The company I work for is looking for a perpetual CFD license we can use in the aero field. The company is really just interested in modeling ECS and avioncs cooling, but may want to extend to the field of anti-icing and possibly decompression. So far we have been reviewing FLUENT, STAR-CD, CFdesign, and on a lower tier, FLOWORKS. The two that stand out are STAR and CFdesign.

My Question: The spokes-person for STAR stated that having the polyhedral (as opposed to tet) mesh would drastically decrease the PC-Processor power and run-time. He also stated that CFdesign would not be able to model basic ECS (duct flow through the aircraft cabin) because it is such a large volume and the enormously large amounts of tetrahedral meshes and iterations would generally freeze a "up-with-the-market" computer. Are there any users (of STAR and/or CFdesign) that can give me any info on your experience with these programs, and if what the STAR rep said was true?

Mmmm. That's quite a wide range of applications! It's not only Adapco that have the polyhedral meshing technology, it is now available from others, such as Fluent, as well. In CFD, as in other aspects of life, talk is cheap. I would ask the sales reps to show you how they would tackle a representative large problem, so that you can judge how well the code fits the application. Unless you are going to be doing many runs on a single mesh, chances are that you will spend more time meshing than solving, so the ease and robustness of the meshing process becomes a very important factor in the decision. I would personally be reluctant to commit to a perpetual license untill I've tried the code in real life. All the vendors I know of will let you have a demo license to experiment with. Try to get at least two months ... and whoever is going to do the work must also be able to set aside the time to work on this full-time.

I use STAR-CD and STAR-CCM+ and am a big fan of both but particually CCM+. CFDesign I have used a little and it seems good for what it is. I think you will find that what STAR will bring you is a lot more flexibilty than cfdesign, better parllelisation and probably more accuracy, models and development effort. Tetrahedrals are far more numerically diffusive than polys and hexas (trim cells in the star world) and you are likely to get better accuracy they are also less memory hungry and converge faster (according to the research papers and my limited experience with them).

If you are new to cfd I think you will probably find that the support/training organisation of cd will be greater than cfdesign, I have heard that cfdesign has very few support engineers and is more of a "heres your software, bye bye, type organisation". I have always found the cda support organisation very good (here in europe at least).

As I say I am commmenting on this as a STAR user so I may be biased, so its a judgment call on your part really.

I would be very wary about what sales agents say as their primary aim is to sell you their code-so they will definitely say that their code is the best while the others fall way behind. I would agree with Charles-get a few demo licences for different CFD codes and run a few "real-problem" test cases. And see how long it takes you to create a "practical but efficient" grid, memory usage, solution convergence, the ease of using the pre-processor and the post-processor (these are just a few things to look at).

the current versions of fluent (till 6.2) can not read polyhedral meshes. Or in other words you can not use polyhedral meshes with Fluent. Having said this, internally fluent can handle the polyhedral meshes almost exactly same as starccm+. Probably the coming versions might support polyhedrals meshes.

Agreed on that. I have seen demo's of Fluent 6.3 (which is out on Beta, and should be released really soon) with polyhedral grids.

6.3 beta is installed on our calculation server, never got time to explore it. When i will get time will do it.

If you are interested in icing calculations for external aerodynamics, have you heard of the SOLAR suite and the icecremo package?

Iain

ps The solver I use runs on polyhedral meshes, (this is the flow solver which is part of the SOLAR suite), the mesh generation involves cutting meshes across each other. This can result in cells with 50 faces, centroids outside of the cells (!!!), etc. Despite, all this, the solver has no problem with this since it is a face-based solver for the integration of the finite volumes.

My point is this:- even if the solver says it is a polyhedral solver, it probably actually means either it can only solve on a variety of shaped elements (tets, cubes, prisms, etc.) Or cells have to be well-behaved (i.e. centroids are located inside a cell).

I have tried a solver which reported itself to be "polyhedral" but it fell over on our grids.

Thank you all for the responses. I requested a few demos but some of them do not have transient capabilities (demo only) which is something we have to evaluate. My biggest issue is the limitations of CFdesign. The rep from CD-ADAPCO says they (CFdesign) have limitations but failed to give any examples.

FLUENT is a great program, but as of now our CFD jobs are very limited, and we don't think (as of now) we would be profiting much (if any) due to FLUENT's high prices.

STAR-CD had a variety of applications, and so did CFdesign. The issue now is that STAR-CD costs approximately twice as much as CFdesign.

Are you going to make design decisions with the CFD code? If so, I would be very careful about selecting a code based on price. CFdesign is cheap, and has a very slick front end which is nice, but the quality of the result will get you in a lot of trouble at your design reviews. Do you want an unhappy program manager and bad reputation because your CFD tool was cheap?

My suggestion, stick to the main players - STAR & fluent who have a well established proven solver and CFD experience. They are more expensive for a reason.