dotted in Figs. 1 and 2. 133, 484, 263. 260, 505, 300, 402. 322, 113,224, 288, 210. The simplest test piece for this test is that shown in Fig. 3, which is turned up after the joint is made. If this is tested in axial-loading shackles satisfactory re- sults are obtained. These test pieces are rather small, and for some purposes it is desirable to use a larger one; the test piece shown in Fig. 4, has been used and gives fairly satisfactory results. A typical set of results, using this test piece, is : Ultimate Stress in Ibs./sq. in. Solid Timber... 1,120, 1,130, 1,310. Joint, Glue ... 700, 916, 806, 924, 1,070, 860, 570. Casein ... 410, 527, 470, 500, 542, 585, 558. The results on solid timber test pieces agree with simple tension tests on the same timber (1,190, 1,300 Ibs- per sq. in.) so that this design secures a reasonably good stress distribution. The results of tests on tins joint are rather irregular, but it is pmbable that witii a better technique for making them they could be con- siderably improved. Similar figures to the above have been obtained for joints in Ash, but joints in Beech generally give much lower results. From some tension tests on a very small piece of glue (turned up from a piece of cake glue), the tension stress appears to be at least 3,000 Ibs. per square inch, it is clear therefore that in the joint tests described above, the failure is in the adhesion of the glue to the timber. End-Grain Joints. It the joint is made on the end grain, the upper limit of strength of the joint imposed by the timber is over 18,000 Ibs. per square inch. A. convenient form of test piece is that shown in Fig. 5, which is turned up after jointing. Preliminary tests on this type of test piece gave promising results, and it was found that good results could be obtained with soft timbers as well as with hard. It should be noted that the high results are probably the only reliable ones. BESULTS OF END GRAIN TENSION TEST. (Scotch Glue). Diameter of test piece approximately 0'5 inch. Ultimate Tensile Strength. Lbs./ square inch. Ash .. 3,150 ... 3,180 ... 2,900 Spruce 4,170 ... 4,020 ... 3,380 Oregon Pine ... 2,540 ... 2,350 ... 1,960 Walnut 3,360 ... 2,460 ... 2,360 Mahogany 3,520 ... 3,070 ... 3,260 3,210 ... 3,830 ... 2,190 (Phillipino) 3,360 ... 3.710 ... 2,980 Beech 1,178 ... 4,080 ... 4,410 Birch ... 4,200 ... 3,990 ... CHAPTER XI. FIG. I bo 12. no. 6 FIQ-4. 133 These tests appear to fix the real tensile strength of the glue itself at over 4,000 Ibs. per square inch. It is notable that Spruce tests are equal to almost any of the others. The objection to this test is the difficulty of turning those specimens which, owing to special experi- mental treatment have been weakened, for example by boiling in water. A modification of the direct tension test is the Spandau test, used in Germany, in which the pieces are glued end-grain to end-grain, and the specimen broken by bending. The German specification for glue calls for a load which produces a calculated maximum stress of about 2,000 Ibs. per square inch; at the extreme fibres of the section. This test tends to give a higher value of the stress than the direct tension test, owing probably to the imperfect elastic properties of the glue. COMPARISON BETWEEN SPANDAU AND TENSION TESTS ON JOINTS IN WALNUT. (Propeller Glue). Tension Test. Spaudau Test. Ultimate Strength, Ultimate Strength, Lbs./sq. in. Lbs./sq. in 3,500 ... 4,900 2,600 ... 3,400 3,100 ... 4,100 3,100 ... 4,100 3,000 ... 3,780 2,800 ... 3,000 Average 3,020 3,880 An advantage of the Spaudau test piece is that it can be used for an Impact test, which will probably be desirable as there appear to be decided differences of strength under impact between different glues. There are difficulties that have not yet been satisfac- torily overcome with any test piece which involves glueing on the end-grain with hot glue The individual results of a series generally vary widely owing to the formation of air bubbles which come out of the fibres of the wood on the application of the hot glue. A number of methods have been tried to minimise this trouble, but so far with no very definite success. It is hoped that some method will be discovered which will enable these joints to he made satisfactorily, as the test pieces have many advantages and are cheap and easy to make. Shear Tests. Several types of test pieces have been used by different experimenters; some such as those shown in Figs. t> and 7 are adapted for compression machines, and others such as the A.l.D. type (Fig. 9) for tension machines, but none of them is satisfactory, for in none is the stress on the joint a simple shear. Tests described below show that the shear strength of a glued joint exceeds that of the strongest timber, so that any failure at a lower stress must be due to the tensile stresses which accompany the shear. A test piece that is more likely to give values approximately to the shear strength of the timber is shown in Fig. 8. A few tests have been made on each of these types using the same timber and glue with the following results. SHEAR TESTS. Since the shear strength of the timber is of the order of 2,000 Ibs., it is clear that only type 8 gives any close approximation to shear strength, while in the other types failure is initiated by the tensile stress. Probably the simplest method of carrying out a shear test is to make the joint inclined at about 15 or 20 to the axis of a tension specimen (see Fig. 10). The stresses on the joint are then a shear and a tension. If the angle is between 15 and 20 the stress causing failure will generally be the shear stress. Test pieces of this type using walnut and good glue almost invariably fail through the timber, showing that in shear the adhesion and the shear strength of good glue are both greater than the strength of the timber. This is shown by the following tests: Ultimate Accompanying Shear Stress. Normal Stress. Solid Timber ... 2100 ... 886 2350 ... 1094 2330 ... 1080 Joint 2300 wood failed. 2200 2100 In a special apparatus designed to apply a shear with a minimum of tension (ace Fig. 13 in Chapter X) a joint made with good glue was invariably stronger than Walnut. So far no inferior glues or casein cements have been tested in this manner. In order to obtain actual breakdown of the joint in shear, the test piece, Fig. 11, may be used, in which the line of pull is along the fibres of the wood. Using Walnut and glue the following results were obtained. Ultimate Accompanying Shear Strength. Normal Stress. 2,910 1,080 Joint failed. 3,235 1,175 This type of test piece is somewhat difficult to make satisfactorily, since it involves glueing on the end- grain. The high results, therefore, are the only ones of value. The A.I D. test piece (Fig 9) is the one that is used for inspection purposes in this country. It is intended to measure the shear strength of the glued joint, but the following considerations show that the results bear no relation to the strength of the glue. If the dimen- sions of the joint are varied it is found that doubling the lap increases the strength very little, but doubling the width doubles the strength. If solid wooden test pieces are made of the same shape their ultimate strength is only about half the shear strength of the wood, as is shown by the following tests made by the A.l.D. : Test Piece (Fig. 7). Ult. Strength. Lbs./sq. in. 1,413 joint failed. 1,130 1,095 1,740 Test Piece (Fig. 8). Ult. Strength. Lbs./sq. in. 1,840 wood failed. 2,270 2,330 1,970 Test Piece (Fig. 9). Ult. Strength. Lbs./sq. in. 1,070 wood faile.l 1,130 1,270 Walnut. Ult. Strength. Lbs./sq. in. 1660 1615 1260 1170 1060 860 Mahogany. Ult. Strength. Lbs./sq. in. 1510 1510 1340 1310 1250 810 Ash. Ult. Strength. Lbs./sq. in. 1260 1140 1080 The higher figures in this series of tests are doubtful; if the grain of the wood is not perfectly straight there will be some fibres connecting the two halves which will take part of the load in tension. If a small portion is cut away to avoid this possibility the results 134 are generally below 1,000 Ibs. In two tests in which a semi-circular groove 3/32" deep was cut at each point ' a ' (Fig. 9a) the average stresses were only 700 Ibs. per square inch. In these tests the diminution of section increased the bending at this place and so lowered the result. The fractures in all tests where it was clear that no fibres had been continuous across the section were typical tension failures across the grain. The strength per square inch is affected by the thickness of the pieoes; if the pieces are made -J" thick, wood failures are almost invariably obtained at average stresses of 600 to 800 Ibs. per square inch; even with the standard thickness considerable care is necessary to obtain tests of the order of 1,100 Ibs. per square inch, and if the pieces are not tangential boards it is almost impossible to get figures above 1,000 Ibs. per square inch. A typical set of results of tests on glued joints of this type is that given below: Results obtained in A.I.D. Laboratories. Lbs./sq. in. Lbs./sq. in LbK./sq in. Lbs./sq . in. Lbs./sq. in. g. & w. 1120 g.&w 1090 g * w. 1130 g. & w. 1130 w. .. 1060 g. ... 1150 g. & w 1160 g. & \v. 1100 g. ... 118o g- llfiO g. & w. 1110 " 1180 w, ... 1100 g. ... 1160 g- 1160 g.&w. 1110 w. .. 986 g. 1180 g.&w. 1100 g. 1140 w. ... 1050 g.kw 1090 g. ... 1160 g.&w. 1120 g. & w 1120 g. = true glue break. g. & w. = break partially in glue and partially in wood. w. = break wholly in wood. The uniformity of these results is doubtless due to the fact that the timber is generally weaker than the joint. From an inspection of the joints and from the facts above cited it appears clear that the failure is initiated by a tension stress across the joint and not by a shear stress. From the nature of the joint there is a tension stress at the point 'a' (see Fig. 9A), due to the bending moment. There is also tension due to the inclination of the joint produced by the bending of the specimen under load. The exact magnitude of the total tension at these points is uncertain, but many good joints show undoubted signs of having failed in tension there; the portions of the timber near these points are undamaged and the middle portions only are ruptured, probably by shear after the tensional failure has been initiated. It is clear that this test piece has very serious limitations and is not suitable for general investigation work where it is necessary to know exactly the kind of stress causing failure. Suggested Programme for Testing Adhesives. From the tests described above it appears that the principal cause of failure in cemented joints in timber is a lack of adhesion in tension between the cement and the timber. In all the shear tests that have been made the wood has invariably failed before the joint. As the adhesion will vary with different timbers it will be advisable to test the adhesion in addition to the actual strength of the adhesives. The following scheme of test will probably meet the requirements : Adhesion to wood. Strength of the cement. Tension. Test piece Fig. Test piece Fig. 5 or 3 or 4 or equivalent Spandau. bend test Fig. 12. Shear. Test piece Fig. 10 Test piece Fig. 11. supplemented by Fig. 8. Test pieces for testing the strength of the cement all involve glueing, wholly or partially, on the end-grain and some further experience is required in this operation. The above tests are all simple static tests, but Impact tests may also be required. It is suggested pro- visionally that tho Spandau test piece may be suitable, but further experiments are required in this matter, and also on the resistance to alternating stress, which may be of importance. The procedure in the testing of any adhesive will be somewhat as follows : -Preliminary adhesion tests should first be made in order to determine the proper technique in the application of the adhesive. The follow ing points will require examination in greater and less detail : (1) Most suitable proportion of adhesive and water. (2) Effect of temperature. (3) Effect of pressure. (4) Effect of kind of surface. (5) Effect of time during which adhesive is kept liquid. (6) Effect of alternate heating and cooling (for glues applied hot). Walnut may conveniently be used for the first test pieces. The best method of use with walnut having been ascertained, other woods should be tried, such as mahogany, spruce, ash, birch, beech, and others that are likely to be used. The best method of using the adhesive having been found, and its strength measured, the next series of tests should be made on joints which have been subjected to high and low temperature, and to moist and dry atmospheres, so as to ascertain how the adhesive will stand extreme variations of climate. 2. GLUE : ITS MANUFACTURE AND APPLICATION. BY LIEUT. KERNOT. (C.I.M. 707). INTRODUCTION. The importance of glue in a large number of the timber-using trades has always been evident. Owing, however, to the fact that the wood-and-glue structure of the past has not been required, as a rule, to with- stand severe mechanical stresses under bad conditions of temperature and humidity, neither the manufacture of glue nor its technical application has received very careful scientific attention. As a general rule the manu- facture of glue has been carried on by rule and thumb methods handed down from generation to generation ; while its workshop application by the wood-worker has been equally devoid of established (or understood) rhyme or reason. This is a particularly unfortunate situation just now when wood-and-glue structures are required to withstand severe and ever-increasing stresses and when, as was pointed out by Lieut.-Colonel O'Gorman, in " Aeronautics," a glue of known composi- tion and unfailingly uniform in strength would be a most powerful aid to rapid production, and would effect at the same time an enormous economy in timber. The reason for this lack of exact knowledge is to be found in the small amount of attention so far paid by organic chemists to products of animal origin; and in this relatively neglected class, the colloids, gelatines (glues) and allied products must be placed. Gelatine is that substance which is extracted by boiling water from animal (and some vegetable) tissues and which has the property of forming jelly or " gela- tinizing " when in cool solution. The gelatines and glues 135 of commerce certainly vary in general appearance, but their source of origin is the same and, within limits, they are similar substances. They differ only in the relative purity of edible gelatine as compared with the relative impurity of industrial glue. In brief, glue can be described broadly as an "impure gelatine"; a fact which directs attention to the need to confine the use of the word " glue " to the product under discussion and to avoid the popular error which applies the word to any substance exhibiting adhesive properties whether that substance is allied to gelatine or not. Until exact knowledge of the composition and struc- ture of gelatine is obtained it would seem to be im- possible to explain why glue, in contrast with pur'? gelatine which is a poor adhesive, has such properties as adhesiveness and tenacity, and why it succeeds in sticking together such dissimilar materials as wood and glass. On close examination under the microscope both gela- tine and glue appear to be perfectly homogeneous; but this would appear to be a deceptive observation due to the very slight difference between the refractive index of the dispersed particles and that of the dispersion medium. A certain amount of work has already been done by Butschli, Stoffel and others* on the structure of gelatine, and preliminary experiments carried out at the A.I.D. Laboratories confirm their results and show that a similar structure is exhibited by any good quality glue. FIG. 13. FIG. 14. Fig. 13 shows the structure of a gelatine which has been allowed to set slowly, while Fig. 14 shows the structure of a gelatine which has been cooled rapidly. (Butschli.) The problem has been tackled in the following way. Gelatine does not form a true solution with water, but forms what in physical chemistry is termed a " hydrosol." If by addition of alcohol or benzol this hydrosol is trans- formed into an " alcoholsol " or a " benzolsol," the conditions of visibility are greatly improved (owing to the changed refractive index of the dispersion medium), and it is possible to' see (Figs. 13 and 14) the varying structure assumed by different gelatines under dis- similar conditions. One other matter may he referred to in these opening notes as a point on which further information would doubtless prove of value, namely, the effect of various reagents in rendering glue insoluble. Glue in aqueous solution is precipitated by formaldehyde as a substance which, on drying, becomes a white powder in- soluble in water. If only very small quantities of formaldehyde are added, no precipitation takes place and the solution will set to a jelly when cold ; * [(1) Tint, uber Mikros. Schaume, Leipzig, 1892 ; (2) Unters. u. Struk. Leipzig, Ic93 ; (3) Verh. d. nat. med. Vereing, Heidelberg, 1894 ; (4) Uber. der Bau quell. Korp., Gottingen, 1896 ; (5) U.S.A. Journal Phys. Chem, 4. 