key: cord-0264007-tzckz3fv authors: Agapiou, John S.; Carlson, Blair E. title: Friction Stir Welding for Assembly of Copper Squirrel Cage Rotors for Electric Motors date: 2020-12-31 journal: Procedia Manufacturing DOI: 10.1016/j.promfg.2020.05.156 sha: 777e5767ad94807962d6a495ce681cec8d1bd4d7 doc_id: 264007 cord_uid: tzckz3fv Abstract The automotive industry is developing designs and manufacturing processes for new generations of electric motors intended for use in hybrid and electric vehicles. This paper focuses on solid-state welding to join copper end rings to copper spokes in the fabrication of copper rotors. Friction stir welding was explored to examine weldability of these copper components. Defect free welds were produced on 8-9.5 mm thick copper plates with copper spokes through slots at a travel speed of 225 mm/min and a tool speed rotation of 2250 rpm using lanthanated tungsten as the welding tool. The H13 and nimonic tool materials underperformed due to higher welding temperatures. The conditions were optimized by monitoring the tool and copper plate temperatures. Several tool geometries were investigated and the scrolled shoulder with tri-flat threaded pin performed the best. The strength and weld characteristics were sufficient for this application. Finally, this work puts forth major welding process for induction copper rotors. Significant efforts have been devoted in developing energyefficient and environmentally friendly technologies for hybrid vehicles. Vehicle electrification efforts demand efficient motors with increased power and minimal losses. Induction motors (IM) are becoming very important to the automotive industry exemplified by efforts to reduce dependency on high performance magnets [1] [2] . Induction traction motors have used aluminum in the rotor to conduct electricity due to its relatively low manufacturing costs and good electrical properties ( Figure 1 ). However, a copper rotor improves power density and efficiency. Copper, which has over 60% higher electrical conductivity than aluminum, is also much more difficult (i.e. costly) to process, given its significantly higher melting point (1080 vs. 660°C), thermal conductivity, density, and price-per-unit (the latter two by a factor of nearly 3.5). However, an increase of motor efficiency by 2 to 4% was obtained with copper rotors [1, 3] . Squirrel cages for induction motors are currently manufactured by aluminum die casting. This process includes direct casting of the aluminum into a steel lamination stack. Die casting of copper is more challenging than aluminum due to Significant efforts have been devoted in developing energyefficient and environmentally friendly technologies for hybrid vehicles. Vehicle electrification efforts demand efficient motors with increased power and minimal losses. Induction motors (IM) are becoming very important to the automotive industry exemplified by efforts to reduce dependency on high performance magnets [1] [2] . Induction traction motors have used aluminum in the rotor to conduct electricity due to its relatively low manufacturing costs and good electrical properties ( Figure 1 ). However, a copper rotor improves power density and efficiency. Copper, which has over 60% higher electrical conductivity than aluminum, is also much more difficult (i.e. costly) to process, given its significantly higher melting point (1080 vs. 660°C), thermal conductivity, density, and price-per-unit (the latter two by a factor of nearly 3.5). However, an increase of motor efficiency by 2 to 4% was obtained with copper rotors [1, 3] . Squirrel cages for induction motors are currently manufactured by aluminum die casting. This process includes direct casting of the aluminum into a steel lamination stack. Die casting of copper is more challenging than aluminum due to END RING Available online at www.sciencedirect.com Significant efforts have been devoted in developing energyefficient and environmentally friendly technologies for hybrid vehicles. Vehicle electrification efforts demand efficient motors with increased power and minimal losses. Induction motors (IM) are becoming very important to the automotive industry exemplified by efforts to reduce dependency on high performance magnets [1] [2] . Induction traction motors have used aluminum in the rotor to conduct electricity due to its relatively low manufacturing costs and good electrical properties ( Figure 1 ). However, a copper rotor improves power density and efficiency. Copper, which has over 60% higher electrical conductivity than aluminum, is also much more difficult (i.e. costly) to process, given its significantly higher melting point (1080 vs. 660°C), thermal conductivity, density, and price-per-unit (the latter two by a factor of nearly 3.5). However, an increase of motor efficiency by 2 to 4% was obtained with copper rotors [1, 3] . Squirrel cages for induction motors are currently manufactured by aluminum die casting. This process includes direct casting of the aluminum into a steel lamination stack. Die casting of copper is more challenging than aluminum due to the higher melting temperature and thermal conductivity of copper. An alternative is to produce the squirrel cage through an assembly of the bars and end rings which are joined together by a welding process. The high thermal conductivity of copper hinders the welding process because heat required for local softening at the weld sections rapidly diffuses. Thus, fusion welding of copper may require special fixturing and high temperature preheating of the copper end rings. On the other hand, friction stir welding offers a low-temperature, solid state process for the joining of materials. The use of inertia friction welding process was evaluated for this application [4] without