APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN-PROCESS FAULT-DIAGNOSIS

Fault diagnosis has been studied very actively during recent years. Es timation methods, rule-base reasoning and pattern recognition techniqu es are the most common methods used to solve problems. In recent years artificial neural networks have been used successfully in pattern rec ognition tasks and their suitability for fault diagnosis problems has also been demonstrated. However, the results presented in the literatu re usually consider very simple example situations. In this paper a re alistic heat exchanger-continuous stirred tank reactor system is studi ed as a test case. The system with 14 noisy measurements and 10 fault situations is studied. The arrangement of different fault categories i s visualized by the principal component analysis. The fault detection and diagnosis is based on the classification of process measurements a nd the classification is carried out using neural networks.