[R.K]

Hi all,

I am simulating flow over an inlet at hypersonic speeds (Mach 7) in 2D using the JST scheme. I have done a mesh convergence study with inviscid simulations and the residuals go to machine accuracy for all the meshes.

I am using the Navier-Stokes solver for the viscous simulations. I have 5 meshes (very_coarse, coarse, medium, fine, very_fine) to do a mesh convergence study. The very_coarse (VC) and coarse (C) meshes converge to a residual rho of -8.5 and -9 with Cauchy on drag = -10 orders of magnitude. The finer meshes barely converge to -4 to -5, before increasing -2 where they plateau. Something I noticed on the very_coarse and coarse meshes, is that there is an abnormal region at the leading edge where the Mach number is higher (11 for VC and 9 for C). I increased the area around the leading edge, but this region still persists.

I played around with the 2nd and 4th coefficients in the scheme; the 4th coefficient has not effect on this region. Increasing 2nd coefficient smooths shocks, so it decreases the Mach number in those regions but its still higher than freestream. Increasing the 2nd coefficient also makes the shock structure in the rest of the flow field very smooth (looks like a smudge on the screen).

Is there a reason why this happens? I did a few first order upwind simulations; they converge almost instantly and the higher Mach region does not exist. Nor does it exist in 2nd order upwind scheme, but these dont converge any further than -4. I want to conduct optimisation, which the JST scheme shows good stability and convergence.

I have attached the config file, history convergence and images of the domain around the leading. The history file also contains information of the meshes (number of cells and y+ values) and images of the high Mach regions. The file is too big to upload, I have uploaded it to google drive: https://drive.google.com/drive/folde...by?usp=sharing

Thank you,
RK

[bigfootedrockmidget]

Do you have a converging and nonconverging mesh for us?
This could be seen as a bug, so you can also report it on github issues. But it might also be something that can be solve by using some different settings or mesh, it's hard to tell at this moment.

[R.K]

I have uploaded two meshes on the drive - one which converges and another which diverges. I will report it on github later today.

Thanks,
RK