success. The primary objective of this work is to evaluate the response of the assembly of copper squirrel cages for induction electric motors (Fig. 1) to FSW. Studies on copper are very limited and important questions remain about effective, economical tool materials and process parameters for such intricate part geometries along the joint. However, tool materials are readily available that are likely to produce acceptable FSWs in copper for the purposes of testing and analysis [5] [6] [7] [8] [9] [10] . An understanding on the effect of FSW parameters, tool geometry, their interaction, various welding strategies for this application, and their implications on joint characteristics is still evolving. The combination of tests using simple plates and rotor end rings, with and without bars, to select the process conditions and tooling allows a better understanding of the FSW process for rotor application. The W-La tool material with convex scrolled shoulder and threaded/tri-flat pin performed the best. This work helps define the process characteristics and limitations to assist the improvement in copper rotor manufacturing. FSW is not a mainstream joining solution within the automotive industry. The process is accomplished by using a rotating tool that is translated along two butt or lap surfaces (as illustrated in Fig. 2 ) to generate considerable amount of friction heat and mixing of the surrounding material that create a solidstate bond. The tool pin below the shoulder rotates within the surfaces while the shoulder makes firm contact with the top surface of the workpiece(s). The friction generated at the shoulder and the pin is converted into heat that reduces the flow stress of the surrounding material. This results in severe plastic deformation and mixing of the material under the shoulder and around the pin so that the material is extruded by the pin and forged by the shoulder into a joint. The pin is designed to stir the material and move it from the advancing side of the tool to the retreating side as the tool travels along the joint. At the end of the weld, the tool is retracted from the workpiece and a pin hole remains in the part. The maximum temperature reached during FSW is 80% to 90% of the melting temperature of the alloys welded. FSW has limited applications for welding copper such as a copper containment canister for nuclear waste [6] [7] and the backing plate for sputtering equipment [8] . Challenges still exist in terms of a narrow range of process parameters, tool material, and the limited shape of the joint. The material flow during the FSW process can be influenced by many factors such as tool geometry, machine tool welding parameters, hardness of copper end rings and bars, weld geometry, end ring temperature distribution during a circular weld, etc. [9] [10] [11] . The relation of the tool rotation speed and the traverse speed is very important to generate a defect-free weld while overcoming the rapid dissipation of heat within the copper workpieces. The material used in this research are tough pitch pure copper C11000 and C10100. The final rotor assembly is made up of a loose silicon steel lamination stack with copper spokes (bars) extending through slots in the outer periphery of the rotor (Fig. 3) . The stack was 72-mm tall with 56 slots in the laminations. Two copper end rings are used to join the spoke ends on either end of the lamination stack to complete the electrical circuit of the rotor (Fig. 4) . The end rings are made either as a single 9.5 mm copper plate (Fig. 4a ) or an assembly of four 2 mm thick copper sheets (Fig. 4b) . The end rings were made rectangular to simplify the fixture and to allow the FSW tool to exit the circular weld. The bar material was extruded and then half drawn. The tensile strength was approximately 320 MPa, and a hardness of 68 Rf. The geometry of the spokes is identical to the slot shape with a very small clearance. The copper plates and sheets with and without shorting bars (shown in Fig. 5 ) were used throughout tests. The tensile strength for all the copper material was in the range of 290 to 320 MPa, while the hardness ranged from 70 to 80 Rf. Tool Pin Retracting Side Pin Entry Author name / Procedia Manufacturing 00 (2019) 000-000 3 Tool geometry has a significant impact on the process performance with respect to the mixing of the material by the pin and the heating function of the shoulder. The uniformity of the weld microstructure and properties as well as process loads are controlled by tool design. The tool material and its geometry are also affected by the FSW workpiece material because the tool must have sufficient strength at elevated temperatures. The rotor end ring joint design is more complex than conventional joints (see Fig. 5 ); it is a joint with interrupted sections and involves both butt and lap joints. This presents a difficult problem since tool design is a critical factor in determining a successful weld. Variants of the available tool geometries were progressively developed in collaboration with the tool manufacturers for the 8 to 9.5 mm thick circular welds. Therefore, some of the standard tools in the market were evaluated and sub-optimized for welding copper material and the joint geometry for the end rings. Three tool materials were tested including H13, W-La (Lanthanated Tungsten -heavy metal), and Nimonic. Nimonic alloy 105 (W. Nr. 2.4634) is a wrought nickel-chromium-cobalt base alloy strengthened by additions of molybdenum, aluminum and titanium that is developed for service at temperatures up to 950 o C. Table 1 provides a more detailed description of the tool geometry. One of the above tool geometries and its characteristics are illustrated in Fig. 6 . The tool holder utilized to hold