Difference between ANSYS and Calculix

Hi everyone,
I’m attempting to run a steady-state thermal simulation of a cross-sectional slice of a heat exchanger (pictures below). The heat exchanger has certain temperature and convective heat transfer coefficient on the hot (bottom) surface and inside the cold channel in the middle. There’s a symmetry in the middle of the channel.

I’ve managed to run the simulation but I’ve run into a peculiar problem. I’ve loaded the mesh and re-created the model in ANSYS. If You use constant thermal conductivity in both, the models agree extremely well. However, if I use temperature-dependent thermal conductivity I can’t get the two models to agree.

Calculix with constant thermal conductivity:

ANSYS with constant thermal conductivity:

Calculix with temp-dependent thermal conductivity:

ANSYS with temp-dependent thermal conductivity:

As You can see, Calculix shows way more dramatic change in temperatures at the top when changing to temperature-dependent material properties. I feel like I’ve checked all the settings and I can’t figure out why that might be. Everything looks identical. I’ve read the same mesh into both programs and made sure that both material properties and boundary conditions are identical (h=2.5e6 W/m2, T=700K on the bottom surface and h=2.5e3 W/m2, T=320K in the middle channel). I would appreciate some help if anyone has some idea where the discrepancy might be or would be willing to run the input file in some commercial FEA software and confirm that I’m not just going crazy and the discrepancy is there.

Link to the Calculix input file:

For a better visual assessment, try dividing the stress map into the same number of colors, cgx: steps 14

1 Like

Hi Rafal,
That’s a good point. I’ve updated the graphs to have the same number of contours. Thanks for looking at it.

Dear Tomasz,

I’ve calculated your model in Calculix and in Abaqus. Results are the same. So you may search for a mistake in the Ansys calculation.

So I am being a dummy after all! Thanks a lot for doing this. At least I know where to look for mistakes now.