Thermal CFD Analysis Software Perfomance 1

[ZP48]

The application: design of electrical power distribution cabinets.
Current software package in use: Maya Heat Transfer module ESC (Electronic System Cooling) within IDEAS CAD/CAE software.
The software under consideration: Flotherm of Flomerics.
The reason: exploring more flexible and of robust capabilities (mesh generation, in particular) softwares - specifically required while modifying a simulation model.

Does anybody have experience with the Flotherm? Perhaps, in applying it for electrical distribution cabinets design. The cooling means - natural convection only. Also, how robust is the Flotherm in terms of reading/processing files created by different CAD programs (Pro/E, IDEAS, Solid Works, etc.).

Thank you.

 

[amorrison]

Post also to

http://www.cfd-online.com

 

[Bgoldstein]

I've got some experience with CFD packages for electronic cooling. I've used FloTherm briefly a few years (and versions ago). Back then it only had a non-conformal, rectangular grid, which I was very unimpressed with, though they might have changed since then. Have also used IcePak - at the time they had a slightly better meshing technique. Both of them (this is as of 2 years ago) did not have great CAD import abilities - everything came in as IGES files, sometimes small gaps were created, sometimes just got a mess if you had complex assemblies. However, both do have good, robust solver engines, especially for natural convection.

I'm currently using a product called CFDesign from Blue Ridge Numerics. I think it has a much better integration with SolidWorks and ProE, and has some neat meshing capabilities. It is a bit lower end than FloTherm and Icepak in terms of complex turbulence models and preset electronic component properties, but may be enough for what you are looking for.

 

[ZP48]

Bgoldstein,

Thank you for the detailed answer.

Few days ago, I had a one hour web demo by the Flomerics. Yes, I did observe a rectangular grid, density of which could be controlled. What do you mean by "non-conformal" grid?

ZP

 

[Bgoldstein]

If you've done much FEA, you've seen a "conformal" mesh. Basically, the old Flotherm and IcePak mesh was purely nonconformal rectangular - so if you had a round object, the mesh would not rotate to better match the geometry. Also, the grid "bleeds" out though the entire extent - if you have a tight grid around a few parts, this grid would extend everywhere. Conformal meshes will change in size depending on the geometry, and better match the shape.

For example, CFDesign (and a bunch of other more advanced packages, I'm not an expert about all of what is out there) has a basic tetrahedral mesh (think of pyramids), with polyhedrals (rectangular shaped) elements at solid to air interfaces.

What scares me about non-conformal meshing is that for good modeling of solid-to-fluid interfaces boundary layers (critical for natural convection, by the way), you want lots of elements near the surface. With non-conformal, this means a lot of extraneous elements from bleedout, and don't even try to think about the effect if your parts are not nice and rectangular (mine never are!). The saving grace for non-conformal meshes - in electronic packaging, often have mostly rectangular shaped objects, or objects that could be modeled as such.

 

[ZP48]

You brought an important factor(s) that I assumed (rightly or wrongly) Flotherm is capable of doing - to mesh according to size of an object and to use tetrahedral elements. I did see that the software, after a CAD model is brought in, goes through geometrical simplification of the model, i.e., removes curves and holes.

I faced the same necessity in IDEAS-MAYA, during the meshing process. Too many holes and radiuses consumed a lot of elements, so, the holes and radiuses had to be either removed or the curves changed into straight lines.

IDEAS-MAYA does autosuggest size of an element, which is based on size of the object. That size and a desired quality of elements can be input directly, i.e., the program's default values can be overwritten.

In IDEAS, you do need to specify flow blockages (by selecting the related volumes, for example), in order to separate fluid from solids. This, I was told, is not necessary in Flotherm. Assuming the program produces valuable results, it is an improvement.

 

[ko99]

I currently use Flotherm and CFDesign and have also used IcePak. I'm not a CFD expert, but I am very experienced at thermal modeling. A few comments:

All three model natural convection and radiation, among other things, quite accurately.

If you need curves to define your objects, IcePak and CFDesign should work much better for you than Flotherm.

I like CFDesign's CAD interface, but of course there's a LOT of overhead in directly importing a complete detailed design. Sometimes I create a more simplified CAD geometry before importing it to CFD.

Flotherm uses an orthographic (X-Y-Z direction only) grid. It's very efficient for geometry that can be defined accurately without curves (heat sinks, ICs, PCBs, fans, vents, etc). By efficient, I mean it's easy to build models, very easy to grid, get quick convergence, and run lots of variations.

Flotherm (and IcePak) now support localized grids to minimize bleedout.

CFDesign is new, though it's growing, and has less features.

IcePak and Flotherm have macros that are very convenient for electronics cooling applications (heat sinks, fans, vents, ICs, etc). Often it's easier to create the geometry directly in the CFD tool rather than importing it and cleaning it up.

Flotherm and, I think IcePak, also have very powerful automated optimizers (I'm less familiar with IcePak).

Good luck,

ko (

http://www.ecooling.biz)