the tool in the spindle is illustrated in Fig. 7 ; it had a small hole on the side to pass a thermocouple for measuring the tool temperature. The FSW method uses a machine resembling vertical milling but the spindle holding the tool carries high torque and thrust loads. The welding tests were initially performed on a load- controlled FSW vertical machine with a spindle having a maximum rotation rate of 1000 rpm. However, most of the tests were performed in the MTS 5-Axis PDS (Process Development System) machine as illustrated in Fig. 8 which could be used in either the force or position-controlled mode. The spindle had roll and tilt capability to adjust the tool tilt in the vector direction of the tool path, necessary for circular welds. Several weld fixtures were used for welding: copper plate setup, end ring setup, and rotor fixture. The simple rectangular plates or laminated sheets were positioned directly on the machine table on top of a support plate. It was clamped with multiple clamps along its length to avoid any movement due to large forces and torque applied the FSW process as illustrated in Fig. 9 . The fixture was developed (refer to Fig. 10 ) to hold the rotor assembly in place to FSW one end ring at a time. The end rings were in a pocket and bolted in the fixture. to prevent it from spreading or lifting during welding. Tool temperatures were measured in a large number of tests by inserting a standard K-type thermocouple (T/C) through the holder in the tool shank 12 mm behind the shoulder as illustrated in a tool cross section, refer to Fig. 11 . The T/C junction was positioned at the bottom of a small hole at the center of the tool. The EMF signal was collected continuously using the TC-Link Wireless Thermocouple Node small sensor that was a complete, cold junction compensated, linearized system. The sensor was attached to the tool holder shank (as illustrated in Figs. 8 and 11 ). It used a 2.4 GHz IEEE 802.15.4 spread spectrum radio with high gain antenna. K-type T/Cs were positioned near to the weld to evaluate the temperature distribution during welding. 1 mm holes were drilled on the retreating side of a linear or circular weld at 100 mm intervals in a 400 mm long weld as shown in Fig. 9 . Likewise, the T/Cs were placed in the outer perimeter of the circular welds in the end ring ( Fig. 9 ). The sensing tip of the T/Cs was approximately 1mm and the holes used to accommodate the T/C have a depth of about 3-4 mm. Therefore, the T/Cs were securely embedded in the holes and they were placed as close to the tool shoulder as possible without being destroyed. A digital thermometer was used to connect the T/Cs to a personal computer that contained a data acquisition system for recording the temperatures during welding. Figure 11 . Tool with a hole to locate a thermocouple near the shoulder. The process parameters that were controlled for the FSW experiments included tool rotation speed, traverse speed, normal force on the tool, plunge depth of the tool relative to the top plate surface, and tool design. One goal of the activity was to identify optimum FSW parameter ranges. Studies have been presented using software simulation for the analysis of the effects of tool design and welding parameters on the amount of heat generated and process characteristics [5, 12, 13] . At present, the development of FSW is more of an art than engineering especially for copper material. Due to the complex mechanism of the FSW process, optimum values of the above parameters were determined solely experimentally. Three tool materials were initially evaluated. Two or three rows of either 250-or 400-mm long welds were performed across the width of the plates. The plate was cooled between welds. The rotating welding tool was slowly plunged into the plate (feed rate of 30 to 60 mm/min) until the tool shoulder forcibly contacted the upper surface of the material with a short dwell at the bottom (preheating time). The welding process was then performed along the weld path (as indicated in Fig. 1 ) at a specified traverse speed (ranging from 200 to 720 mm/min) until the end of the weld was reached on either a load-or position-controlled mode. The tools were tilted by 2 o to drive a smoother material flow. The welding tool was then retracted, while the spindle continued to rotate. The spindle speed for the above tests was either 750 or 1000 rpm. Some welds were evaluated using tension testing and cross-sectioned to determine weld quality. Based on results, a set of nominal welding parameters were selected for the subsequent tests. The testing continued using the best tool geometry from the previous tests and included further optimization of the process parameters. The tool and workpiece temperature histories were also considered during the optimization of the process parameters and tool performance. The linear welds were used to eliminate the spindle roll and pitch tilt necessary for the circular welds. Therefore, the welds in the strips were designed to be representative of the circular weld in the end rings; the circular welds were about 266 mm long. The surface area of the strips was similar to that of the square end rings with respect to heat conduction and dissipation. Likewise, the temperature history for the circular welds in the end rings were evaluated using the setup in Fig. 11 (b). In some cases, prior to drilling the holes for the T/Cs, the hardness of the Cu in the same location as the T/Cs was measured. After welding, the hardness readings in the same lateral position as the T/Cs, but on the other side of the weld, were measured. This would help us correlate the T/C measurements to the hardness readings. Metallographic specimens were sectioned perpendicular to the welding direction, polished