DYNAMIC PROCESS DIAGNOSIS VIA INTEGRATED NEURAL NETWORKS

A generic scheme of integrated artificial neural networks has been stu died in this work for the purpose of fault detection and diagnosis in dynamic systems with varying inputs. Two general types of neural model s i.e, the feedforward networks (FFNs) and the recurrent networks, wer e integrated in the proposed fault monitoring system. To demonstrate t he utility of the proposed methodologies, extensive experimental studi es have been carried out on a pilot plant which simulates the operatio n of a semi-batch storage system. It can be observed that the predicti ons of the normal system behavior are very accurate and also, the diag nostic conclusions obtained with the integrated networks are highly re liable.