Robust Method of Multiple Variation Sources Identification in Manufacturing Processes For Quality Improvement

Abstract

Variation source identification is a critical step in the quality and productivity improvement of manufacturing processes and draws significant attention recently. In this article we present a robust pattern-matching technique for variation source identification. In this paper, a multiple variation sources identification technique is developed by adopting the linear relationship between variation sources and product quality characteristics, which is described by a coefficient matrix. The columns of the coefficient matrix are treated as the signatures of corresponding variation sources. The matching is conducted between the signature vectors and the eigenvectors of the sample covariance matrix of the product quality measurements. Multiple faults are allowed in the matching. Further, both the perturbation of unstructured noise and the sample uncertainties are considered in this matching method. A comprehensive case study illustrates the effectiveness of this method. This robust method can be used for root cause identification of manufacturing processes. The application of this method can significantly reduce the troubleshooting time and hence improve the quality and productivity of manufacturing processes.