What doesn't ABAQUS do well 2

[gardoggie]

I am new to the forums, and am trying to figure out what FEA package to learn. I am thinking about starting my own one man consulting firm to do structural/machine parts FEA. I am an old engineer that wants to do this as a semi-retirement type of thing to bring in some monthly income. I learned FEA years ago and the packages I knew about have changed from back then. I haven't done FEA in a while.

I have looked through the other posts and don't want to rehash what FEA package is "best" as I know everyone has a different opinion. It appears that Abaqus is definitely a good package, but expensive. I thought I would take a different approach to a general question and ask what types of problems Abaqus would not be good to use for. What are Abaqus's weakneses?

Thanks for any help on this...

 

[rb1957]

if you're doing vanilla stress analysis, 99.99% of FE codes use the same solvers ... the biggest difference is the GUI, pick one you're comfortable with.

the more exotic analysis you want to do, the more time and money you'll need to invest.

look for local salemen and get a trial license ... if they won't give you one, move on to the next vendor.

personally, i've found FeMap to be pretty good, and easy to pick up. forget Patran ... IMHO way too clunky.

good luck !

 

[rstupplebeen]

Price and FSI. Abaqus is my favorite by far but the cost makes a 1 man show difficult especially in a part time setting. Depending on your license it will probably run $25K per year and that is a lot off the top. I would probably pose the question as what is the easiest and cheapest software that can do what I need. I hope this helps.

Rob Stupplebeen

 

[FeX32]

I would recommend Comsol Multiphysics. It's relatively new and a lot cheaper. It's GUI is good for a beginner and you can even write your own subset of codes within. It also is great to link to Matlab.
Oh, and it is very powerful.


Fe

 

[johnhors]

If you envisage doing a lot of complex analysis which requires a package like Abaqus with continuity of work from job to job (no breaks in between), then perhaps the cost could be justified and you can factor in the cost of the license fee into your quotes at a reasonable and acceptable level.

However that is a very big IF !

You would need a lot of high end jobs coming in to make economic sense. But as a one man band semi-retired consultant I just don't see this happening. In all probability you would get jobs that could be solved with hand calculations or linear FEA for which Abaqus is total over-kill. I know of some very well established consultancies that struggle to pay for Abaqus and negotiate monthly licenses that get switched on only when a job comes in that specifically requires Abaqus or the complexity justifies its use.

As you stated "to bring in some monthly income" then realistically you should first consider some of the plentiful cheaper alternatives or free-ware like CalculiX (which I'm sure would be capable of doing over 90% of any FEA work you might get). Later once you get established and have a constant stream of work coming in, you can set your sights higher. Otherwise you could be getting yourself into a financial disaster!

 

[rb1957]

then you could ask for a bail-out ... nah, nothing we'd do would be so epically bad as what the professionals can achieve

 

[corus]

To get back to the original question, and what can't Abaqus do? It's not a package that can assess the results for you (as some products claim) and not a package that has material properties built in. I don't think it can handle fatigue type problems easily where you need to run multiple cases of varying loads and combine the results to give a fatigue life, of sorts. It also can't do 'random' crack propogation well either, though I think a new version is out that can. It also can't do cavity radiation in temperature-displacement models either, say where you have a spinning object. There's probably a way round that though, I've not figured out yet.

As others have indicated though you need to know your market before you set off on your own. My guess is that 90% of it is quite ordinary structural type work, or perhaps thermal analyses that need to be done quickly and cheaply. Probably the type of analysis where you read in a CAD file, click a button to mesh, and bingo, look at the biggest stress.

Calculix might be a good free program to use, but I'd be wary of using anything that appeared to have no QA that proprietary codes claim. Companies would be put off trusting any results from such a code, I'd think.

corus

 

[johnhors]

Just to pick up on Corus's point, I've done a lot of comparisons between various FE solvers. For given meshes and boundary conditions for 3D solids, all codes (including CalculiX) produce identical results for linear, thermal and modal analyses that I have tested. Contact can be quite troublesome in many codes (including Abaqus !), but I've managed to get results from various codes to match within 5% to 10% for non-linear contact. Many people in industry use and respect CalculiX (especially for 3D solid analysis).

 

[gardoggie]

Thanks everyone for your input. It all makes sense. I'll look into your suggestions. I'm still open for any more feedback.

 

[Jordonlaw]

You need to benchmark your choice of solvers against an application that your will be analysing. I have been told so many times that FEA Package A will do, say, non-linear contact with friction, but when you give them a realistic job it will just fall down.

My last benchmark activity included Abaqus. The sales guys came in with a great presentation and some excellent GUI operators but when we gave them a bling analysis to do it just came back (twice !) so far out that I and my collegues were shocked. Their pricing too is almost the highest out there and in the UK at least it it is impractical to purchase it.

If you are doing linear stuff then as others have said most solvers will be fine. If you want non-linear and, in my experience, complex contact then you need to set a realstic benchmark for them to solve before you hand over any money.

 

[johnhors]

Jordonlaw , it would be interesting if you could share with us some information on what your benchmark problem was.

In my experience with several (no make that many!) FEA vendors the guys doing the sales presentations are practically always the guys who know the least about their product! Their presentations are well scripted with canned models and demonstrations designed to dazzle, that never fail (apart from one very very well known vendor whose software was so unstable it constantly fell over - some of you can guess who I mean!). The sales guys tend to be the least technically capable, yet they are told to make astounding claims. I would suggest the possibility that in the hands of a very experienced operator the outcome of your benchmark may have been different. I know a number of very experienced Abaqus users, but none of them use the Abaqus pre-processor!

 

[rstupplebeen]

Abaqus's preprocessor in my opinion is on par with FeMap and superior to Patran, and Ansys. I have found that a bunch of the experienced users got used to a preprocessor before Abaqus had one or before it was any good. If they gave it a try now they might switch.

Abaqus is the only vendor that I have spoken with that has actual users doing the sales calls. Ansys and MSC were jokes behind the wheel.

I hope this helps.

Rob Stupplebeen

 

[johnhors]

Rob, you're very closely located to Abaqus HQ, would you expect anything less?

 

[rstupplebeen]

I am even closer to Ansys and have been in 2 different Abaqus regions (East and Erie).

Rob Stupplebeen

 

[Jordonlaw]

johnhors,

Can't be too specific I'm afraid but my benchmark problem involved contact of hundreds of parts which itself causes most FEA solvers to fail. For example, we have had many individuals claim that you can do it Ansys (implicit) but the latest one, a UK help desk veteran, couldn't get it to converge at all.

Abaqus was a surprise. The analyst has done similar sructures in the past and the models ran well with seemingly sensible results when when we compared the numbers they were well out, ie predicting yield in all components whereas the test data showed everything below yield. Material data and geometry was common across all solvers by the way.

In comparison other codes work just fine particularly MSC MARC who in comparison with what rstupplebeen found were and are excellent in my opinion.

CHE

 

[rstupplebeen]

I have no doubt that a problem involving a great deal of contact could be solved Ansys, MARC or Abaqus with enough treaking. My statement regarding MSC.MARC was about the usability of the GUI not the solver. Can you post a CAD model or a picture of your problem? Thanks.

Rob Stupplebeen

 

[rstupplebeen]

SP: tweaking not treaking

Rob Stupplebeen

 

[XFEM]

Please try the Eigen modes problem validation, especially from FEMAP shell and bending plate and compare with others.