Flat Plate RANS Divergence

[James Johnson]

Hello,

I have been trying the k-tau RANS model on a 2D flat plate geometry with boundary conditions are similar to this setup. 

The simulation is diverging when the RANS model is used. I think the issue is centered around the leading edge of the plate, where the symmetry boundary changes to a wall. In my testing I tried to resolve that region pretty well, but that didn't seem to help. Using really small time steps let things run longer, but that does not seem like a good solution and the output still identifies some elements as having negative TKE/TAU even after running for tens of thousands of steps. 

Attached are a couple of screenshots of the k and tau fields with the mesh near the leading edge as well as the log file from that run.

I also tried changing the shape of the mesh in that region, which did seem to help get some boundary layer development, but the problem persists where I need to take small steps (CFL of 0.1) and some elements have negative TKE/tau values.

Any suggestions for resolving these issues would be appreciated.

Thanks,

James