Learning The FE Method - Any Advice? 5

[KnacKeN]

I plan to attend the NAFEMS Practical FE course in March. Does anyone know if its any good?

I'm a beginner with some experience but I really want to understand the basics of FE and not just how to use a piece of software (ie Femap Patran etc). I have a few books which have helped get me started, but I was wondering if anyone has any tips on how they got started. Id like to learn as much as I can on my own before going on the course, just so that I dont get overwhelmed...

Also, ideally I'd like some simple FE software to allow set up some basic models and try and understand the results, something I can compare against traditional hand calcs. Does anyone have any suggestions? By free I mean freeware or something, I don't do cracked commercial software.

 

[rb1957]

congratulations on your approach !

before you start with simple models (i'd start with beams, then shells) model things you can solve be hand (including redundant problems) ...
i'd recommend running single element "patch" tests. simple restraints, single loads, how does the elemnt compare with classical analysis. this teaches you the weaknesses (and strengths) of different elements, when to use them, how to use them, to get good results, when to be carefull of what the "lying little buggers" are telling you !

there are plenty of free software out there for doing what you want, NASTRAN has a demo version. whilst most codes use very similar elements, the devil (as always) is in the details, and each code will have it's own subtle differences (in NASTRAN, don't ever use AUTOSPC !)

get some real help too, at least a text book, maybe a post-grad course too; i'm not really a fan of seminar courses but if you go prepared then you won't be blown out of the water on the fist day.

good luck

 

[johnhors]

KnacKeN, I agree with rb, your approach is refreshing.

A single element test will certainly highlight its limitations, but the strength of the method is that with more elements the results improve and as you further increase mesh density you should see your results converging on the analytical result (for simple problems with an analytical result of course - like a hole in a plate). This process is commonly referred to as "mesh convergence".

A great freeware FE software is CalculiX, there are both windows and linux versions available.

http://www.Roshaz.com

 

[DanStro]

I second CalculiX. It should also a big help that it is open source so you can go through the code to see exactly how it is coming to its answers.

And it doesn't hurt that the input files are almost exactly the same as Abaqus.

 

[rstupplebeen]

I would use a scaled down version of a commercial code. In the end the real value of learning the FE trade is to obtain a job. When I review resumes I place a strong preference for a code that we have in house, then other commercial codes I know and finally for free codes.

As mentioned by rb1957 NASTRAN has a demo version
Abaqus has a $99 student edition:

http://www.simulia.com/academics/purchase.html

In the end it is absolutely great that you are building a strong foundation. Make sure that your resume shows this rigor. Also, show some advanced skills that places need. I specifically look for complex contact, hyperelastic materials, and strains in excess of 100%.

Rob Stupplebeen

 

[johnhors]

.....so you can go through the code to see exactly how it is coming to its answers.

- Just a slightly steep learning curve there Dan!

http://www.Roshaz.com

 

[DanStro]

Can you buy the student editions if you are just trying to learn the software but aren't enrolled in a school?


johnhors- If you don't mind looking through code and take baby steps it's really a good way to get a handle on not only the analysis part but also the limitations. But it can be a decent amount of work and take a little while.

 

[KnacKeN]

I will look into CalculiX.

Bear in mind I am just trying to get as much backround learning now before going on the course, after which hopefully I will have a better understanding. Then perhaps after that I can go into more depth and look into how this method is used in industry.

 

[GregLocock]

also see thread727-258834

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[gwolf2]

Given a choice I would choose the student version of ABAQUS to learn because the manuals are outstanding and the input decks are very well structured.

NASTRAN is better to study if you want a job because more people use NASTRAN but it is a bit of a dog to use, the input deck formats are user-hostile and there are quite a few wierd tricks like AUTOSPC, KROT6 etc ready to trip up the beginner.

A good place to start is modelling a cantilever beam in a) beam elements, b) shell elements and c) solid elements. You have simple equations for the hand calcs to compare with and will learn a lot about the different elements. Then try a plate with a hole and do some stress concentration factors for linear stress. Then add a nonlinear material and see what really happens at a hole when the stress concentration exceeds yield and you get plasticity.

good luck

gwolf.

 

[AeroNucDef]

Hi

There's quite a few good lectures to watch on youtube.

Here's the intro lectures

The first Video is by the Department of Mechanical Engineering, IIT Madras.

http://www.youtube.com/watch?v=djd9-f-onLs

Second video is by the Department of Aero Space IIT Kanpur.

http://www.youtube.com/watch?v=NYiZQszx9cQ

I think there are about 40+ lectures, each lecture lasting about 1 hour.
So go to youtube and type in the search box: Finite Element NPTEL

 

[flash3780]

It's best to start with a bar element. Bar elements can be solved by hand. Find the weak form solution to the finite element problem, then calculate stresses in the bar under various sorts of loading (tension, gravity, spinning, etc).

From there, you can try building a statically indeterminate truss structure and solving for the forces.

After you've gone that far, you may want to try writing a simple program in Matlab to solve some 2-d or 3-d problems.

Good luck!

 

[spongebob007]

I am of the opinion that probably the most important thing to do if you want to add value rather than simply pump out pretty pictures is to have a thorough understanding of the physics of what you are modeling. This will make it easier to spot errors in the analysis and to make suggestions on how to fix things when the model shows that your structure has problems. Making models is easy, interpreting results earns the big bucks.

 

[KnacKeN]

I am of the opinion that probably the most important thing to do if you want to add value rather than simply pump out pretty pictures is to have a thorough understanding of the physics of what you are modeling.

Yes thats what I was hoping to do. I've gone through bar and beam elements and have a reasonable grip on 3D Space Frame type analysis, and the componentns of the stiffness matrix for such problems. However Im moving on to 2D and 3D problems and the learning curves seems steep to me.

Thanks for advice.

 

[KnacKeN]

Sorry by 2D and 3D I mean plates and brick/pyramid/tet elements etc.

 

[Andries]

KnackeN,

The other replies you got all have good info; I especially recommend doing simple analyses that you can check by hand calcs. This, by the way, is also applicable to more complex analyses - always try to check you results using hand calculations, even if only to obtain order-of-magnitude figures. Another important check is to plot the displaced shape of the structure. If it does not correspond to reality then you know there is a problem.

Regards

Andries

 

[KnacKeN]

Just as a follow up I decided to try out FEMAP with NX Nastran. It allows you to install a 300 node limted demo, which should be OK for me to play around creating simple models as described above and run some analysis. Thanks for all advice.

 

[chris9]

I attended the NAFEMS course 'The Basics of Stress Analysis' some years ago. It was the best stress analysis course I've ever been on and made me into the engineer that I am. Highly recommended!

Chris

http://www.value-design-consulting.co.uk

 

[ysm]

I used this book
A First Course in the Finite Element Method by Daryl L. Logan

Wrote simple codes in Matlab and applied them to each problem in the exercise. It was a good learning experience.

Thanks
YSM