key: cord-0287139-5v2r1qb4 authors: Hasbrouck, C. R.; Fisher, Joseph W.; Villalpando, Manuel Rudy; Lynch, Paul C. title: A Comparative Study of Dimensional Tolerancing Capabilities and Microstructure Formation between Binder Jet Additively Manufactured Sand Molds and Olivine Green Sand Molds for Metalcasting of A356.0 date: 2020-12-31 journal: Procedia Manufacturing DOI: 10.1016/j.promfg.2020.05.056 sha: aa5a2631fe02755561e361148ce9ffbf8ac9a1de doc_id: 287139 cord_uid: 5v2r1qb4 Abstract The development of additive manufacturing in the 1980s led to a revolution in various metalcasting processes as early as the 1990s, including the use of polymer 3D printing for the manufacture of sand casting patterns. While additive manufacturing has entered many metalcasting processes, including both expendable and permanent mold processes, this paper specifically examines the sand casting process. ExOne’s development of binder jetting technology over the last 20 years has allowed designers in the sand casting field to directly 3D print sand molds for metalcasting. Use of binder jet additively manufactured sand molds allows for quicker turnaround time for testing prototypes, new gating designs, and for producing one-off parts. Additive manufacturing of sand molds for metalcasting may implement any of the same foundry sands as green sand processes, but uses a furan binder instead of the traditional mixture of bentonite clay and water. The use of a chemical binder had led to questions about the resulting dimensional capabilities and mechanical properties of the castings produced by the mold. This study investigated the dimensional tolerancing capabilities, surface finish, mechanical properties, microstructure, and defects present in identical castings made from both a traditional olivine green sand molding process and a binder jet additively manufactured silica sand molding process. It was concluded that binder jet additively manufactured sand molds are capable of either equal or better dimensional accuracy and tolerance capabilities than traditional olivine green sand molds. The olivine green sand parts had an average of approximately 1 μm better surface finish than the binder jet sand molds; however, it is likely that both the addition of sea coal to the green sand and the difference in final part color significantly affected this result. The mean hardness of the binder jet parts was 58.9 HBW with a standard deviation of 5.6 HBW, compared to the mean of the green sand parts of 47.7 HBW with a standard deviation of 7.2 HBW. The hardness findings were confirmed by the presence of a finer microstructure in the binder jet parts than the green sand parts. While both types of molds produced parts with defects, a greater variety of defects was evident in the olivine green sand molds. Porosity tended to move toward the surface of the olivine green sand parts, but was relatively evenly spread through the additively manufactured sand mold parts. Additive manufacturing has existed since the 1980s [1] , and its potential to revolutionize various metalcasting processes was discussed as early as 1995 by Hull et al. [2] . Additive manufacturing for production of sand molds has rapidly become more present in modern metalcasting processes during the last 20 years as ExOne has continued to develop industrial 3D printing systems using binder jetting technology [3] . Additively manufactured sand molds are appealing to foundries because they eliminate the need for a pattern [4] , thus significantly decreasing lead time and tooling costs associated Additive manufacturing has existed since the 1980s [1] , and its potential to revolutionize various metalcasting processes was discussed as early as 1995 by Hull et al. [2] . Additive manufacturing for production of sand molds has rapidly become more present in modern metalcasting processes during the last 20 years as ExOne has continued to develop industrial 3D printing systems using binder jetting technology [3] . Additively manufactured sand molds are appealing to foundries because they eliminate the need for a pattern [4] , thus significantly decreasing lead time and tooling costs associated Additive manufacturing has existed since the 1980s [1] , and its potential to revolutionize various metalcasting processes was discussed as early as 1995 by Hull et al. [2] . Additive manufacturing for production of sand molds has rapidly become more present in modern metalcasting processes during the last 20 years as ExOne has continued to develop industrial 3D printing systems using binder jetting technology [3] . Additively manufactured sand molds are appealing to foundries because they eliminate the need for a pattern [4] , thus significantly decreasing lead time and tooling costs associated 48th SME North American Manufacturing Research Conference, NAMRC 48 (Cancelled due to with patternmaking for green sand casting. Although multiple methods of additive manufacturing may be used to create sand molds [5] , binder jetting seems to remain the most popular process. Designing and producing patterns for traditional sand casting remains the most time consuming and cost-intensive step of the casting process, especially for parts with complex geometry [6] . Using additive manufacturing for sand molds allows for quicker turnaround time for testing prototypes, new gating designs, and for producing one-off parts [7] . Additively manufactured sand molds also allow for designs impossible with current