Assessing the risk of product infant failure based on Bayesian network and manufacturing quality variation propagation

Prevention of product infant failure is crucial for quality improvement in manufacturing. And infant failure is the concentrated embodiment of manufacturing quality variation. However, how the quality variation propagates to infant failure and what the impacts of the infant failure cannot be explained clearly so as to impede the effective prevention of infant failure. Therefore, in this paper, infant failure risk is firstly analyzed and assessed to control quality variation propagation and reduce failure impacts simultaneously. First, the connotation of infant failure risk is clarified and an infant failure risk formation chain 'Process Quality Variation - Physical Defect - Functional Vulnerability - Infant Failure' is proposed to explain the formation mechanism of infant failure risk. Second, based on the chain, a Bayesian network is further adopted to unfold the chain to assess the infant failure risk. And the method is executed in three consecutive stages: modeling a Bayesian network by qualitatively turning chain nodes into network nodes (KQC nodes, component nodes and function nodes) and quantitatively determining the probabilities of KQC nodes' variations causing failure based on the stream of quality variation and propagation relationships based on BP neural network; analyzing the risks in terms of probabilities of infant failures (PIFs) and impacts of infant failures (IIFs); and developing an infant failure risk assessment map. Finally, a case study of computer board is introduced to verify the applicability of the proposed method. The final result shows that the proposed method has a good performance in infant failure risk assessment.