FEA and CFD "integration"

[ekline]

I am evaluating multiple FEA packages for my company. I have 6 years experience running MSC/Nastran, and a little experience with SDRC and CosmosWorks. I have narrowed the list to ANSYS, ABAQUS, and NEiNastran. The issue is that we make vacuum motors, so in addition to looking at shaft and housing dynamics, we may want to get into CFD to improve the centrifugal fan designs. I don't think the pressures we deal with affect fan shape, so I don't really need coupled physics, but should I want "integrated" FEA/CFD?

ANSYS/CFX seems like the best "integrated" CFD/FEA package. NEiNastran is good with vibration and the FEMAP interface could also be used with MAYA TMG (CFD software, although probably not as good as CFX). ABAQUS has a steady state dynamics solver that was developed for the tire industry. All three of these companies have given great support during the demo stage.

I have no experience with CFD (aside from this evaluation process) and don't know if it is reasonable for my company to get into it with our staff of 5 engineers, who will not likely devote more than 20% of their time to simulation. However, it may be something we attempt in the next year, and if we do, should the CFD package affect which FEA package I choose now?

Also, I'd appreciate any insight you can lend comparing the three packages above for structural and vibration analysis when an unbalanced rotating shaft is involved. Our motors consist of an assembly of sheet metal and injection-molded parts, steel shafts, and non-structural masses.

Much thanks for all the previous posts on related topics!

 

[drjrice]

I have worked in the CFD area for almost 30 years and have used a number of the commercial packages. I am currently using ANSYS/CFX and am biased, but I do believe that it offers the best integrated solution available today.

 

[fkmeyers]

I have worked in the aerospace industry for over 30 years and too have used a number of commercial FEA packages. We use NEiNastran with FEMAP and TMG for CFD and are quite satisfied. My company is a large aerospace corporation that builds rockets and aircraft. We have other packages including ABAQUS and use that as well for specialized types of analysis. I have used ANSYS and am not arguing that it is not a good package. All of these products are good products in their own right. Reading your comments I see 2 things:

1. You have previous experience with Nastran (6 years).

2. You are not sure if you will be doing much CFD.

Hence the question…”should the CFD package affect which FEA package I choose now?” Based on #2 above no it should not. All the packages you mention have good CFD integration. We have found TMG to be one of the best for our applications. But the fact remains you have 6 years experience with Nastran and ANSYS is very different from that. Nastran also is one of the best products for dynamics analysis on the market. I would recommend NEiNastran and yes I too am biased. If you look through the threads in this entire forum you will find that generally NEiNastran is recommended because it is a great product and well supported. That support will be important when you guys pick up a new project and need help getting it started. As for rotor dynamics I have done some of this with NEiNastran for a rotating turbine application and it worked well for that. They support complex eigenvalue analysis which I know can be used for this type of analysis as well.

Frank

 

[cloche]

I may also suggest having a look at ALGOR Multiphysics, which has the advantage to combine explicit dynamics and "static" analyses (though it may sound a contradiction - what I mean is that it performs static analysis at each timestep of a time-dependent analysis where the dynamic effects come from explicit formulation). This would be fine for rotor dynamics although it seems to me it doesn't have dedicated processing functions for rotodynamic / vibroacoustics. Recently the CFD module has been improved and allows for rotating frames (not sure if it also has moving meshes).
But, as Fkmeyers points out, if you already have knowledge on Nastran, why not use it?
Another solution, esp. if you want extremely strong integration with CAD (do you use SolidWorks?) is CosmosWorks + CosmosMotion + CosmosFloWorks (but I wonder if the sum of these packages would be less expensive than Nastran + CFD...)

 

[ekline]

cloche,

Thanks for the advice. We run Pro/E, although I'm a big fan of SolidWorks. The big selling point the Cosmos salesmen were pushing was the integration with SolidWorks, which isn't an advantage to us. Secondly, the current versions of CosmosWorks don't allow integration of shell and solid elements. Our designs use sheet metal and injection molded parts. Cosmos claims that they can handle that, but I don't buy their "use a really fine tet mesh on the sheet metal" solution. DesignStar and CosmosM allow the mixed element models, but at the expense of CAD integration. I believe CosmosWorks will implement mixed element types soon, but that was the main reason I eliminated them from consideration. Also, FloWorks doesn't have the ability to handle non-axisymmetric stators (which is also supposed to be addressed soon).

drjrice and fkmeyers, thanks for your posts. The input is much appreciated and worth more than a truckload of marketing brochures.

 

[cloche]

IMHO you should forget CosmosWorks world. At least here in Italy their products are supported by qualified people that would NEVER, but NEVER and again NEVER say that you can handle a shell-solid model by "using a very fine tet mesh on sheet-metal part"!!!!
If CAD integration is important, Algor is linked to Pro/E. So does ANSYS, if you install the Workbench interface. In our company, ANSYS is THE reference for all FEA analyses and we believe it is still the best by far for our kinds of problems, but of course this statement depends on anyone... I have no knowledge on ABAQUS or NASTRAN.
And as regards CFD, why not having your tasks accomplished by an external service, if you plan to use it seldom? There are at least two cases where outsourcing CFD is valuable:
- either your problem is very complicated (free surface flow, moving meshes,...), so that the programs able to handle it are extremely expensive
- or your problem is simple but not very frequent, so you would have a really under-used license of the CFD program

 

[fkmeyers]

I think the SolidWorks + Nastran integration in NEiWorks is pretty good. We use some SolidWorks here but mostly Catia. I would check out their website (noran engineering) and their 1.1 NEiWorks release. I know for a fact they support solid/shell meshing, surface contact, assemblies, etc. The only disadvantage is their CFD package is for integration with FEMAP and not SolidWorks. You can always get that later. I know for a while they were including NEiWorks for free with NEiNastran. I am not sure if they are still doing that. Still if you are not concerned about integration with SolidWorks then NEiNastran with the FEMAP modeler will work best. They also have multiphysics but simply put it is just nonlinear transient response.