FEM Software: Visual Nastran vs Cosmos?

You could take a product recommendation from me, or you could do what I do: get the vendors in and have them run test analyses on your stuff.

I have used Cosmos/M, DesignStar, Cosmos, FlowWorks, CosmosMotion in the past. Cosmos/M will do just about anything, but the interface is circa mid 80s. My choice in the SRAC family for most work is DesignStar because it is standalone, but links nicely with SW. Cosmos and DesignStar seem to leapfrog each other trying to catch up to Cosmos/M with capabilities. FlowWorks and CosmosMotion are nice, but seem to require a lot of resources. For CFD I use CFDesigner.

Currently I use NE/Nastran and NEiWorks. NE/Nastran comes with FEMAP for pre and post. It has a mid 90s interface and can do a lot. Using NE/Nastran with FEMAP assumes you know something about FEA. NEiWorks on the other had is very similar to CosmosWorks except that it interoperates with it's big brother very well which is not true of the SRAC products which really can't work with each other except on a superficial basis.

There are a couple things I look for in FEA. One is linking to SW without requiring SW to be running while I do FEA solutions. The obvious reason is that SW can still be used during long analyses. The other is having a separate standalone analysis file as a record of what was done. With integrated packages like CosmosWorks, a change in the model after an analysis means that the analysis is no longer valid. Nastran has a nice feature in that once you write the input deck for the solver, if you know what you are doing you can change all kinds of things very quickly and do trade studies and whatnot.

Further, NE/Nastran can do shell models and beam models with ease. And it can do mixed models with beams, shells and solids.

TOP

And as you are probably aware, FEMAP is owned by UGS/Seimens. FEMAP is under continuous development as it is the pre/post processor for NX/Nastran as it continues to develop from SDRC's I-deas.

MSC had a bad run, especially after UGS canceled the agreement to use FEMAP as the pre/post for MSC/Nastran. But just recently MSC has made some great strides in its interface (Nastran is Nastran is Nastran) especially with the SimEnterprise group of products as well as their "MD" group of products compares to standalone Adams, Nastran, Patran, Marc, etc.

When I was in the market for FEA not too long ago, I was looking at Algor, Ansys, Abaquas, and NEI. I even peered into NX/Nastran. Before I made my decision I saw an article about MSC's new MD (multi-discipline) product sets and was convinced enough that they are back in the game and worth a second look.

Just an FYI to make sure you are keeping your options open and looking at all that is available.

--Scott

We also reviewed MSC Nastran and NeNastran

MSC have priced themselves totally out of the market for small to medium sized companies eg $70,000 per MSC fixed seat against $20,000 for an NeNastran seat WITH floating dongle licenses for both the post and preprocessors.

Obviously prices vary per the modules you require but the above price comparison is for the same functionality in each product.

First a comparison of SRAC products:

CosmosWorks Integrated with SW. Basic FEA capabilities. Some more advanced capabilities. Good mesher Very fast static solver Must run SW to run CW

Designstar Very much like CosmosWorks but stands alone. Will integrate with SW for geometry changes. Capabilities of DStar and CW leapfrog each other. Will work with several CAD packages including CATIA, SE and SW. Will read parasolid, acis and I believe STEP and CATIA (with optional module) Saves a standalone model files that doesn't need SW.

Cosmos/M Full featured FEA program with implicit and explicit solvers, non- linear capabilities. At one time meant to be a poor mans ANSYS. Has roots in DOS and Unix. Interface circa 1980s/early 90s. Creates a standalone model file and can be scripted.

NE/Nastran Full featured FEA program with very good solvers and a long heritage. Pretty much all the capabilities of Cosmos/M except for one thing. No true axisymmetric elements. Like the other flavors can handle extremely large models. Has no problem utilizing dual or quad core processors. Fast PCGLS solver. Nastran is used heavily for dynamics problems but this flavor has very good contact and non-linear material capability. Runs from a text file created by a pre/post processor called FEMAP. FEMAP will read parasolid, acis, iges and many input decks from other FEA programs. This means if FEMAP is fed a Cosmos/M command file or an input deck from Ansys, or even another flavor of Nastran it can pull it in and modify it. The FEMAP interface is probably mid 90s technology.

