I successfully ran a RANS simulation of industrial fan with all wall BCs, pimpleDyMFoam solver and spallart-almaras RAS model.

I got results upto some time, I would like to continue that simulation with LES instead of RAS, what are the options how I need to refine the mesh. (one parameter is way off the experimental value) specifically I would like to continue the simulation from where RAS ended say from 100s.

For instance if I am using one equation eddy model, how to accommodate nuSgs for the solution with RAS?

quick help would be appreciated


2 Answers 2


"pimple" is a good solver to work with, as it handles transient cases with RAS or LES turbulence. To change turbulence models, you'll need to specify this in your "turbulenceProperties" file by changing "simulationType" from "RASModel;" to "LESModel;" and providing an appropriate "lesProperties"file.

I'd suggest checking out the channel395 tutorial in tutorials\incompressible\pimpleFoam to see the usage of LES turbulence models. You'll also need the same boundary conditions for the solver (such as nuSgs, nuTilda, etc.)

That should be about it, and if you're missing something the error should point you toward something else to copy/modify.

For grid refinement, you'll want to have a much finer grid in regions with large gradients, because LES needs a finer resolution to resolve the turbulent kinetic energy accurately. It really depends on your flow conditions though. Try solving with your initial RAS mesh, then refine it in areas with sharp velocity gradients to improve your results.

I don't use "pimpleDyMFoam" for my research, but I am studying turbulent jets using "pimpleFoam" as a solver. pimpleDyM just uses a moving mesh.

  • $\begingroup$ this is an old question,unanswered, you make good points. I figured out how to map a RAS solution to LES with 'mapFields' utility! $\endgroup$
    – fedvasu
    Commented May 13, 2014 at 11:32

I used 'mapFields' utility, which can be used to map field values from any arbitrary case to other, irrespective of turbulence model used, just the boundary patches need to be specified in one-to-one correspondence in 'mapFieldsDict' ins system/ directory.

for LES case set-up one can see any of the myriads of LES tutorials (in my case incompressible)

For more information,see this

  • $\begingroup$ Do you just map U and p and leave nuSgs uninitialized? $\endgroup$
    – akid
    Commented May 14, 2014 at 9:13
  • $\begingroup$ yes, only those fields present in RAS simulation gets mapped, untill OF-2.3 you couldn't even selectively map fields you want, all fields from source case would get mapped to dest case. with OF-2.3 you can specify which fields to map. $\endgroup$
    – fedvasu
    Commented May 21, 2014 at 8:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.