(Proceedings of the 1994 Steel Founders' Society of America Technical & Operating Conference, Chicago, IL, Nov. 9-12, 1994) DIMENSIONAL VARIABILITY OF PRODUCTION STEEL CASTINGS Frank E. Peters - Research Assistant John W. Ristey - Research Assistant Wayne G. Vaupel - Research Assistant Edward C. DeMeter - Assistant Professor Robert C. Voigt - Associate Professor Department of Industrial & Manufacturing Engineering The Pennsylvania State University University Park, Pennsylvania ABSTRACT Work is ongoing to characterize the dimensional variability of steel casting features. Data are being collected from castings produced at representative Steel Founders' Society of America foundries. Initial results based on more than 12,5 production casting feature measurements are presented for carbon and how alloy steel castings produced in green sand, no-bake, and shell molds. A comprehensive database of casting, pattern, and feature variables has been developed so that the influence of the variables on dimensional variability can be determined. Measurement system analysis is conducted to insure that large measurement error is not reported as dimensional variability. Results indicate that the dimensional variability of production castin% features is less than indicated in current US (SFSA) and international (ISO) standards. Feature length, casting weight, parting line and molding process all strongly influence dimensional variability. Corresponding pattern measurements indicate that the actual shrinkage amount for casting features varies considerably. This variation in shrinkage will strongly influence the ability of the foundry to satisfy customer dimensional requirements. DISTRIBUTION OF THIS DOCUMENT IS UMUMITEO
INTRODUCTION To meet ever tightening customer demands for closetolerance, near-net-shape steel castings, the foundry must exercise great control over pattern dimensi.ons,.. foundry process. variability, and measurement system variability. Ongoing work by ROSS',' and the authors3 has strongly suggested that past ~tudies~*~*~ have overestimated casting dimensional variability. Past results have been confounded by measurement errors. Widely used steel casting tolerance specifications based on these therefore do not accurately reflect the true dimensional - capabilities of steel foundries. This has limited the market for steel castings in precision applications. Less-than-adequate pattern shrinkage allowances, and traditional trial and error pattern correction methods also add unnecessary cost and lead time to the initial casting approval process. In this paper, first year results for an on-going three year study of dimensional control of steel castings will be summarized. The initial results presented here build on measurement system analysis studies by Ross and early dimensional studies by Peters and Voigt that have been reported previou~ly.'~~~~~~~'~ All of the results presented in this paper are based on dimensional surveys of production castings at member foundries. To date, comprehensive data from 6 molding processes from 4 foundries have been collected. This includes measurements of 498 different casting features from 57 different castings. The corresponding pattern feature measurements have also been 2
taken, when feasible. In most cases, features have been measured on at least 2, and usually 3, different castings to determine a specific feature's dimensional variability. This corresponds to more than 12,5 individual measurements in the database. Data from each casting and each feature have been compiled in a comprehensive descriptors database to assist in the analysis and interpretation of the data. Complete details of the database structure, measurement system analysis techniques used, and casting feature inspection procedures developed have been previously rep~rted.~ All dimensional data collected to date have been compared to existing casting dimensional tolerance specifications--the SFSA I1Tl1 grades and IS specification^.^-^ It is important to note that although these specifications predict similar tolerances for casting features, the fundamental nature of the SFSA and IS dimensional tolerance specifications are different. The SFSA IfTii grades predict feature dimensional variability based on feature length and casting weight. The IS specification uses only feature length to determine expected tolerance for a given molding process. Data analyses presented in this paper are preliminary. They will be modified as additional dimensional data are added to the database. The present database is expected to grow by a minimum factor of 3 or 4 before the end of this focused effort. As the size of the database increases, the statistical certainty with which conclusions can be made will also increase. However, it is 3
