[diwakaranant]

Hi

I was trying to find the gradients for Onera M6 wing for rans case using
finite_differences.py using SU2 v 2.0.5

I am using "FEA" as grid deform method, but the mesh deformation doesn't
seem to converge when the mesh is deformed after each flow run.

----------------- FFD technique (parametric -> cartesian) ---------------
Update cartesian coord        | FFD box: 0. Max diff: 1.46435e-07.

----------------------- Volumetric grid deformation ---------------------
Performing the deformation of the volumetric grid.
=> BCGSTAB residual history
=> residual tolerance target = 1e-08
=> initial residual norm     = 7.61365e+06
     0     3.20299e+12
     50     15819.9
     100     2367.66
     150     851.439
     200     613.143
     250     23395.4
     300     3507.32
# BCGSTAB final (true) residual:
# iteration = 300: |res|/|res0| = 3507.32
End and write output files.

------------------------- Exit Success (SU2_MDC) ------------------------

The mesh deformation code seems to run upto 300 iterations leading
to incomplete convergence.

Why is this happening ?

Thanks
Anant

[fpalacios]

Originally Posted by diwakaranant

Hi

I was trying to find the gradients for Onera M6 wing for rans case using
finite_differences.py using SU2 v 2.0.5

I am using "FEA" as grid deform method, but the mesh deformation doesn't
seem to converge when the mesh is deformed after each flow run.

----------------- FFD technique (parametric -> cartesian) ---------------
Update cartesian coord        | FFD box: 0. Max diff: 1.46435e-07.

----------------------- Volumetric grid deformation ---------------------
Performing the deformation of the volumetric grid.
=> BCGSTAB residual history
=> residual tolerance target = 1e-08
=> initial residual norm     = 7.61365e+06
     0     3.20299e+12
     50     15819.9
     100     2367.66
     150     851.439
     200     613.143
     250     23395.4
     300     3507.32
# BCGSTAB final (true) residual:
# iteration = 300: |res|/|res0| = 3507.32
End and write output files.

------------------------- Exit Success (SU2_MDC) ------------------------

The mesh deformation code seems to run upto 300 iterations leading
to incomplete convergence.

Why is this happening ?

Thanks
Anant

The current 3D implementation of the grid deformation is based on Tets (as a preprocessing the Hex are divided into Tets). This is not optimal, we are working in a real FEM implementation of the Hex element. But this going to take some time unless somebody in the open-source community helps us with this "easy" development.

Meanwhile, our recommendation is use Tets grids or perform small deformations and use the parameter.
LINEAR_SOLVER_ITER=10

Cheers,
Francisco

[economon]

Hi,

I'll also pitch in and say that it may be worth trying with one of the newer versions of SU2, such as V2.0.8, as we have made a few improvements to the FEA since V2.0.5.

T