1900 ; and (6) Diff, versch. in Gallerten Dies Zurich, 1908]. but if the quantity of added formaldehyde exceeds 0-02 %, the dried jelly will not re-dissolve in water. In similar fashion, if a glue solution is treated in the dark with a soluble chromate or bichromate, such as potassium bichromate, no change takes place; but if the mixture is exposed to sunlight the glue is rendered insoluble. THE MANUFACTURE OF GLUE. 1. Raw Materials. Industrial progress depends largely upon the use made of the waste products of manufac- ture, and this is very clearly illustrated in the glue in- dustry, where not only do the raw materials consist of the waste products from other industries, but the waste substances from glue making are themselves also valu- able, being chiefly used as manures or fertilizers. Commercially, glues are known as bone glues or hide glues, according to the raw materials used. When pre- pared from skins, glutin is the main constituent, while bone tissues yield a product containing a large portioL of chondrin. As an adhesive, glues from hides are pre- ferred, owing to the greater binding power of glutin. The bones are chiefly used in the raw condition, and consist of the heads, ribs, shoulder blades, etc., of domestic animals. Hide glue is made from tannery waste, such as skin trimmings, tendons, etc. 2. Bone Glue (Normal Process) . The bones are first sorted and separated from pieces of iron, wood or hoofs that may be present. They are then passed through a mill, which cracks them slightly, and thence to the extracting plant for the removal of fat. This may be performed in several ways : (a) by open boiling, (b) by digestion with steam at 40 Ibs. pressure, or (c) by means of solvents. The last method is the one generally used, as it is the most efficacious. Boiling leaves a good deal of fat in the bones, and the second method, whilst removing more fat, also removes a large portion of the glue-yielding substances. Various solvents are used in the extraction of fat, crude benzol or petroleum being the most common. Carbon bisulphide, although an excellent grease solvent, is too dangerous on account of its volatility and in- flammability. The cleansed bones are then transferred to the glue extraction apparatus, when they are placed in a vertical boiler and treated with steam at a pressure of 15 Ibs. for about two hours, after which the pressure is reduced to 5 Ibs. By thus lowering the pressure, the glue within the bone is brought to the surface from which it is washed down by a spray of water from a coil fixed in the dome of the boiler. This process is repeated two or three times until a sample of the liquid shows a strength of 20% dry glue. The glue liquor is then run into shallow vats and clarified by heating with the addition of substances, such as alum, oxalic acid, albumin or blood, the first named being the. best. After boiling for about ten minutes, the liquor is allowed to settle, the heavier mineral and organic matter falling to the bottom and the lighter forming n scum on the surface. These im- purities are removed by filtering through fine wire gauze or calico, the liquor being kept hot by a steam coil. After being clarified, the liquors are concentrated, bleached and dried as is described in the following pages. 3. Bone Glue (Acid Process). Glue can also be ob- tained from bones by the acid process in which the organic matter may be separated from the earthy phos 136 phates and carbonates by steeping the bones in dilute hydrochloric acid. The mineral salts are dissolved and a soft, flexible substance is left, which after being washed free of acid, is treated with steam in a closed iron digester. A mixture of fat and glue is thus ob- tained, which is run into settling tanks, and after the separation of the former, the glue is clarified and treated in the same way as that obtained by the process already described. 4. Hide Glues. The quality of the hide glues varies considerably with the raw materials used, and experience shows that clippings from the hides, also the tail, head and ear pieces, which are useless for leather making are, together with the residues from the manufacture of kid gloves, rabbit skins, scraps of parchment, etc., very valuable raw material for glue. The hides and pieces are first steeped in milk of lime in wooden vats or cement pits, and frequently stirred with long forks, in order that every part may be acted upon. The treatment takes from two to three weeks. When the skins are firm and free from any greasy feeling, the milk of lime is run off and the vats filled with water containing a small amount of hydrochloric acid. The hides are then thoroughly washed with clean cold water, until free from any trace of acid, and hung up in bundles to dry. Sometimes the excess of water is squeezed out in a press. The glue is extracted by boiling slowly with water either in open or closed vessels. As the boiling proceeds the raw material decreases in bulk, owing to the ex traction of its soluble constituents by the hot water, and more raw material is added until a sample of th-.> liquid shows it to be of the right strength. When the desired density is obtained, the liquor is clarified in the same way as described for the bone glue. Scotch glues are manufactured by placing the hides in a loosely woven sack. This is lowered into a circular cauldron containing water which is gradually brought to the boil. The boiling is continued (with fresh additions of raw material) until the liquor gives a firm jelly on cooling. The liquor is then run off, concentrated and allowed to dry in the open air. 5. Concentration and Drying. After being clarified, the liquor obtained by any of the processes mentioned above, are concentrated in a vacuum pan heated by steam. Solutions of glue are very susceptible to change by prolonged boiling at atmospheric pressure, and if the glue liquors were evaporated in an open vessel at 100 C. the product would be of inferior quality and dark in colour. Hence it is necessary to evaporate at a lower temperature, which is achieved by lowering the pressure. Multiple evaporators are used, in which the steam from one pan is used to heat the next one, and so on. The concentrated gelatinous liquor is allowed to cool and when it sets to a jelly, can be cut up into sheets of different thickness. The processes of cooling and drying the glue require great care, as frequently, through causes not always understood, whole batches of glue are spoilt. The liquor must be cooled quickly, as the liquifying bacteria bacillus subtilis will rapidly injure the glue if the cooling process is prolonged. Refriger- ating machines are therefore used so as to obtain a temperature of about C. Below this temperature the jelly becomes too hard and difficult to cut. The solidified jelly is then cut up into sheets and the sheets dried by spreading them on metallic, netting, preferably on heavily galvanized iron wire. The temperature of the drying rooms (which should be well ventilated) should not exceed 21 C., and as far as possible access of dust ought to be prevented. Dry air is suggested in the German patent 232,715. 6. Bleaching. The colour of the glue obtained by the processes mentioned varies very much, owing to the residues of coagulated blood or to the formation, during the process of concentration of the gelatinous liquor, of a brown substance, allied to casein. To obtain greater homogeneity, some of the best makers resort to artificial bleaching before final concen- tration and drying. Many carpenters have a great prejudice against light-coloured glues. A lighter article, they say, has either b^en made of selected material and therefore is expensive, or it has been bleached and has probably been reduced in strength. But such is not necessarily the case. Bleaching, when properly carried out, not only reduces the colour, but increases the keep- ing qualities of a glue, and sometimes improves its strength. For bleaching, many substances have been used, such as Sulphurous Acid or Hydrosulphites, Chlorine of Hypochlorites, Ozone, also the Peroxides of Potassium, of Sodium and of the Alkaline Earths. An important consideration in the use of different bleaching agents is the effect that they may have upon the viscosity of the gelatine. It is a well-known fact that the glues of high viscosity possess the greatest strength. From the recent investigations of Wo. Pauli (Pfluger's Arch. 71, 1898 and following years) it appears that Sulphates increase the viscosity and strength of gelatine in all concentrations. On the other hand Chlorides and Nitrates lower the viscosity of gelatine and in consequence weaken it. Sulphurous Acid and the Hydrosulphites (Chemiker Zeitung Rep. 1,906 page 418) were the first bleaching agents known, and they are still vised more widely than any other. Besides reducing the colour of the glue, they combine with the mineral matter contained in the glue forming sulphates which, as explained above, in- crease the viscosity and strength of the. glue. Chlorine and hypochlorites react with the nitroge- nous organic matter in glue, generating chloramine hydrazine, and other bodies of high germicidal power, and they, therefore, increase the keeping qualities of the glue. Chlorides however, which are also formed, lower the viscosity of the glue and weaken it. For this reason, therefore, chlorine is inferior to sulphurous acid as n bleaching agent. The action of ozone and of peroxides (/.c.) cannot always be controlled and their use has been abandoned. The peroxides form nitrates by acting on the nitrogenous organic bodies contained in glue, and by lowering the viscosity of the glue, weaken it. 7. Liquid Glues. Bone and hide glues constitute the principal ingredient of practically all the so-called "liquid glues " of commerce. It is to be noted, however, that the majority of these glues are not liquid, but jellies at ordinary temperatures; so that the term "jelly glue" describes them with greater accuracy. In general it may be stated that these jelly glues are simply ordinary bone or hide glues to which the requisite amount of water plus small quantities of other ingre- dients have been added. Carbolic acid in quantities between 5 and 10% by weight is frequently present in these jelly glues. Carbolic acid certainly delays the setting point of a glue solution (thus performing a use- ful function from a " liquid " glue standpoint) and To face page 137.] [CHAPTBE XI. JSK.'f-r- ni { "i - * . ' -^ .' ; ^ FIG 15. Showino- good penetration of glue in a sound joint between two pieces of mahogany. The thick r parts of the joint are occasioned by the " toothing " of the wood before glueing. FIG. 16. Section of a poor glued joint cut from a propeller that failed in use. The large cavity is one of many large air bubbles in the joint. The woods joined together are walnut and mahogany. 137 undoubtedly acts as a preservative so long as it remains in the glue. Carbolic acid, however, is vola- tile and if the glue jelly is exposed to the air, or when the glue jelly dries (as in a joint), the carbolic acid will disappear and its preservative effect will be lost. It has been observed that glue joints, prepared with a liquid glue originally containing carbolic acid, have not escaped attack by liquifying bacteria The only point in favour of jelly glues is their greater convenience in use, i.e., it is unnecessary to go through the soaking process, while the jelly glue, when warmed to a liquid, does not (owing to the presence of the car- bolic acid or other similar ingredient) tend to " chill " or set quickly while making a joint. On the other hand the liquid or jelly glue is yet to be found which gives superior results to those normally obtained by means of ordinary cake glues; and it is to be noted that liquid or jelly glues frequently give very poor results owing to improper or careless manufacture or to the inferior quality of the ingredients employed. It is self-evident that the preparation of a liquid or jelly glue affords opportunity to use a grade of cake glue which would be difficult to dispose of by other means PREPARATION AND USE OF GLUE. 1. Preparation of Glue for Use The best method to prepare glue for use is to break thel cake glue into small pieces (about the size of peas) and to soak them in water at room temperature for 24 hours. The glue in this time will have absorbed enough water to form a solution when heated. Hence, any excess of the soaking water should be thrown away, and the swollen pieces of glue should be transferred to, and melted in, the familiar water-jacketted glue pot. The heating or melting process should be conducted as quickly as possible without boiling the water in the glue pot jacket furiously, and the contents of the glue pot should be raised to a temperature between 60 0. and 80 C. (140 F. and 175 F.) when the melted glue should be ready for immediate use. The bigger the " job " the warmer should be the glue (within the above limits). If glue is heated more than about 6 hours, it becomes steadily weaker. Prolonged heading may cause the formation of substances such as Mucin (a non-adhe- sive), although such products may also be present in a glue owing to its method of manufacture. Hide glues, for example, have this drawback when the hides have not been carefully washed after liming. Stale glue should not be used, and fresh glue should not be added to previously heated glue. Only sufficient glue for the work required should be prepared, and any surplus glue should be thrown away. 2. Method of Using Glue. The glue solution, once it has become hot, should not be allowed to cool, as remelted glue has not the same tenacity as a freshlv prepared solution. Whenever possible the wood should be warmed slightly to avoid chilling the glue. On the other hand the wood must not be too hot, as very hot wood rapidly absorbs the water in the glue, causing the glue to dry too quickly, with the result that a weak- joint is obtained. The faces of the wood to be joined together should be made perfectly " true," after which, if possible, they should be lightly " toothed " with a fine toothing plane The room in which the erlueing is done should be at a temperature not below 20 C. and free from draughts The joint, when made, should be clamped up for at 27264 least 12 hours, after which the clamps may be removed, but the joint must be allowed to set at least another 24 hours before it is subjected to any stress. It is generally recognised that the quality of a glue joint is very greatly dependent on the experience of the operator; but it may be assumed that the more a glue penetrates into the pores of the wood the more efficient is the joint, always providing that sufficient glue remains between the surfaces of the wood in order to form the joint. Fig. 15 shows the penetration of a good glue. Good penetration is always obtained with glues which do not " set " too rapidly. In glueing, great care should be taken to avoid the formation of air bubbles, which are the source of con siderable trouble, and frequently cause the failure of a joint. Fig. 16 shows one of many air bubbles found in one of the joints of a propeller which failed in service. Air bubbles are due sometimes to bad glut; brush manipulation, and sometimes to frothing or foaming, objectionable properties which are peculiar to some kinds of glue. PROPERTIES OF GLUE. It is not an easy matter to state what definite pro- perties a good glue should possess ; but in a general way it may be said that the qualities required are (1) Adhesiveness; (2) Tenacity; (3) Elasticity; (4) Covering Power and (5) Keeping Quality. 1. Adhesiveness. May be defined as the power to stick together two similar or dissimilar materials. 2. Tenacity. May be defined as the power to resist the disruptive effect of any stress. 3. Elasticity has its usual meaning, namely the power to stretch slightly without fracture. In this connection the moisture-content of the glue is an important factor Glues which are too dry are inclined to be brittle. They will resist an enormous steady stress, but will break under very small stress if applied with a jerk. 4. " Covering Power." By careful experimental work, it has been ascertained that the covering power can be determined by estimating the water-absorption, the tenacity of the jelly, and the viscosity of a solution of known strength. In a broad sense it may be said that a glue will give a joint of high quality and strength if the tenacity of the jelly is high and if the viscosity of the solution is high ; but in view of the fact that no two glues are of precisely the same composition or exhibit absolutely similar chemical and physical properties, it is impossible to regard these deductions as anything more than a general guide to anyone attempting to draw conclusions from the tests in question. 5. The " Keeping Quality " is another important property which, unfortunately, is not always appre- ciated. A glue which on keeping turns sour or mouldy will be liable to exhibit similar faults in the joint in damp weather. A glue which has good keeping qualities will improve in strength by storage, as every practical joiner knows. ANALYSIS AND TESTING OF GLUE. Many methods have been proposed for the analysis and testing of glue ; but, owing to the obscure character of the material itself, many of these tests give results that are not comparable one with another. S 138 The following chemical methods may be employed: 1. Estimation of Water. This is an important deter- mination to make in order to obtain comparable results from the same tests applied to different samples of glue. One or two grammes of powdered glue are dried to constant weight at 110 C. Glues may contain from 5 to 18% of water; but a good glue should contain not less than 10%. A lower moisture content indicates that the glue has been spoiled by over-drying and that it will be brittle. In this connection it is to be noted that glue may be over-dried and spoiled in much the same way as timber may be over-dried and spoiled. Subsequent dilution of the glue with water does not improve its quality any more than an over-dried wood has its elasticity restored by subsequent absorption of moisture. 2. Mineral Matter. The mineral matter content may be estimated by very careful incineration of two or three grammes of the powdered glue at as low a tem- perature as possible. The percentage of ash may vary from 1'5 to 3%. The constitution of the ash may afford a clue concerning the origin of the glue. A large percentage of ash indicates that the glue is of impure origin (i.e., made of dirty bones or from hides that have been washed insufficintly after liming), or that mineral substances- have been added as adulterants. The estimation of chlorides in the ash is of import ance, as an excess of chlorides will indicate that chlorine or hypochlorites have been used for bleaching purposes and it has already been shown that the presence of chlorides in excess tends to lower the viscosity of a glue solution and to impair its utility. 3. Absorption of Water. The quantity of water which a glue will absorb varies within wide limits, and affords an indication of the quality of the glue under examination. The absorption value is estimated by soaking a known weight of glue in water for 24 hours and reweighing. According to Schatterman (1845) and Schlossmann Papier Zeitung XVII., 2,484), the quantity in parts by weight of water absorbed by one part by weight of glue after soaking in water for 24 hours is as follows : Glue Water absorbed by one (and origin) Part by weight of Glue in 24 hours. Bird (Skin) Pish (Offal) Veal (Hide) Horse (Hide) Sheep (Hide) Ox (Hide) English Glue (not defined) 23 ... 11 4 ... 3 3 3 ... 2-5 Tendons, Hoofs, etc. ... ... ... 