Proceedings of the 2010 International Conference on Industrial Engineering and Operations Management Dhaka Bangladesh January 9 10 2010

Process Capability Analysis of a Centrifugal Casting Process

S C Mondal Department of Mechanical Engineering, Bengal Engineering and Science University, Shibpur, Howrah 711 103, West Bengal India J Maiti, and P K Ray Department of Industrial Engineering and Management, Indian Institute of Technology Kharagpur, Kharagpur 721 302, West Bengal India Abstract
The concept of process capability was proposed in early 1970 with the introduction of capability ratio by Juran, a world known quality expert. Thereafter almost 40 years have been passed and many indices such as Cp, Cpk and Cpm were developed for measuring process performance. However, many researchers, for example, Kotz and Johnson [1-2] raised concerns about the application of the diversified indices owing to three reasons (i) too theoretical for practitioners, (ii) mismatch with performance values measured through different indices, and (iii) inadequacy to mimic the reality. This study therefore focused on the application of the different indices developed so far with a case study on a centrifugal casting process. Five important quality variables were considered and 12 UPCIs (such as Cp, Cpk, Cpm, ., Spmk, and Ca) values were calculated based on data collected from the plant. Results obtained are compared with percent conforming products as well as amongst the indices themselves. The values differ considerably for some indices. The application shows that the developed indices can create confusion to practitioners unless a clear and concise implementation scheme is provided. Keywords: Process capability, UPCI, centrifugal casting, implementation scheme.

1. Introduction
Casting is one of the important metal forming processes used in foundry. There are several types of casting processes namely die casting, slush casting, investment casting, centrifugal casting and continuous casting. Centrifugal casting is a special type of casting process used for making hollow parts by rotating the mould at a predefined rpm and pouring the molten metal at a uniform rate from a ladle. The molten metal is solidified in the rotating mould. The size and shape of the job depends on the shape of the mould. A particular mould size is used for a particular type of job. If size of job change, the mould size also changes accordingly. The quality of the cast product depends on the rate of solidification, pouring time, pouring temperature, ladle preheat temperature and rpm of the mould. In centrifugal casting, the mechanical properties of material are considered as more critical than the dimensional characteristics (having a wide range of specification). In the centrifugal casting process, the input materials to the crucible are ingot, and virgin metal which are a compound of copper, tin, silicon, nickel and phosphorous. The output quality characteristics are percent copper, percent tin, percent nickel, percent phosphorus and hardness which affect the performance of the cast product during application. The multiple quality characteristics with varied tolerance limits pose challenges for producing good quality end products. It is said that good quality product is a result of a capable process. Therefore, many theories and methodologies were developed for process capability analysis [3-14]. While there are many indices developed to measure process capability, most of the studies are done based on simulated data. Application in real world problems, primarily in manufacturing industry, is less compared to other industries. Though Kane [5] discussed the various applications of Cp and Cpk in Japanese industries and the U.S. automotive industry around 1980s, yet its application in manufacturing has got momentum only recently [15] such as in ball bearing manufacturing [16], in heat treatment [17], in voltage level translator manufacturing [18], in stamping operation [19], in tube-bending process [20], in machining [21], in silicon-filler manufacturing [22], in forging operation [23], in electronics manufacturing [24] and in drug tablet manufacturing [25]. However, many researchers, for example, Kotz and Johnson [1-2] raised concerns about the application of

the diversified indices owing to three reasons (i) too theoretical for practitioners, (ii) mismatch with performance values measured through different indices, and (iii) inadequacy to mimic the reality. This study is another addition to this thrust area with an application in a centrifugal casting process. Five important quality variables were considered and 12 UPCIs (such as Cp, Cpk, Cpm, ., Spmk, and Ca) values were calculated based on data collected from the plant. Results obtained are compared with percent conforming products as well as amongst the indices themselves.