expected that many of the general trends in the data suggested from these first year data will be reinforced with future data. DESCRIPTION OF CURRENT DATABASE The range of feature lengths and casting weights for green sand, no-bake and shell casting dimensional features measured to date at participating foundries is shown below. MOLDING SMALLEST LARGEST PROCESS FEATURE LENGTH FEATURE LENGTH inches (mm) inches (mm) shell.25 ( 6.6) green sand.25 ( 6.5) no-bake.48 (12.3) 5.6 (141.6), 11.6 (295.2). 33.4 (848.4) MOLDING SMALLEST LARGEST PROCESS CASTING WEIGHT CASTING WEIGHT pounds (kg) pounds (kg) shell 5 (2) 16 ( 7) green sand 2 (1) 228 ( 13) no-bake 9 (4) 256 (1137) Figure 1 graphically shows the size and weight distribution of the castings measured for the various molding processes. These feature size and weight distributions are typical of the size and weight distributions of the industry. However, it is clear that additional data are needed for larger castings and for large features on castings of all weights. This data are needed to allow for reliable statistical analyses of the entire size range of castings produced in the industry. At the present time only carbon and low alloy steel producers have participated in the 4
16 T o a. a) 2t 1 1 1 2 3 4 5 6 FEATURE LENGTH (inches) 25 b) 2 4 6 8 FEATURE LENGTH (inches) 1 12 25 2 15 lo C) 5 Q O 3 8 OO h. 5 1 15 2 25 3 35 FEATURE LENGTH (inches) Figure 1: Feature length and casting weight distribution of casting features measured for various molding process a) shell mold castings b) green sand castings c) no-bake mold castings. 5 1
study. Another goal is to obtain a significant database for high alloy steel castings. CASTING DIMENSIONAL VARIABILITY RESULTS In general, many casting-specific and feature-specific factors influence the overall dimensional variability measured in the study. These factors include molding process used, feature length, casting weight, parting line and others. Before the combined influence of these many factors are described, the influence of some important individual factors on dimensional variability will be presented. Figure 2 illustrates the influence of feature length on dimensional variability for the various molding processes. In this and all subsequent figures, the dimensional variability is expressed in terms of the "half tolerance1i or three standard deviations (3a) about the mean. Each data point shown on this figure and subsequent figures is the calculated three standard deviation value determined from individual dimensional measurements of the same casting feature on 18 to 3 castings, The larger features, in general, exhibited more dimensional variability than smaller features. (It should be noted that the length and variability axes on each plot of Figures 2a-c are scaled differently.) Dimensions crossing the parting line show a slightly greater variability than those not crossing the parting line, for each of the molding processes. Figure 3 compiles all of the data from Figure 2 to illustrate the relationship between 6
r s- fl (D 3 STANDARD DEVIATIONS (inches) - K U N E 3 STANDARD DEVIATIONS (inches) PPPP PPPP ot3g8gp---- -lqbo.3 3 STANDARD DEVIATIONS (inches) <e* * 3 L O * a * * ** * *
:.o 3 STANDARD DEVIATIONS (inches) P P 1 4 P d 1 P N. '.8 O a n b o +-. +:.. + -$ ++ +3+ 4.,s *. *......?LO.... e. O *.. e &*o* 4...... 8
the green sand, no-bake and shell data. In general, somewhat less variability is observed for shell and no-bake casting features than for green sand features. The scatter in the data can be largely attributed to factors other than feature length that influence variability. Similar plots can be used to show the influence of casting weight on overall dimensional variability. Figure 4 indicates that casting feature variability, in general, increases as casting weight increases for each molding type. This suggests that the current IS specifications, which do not include the influence of weight, can significantly underestimate or - overestimate steel casting tolerance capabilities depending on the size of the casting. Figure 5 combines the data on the same axes to illustrate the general influence of casting weight on dimensional variability for all three molding processes. The previous figures also suggest a significant influence of parting line on dimensional variability, Figures 2 and 4. The effect of the parting line on the overall dimensional variability is summarized below. CONDITION AVERAGE VARIABILITY (3) NUMBER OF inches (mm) FEATURES crosses parting line.75 (1.9) does not cross the parting line.45 (1.15) 12 5 286 The difference of these two average values is an estimate of the influence of the parting line on dimensional variability. 9 T
