Retrospective Hotelling's T2 Control Chart for Automotive Stamped Parts: A Case Study

Abstract

Traditional univariate control charts designed to monitor single variable quality characteristic have been successfully used in manufacturing processes. However, in manufacturing environment, variable settings are essentially multivariate that univariate control charts are not suitable for monitoring purposes. The Hotelling’s T² multivariate control chart is a powerful statistical tool for modeling multivariate production systems and can shed light on how variables are interrelated to facilitate better understanding of process variations. This study deploys a multivariate control-charting scheme to monitor the quality of a manufactured part in a Malaysian-based automotive parts manufacturing company, as a case study.  Three major steps in the Hotelling’s T² retrospective operation are outlier deletion, variable selection and parameter estimation are methodically described in this paper. When applied to new sample observations of selected quality variables, the T² control chart reveals an ‘out-of-control’ condition, thus confirming the need for quality enhancement in the locally produced stamped part.