FEA software developers guide/forum/book/community?

[bongirs]

Hello all,

I am a Mechanical engineering graduate student with a couple of years of professional experience in FEA, as well as my research project is in FEA. It is my personal desire to develop a basic FEA software of my own. I have the software development skills, but do not know the mathematics required for this. Although I studied variational and galerkin methods, but those were specific cases where differential equations were given. I would like to be directed to an appropriate community/forum/book or anything where I can get the required help.

 

[aerostress82]

Colorado University has a really good FEA lecture notes on one of their professor's website. I don't remember the exact link right now but let me know if you can't find it over the weekend: i'll look it up for you.

As you are aiming at a basic level FEA, an understanding of a 1D 3 rod elements structure with a force applied at the one of the ends and constraint at the other end would give you a really good kickstart on this. You can use Mathcad/Matlab for this with a student license.

If you want to aim more diverse models, you might wanna look into "penalty method" where you will have an understanding of a simple GAP element application.

I think you can acquire a good level into the topic with 5-6 months' commitment. But as you know, your biggest problem will be on Gauss Elimination on your mathematical models - which you can easily work out for small models.

After going through the whole Colorado University lectures and more into the above applications as a beginning, you should be able to have better direction on this yourself. I would advise you to learn coding into already built FEA codes through C+, Python (or Fortran for Patran). It will give you more advantage after your familirization with some simple basic FEA algorithm setup with the above study. But try not to waste more than a year building FEA codes. It really is already found.

If you want to do something perfectly revolutionary, try to build not FEA but FNA (finite node analysis-yes I made the name up) in the long run for the next couple of years. That's where innovation will be when our quantum computers come in play..

Spaceship!!
Aerospace Engineer, M.Sc. / Aircraft Stress Engineer

 

[bongirs]

@aerostress82 Thank you for your valuable advice!
I found the following 3 courses from University of Colorado at Boulder:
1.

http://www.colorado.edu/engineering/CAS/courses.d/IFEM.d/Home.html

2.

http://www.colorado.edu/engineering/cas/courses.d/NFEM.d/Home.html

3.

http://www.colorado.edu/engineering/CAS/courses.d/MFEMD.d/Home.html

On first glance, it seems that the course notes are exactly I was looking for. But there are no video lectures though...

I also found two video courses at MIT OCW by Prof. Bathe who developed the ADINA software:

http://ocw.mit.edu/resources/res-2-...edures-for-solids-and-structures-spring-2010/

A quick Google search shows some useful articles on the penalty method. Perhaps it would be useful when I first understand the basics.

Thanks for the advice again about open source codes. I was reluctant to pursue that path as usually these are developed by a closed community with a specific aim. So they won't entertain an outsider and give the detailed documentation.

Didn't quite get about the FNA part, but I am not innovating in the mathematics. I want to use the same mathematics that's used in software's like ANSYS, HyperMesh, or even the CAD packages FEA modules. The reason I want to develop this software because I am sick of these established vendors not creating user friendly software. They just acquire newer startups and "integrate" that code into that software. Learning HyperMesh was a big headache. Also, I want a totally different business model, hint "Cloud"...
I don't want much profits, I just want to develop a simple software that product designers can use, not aiming FEA analysts at all. Its more about helping humanity...

 

[GregLocock]

Hmm, if you want to help humanity then I suggest you develop a package that FORCES the user to correlate their model. Fat chance.



Cheers

Greg Locock


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

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

 

[bongirs]

Greg, it is always a pleasure to see your comments. Didn't quite get what you meant though. A tool is only as good as its user, if that's what you were suggesting. ;-) But jokes aside, can you give me any guidance?

 

[ShellsPlatesMeshes]

The best way to go would be to obtain working FEA codes capable of solving some real-life problems,together with detailed description of how those codes work. That would serve as a framework for your own software development. Below is a reference to a book that contains all that: theory, detailed descriptions of the algorithms, and the codes themselves:

http://members.ozemail.com.au/~comecau/quad_shell.htm

It is available from the Amazon.com:

https://www.amazon.com/Computational-Geometry-Surfaces-Application-Analysis/dp/0646930818

Hope this helps.

 

[bongirs]

Thank you, it is quite expensive though.

 

[ShellsPlatesMeshes]

Usually, when you buy a textbook, it is just a book with lots of formulas, etc., but with very few really working source codes (if any at all). However, the book that I advised you to consider also contains all the source codes on CD-ROM (and the pre-compiled executables, GUI interface, etc.), which is essentially what you asked for.