green sand methods, including elimination of parting lines and drafts, incorporation of undercuts, novel sprue designs for cleaner castings, and embedded sensors for data collection during metalcasting research and development [6] [7] [8] [9] [10] . Additive manufacturing of sand molds for metalcasting may implement any of the same foundry sands as green sand processes, but uses a furan binder instead of the traditional mixture of bentonite clay and water. This raises questions about the dimensional tolerancing capabilities and resulting mechanical properties of castings created with additively manufactured sand molds. The thermal properties of a chemical-based resin may differ significantly from that of bentonite clay and water, resulting in different casting cooling rates and ultimate different mechanical properties. However, some studies comparing traditional sand molding methods to additively manufactured sand mold methods have found that the mechanical properties of castings produced by additively manufactured sand molds match or exceed those produced by traditional sand molding methods [11, 12] . In addition to concerns regarding mechanical properties, the potential difference in binder saturation levels, mold permeability, and green strength of the sand molds printed with a chemical resin may affect the resulting dimensional accuracy of the parts cast. Current available research lacks information about the thermal properties of binder jet additively manufactured sand molds. Pennsylvania State University is working to quantify the thermal properties of binder jet sand used for metalcasting [13] , and the current study aims to examine the resulting microstructure and dimensional tolerancing capabilities of the process compared to traditional green sand molds. While additively manufactured sand molds show great promise for implementation in traditional green sand processes, this work needs to be completed to assure foundries that the dimensional tolerancing and mechanical properties will not suffer by adopting this technology. This study investigates the dimensional tolerancing capabilities, mechanical properties, microstructure, and defects present in aluminum castings made from both a traditional olivine green sand molding process and a binder jet additively manufactured silica sand molding process. The goal of this study was to investigate the use of additively manufactured sand molds in metalcasting and compare castings produced with binder jet sand molds with identical parts created with a traditional olivine green sand method. A test block was designed with both internal and external features, as well as both round and sharp features. Twenty identical parts were cast with aluminum casting alloy A356; ten were made using additively manufactured sand molds and ten using olivine green sand molds. The parts were compared on the basis of dimensional accuracy and tolerance capabilities, surface roughness, defects such as porosity, and resulting mechanical properties and microstructure. Each step in the process is described more thoroughly in the following sections. The test block was inspired by the NIST additive manufacturing test artifact [14] ; however, it was necessary to consider the limitations of a sand casting process during its design. Because the tolerances of the additive manufacturing processes tested with the NIST artifact are far tighter than those of traditional sand casting, the part included similar features to the NIST artifact, but scaled up for use in a sand casting process. For example, the total tolerance of a one-inch dimension on a steel casting made with a sand mold process is 0.24 inches [15] , while the overall tolerance of high-end fused deposition modeling (FDM) systems is 0.007 inches [16] . Since the part was designed for direct comparison between two sand mold making methods, and not to test the capabilities of sand casting as an overall process, the features were created to be relatively large and non-demanding to ensure the parts filled completely and could be compared between processes without risk of misruns or incomplete features. Additionally, overhanging features, negative drafts, and similar features possible with additive manufacturing but impossible with sand casting were removed. Figure 1 shows all the features of the NIST additive manufacturing test artifact [14] . Drafts were implemented on all features to allow removal from sand molds and all overhanging features were eliminated. The part ultimately contained four types of geometric features: circular cuts (diameters 1.00 inch, 1.25 inch, and 1.50 inch), rectangular bosses (sides of 0.50 inch, 0.75 inch, and 1.00 inch), a triangular boss (sides of 1.00 inch and 2.50 inch), and a conic boss (diameter of 1.00 inch and height of 0.45 inch). The height of the external features and the depth of the internal features was for ease of removal from the mold and was not intended to be part of the measurements. The limitations of the traditional sand casting process also affected the additively manufactured components of this study. Although additively manufactured sand molds eliminate the need for a pattern [4] , traditional green sand methods still require the use of a pattern and its associated mold parting line and draft angles on the parts. Therefore for the present study, although unnecessary in practice, a traditional mold parting line and drafts were added to the binder jet sand molds to allow for