CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   CFD-FASTRAN (

S. Russell May 18, 2001 15:24

Has anyone used CFD-FASTRAN and if so, what were its benefits and downfalls. Thank You in advance for your time.

Charles Crosby May 20, 2001 16:56

I have been using CFD-Fastran for the last twelve months or so. It is the first commercial CFD-code that I have used (worked previously with self-written and other research codes), so I have no objective basis on which to compare the performance to Fluent or CFX, for example.

CFD-Fastran's Strengths:

1. Multi-block structured grid which allows degenerate faces and edges is quite an efficient way of meshing an external domain with boundary layer refinement.

2. Structured multi-block also lends itself to parallel processing, and the parallel license pricing is attractive relative to other CFD codes. (It is double the standard license, independent of the number of processors, unlike most other codes, where you have to buy a license for each CPU, albeit normally on a sliding scale) Given the very low price of high-end PC's at the moment, it is possible to follow a very economical "brute-force" approach!

3. CFD-Geom is now a very capable structured mesh generator, it has in fact improved very significantly over the 12 months that I have used it.

4. Capability to handle overset moving grids (for store-release simulation, for example) is its major competitive advantage, although I haven't actually used it yet!

5. Stability is reasonably good, although I have battled (without any obvious reason) with one or two solutions.

6. CFD-View is a very nice post-processor.

7. CFD-Geom is a pretty good "parametric" mesh generator. It is possible and practical, for example, to build a single mesh, and then distort the mesh for different control surface deflections. This makes it very quick to get meshes for slightly differing configurations.

8. There are currently three turbulence models available, k-epsilon, k-omega and Baldwin-Lomax.

9. Perhaps rather strangely for a density-based solver, it handles low Mach number flows very well.

CFD-Fastran weaknesses

1. On the basis of anecdotal comparison to other codes (like Fluent and CFX, for example), it seems rather slow. I have not accurately benchmarked this.

2. We sometimes battle with transsonic flows to get the residuals down. Integrated forces seem to "converge" very nicely, but the residuals get "stuck". Very difficult to make a confident statement about convergence in a situation like that.

3. CFD-View post-processor lacks the ability to integrate pressures to get forces on arbitrary walls.

4. It can handle unstructured meshes, but at the moment it is not possible to get an "advancing layer" prismatic grid, so practical boundary-layer type meshing is not really feasible with the unstructured mesh. I think the unstructured mesh is currently best suited to inviscid simulations, for this reason.

We will be renewing our license for CFD-Fastran for the next twelve months, which I guess says something about it. However, the best advice I can give is that you do really need to check and benchmark it against the opposition FOR YOUR APPLICATION.

John C. Chien May 20, 2001 18:16

(1). It is an excellant review. Thank you for your detailed information. (2). I think, in the hands of expert, commercial codes can be useful to some extents. (3). But I personally think that the solution must be reliable in the first place for the solution to be useful.

clifford bradford May 28, 2001 17:08

If they fix (3) in the disadvantages section I'd buy it for sure.

Charles Crosby May 28, 2001 17:23


Charles Crosby May 28, 2001 17:26

Sorry, that was a bit of finger trouble ... Ah yes, the force integration in CFD-View. I've asked them to look into adding it into CFD-View. It shouldn't be too difficult, as all the necessary data are readily available. We wrote a custom program to extract the forces, but it is a bit kludgy, and it doesn't work with DTF files or unstructured meshes. It works fine for structured meshes if you don't use DTF, but that kinda rules out parallel processing.

All times are GMT -4. The time now is 20:07.