and etched. Optical microscopy was used to characterize the material flow by the different tool geometries and the joints. The stir (heat affected) zone (HAZ) and the thermo-mechanically affected zone (TMAZ) were evaluated for possible voids along the weld interface. The Vickers hardness profiles of the joints were measured across the welding area in a cross-section. Tensile testing was carried out using an Instron-type testing machine. The tensile test samples were sectioned according to the standard from the linear weld plates. The selection of the proper FSW tool material for Cu workpiece material was the first objective. Several hundred linear welds in solid Cu plates with and without spokes were performed with the tools in Table 1 . The H13 tools, tested at 750 and 1300 rpm and 200-300 mm/min traverse speed along the weld, did not complete a full weld length of 250 mm. The tool pin was deformed and twisted, and in several cases, it separated from the shoulder within the weld. The observations suggested the tool material had been compressed in length and expanded outward in diameter. This was an indication that the tool traverse speed was too high for the flow stress of the Cu plate material. Furthermore, the process required additional heat in the weld zone to reduce the flow stress of the Cu material. The traverse speed was reduced to 22 mm/min to increase the heat input per unit time and reduce the flow stress of the Cu material; the tool didn't maintain strength and dimensional stability at higher temperatures due to thermal softening in the tool material. The H13 tool material didn't have the high deformation resistance at elevated temperatures (creep resistance) to prevent the pin from deforming or collapsing under the stresses by the extruded and plowing copper material within the weld zone. However, the tools performed well with shorter length welds (<100 mm) using lower traverse speeds (22 mm/min) because their dimensional stability was acceptable. Several cross-sections were evaluated for defects within the weld zone, such as cavity/voids or narrow tunnel, and the results did not raise any concern. The distinct microstructural changes in various zones were obvious in the macrographs. The stirred (nugget) zone was distinct but not the thermomechanical affected zone (TMAZ). It was a visually goodlooking weld with minimal flash and a solid stir zone along the full length of weld. In this case, the temperature response exhibits a 340 o C rise over the 100 mm weld length peaking at 370 o C at the tool shank. The low traverse speed decreased the chances of void formation since the higher temperatures in the weld zone resulted in lower flow stress and improved plastic flow of the Cu material. Cu plates with slots and spokes were also tested using H13 tools. The weld was poor because the temperature was insufficient to reduce the flow stress of the Cu spokes in the weld zone for good stirring characteristics. It is hypothesized that the clearance fit for the spokes in the end ring slots resulted in an airgap between the two mating parts, and being a poor conductor of heat, acted as a barrier to rapid heat transfer in the spokes. Therefore, the H13 tool material could not be utilized effectively to weld 8 mm laminated plates with spokes for this application. The Nimonic 105 tools with convex scrolled shoulder and either threaded/tri-flats or stepped spiral pin, initially tested at 750 and 1000 rpm and in the range of 240 to 720 mm/min traverse speed, produced reasonably good welds. At the lower rpm, the tool pin visually began to deform after two or three welds. There was a layer of material pushed at the advanced side of the weld and the pin had built-up Cu material at the front of each of the three flats. The tool at 1000 rpm performed better than the 750 rpm because it generated a larger amount of heat for preheating the plate during the tool plunge and dwell periods. The welds were visually good, and the appearance of the top surface is illustrated in Fig. 12 ; in some cases, there was flash on the advancing or both sides of the weld line which was easily removed. The thrust force was in the range of 23 to 27 kN and the higher force occurred at the lower traverse speeds. After five welds at 600 mm/min traverse speed and 1000 rpm with a thrust force of 24 kN, the tool pin was again twisted. It seemed that the induced heat in the weld zone was not high enough to reduce the flow stress of the Cu plate material sufficiently for the Nimonic tool to travel at higher traverse speeds. It would be preferable to use a higher rpm since the tool material has significantly greater temperature stability (about 950 o C) and creep resistance than the H13 tools. Metallographic analysis revealed that the welds made with the threaded/tri-flats pin tool #5 were good without defects in their cross-sections. Despite the good surface appearance with small amount of flash, metallographic analysis, refer to Fig. 13 , indicated that some welds made with Tool #6 had small void defects at the retreating side of the welds. However, the voids were not present in the welds made with the threaded/tri-flats pin tool, as opposed to those made with the stepped spiral pin tool. It could be that either some of the Cu pieces didn't flow smoothly between the soft material in the nugget zone and the Cu bulk or some of the material at the perimeter of the nugget zone was not stirred into the nugget zone with the stepped spiral pin geometry and possessed difficulty in deforming and flowing in the Cu bulk at the prevailing welding temperature. Contradictory to this, the threaded/tri-flats tool pin provided better stirring function of the material in the nugget zone. It seemed the voids occurred due to the higher flow