2 Feature Does Not Cross Parting tine + Feature Crosses Parting tine.6.5 4 2; 3.4 n o 5 + 8.3 2 Q.2 n z i Q.1 d a, 2 4 6 8 1 CASTING WEIGHT (pounds) 12 14 16 2.18.16.14 2; 3.12 2.1.8 n.6 2.4 Q 2 2.2 r3 b) 2 5 C) f O @ +.5 r3 + + 4 Oel v) + 4 $.15 3 * 25 + 4.2 n o 2 15 CASTING WEIGHT (pounds).25 Q2 >- 1 5 5 8 -I loo 15 2OOo 25 3ooo CASING WIGHT (pounds) Figure 4: Influence of casting weight on overall dimensional variability a) shell mold castings b) green sand castings c) no-bake mold castings, 1
.. 8 In *+ e** *no. ***. an*.. ***.e.* e.'e.. o * + I 8 - - c 11
The data show that the total dimensional variability (6a), averages.6 inches (1.5 mm) for features that crosses the parting line. However, other factors such as feature length and casting weight appear to have a significant influence on the dimensional variability of features that cross the parting line. This is perhaps to be expected; large molds made with no-bake processes could be expected to have more parting line variability than small green sand or shell molds. All castings measured as part of this study were inspected after final heat treatment and shot blasting. The influence of type of heat treatment on overall dimensional variability is shown below. AVERAGE HEAT TREATMENT VARIABILITY (3a) NUMBER OF CONDITION inches (mm) FEATURES quenched & tempered.36 (.92) 79 normalized & tempered.49 (1.24) 57 normalized.57 (1.44) 21 homogenized, quench & temper.73 (1.84) 65 This data must be interpreted with great care because the average weight and feature length for each heat treatment condition was not the same. Future analysis of the data will be performed to determine the influence of heat treatment on casting variability independent of the influence of other significant variables. The effect of core binder type on dimensional variability for the 1 features measured across cores is shown below. 12
CORE TYPE oil sand shell no-bake AVERAGE VARIABILITY (3U) NUMBER OF inches (mu) FEATURE3.41 (1.4) 5.42 (1.8) 7.54 (1.37) aa These preliminary observations must also be interpreted with great care because of the low number of observations for shell and oil sand cores, and because of the confounding influence of core size on variability. In general, the average size of the no-bake cores was signifi.cantly greater than the size of the oil sand and shell cores included in the analysis. The difference in core size is a likely explanation for the increased dimensional variability observed for no-bake cores. The influence of mold wash on dimensional variability has also been evaluated. The data below are for features made from no-bake molds, and are not across a core. WASH CONDITION no mold wash mold wash AVERAGE VARIABILITY (3C) NUMBER OF inches (nun) FEATURES.57 (1.44) 72.69 (1.75) 54 This difference is not expected to be statistically significant. The data are presented primarily to illustrate the ability of the database developed in this study to evaluate the influence of individual variables or combinations of variables on dimensional variability. 13
The dimensional variability obsenred for casting features also depends on the mold and core relationships that affect dimensions. For example, the diameter of a cored hole is controlled by the reproducibility of the core used to make the hole. However, the wall thickness for that same cored feature is also influenced by variability in core placement. The variability in wall thickness can be expected to be greater than for the hole diameter itself. However, the mold to core dimensions are typically smaller, and therefore exhibit less variability. A summary of the overall influence of the dimensional variability observed for some mold and core relationships to features is shown below. Examples on these various types of dimensions are shown schematically in Figure 6. DIMENSION TYPE AVERAGE VARIABILITY (3a) NUMBER OF inches (mm) FEATURES mold to core across casting.41 (1.5) 55 mold to mold across casting.5 (1.28) 126 core to core across core.52 (1.33) 1 mold to mold across casting/core/casting.78 (1.97) 77 Figure 6: Schematic diagrams illustrating dimension types. II, 2 k 4 - l 14-1 1 moldt o core across castlng 2 mold to mold across castlng 3 core to core across core 4 mold to mold across - casting / cc~e / casting