2. Process capability indices

There are many indices developed so far for measuring process capability. In this section, the mathematical brief of each of the univariate process capability indices applied in the case study plant is presented. The indices are Cp [3], Cpk and CPU/CPL [5], Cpm [6], Cpmk [7], PCI [26], Cpp [27], Cs [28], Cpw [29], Spmk [16], Ca [30] and PU/ PL [11]. For details, interested readers are requested to take help of the references. The concept of process capability was first introduced by Juran [3] which is the ratio of specification range (USL-LSL) to the process variation (6) and is known as capability ratio (Cp). It is designated as
C p = USL LSL 6

. Cp does not consider the location of mean () which is captured by Cpk [31, 5] where,
U SL 3

Cpk = Cp (1 k ) and k is termed as the bias factor. For one-sided specification, Kane [5] proposed upper and

lower capability indices as

CPU =

and CPL =

LSL 3

, respectively. Pearn and Chen [11] developed a

C P U = b n -1

procedure to obtain unbiased estimators of PU and PL as

(U S L - y )
3S

and

C P L = bn 1

( y LSL )
3S

respectively, where y , S and n are the sample mean, sample standard deviation and sample size, respectively and bn-1 is the correction factor. Chan et al. [6] proposed Cpm considering specification range, process variation and variation of mean from the target (target deviation) which is defined as C pm = USL LSL . Pearn et al. [7] proposed a process
6 2 + ( T )2

capability index (Cpmk) considering all the characteristics features of Cp, Cpk and Cpm and defined as
USL LSL C pmk = min , 2 3 2 + T 2 3 2 + ( T ) ( )

. It is most suitable for symmetric tolerance and not appropriate in one-

sided specification limit. Greenwich and Jahr-schaffrath [27] proposed Cpp as the reciprocal of the square of Cpm [6]. Spiring [29] developed a similar type of process capability index as proposed by Chan et al. [6] based on weight function approach and is given as C pw = USL LSL , where w represents a weight
6 2 + w( T )2

function. Fraction non-conforming items, process capability and process yields are closely related. Carr
[26] proposed PCI as PCI = 1 1 1 1 p , where p is the expected proportion of non-conforming products and 3 2

denotes cumulative distribution function (CDF) of standard normal distribution. Chen and Ding [16] introduced a new process capability index (Spmk) as
S pm k = 3 p 1 1 2 2 T 1+

. The process capability indices

discussed herein were developed with the assumptions that data follows normal distribution. Indices for non-normal data are not considered in this paper.

3. Case Study
3.1 Data For this study, a centrifugal casting process is chosen which manufactures worm wheel rim of sizes 2763 , and 41 with an average annual production rate of 60 components. The average non-conforming rate in casting is 3 items per annum. The production rate of 41 worm wheel is more compared to other two

products and is considered in this study. 40 observations data for 41 worm wheel component were collected during a period of three years (2004-2007). The casting process for the case study plant is shown in Figure 1. The input materials to the crucible furnace are ingot, and virgin metal which are a compound of copper, tin, silicon, nickel and phosphorous. Cu is the main alloying element in the worm wheel, whereas, Sn increases hardness and anti-fluidity properties, Ni improves toughness and tensile strength and Phosphorus is used to improve fluidity, machinability and antioxidation property. Some impurities (e.g., Zinc, lead, Iron) are occasionally present but in small quantity (less than 0.5%) and do not affect much on product performance. The input materials are heated to a temperature of around 1050 - 11500c and the molten metal is then poured into a rotating mould using a ladle. The quality characteristics along with specification limits of the worm wheel, the key input process parameters, and the input raw material characteristics are shown in Tables 1 3. Control parameters Molten metal Control parameters

Ingot, Virgin metal (Cu, Sn, Ni, P)

Crucible furnace

Centrifugal Casting machine

Worm wheel rim

Noise parameters

Noise parameters

Figure 1: Input-output process model for centrifugal casting Table 1: The worm wheel quality characteristics along with specification limits Quality characteristics Range of values Lower level Higher level Cu (%) 87.22 88.72 Sn (%) 9.75 10.75 Ni (%) 1.25 1.75 P (%) 0.03 (max) Hardness (BHN) 90 (min) Table 2: Process parameters (control and noise) along with their range of values Input process parameter Range of values Lower level Higher level Pouring temperature (0c) 1245 1260 RPM of the centrifugal 110 120 casting machine Charging time (minute) 630 750 Ladle preheat temperature (0c) 350 400 Humidity (%) 70 95 Room temperature (0c) 16 45 Table 3: Quality characteristics of the input materials Quality characteristics Range of values Lower level Higher level Cu (%) 87.70 88.94 Sn (%) 9.90 10.95 Ni (%) 1.30 1.95 P (%) 0.03 (max)