stress of the material at the thermo-mechanical affected zone (TMAZ) and the effectiveness of the tool pin geometry. This indicated that the tool pin profile had significant effect during FSW. Furthermore, no distinct TMAZ was seen in Cu welds. The Nimonic tool #5 was also evaluated in plates with spokes (see Fig. 5(a) ). The tool was run at 1000 rpm and the traverse speed ranged from 480 to 720 mm/min. 20 welds were made and the appearance of their top surface was good. The macrographs in Fig. 14 revealed the effect of the welding process on plates with spokes. The cross-section of the welds didn't indicate any defects at the joints of the spokes with the plate. The threaded/tri-flats pin performed better than the stepped spiral pin in the above process conditions. A square butt joint configuration was welded using two plates adjacent to each other to verify the weld performance through tensile tests. The tensile direction is perpendicular to the welding direction. The cross-section of the welds, which appear similar, is shown in Fig. 15 . Also, Figure 15 shows the fractured specimen from one of the welds that occurred near the HAZ since the weld didn't penetrate through the full joint section. The tensile strength for the specimens from both welds was in the range of 236-242 MPa. These values were somewhat lower than the strength of the base Cu material but since the weld didn't penetrate through the thickness, it is considered a strong weld. The above tests indicate that higher rotational speeds should be considered to induce a higher amount of heat resulting in higher temperature in the Cu plate to improve the material flow and possibly the weld characteristics. The Nimonic tool evaluation followed at higher spindle rpm (using the bridge machine type). In this case, the temperature history in the plate was measured using five T/Cs along the 400 mm linear weld at 100 mm apart (layout of the setup is shown in Fig. 9(a) ). Prior to drilling the holes for the T/Cs, hardness measurements were made (five measurements per location) on the Cu plate in the same location as the T/Cs. The forge position for tool plunge was 6.1 mm at 35 mm/min with a forge force limit of 14 kN and the spindle at 2200 rpm. The traverse speed was 260 mm/min at the same rpm as the plunge. More than ten similar welds were created and the weld surface appearance was very similar among them all. The hardness was also measured after welding in the same lateral position as the T/Cs but on the other side of the weld (the retreating side of the tool). The forge distance and force were very consistent throughout the weld as shown in Fig. 16 . Initially, the forge force increased to 12.5 kN while the tool pin penetrated the plate and increased to 14 kN as soon as the tool shoulder contacted the upper plate surface. The forge force dropped to about 10-10.5 kN during the welding cycle since the flow stress of the material decreased with temperature. The spindle torque raised rapidly to 47 Nm during the tool plunge cycle and it reduced significantly during the welding cycle down to 35-37 Nm. The tool was very stable during the welding cycle because the variation of the X-and Y-forces was small except during the initial contact of the tool shoulder with the plate surface as illustrated in Fig. 16 . It is obvious from the graph that the initial contact of the tool shoulder on the plate was severe since the thrust force, torque, and side forces had the highest variation. The significant increase of the X-and Y-side forces was an indication of the tool orbiting during the initial contact of the shoulder with the part. The tool temperature response is provided in Fig. 17 . There is a significant temperature drop between the pin and the T/C location due to heat transfer from the tool in the tool holder and spindle. The measured temperature increased with time but didn't reach a steady state. There was a marked increase in temperature during the initial 40 mm of weld but still the actual temperature rise was most likely faster; this is due in part to the response time of the T/C and the location of the T/C away from the tool's pin and shoulder. The temperature response of the five T/Cs in the plate during welding is shown in Fig. 18 . The response near the tool shoulder is affected by the heat input in the plate and the loss of heat in the backing plate. The temperature rapidly increased as the tool passed through the T/C location and then decreased exponentially as expected. This was an indication that the heat transfer in the backing plate was significant. The measurement of the plate hardness before and after the weld by the location of the T/Cs is provided in Fig. 19 . The hardness of base Cu plate had a range of 72-75 Rf. However, the hardness near the weld zone at the advancing (AS) and retreating side (RS) exhibited slightly lower values in the range of 38-71 Rf. These results indicate that the weld zone has significantly lower hardness than that of the base plate. The Lanthanated Tungsten (W-La) tool (#4 in Table 1 ) was initially evaluated at 1000 rpm, 600 mm/min traverse speed, and a force control of 30 kN. The weld was good visually and a loose flash was present along the full length of the weld. Several identical welds were made and it was observed that there wasn't any concern with the pin deformation. Therefore, the feed rate was increased to 840 mm/min with 28 kN thrust force and the welds were good but the tool shoulder was covered with a copper burr. The feed rate was increased to 960 mm/min and the tool pin broke in the plate after approxi-mately180 mm of weld length and the force reduced to 24 kN. A new W-La tool was used and several welds were performed at 720 mm/min feed rate at 29 kN forge force. The weld