MLP in Robust Fault Detection of Static Non-linear Systems

Abstract

One of the most desirable features of the model obtained during the system identification is small modelling uncertainty which is defined as a mismatch between the model and the system being considered [10]. It follows from the fact that the effectiveness of the fault detection systems depends on the uncertainty of the neural model and disturbances existing in the industrial system. In the case of the most widely applied the ANNs such as the MLP, the model uncertainty can appear both during structure selection of the neural model and parameters estimation.