Variation source identification of multistage manufacturing processes through discriminant analysis and stream of variation methodology: a case study in automotive industry

Product quality problem is a critical issue for multistage manufacturing processes, especially in continuous production lines whereby quality characteristics are measured at the end of the line. Therefore, it is important to reduce process variation by identifying its sources and eliminating its causes. In this regard, a novel approach, to identify the source of variation in multistage manufacturing processes through integration of the Fisher's linear discriminant analysis and the stream of variation methodology, is proposed. Linear discriminant analysis separates the variation of the quality characteristics through the manufacturing process stages while the stream of variation methodology is used for variation propagation modeling in multistage manufacturing processes. Finally, the future deviation is assigned into the analysis in order to identify the source of variation. With an illustrative case study, it is concluded that the proposed approach improves fault diagnosis of continuous production lines in multistage manufacturing processes.