ANSYS vs. ABAQUS
Emilio P. Calius
with our choice for some years. I'd dearly like to get some opinions on
ANSYS 5.0 versus ABAQUS (with the PATRAN3 pre/post-processor) for a Sun
SparcStation 10 from anybody on the net willing to share experiences, or even
some gossip on the subject.
The objective is to perform structural analysis, initially static linear and
non-linear collapse, of mostly thin-walled composite structures.
It would be nice to be able to do quasi-static contact problems and maybe
even transient thermal analysis, in an effective manner.
None of us has used either of these packages. Except for one person, we are
relatively inexperienced in finite-element analysis. However, we'll have
about 6 monts to develop our proficiency.
We'll be creating most of the geometry directly in the FEA package.
Comments would be very welcome about any and all of the following:
1) Creating 3-D solid geometries with holes, fillets, ply drop-offs, etc.
2) Power and convenience of the automatic meshers for surfaces and solids,
whether they're quadrilateral/hexahedral, or more complex.
3) How good are the tools for estimating the error and refining the mesh.
4) How efficient and/or trustworthy are the nonlinear solvers
5) How convenient are the contact elements
6) Whether sub-modelling/sub-structuring is an effective way of looking at
small (approx. 1-2 inch) details in a larger (approx. 100 ft.) structure.
7) Anything else you think is relevant
Thanks in advance for your assistance.
Emilio P. Calius
FRE Composites
cal...@fre.login.qc.ca
(514)537-3311
--
Emilio P. Calius -- FRE Composites Inc. -- (514)537-3311
> structures for Space Station, the Channel Tunnel & anything in-between <
> structures pour Station Spatiale, Tunnel TransManche et ce qui va entre eux <
Pete Roth
>We're considering the purchase of an FEA software package and we'll be stuck
>with our choice for some years. I'd dearly like to get some opinions on
>ANSYS 5.0 versus ABAQUS (with the PATRAN3 pre/post-processor) for a Sun
>SparcStation 10 from anybody on the net willing to share experiences, or even
>some gossip on the subject.
[...]
I would hope that you will not base your decision on what you get off the
net! Since your team (what, say 5 folks?) will be at this for several years
(what, say 5?), that's 25 manyears (sorry ladies) at say $100,000 per, pretty
soon you're talking real money.
Send some guys to a course in each program, see what they find out. It'll
cost you about $5K for this, but you gotta do it.
###
Grace & peace
ro...@oasys.dt.navy.mil alias: Peter N Roth
"On the Internet, nobody knows you're a dog." - P Steiner.
W. Donald Rolph III
>From: ro...@oasys.dt.navy.mil (Pete Roth)
>Subject: Re: ANSYS vs. ABAQUS
>Date: 3 Nov 93 20:35:22 GMT
>[...]
I have been trying to stay out of this one, but I find I really cant.
First - courses are notriously poor vehicles for deciding on a software
package of this scope. You will have no real sense in the short period of
time with a course on how well the package really works for your class of
problems.
My approach has always been:
1) Lay out the classes of problems you are concerned with in your
facility
2) Look at the theory manuals for the algorithms used in your class of
problems (most egregrious case - for example does the laarge
strain formulation for your materials use velocity
strain/Jaumann rate or hencky strain/material stress)
If the formualtion is bad for your class of problems - drop the code
like a hot potato!!
3) Bring in one copy of each on a 90 day trail (now you want to take
the class) and run both of them. I typically bring in a proposed
new code on a one year trial before I will commit to it
Remember all codes have very severe warts (features ;-) ). You want the code
whose warts are least distracting for your specific class of problems.
{Two additional comments:
A) complex nonlinear FEA is a real challenge to move into production
- ABAQUS and ADINA are probably the two most robust and best
developed nonlinear fea codes for complex nonlinearities
B) nonlinear FEA codes can be a real trial if virtually all you work
(>90+%) is linear FEA
}
Emilio P. Calius
> I would hope that you will not base your decision on what you get off the
No, I'm looking for some more hard questions to ask of the suppliers. We've
given them both example problems and they're coming in and showing how they
would solve them.
What I'm trying to identify is issues that might not come to light until after
some time (months or years) is spent using the program, as well as intangibles
such as support & bug fixes.
> Send some guys to a course in each program, see what they find out. It'll
> cost you about $5K for this, but you gotta do it.
It did indeed cross my mind. I did indeed take a course on one of these
products. I feel that with 1 or 2 courses you will not have enough experience,
not enough practice, to really apprehend some of the long-term difficulties
or advantages that you will run into after you've lived with the product for
a good while.
We believe we can learn more by watching how an expert attacks a "typical"
problem.
I'm hoping that there a few experts on the net willing to share their
hard-won expertise with these packages.
Here's hoping...
anthony.r.rizzo
I find the parametric modeling capabilities of ANSYS incredibly powerful.
I'm unaware of another program with similar capabilities. Release 5.0
of ANSYS gives the user a nearly complete programming languange. In
addition, all quantities entered by the user or calculated by the
program can be assigned to a variable and used for further processing
or branching. Release 5.0 also lets users define 1-dimensional and
2-dimensional arrays, which can be used for either input or output
storage and processing. All this is in addition to the full analysis
capabilities of the program, which include non linearities such as
plasticity, large strain, large deflections, contact with friction,
full dynamic capabilities, fluid loading on submerged pipes and
cables, and a great deal more.
I am convinced that during the next 15 years engineers in industry
will wake up to the fact that the use of computer models in
statistically designed experiments is the only way in which finite
element analysis can be used to SIGNIFICANTLY increase the
VALUE ADDED to products.
