Quantcast
Channel: Solving, solving\ - Latest Discussions - COMSOL Forums
Viewing all articles
Browse latest Browse all 3988

Speed up a Computational Problem in COMSOL involving parametric sweep

$
0
0
Consider the general parametric sweep case:
When I set up a non-linear FEM problem, it takes considerably more time than a Linear FEM case. The reason for this extra time taken is the extra effort by COMSOL which involves :
1. COMSOL takes the step as specified by the user in the parametric sweep section and if cannot converge, takes smaller steps to go as far as possible. SO many step sizes are tried by COMSOL.
2. COMSOL tries different solver strategies and tries to converge for each step.

For my problem, probably the last of 1 and 2 efforts are successful which involves a total of 4 hours to solve. I can see in the progress section that the parametric sweep goes very fast from 1V to 26V and then it takes at least 3 hours near 26V. I have specified parametric sweeping limits from 1V to 50V in steps of 1V. It is fantastic that the error comes at 26V stating inverted mesh elements (owing to system instabilities) retaining the previous solutions of 1V to 26V.

Question 1: Is there a way to find out which solver strategy was successful in the first run and in subsequent runs force the same strategy so that the simulation is sped up in subsequent runs?
Question 2: Can the 3 hours be saved in which nothing useful was done except returning the error, provided that I dont know that the system instability occurs near 26V at the beginning of running the simulation?

Viewing all articles
Browse latest Browse all 3988

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>