Abst Head – Abstract Heading

2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages)

Qualification Case Study for Chemically Bonded Sand System

P. Patel, L. Wells, S. Ramrattan Western Michigan University, Kalamazoo, MI

ABSTRACT INTRODUCTION

Chemically bonded sand systems have been used to BACKGROUND manufacture complex near-net-shape castings for more The United States is the third largest supplier of castings than sixty years. Over this period, many foundries have in the world and the metal casting industry is the sixth substantially invested in advanced technologies for largest industry in the United States, according to melting and filling processes, such as automated pouring American Foundry Society.1 Metal castings are essential systems, spectrometers, etc. Despite these investments, to a majority of industries such as defense, automotive, the casting industry still suffers from high variability and construction, agriculture, aerospace, oil and gas, mining, scrap/rejection rates. One cause of these quality issues is railroad, transportation, and health care. Casting quality is the inability to effectively monitor the quality of incoming a major contributor to the success of these industries. sand-binder systems. Most foundries are dependent on their sand-binder suppliers to detect differences in Casting quality losses and rejection rates are directly incoming sand-binder systems as foundries expect to related to foundry operational costs, such as; labor, receive consistent sand-binder systems. This dependency energy, raw material, and delayed customer delivery. In stems from the fact that foundries use traditional addition, sand consumption and cost continue to increase. chemically bonded sand tests, such as the hot tensile Metal casting industries use about 9% of the industrial strength test, which suffer from excessive levels of sand for foundry purposes.2 Figure 1 shows the dramatic variability that makes them insensitive sand-binder increase in unit price of sand and gravel3 over the last system shifts. century. In addition, since sand is the second most widely consumed natural resource4, there are some concerns Recent research has shown that disc-shaped specimen about sand shortages in the future.5 It can be considered a tests can detect wide range of differences in sand-binder responsibility of all industries in the world to minimize systems. In this paper, a qualification methodology for waste of sand. Foundries can minimize sand waste in chemically bonded sand systems is proposed, that focuses several different ways, such as the widely implemented on combining casting quality to statistical process control recycling or reusing of sand. However, it is impossible to for chemically bonded sand-binder systems. An achieve 100% sand reclamation, as losses are inevitable in implementation of the qualification methodology is casting and recycling processes. A second approach presented through a case study. Principal component towards minimizing sand waste is to reduce rejection analysis (PCA) was applied on thermal distortion test rates. If castings are acceptable every time a foundry (TDT) data to monitor an in-control resin coated sand- pours molten metal into molds, then indirectly, sand waste binder system with a control chart. In this study, two is reduced. different resin coated sand2 -binder systems with the same binder type and binder𝑇𝑇 level as the in-control system but Sand is majorly used either as a green sand or as a different sand origins were considered as out-of-control chemically bonded sand in foundries. Green sand is a systems. These sands are round grain silica sands with mixture of sand, clay, and moisture. Several practices similar silica content, grain fineness number, and sand have been identified to control green sand mold making grain shape. In this case-study, if a sand-binder supplier process. These practices include sand testing, analysis of uses these sands interchangeably, the control chart variance, linear regression, Taguchi method, simulation, detects the out-of-control system while 2the tensile and particle physics. In the United States, green sand strength test cannot. Casting trials suggests𝑇𝑇 that casting testing is used for sand control and these tests can be surface quality from these three sand-binder systems are prioritized according to a foundries need.6 Jacobson significantly different for gray iron, but it is similar for introduced a method of mass balance which calculates aluminum. Therefore, with respect to this case study, the how much new clay should be added for a specific job.7 ability to detect this shift (for a gray iron system) would In addition, a data-driven modelling approach has been help prevent surface-quality related losses. developed that provides rejections predictions, to assist in sand control, based on historic data.8 Despite having Keywords: Qualification, resin coated sand system, many different approaches, foundries still struggle to thermal distortion test, quality control, principal maintain green sand casting quality, primarily due to of component analysis. dynamic nature of a green sand systems.

Page 1 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages)

Unit Value ($/t) 30

0 1910 1914 1918 1922 1926 1930 1934 1938 1942 1946 1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 2010 2014 1902 1906 Years

Figure 1. Unit price of sand and gravel.

