CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   steady state simulation timesteps (

amine March 10, 2008 15:17

steady state simulation timesteps

i have set 500 000 elements on my simulation,i've read on the help that the steady state simulation should converge after 50 to 100 timesteps ,now my question is ,if it exceeds 200 timesteps and it on the way should i wait more or i have to change the meshe,i'm using high resolution scheme.

Dre March 10, 2008 15:44

Re: steady state simulation timesteps
What is your convergence criteria?

Kushagra March 10, 2008 17:21

Re: steady state simulation timesteps
It will depend upon many factors.

1) Do you expect (with reasonable assumption) your flow to be a steady state flow? If not, then it would be very difficult to get it converged with 'high resolution' scheme. You might think about switching to transient solver. If your flow is any where near to steady state nature, your transient simulation will show it. Just monitor velocities/pressure at some points in your domain.

2) If simulation is diverging, reduce the false time step in steady state simulation. But it will slow down the simulation. Hnece the benefit of using steady state solver is lost.

3) Check out the location of high residual in CFX-Post. (You will need to enable to 'out put eq residual' in CFX-Pre -> output control-> results, set it ALL)

basically, it would depend upon nature of your flow, mesh quality, and time step...etc....

Read the documents (modeling advice )on convergence as well...


wan March 17, 2008 01:13

confius on operation setting
For laminar i should set the [default domain modified->terbulence] to [none-laminar], but if i want to run under terbulence condition, which operation should i select [k-epsilon/shear stress transport/BSL Reynold stress/SSG reynold stress]?

Kushagra March 17, 2008 11:29

Re: confius on operation setting
depends upon ur problem... zero equation model->k-e->Reynolds stress(SSG-RSM,LLR-RSM)->LES turbulence models are in the order of increasing computational cost....

for some problems k-e will work, for some even RSM might not work....

search out the papers on your problem and get an the CFX-11 manual as well

All times are GMT -4. The time now is 14:58.