Information fusion in fault diagnosis for automotive fuel cell system based on D-S evidence theory

In order to detect the inner healthy state of 60 kW automotive fuel cell system designed by our group, according to the typical faults of flooding and drying of proton exchange membrane fuel cell(PEMFC), this paper presents an informagtion fusion fault diagnosis method based on both neural networks and D-S evidence theory. Firstly, from the qualitative analysis of water transmission mechanism of PEMFC, the experimental data from multi-sensor such as voltage, current, pressure, temperature and so on is re-processed and utilized to train two independent BP neural networks, then the output values of the neural networks are directly taken as the primary diagnosis results and new fault belief function assignment of the frame of discernment fused with D-S evidence theory to get the final diagnosis results. By comparing the diagnosis results of the two independent BP neural networks with the ones based on D-S evidence theory fusion method, the experimental results demonstrate that the reliability and accuracy of the later one are much higher as to the health diagnosis for flooding and drying of PEM. © 2011 Binary Information Press.