Abstract
Quantitative analysis is an integral part of today's manufacturing industry and process capability indices are being used for performance analysis purposes. Many currently used indices are based on theory of normality and assessing product performance with univariate quality characteristics. Very often, manufactured products have multivariate non-normal quality characteristics critical to customer satisfaction. To evaluate product performance with multivariate non-normal data is still an open challenge for researchers and there is a dire need to develop a simple and straight forward approach. In this paper, Wang's geometric distance approach to reduce high dimensionality of multivariate date to univariate date is deployed. Instead of fitting different distributions to geometric distances, a traditional practice, just one distribution, the Burr distribution is used in this paper. Also a new approach to find fitted Burr distribution parameters using simulated annealing search algorithm is introduced and results are then compared with a commonly used criterion, the Proportion of Non-Conformance (PNC). Finally, a case study using real data is presented