Frank E. Peters Robert C. Voigt. Industrial and Manufacturing Engineering Department The Pennsylvania State University University Park, PA 16802

Transcription

1 Assessing the Capabilities of Patternshop Measurement Systems Frank E. Peters Robert C. Voigt Industrial and Manufacturing Engineering Department The Pennsylvania State University University Park, PA ABSTMCT Casting customers continue to demand tighter dimensional tolerances for casting features. The foundry then places demands on the patternshop to produce more accurate patterns. Control of all sources of dimensional variability, including measurement system variability in the foundry and patternshop, is important to insure casting accuracy. Sources of dimensional casting errors will be reviewed, focusing on the importance of accurate patterns. The foundry and patternshop together must work within the tolerance limits established by the customer. In light of contemporary pattern tolerances, the patternshop must review its current measurement methods. The measurement instrument must have sufficient resolution to detect part variability. In addition, the measurement equipment must be used consistently by all patternmakers to insure adequacy of the measurement system. Without these precautions, measurement error can significantly contribute to overall pattern variability. Simple robust methods to check the adeq uacy of pattern measurement systems are presented. These tests will determine the variability that is contributed by the measurement equipment and by the operators. Steps to control measurement variability once it has been identified are also provided. Measurement system errors for various types of measurement equipment are compared to the allowable pattern tolerances, that are established together by the foundry and patternshop.

2 INTRODUCTION Today's casting customers are demanding metal castings with tighter dimensional tolerances. This is in part due to the increased use of CNC machines. The customer desires a casting that can be quickly and accurately fixtured into a machine tool without the need for pre-machining. To meet the customers' needs, the foundry must use accurate patterns, and control the production variables that affect the dimensional variability of a casting. The pattern must be accurately designed and produced so that when the casting shrinks upon cooling it will be the specified size. Patternmakers can no longer just pay attention to the nominal dimensions on finished part, casting, or pattern drawings, Contemporary dimensional demands also require patternshops to re-evaluate their pattern measuring equipment with respect to the dimensional tolerance requirements of the casting. As casting tolerances become tighter, patterns themselves must be more accurate, and the measurement equipment used to measure the patterns (and final castings) must also be more accurate. For example, if a part or casting drawing has relatively large dimensional tolerances, the traditional patternmakers' shrink rule may be suitable for measuring and producing the respective pattern. This traditional measurement method is not appropriate, however, when narrow casting tolerances are required. Correct pattern shrinkage allowances

3 and accurate pattern measurement systems are both needed to assure the production of accurate patterns. This paper will describe methods to evaluate the appropriateness and accuracy of pattern measurement systems based on final casting dimensional tolerance requirements. These methods are based on fundamental gage repeatability and reproducibility (gage R & R) tests that can be used directly in the patternshop. Before these measurement system analysis procedures are presented, the different types of casting dimensional errors will be briefly discussed. A review of literature pertaining to pattern and casting inspection will also be presented. BACKGROUND The authors are involved in ongoing research to assess the dimensional variability of production steel castings across North America. Data are being collected from production castings and their respective patterns. This data will be used to determine the dimensional capabilities of the steel casting industry, Dimensional data from castings and patterns will also be used to develop improved pattern shrink allowances based on casting geomztry and foundry practice. All measurement systems used to measure pattern and casting dimensions collected as part of this comprehensive study have been assessed to insure that measurement system errors are acceptably small. These precautions insure that the dimensional variability measured in this study is due to casting process variability rather than measurement error. The

4 4 measurement system analysis methods used throughout this effort can and should be used by patternshops to quantify measurement system errors. This is critically important f o r patterns meant to produce close tolerance castings. CASTING DIMENSIONAL VARIABILITY There are many individual processing steps needed to produce a metal casting in a sand mold. Such steps include sand preparation, core making, mold making, mold and core assembly, and finishing. Each of these steps, as well as other factors such as pattern wear, will contribute to the overall dimensional variability of casting features. Figure 1 displays the major categories of casting production variables influencing the dimensional integrity of casting features. The resultant size of every casting feature will not be the same for each casting produced, but will be distributed over a range of values. This variation of feature dimensions from casting to casting is commonly expressed in terms of standard deviations about the mean. A total of six standard deviations ( 6 0 ) centered on the mean accounts for % of all observations made, for normally distributed data. This means that for any particular feature, the feature dimension will be expected to be within the 60 range for 997 out of 1000 castings. As the process variability decreases, the probability of a dimensional casting error also decreases. An example of dimensional variability and its relationship to the customer specified feature tolerance is shown in Figure 2.

