[Xiaosong]

Hi,

I am using SU2 as CFD solver for transonic viscous flow around rae2822 airfoil.

When solving for the flow field, if I use only one processor, it will give me very good results. However, if I use multiple processors, like 4, 16 and 32, it is always diverging. I am really confused about this, because the solver is exactly the same. Since I only change the number of processors, SU2 should give me very similar answer, using less time, correct? Anyone has any ideas to make it convergence also, when using multiple processors, please?

I am looking forward to replies from you guys! Thanks for your time!

Best,

Xiaosong

[fpalacios]

My guess is that you are running the problem with Multigrid, could you please try without multigrid? Unfortunately the MG algorithm depends on the number of partitions. Apart from that, and if you fully converge the linear system at each iteration
LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= LU_SGS
LINEAR_SOLVER_ERROR= 1E-10
LINEAR_SOLVER_ITER= 50
The solution per iteration should be the same independently of the number of cores.

Thanks for using SU2,
Best,
Francisco

[Xiaosong]

Originally Posted by fpalacios

My guess is that you are running the problem with Multigrid, could you please try without multigrid? Unfortunately the MG algorithm depends on the number of partitions. Apart from that, and if you fully converge the linear system at each iteration
LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= LU_SGS
LINEAR_SOLVER_ERROR= 1E-10
LINEAR_SOLVER_ITER= 50
The solution per iteration should be the same independently of the number of cores.

Thanks for using SU2,
Best,
Francisco

Hi, Francisco,

Thank you so much for your reply! That does help me a lot!

Today, I ran the rae2822 mesh in the folder TestCases/rans/rae2822 from SU2 official website. This mesh is fast and accurate. However, I only have the .su2 file, through which I cannot get enough details. So, would you please send me the original mesh, such as pointwise-format file, corresponding to the "mesh_RAE2822_turb.su2" file?

Thanks for your time! I am looking forward to your reply soon!

Best,

Xiaosong

[talbring]

Hi Xiaosong,

you can also look at the mesh by loading the flow solution file in Tecplot/Paraview.

Tim

[Xiaosong]

Originally Posted by talbring

Hi Xiaosong,

you can also look at the mesh by loading the flow solution file in Tecplot/Paraview.

Tim

Hi, Tim,

Thanks for your help! I can open the .su2 file using Tecplot now!

I like this mesh very much, so I hope to know more details, such as the number of nodes on surfaces and spacing at the leading/trailing edge. So, I am afraid only using Tecplot is not enough. Do you have more suggestions, please?

Xiaosong