3D current density reconstruction in a fuel cell stack with external magnetic field measurement
Résumé
Today, optimization of the fuel cell lifetime is a challenge for many researchers. Controlling fuel cell aging by the diagnosis allows us to prevent a possible breakdown and ensure its maintenance at the appropriate time. The originality of the presented diagnosis technique consists in being able to reconstruct the distribution of the current density within the fuel cell without disturbing its operation.
The diagnosis method is based on the measurement of the magnetic field surrounding a PEMFC stack. It relies on the measurements of the magnetic field signature generated by the current distribution in the PEMFC. The sensors are oriented in such a way to detect only the magnetic signature of a default [1]. From these measurements, an inverse problem is solved to get the 3D current density inside the fuel cell. The current density distribution leads us to conclude on the change of the electrochemical behavior of the stack due to cell ageing or failure.
Compared with our previous paper [1], the proposed diagnosis tool is extended to capture 3D default induced by the local properties degradation of one cell or few cells in the stack. Like in [1], a current density basis using 2D Fourier series and completed in 3D with a finite volume method has been built [2]. To properly define 3D current, the proposed approach is reproduced several times along the main current direction. Each vector of this basis is an elementary default which radiates a magnetic field around it. The problem being linear but ill-posed, truncated single value decomposition (SVD) is used to get a stable solution.
Current streamlines
…
Current streamlines
Fig. 1: (a): Original 3D current density, (b): Current density basis, (c): Reconstructed 3D current density
In a first stage, a physics-based model of PEMFC (or direct problem) [2] is used to simulate the 3D distribution of the current density and finally the induced magnetic field around the stack when considering defaults. Second, an inverse problem is solved to build the 3D current density. The two current density distribution obtained from the direct or inverse problem are compared.