CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Running, Solving & CFD (
-   -   How to make it converge, ask for tips (

1/153 August 23, 2012 11:38

How to make it converge, ask for tips
I am running a simpleFoam case for flow around a bluff body (2D Steady-RANS), and for one case, it converged perfect well (Judging from velocity field and other flow field and also the force coeffs).

But when I move to another similar but a different geometry case, (it is bluff body external flow), and the shear layer after separation shows instability, looks like a transient simulation, even though it is not totally unsteady, no vortex shedding, but I can still see the velocity wriggling on the edge of boundary layer.

I am using gamma 0.1 scheme for div, gauss linear corrected for laplacian. RF for p is 0.1, U 0.3, others 0.2.

Any ideas?

1/153 dw

owayz August 25, 2012 11:29

May be you should think about your Domain extents and boundary conditions. If you can share your 0/ constant/ and system/ folder we can have a look and I am sure somebody will suggest a good solution. :)


1/153 August 27, 2012 14:41

Thank you Awais.
I am sorry I can't release the mesh right now. But it is much like flow around a square cylinder, with turbulence inflow.

After googling a little bit, my impression is, for certain problems, by nature the flow is just too transient, and it might be difficult or impossible to have an converged solution????

What do you think?

owayz August 27, 2012 20:55

Well yes, in-general for transient cases you might not get good convergence sometimes.
But if you are using an algorithm like PISO, you should try to converge in one time step (i.e. for a time step your residuals must be small enough or at least they should come to the minimum before showing some steady behaviors).
I mostly use Pimple algorithm and to get good convergence in a transient case I closely monitor residuals of pressure and momentum. And then I increase the nOuter iterations till I achieve a steady state for residuals or they stop decreasing further. Also pressure corrections in one outer iteration is also important for that I use pressure corrector to decrease pressure residual to certain minimum level.
Time step is also another very important parameter in highly transient flows. It could be possible that time step is very high to capture the motion of Large scales, this thins is specially important in LES. But mostly I have observed people using a Courant Number of less that 0.5 for highly transient flows and then they adjust all other parameters (like outer iterations, pressure corrections and nonorthogonal correctors) accordingly.
Hope this will help you,

owayz August 27, 2012 20:59

But now I see that you are using simpleFoam. So the transient nature of the flow should not be problem for the convergence of your setup. What you will get after simpleFoam simulations is a time averaged solution of the transient problem.
I think you should try playing with Boundary conditions and relaxation factors.
If you can tell what boundary conditions and solver settings you are using it would be easier to help you. :)

1/153 August 27, 2012 23:09

Thanks, I am now playing with the convection scheme, hoping the upwind will calm down the oscillating velocity field.

Since I am working with low-re mode of turb model, so the problem is quite stiff, b.c. of omega is of the order of sqr(y_1st). I will keep you posted. Thank you, Awais

immortality February 23, 2013 17:17

dear Barre
I have a high transient problem that pressure seems don't it can be find which values for outer and pressure(inner)correctors and also maxCo in rhoPimpleFoam is more suitable for converging?

All times are GMT -4. The time now is 17:36.