a direct comparison to the parts manufactured with the traditional green sand process. The sand casting pattern was created using additively manufactured polymers; the test blocks were made with an ultraviolet-cured photopolymer in a stereolithography (SLA) process, and the gating and riser system parts were made with polylactic acid in a fused deposition modeling (FDM) process. The machines available to the authors in the additive manufacturing laboratory limited the size of the parts that could be manufactured for the pattern. All pattern parts were split with consideration for the parting lines of the sand molds. Since aluminium castings shrink upon cooling, a patternmaker's allowance of 1.3 was used to oversize the pattern dimensions [17] . In addition to the measurable features on the part, the solid part volume was made sufficient for determination of grain structure and potential extraction of tensile specimens. The original gating design included a central sprue, four parts, and two risers, as shown in Figure 3 (a). Hand calculations performed with Chvirinov's Rule [18] predicted the solidification of the risers after the castings. However, simulation of the pour with SOLIDCast showed that the gating and risers were predicted to solidify before the castings, as shown by the gray portions of Figure 3 An updated gating and riser design eliminated two of the test blocks, gated the risers directly to the castings, and increased the size of the gating and risers into the cope of the sand mold. Gating and riser sizes were tweaked until an optimal directional solidification of the configuration was reached. The updated gating and riser design is shown in Figure 4 . Solidification was modeled using SOLIDCast. The properties used for simulation of the casting material (A356.0) and the mold (silica sand) are shown in Table 1 . The thermal properties were automatically populated by the standard SOLIDCast materials database. The final gating and riser design allowed the risers to solidify after the castings and maintained enough heat in the gating to prevent a misrun. Figure 5 shows a screenshot from the solidification model video. The color bar on the right shows the critical fraction solidification time in minutes, or essentially how long it takes each portion of the part to solidify. The yellow color takes the longest to solidify and the dark blue takes the least amount of time to solidify. Mold filling was modeled using FLOWCast. FLOWCast is an add-on to SOLIDCast that allows for computation fluid dynamics modeling of casting mold filling. A fill simulation was run to ensure there were no points of obvious turbulence and that all parts of the mold filled prior to solidification. Figure 6 shows a screenshot from the mold filling model video. The color bar on the right shows the temperature within the casting as it fills. After the simulation of the gating and riser system yielded satisfactory results for mold filling and solidification, the molds to be binder jet printed were designed. The CAD models of the mold were designed to match the dimensions of the sand casting flasks available at the Pennsylvania State University foundry. The mold was split into cope and drag to match the green sand casting process. Additionally, a mold half alignment system was created using extruded pins and lowered notches. Figure 7 shows the CAD models for the (a) cope and (b) drag halves of the mold. The standard additive manufacturing workflow was followed for both the green sand pattern parts and the binder jet mold parts. Upon completion of any necessary modifications for the chosen additive manufacturing process, all CAD solid part files were converted to STL files through an export from SolidWorks. STL files were examined using Netfabb for errors and ensured to be "watertight." Once the STL files were verified, the build orientation and build plate layout was decided for both the SLA and FDM processes. Additionally, the slicing and layer thickness was considered for each process. The green sand pattern parts were modified to be printed on a Formlabs Form 2 stereolithography (SLA) printer, including support structures. A standard Formlabs grey photopolymer resin was used with a build layer thickness of 0.1 mm. To allow the parts to fit on the build plate and to prevent suction on the features, the SLA process required an angled build orientation and support structures, as shown in Figure 8 . Examples of the finished SLA parts are shown in Figure 9 . Upon removal from the machine, support structures were cut off and the parts were soaked for 10 minutes in an alcohol bath to dissolve any uncured resin. The bottom edges of the parts that contact the pattern matchplate were sanded with 100 grit sandpaper until the remaining support structures were flat. This did not impact any measured feature of the parts. The gating and risers were printed using 3 mm diameter blue PLA filament on a LulzBot TAZ 5 FDM printer, with 0.18 mm layer thickness and 10% infill. The finished FDM parts are shown in Figure 10 . The binder jet sand molds were printed on an ExOne S-MAX with a maximum build rate of 105 L/hr [19] . The layer thickness was 0.35 mm, with an average binder saturation level of 1.2%. The verified STL files were sent to the supplier and the finished sand molds were shipped back to the foundry at Pennsylvania State University; no post-processing outside of loose sand removal is performed on the printed molds. The