In contrast to green sand, chemically bonded sand obvious uncommon sand systems changes, which restricts systems use a chemical binder(s) and catalyst to cure and its effectiveness for use in statistical process control harden a mold/core. There are different chemically (SPC) to detect small sand-binder system changes10, 11. bonded sand processes, such as hotbox, cold-box, no bake, injection transfer molding, and 3D printing. Each of In contrast, recently adopted AFS standardized, disc- these processes have a specific binder(s) used to make shaped specimen tests, such as; thermal distortion test, molds/cores. Chemically bonded sand is widely used for abrasion test, impact test, and hot permeability, developed manufacturing molds/cores because of its high at Western Michigan University, collect data regarding productivity and dimensional accuracy compared to green independent as well as coupled thermal, mechanical, and sand. physical san-binder system properties. In addition, these tests have shown reduced specimen-to-specimen and test- The quality of chemically bonded sand casting is highly to-test variability, compared to traditional tests. Previous sensitive to the sand-binder system being used to make research has also shown that these disc-shaped specimen molds/cores. When a chemically bonded sand mold/core tests are able to differentiate various sand-binder is exposed to molten metal, unnecessary casting defects systems12. Finally, research has shown that an SPC can occur if appropriate sand system was not used. method, based upon principal component analysis (PCA), Therefore, it is important to detect shift (identify can be very effective in detecting small shifts in assignable causes for changes) in a sand-binder system. chemically bonded sand systems13. The premise of this work is that testing procedures that determine chemically bonded sand system properties can A new concept of qualification for chemically bonded also be used to detect shifts in a sand-binder system. sand systems was proposed in 201714. In the following section, the quality control framework component of the In the United States, the American Foundry Society’s qualification process is presented. The use of this (AFS) standardized sand tests, such as tensile strength framework will be demonstrated through a case study. test, loss on ignition, disc-transverse test, etc., are widely used to monitor for changes in a chemically bonded sand QUALITY CONTROL FRAMEWORK system. These tests are primarily based on independent The proposed quality control framework is shown in Fig. physical, mechanical, chemical, and thermal sand-binder 2. The objective of this framework is to combine SPC properties. However, sand-casting processes are with casting trials to ensure sand-binder system shifts do inherently thermo-mechanical, thermo-chemical and not affect quality. After detecting a sand-binder system thermo-physical. Foundry engineers have recognized that shift, a foundry needs to verify that this “new” sand- certain AFS standardized sand tests provide limited binder has no effect on casting quality, as not all process information regarding sand-binder system behavior, shifts negatively quality. In this paper, the quality control which restricts engineers from successfully controlling framework adopts a surface defect casting trial that uses sand-binder related quality losses.9 In addition, tests such disc-shaped specimens to validate casting quality and as hot tensile strength has high variability in their “qualify” a chemically bonded sand system. A sand- measured data,10 which makes them insensitive to small binder system is deemed qualified if casting trials are free or medium-sized process shifts. Previous research has of a specific defect under given process parameters. The shown that this test (hot tensile strength) can only detect proposed quality control framework can help foundries,

Page 2 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages) through a systematic process, efficiently administer a on SPC to identify any shifts that cause the sand-binder mold/core making process. systems to become unqualified.

The quality control framework shown in Fig 2 is, for the most part, self-explanatory. The important aspect of this framework is to perform diagnostic casting trials when a possible sand-binder system shift is detected. Foundries can use disc-shaped specimen tests to perform the SPC component of the framework. For this case study, an SPC method based upon PCA which only relies upon results for the TDT13, is considered. If a potential process shift is detected, a casting trial(s) is performed to observe the effects, if any, of this change on casting quality. Casting trials used in this step will maintain constant casting parameters, such as; metal chemistry, pouring temperature, and height of pouring. If this casting trial generates acceptable result, then this “new” sand-binder system should be considered as qualified, as it will not deteriorate the quality of the casting. Furthermore, if acceptable casting results from this trial and the foundry experiences worsening quality in actual production, then the casting trial can be considered a diagnostic tool that suggests other factors, such as; improper pattern or tool design, improper gating design, improper melting and pouring practices, or improper metal composition are the sources of quality loss.