2'5 4. Acidity of Glue The acidity of a glue may be determined by dissolving 30 grammes of glue in 100 c.cs. of water, and then passing a current of steam through the solution (steam distillation). This serves to distil over the volatile acids which are collected and subse- quently estimated in a flask containing a standard solution of alkali. If a close examination is required the distillate should be examined for sulphurous, acetic and butyric acids. The presence of the latter acids points to a souring of the glue during manufacture and subsequent deterioration. In this connection, however, it is to be noted that glues may actually be alkaline owing to the presence of lime due to improper initial washing of the materials from which the product was made. Alkalinity is a dangerous condition, as bacteria such as the Bacillus Subtilis which cause objectionable changes to take place in glue, develop with ease in a slightly alkaline medium, although they will not develop in a strongly (caustically) alkaline medium. In point of fact, very little work seems to have been done on the putrefaction of glue either by mould or bacteria, but work in this direction in now in hand at the A.I.D. Laboratories. The experiments in question are being made in order to bring to light a suitable pre- servative for glue, the presence of which shall not impair the strength of the glue for practical purposes. With this in view, an attempt is being made to grow mould on glue jellies containing various percentages of preservatives, the amount of preservatives used being the minimum percentage stated by the Bulletin of the Bureau of Plant Industry, U.S.A., No 227, 1915, to be necessary to prevent the formation of mould. In this connection the following results have been obtained by maintaining the glue jellies plus their preservatives at a constant temperature of 25 C. for the times stated Toxic Substance. Amount Bemarks. per cent. 1. No preservative Mould after 3 days. 2. Sodium Chloride 8-0 Glue liquid in 3 days.* 3. Sodium Fluoride 2'0 No mould in 12 days. 4. Copper Sulphate 4'0 No mould in 12 days. 5. Mercuric Chloride 0-08 No mould in 12 days. 6. Zinc Chloride 0'80 Mould after 12 days. 7. Zinc Sulphate 0'80 Mould after 7 days. 8. Cresol 0-125 No mould after 12 days. 9. Phenol 0-125 No mould after 12 days. 10. Thymol 0-01 Mould after 3 days. The' above work has been little more than commenced so that it would be unwise to offer any comment until further evidence is available. 5. Determination of Gelatine. The following method for estimating the gelatine-content of glues is in general use on the Continent, and was suggested by Grager (Dingler's Polytechnick Journal, 126). A quantity of glue, in which the percentage of moisture has been determined, is dissolved in warm water, and when cool, the gelatine is precipated by a 1 in 20 solution of tanniu. The latter is added until no further precipitation is pro- duced. The precipitate is collected on a filter, washed and dried to constant weight at 120 C. This dried precipitate contains for every 100 parts, 43 parts of gelatine and 57 parts of tannin. The quantity of gelatine contained in ordinary glues varies within 70 and 80%; but in the present state of our knowledge of glue it would not appear to be possible to state precisely the part played by the gelatine-content in a glue. Pure gelatine is a non-adhesive; but whether the gelatine in glue remains a non-adhesive in the presence of the albumoses, peptones, sugars, and other ill-defined bodies found in commercial glues is a prob- lem which remains to be solved. An indirect method for estimating the gelatine-content of a glue is provided by a determination of the nitrogen- content by the Kjeldhal method, and by multiplying the total nitrogen obtained by the factor 5'56. This factor Pee p. 138, Section 6, 139 has been derived from the following experimental figures obtained by four different observers : Observer Nitrogen, Factor. per cent. Allen 17-9 ... 5'59 Chittendeu 17-97 ... 5'57 Schroeder and Paessler ... 18'1 ... 5'52 Bideal 18-0 ... 5'57 Estimating the gelatine by determining the nitrogen content by the Kjeldhal method, is, perhaps, sufficiently accurate for commercial purposes, but it is to be remem- bered that by this method the nitrogen-content of the proteids and other nitrogeneous products (always pre- sent in glue) is included in the total figure obtained. 6. Estimation of Fat. The presence of fat in glue is very objectionable, as it causes " foaming." In pro- perly manufactured glue, fat should be completely absent. Fat is estimated by dissolving 20 grammes of broken glue in 50 c.cs. of water to which 5 c.cs. of hydro- chloric acid have been added. The acid is added in order to decompose all soapy material that may he present, such as lime soap, due to careless initial wash- ing of the raw materials. When the glue has dissolved, a quantity of calcium sulphate is added to absorb all the liquid. The mass thus obtained is thoroughly dried and powdered, and the powder extracted in a Soxhlet apparatus by some suitable solvent, such as benzol or carbon bisulphide. The solvent is subsequently driven off and the residue dried and weighed. The chemical tests detailed above are of considerable interest and importance to glue manufacturers and are also of no little importance to glue users. The latter, how- ever, are principally interested in the strength of a glue i.e., its adhesiveness, tenacity, etc., and to estimate those factors physical tests must be resorted to. Visual examination of a cake of glue is not particularly instructive, but in a general way it may be remarked that good glue (free, for example, from chlorides) will not become damp in storage under reasonable conditions, and that the cake should bend reasonably without shattering violently into a large number of small pieces. The following physical tests may be applied : 1. Consistency of the Jelly This test was sug- gested by Lipowitz in 1861, and has been extensively adopted for commercial purposes. Five grammes of glue are soaked in water at room temperature and then dissolved in enough warm water (70 C.) to make the total volume 50 c.cs. when cold. The solution is allowed to stand in cylinders half-an-inch in internal diameter for 12 hours at 18 C. The consistency value of the column of jelly thus obtained is then determined by loading a small pointed plunger (carrying at its upper end a funnel, into which lead shot is gradually added) until the plunger is just persuaded to pass right through the jelly, from the top surface of the jelly to the bottom of the cylinder. The weight of shot necessary to effect this gives the Lipowitz number. It is necessary that this test should be carried out at a definite temperature, and that the other conditions should be kept constant. As was pointed out by Heinze the results obtain under similar conditions of temperature and size of apparatus are reasonablv comparable and do not vary more than 10%. 27264 At the A.I.D. Laboratories a slight modification of the pointed steel plunger adopted by Lipowitz has been made. A plunger is employed consisting of a glass, round-ended rod, and the results obtained have been found to be rather more consistent and satisfactory s a consequence. In passing, it is of interest to note that just as sulphates increase the viscosity of a glue solution, so also does the presence of sulphates increase the consistency of the jelly when examined by the Lipowitz method; while in like manner the presence of chlorides and nitrates diminishes this consistency. 2. Viscosity of Glue solutions. In conjunction wilh the jelly consistency test dealt with above, a determina- tion of the viscosity of a standard solution of glue is re- garded by many glue manufacturers as a test capable of yielding important and reliable results. The test was first put forward by Julius Fels (Chem. Zeitung, 1898, 1899), and in making the determination it is first of all necessary to know the percentage of water con- tained in the glue in order to prepare a solution con- taining a known percentage of dry glue. Fels advises the use of a 15% solution of glue (15 grammes of glue in 100 c.cs. of solution) and recommends the determination to be carried out at 30 C. comparing the viscosity of the glue solution with that of pure water at the same temperature. Any form of viscometer may be used. Fels used the modified Engler viscometer, but other workers have used the apparatus described by Slotte and modified first by Rideal (Journal Society Chemical Industry, 1891) and then by Scarpa (Gazetta Chimica Italiana, 1910). Fels, using a 15% glue solution, found that the viscosity varied from 1'5 to 2'2. Other workers, using different solutions, obtained different figures; but it was generally agreed by all these workers that the viscosity of a glue solution and its adhesive quality were related factors. In measuring the viscosity of a standard glue solu- tion not only must attention be paid to the temperature, but it is important that the viscosity should be deter- mined immediately the solution has been prepared, be* cause as the solution ages it tends to increase in viscosity and may indeed become gelatinous in con- sistency. 3. Plaster-rod Test. A method of glue testing, sug- gested by Kamarsch and modified by Weidenbusch (1859), consists in breaking a small rod made with glue and Plaster of Paris. Small rods of Plaster are satu- rated with glue solutions of known strength and' are then thoroughly dried. They are then placed horizon- tally on two supports and loaded in the centre, the weight required to break them being the Kamarsch figure. 3. CASKIN CEMENT. (This section is based on Lieut. Rernot's reports).. The adhesive properties oJt casein are so well known and appreciated that it is superfluous to deal at length with them. It is, however, necessary to point out the advantages of such cements over ordinary gelatine glues, as their use is daily increasing. Their adhesive pro perties are nearly as good as those of the best gelatine glues ; they are very easily rendered fluid and keep so for a comparatively long time ; they can be used cold, are inodorous, are not inclined to foam and are always ready for immediate use as (hey do not require soaking or heating. Once dry they are not affected by moisture and are very resistant to bacterial action. S 2 140 The raw material for casein cements is the fat-free casein obtained from curdled milk and brought into a liquid or pasty form with the aid of chemical agents, such as alkali hydroxides, alkaline earths, their salts, etc., and water. During the course of experiments in the preparation of casein cements it has been found that commercial casein is by no means a uniform material, and that the variation between lots from different manufacturers, or even between different ship- ments from the same makers, is sufficient to necessitate an alteration in the formula in order to produce a good cement. For this reason the success of a formula for casein cement depends upon the uniformity of the raw material. The following are the agents used for precipitating the casein from skimmed milk. (&) Rennet. (b) Hydrochloric acid. (c) Sulphuric acid. (d) Lactic acid (the self-souring process). (e) Sulphuric acid at 90 C. With the exception of the first, all the other agents are in use for the preparation of the material on a large scale. The cause of the variation in the properties of commercial casein is not exactly known, and it is to be hoped that a practical method will be found, capable of being put easily in operation, which will produce a uniform product. In order to obtain a good cement it is necessary that the casein should be as far as possible free from fat and from acid, and be finely ground. An excess of fat, while not affecting the waterproofness of the cements, causes a very appreciable decrease in strength. An excess of acidity not only affects the strength of the cement, and its waterproofness, but also causes the formation of lumps when the cement is mixed with water for use. The fineness of the casein has a great influence on the quality of the cement; if the casein is too coarse it will dissolve with difficulty and produce a lumpy mixture instead of a smooth paste. Casein cements, as found on the market, are generally composed of casein, an alkali, and slaked lime. To these several substances are added, some of which are detrimental, some indifferent, while others are capable of improving the adhesiveness and the keeping qualities. In the first two classes may be mentioned such substance as sand, resins, camphor or those substances which are added to delay the setting In the third class are all those salts which, in addition to helping the casein to dissolve and therefore to form a more uniform pro- duct, also have a preserving action and improve the keeping qualities of the mixture. Substances that come in this class are the alkaline phosphates, berates, sodium chloride (S. Ryd.-Zeit. fur Elektrochemie, 1913-23, 19), potassium or ammonium oxalate, sodium fluoride, etc. The action of sodium fluoride on casein is very re- markable. Besides being one of the best-known sol- vents for casein, this salt acts as a very powerful germi- cide. Malenkovic found that it is more toxic than zinc chloride and it has only a slight corrosive action on metals. In 1917, at the suggestion of Capt. G. W. C. Kaye, B.A.F., investigations were carried out at the Labora- tories of the Aeronautical Inspection Department on casein cements, and a formula was suggested which gave excellent results when the mixture was freshly prepared (nee Air Board Specification V.2). But later it was found that the keeping qualities of this cement were poor and that it grew weaker with time, especially in tropical climates. This was due to the excess of caustic soda, and therefore experiments were undertaken by the Material Section in order to find a formula which, with good adhesive qualities, should also have good keeping qualities. As a result of a large number of ex- periments the following formula is suggested : Casein 78% Dry Sodium Carbonate ... ... ... 4'5% Freshly slaked Lime 12'5% Sodium Fluoride ... ... ... ... 4% Sodium Arsenate ... ... ... 1% All the ingredients are to puss through a 90 mesh per inch sieve. For use this cement may be mixed with water or with weak solution (5%) of good gelatine glue. It has been found that this cement gives results equal to those obtained with the best propeller glues, while it has the advantage of offering a very remarkable re sistance to the action of water, moulds and bacteria. Analysis of Commercial Casein. It has already been mentioned that commercial casein is by no means a uniform product, but unfortunately no method of analysis for controlling the uniformity of the material is known. It is hoped that the following notes will be of use, and that they may serve as a basis for a specification to replace the present insufficient one. Colour and Odour. Good commercial casein should be white or exhibit at the most a slight yellow colour, and should be free from rancid or cheese-like odour and from foreign matters. Moisture. For the estimation of moisture, 5 grs. of the sample should be dried at a temperature of 100 to 105 C. until constant in weight ; this requires about two hours heating, and the weighing should be made at frequent intervals to avoid the oxidation of casein by prolonged henting. Ash. For the estimation of ash 5 to 10 grs. of casein should be ignited in a silica basin. In some cases the ignition proceeds with difficulty; to overcome this W. Hopfner and H. Burmeister (Chem. Zeitung, 1912, 36, 1,053) suggest that the mass should be extracted with water. This does not seem to be advisable owing to the probable composition of the ash, and it would be better to oxidize the carbon with nitric acid or with ammonium nitrate and sulphuric acid. For compara- tive and accurate estimation the latter method is preferable. Acidity. The acidity of the sample should be esti- mated as follows. Ten grs. of casein should be shaken with 100 c.cs. of water, filtered, and 50 c.cs. of the filtrate N titrated with- KOH using phenolphthalein as indicator. Fat. The estimation of fat is by far the most im- portant of all, and it is most difficult to obtain accurate and concordant results. When the casein is precipitated from the milk each minute particle of the precipitate encloses a small amount of fat. In order to extract all the fat it is necessary to dissolve the curd and liberate the fat, which otherwise will not be completely dis- solved by the ordinary solvents. The extraction should be carried out as usual in a Soxhlet apparatus by a mixture of ether and petroleum ether, but in order to '141 obtain all the fat the ordinary method should be modified as follows : A weighed amount of casein should be mixed with water spread on a filter free from fat, which should then be placed in an extraction cone and without any pre- liminary drying extracted in the Soxhlet apparatus with 1% acetic acid for 2 hours. This removes most of the protein matter. The cone should then be washed with hot water to remove the acetic acid, dried in an air oven at 100 C. for 6 or 8 hours, and extracted with the ether and petroleum ether in the Usual way. Another modifi- cation is suggested by W Hopfner and H. Burmeister (I.e.), namely, to grind the casein after mixing it with sand, and then extracting in the usual way. Of the two methods the first one, modified as shown above, is very reliable as max be gathered from the following table (K. T. Mohan, Sclent. Amer., 1915). Ordinary Calculated Fat. Ether Ext. Modified Method. Per cent. Per cent. Per cent. 7-25 ... 6-85 ... 7-20 7-98 ... 7-46 ... 8-00 8-74 ... 8-13 ... 8-75 9-14 ... 8-33 ... 9-20 9-70 ... 8-67 ... 9-65 10-35 ... 9-13 ... 10-40 Average 8 - 86. Average. 