Topology Optimization of Heat Transfer Problems

[schewef]

Dear @friedenhe

as you know I am interested in topology optimization unsing DAFaom.

Especially I am interested in heat transfer problems. Therefore I would like to ask the following questions:

1. As part of the repository the heat transfer solver "SimpleTFoam" is available. According to your publication you already investigated more solvers like "buoyantSimpleFoam". Is it planned to include that solver to the repository?

2. For topology optimization we need a source term for the porosity in e.g. simpleTFoam which is currently not implemented (as far as I know). The implementation in DAFoam itself should be straight forward. What I am wondering about is, if the porosity term has implications on the AD version of OpenFOAM? Would one have to change/extend the AD-OpenFOAM at all or is the calculation of sensitivities according to the porosity independent from the solver implementation in DAFoam?

[friedenhe]

I just added the alphaPorosity term for DASimpleTFoam. Check this 9b0f251. It should work as is; no need to change the AD version of OpenFOAM or the calculation of alpha sensitivity in DAFoam.

[chairmanmao256]

@schewef How did you check the sensitivity field? I wonder if there's a way to visualize the sensitivity field in Paraview or Tecplot.

[friedenhe]

Check this page and the tutorial.

[chairmanmao256]

@friedenhe Thank you! However, I'm working on Topology optimization these days, the dimension of design variable (alphaporosity) is as large as the number of grid cells. So I wonder if there's a way to visualize this sensitivity 'field' in Paraview or something else.

[friedenhe]

https://dafoam.github.io/mydoc_get_started_faq.html#how-to-get-sensitivity-maps

[chairmanmao256]

@friedenhe Thank you! I should read DAFoam's website more carefully!

[sodabryant]

[friedenhe]

1. HFL0 is the heat flux for the baseline design. You can get this information from the first flow solution
2. Right, it is not consistent... We have not implemented the Nusselt number in DAFoam version 2 yet... The thing we have not added to version 2 is a variable dT that normalizes the heat flux.
3. The reason to include only a portion of the channel is to be consistent with the experiment. Check the reference in the IJHMT paper.

[sodabryant]

I got it!!
Thanks and regards.

[sodabryant]

Dear @friedenhe
Is it true that the topology optimization case of pitzDaily only works with the SNOPT optimizer? To solve the TO problems, any alternative optimizer do we have ?

[friedenhe]

IPOPT may also work, but we have not managed to make it yet...