Best FEA software 1

[stw]

thread727-123750

I came across this post when googling "best fea software" and was quite interested in the comments. Particularly this one:

"The complexities of doing this are enormous. You would just get a garbled mixture of results. Also, you mention "better" a lot, without clearly defining exactly what "better" or "best" means, which leaves it open to interpretation by the user of each software. What defines the "best" result? That which matches closest to someone elses model? Someone's PhD thesis? Experimental evidence/data? Of course, the latter, always. How about access to such data?"

I think what happens when you use modelling software for too long is that you lose sight of what it is that you are actually doing. The answer to the question of which software is "best" is very clear and straight forward. It is the software that is the most predicatively accurate with regard to reality. So the best FEA software is the one that will predict the outcome of a given scenario with great accuracy.

In my office we are currently looking to buy a relatively inexpensive FEA capable modelling package. We design and manufacture steel frames, piping, supports, jigs etc and want to check our designs, initially at least, for stress loads and distortion and to use the software to predict those loads and then optimize the design to a safe working factor. We are looking at Inventor Professional and also Solidworks. We are building a large mechanical, hydraulically actuated, vice and are modelling it with the two packages to compare results. We will then gradually test the frame to destruction by hydraulically loading it and measuring the bending and deformation at set loads. We know what the material and weld properties are so we regard it as a controlled test. We will then compare the results from the test with the simulations and determine the most accurate FEA software for use in that domain (the domain that we will use the software model in). If we begin using the software in a new domain then we will assess its predictive capability and its accuracy in that domain too. It is very important to verify the validity of the modelling output in my opinion.

I am new to FEA modelling but I have done a lot of hydro-code modelling of blast and explosions where the strain rates are enormous and the materials response is non-linear but what I have found is that a lot of the accuracy is in the quality of the materials models.

Just one last point, if the software leaves the result "open to interpretation" then it wouldn't be very good software in my book.

 

[GregLocock]

" So the best FEA software is the one that will predict the outcome of a given scenario with great accuracy"

Shrugs, scarcely a useful definition.

Suppose one is 99% accurate (ie deflections and stresses within 1%) and the next contender is free, and easy to use, and is 98% accurate?





Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[corus]

The theory behind FE software is generally consistent between all of them, so one won't be more 'accurate' than another. Some FE packages handle non-linear analysis better than others, however, and some FE packages are more user friendly than others. But at the end of the day it's the person who constructs the model, who decides what assumptions to make (within the capabilities of the software), and who will ultimately have more bearing on the accuracy of any results than the software itself.

 

[ThomasH]

stw

This type of question comes around every now and then. I would say it depends on your application. There is no single software that is "the best" for everything. Read the tread you used as reference.

If I understand your application correct it is static analysis of models that can be built from beam och plate element. There is several options that can do that. If you need nonlinear it depends om material models but there is still several options.

But remember, the software is a tool. It does not do the analysis. It just does the number crunching. If you don't know what you are doing you can mess up any analysis. But if you know what you are doing any software with a reasonable track record will probably help you obtain accurate results.

Good Luck

Thomas

 

[IRstuff]

as the old saw goes, "garbage in, garbage out." Given the variability of material properties, and the vagaries of meshing, etc., "most accurate" seems to be an over-reach. "Good enough" should always be a major consideration.

To take Greg's example to the absurd, would you buy a package that is 99.99% accurate, because it's the "most accurate," even though you have variabilities elsewhere in the problem set on the order of 10%, or more?

Moreover, design factors are typically in the multiples of 10%, i.e., 50% or 100%, so is something that is 99.99% accurate even meaningful?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[rb1957]

"best" depends on what you value, there is no universal best.

you might choose to spend a lot of money to get an easy interface and maybe lots of support.

you might choose to go cheap and put in time to get things to work the way you want.

"best" for you might be to easily communicate to your CAD program.

hopefully you won't choose a program that gives crappy results.

Quando Omni Flunkus Moritati

 

[IceBreakerSours]

I would think of, at least, the following before purchasing an FEA package:

Pre-processing
Analysis
Post-processing
Documentation
Usability (convenient user-interface, perhaps)
Licensing/Cost structure
User-base
Support

Now, say, if you wanted to use a serious multiscale material model, most FEA packages (the ones I am aware of) are lacking in them which renders the "best FEA software" phrase pointless. [Most major developers happen to have tie-ups with companies that offer such material models but you'll end up paying more.]

Are you new to this forum? If so, please read these FAQ:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[jhardy1]

The previous thread was closed for a very good reason - it is a futile debate!

To the OP:

Do you have any experienced FEA users in your organisation? If so, ask them what software they recommend.

Define what you want the software to do (linear static, non-linear, buckling, post-buckling, natural frequencies, transient response, multi-physics, design to steel / concrete / piping / pressure vessel codes, interface / integration with CAD software, etc, etc, etc). Which of these are "must-have", and which are "nice-to-have" or possible future needs? Cross off any FEA software which is missing any of the must-have functionality you are looking for.

Test some of the packages that meet an acceptable fraction of the functionality you seek.

Ease of use, and support from the vendor, can be very important - especially if your organisation does not have a team of experienced FEA specialists.

"Bang-per-buck" probably also need to be considered (unless you are lucky enough to have an unlimited budget).

Depending on the industry you work in, and the class of problems you need to solve, you will probably find free or very low-cost FEA software that can do at least 80% or more of what you want (e.g. Calculix, Elmer, Lisa, etc), low-cost commercial FEA software that can do 90% of what you want for a few thousand dollars per seat (e.g. Strand7), or higher-cost commercial software that might cost anything from $10k to $30k or more that can do 99% of what you want (e.g. Femap / NX-Nastran, Ansys, Abaqus, etc). All of these options can get the standard classical test problems "right", as they all use the same underling theory, but some of them may be unsuitable for tackling "large" or "complex" problems.

Which is the "best"? Only you can decide!

http://julianh72.blogspot.com

 

[petb]

Hi,

First invest time and money in understanding finite elements or hire an experienced analyst. Without that, FE is useless and potentially dangerous. The software is solving the posed problem it is given by the user, nothing more, nothing less. What you said about "results left open for interpretation", is in my opinion the opposite, it is the essence of what a structural analyst should be doing. Judging the outcome of a mathematical simplification of the real world problem at hand, where he/she knows what limitations his/her model has. I've seen far too many gigantic finite "elephant"-models being reduced to what effectively becomes a 500 kDOF cantilever beam, with no physical meaning whatsoever. In my opionon you can use any FEA software that you feel comfortable with.

This is in no way meant to be rude, english not even my second langage, so sorry if my wordings where blunt.

Just my 0.02 SEK...

 

[MacEd]

Perhaps this is a bit obvious but a huge factor in accuracy is the mesh density yet you can't simply control for that to compare software. One may converge faster to a less accurate solution, while another converges more slowly to a more accurate solution. I've seen this happen for different element formulations within the same software. If the slower one has adaptive refinement and the other requires manual restarting, is it really slower? If the difference in accuracy is small compared to the error in your estimates of loading, do you really care? This is more relevant for shell elements than solids, since there are so many different formulations with different idealizations. Solid elements modelling the same physics should give the same results in any software after you're ensured the mesh is fine enough. If they don't, then one of them is probably broken.

As others have said, usability and feature set are probably more important than some hard-to-define concept of accuracy. Your description sounds like a fairly simple static problem that any software could do just as well. It's when you get to more obscure features that different physical models are used, and then you really need to understand what assumptions go into the model to make a judgement.



FEA software feature chart

http://feacompare.com