5 5 In addition to the casting dimensional errors caused by process variability, there are also errors associated with the design and construction of the pattern. Metal castings contract as they cool from the solidification temperature to the room temperature. This contraction can be predicted and is compensated for by making the pattern oversized -- the pattern shrinkage allowance. If the actual shrinkage of a casting feature is different than the shrinkage allowance used in the pattern design, or if the pattern feature was made incorrectly, then the resultant casting feature will not be of the proper size. Either error in the pattern will cause the average value of the casting feature to be shifted from the specified nominal dimension, as shown in Figure 3. The total dimensional error seen by the casting customer is a combination of the spread of dimensions and the offset of the mean from the nominal dimension. These error types can be classically referred to as random and systematic errors, respectively. Figure 4 schematically illustrates the dimensional error that results from a combination of both error types. In Figure 4, the area under the distribution curve and less than the lower tolerance limit represents the features which are o u t of tolerance for this example. For a casting production run to be within dimensional tolerance, three standard deviations (random error caused by the foundry process variability) plus the systematic error amount (caused by shrinkage allowance and

6 6 pattern inaccuracies) must be less than half of the specified tolerance. If the random dimensional error due to foundry processing variables is equal to the specific tolerance, then the shrinkage allowance used and pattern construction and measurement must be perfect to produce in tolerance parts. Clearly, the foundry and the patternshop must 'share' the total tolerance specified by the customer for the casting feature in question. As the total tolerance specified by the customer decreases, it becomes important to control all sources of dimensional variability both in the foundry and in the patternshop. One important source of error that should be regularly monitored and controlled by the patternshop (and the foundry) is measurement error. This is not simply a matter of insuring that measurement instruments are correctly calibrated. Measurement error also incorporates the accuracy limitations of the measuring instruments themselves and their usage. In many cases, a feature is measured only once and the measurement is assumed to be perfect. However, a range of dimensional values is likely if the same measurement is repeated by a single inspector, or by different inspectors. These measurement errors can significantly add to the dimensional variability during patternmaking. These typically small errors may not be significant for castings produced to wide customer dimensional tolerances. But, as will be shown, these same small measurement errors are unacceptable in patterns for close tolerance castings.

7 Measurement system analysis (commonly called gage repeatability and reproducibility or gage R & R tests) should be conducted to check the acceptability of all pattern shop inspection processes. The importance of these tests, and specific test procedures for use in the patternshop, will be described. Measurement system analysis was first introduced by Grubbs in 1948, but was not widely used until recently.2 Recent foundry industry authors have demonstrated the effectiveness of gage R & R tests for dimensional and other types of measurements. Studies by Ross have indicated that the measurement capability of 3-8 patternshops is often inadequate. However, American Foundrymen's Society references on patternmaking do not include any discussion of pattern inspection or measurement error. MEASUREMENT ERROR One source of measurement error is gage accuracy. To eliminate this source, all measurement equipment should be regularly calibrated to assure that it measures accurately. Calibration should be completed at least annually, but the frequency of calibration is dependent upon the amount of use. However, insuring the accuracy of measurement equipment by frequent calibration is not sufficient however. Of concern is not just the dimensional integrity of the measurement instrument, but the dimensional integrity of the entire 'measurement system'. A measurement system may be as simple as a single patternmaker using a tape measure, or as complex as a CMM with

