Convergence accelaration
Hi SU2 developers,
I have started exploring SU2. Onera-M6 wing simulation (RANS), took around 15000 iterations for convergence. I urge developers to give a try to (LU-SGS+GMRES) matrix free implicit method (which is supposed to give much faster convergence). Also some of the commercial software uses good initialization. For example Fluent uses so called FMG (Full Multigrid) initialization to obtain good initial solution (Euler). With this initial solution, one can quickly arrive at converged results. My experiences with some of the commercial software -->> converged results for forces and moments are obtained within 300 to 1000 iterations for most of the cases. I am sure SU2 will come to this level very soon :) Thanks and Regards Amit (Harry) |
Hi Amit,
Thank you for your suggestions! We are continually working on improving the code's capabilities, and initialization and inversion methods are certainly items we hope to improve in coming releases. Kind regards, Aniket |
All times are GMT -4. The time now is 20:43. |