Learning Forum

[thore.pruss]

Dear all, 

I’m currently developing a CFD-Model for a new kind of Fuel Cell using ANSYS Fluent, which is using both a PEM and an AEM Membrane. Because of this, I can’t use the PEM-Fuel Cell Model which is provided in ANSYS Fluent and instead develop a model using User Defined Functions. The Transport of the electrons is accounted for using a UDS-Scalar Transport Equation (like in the PEM-Model provided by ANSYS) which, after solving it, provides me with a potential distribution. This is the potential distribution i get at the moment: 

This is comparable to potential distributions I found in the literature. (Note that I solve the equation in the Flow-channel as well, which is not physically meaningful, but I didn’t find a way around it). 
The problem I encounter is the following:  
When I calculate the current density in the Y-Direction from the potential distribution (conductivity * gradient of potential in Y- Direction). I get some weird results, as the gradient in the cells right next to the channel is much higher than for the rest. 
You can see the potential distribution in the second photo (phi_e (the first row from the right is the first row under the channel)). As well as the gradient I get (j_e).

At least to me, the difference in potential per cell in the Y-Direction for the third and 4th cell row doesn’t look high enough to explain the significant increase in the gradient, for the 4th row from the left in comparison to the third row.  
In my opinion, a possible explanations for this could be that the Calculation of the gradient of the electrical potential is wrong (I’m using the C_UDSI_G(c,t,0)[1] command to obtain the gradient in the Y-Direction). I think it is odd, that the Gradient in the Y-Direction is so much higher near the Flow-Channel, which leads me to believe that at least some information from the cells in x and z direction goes into the calculation of the gradient? Maybe for the calculation of C_UDSI_G? For spatial discretization, I use the Least Square Cell based method.  In the last photo you can see the code i currently use to calculate my gradient. 

If anyone has encountered this problem before or has some insight that might be useful and can help me out, I would much appreciate it! 

Best regards 

Thore  

[Rob]

Can you retry with a mesh with a much lower aspect ratio? Before looking at the method in too much detail eliminate the mesh issues. 

[thore.pruss]

Thank you for your reply! As you suggested I've retried with a mesh with much lower aspect ratio (According to ANSYS-Meshing: 3,3063 - 7,15 ). Unfortunately the problem remains the same (at least in my opinion). I think the gradient near the edge, where Gas-Channel and GDL meet, has to be slightly higher than in the next cells, but not as high as it currently is. But maybe I am mistaken? I think the gradient in the first cell underneath the Gas-Channel is also odd. My guess would be that the edge is not well-defined for C_UDSI? 

To further explain my problem I've added two photos, where you can clearly see that along the edge Gas-Channel <-> GDL current collector boundary, there is a much higher gradient, both in the cells underneath the channel and underneath the current collector right next to the edge, where the GDL meets the Gas-Channel. 

 

If you can give me any advice why this is the case, I would much appreciate it! Other tips or suggestions are also very welcome!  

Best regards 

Thore 

 

[Rob]

Check scalar diffusivity and zone assignment - ie should the scalar be limited to certain zones. 

[thore.pruss]

Thanks again for your time, and effort and sorry for the late reply. 

Regarding the Zone assignment:

I  try to solve the equation: 

in the entire Computational domain:  

 

The conductivity (diffusivity) is prescribed as 640 S/m (constant) in the relevant zones, which is also the value Ansys shows me as the Diff. Coeff. of phi_e. in these zones.   

[Rob]

Should it be in all zones, or just some zones? Do you have a mix of fluid & solid?  We're a bit stuck in that Ansys staff can't get too technical so I can only give you some pointers. 

[thore.pruss]

It should be physically meaningful only in some zones (GDL, MPL,CL), but for the moment I solve the equation in all the zones. I've modeled the Membranes as a solid and in them the conductivity (diffusivity) of the electrons is set to a very tiny number (1e-16). I tried once with fixed values for phi_e in the membranes, but this lead to very weird results. 

[Rob]

That could be a potential (sorry, I like bad puns) cause of the odd results. Jumps in diffusivity etc may upset the solver maths if the mesh is relatively coarse. 

[thore.pruss]

Haha :D 

Okey, so you would recommend to solve the equation only in certain zones instead? 

Thanks again for your help! 

[Rob]

I would look carefully at what is happening in reality - I know very little about fuel cells!