[Stuntmanbob]

Hi,


i'm trying to simulate an airfoil of an linear cascade.
The mesh contains one airfoil and has periodic boundaries near the pressure and suction side. It is a q3d-case with only one layer, with a slip BC on the one sidewall and a symmetry BC on the other. The thermophysical properties are Perfect gas with constant thermodynamics and Sutherland for transport.



Check mesh only complains about about high aspect ratio cells. But these are necessary for y+ values less then 1.


I'm using the k-w turbulence model and the Smooth Solver for p, U, h, k and omega. The divergence scheme is Lust.


As for the boundary conditions i choosed k = Low Re Wall Function and omega = Standard Wall Function (nu_t = Low Re Wall Function and alpha_T Standard) on the walls.



The simulation runs fine and the residuals went beyond 1e-5 for Ux, Uy, Uz, p and k (h get stuck by ~1e-4). But here is the problem:
The solver won't solve omega from the first timestep.

velocityDampingConstraint velocityDamping damped 0 (0%) of cells
smoothSolver:  Solving for Ux, Initial residual = 3.597191e-007, Final residual = 6.192554e-008, No Iterations 2
smoothSolver:  Solving for Uy, Initial residual = 2.276461e-007, Final residual = 4.008005e-008, No Iterations 2
smoothSolver:  Solving for Uz, Initial residual = 1.496093e-005, Final residual = 5.666532e-008, No Iterations 8
smoothSolver:  Solving for h, Initial residual = 8.749293e-005, Final residual = 1.187253e-008, No Iterations 8
smoothSolver:  Solving for p, Initial residual = 8.486576e-007, Final residual = 5.21028e-007, No Iterations 102
smoothSolver:  Solving for p, Initial residual = 6.523312e-007, Final residual = 4.989128e-007, No Iterations 102
smoothSolver:  Solving for p, Initial residual = 5.678153e-007, Final residual = 4.820913e-007, No Iterations 102
time step continuity errors : sum local = 7.096876e-012, global = 6.886245e-012, cumulative = 0.0002020726
rho max/min : 1.209845 1.162966
smoothSolver:  Solving for omega, Initial residual = 0, Final residual = 0, No Iterations 0
smoothSolver:  Solving for k, Initial residual = 4.188657e-007, Final residual = 2.28007e-008, No Iterations 2
ExecutionTime = 934.098 s  ClockTime = 934 s

So i checked google for solutions. The only useful information i found was this thread in this forum:

https://www.cfd-online.com/Forums/openfoam-solving/81527-problem-sst-model-strange-behaviour.html

, where the user Felix recommends to set user lower values for the tolerance and relative tolerance for k and omega. But in their case omega got solved at the start of the simulation. However, i tried to use lower tolerances for k and omega, but the residual of omega is still zero. I also tried different BCs and solvers, but this won't work either.


Does anyone had to deal with the same error and found a solution? And what does it mean for the results when one field variable isn't solved?


Greetings,
Stuntmanbob




Edit: I'm working on Windows10 with the free Version of SimFlow 3.1, which uses the OpenFoam 16.12. I also tried the same simulation with OpenFoam 7 (19.06) on ubuntu installed via the windows subsystem for linux. But the result is the same

[cryabroad]

Originally Posted by Stuntmanbob

Hi,


i'm trying to simulate an airfoil of an linear cascade.
The mesh contains one airfoil and has periodic boundaries near the pressure and suction side. It is a q3d-case with only one layer, with a slip BC on the one sidewall and a symmetry BC on the other. The thermophysical properties are Perfect gas with constant thermodynamics and Sutherland for transport.



Check mesh only complains about about high aspect ratio cells. But these are necessary for y+ values less then 1.


I'm using the k-w turbulence model and the Smooth Solver for p, U, h, k and omega. The divergence scheme is Lust.


As for the boundary conditions i choosed k = Low Re Wall Function and omega = Standard Wall Function (nu_t = Low Re Wall Function and alpha_T Standard) on the walls.



