CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Unsteady problems - under-relaxation factors (http://www.cfd-online.com/Forums/fluent/115331-unsteady-problems-under-relaxation-factors.html)

stefanos March 28, 2013 09:10

Unsteady problems - under-relaxation factors
 
Dear community,

At unsteady problems where noticeable structures of the fluid interacting with other fluid or body zones might appear (alike vortex shedding) it is often necessary to decrease the under-relaxation factors of pressure, momentum, density, etc for avoiding "divergence problems" of the transferred properties. I would like to ask, if the under-relaxation factors of turbulent quantities are also decreased, is there a possibility that this may influence the appearance of these structures/or not?
In particular, i have been expecting to track a jet somewhere within my FSI problem, but it isn't very clear, beside the dense mesh, etc.

flotus1 March 28, 2013 09:41

The URFs should not affect the final solution. Yet lower URFs mean that the solution advances slower and you need more Iterations to get convergence.

Phenomena like vortex shedding cannot be captured accurately with a RANS-type turbulence models. Maybe this is why your solution is not as you expected.

stefanos March 29, 2013 06:31

Thank you for your notification, sir.

Regards

S

openfoammaofnepo July 21, 2013 15:09

Hi All,

Is it reasonable to use an under-relaxation factor for unsteady simulation? I think it is not, but I found that some of the Openfoam tutorials still have the options for under-relaxation factors for the unsteady solvers. Could anyone tell something about this issues ? Thank you so much!

Quote:

Originally Posted by stefanos (Post 416970)
Dear community,

At unsteady problems where noticeable structures of the fluid interacting with other fluid or body zones might appear (alike vortex shedding) it is often necessary to decrease the under-relaxation factors of pressure, momentum, density, etc for avoiding "divergence problems" of the transferred properties. I would like to ask, if the under-relaxation factors of turbulent quantities are also decreased, is there a possibility that this may influence the appearance of these structures/or not?
In particular, i have been expecting to track a jet somewhere within my FSI problem, but it isn't very clear, beside the dense mesh, etc.


flow_CH July 21, 2013 15:23

For openfoam please go openfoam forum.
For fluent yes you can change under relaxation factors in unsteady simulation in solve -------> control --------> solution
Discretization of equations are same for steady and unsteady.

openfoammaofnepo July 21, 2013 15:29

Thank you very much. But I think this is a general CFD problem. Does anybody know something about this?

Quote:

Originally Posted by flow_CH (Post 441086)
For openfoam please go openfoam forum.
For fluent yes you can change under relaxation factors in unsteady simulation in solve -------> control --------> solution
Discretization of equations are same for steady and unsteady.


flotus1 July 22, 2013 04:25

Why dont you think it is reasonable to use under-relaxation?
Without it, implicit solvers like SIMPLE become unstable.

openfoammaofnepo July 22, 2013 04:32

Because, for example, when i use the time step is 1.0d-04, and then I use under-relaxation factor is 0.1, in terms of time marching, does it mean that the effective time step will be reduced? In other words, the simulation is marching with a time step less than 1.0d-04. Is that so?

Quote:

Originally Posted by flotus1 (Post 441156)
Why dont you think it is reasonable to use under-relaxation?
Without it, implicit solvers like SIMPLE become unstable.


flotus1 July 22, 2013 04:38

This holds true only for the iterations within each timestep.
Given a sufficient amount of iterations per timestep (convergence!) the physical time step size remains unaffected of the under-relaxation.

openfoammaofnepo July 22, 2013 04:41

OK, thank you very much. if I just use one iteration for each time step, but the convergence is reached for that iteration. So the under-relaxation is still can be used for unsteady time marching.

Quote:

Originally Posted by flotus1 (Post 441161)
This holds true only for the iterations within each timestep.
Given a sufficient amount of iterations per timestep (convergence!) the physical time step size remains unaffected of the under-relaxation.


flotus1 July 22, 2013 04:49

No.
An iterative solver with only one under-relaxed iteration does not produce correct results.


All times are GMT -4. The time now is 00:57.