sand used in this process is a foundry silica and the binder is a furan resin. An example of the finished mold halves is shown in Figure 11 . The castings were poured into two different types of molds: the additively manufactured silica molds and olivine green sand molds. To pack the green sand molds, the pattern first had to be finished. The SLA parts required minimal work to be pattern-ready; the small holes added to the parts to prevent suction during printing were filled with steel epoxy and sanded down until smooth. The filled holes were not on any areas of the parts that were to be measured. The FDM parts required considerable work before being glued to the pattern, including rounding with steel epoxy and being sanded until smooth. Various grit sizes were used to round out the risers; starting with 60 grit for fast removal of epoxy and working down to 320 grit for finish sanding. The FDM parts functioned only as gating and risers to the casting, and were therefore removed from the final part. The addition of a layer of epoxy was not significant in changing the cooling rate or dimensional accuracy of the casting. Both the SLA and FDM parts were attached to a metal matchplate using silicone, as shown in Figure 12 . Note that the gates from the risers to the parts as well as the sprue were added to the green sand molds manually before the pour. The binder jet sand molds were glued together using a foundry glue and allowed time to set before pouring. Virgin ingots of aluminum casting alloy A356.0 were used. A356.0 is an aluminum-silicon-magnesium alloy (7.0% Si and 0.35% Mg) considered to have highly desirable casting qualities, such as fluidity, and is heat treatable to increase strength [20] . Several days elapsed between the printing of the additively manufactured sand molds and the pour so that the resin had sufficient time to set. The green sand molds were used immediately after being made in the foundry. The green sand used in the Pennsylvania State University foundry has an average southern bentonite clay content of 9%, an average moisture level of 3.4%, an average American Foundry Society (AFS) grain fineness number (GFN) of 72, and a green strength of approximately 21 psi. The additively manufactured sand molds have an average furan binder saturation level of 1.2%, an average AFS GFN of 80, and a yield 300 psi transverse bar break strength on 1 in x 1 in test bars. All pours were completed with virgin ingots, and each ladle was heated to the same temperature of 704.4°C (1300°F) before pouring. The parts were left to solidify in the molds completely before being broken out. Solidification progress was checked by physically tapping the top of the risers until they were solidified. Since the risers were the last parts to solidify based on the solidification simulation, the castings were assured to be solid once the risers were completely solidified. The parts were then allowed to air cool for several hours before being quenched. An example of a finished casting is shown in Figure 13 . After quenching, the parts were removed from the gating with a band saw. The parts made with the binder jet sand molds were labeled AM1 through AM10, the AM designation signifying additive manufacturing. The parts made with the olivine green sand molds were labeled GS1 through GS10, the GS designation signifying green sand. Data was collected on the parts upon removal from the gating systems. First, the features of the parts were measured in multiple locations for dimensional accuracy and tolerancing using a digital caliper. Next, a slice of each part was removed with a band saw to aid in measurements of surface roughness and hardness. The portion of the casting removed for this testing is shown in Figure 14 . To compare surface roughness of the two processes, five different locations on each part were measured for surface roughness using an optical profilometer. Hardness was measured using ASTM standard procedures for the Brinell hardness scale; five hardness measurements were taken in the same location on each part. To examine the microstructure of the castings, two random parts from each casting process were selected to be sectioned and inspected. A piece was removed from each part from both the slowest cooled area and the fastest cooled area of the casting ( Figure 5 ). The pieces were removed, mounted, polished, and microscopically inspected using standard metallurgical procedures. The specifics of each measurement technique are discussed in further detail in their respective subsections within the Results and Discussion section. Several measurements of each feature on all the finished parts were taken with a Mitutoyo Digimatic caliper with a resolution of 0.0005 inches and statistically analyzed. Slices of each part were removed with a bandsaw for surface roughness measurements via optical profilometry as well as Brinell hardness testing. Additionally, small pieces were removed from the areas of both slowest cooling and fastest cooling, as determined by SOLIDCast analysis (Figure 5 ), from two random parts from each casting mold type. These samples were hot mounted in phenolic resin, metallographically polished, and microscopically examined for grain structure. Since all variables during the pour were kept constant, all measurements of the parts from each mold type were considered as a whole in their respective group instead of on the basis of individual parts. For all features except the cone height