Dynamic Uncertain Causality Graph Applied to Dynamic Fault Diagnoses of Large and Complex Systems

Abstract

Intelligent systems for online fault diagnoses can increase the reliability, safety, and availability of large and complex systems. As an intelligent system, Dynamic Uncertain Causality Graph (DUCG) is a newly presented approach to graphically and compactly represent complex uncertain causalities, and perform probabilistic reasoning, which can be applied in fault diagnoses and other tasks. However, only static evidence was utilized previously. In this paper, the methodology for DUCG to perform fault diagnoses with dynamic evidence is presented. Causality propagations among sequential time slices are avoided. In the case of process systems, the basic failure events are classified as initiating, and non-initiating events. This classification can increase the efficiency of fault diagnoses greatly. Failure rates of initiating events can be used to replace failure probabilities without affecting diagnostic results. Illustrative examples are provided to illustrate the methodology.