code_saturne User's Forum

I would like to know if Code Saturne can function as a low speed virtual wind tunnel to investigate external aerodynamics of small(ish) objects. The expected Re is up to 45000.

I have looked through the user guide and read as much as I could find and understand about cases that have some overlap with what we need, but I still have some questions.

- Does Code Saturne support moving mesh or deformable mesh or some other way to simulate motion besides the supported periodic rotation (also useful)?
- Does the workflow permit rapid geometry changes? In other words can the case be retained and just the mesh swapped out, or does the entire case need to be rebuilt every time a change needs to be made?

I also have a question about performance. How well does Code Saturne scale over multiple CPUs (not just cores)? I understand that Code Caturne runs very well on supercomputer grade machines, but their memory controllers and bandwidth are a bit beyond the performance of Xeon or Opteron workstations.

Thank you,

V.

Hello,

In the sense that low-speed wind-tunnel flows can usually be considered to be incompressible, yes, you should be able to use Code Saturne investigate external aerodynamics of small(ish) objects. Depending on your mesh refinement, you may need to use either high or low-Reynolds turbulence models, but Code_Saturne has both.

- Yes, Code Saturne supports moving or deformable meshes besides periodic rotation. For very large deformations, a domain coupling approach similar to the one used fro periodic rotation may be necessary, otherwise just use the ALE functionnality.

- Yes, the workflow permits rapid geometry changes? If you build several meshes with different geometries but similar boundary and volume zone names, you can swap meshes with the same data setup (you can also define boundary conditions and such on groups which do not exist on a given mesh, so even if certain topological features appear only on some meshes, you can still swap meshes if your data setup is sufficiently well prepared).

Scalability over multiple CPU's is usually good, up to a point: the performance is often more memory-access bound the CPU-bound, so above a certain number of cores, you may not gain much if the case is large enough that it does not fit in cache. If the data set is too small, there is not enough work to divide. In the intermediate region, parallel performance is pretty good. Your mileage may vary, depending on you hardware and the size of the case you are running, so nothing can replace benchmarking here. I have not made detailed benchmarks on a workstation recently, but scalability is better on my office Xeon E5504 than my laptop's I5, which is consistent with the extra cache being more useful than CPU power for this.

Regards,

Yvan

Yvan,

thank you for the quick and informative reply. Can you tell me if there are some freelancers/researchers/students that would be willing to set up a script or specific instruction set to allow me to conduct just external flow simulations using a set of variables, with or without moving parts?

I am an end user, not a researcher or developer so I would like to limit the time it takes to learn how to use Code Saturne just for our cases. I do not require thermal simulations, or coupling - except perhaps solid stresses (FEA) for coupled aerodynamics/structural optimization further down the track.

We in fact already have one student from IPSA working on a dynamic structural analysis problem and we may engage an intern from INSA for general engineering R&D...thus we already have a record in working with French students

Cheers,

Victor

Hello,

I do not know many students, as we only have a few interns already working on specific subjects.

I'll check with colleagues who give courses in engineering schools, and as there are a few users of Code_Saturne at the University of Manchester, you might have chances there, but freelancers for this unfortunately seem quite rare...

You best hopes might actually be to post a with a more explicit title on this forum. There are not many freelancers, but there are a few at some times.

Best regards,

Yvan

OK, I will make the post here and see what happens.

I looked at the Manchester Uni Code Saturne page and it is neat, except that I could not find anyone to contact there and their forum is a bit...dead. Thus can you tell me who might be a good person to email there as a starting point? Personal message will do as I am not sure they would want their email address nuked by spam bots...

Cheers,

Victor