Better Alternatives to Simulation/CosmosWorks? 1

[wpeaton]

I've been using CosmosWorks/Simulation off and on for about 6 years now. For simple problems it's pretty good. But for anything a little complicated, it often turns into a struggle.

Is there anyone out there who has used CosmosWorks/Simulation and another product? I always wonder if I'm wasting my time when I could be using another product and get better results. Or any results at all. Complex assemblies often cause crashes. Either at the mesh stage or the solver stage. I could live with much longer simulation times if I had a warm feeling that I could eventually get a solution. But as it is, I always end up having to futz with the model to make it simulate. Sometimes I just give up, because I don't have days/weeks to set up the problem and try to make it work.

Other Annoyances:
* the way Simulation handles materials libraries has ALWAYS been annoying, and in Simulation
2009, they broke things (SP4 is supposed to fix)
* meshing assemblies is often hit or miss. maybe this is common to other products.
* feedback from the solver has always been unacceptable. sometimes you'll start a simulation and it will be a
few minutes before the dialog box with the progress bar opens up. Or the simulation will be mostly
complete, the dialog box disappears, and you wait a few minutes before you can see the results.

 

[rstupplebeen]

I have used most of the high priced competition: Abaqus, Ansys, MARC, MSC (Nastran, Patran, Dytran), FeMap, Nx Nastran oh yea and Cosmos. Cosmos is by far the easiest to use however it is limited compared to many of these others. My strong preference is for Abaqus however the price tag may deter you. You will need to contact your local Abaqus office to find the actual trice but $20K per year for 1 seat is a rough estimate. The usability of Abaqus and FeMap's GUI is second only to Cosmos. Abaqus's solver can handle nonlinear problems such as contact large deformations and nonlinear material properties better than any one. Also, they are owned by Dassult Systemes same as SolidWorks so hopefully there will be significant integration in the future. Catia and Abaqus already work well together. Remember these are just my opionions and you should get a 1 month free trial of any software you would consider. If they won't give you that have them come in provide some geometry before hand for them to get a canned demo and also have a pop quiz for them after their demo with another set of geometry. I hope this helps.

Rob Stupplebeen

 

[wpeaton]

Rob,

Thanks for your reply. Do you have any feeling for what products are likely to get to a solution with minimal futzing? I find that even for linear/noncontact problems, I often cannot simulate unless I spend days and days setting up the problem just right. For me, it's usually it's pretty straightforward to figure out how to simplify models for simulation. But even after that, it can take days to get it to mesh or find a solution.

Maybe other products are far more pricey, but maybe they're worth it if I get a solution in the end. I have many examples where I just give up because my boss doesn't want me to futz around with a 3 part assembly for a week.

Lately I've even been looking at Calculix. From a user interface perspective, it's hard to love. But if it gives me an answer, I'm happy.

 

[nodek]

I must add something to what u said. Cosmos or SW Simulation are tools for design engineers. They are simple black boxes and u don't have to know much about it. If u'd like to simulate a real world problems in Abaqus or similar software u will have to learn a lot not only to get a result, but to get a result which fits for the purpose. Basically u will have to become more or less a CAE analyst...

 

[rstupplebeen]

If you explain one of your simulations that you could not get to converge I can try to give the sort of estimate that I would give my manager with around 50% error most of the time. A picture is always helpful.

To nodek's point training will be very important. The real upside is that a good analyst is hard to come by so they are paid much better than the average CAD user so after the training update your resume. I hope this helps.

Rob Stupplebeen

 

[rstupplebeen]

"The real upside is that a good analyst is hard to come by so they are paid much better than the average CAD user so after the training update your resume."

I wanted to clarify before a flame war got started. A good CAD user is also hard to come by. The market in general pays analysts more. So if pay is a significant driver in your career the expensive skills are the ones you want to learn. FEA or CAD for that matter can drive some people nuts though so see if you like it.

Rob Stupplebeen

 

[TheTick]

Are you doing any sort of preconditioning of your assembly geometry? Do you analyze weldments where non of the parts touch because of weld gaps built into design? Are you analyzing motion where there is contact between bearing surfaces that are not quite identical?

 

[wpeaton]

Nodek, It has been my observation that as much as they try to dumb down Cosmos, you still have to know a lot of the underlying theory in the end. I think black boxes only work for really simple geometries and loading conditions like -- er -- boxes!

Having to learn even more about FEA doesn't scare me. What scares me is this nagging feeling that maybe there are better tools out there that -- with the same geometry -- would be me solutions more reliably. I'm not even talking about necessarily getting the right answer yet, but actually getting things to solve. It's been my experience that once you can mesh and solve, then it's a little more clear about how the model can be refined to get answers that are reasonably accurate.

Responding to The Tick, I'm not sure what you mean by preconditioning. I never deal with weldments. But often I'll have to massage geometry to deal with small gaps -- I like to either make them disappear or make them bigger so that they don't make mesher/solver barf. In fact, spending a few hours simplifying a model before analysis is to be expected.

 

[ktop]

Noran Engineering has a very good set of products in the CosmosWorks market space.

NEiWorks is very similar to CosmosWorks in the user interface department, however, it creates a proper Nastran input deck so you can see what you created and you can use it in higher end analysis.

NENastran is the big brother and comes with FEMAP for pre/post processing.

If you go beyond NEiWorks you need to learn a lot more about FEA than the average FEA designer.

KTop