Restart option's effect on results
Hello,
I am running a simulation with Acusolve and I try to investigate the effect of the Restart option at the final results. So I run the simulations in 3 different ways:
1) Using the job launcher of Hypermesh and letting Acusolve solve until it reaches convergence at the 31th time step.
2) I make an initial run through job launcher of Hypermesh, I stop the solver at the 20th time step and then I use AcuConsole's Restart tab to continue solving, initializing from the first run. The convergence is achieved at the 30th time step.
3) I utilize AcuConsole's Project tool in order to project the results of the initial (stopped) solution of the 20th step and then continue solving until the convergence is achieved at the 30th step.
The resultant force as shown at AcuProbe is the same for cases 2 and 3 (175N) while for the first case it stops at (207) N. This difference seems strange to me since the models for all the simulations have the same mesh and solver settings. I would expect that the calculated forces by AcuProbe would be the same for all of the 3 cases. Is something that I am missing? Any ideas why does it happen?
Thank you in advance.
Ntinos
Answers
-
There could be some 'path dependence' on the results, but I would also not expect such a large difference in the result, about 15%. Do you also get similar results differences if you reduce the convergence tolerance from 0.001 (assuming you're using the default) to 0.0001? Are you able/willing to share the model files (.inp, MESH.DIR) here? If not, you should contact the local Altair Greece support team.
0 -
I tried to change the convergence criteria but the results didn't change.
However I noticed that the same model gives different results when setted up through Hyeprmesh and run through Acusolve Job launcher and when it is setted up at Hypermesh, exported and run through AcuConsole. I scanned the .inp files to examine if there is any different setting but I didn't find anything.
So the problem is not related to the RESTART option but probably to Launcher/AcuConsole choice of the model's setup. It seems really strange!
You can find attached the .inp file and the MESH.DIR folder for the ACUCONSOLE method and the HYPERMESH.DIR folder and the correspondent .inp file for the LAUNCHER method. The monitored results is the Traction force in y direction.
0 -
It looks like there are two surfaces in the AcuConsole version (acusolve.inp) with both Simple BC (type = wall) and Surface Output commands, where in the HyperMesh version (launcher.inp), those two surfaces only have Surface Output. It's easy to miss the fact that by default, the model after importing the mesh/.arm to AcuConsole will likely have Simple BC active.
inter_CFD_bl001_CFD_tetcore001
and
inter_CFD_tetcore001_CFD_bl001
0 -
Ok I got it. These are parts that are created automatically during export from Hypermesh. Anyway I believe these cause the problem .
Thank you very much!
0
Altair Employee