What are some interesting analyses to do on your own? 3

[KevinDeSmet]

I am about to go on my holiday break and don't feel like partying, feel like breaking into the world of FEA and CFD instead! I would like to do some or multiple little projects on my own to build experience, and be able to show it when applying for myb first simulation jobs in the future.

I am looking for ideas, anything goes, preferably things that are doable on your own in about a month's time and that you can pull off on your own credibly and maybe something fresh, a little original.

Thanks!

Certified SolidWorks Professional

 

[rstupplebeen]

Once I get the free time I plan on analyzing either a sailboat sail or a kite for kite surfing in CFD. Once I get that working and reasonable I was planning on coupling it with FEA since I am very familiar with that.

So the real suggestion is what are your hobbies and analyze that. If you are into fishing the bending of a pole would be cool since it has contact, composites, fatigue, dynamics.

Conversely if you can think of a cool company to work for or one in a cool area try practicing on their product line. I know during an interview I would be impressed that they put so much of their free time into my industry and who knows one of us may help you.

I hope this helps and that the forums can help you model and learn something cool.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/

 

[ESPcomposites]

Not to be a downer, but it seems like you have no experience in FEA/CFD. That's OK, but I would consider this first, if you want it to be effective.

- FEA and CFD are very different, requiring specialized skills for each. How about just starting with one?

- Are you are stress analyst? Judging by the signature, perhaps not? If not, FEA will not do any good until you have "mastered" classical techniques. Of course this is my opinion, but one that I have learned through trial and error myself and witnessed many times from others.

- Before you begin on the computer, and after getting good with the classical approaches, you should also read up on FEM theory. You don't have to get super detailed with it, but a cursory knowledge is important.

- Short of this, you will likely be making pretty pictures. A lot of us have been there and it really is kind of fun, but ultimately you this is counter productive to what the goal of FEM is (which is to enhance classical solutions or go beyond their range). The reality is that MANY (and probably most) FEM solutions are actually less accurate than a pencil and paper in the hands of a good stress analyst.

Brian

http://www.espcomposites.com

 

[KevinDeSmet]

Rob,

Those are two great suggestions. Unfortunately I don't really have a lot of hobbies (unfortunately? maybe, I suppose) but I might run with your second idea and focus on industries I would like to work at.

Brian,

Not at all, I'm a realist myself. FEA and CFD are very different and I'm not that knowledgable on CFD as I am FEA, so I might take that into account when deciding. No I'm not a stress analyst but I've been researching FEA for about a year now and I feel like I got the theory down, I understand it's very difficult to do anything but make "pretty pictures" without the ability to correlate my results (I don't have a lab, right!) so I might just either accept that assumption or stick to the more stable analyses like linear statics and avoid fatigue/nonlin/dynamic but then again that's where the more skill resides so I guess I'll have to weigh those two things against eachother.

Certified SolidWorks Professional

 

[ESPcomposites]

Kevin,

What does "researching" FEA mean exactly? What books do you have for stress analysis? Why don't you correlate the results to some classical solutions? You don't need a lab.

You may not be aware, but the advanced FEM (nonlinear, dynamic, etc) will almost assuredly require very solid fundamentals and classical experience. Yes, you could probably use the tool to get an answer, but it won't be believable and this apparent "skill" would be pretty useless to anyone.

If you are a designer/CAD guy (which I did my fair of earlier in my career) you should know the difference between actually making a good design and making a solid model. You can get pretty good with CAD in days or weeks. But that does not make you a designer. Similarly, getting a FEM to run is really not too difficult and can be learned in weeks. It is not really that special of a skill. The real skill comes from the years it takes to be a stress analyst and applying that to FEM.

That being said, could you get a job with an apparent ability to make pretty pictures? Probably. Sadly, this is "skill" that managers like to see. It is totally counterproductive and cripples the good users. For example, the FAA frowns on FEA because of general poor quality. Hopefully you don't want to be part of the problem, but rather part of the solution.


Brian

http://www.espcomposites.com

 

[KevinDeSmet]

Brian,

Cosmos Companion Series, SolidWorks World Presentations, Autodesk University Presentations, Rice University Mech403 Courseware, SolidWorks Simulation tutorials and various miscellaneous Google searches.

Just getting absorbed in it, you know. I have focused my efforts on SolidWorks Simulation since it might have the most value for starting in FEA. My current SolidWorks knowledge is beneficial there, and unlike the big standalone more expensive packages where the positions are mostly reserved for expert level users.

I see great benefit in this for the more upfront FEA work. And might be able to use that, and grow from there. I've gotten myself certified with the SolidWorks Simulation certificate.

I wasn't aware actually that nonlinear and dynamic relied heavily on classical experience. I thought that was mostly for beam theory and engineering metals?

I totally get where you're coming from and I do intend to be part of the solution, but you've got to start somewhere.

