Finite Element Analysis Software

[Casper]

Good Day to Everyone,

I'm planning to buy a FEA software and I would like to ask some of you who have used this kind of programs, which one is the best for the specific application of boatbuilding materials i.e. FRP, Aluminium and Steel, does this software is applicable for all of the above mentioned? Can we consider the CATIA that has a FEA capability?

I will appreciate any recommendations.

Thanks,
CAS

[PI Design]

All FE packages can handle homogeneous materials like ali and steel. Any package worth its salt should also be able to cope with FRP (which has different material properties in different directions), but determining failure modes for FRP structures is more tricky.
We use ANSYS, which works well. There are several different ANSYS licenses available, but Workbench is the easiest to use if you have little previous experience. CATIA would be fine too, I imagine.

PS ANSYS was used by the winning Alinghi America's Cup team.

[MikeJohns]

Casper
Watch the level of embedded FEA/FEM packages some are basic linear and you will be best serviced with a non-linear capable package.

[NeiNastran]

I recommend Nei Nastran. You can view more marine applications of FEA here:
http://www.nenastran.com/newnoran/consultingMarine

[sami]

G'day CAS,

I think I have a solution to your question. Just to let you know that we just opened a company in the Middle East (Riad) and I am the selling person the UAE. Our company is mandated to sell ANSYS license (most FEA tools convenient for your application) and more than that, we do provide training and technical assistance on how make good use of the S/W. Can you please send me your personnel contact so we can have more talk on this?

Tx

Sami

Originally Posted by Casper

Good Day to Everyone,

I'm planning to buy a FEA software and I would like to ask some of you who have used this kind of programs, which one is the best for the specific application of boatbuilding materials i.e. FRP, Aluminium and Steel, does this software is applicable for all of the above mentioned? Can we consider the CATIA that has a FEA capability?

I will appreciate any recommendations.

Thanks,
CAS

[alidesigner]

I recommend Nastran. I have used it for years for aluminium boat FEA. A 50m monohull has a 13 second solve time and its easy to learn.

Alos there is a great book at amazon that uses Nastran in its examples.
Here is a link to the page at amazon http://tinyurl.com/FEA-Book

[LyndonJ]

Originally Posted by alidesigner

I recommend Nastran. ..... A 50m monohull has a 13 second solve time and its easy to learn....

Thats probably a little misleading without showing the analysis you actually did.
13 seconds if it's very simple but be prepared for a long wait if you do a more detailed analysis.
For example we tried Nastran when we did a 5m by 8m hydro gate simple curved surface steel plate on a frame and only a static pressure load on the face.
It took over 6 hours to run the analysis on a 3GHz PC.

If you really modelled the whole boat and applied proper restraint and loads I think those 13 seconds would very quickly multiply.

Cosmosworks is quite good. Strand/straus is good Ansys is good Algor is not so intuitve , Catia I don't find intuitive either.
A really good test is to import some solid from a CAD package define restraints, loads and run within a few minutes. If you can do this you'll use it much more as a design tool.

[alidesigner]

Not misleading at all. It was a full boat modelled with symetry constraints, 6 load cases using shell and offset beam elements. 6 years ago Nastran 4.5 would solve a similar model in 6 minutes. I cant give you you node and element numbers as I am not at work until next year but I can assure you this is no small model. From start to finish it took 3 weeks to model, evaluate, modify and confirm the design requirements.

If you are getting long run times in Nastran you are doing something wrong.

For those who are serious about FEA the only 2 major players are ansys and nastran and ansys is more expensive and really needs ProE behind it.

Nastran is the best stand alone FEA that I have found.

[MikeJohns]

Originally Posted by alidesigner

Not misleading at all. It was a full boat modelled with symetry constraints, 6 load cases using shell and offset beam elements. 6 years ago Nastran 4.5 would solve a similar model in 6 minutes. I cant give you you node and element numbers as I am not at work until next year but I can assure you this is no small model. From start to finish it took 3 weeks to model, evaluate, modify and confirm the design requirements.

If you are getting long run times in Nastran you are doing something wrong.

For those who are serious about FEA the only 2 major players are ansys and nastran and ansys is more expensive and really needs ProE behind it.

Nastran is the best stand alone FEA that I have found.

Interersting that this popped up and it's worth a comment I think it's often a case of "My wife is the best wife that I am currently married to"
I'd break any review down into the following


Ease of importing models
Flexibility of assigning loads and restraints
Element types supported
Libraries
Meshing ability
Speed
Results tools

As for speed If the model you show took only minutes then the mesh is very coarse since there is no magic bullet when it comes to FEA.

Try redefining your mesh so that the minimum size is one third the minimum material thickness and then do a buckling ( non-linear) analysis and I would bet that model will take close to 3 days to run. Then you really find out which is the better package and which have poor memory management (and which compilers have bugs when it comes to memory management).

You need to do a non-linear analysis to give a buckling indication and you would need to break that large model down considerably into smaller sections to determine max material stress in practical use. Or get run time on a mainframe computer.

[Ad Hoc]

I was asked the very same Question recently, i summarised as thus:

FEA is a great tool, but it is just a tool. They can be very very complex to understand and learn. If you have sufficient structural/engineering theory/background, this will aid you enormously. Since You must always question the results and attempt to validate in any way you can.

Usually by making small simple models and checking against known results from theory.

Having said all that, the most "popular" FEA codes are all validated. They must conform to NAFEMS and produce consistent results, like the "patch test", which is a way of checking how the program works and what errors there are.

I use and selected COSMOS many many years ago. The principal reason for selecting remains the same today. Ease of use.

You need an FE that is easy to model in, like a simple CAD package. Everyone tells you, ahh..no need, model it in your Cad or Rhino etc. This is taking away "control" of how the FE works if you do this. I do all my modelling in the FE program. I can control the way the program works and more importantly the meshing. Since when you ahve a problem, how easy can you correct it, by modelling 'inside' the FE program?

So, my advice to you is to try around ~5 different FE programs, test them see how easy it is to model:
1) a simple shape
2) complex

You probably wont be able to so, But, tell the "salesman" to draw 1) and then 2) in the program using the modelling function. Have your sketches of 1 & 2 in advance but don't tell them. How they react will be a true measure of its ability and the salesman too.


Additionally I concur with M.Johns.

Also processing time....personally i don't care if the time it takes is 10second or 10 hours, so long as I can manipulate and alter the model easily. If 10 hours, that is a thousand times quicker than i would be able to do by hand....so I'm not fussed about processing time.

Alidesigner will know, i used cosmos on a 386 PC years ago. The time to process some of the more detailed models some 15 years ago took 10+hours...now the same model i can run in literally 5 seconds. The way the model is created is where the time is lost, especially in meshing and applying constraints etc and of course, the results and how they are presented.