nuTilda and convergence?
 

I am new to OpenFOAM and CFD in general. Eventually, I want to evaluate some wing designs, but I am starting with simple 2D cases with known results to learn how to set up cases.



First, I set up this NACA 0012 example from the OF User Guide:
https://www.openfoam.com/documentati...irfoil-2d.html
I downloaded the grid generated by NASA, and simpleFoam converged in 1367 steps and got very accurate results.


The next step was to generate my own grid. I find BlockMesh and SnappyHexMesh confusing, so I tried using GMSH instead. I created an unstructured grid with a boundary layer (the NASA grid is a C-grid). The NASA grid has only "inlet" and "outlet" boundaries, so when I generated the grid in GMSH, which assumes a rectangular domain, I grouped the top, bottom, and upstream boundaries to form the "inlet" boundary.


I then used the same case files as before with my grid. However, after 3000 steps, simpleFoam still does not converge. From the output on the screen, it looks like there is a problem with nuTilda - the average value is about 10x larger than the initial guess, which is also the converged value for most of the cells in the solution obtained using the NASA grid (3.42e-5). Also, the residual value seems stuck and won't decrease. This is the screen output for the last few steps:


Time = 2998

smoothSolver: Solving for Ux, Initial residual = 3.54551e-06, Final residual = 2.34393e-08, No Iterations 7
smoothSolver: Solving for Uy, Initial residual = 8.58902e-07, Final residual = 5.89313e-09, No Iterations 7
GAMG: Solving for p, Initial residual = 3.9984e-05, Final residual = 3.36825e-07, No Iterations 6
time step continuity errors : sum local = 4.4999e-08, global = -2.3288e-09, cumulative = 9.06463e-05
smoothSolver: Solving for nuTilda, Initial residual = 0.0214847, Final residual = 0.000213755, No Iterations 217
bounding nuTilda, min: -0.000319118 max: 0.00758093 average: 0.000344346
ExecutionTime = 36871.7 s ClockTime = 43763 s

Time = 2999

smoothSolver: Solving for Ux, Initial residual = 3.30334e-06, Final residual = 2.49857e-08, No Iterations 7
smoothSolver: Solving for Uy, Initial residual = 8.50989e-07, Final residual = 6.09523e-09, No Iterations 7
GAMG: Solving for p, Initial residual = 4.28384e-05, Final residual = 3.36684e-07, No Iterations 6
time step continuity errors : sum local = 4.49822e-08, global = -1.00682e-09, cumulative = 9.06453e-05
smoothSolver: Solving for nuTilda, Initial residual = 0.0214627, Final residual = 0.00021444, No Iterations 233
bounding nuTilda, min: -0.000258275 max: 0.00797989 average: 0.000370175
ExecutionTime = 36885.5 s ClockTime = 43777 s

Time = 3000

smoothSolver: Solving for Ux, Initial residual = 3.54086e-06, Final residual = 2.33953e-08, No Iterations 7
smoothSolver: Solving for Uy, Initial residual = 8.61858e-07, Final residual = 5.94455e-09, No Iterations 7
GAMG: Solving for p, Initial residual = 3.99437e-05, Final residual = 3.63031e-07, No Iterations 6
time step continuity errors : sum local = 4.85004e-08, global = -3.75333e-09, cumulative = 9.06416e-05
smoothSolver: Solving for nuTilda, Initial residual = 0.0214879, Final residual = 0.000213995, No Iterations 217
bounding nuTilda, min: -0.000327578 max: 0.00758115 average: 0.000344323
ExecutionTime = 36903.6 s ClockTime = 43795 s

End



Am I correct that nuTilda is the problem? How is it related to the grid? Any suggestions on how to change the grid?

From another recent post, https://www.cfd-online.com/Forums/op...-made-out.html, it looks like the problem is that I need to change the settings in fvSchemes and fvSolutions for OF to work with an unstructured mesh.