Robust LPV model-based sensor fault diagnosis and estimation for a PEM fuel cell system

In this paper, a model-based fault diagnosis methodology for PEM fuel cell systems is presented. The methodology is based on computing residuals using a LPV observer. The problem of robustness is faced using adaptive thresholds generated with interval observers. Fault isolation is performed using structured residuals once the fault has been detected. The algorithm proposed here is able to identify and estimate multiple sensor faults presented at different time instants (k). To illustrate the results, a commercial fuel cell prototype (Ballard, 1,2kW Nexa^®) is used in simulation where a set of typical sensor fault scenarios have been considered. Finally, the diagnosis results corresponding to those fault scenarios are presented. It is remarkable that with the proposed 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.