A prototype expert system for fault detection of the incinerator control system

Abstract

A prototype expert system is presented in this paper for detecting failures of the incinerator steam temperature control system. The causal relationships of the steam temperature instrumentation system are represented by a Bayesian network which is a directed acyclic graph in which the nodes represent failure events, the arcs signify the existence of direct causal influence between the linked failure events, and the strengths of these influences are quantified by conditional probabilities. Based on the created Bayesian network, a reasoning inference engine is formulated to infer the most probable causes of instrumentation system malfunctions, predict the most probable future upsets or failure events for given initiating events, and interpret the observed evidence of the incinerator steam temperature control system. The inference engines for diagnosis, prediction, and interpretation presented in this paper are system or process independent. In other words, they can be applied to any other fault detection problem as long as the problem can be well defined by probabilistic causal networks.