3.2 Analysis and Results The analysis consists of computation of process capability index values for 12 different indices for 5 quality variables, namely Cu, Sn, Ni, P and hardness. The computed values are shown in Table 4. Of the capability indices, Cpp measures the process incapability and therefore, its reciprocal can be compared with other indices. For the quality characteristic Cu, there is not much variability in the index values (0.007). The variability in the index values increases for Sn (0.03), Ni (0.04), P (0.15) and hardness (6.74). Table 4: The values of univariate process capability indices for the casting process Quality characteristics References Juran (1974) Kane (1986) Kane (1986) Chan et al. (1988) Pearn et al.(1992) Carr (1991) Greenwich and Jahr Schaffrath (1994) Wright (1995) Spiring (1997) Chen and Ding (2001) Chen et al.(2001) Pearn and Chen (2002) Proposed indices Cp Cpk CPU/CPL Cpm Cpmk PCI Cpp Cs Cpw Spmk Ca PU/ PL 0.799 0.79 0.75 1.55 0.76 0.79 0.76 0.99 0.58 0.48 0.65 2.98 0.47 0.51 0.63 0.99 0.62 0.38 0.75 2.57 0.38 0.55 0.53 0.92 Cu 0.8 0.79 Sn 0.61 0.5 Ni 0.89 0.55 P 1.59 1.47 1.7 1.5 1.39 1.33 0.44 0.75 1.47 1.22 0.93 1.67 Hardness 8.81 3.32 3.32 0.24 0.1 1.33 17.36 0.1 3.32 0.04 0.38 3.25

Variability in index 0.007 0.03 0.04 0.15 6.74 values* *Variance of index values for the quality variables. The reciprocal of Cpp is considered in variance calculation. In comparing the capability indices (Cp, Cpk, Cpm and Cpmk), the general conception is Cp>Cpk>Cpm>Cpmk which is not achieved for the casting process (except Cp>Cpk). For Cu, Sn and Ni, all the capability index values are less than 1, whereas for P and hardness a mixed result is obtained. A closer look into the data reveals that there are 2 nonconforming (NC) items attributed to Sn and one item attributed to each Cu and Ni. No NC is attributed to P and hardness. The value of Cp for Ni (0.89) is more than for Cu (0.8) though both producing same number of NC items. This is because Cu (s = 0.310) poses more process variation than Ni (s = 0.092). The other notable feature from this study is that for quality characteristics with two sided specification limits (e.g., Cu, Sn and Ni), Cp, Cpk and CPU/CPL are the better measure of process performance and for unilateral specification limit, these are inappropriate. Of the quality variables considered, the casting process is the most capable for the variable P as it produces no NC items, poses minimum variance (s = 0.003), minimum Cpp (0.44), and its mean is very close to target (-T = -0.0161). The process is the least capable with respect to Sn as it produces two NC items, poses more variance (s = 0.275) and maximum Cpp (2.98) (except hardness).

4. Conclusions and Future Scope of Research

The case study results provide the insights about the casting process in terms of process variation, target deviation (bias), location of the mean in the specification interval and fraction non-conforming. The effect of mean shift is more compared to improper centering of the specification (target) limit. The general tradition is to take the average of the specification interval to find the target. This concept is useful when the data have two sided specification limits. When the process mean moves away from the center of the

