Discussion Forum

Hi,

since then I have good experiences calculating free convection. Because of expansive simulation time I'm looking for optimization of solver settings.

First I run a stationary solution for p,u,v,c (dissolved salt concentration) andT without buoyancy force. Then I run a transient solution and I switch on buoyancy force over time (using a step function).
For the transient simulation all variables are manually scaled for their magnitude. I use a segregated solver with 2 groups p,u,v and c,T (because of the couplings c,T). Buoyancy force has a strong influence of the flow, it is very useable to seperate p,u,v and c,T and get buoyancy force by iteration of the segregated groups and via time stepping of the transient solver. It works very well but it needs much calculationg time. Time steps are depended on transient force and 25sec gets a stable time stepping.

I use the following solver settings:
* absolute tolerance of manually scaled variables is 0.001
* relative error of transient solver is 0.1
* tolerance factor of segrgated solver is 1
* tolerance factor in each group is set to 0.01

What I want to achieve is an accurate and stable result in each group with a lower termination criterion for the iterations of the transient segragated solver. Because transient simulation is only a trick for switching on buoyancy force there is an accepted relative error of the transient solution.

* outer iteration: time stepping wiht an accepted relative error (10%)
* iteration of segregated solver with accepted tolerance
* inner iteration of two groups (p,u,v);(c,T) with an accurate result in each group

What are your experiences? Are there any recommendations for improving calculation speed?

Thanks,
MIchael R.