Efficient plant supervision strategy using NN based techniques

Abstract

Most of non-linear type one and type two control systems suffers from lack of detectability when model based techniques are applied on FDI (fault detection and isolation) tasks. In general, all types of processes suffer from lack of detectability also due to the ambiguity to discriminate the process, sensors and actuators in order to isolate any given fault. This work deals with a strategy to detect and isolate faults which include massive neural networks based functional approximation procedures associated to recursive rule based techniques applied to a parity space approach. © 2010 Springer-Verlag.