How to restart a URANS simulation using transition model

[MaganS]

I am doing transition simulations on an airfoil.

I am using the LM transition model and the SA Edwards turbulence model.

When I restart my simulation from a specific time step, the results do not match those obtained without a restart, as shown in the image.

SU2 version: SU2-8.3.0; Transition model: LM-Medida-Beader; Turbulence model: SA Edwards; Flow: Mach 0.735, AoA 4, Re 3x10^6, Airfoil: OALT25

I commented the following when I restarted, but that does not solve the problem.
FREESTREAM_TURBULENCEINTENSITY = 0.001

I am using only the above parameter in my config file. My config file
su2_unstdy.txt

The following parameters are set by default for my simulation because I have not specified them in my config file.
% Value for freestream intermittency
FREESTREAM_INTERMITTENCY= 1.0

Free-stream ratio between turbulent and laminar viscosity
FREESTREAM_TURB2LAMVISCRATIO= 10.0

Please suggest. Any help can make a big difference.

Thank you

Best,
Magan

[bigfooted]

Does this also happen with a first order restart (start with first order timestepping, then restart with first order timestepping)?
We probably recompute some values at restart, which causes the difference. Maybe you can narrow it down by creating some simpler testcases, using only the EULER or NAVIER-STOKES solver, or RANS without transition.

[MaganS]

I did not try first-order time stepping. And also, I did not do any EULER or N-S solver simulation. But I checked the same using URANS with SA Edwards without transition. I found a perfect match, no issue.

[MaganS]

You can see how the restart of the transition simulation is affecting the result.
Image 2:
The image shows decaying oscillations of cl on OALT25 when the simulation is restarted, but the oscillations are sustained in image 3 when the simulation is not restarted.
SU2 version: SU2-8.2.0; Transition model: LM; Turbulence model: SA Edwards; Flow: Mach 0.730, AoA 4, Re 3x10^6, Airfoil: OALT25

Image 3:
SU2 version: SU2-8.2.0; Transition model: LM; Turbulence model: SA Edwards; Flow: Mach 0.730, AoA 4, Re 3x10^6, Airfoil: OALT25

[bigfooted]

OK, thanks, it is strange the behavior is completely different with a restart... maybe it's more than just a variable that is incorrectly re-initialized...

First thing to check I guess is if we correctly save and restore the LM. You can write the residuals to screen of RMS_INTERMITTENCY and RMS_RE_THETA_T, in addition to the other residuals (it should be part of the SOLUTION variables, so if you write all SOLUTION variables, that should be fine).
Then you can do the standard test:
Run 10 iterations -> RESIDUALS at i = 10 iterations
Run 9 iterations -> save -> restart -> Run 1 iteration -> restarted RESIDUALS at i=10 iterations

[rois1995]

I haven't tested the Transition model with URANS. If you keep restarting from the same iteration, are the results the same each time? Or are there differences?

Setting the value of FREESTREAM_INTERMITTENCY won't change anything since the freestream value for gamma should always be 1.0 . The SA model is also unaffected by changes in the FREESTREAM_TURB2LAMVISCRATIO (the way to prescribe different values of nu_tilde at the farfield is with the FREESTREAM_NU_FACTOR option) . On the other hand, changing the FREESTREAM_TURBULENCEINTENSITY will lead to different results (the default value is 0.05, thus 5% TI).

[MaganS]

Hi

@bigfooted
I ran the transition simulation as you mentioned. The following is the output. The 2nd line residue output is from the 10th step of the simulation. 3rd, 4th, and 5th lines are from the restart simulation for the 1st, 2nd, and 3rd times, respectively.

"relrms[Rho]", "relrms[RhoU]", "relrms[RhoV]", "relrms[RhoE]", "relrms[nu]", "relrms[LM_1]", "relrms[LM_2]"
-4.158509228, -4.477456879, -4.462767335, -4.225002142, -4.314942179, -3.843988087, -3.92503119
-4.158819104, -4.473892097, -4.4628482, -4.225445877, -4.00584174, -0.2369005737, -2.882601448
-4.158819104, -4.473892097, -4.4628482, -4.225445877, -4.00584174, -0.2369005737, -2.882601448
-4.158819104, -4.473892097, -4.4628482, -4.225445877, -4.00584174, -0.2369005737, -2.882601448

SU2 version: SU2-8.3.0; Transition model: LM-Madida Beader; Turbulence model: SA Edwards; Flow: Mach 0.730, AoA 4, Re 3x10^6, Airfoil: OALT25

Upon restart, RELRMS_INTERMITTENCY and RELRMS_RE_THETA_T change significantly. RELRMS_NU also changes.

How to solve this?

@rois1995
Restarting the simulation multiple times does not alter the solution every time, as shown in the table above.

[rois1995]

It may be related to the compact restart file. Can you please try again with the option WRT_RESTART_COMPACT= NO ? Either this, or there is a problem when initializing the TransSolver that might be related to the intermittency_eff variable.

[MaganS]

No change in solution with this option

[rois1995]

For now, please stick with that option. I will try to fix it in the following days.

[MaganS]

Thank you