Multi-source fault detection and diagnosis based on multi-level knowledge graph and Bayesian theory reasoning

Abstract

In complex industrial processes, complex associations are often involved. This complex relationship makes traditional fault detection and diagnosis methods difficult to achieve satisfactory results in multi-source fault detection and diagnosis. Therefore, this paper proposes a new multi-source fault detection and diagnosis framework. The method can successfully detect the state of the system, and at the same time, can locate the fault source simply and quickly. This method firstly constructs the multi-level knowledge graph in the complex industrial process, and then use the discriminant coefficient R to detect whether the system has failed. If the system fails into the fault diagnosis stage, the probability of the fault is derived based on Bayesian theory. This article describes the framework in detail. The TE process is taken as an example to prove the effectiveness of the method.