METHODOLOGY

The effectiveness of the proposed quality control framework is exemplified through a case study. In this study, three sand-binder systems with same sand type (round grain silica), GFN (60), binder type, binder level (3% resin coated shell) were considered. The only difference between these systems is the sand’s origin. In this paper these three sands are denoted as X, Y, and Z. Sands X & Y, X & Z, and Y & Z originated about 65, 170, and 140 miles apart, respectively. Foundries, or sand-binder suppliers, may use these sands interchangeably as these are very similar. For the purpose of this case study, these sands represent batch-to-batch variability that occurs in working foundries. The resin coated sands, for all three sands, were prepared in the same coating facility using the same binder type and additives level to avoid any inconsistencies. Identity of 14 Figure 2. The Quality Control Framework sand origin and coating facility is kept anonymous as the intent of this paper to demonstrate the usefulness of the It must be noted that this framework only considers SPC method using PCA with TDT data in detecting shifts defects caused by improper or shifted sand-binder from on sand to another. In addition, this case-study systems and assumes that other factors, such as; improper illustrates the power of incorporating this SPC strategy pattern or tool design, improper gating design, improper with a casting trial to verify casting quality when a sand- melting and pouring practices, and improper metal binder system shift occurs. It is not the intent of the composition are do not result in defects. More authors to criticize any sand-binder systems. specifically, for this framework, these factors are considered constant and properly designed. As a result, SAND SYSTEMS this framework qualifies a chemically bonded sand Table 1 shows the percentage sand retained in GFN for system for a given alloy, metallostatic head pressure, the three sands. GFN results show that all three sands are pouring temperature, and minimum section thickness of similar in size which is AFS-GFN 60. In addition, as can the mold/core using disc-shaped specimen tests and relies be seen in Table 2, sand particle shapes are quite alike.

Page 3 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages)

Altogether, these results suggest that differences between Subsequently, disc-shaped specimens were made for these these sands are minimal. sand-binder systems using the same parameters 250°C and 40 seconds. To detect a process shift of going from Table 1. Percentage Sand Retained in GFN for Sand one sand-binder system to another, the SPC method using Systems PCA with TDT data was implemented. It should be re- System X Y Z emphasized that this SPC, in previous research, was Sieve shown to be capable of detecting small shifts in sand- 6 0 0 0 binder systems. Five different TDT test settings 12 0 0 0 (combinations of temperature and pressure), as shown in Table 4, were considered to determine if there was an 20 0 0 0 optimal setting for quickly detecting this specific shift. 30 0.1 0.1 0.1 40 2.5 2.7 6.0 Table 4. TDT Settings Test Setting Temperature Pressure (N) 50 22.5 23.1 17.6 (C/F) 70 36.2 34.0 29.9 1 750/1382 3.5 100 25.8 26.5 35.8 2 750/1382 5.5 140 11.3 11.7 9.3 3 850/1562 4.5 200 1.5 1.8 1.2 4 950/1742 3.5 270 0.1 0.1 0.1 PAN 0 0 0 5 950/1742 5.5

Table 2. Microscopic Image of Sand Particles In this study, sand X is considered as the in-control sand, X Y Z because the available quantity of this sand was greater than ands Y and Z. It should be noted that during TDT, specimens can often crack. Any specimen that cracked during TDT was disregarding in this study. In addition, because of the limited quantity of these sand systems; 15, 5, and 5 TDT observations of specimens (without cracking) were taken from sands X, Y, and Z; respectively for each test setting (Table 4).