and the sides of the triangular boss, four separate measurements were taken based on where the nominal dimension was established on the surface of the part. When analyzing cone height and the sides of the triangular boss, it was only possible to take one measurement on each part. Using Minitab, the binder jet mold part dimensions (designated "AM") and the green sand part dimensions (designated "GS") for each feature were tested for equivalence of variances. If the variances were equal according to a multiple comparisons test, a one-way ANOVA was carried out on the data set. If the variances were deemed unequal by a multiple comparisons test, a Welch's ANOVA was carried out on the data set. An ANOVA analysis was used in place of a paired t-test and normality of the data was ignored per Norman's study on the "laws" of statistics [21] . Using standard statistical analysis methods, the null hypothesis stated that the means of the dimensions of the AM parts and GS parts were equal. A 95% confidence level was used. Therefore, any pvalue lower than 0.05 provides evidence against the null hypothesis and suggests that the means of the AM parts and GS parts are not equal. The results of the statistical analysis for each measurable part feature are shown in Table 2 on the following page. Only two features had the same mean dimension for both the AM and GS methods: the largest circular cut and the smallest square boss. For these two features, the AM parts were found to have a mean closer to the nominal dimension than the GS parts. The AM parts had a smaller standard deviation than the GS parts for all but three features: the smallest circular cut, the height of the conic boss, and the long side of the triangular boss. However, for those three features, the variances were found to be equal through a multiple comparison test. It can be reasonably concluded that binder jet additively manufactured sand molds produce either the same or better dimensional accuracy and tolerance capabilities as traditional olivine green sand molds. It is important to recognize that having a physical pattern adds an extra amount of tolerancing issues. The SLA process used to produce the pattern parts of interest is considered highly accurate with very good surface finish compared to other additive processes due to its exceptional resolution. However, removal of the pattern from the molds and re-assembly of the cope and drag can lead to mold collapse and dimensional issues depending on the force used to remove the pattern and how much the pattern shifts during removal. Therefore, the elimination of the pattern with binder jet sand molds may inherently lead to better dimensional tolerancing capabilities. The same portion of each part was removed with a band saw for ease of measuring surface roughness through optical profilometry. Using a Zygo Zegage optical profilometer, five measurements of surface roughness were taken for each part. A 10X Mirau lens was used with the optical profilometer and was calibrated with a standard silicon carbide reference flat. An example of the surface profile of a part is shown in Figure 15 . The roughness measurement recorded was the average roughness evaluated over the complete 3D surface sampled, or Sa. The variances in surface roughness of the AM and GS part groups were found to be unequal based on a multiple comparisons test. Therefore, a Welch's ANOVA was carried out on the data sets. The null hypothesis stated that the means of surface roughness of the AM parts and GS parts were equal and a 95% confidence level was used. The resulting p-value of 0.031 suggests that the means of the surface roughness of the AM parts and GS parts are not equal. The mean surface roughness of the AM parts was 15.4 μm with a standard deviation of 3.3 μm, compared to the mean of the GS parts of 14.2 μm with a standard deviation of 1.8 μm. Therefore, the parts made with the olivine green sand process were found to have both a smoother and less variable surface finish than the parts made with the binder jet additive manufacturing process. In addition to the difference in types of sand used (silica for binder jetting and olivine for green sand), there are a few additional points worth noting about this conclusion on surface roughness. Pennsylvania State University's foundry sand contains an additive of sea coal to improve surface finish. Sea coal generates a gas film upon contact with hot metal, which keeps the sand from adhering to the casting and thus improving surface finish [22] . However, the addition of the carbon in the coal produced a dark, dull surface on the green sand parts compared to the shiny, bright surface of the binder jet parts. This is illustrated by Figure 16 . Since optical profilometry measures surface roughness through interferometry, it is important to note that measurements may differ between dull surfaces and shiny surfaces of the same baseline roughness. This is especially well-highlighted in the range of data collected for the AM parts in this study; the highest value measured was 23.1 μm and the lowest was 5.6 μm, a range of values of over 17.5 μm. Green sand casting processes are known to have a rougher surface finish than other types of casting, which is confirmed by the data collected in this study. Sand castings that require particularly tight tolerances or smooth surfaces are typically oversized to allow for a machining pass to create the final dimensions and surface finish. Brinell