[AlexF]

Hi SU2 community,

I am trying to simulate laminar flow over a 2D supersonic wedge.

After forming the shock, the solution seems to converge extremely slowly, and I have been unable to adequately resolve the boundary layer. The inviscid shock region values agree with analytical results; however, the boundary layer and surface values do not (e.g. skin friction values from the simulation are much larger than those estimated using reference temperature methods). I have tried strategies such as adjusting the CFL, adjusting the relaxation factor, and using multigrid to increase convergence speed. This has sped up the time it takes to form the shock, but the boundary layer values still change extremely slowly.

I have had better success converging the solution using first order methods; however, the skin friction and pressure values at the wall (quantities of interest to me) are not accurate enough. Plotting the residuals show that they decrease but begin oscillating after the shock is formed (see attached image).

Why might this be happening, and does this signify a problem in the setup?

I attached my current configuration file and an image of the current grid I am using. The first cell height is on the order of 1e-5. Boundary conditions are as follows: supersonic inlet (freestream conditions correspond to 100,000 ft), supersonic outlet, symmetry, no-slip adiabatic wall, and Euler for the upper boundary. Since previous solutions showed the wall at a higher temperature than expected, I have also tried using an isothermal wall boundary condition with the temperature set as an estimate for an adiabatic wall (via recovery factor relation), but I still received similar non-converged results.

Do I have any obvious errors in my setup or configuration file that are preventing a better and more efficient converged solution?

Any advice or help is greatly appreciated,
Alex

[aero_amit]

Originally Posted by AlexF

Hi SU2 community,

I am trying to simulate laminar flow over a 2D supersonic wedge.

After forming the shock, the solution seems to converge extremely slowly, and I have been unable to adequately resolve the boundary layer. The inviscid shock region values agree with analytical results; however, the boundary layer and surface values do not (e.g. skin friction values from the simulation are much larger than those estimated using reference temperature methods). I have tried strategies such as adjusting the CFL, adjusting the relaxation factor, and using multigrid to increase convergence speed. This has sped up the time it takes to form the shock, but the boundary layer values still change extremely slowly.

I have had better success converging the solution using first order methods; however, the skin friction and pressure values at the wall (quantities of interest to me) are not accurate enough. Plotting the residuals show that they decrease but begin oscillating after the shock is formed (see attached image).

Why might this be happening, and does this signify a problem in the setup?

I attached my current configuration file and an image of the current grid I am using. The first cell height is on the order of 1e-5. Boundary conditions are as follows: supersonic inlet (freestream conditions correspond to 100,000 ft), supersonic outlet, symmetry, no-slip adiabatic wall, and Euler for the upper boundary. Since previous solutions showed the wall at a higher temperature than expected, I have also tried using an isothermal wall boundary condition with the temperature set as an estimate for an adiabatic wall (via recovery factor relation), but I still received similar non-converged results.

Do I have any obvious errors in my setup or configuration file that are preventing a better and more efficient converged solution?

Any advice or help is greatly appreciated,
Alex

Can you share cfg and mesh file