Signal-Based Diagnostics by Wavelet Transform for Proton Exchange Membrane Fuel Cell

In order to exploit all the benefits from the Proton Exchange Membrane Fuel Cell (PEMFC) technology and to gain a deeper understanding of operating faults during fuel cell operations, Investigation of the origins of faults is necessary. In this work, a diagnosis approach consisting of a method using signal-based pattern recognition is proposed. It is aimed at a minimization of efforts and costs in acquisition and evaluation of data for diagnostic purposes. All information needed to locate the faults is drawn from the recorded fuel cell output voltage, since certain phenomena leave characteristic patterns in the voltage signal. A signal analysis tool, namely the Wavelet Transform (WT), is employed to identify different patterns or faults signatures. The approach has been applied to voltage data recorded on a PEMFC suffering from dysfunctions related to inappropriate humidity levels inside the cell (two different faults are simulated : flooding and drying out). Characteristic features in the output voltage signals were outlined, so a distinction of several states of health was accomplished. The results show the efficiency of the proposed approach, and the WT can be considered as a reliable method to localize the dysfunctions. A comparison between the Discrete Wavelet Transform (DWT) and the Continuous Wavelet Transform (DWT) has shown that the DWT is more efficient in detecting and localizing faults in fuel cells.