Failure detection in PEM fuel cells for diagnosis
Résumé
In this paper, two approaches for PEM fuel cell diagnosis are explored. Both use non-invasive measurements. The first to be studied is fuel cell response to current step excitation, which can occur at startup or shutdown of the fuel cells. This approach leads to diagnostic methods based on models of cell and parametric estimation. The second approach is based on magnetic field measurements. These measurements are representative of current distribution inside the fuel cell stack. Magnetic field is measured by a set of 24 sensors placed in a circle around the middle of the stack. Their location allows detecting inhomogeneities in current density distribution. This approach leads to diagnostic methods based on magnetic signatures that are defined for low air stoichiometric ratio and low humidity of gas. The first approach has been applied to 8-cells PEM fuel cell stack with new cells and after six months of use. The evolution of the model's parameter due to aging could be attributed to the degradation of the active layers. In the second approach, the magnetic signatures corresponding to low air stoechiometric ratio and magnetic signature of low humidity of gas have been characterized. Therefore, these two approaches are complementary. To detect aging damage, diagnosis by current step allows observing time evolution due to aging of the model parameters. To detect a change in the operation mode, it is preferable to use a non-active measurement. That is why the diagnosis by magnetic field seems more appropriate.