hardness testing was performed according to ASTM E10-18: Standard Test Method for Brinell Hardness of Metallic Materials [23] . Testing conditions of 10 mm/500 kgf were used. Calibration was checked with a 339 ± 10 HBW gauge block at 10 mm/3000 kgf. A Brinell optical scanning system (B.O.S.S.) was used to measure the diameter of the indents. The variances in hardness of the AM and GS part groups were found to be unequal based on a multiple comparisons test. Therefore, a Welch's ANOVA was carried out on the data sets. The null hypothesis stated that the means of the hardness of AM parts and GS parts were equal and a 95% confidence level was used. The resulting p-value of 0.000 suggests that the means of the hardness of the AM parts and GS parts are not equal. The mean hardness of the AM parts was 58.9 HBW with a standard deviation of 5.6 HBW, compared to the mean of the GS parts of 47.7 HBW with a standard deviation of 7.2 HBW. Therefore, the parts made with the binder jet additive manufacturing process were significantly harder than those made with the olivine green sand process. For metallographic analysis, the areas on the cast parts where both the slowest cooling rate and fastest cooling occurred were determined with SOLIDCast solidification analysis ( Figure 5 ). Small samples were cut out of each area of two random parts from each casting mold type; specifically sampled were AM1, AM10, GS2, and GS5. These samples were hot mounted in phenolic resin and polished with standard metallographic procedures. Examination of the grain structure of each sample under a metallographic microscope yielded the pictures shown in Figures 17 and 18 . All micrographs were taken at the same magnification and show a 200 μm scale bar in red in the bottom right corner. Figure 17 shows the grain structure of the area of the castings with the slowest cooling rate. Figure 18 shows the grain structure of the area of the castings with the fastest cooling rate. In all samples, microporosity is visible and the secondary eutectic silicon phase can be seen forming between grains. It is worth noting that secondary phases in the structure of aluminum castings does not limit the resulting mechanical behavior significantly when nonmetallic inclusions or microporosity are also present [24] . The AM parts exhibited a finer grain structure than the GS parts in both areas examined. This further supports the hardness measurements from section 3.3. It can be concluded that the cooling rate for the binder jet additively manufactured sand molds is faster than that of the olivine green sand molds. It is known that the specific heat capacity of olivine is similar to that of silica [25] . Therefore, the methods of binding the sand must be the factor that changes the heat transfer rate of the molds. In green sand molds, the binder is a combination of bentonite clay and water. The vaporization of water during the period of mold heating upon contact with molten metal strongly influences thermal conductivity of the mold, but the thermal properties appear to remain mostly constant after this period [26] . The thermal conductivity value of green sand stabilizes around 0.5 W/m-K in temperatures between 100-500°C [26] . In binder jet additively manufactured sand molds, the binder is a furan resin. Bate's study of the thermal properties of binder jet additively manufactured sand molds [13] found that, according to simulation results, the thermal conductivity of these molds should lie between 0.29 W/m-K and 0.42 W/m-K. However, these theoretical results directly conflict with the experimental results obtained in this study; the smaller grains present in the parts cast with the additively manufactured furan-bonded sand molds indicate that the thermal conductivity of these molds is higher than that of traditional bentonite clay-and water-bonded green sand molds. Bate also found that the amount of furan binder present in the sand molds did not significantly affect the casting solidification time or mold heat transfer rate [13] . Therefore, the authors suspect that only the type of binder used affects the heat transfer rate in sand molds, but not the amount. The results of this study suggest that further research into the thermal properties of binder jet additively manufactured sand molds is necessary to get an accurate comparison to traditional green sand mold thermal properties. Both types of the casting mold types showed defects in the final parts. Most noticeable was the porosity present in all of the parts. The olivine green sand parts showed porosity mainly toward the surface of the parts while the binder jet sand parts showed porosity throughout the entire part. This is shown in Figure 19 . Porosity is undesirable in castings because it greatly decreases the mechanical properties of the part. Porosity decreases the density of a part and subsequently decreases strength and hardness by introducing voids in the metal matrix. It also decreases strength by acting as stress concentrators and crack initiation sites during loading. The difference in location of porosity may have multiple causes. For example, it is known that sand cores, which are used in green sand casting and may use a furan binder or similar chemical binder, tend to off-gas during casting [27] . In fact, in a similar study, Snelling et al. saw significant off-gassing in binder jet additively manufactured sand molds [28] . This additional gas created may have promoted additional