Improve FEA Skills in SolidWorks | SimulationXpress (2011)

[Curl]

I want to learn/practice using Solidworks' FEA tools. My version only has the SimulationXpress (2011) package which seems weak.

I want to know if it is possible to specify roller/slider/pin conditions when using the structural analysis tools. So far I've only seen fixture condition, so is this the only option or am I not looking hard enough?

 

[xxChrisxx]

No idea what SimExpress is, but if its just the new name for CosmosExpress, you'll have a hard time finding anything useful.

It's designed for nubs to use, so they can press the play button and get a semi decent answer. I've found it inadequate for almost everything I've ever done.

 

[Curl]

Sounds right...

So is it worth going into the pro/premium versions or better to just ditch SW and learn Comsol?

 

[Mech_Engineer]

I found Solidworks' built-in FEA capabilities to be sub-par compared to any decent off-the-shelf FEA package. The meshing, solver, and boundary condition controls are all way behind ANSYS (which is what we use for FEA work).

 

[xxChrisxx]

FEA is FEA is FEA.

If you know the theory and what you want to achieve, actually doing it is only a click of the F1 key away. The actual package used is largely irrelevent, as they only differ slightly for basic problems.

 

[Mech_Engineer]

FEA is FEA is FEA.

If you know the theory and what you want to achieve, actually doing it is only a click of the F1 key away. The actual package used is largely irrelevent, as they only differ slightly for basic problems.

This is absolutely not true, especially with the package we're talking about (http://www.solidworks.com/sw/products/solidworks-simulation.htm"). Mesh quality, boundary condition options, and solver choices are largely software dependent; the simple FEA in Solidworks is way behind anything you can do in ANSYS Multiphysics or Comsol. Doing a simple mehcanical model of a single part works ok, but once mesh density goes up and parts have nonlinear contact conditions, you're out of luck.

If you want to learn an "industry standard," learn ANSYS in an FEA class. No one's using Solidworks Simulation to teach people FEA.

 

[Curl]

Is ANSYS that much better than Comsol? The version we have here looks garbage, while Comsol we have 4.2 and seems much more complete. Is it okay to learn comsol or is it worth to switch to ANSYS?

I don't care about industry standards, I just want it for myself but I don't want garbage results.

 

[Astronuc]

FEA is FEA is FEA.

If you know the theory and what you want to achieve, actually doing it is only a click of the F1 key away. The actual package used is largely irrelevent, as they only differ slightly for basic problems.

Not so. Some FEA systems are better than others, and some are plain wrong, while others are correct, or perhaps more correct.

We develop our own proprietary methods, particularly non-linear systems and for mechanics/dynamics with large strains and high strain rates.

I found Solidworks' built-in FEA capabilities to be sub-par compared to any decent off-the-shelf FEA package. The meshing, solver, and boundary condition controls are all way behind ANSYS (which is what we use for FEA work).

That is our experience.

 

[Mech_Engineer]

Is ANSYS that much better than Comsol? The version we have here looks garbage, while Comsol we have 4.2 and seems much more complete. Is it okay to learn comsol or is it worth to switch to ANSYS?

I don't care about industry standards, I just want it for myself but I don't want garbage results.

It's hard to quantify what is "much better," both are powerful tools in the right hands and complete garbage in the wrong ones. Which version of ANSYS do you have? I have found ANSYS's technical support and documentation to be top-notch, and its solvers and meshing are second to none. Visualization-wise it might be a half-step behind COMSOL, but making a picture prettier doesn't make it more correct if you catch my meaning.

 

[xxChrisxx]

Is ANSYS that much better than Comsol? The version we have here looks garbage, while Comsol we have 4.2 and seems much more complete. Is it okay to learn comsol or is it worth to switch to ANSYS?

I don't care about industry standards, I just want it for myself but I don't want garbage results.

ANSYS is a beast of a programme, Workbench combines user friendliness with top quality solvers. I find myself keep going back to Ansys classic though. Also learn to write batch files, they are a godsend.

However everyone seems to be hung up on learning to use a single programme. We never learned like that at university. FEA was all very theoretically taught and it's all basically the same. Discretise, create matrix, solve.

With a bit of reading and playing you can find individual solvers strengths and weaknesses. For example we always used Ansys for linear and model analysis and Abaqus for non linear, dynamic. The solvers work differently, but no so differently that you can't pick it up.

On saying that I never really do any taxing FEA, it's all just basic linear structural. So my opinion of it is skewed somewhat. It's cheaper to subcontract the comploicated FEA work to a consultancy.

As an analogy, if you learn how a Ford engine works, you will have a fairly good idea how any engine works, as it's all basically the same.EDIT:
Regarding garbage results. The specific programme isn't going to get you good or bad results. Good modelling and simulation practises are.

 

[Curl]

However everyone seems to be hung up on learning to use a single programme. We never learned like that at university. FEA was all very theoretically taught and it's all basically the same.

That's because learning new UIs is annoying and takes time which I don't have. Plus, FEA class from the ME department is a disgrace (just like any other class in the department). If I want to learn FEM I'd take a course from applied mathematics or study the course notes myself; but again no time.

 

[xxChrisxx]

If you want to learn a user interface that will be most useful to you, learn Ansys. If you already know a decent amount of Comsol, stick with that.

It's really not the right way to learn to use FEA though. As it boils down a problem into the robotic clicking of buttons you are used to seeing. And you'll end up being like one of the annoying people who hands me a report filled with utter crap becuase of it.

Also, you don't need to extensively learn the maths behind it unless you will eventually write bespoke code. Learning good modelling techniques and practises are most important to an off the shelf package proficiently.

 

[AlephZero]

FEA was all very theoretically taught and it's all basically the same. Discretise, create matrix, solve.

That suggests you weren't taught anythnng about creating practical (i.e. efficient and accurate) FE software - not for structural analysis, anyway.

Go and read 100 papers in peer reviewed journals about formulations curved shell elements. for example. (Any decent search system will find WAY more than 100 papers on that topic to choose from). Then decide for yourself if "it's all basically the same".

Regarding garbage results. The specific programme isn't going to get you good or bad results. Good modelling and simulation practises are.

That's a self-contradictory statement, since "good modelling and simulation practice" includes making an objective choice of what analysis software to use for the simulation.

Just because it's called NASTRAN, or ANSYS, or ABAQUS, or any other well known brand name, doesn't mean it's 100% right. They all contain things that are just plain stupid, even completely wrong. Last time I looked at the list of known bugs in one of the above-named programs (and the list is freely available to any customer who wants to see it) it contained more than THREE THOUSAND entries.

Actually, publishing a list of >3000 bugs is GOOD for quality, not bad. I wouldn't use any software from a company that wasn't honest about its own fallibility.