8 8 its fixtures, software, and associated personnel. In any case, the requirement of the measurement system is that it is can obtsin measurements with minimal added variability. Measurement error in the patternshops must be compared with either the tolerance allowed for the pattern or the actual variability of the features aeasured. A feature with a small dimensional tolerance cannot afford to give as much variability to the measurement system as a feature with a larger dimensional tolerance. There are two common definitions of measurement error: Measurement Error ( % I = Measurement System Variability Variability of Feature Measured x 100 = Measurement System Variability Tolerance of Feature Measured x 100 and Measurement Error ( % ) The ability of a measurement system to produce acceptable measurements is evaluated using the following criteria:l2 Under 10% measurement system error % measurement system error -Over 3 0 % measurement system error -- Acceptable Marginally acceptable Measurement system needs improvement. If measurement system errors are greater than 3 0 %, actual dimensional variability (real pattern variability) cannot be adequately distinguished because of measurement system variability. If two inspectors accurately measure the same part several times using the same instrument, it is very unlikely that they will always obtain the same measurement. This is due to

9 9 variability in the measurement system. Measurement system errors can be further divided into separate repeatability (equipment) and reproducibility (operator) errors. Note that repeated identical measurements of a feature, rather than indicating adequate dimensional control, may be an indication that the measurement instrument used is not appropriate for the inspection task. As a general rule, there must be ten, and preferably fifty, measurement increments between the tolerance allowed for the pattern. Table 1 provides examples of the required and preferred measurement increments for some sample pattern tolerances. It is important to note that a measurement instrument with acceptable resolution is a requirement to have acceptable repeatability, but not a guarantee. Gage repeatability is the variation of repeated measurements taken by one person using a single acceptable gage. Repeatability error will be large if the measurement increments are large compared to the part tolerance or part variability. Therefore, fifty measurement increments between the part tolerance are preferred- There are other causes of poor repeatability including the inconsistent use of a measurement instrument by a single inspector. The variation caused by different inspectors, measuring the same parts, using the same gage, is defined as gage reproducibility. This is primarily caused by the different inspectors using the gage inconsistently.

10 10 Both repeatability and reproducibility errors contribute to pattern errors and must be characterized and controlled by the patxernshop. Simple robust procedures have been developed to measure the amount of repeatability and reproducibility errors. Gage repeatability and reproducibility (gage R & R) tests measure the variability contributed by these two error types. The simplest method of determining the measurement system variability would be to have two or more patternmakers measure the same pattern dimension several times. The variability of the measurements could be calculated directly. However, the chance of testing bias is great, because inspectors remember previous results. Therefore, the preferred gage R & R test involves at least two persons taking at least two measurements on several parts or patterns. In order to separate actual part variability from the measurement variability, the exact measurement location must be identified so that each subsequent measurement is taken at the same location. To minimize the chance of a biased test, the order that the parts are measured must be randomized and an extra person is required to record the measurements. An example of a completed gage R & R test is displayed in Figure 5. A brief review of gage R & R procedures follow. The reader is referred to a more comprehensive source on measurement system analysis for all of the pertinent details.12 In this example, two inspectors each measured ten different part dimensions, two times. The difference in each pair of measurements taken for a single feature by an inspector is used

11 11 to calculate the repeatability portion of the overall measurement variability. The difference between the average value of all measurements taken by each inspector is used t o calculate the reproducibility variation. The values calculated in the worksheet for repeatability, reproducibility, and total gage R & R are within a 5.15 standard deviation spread. Statistically, these values will account for 99% of the measurement variability. A s shown in the worksheet of Figure 5, an estimate of the process variability can also be calculated. For each feature, the average feature size is determined from all of the measurements for that feature ( 4 in the example). The process variability is calculated from the difference in the maximum and minimum average value. The gage R & R value can be compared to the process variability or the specified tolerance on the part drawing. the gage R & If R is less than 30% of the process variability and of the feature tolerance, than the measurement system is adequate to measure this particular pattern feature. Because ten identical patterns are typically not available, the standard gage R & R procedures just described must be modified for use in the patternshop. Table 2 is a guide to select the appropriate modified gage R & R test for these situations. Measurement option 1 from Table 2 allows for gage R & R to be calculated based on measurements across the same feature of a pattern. For instance, if only t w o identical patterns were made, five measurements at different locations of a flange thickness could be taken from each pattern. Again, the