These sands were coated with phenolic novolac resin at CASTING TRIAL 3% binder level. The resin coating was done at one of the The casting trial procedure consisted of the following sand-binder supplier’s laboratories (whose identity is kept steps: anonymous), where the sands were treated similarly to 1. Temper the green sand to the desired compactability ensure that no variation in the sand-binder systems’ were level in a muller or mixer. caused by the coating process. Disc-shaped specimens 2. Squeeze the sand within the flasks against the were made at WMU’s laboratory using the supplier’s matchplates to produce molds. recommended parameters of 250C (662F) and 40 seconds. 3. Pour molds. 4. Shake and inspect casting. DETECTING SAND SYSTEM CHANGES The hot tensile strength test, which is a commonly used Note: All molds were prepared and tested at WMU Metal AFS standard test use by most foundries in the United Casting Laboratory. Ambient conditions were temperature States to monitor sand-binder systems, was performed on controlled at 20 ± 1°C and relative humidity was these three systems. Historical data (supplier provided) controlled at 50 ± 2%. suggest that the hot tensile strength for X 3% RCS should be between 300-450 psi. Results of the hot tensile strength Preparation of Green Sand are shown in Table 3. It was important to keep the bonding formulation simple to reduce potential errors in preparing the green sand Table 3. Hot Tensile Strength Test Results batch and simplifying the analysis. Apart from sand, clay, 3% Hot tensile strength (psi) and water; no additives were introduced to the green sand RCS systems used in this study. Compactability was monitored X 335 322 425 390 continuously, while water additions were raised or Y 330 352 305 415 lowered as needed to produce the 35% target Z 410 320 430 330 compactability. The sand was not discharged until the compactability was on target. Thus, the green sand

Page 4 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages) systems used in this study were tempered to a desired compactability and tested for other green sand properties.

The silica base aggregate (lake sand, 62 GFN, 4 screen) used in the study came from Michigan. The green sand system used in the study was mulled (200 kg Simpson Sand Muller) at WMU Metal Casting Laboratory. The green sand system used a clay bond pre-blend made up of 20% Southern Bentonite / 80% Western Bentonite where total clay added was 8.0% BOS (methylene blue clay was 7.45%), and water added to produce the desired compactability.

Surface Defect Casting Trial Figure 3. Matchplate pattern for the surface defect One of the casting trials used at WMU investigates casting trial. surface anomalies and finish at a specimen/metal interface. This so-called surface defect casting trial was implemented in this case-study to compare casting surfaces for the three chemically bonded sand systems. Veining and penetration are among the most common metal casting surface defects. Therefore, the proposed casting trial aims to qualify against these two surface defects. Ramrattan et al. (2011) and Derrick et al. (2012) postulation that one mechanism for veining is stress cracking in the sand-binder system where metal can penetrate. The aim is to wet the surfaces of the disc- shaped specimens while being poured with aluminum or gray iron (in this case-study) to a 6 inch and 9 inch head height, respectively. The purpose of this work is to determine whether or not a shift, detected using TDT data, in a sand-binder system affects casting quality. In Figure 4. The disc-shaped specimens placed on the essence, common casting surface defects that will occur core prints in the drag half of the mold. in this casting trial, such as penetrations and veining, can be related to thermal-mechanical issues measured by Melting and Pouring TDT. During preparation for the casting trial, the chemically bonded disc-shaped specimens were placed randomly. Procedure The sand-to-metal weight ratio for all molds was 2:1. The Green sand cope and drag mold halves were fabricated molds were manually poured (temperature at pour ladle according to an experimental matchplate pattern, shown was 700C (1292F) for aluminum and 1427C (2600F) for in Fig. 3. The chemically bonded disc-shaped specimens gray iron) and molten metal was delivered through a set on core prints in the drag mold are pictured in Fig. 4. direct pouring sleeve fitted with a ceramic filter. The The gating was a central 6 inch and 9 inch sprue pouring metal chemistry is shown in Table 5 and Table 6 for gray sleeve fitted with an appropriate filter prior to the gate for iron and aluminum, respectively. wetting specimen surfaces.

Table 5. Gray Iron Chemistry C Mn P Si Cr S Al Pb Ti Mo Cu Ni Sn Sb Mg Fe

3.59 .65 .097 2.39 .103 .101 .001 <.001 .011 .02 .13 .06 .009 .002 .0025 92.94

Table 6. Aluminum Chemistry Al Si Fe Cu Mn Mg Cr Ni Zn Ti Ca Cd Na Pb Sr 92.2 7.07 0.16 0.05 0.02 0.33 0.007 0.006 0.02 0.10 0.0007 0.002 0.001 0.02 0.01

Page 5 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages)