NEiWorks Runs within SW like CosmosWorks, but writes a Nastran text based input deck for the solver. Therefore it has the benefits of documenting FEA without danger of accidental change and of running solutions independent of SW. NEiWorks Nastran text files can be read into FEMAP and NE/Nastran making this a very flexible system.

My preference is for FEA solutions that can integrate with SW for geometry but that create independent analysis files. This is for two reasons. First for documenting the FEA the standalone files create a fixed record. Second they can run the solver outside of SW leaving SW available for modeling. My candidates then are:

DesignStar NEiWorks NE/Nastran with FEMAP

The first two can be used by users with beginning to moderate skills in FEA while the later is for experts.

TOP

It also depends on what industry you work in as to what is acceptable to any certification authorities you need to rubber stamp your engineering.

In the aerospace industry Nastran is the only acceptable FEA analysis for the FAA, why? because the results have been backed up by numerous physical testing. The NAS in Nastran stands for NASA, they developed the software.

Some clarification on CW:

If you have a multi-cpu system and enough memory, you can run a second session of SolidWorks while SW & CW are running in the background. (You can't run two sessions of COSMOSWorks, though.)

DesignStar is now based on an OEM version of SolidWorks. Funcionality is pretty much identical to CW, except many SolidWorks functions (e.g., Simplify, Configurations), Optimization and Motion are not included. You cannot install DesignStar on a system that is running SolidWorks.

You can now mix beam, solid and shell elements in CW.

CW now supports non-linear materials and contact conditions. Advanced Dynamics will be in CW in 2008.

CW Advanced Professional includes COSMOS/M.

$20K for NeNastran! Wow. I bought my seat for under $6K about 6 years ago.

=========================================================================== Chris

How much did you pay for Solidworks 6 years ago?

As I said it also depends on what modules you want, non linear, composites etc.

Jim,

I have been away from SRAC products for a couple years. From what you have said Cosmos and DesignStar development has followed more along lines that marketing wants, rather that good engineering. If I had kept my DesignStar subscription then I would no longer be able to use it simply because SWinc didn't want it to compete with CosmosWorks. DesignStar used to be a Swiss Army knife because it would work with SW and many other CAD packages, SW just removed the main blade. (Does this give any hint where Greg Ray is going with things?). It also suggests that getting any kind of certification is not going to happen because of the radical changes the software is undergoing.

Jim, you bring up a good point about running outside of SW. I can fun FEMAP on one machine (the one with SW) and have the solver run on a completely different machine (maybe a Linux or Unix box, maybe Windows). Because of this, the other machine can also be 64 bit, whereas the SW would be on the faster for SW 32bit machine, whereas the FEA solver can run on a 64 bit multicore/multiprocessor machine with more memory allowing huge non-linear models to run in memory. Or they can all run on the same box (as they would for most people). I do the separate box install simply because I can strip Windows down in the solver machine to the bare essentials and occupy as many processors there as I can afford. From what I understand running Nastran on Linux provides even more speed. If I go with NEiWorks I can build the model in SW and solve it on another machine (except that NEiWorks is a one dongle software whereas NE/Nastran is a two dongle software).

It remains to be seen how well SW integrated beams, shells and solids. When I do this kind of work I prefer using quads for shells, and bricks for solids. Cosmos does not support this at all. I know in the past that when I tried to use shells in Cosmos I would frequently have serious difficulties. Since I had to have shells for the thin plates I now use this kind of unreliability was not acceptable. NE/Nastran addresses any problems quickly and to the customers satisfaction. SW takes a vote and if the other 599,999 customers don't think you need it fixed, it doesn't get fixed.

So from what you have said, SW has removed DesignStar from the running for anyone using SW which makes the choice that much easier. If you are serious about FEA, get Nastran, if you are a beginner and need something to get started with get CW or NEiWorks. If you later want to expand, Cosmos/M is not the route to go simply because NE/Nastran will read SW generated FEA models from NEiNastran.

TOP