Availability of Transition Model (Gamma-Re_theta) for Hybrid RANS/LES methods

[gauravity30]

Hello !
I am planning on doing hybrid RANS/LES simulations in SU2 and wanted to confirm whether the LM transition model available for those too ?

Thanks,
Gaurav

[rois1995]

Hi, yes, the LM model is available for RANS/LES computations in the sense that it is usable. However, I am not sure if the implementation has to be changed or not to account for LES regions.

[gauravity30]

Hey !
I think the case becomes non-physical or blows up way before the big oscillations (from 19s to 23s). I am not sure where the oscillations are. I have attached a video showing the time variation of cp on the sphere surface.

I have done the current simulation (whose results are shown in the previous and current post) with EDDES. I don't think this variation of DES requires any transition model since it handles the transition through LES. Since, the flow separates laminarly over the sphere and then the shear layer transitions to turbulent, the LES part should be able to do that. I am not sure, please correct me if I am wrong.

Is the SLAU2 better in such cases ?

cp.mp4

Thanks,
Gaurav

[rois1995]

The transition will be predicted by LES only if the BL is laminar and the LES region can enter the boundary layer, which should not happen in principle due to the shielding function fd. The EDDES has been designed to avoid this phenomenon, thus at wall you should always have the RANS model active, which is fully turbulent by definition. The LES region can enter the boundary layer only if the grid spacing is fine enough. You can try and output the shielding function value by slightly modifying the code as done in the branch feature_SST_IDDES, or you can output the DES_lengthscale and try understanding from there where LES region is active in your volume solution (although I prefer the first approach as it clearly shows the distinction between the two zones).

[gauravity30]

I will and then probably I will reevaluate my mesh and try with SLAU2 instead of the modified JST.
Thanks,
Gaurav

[gauravity30]

Hey @rois1995 !
I did some reading related to mesh design for DES. I made a new mesh and did the simulation. The simulation didn't diverge and predicted the drag accurately but the flow had characteristics of supercritcal regime of the sphere suggesting the flow had an early transition. I verified this by checking the contour of $\nu_t/\nu$ and it was non-zero before the expected transition point ($\theta=82$). I checked the config file and found that I had missed to set the value of FREESTREAM_NU_FACTOR although the FREESTREAM_TURBULENCEINTENSITY was equal to 0.5% i.e. 0.005. So, I ran again the simulation with setting FREESTREAM_NU_FACTOR equal to 10^{-10}. Is it the correct way for setting initial condition for SA-EDDES ?

I am little confused whether both $\nu_t/\nu$ and T.I. should be set or only $\nu_t/\nu$. Since, the factor is required only for the SA model. Also, how should I choose a value of the factor ?
Thanks,
Gaurav

[rois1995]

Hi @airwarriorg91, I am sorry for the late reply.

The TI value is useful only if the transition model is active. I am not sure about how to set the freestream nu-factor, but I usually leave the default value. If you experience early transition and cannot change the TI at freestream then you can surely decrease the nu-factor and see what happens. I don't know if decreasing the nu-factor automatically delays transition or if other factors come into play and lead to a different outcome. I am definitively interested in your results, thus, please, keep me updated on this!