observations pointed out that these welding parameters were good for the W-La tool as illustrated in Fig. 20 . In addition, 00 05 09 13 17 21 26 30 34 38 43 47 51 55 59 63 67 71 75 79 83 87 91 95 these welds suggested that the temperature of the plate is very critical to weld quality and the presence of flash. The flash was mostly generated when the plate was very hot after multiple welds were made sequentially in the same plate. In addition, the tool rpm and plunge feed rate parameters were very critical to preheating of the Cu plate to avoid breaking the tool either at the shank or the pin because the higher required torque could not be supported by the tool geometry. The testing of the W-La tool continued at higher spindle Speed of 2100 rpm and 340 mm/min traverse speed while performing 365 mm long welds. The plunge rate was 35 mm/min. Fourteen thermocouples were embedded in holes filled with hardened epoxy in the plate along the weld, seven in the AS and the remaining in the RS of the weld (as indicated in Fig. 21) . Figure 22 plots the measured temperature signals for the weld which increase as the tool passed by the T/Cs and reduce exponentially as the tool travels away from the T/Cs. The temperature rise for the T/Cs is not identical because their depth location in the drilled holes may have varied. There wasn't a marked increase in temperature prior to traveling in front of each T/C, possibly due to the high capacity of the plate to conduct heat away from the tool point in the subplate as was observed with the Nimonic tools. Likewise, the rise of temperature at the T/C locations near to the shoulder was not as high as expected due to the rapid heat dissipation in the subplate and clamps. The temperatures of the T/Cs for the second weld in the plate had similar response to that of the first weld. The temperature of the T/C in the tool shank was in the range of 600 to 650 o C at the end of the weld, an indication of high temperature at the tool pin and shoulder. The response was similar to that of the Nimonic tools in Fig. 17 at similar process conditions, as expected. The performance of the W-La was better than that of the H13 and Nimonic tools to sustain high temperatures throughout the full weld length as expected. The testing was shifted to end rings to further optimize the process parameters for circular welds similar to those that will be required for the rotor (Fig. 5(b) ). Thermocouples were utilized around the weld path since the heat transfer in the circular path would be different than the linear weld. The setup is illustrated in Fig. 9 (b). Solid and laminated plates with and without bars were used for the tests. The welds were performed at various rpm and traverse speeds to optimize the weld quality. The weld tool path is shown in Fig. 23 with approximately 60 degrees overlap before exiting the tool. The diameter of the circular FSW tool travel is 86 mm. Several welding parameters gave rise to welds having good surface appearance. Some exhibited more flash than others in the advancing and/or retreating side of the welds. The range of parameters were evaluated to reduce the amount of flash and generate uniform welds. Effectively, from Fig. 24 , which illustrates the surface of two welds produced with the different parameters (in a solid and laminated end rings), it can be observed that good surfaces with flash were produced. In spite of the good surface quality, metallographic analysis revealed that some welds had internal void defects especially in the welds produced at lower spindle speed. Nevertheless, as observed in the weld nugget cross-section macrographs refer to Fig. 25 , for higher rotational speeds, these defects disappeared. Some of the cross-sections exhibited a joint line remnant caused by the oxides of the plates, but no defects. Investigation of the temperature distribution during welding was performed using 14 thermocouples around a 109 mm circumference (refer to Fig. 2(b) ). The distance of the thermocouples from the center of the tool's pin was 11.5 mm and 4 mm from the outside diameter of the tool's shoulder. The tool started at zero degrees (Fig. 23) , travelled one full revolution, and overlapped for about 60 degrees before it exited. The temperature response for one of the welds in a plate with bars, at 2250 rpm and 225 mm/min conditions, is given in Fig. 26 . In this case, the temperature was rapidly distributed through the plate since the T/C #14, located at about 330 o in the circular weld path, was at 170 o C as the tool approached it. The temperature for all the T/Cs reached the range of 400 to 470 o C in the above weld and the temperature throughout all the 20 circular welded rotor joints ranged from 350 to 500 o C. The distribution of the temperatures indicate softening (lower deformation resistance) of the Cu bulk material along the tool path. Therefore, either the traverse speed or the spindle rpm should be changed along the tool path to compensate for the softening of the plate material. The traverse speed was changed in three segments along the circular path, 225 mm/min from zero to 180 o , 340 mm/min to 360 o , and 375 mm/min to 420 o during the overlap, while the spindle speed was 2200 rpm throughout the weld cycle. The weld quality was improved significantly using the above process parameters implying that the temperature responses of the T/Cs are quite steady and reasonable. Vickers hardness tests were conducted to evaluate the hardness distribution at several traverse cross-sections along the circular path. The hardness distribution was also measured at two depths down the stirring zone to evaluate differences within the nugget. Figure 27 shows the distribution and the left side is the advancing side, while the right side is the retreating