8-09. Average 8-86. These measurements were carried out with evaporated milk, but experiments have proved that the method is equally reliable for precipitated casein. Estimation of the Quantity of Casein in a Commercial Sample. (p) By estimation of total Nitrogen. The total nitrogen should be estimated by the Kjeldahl method. To obtain from the amount of nitrogen the quantity of casein ; the percentage of nitrogen is multiplied by a factor which varies according to the different" authors, from 6-M4 to 6'99. Hopfner and Burmeister (I.e.) find that the factor 6' 61 probably is most correct, as the quantity of casein calculated by its use generally approximates to 100%. From the writer's experience it was found that this factor always gave high results. An important series of experiments was carried out by H. Droop-Richmond (The Analyst, 1908, 179) with the object of determining whether the factor G'37, used for " milk proteins," was applicable to both casein and albumin, when the nitro- gen determination was made by the Kjeldahl method. The starting point was the fact observed that whilst preparations of laetoalbumiu gave results for nitrogen very closely agreeing with those required by the above factor (i.e., 15'7% N.), several preparations of casein gave results seriously below this, the mini- mum being 14'2%. This is due to the impurities that are always to be found in the precipitated casein. As the casein used in H. Droop-Richmond's experiments (I.e.) liad been extracted with ether, it was not thought necessary to look for fat as an impurity, but, as it has been said before, if the ordinary method of extraction is used, all the fat is not extracted and this is a possible explanation of the low results. The same author, with Miller (Analyst, 1906, 31-321) found that milk proteins take small quantities of aldehyde from the ether. By estimating the aldehyde figure of casein which has been treated and comparing it with that of casein which has not been treated, it appears that a small correction should be made for this. It is probably due to all these causes that there is such a disagreement between the different experimenters as to the factor that should be used. In a large number of experiments, in which all these sources of error were taken into consideration, results were obtained corresponding always to the factor 0-37 (N% in casein = 15-75). (6) By the Ferric Alum Method. This method is based on the precipitation of casein with an excess of a solution of ferric alum of known standard; estimating the unused iron in the filtrate and then calculating the amount of iron used by the casein in the act of precipitation. The standard iron solution contains 48'2224 grs. of Fe (NHJ (SOJ 2 -12H 2 Otothelitre, and it is standard- ised by the usual method of volumetric ferric estimation ; viz., adding potassium iodide and hydrochloric acid N and titrating the liberated iodide with JQ sodium thio- sulphate. One cc. of this standard solution liberates 12-692 mmgrs. of iodine which in turn requires 1 cc.^W sodium thiosulphate. CHAPTER XII. BUBBER SHOCK-ABSORBER CORD. The function of the shock-absorber of an aeroplane is to provide a means whereby the energy due to the vertical component of the velocity of the aeroplane on landing may be obsorbed and then dissipated. If the velocity just before touching the ground is uniform the work done by gravity is balanced by the work of the drag W forces, and the energy to be absorbed is g (V sin n)' 1 where W= the weight, V= the velocity, and a = the angle which the directior. of motion makes with the horizontal. When the vertical component of the velocity is zero this energy is stored in tin- various parts of the aeroplane, particularly th,- absorber, the tyre, and the axle. The energy in the tyre and axle is almost entirely given back in the rebound, as it is mainly energy of elastic strain. Of the energy in the absorber, some part, though generally not the whole, is not given back, but is converted into heat. The load deflection diagram of a typical rubber shock-absorber is somewhat as shown in Tig. 1, diagram 4. Area A IU'I)E = the energy put into the absorber (resilience). Area AFDE A = energy given back by absorber. Area ABCDFA = energy dissipated, i.e., hysteresis DFj = maximum load. Since the maximum load DE is limited, the capacity for absorbing energy can only be increased by increasing the deflection, or by the use of a flatter load deflection 142 diagram. The deflection in many designs is limited by the diameter of the wheels, or propeller, or other details. The problem in the design of the complete absorber is to absorb and dissipate as much energy as possible for a given maximum load and deflection, subject to the condition that the machine is also reasonably sprung for running on the ground. One of the most convenient materials for absorbing energy and dissipating it is rubber, which is generally used in the form of a braided rubber cord. Braided Rubber Cord. Ordinary braided rubber cord is made up of a large number of strands of rubber, usually about 1-30 inch square, covered by two layers of braiding. The strands are made of fine hard Para rubber, and are very extensible, the ultimate strain at fracture being generally between 7 and 8 (i.e., length at fracture eight to nine times initial length). The braiding is put on whilst the rubber is considerably extended. As compared with the rubber the braiding is practically inextensible, so that when the tension on the rubber core during braiding is removed the braid exerts a radial com- pression on the rubber, which prevents its returning to its original length. The rubber core is, therefore, per- manently held with an initial strain in it. Tests of f inch Cord No. 2 (Figs. 1, 2, 3) The loads were applied by dead weights. Two elastic bands were fixed 4 inches apart, and the extension measured at each load. A series of tests was made for different maximum loads, and in two cases (at 60 and 120 Ibs.) a test was also made after 10 repetitions of the same load. The outer braiding broke at 180 Ibs., and two tests, Nos. 6 and 7, were done on the rubber core alone. The curves plotted from the observations are given in Fig. 1, and a summary in the following table : 3/8 INCH SHOCK ABSORBER CORD. 2 Layers nf Braiding. 110 Strandt RvXber, approx. -03" square. From another specimen of similar cord. Length before braiding cut off = 4". t Original Length length... 22-7" g 11. nil Length. Weight. of rubber with braidin d Real initial strain in rubber ... 1 Gauge length used 4" Braiding cut off < 5 U v? r i tsrAifiiutr grms. 1-75" 3'74 fi 90" S 40 Weight of cord per foot = '047 Ibs. Weight of rubber = 52-5%. 1 Resilience. Hysteresis. % Taken by Rubber. fc 43 _p 3 g o 02 c & "o a a 'S 5^ V B 4 c s 1 - i tj 83 al jj bd d a B E E o a _, o I i 35 4 9 3^ < ' ^4 S '53 an H '|5- 8 * K w" 4 . H PH Ibs. in. Ibs. in. Ibs. in. Ibs. in. Ibs. 1 20 7 2-4 2-52 40-6 10-2 1 1-6 4 87-5 91 63 2 40 1-13 3-25 5-27 85-2 21-3 58 9-35 2-1 57-5 84 36 81 60 1-25 3-5 6-5 104 26-0 1-24 20 5-0 42-5 81 22 SB 60 after 1-28 3-56 6-32 101 25-3 85 13-7 3-4 10 loadings. 4 80 1-33 3-65 7-11 114 28-5 1-47 23-4 5-85 i 33-8 80 24 5A 120 I'M 3-82 8-93 143 36-0 2-95 47-4 11-9 23-8 70 16 SB 120 after 1-44 3-87 8-21 132 33-0 1-85 29-8 7-45 10 loadings. 180 1-58 4-16 17-0 274 68-5 __ 18-0 Rnpture of braid. Per J" of core, Rubber Strands only. On 2" gauge length. Rubber in 1" of i cord. 6 38 4-4 1-33 " 21-5 5-4 7 .52 ~~ 4-9 2-U6 47-5 11-9 In Fig. 2 are given similar data from tests on a single strand of the rubber core and in Fig. 3 are given the load, resilience and hysteresis diagrams of the complete cord and also of the rubber core plotted on a strain base. Since the rubber when made up into the core of the cord is initially strained, the real strain in it has been calculated and a scale for it is also given. The typical load strain diagram of a single strand of rubber is shown in Fig. 2. There is hardly any portion where the load is strictly proportional to the strain. The first part of the diagram is curved, but between a strain of 1 and 3 the relation of stress to strain is almost linear. Beyond the strain of 3 the loads for equal increment of strain get increasingly greater, and above a strain of fi the curve is very steep indeed. The hysteresis is very small up a strain of 3, i.e., the end of the straight por- tion of the load extension curve, but increases fairly rapidly for increased strain beyond this point. Chapfcer.XH " SHOCK ABSORBER CORD, N ? 2. LOAD-STRAIN DIAGRAMS. FIG. REAL STRAIN IN RUBBER SINGLE STRAND Or RUBBER ftf FRO* DIFffKCHT MAKCffS CO 6 CM *Z o o jo. i tru ESI MOM auow. 143 The typical load strain diagram of the complete cord as given in Fig. 1 (Curve 4) has three well-marked portions : A. to B. In this region the load is taken entirely by the rubber core. On account of the relatively inexten- sibility of the braid the small strains, by reducing the lateral compression on the rubber, are accompanied by relatively large increases of load. At the point B, the braiding is quite slack, and does not constrain the rubber. B. to C. In this region the load is still almost entirely taken by the rubber, the shape of the diagram follows closely that of the rubber alone. 0. to D. In this region the influence of the braid becomes very marked. On account of the large exten- sion and the resulting small diameter of the rubber core, the threads of the braid are now inclined at small angles to the length, and being much less extensible than the rubber they take increasingly greater proportions of the load. In this particular sample the rubber takes most of the load until the nominal strain of the cord is about 1. With greater strain than 1-25 the braiding takes the larger share, and at rupture this amounts to 82% of the total, i.e., 148 Ibs., while the rubber takes only 32 Ibs. In Fig. 3 the resilience is plotted against the strain. For strains less than 1-25 the major portion of the resili- ence is taken by the rubber; beyond this point the curve rises very rapidly, largely due to the fact that the braid- ing is being overstrained by permanent extension. It may ba noted that the resilience at a load of 60 Ibs., and also at 120 Ibs., is not seriously decreased after a repetition of 10 loadings. . The hysteresis curves in. Fig. 3 show that for small strains, less than 0'5, the rubber accounts for the major part, but at a strain of 1 it accounts. for less than $ the hysteresis. Since for small strains the hysteresis, of the individual threads of the braid itself is inconsiderable, this indicates that energy is being dissipated in friction between the strands of rubber and the braiding,. For the large strains the big hysteresis values are due mainly to permanent overstraining of the braiding. The hysteresis decreases after repeated loadings, thus with a load of GOlbs. the hysteresis after 10 loadings was-3'4 inch-lbs. per inch, or only about 68% of that with the first loading; and with a' load of 120 Ibs. the hysteresis after 10 loadings was T'45 inch-lbs. per inch of cord, or only about 63% of that with the first loading. This change is characteristic of rubber. After a lapse of a little time, however, the rubber recovers its original pro- perties, and the original resilience and hysteresis curves are again reproduced. The remarkable extensibility of this high-grade rubber thread makes it peculiarly fitted for absorbing energy. The amount of energy that 1 Ib. of rubber thread can take up in being strained between the limits 1 to 4| (the working limits off the core of shock-absorber cord) is about 14,000 inch-lbs. This is many times that of a spiral steel spring, which even when stressed up to the high shear stress of 40 tons per square inch, only absorbs 616 inch-lbs. per 1 Ib. of material. When made into cord the rubber is accompanied by approximately two-thirds of its own weight of braiding, so that the energy per Ib. of cord is approximately 8,400 inch-lbs., i.e., the braided cord takes up approximately 13 times as much energy as a steel spring of the same weight. The braiding will only allow of a certain extension before it begins to take considerable load. If therefore the initial strain in the rubber is increased the load taken by the rubber at the point where the effect of the braiding becomes very marked may be considerably in- creased. These tests indicate that an initial strain of 1'5 to 2 would be suitable. After the above tests were completed samples of f inch cord with larger initial tension were obtained and tested. Test of f inch Cord No. 3. (Fig. 4.) In this cord the initial strain is considerably increased above that of .No.- 2, i.e., to 1'45, and 200 strands of it are required to make up the f inch diameter. Particulars of the tests are given in the following table, and the load-strain, resilience-strain, and hysteresis-Strain diagrams 'are given in Fig. 4. No. 3. 3/8 INCH SHOCK ABSORBER CORD. 2 Layers Braiding. 200 Strandt Rubbtr, approx. -03' X 03". iLength before braiding cut off = 6". Original length 17-4" Length. Weight. Length of rubber with braiding cut off ... ... v.. ... 7-1" ' grammes. Real .initial strain in rubber ... 1-45 Gauge length ... 2" Braiding cut off | 3"^^ 2-08" 10-6" 6-16 4-14 Weight of cord per foot = -045 Ibs. Weight of rubber = 609r. ' a ri 3 Resilience. Hysteresis. % Taken by Rubber. *F 2 H i >M i . tr a O "3 a E 1 S 8 .JJ K-o'. - i L- 13 g 1 s E . S* JJ 1 ' f ,1 o O ' W HI in Ibs. in llm. i 60 78 3-36 2-2 _ 28-7 2 2-59 73 88 69 2 96 1-05 4-02 3-8 _ 48-3 83 10-7 59 81 44 3 124 .1 17 4-32 . .4-8 . 62-2 1-34 17-4 54 76 40 4 204 1-34 4-73 6-8 87-8 2-47 32-0 41 70 36 5 304 ' 1-46 5-08 9-3 120 4-23 54-9 32 63 28 i 144 In determining the contribution of the rubber the data obtained from the single strand of No. 2 have been used. The two cards, No. 2 and No. 3, are from different makers and may have differed in quality of rubber. The No. 3 cord although its weight per lineal foot is practi- cally the same as No. 2 sustained a load of 304 Ibs. without rupturing. In Fig. 5 the load strain diagrams up to 120 Ibs., and also the resilience and hysteresis curves for the two cords are plotted side by side for comparison. It will be seen that the resilience of No. 3 is more than twice, and the hysteresis several times, as great as for No. 2 over the whole run; for example, at a strain of 1 the resili- ence of No. 3 is 2 - 7 times, and its hysteresis 13 times that of No. 2. Function ol Braiding. From the nature of the work which is required from a shock absorber it is obviously inadvisable (even if it does not mean damage to the braiding), to stretch the cords to a greater nominal strain than 1 (i.e., 5 inches of cord stretched to 10 inches). Bound about this strain the load extension curve for the rubber itself, as well as for the complete cord, becomes very steep, and comparatively little work is required to increase the maximum load very con- siderably. The function of the braid practically amounts to First. To maintain the initial strain in the rubber. By initially straining the rubber it is possible to take full advantage of the extensibility of the rubber without the abnormal extension which would otherwise be required. Second. To protect the rubber against oil, sun- light, accidental cuts, etc., to all of which the rubber is very susceptible. Relation between the real strain in the rubber of the core and the strain of the cord as a whole. Since the action of the complete cord within the strains that will be used is practically that of the rubber alone, it is necessary to be able to calculate the real strain in the rubber. The initial strain can be ascertained by cutting a definite length of the cord, stripping the braid and measuring the unstrained length of rubber. Let A be the length off the unstretched rubber, before braiding; let B be the length of the unstretched rubber cord, after braiding; let C be the length of the rubber cord, after stretching. Then the. strain in the rubber due to braiding" is T> A = S say; and the strain of the cord due to C-B r " stretching is ^- =S 2 say; and the final real strain of the rubber is v = S 3 say. Then it is easy to show that S 3 = S, + S^ + SjSj. Use of Rubber without braiding. In designs in which the rubber can be adequately protected from all damage, bare rubber can be used with advantage. Using the kind of rubber employed in the manufacture of these cords, it would require to be put on with an initial strain of 1J, and be worked up to a strain of 4 or 4. Other types of rubber are probably available which could be used in slightly different ways ; further experiments are very much needed on this subject. " Initial Tension " of Cord In an early specification for shock-absorber cord the term " initial tension " was used to define the load necessary to extend a specimen having an acting length of 12 inches through 0'06 inch. This, however, does not correspond with any well marked point in the load-deflection diagram, and in practice it is almost impossible to determine it satis- factorily. If, however, the term be used to express the load at a strain of O'l, a well-marked point is obtained, i.e., the point at which the braiding is just slack, which corresponds with the end of the first portion of the load extension diagram. The value of the tension at this strain enables a good estimate to be made of the initial strain of the rubber. It is also a useful quantity for design purposes. Assuming that the load deflection curve of the single cord represents that of the complete shock absorber, the number of acting cords should cer- tainly be equal to, and preferably be greater than, the total weight divided by the initial tension defined as proposed. Cords can be made with the initial real strain in the rubber as high as 2'33. In Fig. 6 are given three curves for cords made by Messrs. Luke Turner & Co., of Leicester. The tests were made by Messrs. Turner exactly on the lines of those already described. In general it does not appear to be advisable to use as high an initial strain as 2 - 3 and for most cases the cord No. 3 with an initial strain of 1'45 will suffice. Complete Shock Absorber. Fig. 7 gives the results of tests made on a complete shock-absorber supplied for the purpose by Messrs. A. V. Hoe & Co. In this absorber the cord was put on with considerable (though indefinite) initial tension and in two layers. The inner layer was found to be much tighter than the outer. The cord is first made into one long endless loop 9 feet in circumference, the joint being made by three rings of steel wire pinching the two ends. To make sure he will get the right number of turns on, the workmen naturally puts the first ones on very tight and usually has some- thing " left over " at the end. The effect of this un- equal tension is clearlv seen in the Fig. 8 ; the hysteresis loops get much smaller for the same load after a few repetitions of load have equalised the unequal distribu- tion. A further set of tests was made with the cord put on in what was sajd to be a much more satisfactory manner; the results are shown in Fig. 9. It will be seen that the resilience and hysteresis are less than in the former tests. The same length of cord was used, but was put on so as to make 14 coils instead of 11. thus the initial tension was considerably increased and the steep part of the curve occurs after a very little elongation has taken place. In the drawings the lines "orresponding with the actual dead load and 3, 4 and 5 times this load are shown. Neither of these load deflection diagrams is a good one, the arrangement is not suitable for absorbing energy. In the first set of tests the deflection is very considerable before anv appreciable load is taken and afterwards the load-deflection curve becomes very steep. In the second set whilst the loads are better the deflec- tions are much less and the actual resilience and hy- steresis have decreased. The practice of stretching these cords when putting them on should not be fol- lowed. It is preferable to use a cord in which the strain has been correctly adjusted in manufacture. From a number of tests on under carriages the data referring to the shock absorber alone have been extracted and the curves given in Figs. 10 and 11 show the results It appears that in several of these a'so full advantage is CHA PTER.XII. F.G.5. COMPARISON BETWEEN SHOCK ABSORBER CORDS >V<*2 o ^3 * cr> 19 4 I fc '9MOJ. SNOJL QVQ-1 CHAPTER. XII. TISTS OF COMPLETE VNDR CARR/AGZS LOAD DEFLECTION OAORAMS OF SHOOK ABSORBER 11,000 10.000 MM * 1 KffVfTMOe \ /w "pofrr ASH. r *po#r A ess RF*Oftt A gSO aatnt rt*m rr/r rtAl i IM Wt/'%T i 'li.jJ 1 / ' fa 7>W &. 