specification limit, particularly in case of unilateral specification limits (e.g., hardness), the process capability index values reduce drastically. Future research should be conducted to adequately define the target for unilateral specifications. From development point of view, both Cp and Cpk provide information about the process variation and fraction NC. Cpm emphasizes on target deviation and Cpmk considers all the criteria that Cp, Cpk and Cpm pose. Cp measures process potential capability where Cpk measures actual capability (process performance). Cp, Cpk, Cpm, Cpmk, Cpw, Ca and Cs are appropriate to measure process performance with two sided specification limits, however, CPU/CPL and CPU/CPL are used for processes with one sided specification limits. Further, these indices assume that data follows normal distribution. The developed PCIs can be classified in two categories, namely (i) variation based and (ii) fraction NC based. There are ten PCIs that use process variation and target deviation in the calculation. Two PCIs, namely PCI (Carr, 1991) uses fraction nonconforming and Spmk (Chen and Ding, 2001) uses fraction NC, process variation and target deviation. The PCI is independent of specification limit and process variation and target deviation. Among the existing capability indices, the index Spmk is theoretically more robust as it considers most of the criteria of capability analysis and is able to measure process precision (variation based) and accuracy (process yield). However, the Spmk does not consider the location of the mean with respect to the specification interval. For the case study, it didnt give better result for unilateral specifications. Future studies incorporating more industry cases are required to properly realize the theoretical merits of Spmk. From application point of view, the plant is not performing well. The causes of low index values for Sn and Ni need to be explored. As the quality of the end product very much depends on the quality of the input materials which are supplied directly from vendors or from preceding manufacturing stages, a boarder quality assurance scheme involving suppliers seem to be missing for the plant studied. Other criteria that to be looked into for the casting process, are human and machine tool capability. Amongst the quality variables, hardness is the key variable which to a larger extent determines the performance of the worm wheel in industry uses. The process engineers are therefore more inclined to measure and interpret process capability indices for hardness. The capability index values vary significantly from one index to another (s = 6.74). This creates confusion in the practitioners mind which was posed by Kotz and Johnson. Functionally, the indices that were developed for unilateral specification limits should be used for hardness. For the case study, the authors recommend the use of those indices that have almost similar values and that better represent the characteristic hardness. It is observed that Cpk, CPU, CPU and Cpw values match closely for hardness while the other indices differ significantly. Moreover, the later indices also vary amongst themselves. So, the process engineer can chose any one of Cpk, CPU, CPU and Cpw indices for measuring the capability value for hardness. This finding also supports Kotz and Johnsons concern. Of the recommended indices for hardness, the mathematical complexity varies. For example, in order to calculate CPU, one requires estimating bn-1 which is a correction factor. Eventually, the recommend is then to use Cpk, the simplest in the lot. The questions of why there are so many indices or which one to chose? are still unanswered. Other notable feature of this study is that the variability in index values is more for one sided specification limits (see Table 4). In order to overcome the concerns raised by Kotz and Johnson [2], the authors, based on this case study, suggest the following for development and use of process capability indices: (i) A general scheme with step-by-step procedures for selection, execution and interpretation of process capability indices needs to be built that can be used across industries with little modifications, and (ii) As different indices were developed to overcome the limitations of the former, the indices should be grouped functionally, i.e., indices for unilateral, bilateral specification limits. The process engineers should be trained with the functional characteristics of variables vis a vis process capability fundamentals. Acknowledgements: The authors gratefully acknowledge the kind support and cooperation provided by the case study plant.

