An Integrated Approach to On-line Fault Detection and Diagnosis - Including Artificial Neural Networks with Local Basis-Functions

Abstract Operator support systems for on-line fault detection and diagnosis rely on the plant knowledge compiled into the algorithms. In this paper we emphasis that several complementary representations of this knowledge should be used together to increase the reliability of the system. This leads to an integrated approach where first principles dynamic models, rule-bases and pattern recognition techniques are integrated. The integrated approach is discussed in general, and in particular we consider a pattern recognition module based on an artificial neural network with local basis-functions. With a simulation example we show how pattern recognition and a dynamic model of the plant can be combined.