Learning Forum

[pgl]

Any suggestions on speeding up convergence behavior with surface-to-surface radiation?

[cbhavsar]

Here are some checks that can be useful to speedup convergence 1. If the initial temp of parts/regions close to the final temperature, it can decrease total # of iterations. (For Mechanical, the initial temperature can only be defined for the entire system, not per part. Either link another analysis where you solve a trivial condition of temp=X for part Y to be used as initial temperatures, or use a "Commands (APDL)" object with IC commands. 2. Tighten the radiosity solver tolerance. This may sound unintuitive, but the tighter the tolerance, the more accurate the heat fluxes that are passed back to MAPDL, meaning that you have fewer total iterations. 3. Try changing over-relaxation factor to 0.5. The radiative flux over-relaxation value uses a portion of the previous solution. The default is 0.1, which makes convergence slow since 90% of the previous solution is used. Since you have a radiation-dominant problem (i.e., only heat flow path between two bodies is via radiation), you probably don't want to set it to 1.0 (use only currently-calculated values and not previous solution), but using a higher value than the default of 0.1 can help reduce total number of iterations. It is advised to try each of these things, one at a time, to see which option has the biggest (positive) effect for particular case rather than changing all 3 at once.

[cbhavsar]

testing1

[cbhavsar]

[cbhavsar]

[cbhavsar]

testing11

[cbhavsar]

testing2