References
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. Kotz, S., and Johnson, N.L., 1993, Process Capability Indices, Chapman & Hall, London. Kotz, S., and Johnson, N., 2002, Process Capability Indices a review 1992-2000, Journal of Quality Technology, vol. 34, No. 1, pp. 2-19. Juran, J. M., 1974, Quality Control Handbook, third edition, McGraw-Hill, New York. Hsiang, T.C., and Taguchi, G., 1985, A tutorial on quality control and assurance the Taguchi methods, In: ASA Annual Meeting, Las Vegas, Nevada, USA. Kane, V.E., 1986, Process Capability Indices, Journal of Quality Technology, vol. 18, No. 1, pp. 4152. Chan, L.K., Cheng, S.W., and Spiring, F.A., 1988, A new measure of Process Capability: Cpk, Journal of Quality Technology, vol. 20, No. 3, pp. 162-175. Pearn, W. L., Kotz, S., and Johnson, N.L., 1992, Distributional and inferential properties of process capability indices, Journal of Quality Technology, vol. 24, No. 4, pp. 216-233. English, J.R., and Taylor, G.D., 1993, Process Capability Analysis: a Robustness Study, International Journal of Production Research, vol. 31, No.7, pp. 1621-1635. Yeh, A.B., and Bhattacharya, S.K., 1998, A robust process capability index, Communications in Statistics Simulation and Computation, vol. 27, No. 2, pp. 565-589. Frey, D. D., Otto, K. N., and Wysocki, J. A., 2000, Evaluating Process Capability during the Design of Manufacturing Systems, Journal of Manufacturing Science and Engineering, vol. 122, pp. 513-519. Pearn, W. L., and Chen, K. S., 2002, One-sided capability indices CPU and CPL: decision making with sample information, International Journal of Quality and Reliability Management, vol. 19, No. 3, pp. 221-245. Wu, C.C., Kuo, H. L., and Chen, K.S., 2004, Implementing Process Capability Indices for a complete Product, International Journal of Advanced Manufacturing Technology, vol. 24, pp. 891-898. Hsieh, K-L., and Tong, L-I., 2006, Incorporating process capability index and quality loss function into analyzing the process capability for qualitative data, International Journal of Advanced Manufacturing Technology, vol. 27, pp. 1217-1222. Hsu, Y.-C., Pearn, W.L., and Wu, P.-C., 2008, Capability adjustment for gamma processes with mean shift consideration in implementing Six Sigma program, European Journal of Operational Research, vol. 191, pp. 517-529. Wu, C-W., Pearn, W.L., and Kotz, S., 2009, An overview of theory and practice on process capability indices for quality assurance, International Journal of Production Economics, vol.117, pp. 338-359. Chen, J.-P., and Ding, C.G., 2001, A new process capability index for non-normal distributions, International Journal of Quality and Reliability Management, vol.18, No.7, pp. 762-770. Dorward, R. C., 2002, Integrating product capability statistics with process models, Journal of Materials Processing Technology, vol. 122, pp. 331-337. Lin, P. C., and Pearn, W. L., 2002, Testing process capability for one-sided specification limit with application to the voltage level translator, Microelectronics Reliability, vol. 42, pp. 1975-1983. Ahmed, S.E., 2005, Assessing the process capability index for non-normal processes, Journal of Statistical Planning and Inference, vol. 129, pp. 195-206. Chen, S.K., Mangiameli, P., and Roethlein, C. J., 2005, Predicting the output of a tube-bending process: A case study, International Journal of Production Economics, vol. 95, pp. 307-316. Motorcu, A.R., and Gullu, A., 2006, Statistical process control in machining, a case study for machine tool capability and process capability, Materials and Design, vol. 27, pp. 364-372. Chen, K.S., Yu, K.T., and Sheu, S.H., 2006, Process capability monitoring chart with an application in the Silicon-filler manufacturing process, International Journal of Production Economics, vol. 103, pp. 565- 571. Chang, Y. C., and Wu, C.-W., 2008, Assessing process capability based on the lower confidence bound of Cpk for asymmetric tolerances, European Journal of Operational Research, vol. 190, pp. 205227. Wu, C.W., 2008, Assessing process capability based on Bayesian approach with subsamples, European Journal of Operational Research, vol. 184, no. 1, pp. 207-228. Merwe, A. V., and Chikobvu, D., 2009, A process capability index for averages of observations from new batches in the case of the balanced random effects model, Journal of Statistical Planning and Inference. www.elsevier.com.

26. Carr, W. E., 1991, A new process capability index: parts per million, Quality Progress 24, vol. 24, pp. 152. 27. Greenwich, M., and Jahr-Schaffrath, B.L., 1994, A process incapability index, International Journal of Quality and Reliability Management, vol.12, No.4, pp. 58-71. 28. Wright, P. A., 1995, A process capability index sensitive to skewness, Journal of Statistical Computation & Simulation, vol. 52, pp. 195-203. 29. Spiring, F.A., 1997, A Unifying Approach to Process Capability Indices, Journal of Quality Technology, vol. 29, No. 1, pp. 49-58. 30. Chen, K.S., Huang, M.L., and Li, R.K., 2001, Process capability analysis for an entire product, International Journal of Production Research, vol. 39, No. 17, pp. 4077- 4087. 31. Sullivan, l. P., 1984, Targeting variability a new approach to quality, Quality Progress, vol. 17, No. 7, pp. 15-21. 32. Johnson, N.L., 1949, System of frequency curves generated by methods of translation, Biometrika, vol. 36, pp. 149-176. 33. Box, G., and Cox, D. R., 1964, An analysis of transformation, Journal of Royal Statistics Society B, vol. 26, No. 211-243. 34. Somerville, S. E., and Montgomery, D. C., 1996, Process capability indices and non-normal distributions, Quality Engineering, vol.19, No. 2, pp. 305-316.