hi..

i have some doubts about the convergence criteria of some of the solutions produced by fluent..

how can we be sure or at least be satisfied with the results..?? this is because.. if the solution of the problems by fluent doesnt converge for some time..that the residual plot flattens... i will always modify the relaxation factor n run again untill the residual plot flattens again..n i will continue to reduce the relaxation factor right up untill convergence..

is my method correct or not??please correct me.. i am rather confused.. untill what extent may i reduce the relaxation factor..?? what happens if the solution doesnt converge at all even after i reduced the relaxation factor so much??

please help me on this...thank you very much...

Well, I would say, that for a given problem setup (mesh, BC, models) your solution can be considered "converged" if the following are occuring - your residuals have flatlined - avea-weighted and volume-weighted averaged velocities/pressures/temperatures at selected surfaces/volumes are not changing significantly with any further iteration - there is a global mass and energy balance as reported in the flux reports within +/- 0.1%.

I think that's the best you can do.

thank you very much anton... but what about the relaxation factors?? how much can i reduce the relaxation factors so that it gives considerable results?? is there a guideline that i can use?? coz maybe for each different model, there is some different limitations on the reduction of the relaxation factors??

thank you..

Hi!

You should be carefull with underrelaxation factors. It is not good to change them at all. The only reason to do that is to achive stabillity of he solver and not to achieve better convergence. If they are too small the residuals will get smaller but that will no longer be a proof of convergence. I woluld recomend not to use values smaller than 0.4 or 0.3 (in some cases VoF) 0.1. There are also some limitations...I'm not shure but I think that the sum of underrelaxation factor for pressure and momentum should be at least 1.

Hope this helps

MATEUS

Hi,

As Mateus has said, you have to be carefull about convergence when you decrease the underrelax. factors too much.This is because that, they are too small to result a significant change in the solution, therfore they may lead you to a false convergence. Try to folow the suggestions in the manual about how big you have to use. I recommend you also to use monitors of surface or volume integrals, in addition to the residuals.

Ozgur

Under-relaxation is a tecnique to stabilize the solution in order to achieve convergence.

When you apply it, you don't allow to the under-relaxed variable to reach its final value directly, but you divide this process in steps.

Keeping this in mind, when you under-relax, you have to do the following things:

1. Do a number of iteration on the under-relaxed variable which is high enough to allow to the same variable to reach its final value.

2. Use a convergence criterion which is strict enough to avoid problems of false convergence.

If you respect these conditions, you can under-relax without any problems.

Anyway, a well-posed problem, with proper boundary conditions and grid, usually doesn't require a high reduction of under-relaxation factors. Things changes if you're working with stiff problems.

Hi

ap

Hi,

Please remember the meaning of underrelaxation factor in the textbook of numerical methods. So it will guide you both to stabilize and converge the solution. The best strategy to find a best solution is by controlling the underrelaxation factor while you keep the convergence criteria. Once you find it converged, you may use it a the best solution.

Thank you

Sorry if I didn't explain properly. I didn't say you have to change both under-relaxation factor and convergence criterion.

I just wanted to point out that you can underrelax until you allow to the variable to increase (if it has to) of a quantity which is bigger than the limit defined fixed by the convergence criterion. This means your convergence criterion has to be strict enough.

Hi

ap