Faults diagnosis for PEM fuel cell system based on multi-sensor signals and principle component analysis method

Fuel cell vehicles are becoming more popular and attracting more attention from industries, but stability and reliability of the fuel cell system (FCS) are still problems for its commercial progress. Therefore, a fault diagnosis system is essential for a reliable and long working lifetime FCS. In this work, a fault diagnosis method based on multi-sensor signals and principle component analysis (PCA) is proposed to improve FCS performance. By using this method, the correlation among different sensor signals are analyzed based on multi-sensor signals, and a simplified statistic index for fault diagnosis is deduced based on the PCA. The FCS operation conditions are monitored online, and faults in sensor and system levels are diagnosed. Experimental results show that, two typical fault scenarios, i.e., a single sensor fault and a serious system failure, can be successfully diagnosed and distinguished. For the single sensor fault, the sensor signal is reconstructed immediately to ensure that fuel cell vehicles operate normally. For the system failure, the fault can be detected in 17 s and the fault source signals can be located in 31 s, so the fuel cell stack can be protected timely. The main contribution of this work is to deduce a simplified statistic index for fault diagnosis based on multi-sensor signals and PCA method, and to provide an experimental study on identifying faults in sensor and system levels of a PEM fuel cell system.