This clearly shows that Itall dimensions are not created equal". The data must again be interpreted with great care, since other variables such as parting lines, influence only some of the dimension type'categories. ASSESSMENT OF FOUNDRY DIMENSIONAL CAPABILITIES To date, extensive data sets have been collected for green sand molding lines at two foundries, for no-bake molding lines at two foundries, and for shell molding lines at two foundries. (Work at other foundries is in progress.) Participating foundries will receive a comprehensive proprietary report detailing their specific dimensional capabilities and an assessment of their general dimensional control capabilities compared to the rest of the industry. A general comparison of the overall dimensional capabilities of all of the first year participating foundries compared to the existing SFSA tolerance standards is shown here. PERCENTAGE OF FEATURES WITH DIMENSIONAL VARIABILITY LESS THAN THE EXISTING SFSA "T" TOLERANCE GRADES SHELL ----- T3 T4 T5 T6 T7 89% 1% 1% 1% 1%. 86% 1% 1% 1% 1% GREEN SAND NO-BAKE 42% 75% 88% 94% 1% 15% 62% 1% 1% 1% 64% 9% 98% 1% 1% 68% 81% 96% 1% 1% This table summarizes the percentage of features for each foundry molding process that exhibited dimensional variability that was 15
less than the existing "T grade" tolerance. The current ot311 grade is considered appropriate for castings produced from special molding processes, the current trt511 grade is considered appropriate for typical (green sand) processes, and the current 11T711 grade is considered appropriate for complex casting configurations. When these llt1l grades were initially established they were based loosely on the average dimensional capability (5% conformance). Clearly, the foundries' ability to control dimensions is considerably better than indicated by the existing SFSA standard for all molding processes. Linear regression analysis techniques can be used to develop improved models for predicting the dimensional variability of each individual molding process for each foundry. The following are multiple linear regression models for each of the six molding processes. Feature length, casting weight, and the presence or absence of a parting line are the independent variables used in this simple analysis. MOLDING PROCESS REGRESSION MODELS TO PREDICT DIMENSIONAL VARIABILITY SHELL : 3 StDev =.3 +.1 * W +.3 * L +.8 PL 3 StDev = -.13 +.3 * W +.1 * L -.7 PL GREEN SAND: 3 StDev =,48 +.2 * W +.5 * L +.2 PL 3 StDev =.27 +.1 * W +.5 * L +.15 PL NO-BAKE: 3 StDev =.34 +.2 * W +.3 * L +.15 PL 3 StDev =.24 +.5 * W +.2 * L +.24 PL 16
Where 3 StDev = 3 standard deviations (inches) W = casting weight (pounds) L = feature length (inches) PL = 1 if feature crosses parting line = if feature does not cross parting line There are some important general conclusions which can be made by comparing these equations: - In general, foundry-to-foundry dimensional capability variations for a given molding method are less than the variations between molding processes. - Coefficients on the weight terms in the regression equations differ significantly for the different molding processes. Casting weight is much more of a dimensional variability factor for green sand than for no-bake molds This is no doubt due to the rigidity of no-bake molds compared to green sand molds. This also suggests that control of green sand molding processes are key to insuring casting dimensional control. An example of a dimensional capability table generated for a particular foundry based on its linear regression table is shown in Table 1. This type of information can be effectively used by participating foundries to better predict their dimensional capabilities. In order to put the dimensional variability measured to date in this study in perspective, a comparison was made between the dimensional variability predicted using the new equations and the current SFSA "T" equations. Comparisons for some representative casting features are presented in Table 2. This shows some differences in the ability of foundries to control dimensional variability. In most cases, foundry dimensional control is significantly better than indicated by current tolerance guidelines. 17,;, - -
