Grid deformation for high aspect ratio grids

praveen · January 6, 2013, 02:51

While trying a shape optimization with v1.1, I find cells get all tangled up after using SPRING method for deformation. The grid is hybrid with triangular prisms in the boundary layer. My grid does have high aspect ratio cells, maybe as high as 10^5 in the boundary layer. Trying TORSIONAL_SPRING option, the code seemed to run forever even for almost zero deformation case, and I had to kill it. It is possible my grid is very bad, though I have been able to run su2 cfd on it. I would like to know if there are any known issues with such high aspect ratio case for grid deformation.

Also, since FFD already provides volumetric deformation, is it possible to use FFD itself for grid deformation ?

economon · January 6, 2013, 20:45

Hi Praveen,

Mesh deformation with 'viscous' grids can be more difficult simply due to the highly stretched cells, like you mention. I have had some luck in the past with using the TORSIONAL_SPRING option while also turning down the error tolerance of the grid deformation solver using the GRID_DEFORM_ERROR option. Note that this controls the tolerance of the volume grid deformation and not the surface movement. By default, this value is 1e-14, and relaxing it will often allow for the code to converge. Lastly, there is also a third deformation technique, ALGEBRAIC, that you might try.

At the moment, FFD is not available as a mesh deformation technique on its own, however, this has been discussed by the dev team and could certainly be implemented in the future.

Hope this helps!

praveen · January 17, 2013, 03:32

I am now experimenting with v2. I increased the tolerance to 10^(-8) but still torsional spring does not finish. This happens even on initial grid where the deformation should be zero. Can I further increase it considering that I have highly clustered grids ?

economon · February 1, 2013, 00:44

Have you had any luck since with the mesh deformation?

You could always try some tests by running SU2_MDC directly with your config file and varying some of the parameters (raise the tolerance even further, try algebraic, try a different linear solver, etc.)

Also, one other note is that the algorithm will often fail if a zero deformation is imposed. It is best to at least impose a small epsilon for the design variables.

praveen · February 12, 2013, 23:42

I did try algebraic in the past but did have any luck with it. I will check again. In v2 how do I select linear solver. The old variable does not seem to work since I get this warning

WARNING: unrecognized option in the config. file: GRID_DEFORM_SOLVER.

fpalacios · February 15, 2013, 19:50

Quote:

Originally Posted by praveen

I did try algebraic in the past but did have any luck with it. I will check again. In v2 how do I select linear solver. The old variable does not seem to work since I get this warning

WARNING: unrecognized option in the config. file: GRID_DEFORM_SOLVER.

In the current version, conjugate gradient is the solver by default for grid movement.

January 6, 2013, 02:51	Grid deformation for high aspect ratio grids	#1
praveen Super Moderator Praveen. C Join Date: Mar 2009 Location: Bangalore Posts: 342 Blog Entries: 6 Rep Power: 18	While trying a shape optimization with v1.1, I find cells get all tangled up after using SPRING method for deformation. The grid is hybrid with triangular prisms in the boundary layer. My grid does have high aspect ratio cells, maybe as high as 10^5 in the boundary layer. Trying TORSIONAL_SPRING option, the code seemed to run forever even for almost zero deformation case, and I had to kill it. It is possible my grid is very bad, though I have been able to run su2 cfd on it. I would like to know if there are any known issues with such high aspect ratio case for grid deformation. Also, since FFD already provides volumetric deformation, is it possible to use FFD itself for grid deformation ? __________________ http://cpraveen.github.io http://twitter.com/cfdlab http://github.com/cpraveen

January 17, 2013, 03:32		#3
praveen Super Moderator Praveen. C Join Date: Mar 2009 Location: Bangalore Posts: 342 Blog Entries: 6 Rep Power: 18	I am now experimenting with v2. I increased the tolerance to 10^(-8) but still torsional spring does not finish. This happens even on initial grid where the deformation should be zero. Can I further increase it considering that I have highly clustered grids ? __________________ http://cpraveen.github.io http://twitter.com/cfdlab http://github.com/cpraveen

February 12, 2013, 23:42		#5
praveen Super Moderator Praveen. C Join Date: Mar 2009 Location: Bangalore Posts: 342 Blog Entries: 6 Rep Power: 18	I did try algebraic in the past but did have any luck with it. I will check again. In v2 how do I select linear solver. The old variable does not seem to work since I get this warning WARNING: unrecognized option in the config. file: GRID_DEFORM_SOLVER. __________________ http://cpraveen.github.io http://twitter.com/cfdlab http://github.com/cpraveen

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
MUSCL shock capturing on high aspect ratio meshes	vijay_paul	Main CFD Forum	1	September 29, 2012 00:16
[ICEM] Aspect ratio vs quality of a tetra mesh	vmlxb6	ANSYS Meshing & Geometry	5	September 25, 2012 13:21
[ANSYS Meshing] How to Control Aspect Ratio in Ansys meshing	mecbe2002	ANSYS Meshing & Geometry	0	May 28, 2012 07:56
High Aspect Ratio elements	Flavio	CFX	2	November 24, 2006 12:01
Grid generator for structured grids	Tore Myhrvold	Main CFD Forum	5	January 21, 1999 08:09

January 6, 2013, 20:45		#2
economon Super Moderator Thomas D. Economon Join Date: Jan 2013 Location: Stanford, CA Posts: 271 Rep Power: 14	Hi Praveen, Mesh deformation with 'viscous' grids can be more difficult simply due to the highly stretched cells, like you mention. I have had some luck in the past with using the TORSIONAL_SPRING option while also turning down the error tolerance of the grid deformation solver using the GRID_DEFORM_ERROR option. Note that this controls the tolerance of the volume grid deformation and not the surface movement. By default, this value is 1e-14, and relaxing it will often allow for the code to converge. Lastly, there is also a third deformation technique, ALGEBRAIC, that you might try. At the moment, FFD is not available as a mesh deformation technique on its own, however, this has been discussed by the dev team and could certainly be implemented in the future. Hope this helps!

February 1, 2013, 00:44		#4
economon Super Moderator Thomas D. Economon Join Date: Jan 2013 Location: Stanford, CA Posts: 271 Rep Power: 14	Have you had any luck since with the mesh deformation? You could always try some tests by running SU2_MDC directly with your config file and varying some of the parameters (raise the tolerance even further, try algebraic, try a different linear solver, etc.) Also, one other note is that the algorithm will often fail if a zero deformation is imposed. It is best to at least impose a small epsilon for the design variables.