porosity in the additively manufactured sand mold parts. Additionally, a slower cooling rate of a casting, such as that seen by the olivine green sand parts, allows time for gas bubbles to rise to the surface of a casting. Both types of molds had issues near the parting lines. The green sand matchplate had a slight pattern misalignment, which can be seen in Figure 20 . The binder jet sand mold parts had considerable issues with flash at the parting line. This is most likely due to the alignment pin system used and differences in levels of the foundry glue used. An example of parting line flash is shown in Figure 21 . Several defects were only seen in the olivine green sand mold parts. These defects include a generally bad fill of surface features, large pores in the surface features of the castings, and areas of mold collapse. Each of these issues can be significantly attributed to the lower bonding strength of the olivine green sand molds than that of the binder jet sand molds. These defects are shown in Figure 22 . This study investigated the dimensional tolerancing capabilities, mechanical properties, microstructure, and defects present in castings made from both a traditional olivine green sand molding process and a binder jet additively manufactured silica sand molding process. It was found that binder jet additively manufactured sand molds produce either the same or better dimensional accuracy and tolerance capabilities as traditional olivine green sand molds. The olivine green sand parts had an average of approximately 1 μm better surface finish than the binder jet sand molds; however, it is likely that the addition of sea coal to the green sand significantly affected this result. Both the higher hardness values and the finer microstructure of the binder jet sand mold parts indicated a faster cooling rate than that of the olivine green sand parts. While both types of molds produced parts with defects, a greater variety of defects was evident in the olivine green sand molds. Porosity tended to move toward the surface of the olivine green sand parts, but was relatively evenly spread through the additively manufactured sand mold parts. The Ultimate Guide: Binder Jet 3D Printing Metal for Manufacturing Rapid prototyping: Current technology and future potential Patternless sand mold and core formation for rapid casting A review on additive manufacturing of sand molds by binder jetting and selective laser sintering Re-thinking design methodology for castings: 3D sand-printing and topology optimization A review on additive manufacturing of sand molds by binder jetting and selective laser sintering Application of Six-Sigma DMAIC methodology to sand-casting process with response surface methodology Novel Sprue Designs in Metal Casting via 3D Sand-Printing Measurement of Metal Velocity in Sand Casting during Mold Filling The Effects of 3D Printed Molds on Metal Castings Comparative Study of 3D Printing Technologies for Rapid Casting of Aluminium Alloy FUNDAMENTAL STUDY ON 3D SAND PRINTED MOLDS: METAL FLOW AND THERMAL PROPERTIES National Institute of Standards and Technology SFSA Supplement 3: DIMENSIONAL CAPABILITIES OF STEEL CASTINGS Accuracy in Additive Manufacturing Design for Casting Theory of Solidification of Castings Sand 3D Printers Properties and Selection of Cast Aluminum Alloys Likert scales, levels of measurement and the '"laws"' of statistics ASTM E10-18 Standard Test Method for Brinell Hardness of Metallic Materials The Influence of Microstructure on the Fatigue and Fracture Properties of Aluminum Alloy Castings MOLD SAND, SAND ADDITIVES & SAND PROPERTIES SAND PROPERTIES: Collection and Selection Thermal Properties of Foundry Mould Made of Used Green Sand Core-making in the Foundry: Cores support complexity in sand casting Mechanical and material properties of castings produced via 3D printed molds This work would not have been possible without the cooperation of many individuals across several locations. The authors would like to acknowledge the following people from Pennsylvania State University: Mr. Santosh Sama for his assistance and advice with running simulation software and preparation of sand mold CAD files; Dr. Sanjay Joshi and Ms. Xinyi Xiao for instruction on using the machines in the additive manufacturing laboratory; Dr. Guha Manogharan for supplying literature regarding additively manufactured sand molds for metalcasting; and the Factory for Advanced Manufacturing Education Laboratory technicians Mr. Chris Anderson, Mr. Travis Richner, and Mr. Brent Johnston for their constant support throughout the process of creating and collecting data on the parts studied for this research. The authors would like to thank the following undergraduate research students at Penn State Behrend: Mr. Christopher Lang for his help hot mounting the samples for metallurgical analysis, and Mr. Matthew Gielarowski for his work polishing the samples for metallurgical analysis. Thanks also to those outside of the Penn State system, including Mr. Brandon Lamoncha from Humtown Products for producing the additively manufactured sand molds, Mr. Tanner Mengle from the Materials Characterization Laboratory for the preliminary optical profilometry work, Mr. Chris Jabco from the Materials Research Institute for cutting the samples for metallurgical analysis, and Mr. Rob Hasbrouck of R/H Specialty & Machine, Inc. for his help preparing the metalcasting pattern and cutting samples for surface profilometry.