9. TYP* JQPWTK scewr LOAD ATceu.APSt /< r* WI> (!%3SZ>"Z LOA & AT COLLARS f fttQfJt* (frm*- Tt+tJ A*CJ*& LOAD ATCOtLAftt *S4t*l (AtirSWfftxHldl TOT, "/SK7 Y At *VfJ I to 79rALVf**r AOMfffi mtom.L**. :/ TOT " """'Hxzy'ij TPTAL EMfJt+Y A ^ '* ttt / f J t 7 j * ^ / i / ^ 5 "" t.00 J f f ^ / y J f " t I j L/ j 1 f \ \ 1 fau A/A TAf f H *O ?_ '"" MIL I "* f j. 9 ( fl'L \ ITAt r^A ^ \ X r* r\ T ///* t 0AD MfLECT/ONS M IMCMCS Fi'q.lO. TESTS OF COMPLETE UNDER CARRIAGES LOAD DEFLECTION DIAGRAMS OF SHOCK ABSORBER \ftePORTA. ei ttePOAT J(A3OO \ 'AVfiO ' v Trttf rxtrstt F. m f^i fHJ- itlLITAKY t AO Htfitt. T yyff Sff K-fftOAO- IttfLU FULL MILITARY ti^/J 1*00 LB* IMM LOAO AT [COUJkPSttllttH* (fft*nf {of*-) LOAO AT COLL A/*S-- /* $ ooi SJ HUM TOTAL SMIHY MfUtfff tai at. itn I TOTAL /* IfH&w Aetango iff- L99. TOTAL N*6V Amsoftmett ttt ias. i '' J -4 7 t 1 $ w ^ ^ l\ 2 x s f ^ f ^ / f, - I / ~*T~ fKJ. 1 HUTA,, UA A \ VIM 1 T II ) 1 t 1 1 oeru 1 30 I t 3 'CTIOtIS IN 1HC#CS Fi'q.ll Chapber.XII y ed I I QQ CO $ xl 5 HONI-Xi/O/4 1VJ.GLL ro Co B BS CO <* $ l <^ Chapker.XII. FlG.15. FATIGUE TEST ON RUBBER STRAND FINAL STRAIN 7 6 5 4 3 2 ! * /ND/CATS RUBBF/? &/?Grt'/V AT G&/P "* . S -- , * iii -'"V ^Mi yr A>/ /,v> e = tf*"fc_^ 1 ___ . 105 A'.JV. 115 X 32'* double catered. Initial Strain of Rubber in Cord 1-22. 12 to 1-75 ,. 1-62 .. .. 1-ii .. 1-37.-. i- as 5-11 4-81 4 -fir. 4-27 4-0 250 Complete. 730 ii 1000 67 strands broken. 5000 70 10,000 3 No. 106 K.N. 74 X 18'* single covered. Initial Strain of Rubber in Cord. 1-61. 12 to 1-25 4-87 10,000 2 strands broken. , ., 1-6 5-53 700 Complete. 1-375 5-2 900 38 or more strands bnken. These results agree fairly well with those obtained on the single strand. The lowest final strain is about 3-9 and the highest 4'8 to 4-9. All these tests were done at a speed of 16 cycles per minute. It is probable that the speed at which the 147 tests are made will affect the results. Captain Scoble lias taken photograph records of the load-strain diagram tor a number of tests and has found that the loading curve of a f inch cord is not greatly altered by a change of speed from 16 to 115, but that the hysteresis loop is somewhat reduced at the higher speed. From these tests and those on the rubber strand it appears that so long as the maximum strain in the rubber is not greater than about 4-5 a reasonably long life can be obtained from a cord. Messrs. Rowland and Turner have also carried out some tests on the effect of repetition of strain. They made two cords, one of 3 ',.- " rubber and the other of o 1 ,. " rubber and found little difference in their beha- viour, thus verifying again that considerable variation of size of strand may be allowed. Their experiments were done at a rate of 30 cycles per minute. The' cords had a high initial strain of 2'07 and it was found that straining to double extension, i.e., to a real strain in the rubber of 5'1, the cords were destroyed after about 500 repetitions. When the real strain in the rubber was reduced to 4'1 by only extending the cord to 68% instead of 100% the cords were found to be in perfect condition after 10,000 repetitious. The load-strain diagrams after different periods during the test are of interest; they are given in Fig. 16. It will be seen that the diagrams after 500 and 10,000 repetitions are practically identical. It appears, therefore, that the cord remains practically constant after the first few cycles. These results are in fail- agreement with those of Captain Scoble and all the tests point to the desirability of not exceeding about 4 to 4J for the real strain in the rubber core. Decrease of strength with time. Messrs. Rowland and Turner have pointed out the effect of time on the strengtli of the complete cord. After two months a cord of n \ " rubber which originally gave 200 Ibs. at double extension gave only 164 Ibs. and a cord of ^ l n " rubber which originally gave 200 Ibs., gave only 174 Ibs. This decrease is doubtless due to the gradual stretching of the braid which is of course under considerable cir- cumferential tension; this allows the rubber to contract and reduces the initial strain. This shrinkage is pro- bably affected by temperature and humidity, but no data are at present available. Effect of Immersion of Rubber Cord in Oil. Several experiments have been made by the A.I.D. on the effect of oils on rubber cord. Three pieces of cord ( inch diameter, initial strain i'45) were subjected to the following treatments: Flrni. Pulled to double extension in the usual way. Second. Immersed in Vacuum 1313 oil at 100 C. for eight hours, allowed to cool and then pulled to double extension. Tliird. Outer braid carefully removed and then submitted to the same treatment as the second, the length being measured before removing the braid. In each case the load-strain curve was plotted. The results are summarised as follows : Load at double extension. 1st ... 118 Ibs. 2nd ... 106 Ibs. 3rd 104 Ibs. Hysteresis. 21% 21% 21% it will be seen that the elasticity of the cord has been only very slightly diminished and that the hysteresis has not been affected by the action of the oil. A piece of the cord was cut open after immersion in oil and it was found that the oil had not penetrated more than about jj " into the rubber core. Three more pieces of cord, f inch diameter, were taken. One piece (4th) was stretched to double extension for eight hours, and the load measured at the end of that period. Another piece (5th), was stretched to the same extent and soaked for eight hours in oil at ordinary temperature. A. similar experiment was made on the last piece (6th) after removing the outer braid. The following loads were recorded : 4th 42 Ibs. 5th 40 Ibs. 6th 37 Ibs. From these figures it is clear that the effect of the oil on the cord when strained is negligible. A piece of \" cord was stripped of its braid except at the ends. The rubber was then pulled to 400% elon- gation the load strain curve being drawn. The same rubber after being allowed to rest for 24 hours was heated in oil for eight hours, and then pulled to 400% elongation when cool. The load at this elongation was found to be practically the same as before the treat- ment in the oil, but the hysteresis diminished from 8% to 7%. It is probable that the above mentioned experiments reproduce conditions that are more drastic than those likely to be experienced by shock-absorber cord in use; they show that no serious damage is likely to result from oil in practice. CHAPTER XIII. FABRIC. Report on the Work done during the War by the Experimental Fabrics Establishment, by Professor A. J. Turner. Laboratory, Royal Aircraft 1. From the beginning of the War until about the end of 1915, the Fabrics Department of the Royal Aircraft Establishment conducted the testing and inspection of a deal* of the aeroplane and aeroplane tent fabrics * For some time all this work was carried out at R.A.E. 27264 ordered by the War Office. The aeronautical Inspection Department, however, gradually took over this work, and although a considerable amount of routine fabric- testing was done at R.A.E. during 1910, this practically ceased towards the end of that year. While, therefore, much labour was necessarily devoted to routine work T 2 148 during this period, there remained but little time in which to prosecute experimental investigations. As the routine work fell off, so the research work developed, until from the latter part of ]916 onwards the work was almost entirely of this kind. In May, 1917, the work was further hindered by a fire, which destroyed a num- ber of records and samples of fabrics. 2. The experimental work undertaken was confined primarily to such investigations as would be useful in their application to the two classes of fabric mentioned above, with which the Department was most concerned, viz. : (1) Aeroplane fabrics. (2) Aeroplane tent fabrics. A review of previous investigations showed that little or no useful scientific work had been carried out with a view to determining the best fabric even for the first of these purposes. (H. Dept. Beport, No. 607). As a starting point, therefore, the conditions were studied which must be fulfilled by a satisfactory aeroplane fabric; these were that the final wing covering should be: (1) Airtight and water-tight. (2) Taut. (2) Strong. (4) Light in weight. (5) Tear-resisting. (6) Unaffected by prolonged exposure to weathering action. Moreover, the above features should retain their values throughout the range of conditions which might occur in practice. The desiderata of a tent fabric were practically iden- tical with those enumerated above, except that the pro- perty, of tautness was undesirable. The current practice was, and is, to obtain (1) and (2) by the application of dope and varnish, and to test the resulting doped fabric by determining its weight, breaking strength, tearing strength, extensibility and weatherproofness. These tests were open to certain objections; first, the tests were made under arbitrarily fixed conditions instead of through the possible range of conditions; secondly, the general method of determining tearing strength by find- ing the breaking strength of specimens containing a given transverse wound did not cover all possible cases of damage : and thirdly, although it appeared that the best laboratory reproduction of the practical conditions was by means of some form of bursting test, yet such a test was not in general use. It was concluded, there- fore, that one branch of the work must be further inves- tigation of the physical properties of the materials and of the methods of testing. The results of such work would yield criteria to be applied to any fabrics in order to decide which were the best fabrics for the purposes required. 3. For the ultimate aim of providing the best possible fabrics, a general research scheme was planned which should reveal the most suitable (a) Fibre; M|) Yarn; (c) Fabric structure; (d) Finish of fabric. The ideal complete method of attacking the problem would have involved making a large number of fabrics differing systematically in the features (a) to (d) above; thus each of the different kinds of fibre would have been made into yarns of different counts, each count being made from the best fibre available spun with a complete range of spinning and doubling twists. The resulting yarns would then have been used in the grey and after various treatments for the manufacture of fabrics of various weaves, the same or different yarns being used m any one fabric. The resulting fabrics would finally have been tested in the loom state and with various finishes, according to the methods shown to be necessary by the investigations referred to above. 4. It is manifest that the carrying-out of such a com prehensive scheme as that outlined above was not a " practical proposition " during war-time, especially when it is considered that many other questions of a quasi-experimental nature were also submitted to the Fabrics Laboratory. Such modifications were therefore introduced as appeared likely to give the maximum of result with the minimum expenditure of time and labour. By this process of selection it was of course possible that an erroneous judgment might lead to the elimination of something of real value, but this was a risk which in the circumstance had to be taken. 5. The method of application of the above principles is illustrated by the attempt, begun in 1915 at the in- stance of the Superintendent, R.A.E., to produce a light- weight aeroplane fabric.* At the time about 100 square yards of fabric weighing 4 ozs. per square yard were in use on an aeroplane, i.e., about 25 Ibs. weight; the weight of dope and varnish amounted to about another 15 Ibs., or about 40 Ibs. in all. It appeared that if a fabric could be made weighing 2 ozs. per square yard, and so treated that it required only 1 oz. per square yard of dope and varnish, then the total weight of fabric, dope and varnish would be about 20 Ibs. While not very much, of course, the saving of 20 Ibs. weight appeared to be a useful economy if it could be achieved without the sacrifice of safety and efficiency. The following extracts from H. Dept., Keport No. 607 show the plan proposed : " (a) Yarn Tests. Tests will be made on the yarns of linen, cotton, silk and ramie. Six counts of yarn of each fibre will be used, four being finer, and one being coarser than the stan- dard, which will be the same count as that used in the preparation of E.A.F. linen. Equivalent counts of the yarns of the dif- ferent fibres will be employed. For eacli yarn the twist will be such as to give maxi- mum strength. The results of these tests will show the relation between the count and the strengh of the yarn. In this manner will be determined whether these two quantities are directly dependent one upon the other, or whether there exists a ' best ' yarn, i.e., one having a maximum strength for a given weight." " (b) Fabric Tests. Each of the yarns tested will be made up into a plain fabric weighing from 2 to 2J ozs. per square yard. The fabrics will be made so that for each fabric the strength of the weft is approximately equal to the strength of the warp. The coarser yarns will also be made up into fabrics of greater E.A.K., H. Dept., Keport No. 607. 149 weight. Tests on these fabrics will enable decisions to bo made on the following points : (i) Whether the best yarn gives rise to the strongest fabric having a given weight of 2 to 2-J ozs. per square yard. N (ii) The influence of fineness of weave upon the specific strength of a fabric. (iii) The difference in specific strengths of fabrics composed of linen, cotton, silk and ramie yarns respectively. The most promising of these fabrics will be studied in relation to doping, weathering and wounding." It may be observed that the ramie was included in the scheme owing to claims having been advanced that it was in many respects (e.g., strength, elasticity, etc.) superior to most other fibres and owing to its reported use for the earlier Zeppelin fabrics; while an examina- tion of trade returns had made it appear that the avail- able supply was more than sufficient to meet any conceivable demand. YARN TESTS. 0. In order to carry out the above plan it was neces- sary first of all to determine the optimum amount of twist which a yarn should have so as to possess the maximum strength. Although certain information was available on this point, it was of a fragmentary character and not directly applicable. After some con- sideration it was decided to study the effect of twist in yarns having the same equivalent count as those used in the manufacture of the standard linen. Unfor- tunately it was not found possible to do this in all cases. Thus the desired linen yarns with various degrees of twist could not be obtained at the outset, nor could ramie yarns of sufficient fineness be obtained; yet exj. .>rience with linen yarns of different counts, e.g., thre e linen yarns supplied by the York Street Flax Spin- ning Co., Ltd., together with the results subsequently obtained in the cases of cotton, silk, and ramie lead to the conclusion that the twist normally given to the dif- ferent yarns was practically an optimum. 7. The testing of the varns involved determinations of: (1) Count; (2) Breaking strength; (3) Extensibility; (4) Twist. From the weight (count) and strength of each yarn the corresponding breaking length was calculated, and this, being the true criterion of specific strength, was used in plotting curves showing the effect of twist on strength. The results obtained may be summarised as follows. 8. Cotton Yarns.* The production of these yarns was kindly undertaken by the late Prof. Fox of the College of Technology, Manchester. The varieties of cotton used were the best Sea Islands and the best Egyptian respec- tively. Single yarns of 80s count were spun from each * E.A.B., H. Dept., Report No. 540. variety with 20, 30, 40, 50, 60 and 70 turns per inch respectively. Provision was n.ade to ensure that, irre- spective of the amount of twist, the cross-sections of all the threads should contain the same number of fibres. Each of the twelve single yarns thus obtained was con- verted into six two-fold threads containing 20, 30, 40, 50, 60 and 70 doubling twists respectively : thus, in all, there were 12 single yarns and 72 two-fold yarns. The chief conclusions reached as a result of these tests were : (1) The strongest 80s singles have 40 turns per inch ; it is not from these yarns, however, that the strongest two-fold yarns are made, but from singles with only 20 turns per inch (and 35 doubling twists per inch). (2) The greatest breaking length possessed by any of the singles yarns is 22,500 yards, tested wet; the greatest breaking length of any of the two-fold yarns is 25,000 yards, i.e., 14% greater than the best singles yarn. (3) Comparatively large variations of the twist can take place in the neighbourhood of the opti- mum degree of twist without causing difference in strength. much 9. The last conclusion appeared to be of 3onsiderablo importance, and from it the deduction was drawn that so long as the twist is kept within rational limits the strength will remain practically unaltered. On the other hand it appeared that the quality of the fibre and possi- bly the quality of spinning had a great effect on the strength of the yarn. Evidence of this was afforded by a number of fine cotton yarns which were tested about the same time as the above; a singles 75s (gassed) had a breaking length, wet, of 28,500 yards, no less than 27% better than the best singles 80s of the twist series; this 75s had 31 turns per inch. Again, 2/94s and 2 /108s gave breaking lengths, wet, of 29,700 yards and 29,800 yards respectively; in each case the singles had 29 spin- ning twists per inch, while the doubling twists were 38 and 37 per inch, for the 2/94s and 2/lOSs respectively. These two-fold yarns were therefore over 16% better than the best fwo-fold yarns of the twist series. The tests on these superfine cotton yarns afforded an answer, moreover, to the wider question concerning the relation between counts and strength. It was found that the differences between the breaking lengths of the various single yarns were comparatively small, as were those between the breaking lengths of the various folded yarns, although in general the breaking lengths of the folded yarns were higher than those of the singles. Hence it was inferred that for yarns made under the best condi- tions and within the limits proposed, the strength of a cotton yarn was practically inversely proportional to the counts. 