A reasoning algorithm of applying causality diagram to fault diagnosis of complex hybrid systems

Fault diagnosis of a complex hybrid system, which bears causality loops, non-deterministic system dynamics, presence of not-directly-observable system states, delay of fault effects, and the mixture of discrete and continuous variables in the observed signals, is a very challenging task. This paper presents a novel method based on the causality diagram model for this problem. The method extends the causality diagram to the 2-time-slice causality diagram (2-TSCD), which represents both the fault propagation and the conditional probability distribution of system states in two consecutive time-slices. In the 2-TSCD, basic event variables represent failures; node event variables represent system states at time-slice t+l; some system states at time-slice t are introduced as basic event variables to represent the delay of fault effects. We develop a reasoning algorithm for the 2-TSCD: firstly, the candidate fault modes at time-slice t+l are acquired by using fault propagation; then all system states at time-slice t, which are introduced as basic event variables and cannot be observed directly, are predicted based on the 2-TSCD in previous time-slice; and the posterior probabilities of candidate fault modes are calculated and ranked at last. The advantages of the proposed method are the flexibility of knowledge representation and the rapidity of diagnosis.