CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Mesh Generation & Pre-Processing (
-   -   PEM fuel cell simulation (

pchoopanya March 11, 2013 14:22

PEM fuel cell simulation
5 Attachment(s)
Hello guys,

I am trying to mesh a PEM fuel cell and want to simulate its operation in FLUENT and its add-on module PEMFC and Electrolysis.

Using ANSYS Meshing (Workbench) I always get an unstructured, tetrahedral, and highly skewed elements when the domain is built (ANSYS DesignModeler) as a multi-body part. However, the tutorial given by ANSYS features a very simple single-channel straight flow field assembly, and hence when it can be built as a multi-body part and meshed quite easily. But my domain is bigger and more complex.

I resolve this by building 3 separate parts, multi-body, multi-part approach;

1. anode channel (single body, single part)
2. MEA (GDLs+CLs+membrane as multi-body part), and
3. cathode channel (single body, single part)

All elements can then be meshed hexahedrally. I obtain good-looking orthogonal mesh elements like the ones shown in many published papers. In fact, I have never seen any research group mesh the cell with tetrahedral mesh. They always show a nicely meshed cell without details of how they did it.

This is what I want, all zones are hexahedrally meshed and neat.
Attachment 19781

The mesh on anode-side GDL surface, the anode flow channel is hidden.
Attachment 19782

Please note that the mesh on GDL side does not match the mesh on the bottome of the flow channel (non-conformal mesh) as is shown; the two different meshes (same interface) is overlayed on top of one another
Attachment 19783

The problem with this 3-part assembly approach arises when I set up the simulation case in FLUENT. The interface between the gas-diffusion-layers (GDLs) and the bottom surface of the flow channels (both anode and cathode) are meshed non-conformally as they are different parts. These interfaces (or surfaces) need to be defined as INTERFACE-type boundary condition.

With INTERFACE BC, I cannot find a way to specify a fixed potential on this surface (0V at anode surface, 1.16V at cathode surface).

I then have to go back to a multi-body part approach where the loops begin - I get ugly-looking mesh and sometimes the Mesher fails to mesh at all.

I even try imprinting a face of projected flow channel onto the GDL surface so that I can divide the GDL surface into 2 parts; the one that is not covered by the flow channel (real external surface of the entire domain) and the one that is hidden under the flow channel (interface of two separate parts) hoping that I will only have to do the MESH INTERFACE of bottom surface of flow channel and the GDL surface directly underneath it, leaving another one as external wall boundary and define a potential and then, problem should be solved!!

However, with an imprinted face, even the domain is not single part (3 parts), this even complicates the Mesher. I still get an unstructured grid.

An imprinted face onto a GDL surface, flow channel is hidden
Attachment 19784

This is what I get when an imprinted face is carved onto the GDL surface. From top-to-bottom; anode flow channel, anode GDL, anode CL, membrane, and cathode CL. All layers are perfectly meshed, but the anode GDL where it has an imprinted face on it (even though the GDL and flow channel are 2 distinct parts) is poorly meshed.

Attachment 19786

An imprinted face seems to have an influence on the mesh algorithm even though the GDL and flow channel are 2 separate parts, is this true? (channel hidden)

Anyone has had some experience regarding this? What is your trick to deal with the domain so I can simultaneously obtain a perfect hexahedral element all over the entire domain and can still apply a cell potential value on the GDL surface?

My other questions are;

1. When in FLUENT, MESH INTERFACE... do the two surfaces have to overlap each other entirely? Can they be partially overlapped? If so, the method to deal with MESH INTERFACE different?

2. Providing my domain excludes the current collector plate, when specifying a fixed potential at the GDL surface, is it the portion that is exposed to the external? What about the portion that is underneath the flow channel? Or both?

3. Shall I move to ICEM CFD instead then? Do you still recommend a single multi-body part approach with conformal mesh is desirable? I find it is more easy though as I do not need to define the INTERFACE boundary condition compared to a non-conformal mesh approach.

4. What is the difference between conformal and non-conformal meshing methds in terms of the results? I know the mesh topology is different, but what about the accuracy of the results?

5. GDLs, CLs and membrane each is a very thin layer (in the range of 10-100 micron). What is the best way to mesh them as the resul always features high aspect ratio elements?

6. When imprinting a face, one parameter to be selected is Merge Topology: Yes or No... what is the difference between these two options?

Please ask if you need some clarification on my questions though. Thank you very much.

Best regards,


beehuah March 1, 2016 00:56

Hello Pattarapong,

May I know how do you solve this problem? Did you use ICEM CFD to generate a hexahedral mesh for your fuel cell module? I am currently facing the same issue with add-on fuel cell module in fluent. My solution are diverge and multiple error occurred. Would like to know how do you solved this unstructured mesh issue? thanks

All times are GMT -4. The time now is 14:29.