[crevoise]

Hello

I am getting some convergence problem due too a sharp corner on my geometry. More precisely, the turbulent dissipation rate get residual up to 10, while getting high value at that corner on few cells.
I guess my mesh at that corner is not the best (see attached), and I wanted to know if someone has some hint on how to improve it.

Thank a lot for any help

Corner_Mesh.jpg

Corner_TurbulentDissipationRate.jpg

[Gary6891]

Did you try Cell quality remediation option? Or your initial values might be worth rechecking

[cwl]

Finer mesh based on Volume Shape might help

[crevoise]

Thanks a lot for your answers.
I did tried the cell remediation, but did not realy influence that much.
However, as suggested, I checked my initial values and modified them with some values obtained aith some not well converged computations. Thus the residuals are much better. (see pictures)

I have then a question, to get more light on this as I am not an expert.
I was thinking that the inital values should have an influence on the 'rapidity' of the convergence, but not on the accuracy of the solution.
However, but having good inital values shows that the solution is converging properly, while wrong values do not allow convergence to accurate solution.
So it seems that the initial values have an influence on the final solution, which sounds strange to me.
Am I wrong in my understanding on solution independency toward initial values?

Thanks again for your help and your info.

[cwl]

In case of convergence of the simulation - changing initial values should result only in rapidity of convergence; but convergence itself is highly affected by Solver Settings like CFL, Under-Relaxation Factors.
The convergence in terms of (low) Residuals is not a strict convergence; i believe that simulation run can be considered complete, converged and ok - when mass and heat (at least) ballances (computed using Reports and Field Functions) do converge, - and that i consider a criteria.

[me3840]

Don't touch the URFs or use cell quality remediation. The problem is your mesh. Your prism mesh is far too coarse, you only have 2 prisms, and it looks like it's not thick enough. What's the y+ on that surface? It's probably through the roof. Your boundary layer looks like it's well outside the prism layer on that cut.