Robust LPV model-based sensor fault diagnosis using relative fault sensitivity signature and residual directions approaches in a PEM fuel cell

Abstract

In this paper, a model-based fault diagnosis methodology for PEM fuel cell systems is presented. The methodology is based on computing residuals using an LPV observer. Sensor fault detection faces the problem of robustness using adaptive thresholds generated with interval observer. Fault isolation is performed using the Euclidean distance between the observed relative residuals and theoretical relative sensitivities. To illustrate the results, a commercial fuel cell Ballard Nexa© is used in simulation where a set of typical fault scenarios have been considered. Finally, the diagnosis results corresponding to those fault scenarios are presented. It is remarkable that with this methodology it is possible to diagnose all the considered faults in contrast with other well known methodologies which use the classic binary signature matrix approach.