12 12 measurement locations should be marked and the order of the measurements should be randomized. but adequate way to measure gage R places to take measurements. Option 3 is a less desirable & R when there are limited When using Option 3, care must be taken to insure that the inspectors do not 'remember' previous measurements and bias the test. Measurement system analysis needs to be done f o r each operator and piece of measurement equipment. Individual measurement techniques also need to be qualified. For example, there is a greater chance of measurement variability using a caliper for an inside diameter than an outside measurement of a square feature. The gage R & R tests should be repeated at least annually to insure that no changes have occurred in the measurement system. ACCEPTABLE PATTERNSHOP MEASDREMENT ERRORS An important consideration that remains is that the foundry and patternshop must share the total tolerance called o u t on the casting drawing. When measuring castings, the total measurement variability is compared to the tolerance specified on the print by the customer. The tolerance on the pattern features, however, can only be a fraction of the tolerance specified for the casting features. The amount of variability within which the foundry expects the patternmaker to work should be agreed upon by both parties at the time of the pattern order. A percentage of the casting tolerance may be an appropriate method for establishing the pattern tolerance. This smaller pattern tolerance value is

13 13 the baseline from which pattern measurement variability must be compared, not the total casting tolerance range. The following is an example of a measurement system analysis application. The Steel Founders' Society of America tolerance guidelines will be used to select an example steel casting dimensional tolerance.l3 The total tolerance f o r an 1 inch (25.40 mm) dimension on a 10 pound (4.5 kg) casting under normal production conditions is inches ( m ). (This value and all other casting tolerance values used this example are the total tolerance, 60). I f the pattern is to be made with a dimensional tolerance equal to 20% of the casting tolerance, the total allowable pattern tolerance is inches ( m m ). An acceptable method of measuring this pattern can consume only 30% of the pattern tolerance, or inches (0.175 m m ). (The upper limit of the measurement system acceptability criteria, 30% of the tolerance, is used in this example.) In this example, the allowable measurement variability for the patternshop inspection is only 6% of the overall feature tolerance. This is typical since, in general, the acceptable measurement variability available to the patternshop is only a fraction of the foundry's casting feature tolerance. Table 3 lists similar pattern tolerance calculations for other casting feature tolerances. These measurement system requirements can be compared to data on the repeatability portion of measurement variability f o r many commonly used patternshop measurement instruments, Table 4. This measurement repeatability data were collected as part of an

14 14 ongoing casting dimensional control study by the authors. Note that Table 4 only includes repeatability (equipment variability), but not reproducibility (operator variability). This was done because repeatability is the limit of performance for a particular measurement instrument. Reproducibility errors, although important, can usually be minimized by proper training of all inspectors. I n the previous example the measurement error was limited to inches (0.175 m m ). Therefore, from Table 4 the only acceptable measurement instrument is a micrometer, or possibly a CMM. Table 4 is provided only as an example, and not meant to take the place of a patternshop conducting gage R & R tests on the equipment it uses. Careful comparison of the data contained in Tables 3 and 4 clearly shows that current casting tolerance requirements place exceptional demands on the patternshop to control measurement errors. This is accomplished by using measurement systems with adequate repeatability and reproducibility. The repeatability listed in Table 4 for a scale is inches (1.40 mm). Calculations similar to above indicate that this measurement instrument (or a shrink rule) can only be used to measure patterns in which the overall casting tolerance is at least inches ( mm). The choice of equipment for measuring pattern dimensions should be decided upon when planning the production of a new pattern. To determine the acceptability of a given pattern shop measurement method, the measurement system analysis calculations

