CFD to FEA data transfer procedure

[blues]

Hi guys,

I'm looking for people with experience in transfering data from CFD simulation to FEA analysis software. The packages involved are Fluent 5.5 for CFD and Ideas 9 for the FEA.
I've had a few (apparently succesful) tries already but the accuracy of the results is very critical so I need to check that the procedure I followed is correct. Furhtermore, this same procedure will be used on other projects should it proves its worth.

The end result required:
Check the level of stress is below failure (for a given material) and that deflections are kept to a minimum for a cone (2.8m big dia, 0.4m small dia and 0.85m heigh) submitted to winds up to 10 m/s (at 5200m altitude) and orientated in different ways relative to the oncoming wind (cone horizontal, tilted 45 degrees and vertical).

Procedure followed:
1. Using CFD, the velocities and pressures on/around the surface of this cone are computed. 'Forces' at nodes are exported from Fluent as an Ideas universal file.
2. The UNV file is re-written (using a bit of Idl code) so that the 'Force' result set (written by fluent) becomes a Load Set (to be used by Ideas) which specifies forces at nodes.
3. The later processed unv file is then imported into Ideas, a BC is then created including all these nodal forces.
4. The model is 'cleaned' to get rid of unwanted elements, nodes and forces (i.e. walls and other surfaces that were used strictly for the CFD model).
5. Apply material to the elements and solve.


Please let me know of your thoughts.
Thank you,

Oliver
Mech. Eng.
Cardiff University, Wales
spxoem@astro.cf.ac.uk

 

[sanab123]

Hi Blues
i m in the same boat. But i m using Panel Method(VLM) for Aerodynamics. I have written a Fortran program on the guidelines given the MSC/NASTRAN manual. But still the Question is how one can transfer the pressure loads of CFD to nodal loads of FEM both having different meshes. Since the method given in MSC/NASTRAN is for point-load to point-load transfer i m getting large local deflections due to large point loads coming on the mesh.
Further i m doing aeroelastic Optimization which causes divergence to occur for low values of skin thicknesses.

 

[pja]

If the aero grids and structural grids aren't the same
the transfer must be done using some sort of mapping(interpolation,spline,etc) In my aeroelastic code which uses a nonlinear VLM and a modal solution of nonlinear plate equations I use a thin plate
spline to map from one "grid" to the other.