Table 1: Sample tolerance table for a particular foundry molding process CASTlNG WEIGHT (pounds) 2 4 6 8 1 15 2 3 1 2 5 75 1 2 25 5.4.4.4.41.41.43.43.47.46.46.47.47.47.49.5.54.52.53.53.53.54.55.56.6.59.59.59.6.6.62.63.67.65.65.66.66.67.68.69.73.81.8 1.82.82.83.84.85.89.97.97.98.98.99.1.11.15.129.129.13.13.131.132.133.137 Table 2: Comparison between predicted feature variability using the new equations and the current SFSA tolerance guidelines for selected casting features. Note values in bold are the result of extrapolation. LENGTH (inches) WEIGHT (pounds) 18
PATTERN SHRINKAGE ALLOWANCES Corresponding pattern dimensions were also measured for many of the casting features in this study. By comparing pattern dimensions to average casting feature dimensions, an accurate estimate of the casting shrinkage amount for each feature can be obtained. Because close attention has been paid to controlling measurement errors, these data are perhaps the first reliable set of shrinkage allowance information developed for the industry. Figure 7 shows the distribution of actual shrinkage values for the features measured in this study. The average shrinkage factor measured (2.2%) is virtually identical to the llstandardll pattern shrinkage allowance used throughout the industry (2.8% or 1/4 in. per ft). However, of great concern is the wide range of the pattern shrinkage values obtained. Clearly, these wide variations in pattern shrinkage amounts have a substantial effect on the foundries' ability to produce castings that meet customer dimensional requirements, and must be carefully studied. Many features had shrinkage allowances considerably greater than and considerably less than 2.8%. The average shrinkage amount for each of the molding processes is as follows. Analysis of the data is continuing.. MOLDING PROCESS shell green sand no-bake AVERAGE SHRINKAGE PERCENTAGE 2.4 1.7 2. NUMBER OF FEATCJRES 26 41 83 19. --
INCHES / FOOT 45 4 35 15 1 5 o m 9 c u 9 P 3 9 m F c I o l n 9 c u s 7 4 o -cum 2 o 2m : o 2 SHRINKAGE PERCENTAGE Figure 7: Distribution of observed shrinkage amounts. ANALYSIS OF FOUNDRY MEASUREPENT SYSTEMS Throughout this study, the repeatability of measurement systems used to measure casting and pattern dimensions at the various foundries has been characterized. Table 3 summarizes the repeatability obtained from a wide variety of foundry inspection department measurement systems. This table does not include reproducibility errors that may be introduced from operator variability. Rather, it simply indicates the equipment's baseline repeatability when used by a trained operator. The measurement error (gage R & R) must be less than 3% of the feature tolerance for the measurement system to be adequate.11 Portage layout machines have considerably more repeatability 2
Table 3: Repeatability errors for some commonly used measurement instruments MEASUREMENT INSTRUMENT micrometer CMM Digital Caliper - not a diameter Digital Caliper - diameter Portage Machine Scale / Ruler REPEATABILITY (inches) SAMPLE AVERAGE MINIMUM MAXIMUM SIZE.4 1.8.4.1 1 2.12.6.29 12.23.19.26 4.28.12.41 4.56.55.57 2 Tolerances (?inches) for Tolerance Grade T3 (S) Casting Weight. Ib L* 2 5 1 2 5 75 1 15 2 25 5 75 1 I25 I5 2 3 4OOO 5.47-49.OS7-63 -68-71.75.8 I.9.97.I3-51.OS4.62.68.72-76.79.OS5.94.IO1.I8.OS7.OS9.68.73.78-82 -85.9 I.I.I7.I 13.64.67.75.81-85.89.92.98.I7. I 14.I21 6..69.72.8-86 -9.94-97. I3-1 I2.I2-126 8..OS.OS3.OS5.OS8.63.65.67-71.73.76-84.9-94 -98.lo1,17.116-124.I3 1..OS4.OS6..OS8.61.66.69.71-74.On.79.87-93 -98-11.IO5.I11.I2.I27.I33 15..61.63.65-68.73.76.78.81.84-86.94-1.I@.I8.I I2.I17-127.134-14 2. -66.69-71.74.78.81.83.87-89.91.I.IO5.I 1.I14.I 17.I23.I32.I39.I45 3..75.77.79.82-87.9-92 -95.98.I.I8.I14.I19-123.I26.I32-141 -148.154 4..82.84,86.89-94 -97.99.I2.IO5.I7-115 -121-126.I29.I33.I39.I48.I55.I61 5..88.9.92.95-1.I3-15.I8.111.I13.I21.I27-131 -135.I39-144 -154-161.I67 6..93.95.97-1.IO5.I8.I1-113 -116-118.126.I32-137.I4-144 -15.I59-166.I72 'Dimension leqth in inches Figure 8: SFSA 1tT3tt tolerance table with the tolerances that cannot be measured with a measurement instrument with measurement error of.28 inches (.71 mm) indicated 21
