Detecting multiple special causes from multivariate data with applications to fault detection in manufacturing

Abstract

Several articles in the recent literature propose linear models of product quality for both single station and multistation manufacturing processes. We show how these models may be used in conjunction with statistical methods to design a procedure for multivariate Statistical Process Control (SPC) that outperforms direct application of multivariate SPC. We show how to design the procedure and evaluate its performance in shift detection for models with and without singularities. The use of the procedure is illustrated using two examples from automobile body assembly.

Keywords::

• Multivariate exponentially weighted moving average
• multivariate process control
• state-space models

Acknowledgement

The authors thank the anonymous reviewers and the Department Editor for several helpful comments and suggestions that have resulted in an improved article.

Notes

*A cumulative sum, rather than an EWMA, is used.