side of the weld. This figure indicates no significant difference in hardness between the advancing and retreating sides and along the depth of the nugget. The hardness variation from side to side was within 8 to 15 Hv for each of the seven sections. They were very consistent among the two measured levels for each section and there was no distinct trend either within each Finally, the above selected weld parameters and W-La tool were performed in actual rotors. The fixture in Fig. 10 was setup in the MTS machine (Fig. 8) to FSW one end ring at a time. Several weld trials were made using rotors with 9.5 mm thick end ring plates and with 8 mm laminated end-rings. One of the weld trials is presented in Fig. 28 using laminated end rings. Several signals were recorded including the end ring temperature and their response is plotted in Fig. 29 . The torque increased during the plunging of the tool's pin in the plate and especially at the initial contact with the tool's shoulder (37 kNm). Then the torque decreased (20 kNm) because the plasticized material with reduced flow stress, reduced the load. The torque decreased to 16-17 kNm slowly after travelling the first 180 o in the circular path. The forge force was stable (9 kN) throughout the weld, except during the initial contact of the shoulder with the end ring (14 kN). The side forces were higher at the entry and exit process steps than along the welding path. The tool temperature (behind the shoulder) reached 500 o C at the end of the weld. However, the process parameters were adjusted properly because the tool temperature levelled off during the last one-third of the weld length. The tool temperature related well with the torque response. In addition, the temperature response for the T/Cs in the end ring was in the range of 300 to 350 o C as shown in Fig. 30 . In this case, the temperature was controlled within a narrow range throughout the tool path. The welded rotor after deburring the welded sections on the end rings, is depicted in Fig. 4 . The end rings were machined to the outside diameter of the rotor and then tested for some of the rotor characteristics. The rotor cage resistance was measured as an indicator of the resistance of the welds for the 56 bars in both end rings (112 individual connections); the resisti-vity of the rotor cage was 4-4.5 m that was as low as expected based on similar measurements of identical cast copper rotors. The weld quality was also evaluated by inductive measurement by rotating the rotor through a magnetic field (current is induced in the rotor bars). This rotor test evaluated defects at the welded ends of the copper bars with the end rings. The graph in Fig. 31 had uniform amplitude among the 56 bars which was an indication that all the bars were welded uniformly in the end rings. The end rings from several rotors were sectioned to evaluate the joints between the bars and the end rings. Figure 32 illustrates the cross-section of four welded bars with the laminated end ring. The welded sections didn't indicate any disconnections or joint defects between the bars and the laminated four 2 mm sheets forming the end rings. The lap weld between the four sheets was good and the butt weld of the bars with the end rings was defect free. The rotor in Fig. 4 was assembled on a shaft and machined to the finish diameter as illustrated in Fig. 33 . Its electrical performance from the no-load and locked rotor tests using the equivalent circuit was good. FSW of pure copper was successfully obtained for joining squirrel cage rotor using copper end rings and bars and acquired results are as follow: 1. Input FSW process parameters were adjusted to control the heat generation affecting the plastic flow stress of the copper material of the bars and the end ring. The investigation incorporated the thermal effects of circular welds to define the process window for this application for a defect free joint. The torque that determines the weld energy and forge force are important control parameters for generating a solid-state weld integrating the copper spoke ends into the rotor solid and laminated end ring conductor. 2. The thermal histories in the tool shank and copper end ring were determined experimentally during a FSW lap and interrupted butt joints configuration of 56 copper bars in slots within the laminated copper end ring. The appropriate temperatures of the end ring for a successful FSW process were identified to be between 350 o C and 450 o C. The tool temperature behind the shoulder was controlled to a maximum of 500 o C for the W-La tool material. 3. The tool geometry made a significant difference in the stirring of the copper material. The convex scrolled shoulder performed better with respect to material flow than the concave flat shoulder. The threaded/tri-flat tool's pin performed better than the stepped spiral with respect to stirring and filling the voids produced by the tool motions. More importantly, the tool material was critical for welding longer than 100 mm sections or at higher traverse speeds. The W-La was superior to H13 and Nimonic alloy because it maintained the tool shape throughout the weld. 4. No special preparation was necessary for the combined lap and butt configuration joint with interrupted sections. If the surfaces didn't have significant oxides, the end ring and bars can be easily joined together with FSW process. 