10. Silk Yarns.* As the silk thread is manufactured from a fibre which is by nature continuous, it was not expected that its specific strength would be dependent to any extent, if at all, upon the degree of twist or the counts. This view was substantiated by the results of the tests. The silk yarns were made by Messrs. J. & T. Brocklehurst & Sons. All the experiments were con ducted in the first place on yarns which had been made by twisting together 12 elementary threads (10-12s * E.A.E., H. Dept., Report No. 972. 150 deuier) of Extra Grand Classical Japan Silk. Accord- ing to the method of combination the resulting threads were single, tram or organzine; they were prepared in the gum, souple and discharged states in order that information might be obtained respecting their com- parative merits in these states. The singles yarns were made by twisting together the 12 elementary threads with 2, 4, 6, 8 and 9 turns per inch respectively; to' make the tram silk yarns pairs of elementary threads were twisted together with 2, 4, 6, 7, 8 and 9 turns per inch respectively, and the respective singles thus formed were six-plied with 2i turns per inch. For the organzine silk yarns pairs of elementary threads were twisted to- gether with 10, 12, 15, 18, 21 and 24 turns per inch respectively; pairs of the singles thus formed were twisted together so as to give 2x2 cord organ with 8, 10, 13, 16, 19 and 22 turns per inch respectively; and finally, each 2x2 cord organ was three-plied with 1\ turns per inch. The results of these tests showed that dis- charged silk had a specific strength which was distinctly- greater than that of the souple or gum silks, and about 25% greater than that of linen yarn of about the same count; moreover, the silk yarns were much more regu- lar than the linen yarns. As mentioned above, the degree of twist was found to be practically immaterial as regards strength. 11. A second series of silk yarns* was accordingly made in which three types of elementary threads (13/15s, 10/12s, 20/22s, denier) were progressively folded. A range of each of these was prepared by folding the elementary threads in groups of 4, 6, 8, 10, "l2, 14, 16 and 18 as far as was necessary to give the desired maximum coarseness (about 130s denier). Two series of each compound thread were made witli 5 and 10 turns per inch respectively. The tests on these yarns confirmed the high specific strength of silk, and, as already stated, led to the conclusion that for all practical purposes the specific strength was independent of the counts of yarn. It will be seen hereafter that certain defects of the silk fibre rendered it impossible to take advantage of this high specific strength. 12. Ramie Yarns. Enquiry showed that the finest ramie yarns spun in England were much coarser than the standard linen yarns. Certain dry-spun yarns were supplied from stock by the Eamie Co., Bredbury, Nr. Stockport; these were 30s, 35s, 40s, 45s, and the same yarns doubled, i.e., 2/30s, 2/35s, 2/40s and 2/45s, the counts being reckoned on the worsted scale. Wet-spun yarns were similarly supplied from stock by the Yorkshire Ramie Co., Leeds; these were 16s, 36s, 45s, 2/22s and 2/28s. Each series of yarns was in the half-bleached condition. It was found! that the two-fold yarns had a much greater specific strength than the component singles, and that the dry-spun yarns were rather stronger than the wet-spun. The chief conclusion was, however, that there was no great likelihood that the ramie fibre would yield a yarn superior in specific strength to its count equivalent made from flax. Subsequent experi- ments I with 35s ramie yarn spun with various degrees of twist ranging round 14 turns per inch, both in the unbleached and the half-bleached condition, dry-spun and wet-spun confirmed this conclusion. * R.A.E., H. Dept., Report No. H404. t R.A.E., H. Dept., Report No. 700. J R.A.E., H. Dept., Report No. H396. 13. The position readied as a result of the yarn tests in general was thus the same as that stated in regard to the cotton yarns. Twist appeared to be of subsidiary importance so long as the yarns were " made for strength," while grade of fibre and the quality of the spinning were of primary importance; and, finally, when these fundamental conditions were attended to, it appeared that the specific strength of a yarn of a given fibre was a function only of the fibre and that it was independent of the counts. Hence by testing a large number of high-class yarns it was possible to fix the maximum practicable limit for the breaking length (specific strength) of any yarns of a particular fibre. The recognition of this fact effected a considerable sim- plification of the work in connection with fabrics, which will now be dealt with. FABRIC TESTS. 14. While the tests on the various yarns were proceed- ing, investigations were also made of a large number of fabrics in order to ascertain what were the best fabrics immediately available. The results obtained may be most conveniently classified according to the fibre from which the fabrics were made; they will be considered in the order (1) Linen fabrics; (2) Cotton fabrics; (3) Silk fabrics; and (4) Ramie fabrics. 15. Linen fabrics. The three linen yarns referred to in paragraph 6 (footnote) were also supplied in the form of fabrics weighing 3} to 3f ozs. per square yard respec- tively. B.A.P. Specification 17A then in force for linen aeroplane fabric, required a fabric having the following characteristics : Maximum weight Ends per inch ... Picks per inch ... 4'0 ozs. per sq. yard 96 100 Minimum strength tested wet. Warp 91 Ibs. per inch Weft ... ... 102 At the time, however, this minimum strength wan usually much exceeded; thus, a fabric tested towards the end of 1915 had a strength, wet, of 119 Ibs. per inch in each direction, and such a case may be regarded as typical. The three experimental fabrics differed from the standard only in being made with finer yarns, the ends and picks per inch remaining unaltered. When the results of the tests on the three fabrics were compared with this common standard, it was found that there was practically no alteration in the specific strength, the ten- dency being rather towards a loss than a gain. Other examples may be cited, in which the same tendency was observed: light weight plain linen fabrics weighing about 2J ozs. per square yard were obtained from the York Street Flax Spinning Co. and from Messrs. Stevenson & Son, Dungannon. In order to obtain this reduction in weight, the specification was so drawn up that equal parts were contributed by a reduction in the number of threads per inch and by an increased fineness in the yarns used. This arrangement was necessary so as to preserve the dopability of the fabric; another necessary provision to this end was that the fabric should be highly calendered. In spite of the greater fineness of yarn called for, which, generally speaking, would involve the 151 use of a higher grade of fibre, each of the resulting fabrics had a breaking strength which was barely up to the specification requirements, and a specific strength which was below that of the typical standard mentioned above. It is probable that the reduction in specific strength was occasioned at least partly by the .greater openness of the cloth structure ; it is probable, too, that the fine yarns used were not of the highest class, possibly owing to the cutting off of supplies of Courtrai flax mak- ing itself felt first in the finest counts of yarns and later in the less fine counts. Fairly extended experiments were made with this light-weight linen and in all of them it proved satisfac- tory; it was used on alternate wings of some experi- mental machines and it gave service equally as good as that of the standard fabric used on the other wings, no defect arising after some months' flying. The general adoption of this fabric was, however,' inhibited by the restriction of supply of fine flax yarns which had its incidence about the time these experiments reached a successful conclusion. 16. Cotton fabrics. Tests on cotton fabrics were made in the first place with the idea of investigating their suitability for light-weight aeroplane fabrics. It soon became apparent that there was not much likelihood of cotton displacing linen fabrics, but the experiments were continued so that information on the subject would be at hand if, owing to a shortage in the supply of linen, it were desired to use a cotton wing-covering. Such an eventuality actually came to pass in 1917 vide A.C.A. Eeport, No. 346. The results of tests made on some of these fabrics are given in the table below: Fabric No. Description. Weight, ozs. per Threads per inch. Breaking Strength (wet). lbs./inch. sq. yd. Ends. Picks. Warp. Weft. AF 19 Plain scoured cotton 2-1 152 148 45 50 fabric. Singles warp and weft, AF 22 do. 2-9 152 148 63 65 AF 208 Plain scoured cotton 3-7 134 126 66 78 fabric. Two-fold warp and weft. A 616 Plain cotton fabric 3-6 108 92 79 73 mercerised. Two- fold warp and weft. A 617 Plain cotton fabric. 3-6 112 96 67 60 mercerised. Three- fold warp. Two- t fold weft. A 62a Plain scoured cotton 4-2 104 100 73 77 fabric. Singles warp and weft. A63a do. 4-4 96 102 86 90 A 64 do. 4-2 96 98 93 86 9925 do. 2-6 136 134 6H 6S 9926 do. 2-8 128 118 67 55 9927 do. 3-0 108 108 69 71 at any rate as far as breaking strength is concerned. The same is true of the fabrics in the doped state. When tearing strength was the criterion used, the com- parison was even more unfavourable to the cotton fabrics, whether undoped or doped. Judging by state- ments appearing in various quarters it .was thought that the use of a mercerised cotton fabric might obviate some of these disadvantages, but as will be seen from the table above, the mercerised fabrics gave no better results than the unmercerised ; other tests in which the same cotton fabrics were tested both in the unmercer- ised and in the mercerised states pointed to the same conclusion. The more extended investigation of cotton fabrics will be dealt with later. It may here be observed, however, that the tests on these cotton fabrics confirmed the conclusion arrived at from the study of cotton yarns as to the interdependence of weight and strength when the quality of the fibre and of the me- chanical operations were of the best. 18. Silk fabrics. A number of silk fabrics were sup- plied by Messrs. Brocklehurst in the discharged state; tests on these gave the following results: No. Descrip- tion. -*. 4 b s t ) d -3L Threads per inch. Breaking Strength Ibs./in at R.H. SOo/c. Specific* Strength. Ends. Picks. Warp Weft. Warp. Weft. Mean 1 Plain silk. 1-4 124 4 45-4 57-1 32-4 40-8 36-6 2 do. 1 9 160 90 64-7 67-1 34-0 34-2 34-1 3 do. 1 9 160 90 52-8 62-4 27-7 32-9 30-3 4 do. 3.0 140 70 146-1 52-8 48-7 17-6 33-1 5 do. 3-1 140 82 97-8 92-3 31-6 29-8 30-7 6 do. 3-1 140 78 143-6 63-5 46" 4 20-5 33-4 7 Silk twill. 3-2 136 ll 98-7 94-8 30-8 29-6 30-2 8 do. 3-6 120 116 151-9 124-7 42-2 34-6 38-4 9 do. 3-8 140 74 133-5 136-1 35-1 35-9 35-5 10 do. 4-2 144 112 140-8 120-0 33-5 28-6 31-0 17. The cotton fabrics described in the above table had been made from the best qualities of either Sea Islands or Egyptian cotton; the results obtained are therefore a demonstration of the superiority of linen, * Specific Strength = l^kingStrength (Ibs./inch) Weight (ozs./square yard) As shown by the above table, the silk fabrics langed in weight from 1'4 to 4-2 ozs. per square yard. The counts of yarns used covered a considerable range, being as low as 33s denier for the warp of the plain fabric No. 1 (and 44s denier for the weft) and as high as 176s denier for the weft of the twill No. 9 (88s denier for the warp). The strengths in the two main directions of any fabric, however, bore a close relation to the counts of warp and weft yarns combined with the ends and picks per inch respectively. Owing to the great differ- ence in the strength of the warp and weft of some of the samples the mean specific strength was used for purposes of comparison. This invariably came out high, the extreme values being 30 and 38. and the grand mean for all the silk fabrics, 33; for B.A.E. linen at the same relative humidity (80%) the mean was about 23. The lluctuations of the mean specific strength did not show any correspondence with the changes of weight. thus indicating that as with cotton fabrics the mean breaking strength is approximately proportional to the weight, and that any reduction in weight involved ?. proportional reduction in strength, other characteristics such as counts of yarn, number of threads per inch. * B.A.E., H. Dept., Report No. 859. 152 structure of cloth, having only a subsidiary influence, if any. 19. A notable feature of all silk fabrics is their high extensibility due in large part to the extensibility of the thread itself (about 20%). The mean total extensi- bility for the various plain silks above was 35% warp, 30% weft, at 80% relative humidity; under the same conditions the standard linen extended 25% warp, 7% weft. This high extensibility is doubtless related to a property causing slackness of the fabric, a feature which proved to be a characteristic of silk fabrics and which no treatment could be found to overcome. In order to pursue the investigation further, however, two fabrics were ordered from Messrs. Brocklehurst some- what similar to Nos. 2 and 5 above. Each fabric was a 2 x 2 mat, made from discharged silk singles with 100 threads to the inch in each direction; for the lighter fabric, designed to weigh 2 ozs. per square yard, the yarns were 80s denier and for the heavier fabric, to weigh 3 ozs. per sq. yard, the yarns were 120s denier. The mat ^ r eave was chosen as other tests had pointed to high tearing strength being derived from such a weave. These fabrics when tested gave the following results : in the yarns. The results of the tests* may be sum- marised as follows : Breaking Strength. (Ibs./inch.) Mean Specific Strength. Weight. Description. ozs. per sq. yd. Dry.R.H.64^ Wet. Dry.RH. 64% Wet. Warp. Weft. Warp. Weft. Silk Mat. ... If 86 56 58 45 49 35 29 2 X 2 (2 oz.) Silk Mat. ... 2-75 81 86 75 75 80 57 2 x 2 (3 oz.) These mat fabrics were rather under the weight speci- fied; but nevertheless their properties were as expected. The chief difficulties encountered as regards their adop- tion for wing coverings were: (1) The silk is very sensitive to weathering, as will be shown later; the loss of strength thereby occasioned, however, can be prevented by covering the fabric with a suitable dope and pigmented varnish. (2) The tendency to slackening previously referred to; no proof was found which would elimin- ate this undesirable feature. (3) The high cost of silk; this objection is deci- sive of course when no compensating advan- tages could be derived from the employment of a silk wing covering. 20. Ramie Fabrics. It was not an easy matter to obtain samples of ramie fabrics, as these are not woven in England. Three samples of plain bleached fabrics which had been hand-woven in China were presented by the Yorkshire Ramie Spinning Co., Ltd. The com- ponent yarns had also been hand-made, the fibres being laid side by side and twisted together merely at their ends so that in the ordinary sense there was no twist Description. Weight, ozs. per sq. yd. Breaking Strength. (Ibe./inch.) Jlean Specific Strength. ThreadB per incb. Dry. 80 G. a RH. % Wet. Dry. RH. mo/o Wet. a 1 DO 1 .*> W 90 55-0 52-9 48-4 42-0 60-0 56-6 52-5 48-0 4c. t MI ) ~t~ P.C.10 | 90 9U-5 95-7 101-7 101-0 92-0 '.)'. -1 101-9 101-0 5a. USA 1 )) J) VJ .*J.il. mercerised ', 30 72-4 75-8 72-0 69-3 74-3 75-3 72-2 63-7 5b. 1* ! 60 67-5 65-3 62-8 57-0 69-4 68-7 63-4 59-0 5c. I ) 1) 1 90 60-1 58-0 50-1 47-5 65-0 64-0 r.li'7 .-,3-0 6a. -Mope 30 78-4* 77-1 78-7 76-4 99-0* M*l '.13 4 91-4 6b Tt ) *) ; fj 60 74-4- 73-2 71-5 70-0 94-9* 90-1 86-0 84 6c. )5 *j ) i 90 69-1* 66-6 63-4 59-8 88-8* 80-2 77-5 75-3 7c. ) )J ji ) ) + P.O. 10 } 90 83-7* 82-S 90-5 91-5 101-0 95-0 . 101-3 103-3 9a. Linen fabric A.F. 8676 ... 30 55 ' 5 58-3 61-1 61-4 57-9 56-3 6(1-9 61-5 9b. }1 ) 60 42-3 41-5 39-3 36-6 50-5 48'3 47-1 44-1 9c. ) > 90 34-8 33-9 31-0 28-8 35-7 ' 34-1 31-6 28-6 lOa. ., + dope 30 47-4 49-8 49-9 47-4 85-6 86-1 85-6 81-5 lOb. f . M ) 60 40-4* 40-6 38-4 35-1 72-7* 73-8 66-8 62-3 lOc. *J M JI 'JO 29-1* 27-9 23-7 iiO-2 59-4* 59-9 51-6 46-4 lie. )T ) JJ P.C. 10 j 90 63-5* 71-9 91-7 94-6 104-8* 110-5 120-5 126-2 12a. Ramie fabric 40 55-3 55-9 53-3 47-2 84-5 78-6 77-:t 66-8 13a. Silk twill 15 68-6 62-3 47-0 35-8 96-6 78-0 63-4 52-6 Ha. Jute Hessian 40 62-5 62-4 58-8 47-8 84-5 82-5 7S -0 64-7 R.H. 30% approx. (3) '24. Of many noteworthy features presented by the above table, the most interesting is the reversed sigu after weathering of the humidity coefficient for the vege- table fibres, and its accentuation in the case of silk for which it was already negative. Thus, instead of the points noted previously in respect of the unweathered fabrics, we have the following: (1) After 90 days' weathering, linen fabric is about 20% weaker when wet than at R.H. 40% (2) After 90 days' weathering scoured cotton fabric (H) is about 30% weaker when wet than at E.H. 40%. After 90 days' weathering mercerised cotton fabric is about 20% weaker when wet than at R.H. 40%. (4) After 40 days' weathering ramie fabric is about 20% weaker when wet than at R.H. 40%. (5) After 40 days' weathering jute fabric is about 50% weaker when wet than at R.H. 40%. (6) After 15 days' weathering silk fabric is about 25% weaker when wet than at R.H. 40%. The importance of these results is brought out by a comparison of the linen fabric and the mercerised cotton fabric. Suppose the loss of strength after 90 days' weathering is to be calculated, then for the various humidities the following values hold : PERCENTAGE LOSS OF STRENGTH (WARP) AFTEE 90 DAYS' WEATHERING ( M.