STATISTICAL METHODOLOGIES FOR QUALITY ANALYSIS : TWO APPLICATION STUDIES

Abstract

This thesis presents two studies related to statistical methodologies for quality analysis. The first part deals with a statistical methodology for the analysis of multi-components repairable systems while the second is a study on non-normal process capability analysis. The reliability study considers the problem of characterising the failure of a repairable system through proportional hazard modelling. The application of modified Leemis nonparametric estimation in determining the baseline hazard is discussed. A generalised imperfect maintenance model that allows quantification of the effects of corrective and preventive maintenance on the system is also derived. In the analysis on process capability, an overview of some popular process capability indices (PCIs) is given. The simplicity of such indices has led to much abuse, in particular, their application to process distributions that deviate from the usual assumption of normality. Several existing methodologies that handle non-normal PCIs are discussed. The results obtained shown that the Box-Cox transformation is feasible for such situations.