5. The macrostructure of the weld zone either slightly softened or kept the same microhardness compared to base copper material. There was some variation of the hardness in the stir zone but it didn't affect the weld characteristics with respect to resistivity of the joints and the weld strength. The rotor in Fig. 4 was assembled on a shaft and machined to the finish diameter as illustrated in Fig. 33 . Its electrical performance from the no-load and locked rotor tests using the equivalent circuit was good. FSW of pure copper was successfully obtained for joining squirrel cage rotor using copper end rings and bars and acquired results are as follow: 1. Input FSW process parameters were adjusted to control the heat generation affecting the plastic flow stress of the copper material of the bars and the end ring. The investigation incorporated the thermal effects of circular welds to define the process window for this application for a defect free joint. The torque that determines the weld energy and forge force are important control parameters for generating a solid-state weld integrating the copper spoke ends into the rotor solid and laminated end ring conductor. 2. The thermal histories in the tool shank and copper end ring were determined experimentally during a FSW lap and interrupted butt joints configuration of 56 copper bars in slots within the laminated copper end ring. The appropriate temperatures of the end ring for a successful FSW process were identified to be between 350 o C and 450 o C. The tool temperature behind the shoulder was controlled to a maximum of 500 o C for the W-La tool material. 3. The tool geometry made a significant difference in the stirring of the copper material. The convex scrolled shoulder performed better with respect to material flow than the concave flat shoulder. The threaded/tri-flat tool's pin performed better than the stepped spiral with respect to stirring and filling the voids produced by the tool motions. More importantly, the tool material was critical for welding longer than 100 mm sections or at higher traverse speeds. The W-La was superior to H13 and Nimonic alloy because it maintained the tool shape throughout the weld. 4. No special preparation was necessary for the combined lap and butt configuration joint with interrupted sections. If the surfaces didn't have significant oxides, the end ring and bars can be easily joined together with FSW process. 5. The macrostructure of the weld zone either slightly softened or kept the same microhardness compared to base copper material. There was some variation of the hardness in the stir zone but it didn't affect the weld characteristics with respect to resistivity of the joints and the weld strength. Induction Motors with Die-Cast Copper, Motors with Die-Cast Copper Rotors Design and Comparison between IM and PMSM for Hybrid Electrical Vehicles Designing Squirrel Cage Rotor Slots with High Conductivity Inertia Welding for Assembly of Copper Squirrel Cages for Electric Motors An improved friction stir shear localization model and applications in understanding weld formation process in alloy Ti-6-4 Friction Stir Welding and Processing VII Proceedings of the First International Symposium on Friction Stir Welding Proceedings of the Third International Symposium on Friction Stir Welding The joint properties of copper by friction stir welding Effect of heat input conditions on microstructure and mechanical properties of friction-stirwelded pure copper Elucidating of rotation speed in friction stir welding of pure copper: Thermal modeling Finite element modeling of friction stir welding-thermal and thermomechanical analysis An analytical model for the heat generation in friction stir welding Author name / Procedia Manufacturing 00 Induction Motors with Die-Cast Copper, Motors with Die-Cast Copper Rotors Design and Comparison between IM and PMSM for Hybrid Electrical Vehicles Designing Squirrel Cage Rotor Slots with High Conductivity Inertia Welding for Assembly of Copper Squirrel Cages for Electric Motors An improved friction stir shear localization model and applications in understanding weld formation process in alloy Ti-6-4 Friction Stir Welding and Processing VII Proceedings of the First International Symposium on Friction Stir Welding Proceedings of the Third International Symposium on Friction Stir Welding The joint properties of copper by friction stir welding Effect of heat input conditions on microstructure and mechanical properties of friction-stirwelded pure copper Elucidating of rotation speed in friction stir welding of pure copper: Thermal modeling Finite element modeling of friction stir welding-thermal and thermomechanical analysis An analytical model for the heat generation in friction stir welding Induction Motors with Die-Cast Copper, Motors with Die-Cast Copper Rotors Design and Comparison between IM and PMSM for Hybrid Electrical Vehicles Designing Squirrel Cage Rotor Slots with High Conductivity Inertia Welding for Assembly of Copper Squirrel Cages for Electric Motors An improved friction stir shear localization model and applications in understanding weld formation process in alloy Ti-6-4 Friction Stir Welding and Processing VII Proceedings of the First International Symposium on Friction Stir Welding Proceedings of the Third International Symposium on Friction Stir Welding The joint properties of copper by friction stir welding Effect of heat input conditions on microstructure and mechanical properties of friction-stirwelded pure copper Elucidating of rotation speed in friction stir welding of pure copper: Thermal modeling Finite element modeling of friction stir welding-thermal and thermomechanical analysis An analytical model for the heat generation in friction stir welding The authors would like to thank Robert Szymanski at GM R&D for his great help on the experimental tests and data collections. The authors would like to thank Robert Szymanski at GM R&D for his great help on the experimental tests and data collections. The authors would like to thank Robert Szymanski at GM R&D for his great help on the experimental tests and data collections.