The simulation runs fine and the residuals went beyond 1e-5 for Ux, Uy, Uz, p and k (h get stuck by ~1e-4). But here is the problem:
The solver won't solve omega from the first timestep.

velocityDampingConstraint velocityDamping damped 0 (0%) of cells
smoothSolver:  Solving for Ux, Initial residual = 3.597191e-007, Final residual = 6.192554e-008, No Iterations 2
smoothSolver:  Solving for Uy, Initial residual = 2.276461e-007, Final residual = 4.008005e-008, No Iterations 2
smoothSolver:  Solving for Uz, Initial residual = 1.496093e-005, Final residual = 5.666532e-008, No Iterations 8
smoothSolver:  Solving for h, Initial residual = 8.749293e-005, Final residual = 1.187253e-008, No Iterations 8
smoothSolver:  Solving for p, Initial residual = 8.486576e-007, Final residual = 5.21028e-007, No Iterations 102
smoothSolver:  Solving for p, Initial residual = 6.523312e-007, Final residual = 4.989128e-007, No Iterations 102
smoothSolver:  Solving for p, Initial residual = 5.678153e-007, Final residual = 4.820913e-007, No Iterations 102
time step continuity errors : sum local = 7.096876e-012, global = 6.886245e-012, cumulative = 0.0002020726
rho max/min : 1.209845 1.162966
smoothSolver:  Solving for omega, Initial residual = 0, Final residual = 0, No Iterations 0
smoothSolver:  Solving for k, Initial residual = 4.188657e-007, Final residual = 2.28007e-008, No Iterations 2
ExecutionTime = 934.098 s  ClockTime = 934 s

So i checked google for solutions. The only useful information i found was this thread in this forum:

https://www.cfd-online.com/Forums/openfoam-solving/81527-problem-sst-model-strange-behaviour.html

, where the user Felix recommends to set user lower values for the tolerance and relative tolerance for k and omega. But in their case omega got solved at the start of the simulation. However, i tried to use lower tolerances for k and omega, but the residual of omega is still zero. I also tried different BCs and solvers, but this won't work either.


Does anyone had to deal with the same error and found a solution? And what does it mean for the results when one field variable isn't solved?


Greetings,
Stuntmanbob




Edit: I'm working on Windows10 with the free Version of SimFlow 3.1, which uses the OpenFoam 16.12. I also tried the same simulation with OpenFoam 7 (19.06) on ubuntu installed via the windows subsystem for linux. But the result is the same

Have you checked the omega field in the flow? Is it a uniform distribution? What about k?

I've never used k-omega, but used k-epsilon quite a lot, which I think is pretty similar to k-omega in terms of how things are solved.

[Stuntmanbob]

Hi,


thanks for the reply
ich checked the omega and k distribution (see pics in the attachment). The omega values near the surface of the blade and in the wake are lower than in the main flow. The same with the k distribution. That seems fine to me.


Another point is, that the min, max and average values of omega change during the simulation:

    
volAverage() of omega = 2645.706


min(omega) = 1.153726 at location (0.2775581 0.01820756 0.0694632)
max(omega) = 3.789602e+008 at location (0.1311357 0.2398932 0.0694632)

the maximum seems very high to me. Does it mean something?


Greetings,
Stuntmanbob

[cryabroad]

Okay, so omega does have a non-uniform distribution, which would mean that the omega equation is actually solved. Not sure what's going on. One thing that is similar to this is when you use the compressible solvers, for density, the log file always read

solving for rho, Initial residual = 0, Final residual = 0, No Iterations = 0

. The explanation for it would be that the matrix for solving density is a diagonal one, so there's no need to do iterations. Maybe the omega field is the same?

In the k-epsilon model, epsilon is the turbulence dissipation, so its maximum value could be very high. I've had something in the order of e+006, so I guess your max(omega) should be reasonable? Need some k-w experts here!