Channel_TopoCHT tutorial solution issue

[waterlooMiner]

Hello, In the tutorial:
https://github.com/DAFoam/tutorials/tree/main/Channel_TopoCHT
As written it converges to a solution that is very different that what is shown on the tutorial web page. I then ran it for 1000 iterations, and it does not move towards what is shown on the page. The result I get is non-physical for an optimum for the problem, and what is shown on the web page makes intuitive sense.

The page states: “opt_SNOPT_summary.txt” and " The SNOPT optimizer works better than IPOPT for this case", so it appears the solution shown was found using SNOPT. However, if IPOPT finds a non optimal, incorrect solution, "works better" should be phrased "does not work" I would think.

Can anyone explain what is happening, with two very different solutions, and if non-proprietary solvers work, or can be coaxed to work?

[friedenhe]

@ChrisPsenica Can you take a look at this?

[friedenhe]

I believe you need to run IPOPT for a very long time; the objective will first increase and then decrease.

[waterlooMiner]

I appreciate it, but could you run the simulation and verify what I get? The tutorial ran for 27 steps, and I ran 1000+, it is clearly stable with a very low objective after ~50 iterations. The result is different than the tutorial web site, and non physical. I've ran dozens of runs with different materials and settings, with results stabilizing at odd, non physical results.

[Ananth-Narayan-IITM]

H, you must be running through IPOPT optimizer which works, but in tutorial case its SNOPT optimizer which requires trial version.

[waterlooMiner]

Bump. following up, could anyone confirm that this tutorial works? This concept has potential, but I am not seeing that this is working, given the results I see and limited feedback.

[Ananth-Narayan-IITM]

Hi, I have just started working (1 week back) on this topic, I feel this have some potential. I think we can have some good discussion.