Filtering-Based Actuator Fault Diagnosis using MLP Neural Network for PDFs

In many practical processes, the measured information is the stochastic distribution of the system output rather than its value. In this paper the fault diagnosis (FD) problem is considered by using the output stochastic distributions. A multi-layer perceptron (MLP) neural network is adopted to approximate the probability density function (PDF) of the system outputs and nonlinear principal component analysis (NLPCA) is applied to reduce the model order for a lower-order model. For such a discrete-time dynamic model with nonlinearities, uncertainties and time delays, the concerned FD problem is investigated. The measure of estimation errors represented by the distances between two output PDFs, would be optimized to find the diagnosis filter gain. Simulation example is given for the weighting dynamics to demonstrate the effectiveness