A Concurrent Fault Diagnosis Model via the Evidential Reasoning Rule

Abstract

Fault diagnosis is an important tool to improve system reliability, and multiple faults always occur simultaneously in a complex engineering system. In most of the current methods for fault diagnosis, single-fault is mainly involved and concurrent faults are usually ignored, which may cause hidden danger to the safe and reliable operation of the system. Thus, it is significant to diagnose concurrent faults. However, both multiple indexes of complex systems and harsh environmental interference bring a great challenge to this task. Therefore, in this paper, a new concurrent fault diagnosis model is proposed based on the evidential reasoning (ER) rule. Specifically, multiple sub-ER models are established to form a parallel diagnosis mode. A parameter optimization model is proposed to update evidence weights and referential values. By synthesizing the outputs of all the sub-ER models, the single fault or concurrent fault can be diagnosed. The performance of the proposed diagnosis model under disturbance is analyzed theoretically, which can provide some guidance for its applications in engineering practice. A case study of the machine rotor system is carried out to validate the effectiveness of the proposed ER rule-based model.