|
[Sponsors] |
Convergence Problems for Transition SST Turbulence Model |
![]() |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
![]() |
![]() |
#1 |
New Member
Alex
Join Date: Jun 2015
Posts: 22
Rep Power: 0 ![]() |
Hello,
I run through a 2D simulation of a flow through cascade and wanted to compare the results between calculations based on Spalart-Almaras and Transition SST turbulence models. With Spalart-Almaras I achieve convergence for a wide range of different inflow angles, while it is very hard to get the Transition SST based calculation to converge. I have great trouble with the residual behavior for Transition SST. Could somebody explain this to me? Greetings |
|
![]() |
![]() |
![]() |
![]() |
#2 |
New Member
Alex
Join Date: Jun 2015
Posts: 22
Rep Power: 0 ![]() |
Could someone help me please?
This is very important. Greetings |
|
![]() |
![]() |
![]() |
![]() |
#3 |
New Member
Join Date: Apr 2015
Posts: 28
Rep Power: 11 ![]() |
for SST, change the spatial discretization schemes to 1st/2nd order for everything, it should help.
S-a model is converging quite well because it is a very simple model (1equation). SST is a much more complex model (4equations) therefore you must help him by taking simple spatial scheme |
|
![]() |
![]() |
![]() |
![]() |
#4 |
New Member
Alex
Join Date: Jun 2015
Posts: 22
Rep Power: 0 ![]() |
Thank you very much for your help my friend!
I achieved convergence for Re = 5e05 and for one particular inflow angle (40 degrees) using the following settings: Pressure-Velocity Coupling: SIMPLEC with Skewness Correction set to 1 Spatial Discretization: -Pressure: Second Order -Momentum: Second Order -k, epsilon, intermittency and momentum thickness Re: First Order So you mean that I need to set every scheme to either First OR Second Order, right? Also the simulation is steady-state. I´m not able to obtain convergence for different inlet-velocities or -angles with these settings, so do you have an idea? Should I use another Pressure-Velocity Coupling or run the simulation time-dependent? Regards |
|
![]() |
![]() |
![]() |
![]() |
#5 |
New Member
Join Date: Apr 2015
Posts: 28
Rep Power: 11 ![]() |
"Spatial Discretization:
-Pressure: Second Order -Momentum: Second Order -k, epsilon, intermittency and momentum thickness Re: First Order " It is a non-sense to use different order for spatial discretization because the accuracy is defined by the worst one : -k, epsilon, intermittency and momentum thickness Re: First Order. You should try with first order everywhere or second order everywhere. I'm not a fan of the first order but if your problem is not converging with 2nd then you have no choice... Pressure-velocity coupling : SIMPLEC would not be very useful, I think SIMPLE is better in your case. So basically I would recommend to put first order everywhere even though the results should not be very very accurate. For the steady-state simulation => transient or steady has nothing to do with convergence. If it is not converging in steady, it will most likely be very bad in transient. There is a technic that you could use though to improve the accuracy of your solution : 2steps 1. run first order spatial schemes with simple pressure velocity coupling for several iterations. When you see that the residuals are decreasing under 0.001 go to step 2 2. switch to 2nd order spatial scheme with simple pressure velocity coupling for the rest of the simulation until convergence. Step 2 is done by taking the last .dat of step1 and use it as initial data for step 2 with the new fluent .cas |
|
![]() |
![]() |
![]() |
![]() |
#6 |
New Member
Alex
Join Date: Jun 2015
Posts: 22
Rep Power: 0 ![]() |
Thanks, I will try to do that.
Why should I use SIMPLE instead of SIMPLEC. I am quite new to CFD and thought they were pretty similar. I chose SIMPLEC, because of the Skewness Correction option. Regards |
|
![]() |
![]() |
![]() |
![]() |
#7 |
New Member
Join Date: Apr 2015
Posts: 28
Rep Power: 11 ![]() |
because when you have convergence problem, adding features is not helping. The skewness correction factor should help solving the gradients between skewed cells but it is most likely not an issue for you at this point if your mesh is good enough (skewness <0.9).
You really have to focus for the first iterations on the most simple schemes and approach to solve the equations and after that change to better schemes with an already "good" initial solution. For step 2, you should also switch from SIMPLE to COUPLED. This robust scheme should help not diverging again. |
|
![]() |
![]() |
![]() |
![]() |
#8 |
New Member
Alex
Join Date: Jun 2015
Posts: 22
Rep Power: 0 ![]() |
Ok, I understand. Thank you Manathan.
So I just run my case until the residuals fall below 1e-03, switch to COUPLED and Second Order Upwind and go on with calculating without initializing, right? Can you explain to me why the residuals jump up again, when I choose the other settings? Regards |
|
![]() |
![]() |
![]() |
![]() |
#9 |
New Member
Join Date: Apr 2015
Posts: 28
Rep Power: 11 ![]() |
It's normal => your solution should evolve because 1st order is not accurate. you jsut have to wait until convergence with step2 that's it. This method is just helping you getting a "first solution" not so far away from the good one provided by the 2nd order.
|
|
![]() |
![]() |
![]() |
![]() |
#10 | |
Senior Member
Reviewer #2
Join Date: Jul 2015
Location: Knoxville, TN
Posts: 141
Rep Power: 11 ![]() |
Quote:
|
||
![]() |
![]() |
![]() |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
bad convergence of k-omega SST model | Many | OpenFOAM Pre-Processing | 3 | August 10, 2015 15:15 |
Force can not converge | colopolo | CFX | 13 | October 4, 2011 22:03 |
sst turbulence model with OpenFoam | HerveAllain | OpenFOAM | 1 | September 28, 2010 10:15 |
question about turbulence model selection and sensitivity | karananand | Main CFD Forum | 1 | February 26, 2010 04:41 |
Turbulence model convergence problem | Andrew | CFX | 7 | August 17, 2008 19:35 |