15 15 can be done by measuring similar features on an existing pattern. The assumption being made is that the measurement variability will be the same when measuring similar features. SUMMARY Foundries themselves have an important responsibility in the overall cssting dimensional control scheme. Foundries must insure that their processes are in control and work with the patternshop on using proper shrinkage allowances. Ongoing research has shown that further work has to be done in the application of the proper shrinkage allowance. It must be emphasized that only by using measurement equipment with small variability, can accurate shrinkage information be obtained from measuring patterns and resultant castings. Accurate historical data on casting shrinkage can then be used to predict shrinkage for future pattern designs. Measurement errors can contribute to the overall pattern variability. Gage R & R tests need to be performed on all of the different patternshop measurement equipment and inspection personnel on a regular basis. Figure 6 summarizes the steps for selecting and qualifying a measurement system. These tests are necessary to insure that the variability measured is due to actual part variability and not measurement system variability. Since measurement variability is compared to the allowed tolerance, clearly the measurement system requirements are more severe when producing pattern equipment for castings with tight tolerances. If measurement error is unacceptable for any

16 16 measurement system used in the patternshop (including shrink r u l e s ), steps need to be taken to correct it. Reproducibility errors can likely be corrected through training of all personnel in the proper and consistent use of the measurement equipment. Repeatability errors will probably have to be solved by using measurement equipment with better resolution, or by fixturing the pattern during inspection. ACKNOWLEDGEMENTS 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-FC07-93ID13235, for providing funds to support this research. 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 thereof, 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.

17 17 REFERENCES 1. Peters, F.E. and R. C. Voigt; "Casting Inspection Strategies for Determining Dimensional Variability," Steel Founders' S o c i e t y of America T & 0 C o n f e r e n c e, Chicago, IL (November 1993). 2 Grubbs, F. E. ; "On Estimating Precision of Measuring Instruments and Product Variability," Journal of the American S t a t i s t i c a l A s s o c i a t i o n, vol 43, p 243 (1948). 3 Hubbard, A. R., D.R. Moneymaker, and J. V. Shah; "Gage R &. R for Brinell Hardness Testing," S t e e l Founders' S o c i e t y of America T & 0 C o n f e r e n c e, Chicago, IL (November 1992). 4 Lively, D. M. ; "Measurement System Analysis and Control," Modern C a s t i n g, vol 82, no 5, p 30 (1992). 5 Marshall, D. and J. McGhee; "Gage Study to Determine Process Capability of Drilling and Boring," Steel Founders' S o c i e t y of America T & 0 C o n f e r e n c e, Chicago, IL (November 1993). 6 Ross, P.J.; "Measurement System Capability Project," Steel Founders' S o c i e t y of America T & 0 C o n f e r e n c e, Chicago, IL (November 1990). 7 Volkmar, A.P.; "Compactibility and Measurement Systems Variability," A F S T r a n s a c t i o n s, vol 101, p (1993). 8 Volkmar, A.P.; "Effectively Using Gage R & R and Measurement Systems Variability," Modern C a s t i n g, vol 8 3, no 11, p 30 (1993) * 9 Hamilton, E. ; Patternmaking G u i d e, 2nd ed, American Foundrymen's Society, Des Plaines, IL (1990). 10 P a t t e r n m a k e r ' s Manual, American Foundrymen's Society, Des Plaines, IL (1986). 11 Ross, P.J.; "Measurement System Capability Project Update," Steel Founders' S o c i e t y of America T & 0 C o n f e r e n c e, Chicago, IL (November 1992). 12 Measmrement Systems A n a l y s i s - Reference Manual, Automotive Industry Action Group (1990). 13 Wieser, P.F., ed.; Steel C a s t i n g s Handbook, 5th ed, Steel Founders' Society of America, Rocky River, OH (1980).

18 Figure 1: Fishbone diagram showing the major factors which affect the dimensional variability of castings. Figure 2: Distribution of feature dimensions illustrating both narrow and wide process variability with respect to the casting customer's tolerance limits. Figure 3: Shift in the mean casting size due to pattern errors. Figure 4: A casting feature made with a process with relatively wide dimensional variability, and an incorrect pattern, results in many out of tolerance castings. Table 1: The minimum and preferred size of the measurement increment for the equipment used to inspect patterns with various pattern tolerances. Figure 5a: An example of a completed gage R & R test. 12 Figure 5b: An example of a completed gage R & R test. 12 Table 2: Modified gage R & R test options that can be used to assess measurement variability when ten identical patterns are not available. Table 3: The allowed pattern tolerance and pattern measurement variability for a given casting tolerance. Pattern tolerance is assumed to be 20% of the overall casting tolerance. Table 4: The repeatability variation for some commonly used measurement instruments. The values listed are che average of several repeatability calculations. Figure 6: A summary of the steps to determine if a measurement system is adequate to inspect a pattern.

