Classification based method using Fast Fourier Transform (FFT) and Total Harmonic Distorsion (THD) dedicated to Proton Exchange Membrane Fuel Cell (PEMFC) diagnosis

In this paper, we present an approach for Polymer Electrolyte Membrane Fuel Cell (PEMFC), drying out and flooding diagnosis. The present approach is signal-processing-based pattern recognition. The voltage signal is processed by a Fast Fourier Transform, in order to carry out a frequency analysis and to calculate the Total Harmonic Distortion (THD). The TDH is used, amoung other variables, as descriptor for a global Pattern Recognition approach (fault detection, and identification through a classification approach). In this work, a supervised and non-supervised classification, by means of the K-nearest neighbor and kmeans methods, are achieved to identify the fault. This work resulted in, first, the detection of defects, through the change in the total harmonic distortion level and the frequency spectrum, secondly in the fault identification, through supervised and non-supervised classification methods with, respectively, a good classification rate of 84% and 98.5%.