Assesment of multivariate process capability indices for ceramic manufacturing process

Abstract

Process Capability Jndices are intended to provide a single-number asses,ment of the consistency of manutllCturing process relative to the engineering specification limits on qUlllity characteristics. In many indui>trial instances, product quality depends on a multitude of dependent characteristics. As a consequence, choice of capability indices shifts from univariate to multivariate domain. However, sclccting the correct multivariate process capability indices (MPCI,) for a manufacturing process is always a problem for its mathematical complexity and different inferential ability. In this study five different multivariate methodologies arc used for measuring and comparing MPCIs of ceramic manufacturing processes. The methods are proposed by Taam (1993), Chen (1994), Shahriari (1995), Braun (2004) and Ca,tagliola (2005). The study has heen performed in two ceramic industries, i.e. Shinepukur ceramic industries and Mirpur Ceramic Works Ltd. The mathematical software MATLAB is used for generating simulated data with bivariate normal distribution and observed covariance. It wa, also used to examine the sensitivity of the MPCTsif the processes mean is shifted towards the target. For both the real and simulated cuvironments, only Castagliola's method shows consistency of results for two manufacturing processes. The rejection rate estimated by this method closely resembles with that claimed by the management. The Castagliola's method can best estimate the capability for both industries. Other four methods cannot correctly reflect the real ,ituation and do not react accordingly with simulated situation. However, based on their MPCI values and expected rejection rates, the method by Chen is the better perfonner followed by Taam, Shahriari and Brauns, respectively. This project work provides a comprehensive insight of the five multivariate process capability indices for applying in ceramic manufacturing industries. The outcome of this research work will provide guidance to the personnel involved in dealing with processes in similar manufacturing industries.