Bayesian Network Approach for Dimensional Variation Diagnosis in Assembly Process

The assembly process with hundreds of compliant sheet metal components jointed in body shop is a complex process with uncertainty. One of the issues in mass manufacturing stage is fast diagnosis of dimensional variation root cause according to the fault symptoms. This paper presents a probabilistic fault diagnosis method based on Bayesian Networks, replacing the traditional deterministic linear diagnostic model, to diagnose the root cause of dimensional variation. First, the BN structure is acquired based on the process knowledge and expert experience. Besides, according to the small sample measurement strategy of assembly process, the parameter learning method based on Method of Influence Coefficients(MIC) is utilized and particular considerations are given to the diagnostic procedures for assembly process.