CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Convergence problem (

Sam October 7, 2004 22:10

Convergence problem
I'm simulating air flow (at STP) thru a 90 deg bend diffuser. Inlet to Outlet area ratio is about 5. Outlet static pressure=atm, inlet velocity profile: assume fully developed profile follows 7 power law with Umax=20m/s. No inlet swirl for simplity. I'm hoping to get some steady state solution from CFX 5.7. Target RMS residual set to 1E-7.

I get a converged solution within 300 steps if K-E turbulence model is used. No oscillation at all.

When I tried SST model, it manage to converge to target RMS residual but with sinusoidal oscillation after 200 steps. RMS residuals still show a downward trend (based on 4 osicallatory cycle observed in the residual plot).

When SSG model is used, the solution doesn't converged at all. RMS residuals oscillate sinusoidally (peak-to peak amplitude about the same) at a mean value of about 10-4. I run it to about 1500 steps but still no sign of convergence.

I read the manual. SSG model could produce unsteady solution in the steady state simulation. Is that the reason for my case? Any ideas how to overcome this? Secondary flow is observed downstream in the experiment. SSG model could be best for my case. Can I use it for steady flow simulation?

Thanks, Sam

NymphadoraTonks October 8, 2004 06:40

Re: Convergence problem
You cant insist that a solution be steady state unless you know for sure it is e.g. from experimental data.

Give it a trial run with in the transient mode and observe the flow behaviour.

SSG can be used with steady state and transient solutions . Even if a problem has a steady state solution obtaining an initial guess from a partially converged transient solution can assist convergence in the final steady state run. This is especially true when using advanced reynolds stress turbulence models such as SSG.

All times are GMT -4. The time now is 12:58.