Comparative study between neural network and sliding mode observer for process diagnostic: Application to three tank system

In this paper we are interested in comparing the neural network method and the sliding mode observer method for fault detection. In this work we present the observer technique whose principle is to determine the difference between the output of nonlinear system and the output of the observer. In the other hand fault diagnosis using neural networks has the same structure as model-based methods: a set of signals, carrying fault information is fed into a neural machine which outputs a vector fault signal indicating normal or faulty system operation. These methods are applied to fault detection in hydraulic system with three tanks.