RESULTS AND DISCUSSION 0.08 0.07 An analysis of variance (ANOVA) was performed on the 0.06 hot tensile strength test results, provided from the sand- 0.05 binder supplier, to determine whether or not their means 0.04 differ. Figure 5 shows the interval plot for hot tensile 0.03 0.02 strength for X, Y, and Z sand-systems. The plots show 0.01 Displacement (mm) that the 95% confidence interval of these three sands 0 overlap each other. Results also showed that the p-value -0.01 0 10 20 30 40 50 60 70 80 90 for the three pairwise comparisons were all greater than -0.02 0.05; therefore, there is not enough evidence to conclude Time (sec) X2 X3 X5 X8 X13 X15 that hot tensile strength for the three sand-binder systems are significantly different. Figure 7. Sand system X 3% RCS LD curves for 750C 5.5 N TDT setting.

0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 Displacement (mm) 0 -0.01 0 10 20 30 40 50 60 70 80 90 -0.02 Time (sec) X1 X2 X3 X6 X10 X15 Figure 5. Interval plot of hot tensile strength for sand systems x, y, and z. Figure 8. Sand system X 3% RCS LD curves for 850C 4.5 N TDT setting.

Results of this analysis show that the hot tensile strength test is not capable to discriminate between these three 0.08 sand systems, which further suggests that they are 0.07 0.06 equivalent. As a result, any foundry that relies solely upon 0.05 hot tensile strength to monitor their sand-binder systems 0.04 will not be able to detect a shift from one sand (used in 0.03 this case-study) to another. 0.02 0.01 Displacement (mm) 0 Figures 6 to 10 show representative range of longitudinal -0.01 0 10 20 30 40 50 60 70 80 90 distortion curves for X 3% RCS at each TDT settings. -0.02 Time (sec) These curves suggest that as test temperature and pressure increase, variability in LD curves decreases (spread X2 X3 X4 X5 X6 X9 between these curves reduces) for X sand systems. The Figure 9. Sand system X 3% RCS LD curves for 950C same is true for Y and Z sand systems as well. 3.5 N TDT setting.

0.08 0.08 0.07 0.07 0.06 0.06 0.05 0.05 0.04 0.04 0.03 0.03 0.02 0.02 0.01 Displacement (mm) 0.01

0 Displacement (mm) -0.01 0 10 20 30 40 50 60 70 80 90 0 0 10 20 30 40 50 60 70 80 90 -0.02 -0.01 Time (sec) -0.02 Time (sec) X1 X3 X4 X7 X8 X15 X3 X4 X7 X8 X9 X13

Figure 6. X 3% RCS LD curves for 750C 3.5 N TDT Figure 10. Sand System X 3% RCS LD Curves for 950C setting. 5.5 N TDT setting.

Page 6 of 11 2020 AFS Proceedings of the 124th Metalcasting Congress Paper 2020-074 (11 pages)

To detect a change, principal component (PC) model was control charts correctly identified out-of-control sand developed using samples sizes of one ( = 1) and systems. It is important to mention that 3% RCS is thirteen samples of X sand system were chosen as typically used in foundries to make medium sized iron historical data for developing control limits𝑛𝑛 ( = 13). PC castings. Usually, TDT is performed to mimic the actual model was applied to two remaining observations of X as casting process which would be at or above 1350C well as all the observations of Y and Z sand 𝑚𝑚systems to (2462F) and 4.5 N. The results of this study suggest that collect 4 PCs. Figures 11 to 15 show control charts for TDT can be performed at lower temperature than the the scenarios of shifting from X sand system2 to Y and Z actual casting process to detect a change in the considered sand systems at each TDT settings. 𝑇𝑇 sand system. However, to study the behavior of the sand system under actual casting parameters, it is In these control charts, as first thirteen samples were recommended to perform TDT that mimic the casting considered historical data, sample 14 and 15 are the process. remaining two samples for X sand system while samples 16 through 20 are for either Y or Z sand system as Results of this study show that SPC method using PCA depicted in control chart. can effectively detect a change between these considered sand systems at lower temperature than the actual casting Control charts suggest that as temperature and pressure process. It is important to reiterate that the difference increase in TDT, ability to detect a change in sand system between these sands is very subtle and it cannot be increases. For instance, at 950C (1742F) TDT settings, detected by conventional hot tensile strength test.

10000 10000

1000 1000 2 2

T 100

10 10