Functional Failure Diagnosis Method of Manufacturing System Based on Dynamic Bayesian Network

Abstract

The core function and mission of the manufacturing systems are to produce high-quality and high-reliability products continuously and steadily, and fault diagnosis of a manufacturing system under operation is a precondition to ensure its normal operation and complete the expected functions. Most of the previous studies mainly focus on the fault diagnosis of production equipment in manufacturing systems. Over the years, manufacturing systems have become increasingly large and complicated, making physical failures of production equipment in operation have become rare, and the proportion of functional failures has increasingly increased, which are manifested by the dynamic degradation characteristics of manufacturing systems. Therefore, a functional failure diagnosis approach of the manufacturing systems based on the SQR chain and the dynamic Bayesian technique is proposed in this paper. Firstly, the functional fault connotation and formation mechanism of manufacturing systems are defined based on the SQR chain theory and the deviation flow theory. Secondly, the overall dynamic degradation characteristics of systems are clarified from the dynamic Bayesian network (DBN), and the multi-station manufacturing process quality deviations are used to evaluate the functional failure state for a manufacturing system. Third, the underlying cause of equipment or process change can be analyzed by the SQR chain and the professional knowledge, and a functional fault diagnosis strategy for manufacturing systems is provided. Finally, a case study of the cylinder head as a production system is conducted to illustrate the effectiveness and accuracy of the proposed approach.