It's fine to do one or two analyses to "verify" (ARRGGHHH) a design
or to identify the mode of failure after a disaster. But one or
two analyses don't yield a great deal of engineering information.
To extract the full benefit offered by FEA technology, it's crucial
that we use parametric studies. ANSYS lets you do this, and
ABAQUS does not.
But hey! That's just my opinion.
I strongly suggest that you call Swanson Analysis Systems, Inc,
at 412-746-3304 and ask for the distributor nearest you. Then you
can buy an educational version of the program for your 486 PC.
The educational version is a fully capable program, limited only
in the number of elements and in the maximum wavefront size that
it can solve. But it has ALL the regular features of ANSYS.
The price is approximately $190. No sh&t!
Tony Rizzo <--- happy customer.
Emilio P. Calius
a72...@pan.mc.ti.com (W. Donald Rolph III) writes:
> Remember all codes have very severe warts (features ;-)
Which is why I'm trying to get some feedback from the net. With little success
so far. Am I in the wrong group ?
In our case (laminated composites) we expect to almost always remain within
the linear elastic domain. Also our strains will usually be small since carbon
fibers can't take more than 1-2% elongation to fracture. We are concentrating
on large displacements and contact type non-linearities.
> A) complex nonlinear FEA is a real challenge to move into production
> - ABAQUS and ADINA are probably the two most robust and best
> developed nonlinear fea codes for complex nonlinearities
This is indeed a very important issue, but I'm not sure how to quantify it.
Are there publications which compare the effectiveness of these solvers with
ANSYS' ?
How do you rate the tools provided by ABAQUS for the initial meshing, mesh
refinement and estimation of error ? The more automatic the better.
How convenient are geometry modifications and the resultant re-meshing in
ABAQUS ?
I haven't seen ADINA included in any list of packages that can deal with
laminated composites. Is it a strictly isotropic material code ?
> B) nonlinear FEA codes can be a real trial if virtually all you work
> (>90+%) is linear FEA
Nonlinear collapse is expected to be the design driver for several projects
that we are about to start. We expect to be spending at least 40-50% of the
analyses will be of the non-linear variety.
Venkatesh Kannan
: We're considering the purchase of an FEA software package and we'll be stuck
: with our choice for some years. I'd dearly like to get some opinions on
: ANSYS 5.0 versus ABAQUS (with the PATRAN3 pre/post-processor) for a Sun
: SparcStation 10 from anybody on the net willing to share experiences, or even
: some gossip on the subject.
I suggest the following, though may be little different.
Since your folks are new to FEM, I suggest you to keep an option with Cosmos/M
also.
Cosmos/M has good capability for the following.
- It has very good processor for modeling contact problems, the
latest algorithms are available. I think even Ansys has something
like node-node gap elements. Cosmos/M has node to line, node to
surface elements which helps you model quasistatic contact problems
easily.
- Cosmos/M has substructuring, but no submodeling/subregion analysis.
Ansys has both. Abaqus, I dont think so. However, it is only the
preprocessor which you are using.. If patran has this, then well ok.
- Cosmos/M is very user friendly and those with no prior knowledge of
FEM can do it. (We have tried it, by teaching the little about FEM
and making students learn more in using the available paackages, cosmos
is the easiest on can learn).
- Cosmos has excellent preprocessor, called GEOSTAR. It is very good.
Lastly,
if you have a vacancy, please call me. I will be happy to join you !!
I have hands on experience with all these three packages.
Thanks.
Terence T. Lung
>Which is why I'm trying to get some feedback from the net. With little success
>so far. Am I in the wrong group ?
I have experience using ABAQUS on laminated composites; but no experience
with ANSYS so I cannot give you a comparison just yet. Many of my
colleagues here use ANSYS on laminated composites; but none using
both ABAQUS and ANSYS.
>In our case (laminated composites) we expect to almost always remain within
>the linear elastic domain. Also our strains will usually be small since carbon
>fibers can't take more than 1-2% elongation to fracture. We are concentrating
>on large displacements and contact type non-linearities.
>
>> A) complex nonlinear FEA is a real challenge to move into production
>> - ABAQUS and ADINA are probably the two most robust and best
>> developed nonlinear fea codes for complex nonlinearities
I've seen ADINA used with some popularity, especially for 'dynamic'
applications.
>How do you rate the tools provided by ABAQUS for the initial meshing, mesh
>refinement and estimation of error ? The more automatic the better.
I don't like ABAQUS' meshing; but after having used PATRAN quite a bit.
The guys around here, who use ANSYS, seem to often use MENTAT to do
their meshing.
>How convenient are geometry modifications and the resultant re-meshing in
>ABAQUS ?
IMO, very inconvenient. I keep my meshes and analyses very very simple
now that I no longer have PATRAN at my disposal. PATRAN (2.5 at the time)
made life much much easier (with SGI workstation).
>I haven't seen ADINA included in any list of packages that can deal with
>laminated composites. Is it a strictly isotropic material code ?
>
>> B) nonlinear FEA codes can be a real trial if virtually all you work
>> (>90+%) is linear FEA
>Nonlinear collapse is expected to be the design driver for several projects
>that we are about to start. We expect to be spending at least 40-50% of the
>analyses will be of the non-linear variety.
I used ABAQUS in some elastoplastic analyses (isotropic material in that
work) and thought it was very useful. I've also used it in dynamical
analyses (isotropic and laminated composite); and stability analyses
and thought it still very useful. I give ABAQUS a thumbs up for nonlinear work.
I might note that their support staff was very helpful as well.
Terence