Application of the MLP neural network to the robust fault detection

Abstract

This chapter presents an identification method based on artificial neural networks, which can be applied to the robust fault detection. The chapter shows how to estimate parameters of the multilayer perceptron (MLP) neural network and the corresponding uncertainty. Based on the neural model, a novel robust fault detection scheme is proposed, which supports the diagnostic decisions. During the testing of the proposed approach almost all faults defined in the development and application of methods for actuator diagnosis in industrial control systems (DAMADICS) benchmark were detected. However, the feasible parameter set obtained with this approach is only an approximation of the original one.