Two failure modes related to water management in Proton Exchange Membrane fuel cells (dehydration and flooding) were investigated using electrochemical impedance spectroscopy as a diagnosis tool. It was hypothesised that each failure mode corresponds to changes in the overall stack impedance that are observable in different frequency ranges. This hypothesis was corroborated experimentally.

The experimental implementation required new testing hardware and techniques. A four-cell stack capable of delivering individually conditioned reactants to each cell was designed, built, tested, and characterised under a variety of operating conditions. This stack is the first reported prototype of its type.

The stack was used to perform galvanostatic, impedance measurements in situ. The measurements were made at three different temperatures (62, 70 and 80°C), covering the current density range 0.1 to 1.0 A cm⁻² , and the frequency range 0.1 to 4 × 10⁵ Hz. The recorded data represent the first reported set of measurements covering these ranges.

The failure modes were simulated on individual cells within the stack. The effects on individual cell and stack impedance were studied by measuring the changes in stack and cell impedances under flooding or dehydration conditions.

Dehydration effects were measurable over a wide frequency range (0.5 to 10⁵ Hz). In contrast, flooding effects were measurable in a narrower frequency range (0.5 to 10² Hz). Using these results, separate or concurrent impedance measurements in these frequency ranges (or narrow bands thereof) can be used to discern and identify the two failure modes quasi-instantaneously. Such detection was not possible with pre-existing, do techniques.

The measured spectra were modelled by a simple equivalent circuit whose time constants corresponded to ideal (RC) and distributed (Warburg) components. The model was robust enough to fit all the measured spectra (for single cells and the stack), under normal and simulated-failure conditions.

Approximate membrane conductivities were calculated using this model. The calculations yielded a range from 0.04 to 0.065 S cm−1 (under normal humidification), and conductivities that deviated from these nominal range under flooding or dehydrating conditions. The highest conductivity value (was ∼0.10 S cm⁻¹) was measured under flooding conditions at j = 0.4 A cm⁻². The lowest conductivity (∼0.02 S cm⁻¹) corresponded to a dehydrated cell at j = 0.1 A cm⁻². These values fall within the ranges of published data for modern proton exchange membranes.

The phenomenological and numerical results reported in this work represent the first demonstration of these techniques on a PEMFC stack under real operating conditions. They are also the basis of ongoing research, development, and intellectual property protection.