Multivariate Process Capability Analysis Using Non-Parametric Model...

Current process capability analysis methods rely on the normality and univariate assumptions. This paper aims at developing a novel method general multivariate process capability analysis. Conformity of product or yield is proposed as an index for process capability evaluation. A nonparametric technique, i.e. kernel density (KD) estimation and a sampling technique are proposed for model fitting and yield estimation. The unique features of the new method include: 1) conceptually transparent to practitioners, 2) easy to be implemented, 3) no assumptions on the statistical model of the process data. Case study is presented for illustration. Published in the Transactions of NAMRI/SME, Vol. 37, 2009, pp. 589-596.