error than some other types of instruments and must be used with great caution. Based on the average repeatability of.28 inches (.71 mm), a measurement error requirement of less than 3%, and assuming negligible reproducibility error, the smallest total feature tolerance that the Portage can be used for is.93 (2.37 mm). Figure 8 is a copy of the SFSA T3 table with tolerances that cannot be measured with the Portage machine indicated. Work is continuing to characterize the gage R & R of this and other commonly used foundry measurement analysis instruments and to determine the influence of surface characteristics on gage R & R. SUMMARY Work with SFSA member foundries is continuing so that improved casting tolerance guidelines can be developed, improved pattern shrinkage allowances can be obtained, and dimensional control strategies can be developed for the membership. The data presented in this paper will continue to be analyzed in order to fully understand the influence of the many interacting variables on dimensional variability. Additional data need to be collected during the second and third years of the project so that a comprehensive database can be developed for the industry. This database will provide the dimensional information and the dimensianal CQILLZQL strategies for satisfying the dimensional needs of the steel casting customer. 22
As a result of the first year of study, the following preliminary conclusions can be made: - Adequate measurement system gage R & R is attainable but requires vigilance. - Initial tests on Portage machines indicate relatively poor repeatability. More information is being collected. - Observed casting feature variability is considerably less than indicated by current SFSA standards. - Many processing and geometry variables have measurable effects on dimensional variability. - More data and analysis is needed to determine if there are significant foundry-to-foundry dimensional variability differences and to identify individual sources of variability. - Current shrinkage allowance factors are inadequate. Work is ongoing to develop improved allowances. REFERENCES 1 ROSS, P. J. ; "Measurement System Capability Project, It Steel Founders' Society of America T & Conference, Chicago, IL (November 199). 2 3 4 5 6 Ross, P.J.; ##Measurement System Capability Project - 1992 Update, Steel Founders' Society of America T & Conference, Chicago, IL (November 1992). Peters, F.E. and R.C. Voigt; "Casting Inspection Strategies for Determining Dimensional Variability," Steel Founders' Society of America T & Conference, Chicago, IL (November 1993). Law, T.D.; "Dimensional Tolerances in Steel CastingsIii The British Foundryman, vol 71, part 9, p 223 (1978). Aubrey, L. S. et a1. ; Dimensional Tolerances, Research Report No. 84, Steel Founders' Society of America, Des Plaines, IL (1977). IBF Technical Subcommittee TS71; "Second Report of Technical Sub-Committee TS71 - Dimensional Tolerances in Castings," The British Foundryman, vol 64, part 1, p 364 (1971). 23 T - -- - - -
7 Wieser, P.F., ed.; Steel Castings Handbook, 5th ed, Steel Founders' Society of America, Rocky River, OH (198). 8 ISO; Castings - System of Dimensional Tolerances, IS 8 6 2 9 Voigt, R.C. and F.E. Peters; lfdimensionaltolerances and Shrinkage Allowances for Steel Castings,iiSteel Founders' Society o f America T & Conference, Chicago, IL (November 1992. 1 Peters, F.E. and R.C. Voigt; lldimensionalcapabilities of Steel CastingsIii Proceedings o f the Near-Net-Shape Manufacturing: Examining Competitive Processes Conference, Pittsburgh, PA (September 1993). 11 Measurement Systems Analysis - Reference Manual, Automotive Industry Action Group (199). (1984). The authors wish to thank the U.S. Department of Energy, the Assistant Secretary for Energy Efficiency and Renewable Energy, under DOE Idaho Operations Office, Contract DE-FC793ID13235, for providing funds to support this research. The authors would also like to thank the SFSA Carbon and Low Alloy Research Committee, and the participating foundries, for their support and guidance for this project. Finally, technical support provided by Chad Hafer is greatly appreciated. DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thcreof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. 24
Prepared for the U.S. Department of Energy Assistant Secretary for Energy Efficiency and Renewable Energy Under DOE Idaho Operations Office Contract DE-FC7-93ID13235