\Y-JULY ) . 40% 60% 90% Wet Linen fabric, A.F. 8676 ... 47-5 53-5 66-5 70-2 Mercerised cotton fabric ... 20'2 24- 1 35-0 39'5 Thus not only is breaking strength a function of the humidity and the material, but the loss of strength also is a function of these and of the duration of weather, ing, the humidity coefficient itself depending both upon the material and the duration of weathering. These facts have to be borne in mind in the interpretation of all results of weathering experiments. 25. Tearing Strength. As previously mentioned, the exploration of the possibilities of the introduction of a light-weight fabric led to the concluson that although there was a fairly definite maximum specific strength which could be attained by a fabric made from any given fibre, yet some improvement might be obtained by alteration of the yarns or weave. An attempt in this direction was begun in 1915 with an investigation of a linen mat fabric,* such a weave according to A.C.A. lieport No. 139 (O'Gorman and Smeaton) being most promising in regard to resistance to tearing. Experi- ments were therefore undertaken to determine the several influences of doping, weathering and wounding on an unbleached linen mat fabric (4x 4). For purposes of comparison standard linen was subjected to parallel treatment. Details of the two fabrics are given below : Counts (warp and weft) Ends per inch Picks per inch Weight (ozs. per sq. yd.) Plain linen. 92s 92 102 4-1 Linen mats. 106s 124 120 4-3 The fabrics were examined as regards (1) weight; (2) breaking strength ; (3) tearing strength ; (4) stress- strain diagrams; (5) extensibility; and (6) tautncss. * R.A.E., H. Dept., Report No. H 522. 155 These properties were studied in the doped as well as the undoped fabrics, both initially and after weathering periods of 50, 150 and 300 days respectively. It was concluded that in all essential features the mat fabric was as good as the plain, having in addition rather higher tearing strength, but suffering from the practical drawback of being rather more difficult to manipulate and dope satisfactorily. The results of later experiments however, indicated that the advantage of higher tearing could be obtained in a plain fabric by the use of coarser yarns, without the disadvantages attaching to the mat fabric. 26. From a consideration of the function of aeroplane fabric it gradually became clear that the resistance to tearing and bursting were desiderata of even greater importance than simple breaking strength. But little work had been done on the subject of tearing strength, the chief being by O'Gorman and Smeaton (loc. cit.) and by Page, who, however, except for the study of a rein- forced fabric, confined his attention to the aeroplane fabric then in use. A systematic investigation was therefore undertaken of a large number of fabrics and other materials* to determine what properties give en- hanced resistance to tearing, and in what manner, if any the results might be utilised for improving the standard fabric. The following list of fabrics, etc., tested shows the range covered : LIST OF FABRICS TESTED FOR TEARING STRENGTH. No. Description. Ends per in. I _g b I GO M 1 Weight (ozs./ yds.) a j> 0$ a T3 s. O a 1 Plain unbleached linen to B.A.E. Speci- 92 102. 4-0 5-3 fication No. 17c. 1 2 Plain unbleached linen, diagonally woven 80 84 2-9 4-0 3 heavily hot calen- 96 96 3-9 4-7 dered. 4 heavily hot calen- 88 92 3-2 4-0 dered. 5 coarse linen, uncalendered ... 50 56 3-4 5-8 6 ., ,, heavily hot calendered 50 54 3-5 5-3 7 unbleached linen mat (4 x 4) 124 120 4-0 7-1 8 Fancy linen mat (7 x 7) to R.A.E. Speci- 96 100 3-5 7-0 fication No. 167 A. 9 Unbleached flax drill 48 64 11-7 , 10 canvas 32 28 17-6 11 Plain half-bleached ramie, 2-fold warp and 34 54 6-2 7-5 weft. 12 .. mercerised cotton 78 84 4-2 6-2 13 scoured cotton (H.) 106 100 3-7 5 * 5 14 .. (A.F. 20) 100 110 4-6 5-7 15 (A.F. 208), 2-fold 134 126 3-7 5-3 warp and weft. it; (A.F. 22) 152 148 2-9 4-0 17 (A.F. 19) 152 148 2-2 3-4 18 ., ., rubbered, single ply 140 120 ; 5-7 18a ., rubbered, parallel- 136 129 10-7 doubled. 18b ,. .. .. diagonal-doubled 124 116 8-6 18o .. .. ., ., 3-ply Outer plies 140 116 15-9 ,, ,, Inner ply 120 118 15-9 19 Bleached cotton satin drill, 2-fold warp, 164 98 6-3 _ singles weft. 20 Plain grey cotton duck 50 46 10-2 2! Plain silk twill to R.A.E. Specification 144 78 3-i? 4-9 No. 64 A. 22 Plain silk mat (2 x 2) 104 108 2-9 5-4 23 " " 100 100 1-6 4-0 No. Description. Weight (ozs./yd.) Undoped. 24 Japanese watsrproofed paper 2-1 25 5-3 26 Cellon sheet (0-02 in. thick) 20-0 27 Raftite film A. (0-0073 iu. th ck) 2-5 27a Raftite film B. (0-0093 in. th ck) 3-3 28 Steel gauze (24 mesh) 32-8 29 Brass gauze (30 mesh) 38-2 29a L>o. (90 mesh) 18-6 30 Copper gauze (30 mesh) 35-2 30a Uo. (50 mesh) . 29-1 30b Do. (100 mesh) 82:6 30e Do. (200 mesh) 12-5 31 Aluminium sheet (0-004 in. thick) . 8-0 31a Do. (0-012 in. thick) . 25-4 32 Aluminium alloy sheet, " T S 5 Alloy " (0-0035 8-1 in thick). 32a Do. do. (0-010 23-4 in. thick) 32b Do. Duralumin (0-016 33-4 in. thick). 33 Brass sheet (O'OOo in. thick) 29-3 33a Copper sheet (0-005 in. thick) 42-4 * A.C A.. R. & M., N'o. 4X7. 27264 27. As a result of these tests it was found that two methods of determining tearing strength are necessary and that the two methods are independent of one another; one method is to determine what reduction of breaking strength results from a wound of given dimen- sions (wound strength) while the other is analogous to the ripping of fabrics by hand (rip strength). The wound strength tests were made on specimens of two different sizes in each case : (a) Specimens 3 ins. wide x 8 ins. long; in these central transverse cuts were made, J in., \ in., 1 in. and 2 ins. respectively. (6) Specimens 6 ins. wide x 12 ins. long; in these the cuts were \ in., 1 in., 2 ins. and 4 ins. respectively. Curves were then drawn showing the dependence of wound strength (plotted as ordinates representing the percentages of the breaking strength) on the lengths of wounds (plotted as abscisses). For the determination of rip strength two simple tests were employed: (a) A "rip test"; for this a specimen 2 ins. wide x 6 ins. long was cut lengthwise down the centre from one edge for 4 ins. ; the two limbs thus formed were placed one in the upper jaw of the Avery machine and the other in the lower jaw, so that the line of of cut was vertical. (6) A " tongue test "; for this a specimen 6 ins. wide x 8 ins. long was. cut so as to give a tongue, centrally situated, 2 ins. wide x 4 ins. long. The top of the specimen was gripped in the upper jaw and the bottom of the tongue in the lower jaw. 28. These tests were made on materials exhibiting such differences as would reveal the separate or joint effects of (1) fibre (or other substance) ; (2) yarn; (3) structure; (4) weave; and (5) proofing or doping. It was concluded that : 1. Tearing strength is chiefly conditioned by: (1) The sliding capacity of, or freedom of move ment possible to, the component yarns. 2 156 (2) The coarseness of the yarns. (3) The uniformity of the yarns. (4) The extensibility of the fabric. (5) The brittleness of the fibre (or other substance), this being of especial importance in respect of rip strength. 2. The effect of doping is so profound in its action on 1, (1) that the advantages of any fabric as regards tearing strength derived from this characteristic are made of little avail, and consequently the type of weave employed is practically immaterial. 3. There is little to choose between textiles and con- tinuous material in respect of wound strength, but in rip strength the continuous materials are so inferior as to be unsuitable for covering wings unless some form of reinforcement is introduced. 4. Among textiles, linen appears to give rather better results than silk, undoped or doped, for rip strength, but rather worse results for wound strength; while linen and silk each show a decided superiority over cotton fabrics in both respects, although the superiority is less over a cotton fabric made from mercerised yarns than over a cotton fabric made from unmercerised yarns. 29. Illustrations of the above conclusions are afforded by the following selected results : No. Description. Breaking Strength (Ibs. per inch). Tearing Strength (Ibs. / inch). Tearing Strength (% Breaking Strength). i-inch. Wound.* 1-ineh. Wound.* Rin Hha ^1 >Im S^ e - Rip (Ibs.) (lbg } i-inch. 1-inch. Wound.* : Wound.* Rip. Tongue. j Uudoped Wp. 78 48 25 12 22 61 32 15-5 Bfl 1 Linen (17 c.)...S r> " j ,^ f ' v ' ) Doped Wp. 90 9R 63 54 43-5 24-5 13 11 24 19 70 66-8 48 25-5 14-5 11-5 27 20 Wf. 112 68 33 10-5 20 61 29 ' 5 9-5 18 j Undoped Wp. 52-5 46 35 17 27-5 88-5 67 32 52-5 6 Linen (Coarse) , " , Wf - ' Doped \V p. 48' 5 64 40-5 57 28-5 41 16'8 16 27 28 83-5 89 59 64-5 29 ' 5 25 .-,]:, 44 Wf. 60 47 34 14 30 78 56-5 23 :, I Undoped Wp. 75-5 77-5 52 33 65-5 100 69 37 72 7 Linen (Mat) ... < " , j Doped Wp. 68 128-5 67 85-5 44 50 37 8 62 12 98 66-5 64 39 41 6-5 75 9 ( Wf. 120-5 74 41-5 6 13 62 34-5 5 11 Undoped Wp. 70 36 ' 5 21 11 17 52 30 1(1 24 1 9 Cotton (Mer-J Wf. 77 49 30-5 1 3 ' 5 64 39 ' 5 12 17-5 i & cerieed). i Doped Wp. 76'5 45 21 - 7 12 59 27 9 16 Wf. 104 57-5 32 5 ' 5 9 55-5 31 5 8-5 j Undoped Wp. 69 36 21 4-5 7 ' "> 52 80' 6-5 11 13 Cotton (H.)...< Dop ^ Wf. 66 88 29 ' 5 43 17-5 21 fi 3-1 8 6-6 44 - 5 49-5 26- 24 7-5 3-5 12 6-4 Wf. 96 40 19-5 2-9 4-9 41-5 20- 3-0 5-1 j Undoped Wp. 118-5 69 45 9 14 58 38 7-:. 12 21 Silk (Twill) .... _. " , , Wf ' j Doped Wp. 134-5 108 72-5 64-5 49 44 13 6-5 25 11 54 60 36- 40- 9-5 6 18-5 10-5 ( Wf. 132 78-5 44 13 24 HO 38- 10 18 i Undoped \\p. 46 42 34 20 4S-5 91-5 74 43-5 105 23 Silk (Mat) .... n j Doped Wp. 4H-5 67 40 ' 5 42 34 28T> 81 -8 4-8 41-5 9-6 100 63 69- 42- 44-5 7-2 86 14-3 Wf 83 48-5 29 4-6 7-2 77 46 7-3 11-5 25 Brown paper 23 18-5 12 0".l 1-0 85-5 68 3-9 4-3 26 Cellon sheet 203 123 66 0-9 2-0 61 33 0-4 0-9 27a Raftite film 27-8 15-7 7-1 56-5 25-5 29 Brass gauze (30 mesh) . . 151-5 91 66-5 8-5 Hi 60 44 5-5 10-5 29a Do. (90 mesh) . . 47 80 21 1-1 2-4 M 45 2-3 5-1 33 Brass sheet 338 283 133 1-U 3-9 83-5 39-5 0-6 1-2 S2 Aluminium alloy sheet 177 107-8 58-5 0-7 61 38 0-4 * Wound tests on specimens 3 inches wide. From the above conclusions and from a consideration of the other desiderata of a wing-covering, it was finally decided that the chief possibility of improvement lay in the use of coarser yarns. It was shown that a " 4-oz/' coarse linen fabric, having only about 50 threads per inch in each direction was rather superior to the stand- ard linen in wound strengtli and much superior in rip strength, while the use of such a fabric is quite practi- cable if it can be given a highly calendered or highly beetled finish. In Appendices 1 and II are given specifications of standard linen and cotton fabrics which were recommended on the basis of these investigations. 30. Bursting Strength The use of tests for bursting strength and their approximation to working conditions were referred to in paragraph 2. Work in this direction had been done by Fage who came to the conclusion. A,C,A. Report, No, 182, February, 1915, that " it does not appear possible to derive any very useful conclusions from bursting tests on aeroplane fabric." Most of the reasons given for this conclusion, however, arose from difficulties in connection with the method of testing; such difficulties were almost entirely overcome by the use of an apparatus subsequently designed and con- structed at the E.A.Fj. The improvement was largely due to the use of water as a medium for applying the pressure necessary for bursting the fabric, which was clamped on the top of a cup-shaped cast-iron vessel, the fabric being protected from contact with the water by the insertion of a thin rubber sheet underneath the fabric. Various minor difficulties were eventually over- come and an attempt begun to correlate the quantities breaking strength, extensibility and bursting strengtli. 'I'll is investigation is still in progress. This bursting apparatus has been used for comparative purposes, as in the study of the suitability of cotton 157 fabrics for covering aeroplane wings (A.C.A. Eeport, No. 346. September, 1917). Specimens were used which had a circular testing area of 12 ins. diameter; below are given some of the results obtained ; the numbers in the first column refer to the numbers given in the list of fabrics tested for tearing strength, which gives details of these fabrics. Bursting Pressure. (Ibs./sq. inch.) No. Fabric. Un- s-s -si ^i &i wounded. .2 g .2 g 2 a z "7* c -iUt^ ' r^t fe N^ -*'? 17. Cotton Fabric AF. 19 9-0 3-9f> 2-9 2-1 1-85 16. ,, 22 13-9 4-1 2-9 1-9 1-4 15. ,, 208 15-5 5-5 2-65 I" 9") 1-65 1. Linen ,, 12-0 9 4-3 2-7 1-9 Relative humidity during- above tests, !Q/i The wounds in each case were clean central cuts which severed the weft yarns only. Each fabric had received 5 coats of Eaftite, while each of the cotton fabrics had received 2 coats of P.O. 10 in addition. 31. Weathering Experiments. It has long been cus- tomary to make weathering tests, with the object of determining what alteration of properties of the doped fabric is brought about by exposure.. Primarily these tests were made for strength, but since the general in- troduction of P.O. 10, followed by that of pigmented varnishes and dopes generally, this determination has for aeroplanes become subsidiary to that of tautness. This development is due to the very marked protective power of such varnishes, and whereas the deterioration of aeroplane fabric was formerly extremely rapid, es- pecially in the summer months, under the present con- ditions it is almost negligible, at any rate in temperate regions. The possibility of loss of strength, however, remains an important matter in connection with tent fabrics, and probably also for aeroplane fabrics when machines are required for service in the tropics. 32. In December, 1914, Dr. A. J. M. Paget, then Senior Medical Officer of the Meteorological Department in Somaliland, wrote to the Superintendent, "R.A'.E., kindly offering to expose at Berbera, Somaliland, any samples which might be sent to him of fabric used or aircraft. Five series of samples were accordingly pre- pared and despatched in March, 1915: (1) Plain unbleached linen to E.A.F. Specification, No. 17A. (2) Series (1) doped with 4 coats of Cellon dope. (3) Series (2) covered with . 2 coats of V 114 (4) Plain unbleached linen, proofed. with a khaki coloured pegamoid varnish, to E.A.F. Speei- cation, No. 65 A. (5) Cuprammonium flax drill, for aeroplane tents. to E.A.F. Specification. No. 57. The samples were tested on their return to the E.A.E. and the results obtained confirmed in a striking manner the conclusions arrived at after the samples had been despatched to Berbera. viz., that the most active agent causing the destruction of doped fabric is sunlight, and that a khaki pigmeTitod varnish provided an efficient protection against this. 33. As a result of these tests it was deemed desirable to make such exposure tests on a more comprehensive scale* and to send to Somaliland further samples for this purpose. Dr. F. E. Whitehead, who had succeeded Dr. Paget at Berbera, kindly undertook the supervision of these exposures. At the same time parallel series of fabrics were prepared for exposure during similar periods at the E.A.E. , and at Malacca, where Dr. A. H. Keun, the Chief Medical Officer had kindly offered to conduct the exposures. Sets of samples were accordingly despatched in April, 1916, for exposure periods of 10, 25, 50, 100 and 200 days respectively. It was hoped that these parallel exposures would throw light on the effect of climatic differences on the deterioration of fabrics, the outstanding climatic differences being: Farnborough : temperate, moderate rainfall. Malacca: tropical, very humid (very large rainfall). Berbera: tropical, very dry (practically no rainfall). A little later a somewhat, similar series was despatched to Lieut. W. E. G. Atkins, 0. I/C. Experiments, Egypt, for exposure in that country. Unfortunately, however, some of the exposed samples, when ready for testing in 1917, were accidentally destroyed by fire, the Farn- borough series of 50, 100 and 200 days' exposures and the Malacca series of 10, 25 and 50 days' exposures being lost in this way. A second series was, therefore, exposed at Farnborough during 1917 (July, 1917-January, 1918), comprising somewhat similar fabrics to the first series. The following is a list of the fabrics: A. Plain unbleached linen. B. Ditto, covered with 4 coats of Eaftite and 2 coats of P.O. 10. C. Plain scoured cotton. D. Ditto, covered wth 4 coats of Eaftite and 2 coats of P.O. 10. E. Silk twill. F. Ditto, covered with 4 coats of Eaftite and 2 coats of P.O. 10. (r. Silk twill, oil-proofed. H. Unbleached flax drill, unproofed. K. Ditto, coated with P.O. 10. L. Unbleached flax drill, cuprammonium-proofed. M. Ditto, coated with P.O. 10. N. Unbleached flax canvas, unproofed. P. Ditto, bitumen-proofed. Q. Cotton duck, green dyed. E. Cotton duck, " grey " state. S. Cotton canvas, grey-dyed and Astoli-proofed. T. Cotton canvas, bitumen-proofed. IT. Cotton twill, olive-dyed, two-ply, with inter- mediate rubber layer. V. Plain ramie fabric. W. Unbleached flax canvas, unproofed. Z. Cotton canvas, olive-dyed and Astoli-proofed. 34. These fabrics, after being exposed, were returned to the E.A.E. for testing, and determinations were made of: (1) weight: (2) breaking strength; (3) wound strength; (4) rip strength; (5) extensibility and (6) watertightness (of proofed fabrics). It was found that the rate of deterioration was about the same at Malacca as a Berbera, being much greater at both places than at Farnborough. The extremely humid conditions prevalent at Malacca resulted in pro- lific mould growth, which appeared to be responsible for an appreciable amount of deterioration of the fabrics : * R.A.E. H. Dept.. Report No. H 875. 