FDI based on WD combined With PEM and RBF networks : Application to the diagnosis of TECP reactor

The objective of our work is to detect and isolate the faults that occur in large scale systems with many measurements. An advanced fault detection and isolation (FDI) method is proposed based on wavelet decomposition, Parameters Elimination Method (PEM) and Radial Basis Function (RBF) networks. Input signals are decomposed into approximations (low frequencies) and details (high frequencies). The extracted statistical parameters from approximation and detail signals pass through Parameters Elimination Method (PEM) to get rid from parameters that are useless for accurate classification. The selected parameters are used to classify the faults using a supervised RBF network. The performance of our algorithm is compared with a usual method based on classification without decomposition technique and parameters selection. The two methods are compared on the simulator of a chemical reactor: the Tennessee Eastman Challenge Process that is a well known benchmark for large scale systems.