Proton Exchange Membrane Fuel Cells non-invasive data-based diagnosis through a new external magnetic field measurement methodology

This article addresses a new approach to non-invasive diagnostics of proton exchange membrane fuel cells. More specifically, it deals with the detection of several faults generated by inappropriate operating conditions using an original non-invasive methodology based on the measurement of the external magnetic field around the fuel cell stack, thanks to a specially designed ferromagnetic circuit analyzer. A large experimental campaign has been released on a fuel cell stack associated with this magnetic circuit analyzer for normal and faulty fuel cell operating conditions at different current values. These experiments provide a vector of the measured magnetic field at different positions around the fuel cell stack. Each experiment was labeled based on accurate electric and fluidic measurements realized on a laboratory fuel cell test bench. These magnetic field measurements were used as original variables for a data-driven diagnostics approach involving two steps: feature extraction and classification. Finally, the high accuracy given by this diagnostic methodology proves the ability of the new magnetic field analyzer to provide a significant (highly sensitive) information to detect local faults inside fuel cell during operation, and therefore to allow accurate discrimination between different fuel cell normal and faulty operating conditions.