19 c

20 L 0 L w' E 0 c -0 cd --.- E?

21

22 Nominal Dimension Lower Tolerance Limit Upper Tolerance Limit Pattern Err;; rb==-)l

23 Total Pattern Minimum Measurement Preferred Measurement Tolerance Increment Increment inches mrn I inches mm inches mrn I I I I o.oo O.OOO O.OO ,

24 OPERATOW TRLAL# PART 14. AVG. 16. PART AVC (X, 'D. = 327 fw 2 trials and 2.58 for 3 trials. UCLR represents the Emit of individual R's. Circle those that are beyond this limit. Identify the cause and correct. Repeat these readings using the same appraiser and unit as originally used or discard values and reaverage and recompute R and the limiting d u e from the remaining observations. Notes:

25 Part No. & Name: r h k Characteristics: FJ~,,~C Tllxl;ne~~ Specification:.O O L r From data sheet: IT X Diff =. W ~ O GageName: Gage No: Gage Type: R, L Date: 613 Performed by: w a.0274/ = Measurement Unit Analysis % Process Variation ~~ Repeatability - Equipment Variation (Ev) I Reproducibility - Appraiser Variation (AQ %AV J [ ~ x 3 ~ 2 - ( t / & X Z ) j n r = = = 100 [AV/TVI = 1x4. % number of parts numberoftrials Repeatability & Reproducibility (R & R) RtkR = - -j JcC&+.cE)/*:.0091 Part Variation (PV PV = Rp x K3 =.o;lwx /. c A = -0 f/v3 = Total Variation (TV) TV = = J(R~LR~+Pv~) -J Parts K = = 100 [.@LuYfd rnk% ~ = *05w --. AU calculations are based upon predicting 5.15 sigma (99.0% of the area under the normal distribution curve). K, is 5.15/d2.where dr is dependent on the number of trials (m) and the number of parts times the number of operators (g) which is assumed to be greater than 15. AV - If a negative value is calculated under the square root sign, the appraiser variation (AV)defaults to zero (0). Kzis 5.15/dz. where d, is dependent on the number of operators (m) and (g) is 1, since there is oniy one range calculation. & is 5.15/dz. where d, is dependent on the number of parts (m) and (g, is 1, since there is only one range calculation. d, is obtained from Table Dh Quality Control and Industrial StatisticS. A.J. Duncan.

26 [OPTION 1 5 Measurement locations from a single pattern feature 2 Identical patterns 2 Measurementsat each location 2 or 3 Inspectors IOPTlON 2 10 Measurement locationsfrom a single pattern feature 1 Pattern 2 Measurements at each location,2 or 3 Inspectors t 0 IOPTlON 3 5 Measurement locations from a single patternfeature 1 Pattern 3 Measurements at each location 2 or 3 Inspectors I I U I I

27 Total Casting Feature Tolerance inches -TE+E I O.Oo Pattern Tolerance c Maximum Pattern Measuremt inches O.OOO t Variability mm

28 c Measurement Instrument Micrometer CMM Digital Caliper - Not a diameter Portage Machine Digital Caliper - Diameter Scale Transfer Caliper a n d Scale Repeatability 1 mm inches I

29 + casting tolerance f determine pattern tolerance select measurement equipment YES 1 - performgage R &R 4 I YES measurement system is acceptable for measuring this pattern feature

30 Prepared for the U. S. Department of Energy Assistant Secretary for Energy Efficiency and Renewable Energy Under DOE Idaho Operations Office Contract DE-FC07-93ID13235

31