Design of An Intelligent System for Fault Diagnosis in Chemical and Petrochemical Processes

In this paper, we present an intelligent operator support system (IOSS) for detection of process upsets and determining their causes. The system uses a simple conventional back-propagation (BP) artificial neural network (ANN) to achieve its diagnostic purposes. However, the capability of the proposed methodology is enhanced to simultaneously diagnose multiple faults and their severity levels via employing a number of functional units to the network input layer. The effectiveness of the proposed approach has been examined through simulation study of a Heptane-to-Toluene process in steady state operation.