158 for the respective rates of deterioration of similar cotton and linen fabrics were reversed in Berbers and Malacca, from which it is deduced that cotton is less sensitive to light than flax, but more sensitive to mould. Silk rotted very quickly when weathered; jute, too, deterior- ated far more rapidly than either cotton or flax. The pigmented varnish, P.O. 10, while affording great pro- tection to the doped fabrics at Farnborough and Berbera was found to be of little avail at Malacca where the deterioration is presumably largely due to mould action. 35. The relations between the different fibres are illus- trated by the table below of results obtained from the second series of exposures at Farnborough. Some com- plication is introduced by the effect of cloth structure but the figures demonstrate the greater resisting power on the part of cotton, and also show the comparative inferiority of the other fibres : PERCENTAGE STRENGTH RETAINED AFTER 100 DAYS' EXPOSURE AT FARNBOROUGH. Threads - . Threads Aeroplane fabrics. -*s * '3 B C per inch. Tent Fabrics. 6x-e n 8 oc.a g 5 per inch. 5 or> ; K 'O o :* D = double. Thus the protection against weathering action afforded by the use of coarse yarns is considerable, indicating that the yarns used should be as coarse as possible, compati- ble with the desired compactness of the cloth. 38. The Effect of Weathering on Tearing Strength These comparative weathering experiments also threw some light on the effect of exposure on tearing strength. Three important deductions were made: (1) For any given fabric the wound strength bears a practically constant relation to the break- ing strength, the proportion being unaffected by the separate or joint action of proofing or weathering. A reduction in the proportion is, however, caused by doping. (2) For any given fabric, the ratio of the rip strength to the breaking strength in general falls gradually on weathering; i.e... the rip strength diminishes at a faster rate than the breaking strength. The relative rip strength remained practically unchanged by any of the proofings tested, although doping caused a reduction which was often considerable; on weathering, the nature of the proofing may exercise great influence on the fall in the rip strength; thus the relative reduction in rip strength of the bitumen-proofed flax can- vas (P) was comparatively small, and much smaller than that of tho same fabric un- proofed (N). 159 As showing how much more sensitive rip strength is to exposure than breaking strength, an outstanding example from the Farnborough (second) series may be quoted, viz., the unproofed flax canvas. After 200 days' weathering its breaking strength only fell to 80% of its initial value, while the rip strength fell to 25% of its initial value, and whereas the rip strength was initially 40% of the corresponding breaking strength, at the end of the exposure it was only 12%. (3) The order as regards resistance to tearing of fabrics made from different fibres remains unchanged by weathering, flax retaining its initial superiority. 39. The Effect of Weathering on Prooflngs. The various proofings were tested more particularly as to their suitability for tent fabrics. The pigmented nitro- cellulose, oil and cuprammonium proofings respectively were found to be unsatisfactory; the use of rubber introduced undesirable complications, while the bitumen, basic-aluminium acetate and Astoli proofings respec- tively were fairly satisfactory. The following notes show the basis of these conclusions: (a) Pigmented nitrocellulose proofing (P.O. 10). The normal use of this proofing is for doped aeroplane fabric. The amount of proofing given did not render the fabrics watertight, so that a large quantity would be needed to accomplish this, causing the cost to be prohibitive. Cotton Fabric (Mercer.), E/r = 44/42, 2-fold warp and welt. Weight (ozs./sq. yd.). Undoped, 4-27 ; Doped, 7-68 (58% R.H.). SUMMARY. (3" Wound Specimens.) Tearing Strength Tearing Strength % Extensibility (Ibs. /in.). of Breaking Strength. % Size of Wound. Warp. Weft. Warp. Weft. Warp. Weft. Undoped 74-6 74-9 100 100 16 24 i" 62-2 65-2 : 83 87 18 19 1" 54-9 5-9 73 76 19 18 I" 44-5 46-1 59 61 18 17 2" 23-3 23-7 31 32 17 17 1 Tongue (Ibs.) 48-0 49-0 64 65 Ripping (Ibs.) 27-5 27-5 37 37 Doped " 88-9 95 5 100 100 18 17 1" 57-2 67-6 64 71 16 15 1" 43-6 47-9 49 50 15 13 1" 29-9 36-1 34 38 12 11 2" 15-2 16-4 17 17 10 8 Tongue (Ibs.) 18-75 18-75 21 20 Kipping (Ibs.) 10-25 10-25 11-5 11 Tensile Strength (Wet). Warp, 75-2 Ibs./in. Mean E.H,, 66%. Weft, 77-31bs./in. (b) Oil proofing. Initially satisfactory as regards watertightness, on continued exposure fabric proofed in this way gradually loses its water- tightness, is attacked by mould, and, in the case of the vegetable fibre fabrics, quickly loses its strength. (R.A.E. Eeport, No. H. 792). (c) Cuprammonium proofing. The watertightness of the proofed fabric was satisfactory before exposure ; the action of weathering, however, soon caused the fabric to become leaky. Moreover, deterioration of the proofed fabric took place at a more rapid rate than that of the unproofed. (d) Bitumen proofing. The fabrics on which this proofing was tested displayed a good deal of irregularity. The watertightness test results were peculiar in that they pointed to a gradual improvement in watertightness on exposure; thus the 200 days exposure samples of the proofed flax and cotton from Berbera were each the best of their respective series. Possibly a maximum efficiency is reached after a particular interval, after which a decrease takes place; this was suggested by the fact that the 200 days exposure sample of proofed flax canvas from Malacca was less watertight than the 100 days exposure sample ; the reverse was true of the proofed cotton fabric. The effect of exposure on strength appeared to be about the same for the proofed fabrics as for the same fabrics unproofed. (e) Aluminium acetate proofing. This was applied to a fabric comparatively light in weight of which the watertightness was satisfactory initially and for some time subsequently, but tended to diminish gradually as the exposure increased. Better results would be expected with a heavier fabric. Deterioration in strength of the unproofed and proofed fabrics proceeded to about the same extent. (/) Astoli proofing. The cotton fabrics proofed by this process were the most satisfactory as regards watertightness ; they also retained their strength better than any of the other fabrics. This shows that the proofing at least has no injurious action, although the excel- lent performance of these fabrics must be attributed partly, at any rate, to the fabric itself which weighed about 21 ozs. per sq. yard and was made from very coarse yarns. 40. The above results which were obtained from the tests on fabrics exposed at Berbera, Malacca, Farn- borough and in Egypt afforded confirmation of conclu- sions which were arrived at as a result of other tests carried out at the R.A.E. during the interval which elapsed between the despatch of samples to their tropical destinations and their return. These other tests were the subject of R.A.E. Reports, Nos. H. 792 and H. 793. The former dealt with " The Effect of Weathering on Various Proofed Fabrics for Tents, etc." and the latter with " The Effect of Structure of a Fabric on its Resistance to Weathering Action." In these tests, besides the proof- ings already mentioned, two patent proofings were exam- 160 inud, via., Marshall's proofing and the Millerain proofing, which gave results as follows : ( used as wing-coverings, which, by virtue of their trans- lucent nature, would reduce the visibility of an aero- plane. Each of the samples had for its foundation an open material composed of linen threads, with 9 per inch in each direction. In one case a transparent paper and in the other a fine cotton fabric, had been placed over the meshwork and caused to adhere to it by the appli- cation of dope. These samples, however, did not yield much improvement over ordinary doped linen as regards visibility and tests made later with panels of a wing covered with the fine cotton fabric type of translucent fabric were so unpromising that the work was discontinued. The question of the possibility of rendering the wings practically invisible again cropped up early in 1917, and various tests were made on two lines : (1) A series of reinforced fabrics was made with meshworks of -J-inch. J-inch, 1-inch and 2-inch and with very fine and open groundworks (80s singles, 40 threads per inch warp and weft). (2) A number of experiments were made with cellon sheets. The reinforced fabrics represented the limit of dop- ability, but their visibility was much too great to allow of tlieir being recommended. The cellon sheet was very- brittle, and experience in its use for centre-sections showed that it was too unreliable for any more extended employment; attempts were made to reinforce it in some way (e.g., with linen threads) but these experi- ments too proved abortive, and in view of these failures the experiments on the invisible wing-covering were o,gain stopped. 45. Initial tensions in the attachment of fabric. (U.A.E., H. Dept., Keport No. H. 523.) These tests were undertaken in order to determine the influence of the initial tensions employed in the application of aero- plane fabric to the wings previous to doping. Samples .if fabric were, stretched and fixed to frames under ten- sions ranging from 1 Ib. per inch to 6 Ibs. per inch in each direction; for comparison samples were also 161 stretched by hand, which gives a tension approximating to that obtained with a tension of 2 Ibs. per inch in each direction. All the samples were then doped and deter- minations made in the unweathered and weathered states of (1) weight; (2) tensile strength; (3) tearing strength; (4) stress strain diagrams; and (5) tautness It was found that the initial tensions employed, if with- in the above range, had a negligible influence over these properties, and it was therefore concluded that the ordinary method of stretching the fabric by hand was quite satisfactory both in the covering of planes and in experimental work. APPENDIX I. PROPOSED SPECIFICATION FOE STANDARD LINEN AEROPLANE FABRIC. 1. QUALITY AND MANUFACTURE. (a) The fabric shall be made of good grade flax fibre, dew or water-retted. (4) The yarns shall be wee spun. (e) The yarns shall be boiled. (d) 1 he weave shall be a plain weave. (e) The fabric shall be uniform, and as free as possible from slabs, snarls, knots, loose ends and other defects of preparation, spinning, weaving and finishing. (/) The selveges shall be evenly and well made. (j) The fabric shall be heavily beetled, mangled or chested after weaving. (A) When the standard reference dope is correctly applied to the fabric, it snail form a smooth, coherent and properly adhering con- tinuous film, free from any tendency to brittleness. 2. SIZE-SOFTENING AGENTS. Where a size-softening agent or lubricant is used in dressing the warp it shall be i allow, palm oil or Japan wax, each free from paraffin wax and other unsaponifiable matter. No other size-softening agent shall be used. 3. WEIGHT. The weight of the fabric, including normal regain, (12 per cent.) shall not exceed 4 ozs. per square yard. 4. ENDS AND PICKS. The fabric shall have 50 ends per inch and 50 picks per inch. 5. WIDTH. The width of the fabric shall not be le^sthau 36 inches. A variation of not more than plus or minus J-inch shall be permitted in the width. 6. STRENGTH. The mean breaking strength of the finished fabric shall not be less than 75 lus. per inch of warp or weft when tested in the m inner described in Appendix A (below). APPENDIX A. Six specimens shall be cut from any selected part of the roll of fabric in such' a manner as to be representative of the material. Three specimens shall be cut in the direction of the warp and three in the direction of the weft. No two specimens cut in the same direction shall contain the same longitudinal threads. The specimens snail be 2J inches wide and the threads frayed out from each side to reduce the widih to 2 inches and shall be soaked in water for half-an-hour and the excess of adhering water drained oft. Each specimen shall be placed evenly in the jaws of a suitable testing machine so that the unstretched length of the fabric between the jaws is 7 inches, and shall be broken without delay. The load shall be applied at the rate of 150 Ibs. per inch width per minute. If a specimen breaks in or at the jiws at a load much lower than that required, or if a specimen " tears," a, duplicate test shall be made on another test specimen containing the same longitudinal threads. APPENDIX II. PROPOSED SPECIFICATION FOR STANDARD MERCERISED COTTON AEROPLANE FABRIC. 1. QUALITY AND MANUFAOTUKE. () The fabric shall be made from cotton fibre. (i) The length of the staple in the yarns shall not be less than 1 i inches. (<;) The yarns shall be 2-fold 38s. single or double combed. (d) The single yarns shall have 14-16 twists per inch and the 2-fold yarns shall have 23-27 doubling twists per inch. (e) The doubled yarns shall be mercerised under tension and the fabric shall conform with the tests described in Appendix A (below). (/) The weave shall be a plain weave. (g) The fabric shall be uniform, and as free as possible from slubs, snarls, knots, loose ends and other defects of preparation, spinning, weaving and finishing. (/) The selvages shall be evenly and well made. (i.) The fabric shall be heavily beetled, mangled or chested after weaving. (J) When the standard reference dope is correctly applied to the fabric it shall form a smooth, coherent and properly adhering continuous film, free from any tendency to brittleness. 2. SIZE-SOFTENINW AGENTS. Where a size-softening agent or lubricant is used in dressing the warp, it shall be tallow, palm oil or Japan wax, each free from paraffin wax and other unsaponifiable matter. No other size-softening agent shall be used. 3. WEIGHT. The weight of the fabric under normal moisture conditions shall not exceed 4 ozs. per square yard. 4. ENDS AND PICKS. The fabric shall have 50 ends per inch and 50 picks per inch. 5. WIDTH. The width of the fabric shall be not less than 36 inches. A variation of not more than plus or minus } inch shall be permitted on the width. 6. STBENGTH. The mean breaking strength of the finished fabric shall not be less than 75 Ibs. per inch width of warp or weft when tested in the manner described in Appendix B (below). APPENDIX A. A sample of fabric having an area of 36 square inches shall be taken from any part of the piece, and boiled in 200 c.cs. of neutral distilled water for 15 minutes. The volume of the water, on cooling, shall be made up to 200 c.cs. again ; 100 c.cs. shall be withdrawn, to which 0'5 c.c. of 1 per cent, neutral phenol phthalein solution shall be added. If the 100 c.cs of water withdrawn shall give an alkaline reaction, the amount of alkalinity therein shall be determined by titration with decinormal sulphuric acid solution, and shall not exceed O'OOS grams Na^O. If the 100 c.cs. of water withdrawn shall not give an alkaline reaction, the addition of 1 c.c. of centinormal caustic soda solution shall do so. APPENDIX B. Six specimens shall be cut from any selected part of the roll of fabric in such a manner as to be representative of the material. Three specimens shall be cut in the direction of the warp and three in the direction of the weft. No two specimens cut in the same direction shall contain the same longitudinal threads. The specimens shall be 24, inches wide and the threads frayed out from each side ta reduce the width to 2 inches and shall be soaked in water for half-an-hour and the excess of adhering waier drained off. Each specimen shall be placed evenly in the jaws of a suitable testing machine so that the unstretched length of the fabric between the jaws is 7 inches, and shall be broken without delay. The load shall be applied at the rate of 150 Ibs. per inch width per minute. If a specimen breaks in or at the jaws at a load much lower than that required, or if a specimen " tears," a duplicate test shall be made on another test specimen containing- tb same longitudinal threads. 27J64 162 APPENDIX III. LIST op CHIEF REPORTS ISSUED BT R.A.E. EXPERIMENTAL FABRICS DEPARTMENT 19151918. uate. ouujeuc. iveiereuce no. August, 1915 ... Proposed Programme of H. Report. No. 607 Research on Aeroplane Fabrics.' October, 1915 ... Fabric from German H. Report. No. 678 " Fokker " Aeroplane. October, 1915 ... Ramie Yarns H. Report. No. 700 November, 1915 . Weathering of Fabrics in A.U.A., R. & M. Somaliland. No. 224 December, 1915 . Three hand-made Ramie H. Report. No. 751 Fabrics. December, 1915 . Fabric from German "Alba- H. Report. No. 755 tross " Aeroplane. December, 1915 . Preliminary Tests on Non- Tear Translucent Fabrics. January, 1916 ... Silk Fabrics H. Report. No. 859 February, 1916... Effect of Various Wounds on H. Report. No. 875 Tearing Strength. March, 1916 ... Silk Yarns. Part I H. Report. No. 972 March, 1916 ... Visibility Tests on Non-Tear A.C.A., T. 676. Translucent Fabric. March, 1916 ... Contraction of Fabric A.C.A., R. & M. No. 225 March, 1916 ... Action of Air and Water on A.C.A., R. & M. Fabric. No. 223 March, 1916 ... Effect of Water-shrinkage of A.C.A., R. & M. a Fabric on the Component No. 221 Yarus. October, 1916 ... Ramie Yarns. Part II. H. Report. No. H. 396 Date. Subject. Reference No. October, 1916 ... Silk Tarns. Part II H. Report. No. H. 404 December, 1916 . Calendering of Aeroplane A.C.A.. R. & M. Fabric. No. 315 January, 1917 ... Linen Mat Fabric H. Report. No. H. 522 January, 1917 ... Influence of the Initial Ten- H. Report. sions employed, in the At- No. H. 523 tachment of Fabric. February, 1917 . Cotton Yarns H. Report. No. H. 540 March, 1917 ... Weathering of Fabrics in H. Report. Egypt. No. H. 576 September, 1917. Cotton Fabrics A.C.A., R. & M. No. 346 November, 1917. Seams in Aeroplane Fabric... H. Report. No. H. 718 May, 1918 Fabric Gasholders for supply- H. Report. No. 786 ing 1 Oxygen to Aviators. May, 1918 Weathering of Tent Fabrics... H. Report. No. H. 792 May, 1918 Weathering of Fabrics of H. Report. different structure. No. H. 793 July, 191S Tearing Strength of Fabrics, A.C.A., R. & M. etc. No. 487 February, 1919 . Weathering of Fabrics at H. Report: Farnborough, Berbera and No. H. 875 Malacca. THIS BOOK IS DUE ON THE LAST DATE STAMPED BELOW AN INITIAL FINE OF 25 CENTS WILL BE ASSESSED FOR FAILURE TO RETURN IIS BOOK ON THE DATE DUE. THE PENALTY WILL INCREASE TO SO CENTS ON THE FOURTH DAY AND TO S1.OO ON THE SEVENTH DAY SET JUN 14 1953 OCtg 2 KK ? ( M RECEIVED" DEPT - LD 2i-inom.7.'::'.n 102 YF 00258 THE UNIVERSITY OF CALIFORNIA LIBRARY