A Multivariate Quality Control Procedure in Multistage Production Systems

Abstract

In this paper a multivariate quality control procedure is offered in which several correlated stages are present in production systems and in each stage there are several correlated control variable. Using Hotelling's T2 statistic, first, each stage is tested for being out of control. Then out of control variables are selected using Murphy's method. The remainder of the research involves evaluating each stage, when the present and the previous stages are both out of control. In this regard, regression of the present stages' T2 statistic on prior stages' P statistics is considered and a residual test between the present stage estimated and observed T2 statistics is made.