[voidcrow]

Hello. I am trying to get running a PEMFC simulation (very simple geometry, essentially several layers of rectangular prisms, with one flow channel on each side) using the PEM cell add-on module. I am trying to follow the tutorial I found (https://www.youtube.com/watch?v=HX9a5uijvF0) using the same operating and boundary conditions:

• 353 K inlet temperature
• 3e-7 kg/s mass-inlet BC at the anode (mass fractions 0.6 H2, 0.4 H2O)
• 2e-7 kg/s mass-inlet BC at the cathode (0.21 O2, 0.15 H2O)
• No slip boundary conditions for walls
• 0 V applied at the anode tab and 0.65 V at the cathode tab

I assigned the parts of the geometry accordingly with the exception of the microporous layer (like in the video, I get a warning message about missing microporous layer). I also made two types of mesh, one more refined and another that is coarser, and both are conformal meshes.

I initialize the solution with the hybrid method and I get the following message: "Warning: It is highly likely that the Anode GDL/CAT are not assigned correctly!" (even though it is assigned in the PEM fuel cell model panel). When I start the solution, it diverges within 3 iterations with the error "Divergence detected in AMG solver: temperature. Error: floating point exception". I disabled the energy equation, and restarted initialization and it also diverges: "Divergence detected in AMG solver: uds-3" (uds-3 is water content). I disabled the water content equation. Only then the solution converges, but I end up with essentially 0 current density magnitude in the anode and cathode tabs.

I did not change any of the source defaults in cell zones. But I noticed that in the video there are more sources (udf :: pemfc) being activated for each zone in the version of Ansys that the individual is using. I attached a screenshot from the video of the sources being defined in the anode flow channel to contrast with the ones that are active by default when I enable the model (on the right). I am using the 2023 R2 version. I tried using a finer mesh and the outcome is the same.



I don't understand how this person was able to get the model to converge with everything enabled. I tried following everything exactly, including replicating the geometry. But the geometry won't be identical since I don't have the dimensions of the cell he used, and also Fluent creates several walls and shadow-walls when I import the .msh file. Any help is appreciated.