Hi-

We are considering a license of a general purpose CFD code within our group at a major vehicle manufacturer. We are currently considering Star-CD, Fluent, and AVL's codes FIRE and SWIFT. Because we will initially be doing mainly in-cylinder flow modeling, we are attacted to the AVL FIRE code, due to its promise of easier meshing. We also plan to be involved in climate control, underhood and external aerodynamics calculations, and are therefore also considering a license of AVL's SWIFT code. Can anyone comment on the various strengths and weaknesses of FIRE and SWIFT relative to Star-CD and Fluent? I am particularly interested in issues of model building, solver stability, parallel scalability, physical models for engine work and the availability and quality of technical support from the various organizations. Thanks in advance for your input.

Sincerely,

George Hampel

Mr. Hampel,

I have had a small amount of experience using STAR-CD for in-cylinder flows. In my particular case I used the ICE code from Adapco to set-up my model for running with STAR. ICE maps a template mesh to user geometry and then generates all of the mesh commands necessary to move the mesh during the solution.

It took me several weeks to get up to speed on ICE at first, but now using ICE I can go from geometry to a ready-to-run model in about 10 days. Without ICE, the mesh set-up for STAR can take several MONTHS. I should note that ICE carries a pretty hefty additional price tag, but if you are serious about in-cylinder flows it might be worth it to you.

I have no experience with Fluent, however I did look into the moving mesh capabilities of FIRE about 6 months ago and the AVL rep. indicated that problem set-up time was similar to STAR-CD without ICE.

Regarding your question about parallel scalability. I routinely use the parallel version of STAR-CD, STAR-HPC for my larger models. It is extremely easy to use - mesh partitioning, CPU selection, and job submission is all handled by a simple GUI, and takes about 5 minutes to go from a single-zone mesh to a running parallel job. I get a speed-up of about 6 on 8 workstations connected via a non-dedicated 100BaseT network. On a multi-CPU machine, STAR-HPC achieves a speed-up of 7.5 out of 8.

STAR-CD tech support is handled by Adapco in the US and I have found them to be helpful, knowledeable, and responsive to bugs, my ignorance, and how to tackle new problems.

Sincerely,

Jeff Hank

Dear George,

Unfortunately, in-cylinder flow analysis is still a challenging problem for CFD. All of the solver you had mentioned have their own strength and weaknesses. We, at ICEM CFD Engineering are working with all of these solvers to minimize the efforts of model construction and setting the transient moving mesh analysis. Please check our web site at www.icemcfd.com. We will be more than happy to discuss our software capabilities in much greater details.

Dear George,

I have been using STAR-CD for full-cycle in-cylinder flow calculation (laterly with diesel spray and combustion) for quite a while. Previously I have had extensive experience with CFX and Fluent. My experience shows that as far as in-cylinder flow simulation is concerned, STAR-CD is the best available, although its solvers can't match those of CFX. To facilitate moving mesh/event generation for in-cylinder flow simulation, we have developed a program (for internal use) which I believe is similar to ICE but it allows us to preview the moving mesh at any crank angle prior to solving the flow field. This means that we can check the mesh interaction between valve and cylinder mesh. With this I am able to set up the model within 2-3 days depending on the mesh quality for the ports.

I talked to AVL about Fire and believe that the mesh movement is based on rezoning method. This means that meshes at various cranks may have to be developed in advance. Moreover STAR-CD uses mesh addition/removal technique which ensures that the mesh aspect ratio is not exceeding large near TDC, which alleviate the need for very small time steps.

Hope the above information can be of some help.

Regards,

Z. Xu