Abstract
There are many instances in which the quality of a product or stability of a process is determined by several variables. Most often there exist correlations among the variables monitored. In such situations one wishes to take advantage of the correlations among the variables by monitoring all the variables simultaneously using a single control procedure rather than monitoring the variables on an individual basis. Multivariate control procedures are commonly used for monitoring such process. The relation between the variables, however makes it difficult to identify variables that are responsible for the off-target signal. This thesis presents a diagnostic procedure to aid in identifying variables responsible for the off-target signal. The proposed diagnostic procedure is triggered by off-target signals from the multivariate control chart. The performance of the procedure are investigate in conjunction with the two control charts, the X2 chart and the MC1 chart. Two performance measures are proposed to evaluate the ability of the diagnostic procedure to identify off-target variables. The estimate of these performance measures are obtained for various shifts of the process mean in different directions and distances. The results are presented for three and five variable cases. Simulation results based on performance measures are then discussed. Finally a summary of the outcome of this research effort is presented.
Keserla, Adhinarayan A. (1993). A diagnostic procedure for multivariate quality control. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1993 -THESIS -K42.