Make all problems work on Euler, running with slurm

[g-braeunlich]

Current status:

• ✔️ airfoil
• ✔️ heatconduction2d
• ✔️ heatconduction3d
• ✔️ photonics
• ✔️ power_electronics (only simulate available)
• ✔️ thermoelastic

power_electronics

ngspice available on Euler:

module load stack/2024-06 ngspice/44

[ffelten]

@g-braeunlich is this really closed?

[g-braeunlich]

No, github closed it because I had "partially solvers" in the PR

[ffelten]

@g-braeunlich for power electronics, we have an automated install procedure in https://github.com/IDEALLab/EngiBench/blob/main/.github/workflows/test.yml#L26

and this is documented:
https://engibench.ethz.ch/problems/power_electronics/#simulator

Now it probably does not work on the cluster 🙃

[g-braeunlich]

ngpsice available on Euler now (see top message)

[g-braeunlich]

With a couple of improvements to error handling in airfoil, I could identify another problem:

File "/home/mdolabuser/mount/engibench/engibench_studies/problems/airfoil/study_42/airfoil_opt.py", line 234, in main
    DVCon.addVolumeConstraint(leList, teList, 2, 100, lower=area_ratio_min*area_initial/area_input_design, upper=1.2*area_initial/area_input_design, scaled=True)
TypeError: unsupported operand type(s) for *: 'float' and 'NoneType'

Indeed, as in v0.py, we have:

        base_config = {
           ...
            "area_initial": None,  # actual initial airfoil area
          ...
        }

just before templates are filled, area_initial=None in airfoil_opt.py in the container.

I will shortly push what I have so far (improved error handling).
@cashend @ffelten what should we do about the above issue?
Another question:
May I replace the py templating approach by something similar to what we use in heatconduction?
I have almost no support from by IDE for the templated py files which is a pain.

[ffelten]

May I replace the py templating approach by something similar to what we use in heatconduction?
I have almost no support from by IDE for the templated py files which is a pain.

What do you have in mind? Airfoil has a lot more parameters than heat conduction. I would avoid passing everything via command

[g-braeunlich]

I would have in mind passing a json serialized dict via command line as the only parameter