Fault sensitive modeling and diagnosis of PEM fuel cell for automotive applications

In this paper the PEMFC fault diagnosis is based on neural network modeling approach combined to numerical simulation in which a new developed sensitive model of PEMFC has been especially used. In literature the voltage changing is evaluated according to a variation of the total electrical resistance by assuming the same physical parameters (temperature, pressure...) in whole area of the FC cells. The main contribution of this work is to consider a 2D variation of temperature, pressure and humidity within cathode-membrane-anode zone of the PEMFC by using a multi-loops/nods circuit model. Then a NN model has been developed to classify faults and to recognize them on line during the FC operating.