[ViLaks]

Dear All,

Sorry for restarting the thread again. I just wanted to know about the URF in fluent. How to control the residuals using URF? Suppose, I say, the K and epsilon seem to go up i need to reduce the corresponding URFs. But if I do so, the velocities go up or some other parameter. SO I just need to have a basic understanding of the URFs (in solution controls) in fluent, like by how much amount should I reduce the URFs, can I change more than one urf at a time. I cannot grasp much from the user manual though

[ViLaks]

Dear All,

Sorry for restarting the thread again. I just wanted to know about the URF in fluent. How to control the residuals using URF? Suppose, I say, the K and epsilon seem to go up i need to reduce the corresponding URFs. But if I do so, the velocities go up or some other parameter. SO I just need to have a basic understanding of the URFs (in solution controls) in fluent, like by how much amount should I reduce the URFs, can I change more than one urf at a time. I cannot grasp much from the user manual though

[KevinZ09]

URFs are used to control how much the updated computed variables depend on the previous iteration and the current iteration. In equation for:

u_{k+1} = (w-1)*u_{k} + w*u_{k+1},

where w is the URF. If w < 1 then it's usually called an under-relaxation factor. And you can understand from this equation that the smaller w is, the less your residuals will change (smaller change in computed variables means smaller difference between previous solution and current solution).

As to how to choose them, it's (sometimes) a trial-and-error. I'd generally suggest keeping the default values, but it could happen that your solution diverges, especially at the start of the run. So you may want to reduce them slightly. I sometime reduce the momentum and pressure URFs to 0.25 and 0.6. But that's only for the start, as it will slow convergence down. So I try to increase them again later on.

But two things:
- your final solution should generally not depend on the URFs you use. Sometimes you may think that using lower URFs keeps the solution from diverging, but what it generally does is make the divergence slower, i.e., if with a URF = 1 it converges after 100 iterations, it could take several thousands of iterations for the solution to diverge with a URF = 0.5.
- More importantly, don't judge convergence of residuals. Sure, it's nice when they are low, but they don't tell you much. Better look at monitor points, fluxes, etc.

But in short, you can change them as you like, as many at a time as you want. But don't think it will necessary make your solution better. It will make it slower, that's for sure. So lower them for a bit if your simulation diverges, but raise them again later, if possible.

[ViLaks]

Dear Kevin,

Thanks or your reply

With Regards
Vignesh

[Rana shaharyar]

kindly tell what is the reasonablerange of valuesfor scaled residuals for convergence for a 3d wing problem. please give rangein numerical value

[ViLaks]

Originally Posted by Rana shaharyar

kindly tell what is the reasonablerange of valuesfor scaled residuals for convergence for a 3d wing problem. please give rangein numerical value

Generally the accepted values are of the order 1e-03 except the energy residual, whose value should be of the order 1e-06.

[walakaka]

Originally Posted by Dan Williams

If Fluent is defining the residuals as the change in solution variables between iterations then this has nothing to do with the conservation balance, and in fact is a poor technique for defining a residual.

Why would this be a poor technique for defining convergence?

Shafik