Stream-of-Variation Theory for Automotive Body Assembly

Abstract Manufacturing systems usually consist of processes and machines in a multi-leveled hierarchy. As a result, dimensional variation in the final product is accumulated as the product moves along the manufacturing system. This paper discusses the prediction and diagnosis of dimensional variation in a multi-leveled automotive body assembly system. By combining engineering structural models with statistical analysis, the evolution of the variability and stiffness characteristics of sheet metal parts is studied in terms of assembly configurations, i.e., serial or parallel. The diagnosability of serial and parallel assembly systems is evaluated. These results are integrated with correlation clustering to form a complete diagnostic strategy. Predicting and diagnosing variation in a multi-leveled manufacturing system constitute the two aspects of what we call the “Stream-of-Variation Theory.”