Certified SolidWorks Professional

 

[ESPcomposites]

Kevin,

Since you have CAD experience, it is tempting to just run Cosmos, and all of a sudden you are doing FEA. This is what the software was setup for in the first place. Unfortunately, this type of approach is what also gave FEA a bad rap and hurts the stress analysts who do use it properly. I think you would be wise to rethink what materials to learn from. Hint, they are not from the software companies, but probably from Timoshenko and the experts in the industry you are interested in.

Most analysis done in packages like Cosmos are highly suspect because they are often done by designers and not analysts. These software are seemingly built on the premise that the computer can supplement the years of training that it takes to be an analyst. While this was a commendable goal when these sorts of programs were conceived, it was quickly realized that it did not turn out as well as hoped. The benefit and the blame lie in the convenience of such applications, which is a real dilemma. They have their place, but not really as general purpose FEA programs.

On a side note, true stress analysts do no typically use Cosmos (or least not as their bread and butter program). I think many of them think of it as a toy (note this is not necessarily my opinion, which I will refrain from sharing). I have used Cosmos, Mechanica, ANSYS, ABAQUS, Nastran, and CalculiX. I used none of them properly until I improved my skills in the classical methods.

Brian

http://www.espcomposites.com

 

[corus]

As others have said, it would be best to model those cases for which there are analytical solutions and you can compare reaults with hand calculations. Look in Roark's Forumalae for Stress and Strain at cases for thin plates/shells etc. as a good start.

Tara

http://tinyurl.com/4ydjg7m

 

[ZeroExperience]

Here are ten Patran and Nastran demonstrations on youtube. They are not necessarily projects, but they will help a lot in learning to use Patran and Nastran. Note that step by step instructions are given in the description of each youtube video.

http://www.youtube.com/user/PatranNastranAmateur#g/u

The developer of Patran and Nastran has posted some video tutorials at the following link.

http://www.mscsoftware.com/Academia/Learn/NastranPatran-tutorials.aspx

 

[Twoballcane]

I agree with Kevin. I feel that FEA and any other analysis tool is a ticking time bomb in the hands of people who do not know how to do the math and physics of what the software is suppose to calculate. Software is garbage in / garbage out. I would suggest that instead of learning new software that you go and do problems by hand out of text books and then try the new software. The classical theory is your base and where you start in any problem. Then when you do dynamic or non-linear problems, that is where calculus comes in and even differential equations. I actually like to do math modeling to come up with equations for everyday things and see what kind of slopes I can get on a graph and from there you can use calculus such as derivatives for finding max/min and integrals to find areas under the curve(that’s right I’m a geek).

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

 

[johnhors]

Twoballcane - you mean "I agree with Brian" ?

http://www.Roshaz.com

quality, cost effective FEA solutions

 

[Twoballcane]

*&^%&*%$ age is catching up with me...thanks johnhors, I meant Brian...

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

 

[rstupplebeen]

I fully agree that to be a real analyst knowing the theory is key. However, I would much rather have a few wrong assumptions than no model at all.

Many times when I ask "Why is it that size?" the answer is "It looked good."

I would much rather hear "I did a quick FEA and figured 1# was a good load and I fixed the end and there is no plastic deformation"

Kevin,
Model something up, analyze it and then post here in the forums for help. Also, checking out other people's posts and trying to see what makes their model tick is a great way to learn what and what not to do.

I hope this helps.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/

 

[IDS]

Analysis of buckling would provide an interesting line of investigation, where fairly simple models can come up with worthwhile results. Compare non-linear analysis with linear buckling analysis and classical theoretical methods (Theory of Elastic Stability by Timoshenko and Gere is a good reference).

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[Kwan]

Another exercise that will show you how mesh size can affect your results is to model a notched plate. Compare the fem to Peterson's Stress Concentration Factors (this is a book). This will show you why model convergence is critical.

 

[fea123]

Simple recipe that I prescribe to people with a design backgroud in my company wanting to learn FEA, maybe this will work for you.
I know that if I was hiring someone who said they went through the process below would earn a lot of credibility points rather than someone who has worked on an isolated cool project.

i) Select any of the problems at the back of one of Den Hartog's books (Mechanics, Strength of Materials, Advanced SOM, Mechanical Vibrations)

ii) Read up the theory in the relevant chapter and see what principles are applicable and reread till you think you have a rough idea of how you would solve the problem by hand

iii) Build the model and solve it using your FE tool, you might have to select dimensions, material properties etc. Does your answer agree with Den Hartog's answer to within 10% or so? If not, try to figure out why and refine and change your model till it does.

iv) Some problems will not be able to be efficiently solved by FE, this will be an education in itself!

v) Repeat this process about a 100 times.

I can almost guarantee once you have gone through this process, you will be a competent finite element analysts with the ability to build robust models!

 

[loki3000]

iii) Build the model and solve it using your FE tool, you might have to select dimensions, material properties etc. Does your answer agree with Den Hartog's answer to within 10% or so? If not, try to figure out why and refine and change your model till it does.

it is also beneficial not to only read thetheory, but to LEARN it and verify the results by analytic calculations.
if you really want to convert to stress, it would be better to show that you can do something else then look at nice drawings.