3D RANS based shape optimisation mesh deformation
 

Hi everyone,

I am trying to setup a 3D RANS based shape optimisation on a wing. For this I use the Free-Form Deformation (FFD) method and placed a box around the part of the geometry that I want to deform. While the mapping of the FFD box onto the mesh seems to be okay, the subsequent mesh deformation results in unfeasible meshes. In particular, as the optimisation is RANS based the mesh contains prism layers, the mesh deformation method does not correctly deform these prism layers resulting in a completely wrong estimation of the friction drag. Some more information on my mesh and the used deformation settings:

Mesh:

• Around 6 million cells
• Surface mesh with only triangular cells
• Volume mesh with only tetrahedral cells
• Prism layers extending 10-25 cm above the surface

Deformation settings:
DEFORM_LINEAR_SOLVER= CONJUGATE_GRADIENT
DEFORM_LINEAR_SOLVER_PREC= ILU
DEFORM_LINEAR_SOLVER_ITER= 350
DEFORM_NONLINEAR_ITER= 1
DEFORM_LINEAR_SOLVER_ERROR= 1e-12
DEFORM_CONSOLE_OUTPUT= YES
DEFORM_STIFFNESS_TYPE= INVERSE_VOLUME
DEFORM_LIMIT = 1E6
DEFORM_COEFF = 1E6
DEFORM_ELASTICITY_MODULUS= 3E11


As many design variables (FFD box control points) are involved, the conjugate gradient method seems to work better in terms of memory usage compared to FGMRES. Moreover, both the wall distance and inverse volume stiffness deformation methods do not result in proper mesh deformations. Though I have read in other threads and thesis papers on best practices for mesh deformation that the inverse volume method is the best when prism layers are involved.

The last 3 inputs (DEFORM_LIMIT, DEFORM_COEFF, DEFORM_ELASTICITY_MODULUS) were not seen in the standard configuration file posted on the Git of SU2, I obtained these inputs via several other threads/google searches. While the solver does not throw an error saying they are invalid inputs, they also do not show up in the config_DEF.cfg file. So I am wondering whether these actually have an influence / are valid inputs. (Side note, I am using the newest version of SU2 V7.3.1)

Is there someone who has experience with 3D viscous mesh deformations who can provide me with some tips and tricks on how to move forward, and maybe highlight some possible mesh deformation settings that I have not included yet?

Thanks in advance!