How to learn Finite Element Analysis?

This is a topic I get asked about quite often. Usually, it is phrased more like “where to start?” or “what is important to know”. All in all, however, All those questions can be gathered into one coherent problem. How should you learn FEA!

At the start, I feel obliged to point out that I have a very specific view on this subject. A view that isn’t “mainstream” I’m sure.

This is why I will start with a bit of a different question and that is:

Why do you want to learn FEA?

In all honesty, there are most likely millions of reasonable answers, but I will divide them into two groups.

Reason A: I want to write my own FEA solver!

First of all – congratulations! This is a very nice goal to have. As far as I know, this can be a challenge, but all rewarding things are! Sadly, I won’t help you at all with that – or at least not now. Truth be told I wouldn’t be able to write my own solver today without quite some time and internet access… and I have never even tried to do such a thing!

In all honesty, if that is your goal you can skip the rest of the article… it won’t really help you I think. Also, I would suggest the book by Crisfield (among others) “Non-linear finite element analysis of solids and structures” (Second edition from Wiley). It has some python code snippets in it, so it might be helpful even if they use their own FEM code called PyFEM : )

Reason B: I want to design stuff!

Now we’re talking! This is what I do and what I love! I don’t really know how mathematics works in the solver itself, but I definitely understand what certain parameters do, and how to set up the analysis. Designing stuff in FEA (regardless if you do it at work or just for fun) is a completely different skill set than writing your own FEA solver.

If you ever opened a “typical” FEA book (like the one I mentioned above), and instantly got tired about all that math this is your place! I will strongly discourage you to learn those equations (needed in solver writing). In truth, you don’t have to learn FEA math at all! I have never learned it, and I’m completely fine with using FEA in solving super complex problems.

Where it all goes wrong!

My approach may seem “dangerous” to you. I mean, whenever you go people tried to teach you the math behind FEA. I do something completely opposite, so of course, you may be doubtful. But I would encourage to try – I think you will quickly learn to enjoy my approach.

As an academic teacher with 10 years of experience, I’m fully aware where the problem starts. It is much easier to teach you how to write your own solver than how to use one. You don’t need a license, practical application experience and everything is in the books (with nice “scientific” complex math to back it all up!). So what you were taught at the Uni is the “write the solver approach” – it was so much easier for the Academians!

No worries, nothing is lost!

If I ever get challenged about not knowing math I usually say that I’m a driver, not a mechanic. I don’t care how the engine of my car work, I learn the traffic rules! And, I usually politely ask for a very specific example where the knowledge of math actually made someone’s design possible/better. So far I only got one answer, and now I know that if I need to do the very complex math to even describe what I try to calculate… stuff will get serious!

I think that the best argument on how strongly I believe in that is… that I’m doing this for 10 years now. As you know engineering is a “risky business”. If I make a mistake I can get sued for a lot of money. Which means that I bet a well-being of my family on the fact that my approach is actually reasonable – and with 4 kids on board, this actually means something.

So far, so good – nothing happened : )

Learning FEA for practical designers!

All right, if you are here, then most likely you want to learn how to do design in FEA. This is GREAT!

Let me start by saying this is a wonderful skill to have, and without a doubt, it had a huge impact on my career. I hope it will be as good or even better for you!

Finally, we can focus on how you can actually learn FEA!

Don’t learn the software – learn FEA!

Software Vendors are on the other side of the spectrum from “academians teaching math”. They will tell you what every button in their software does, but this isn’t as practical approach as you would think. I still have no clue what a lot of buttons in Femap do. There is a chance I will never know…

FEA is far more about the ability to figure out how to model something properly, rather than how to quickly model it in the software. You can be very efficient in soft, and still, have no idea how to model and solve a problem!

Focus on how things work, how stuff interacts with supports etc. This is not an easy task, but it’s a critical skill!

Think about it this way:

If you know what you want to model, but you don’t know where to click you will ask “how to apply an XYZ to my model in software ABC?”

But if you know how to use the software, but don’t know how to support the model you will ask “how to support my model?”

What do you think – which google search gives you a higher chance of finding an answer?

Decide on the software early on!

This may sound counterintuitive. I’ve just told you to learn FEA rather than software. And while this is completely true using one software makes sense. At least you won’t have to search all the time how to apply loads and boundary conditions in whatever soft you want to use now.

Using one package means that in time you will learn how to use it “completely on your own”. This is a nice thing! I’ve changed software before – this is not a tragedy. However, you do waste some time for searching “the same options” in the new software. Each new soft you learn goes quicker… but in the beginning, there is so much to learn it would be wise to stick to one soft!

Just ask around in your target industry what software is popular there… and use that one : )

Solve a real life problem!

This is an important one! Don’t learn FEA from books. Just try to model a simple real-life example using household items. A simple deflection of something you have handy would do. Try to model it (use stuff with simple geometry and made from popular materials so you can google parameters!) and check if FEA outcomes match the real-life tests you do at home.

Don’t learn about “everything” – focus only on the things that will be important in solving your first problem.

This means mostly boundary conditions and loads. Also meshing to some degree (but don’t get suck in into shape functions and all that… it’s not that important at the start). Of course, some 3D modeling skills will be useful (especially for more complex geometries) but don’t burden yourself too much with that at first – use simple geometries!

Repeat!

If you did FEA design and outcomes are matching the household test – success! You can make a short video of the test, animation of the FEA outcomes, put them one near another on your portfolio webpage… and search for another problem to solve!

Start from the most basic problems you can think of (and measure), and progressively do more and more difficult stuff. At some point, you can check videos of various experiments on YouTube, or search for FEA benchmarks of more complex problems.

If I could learn FEA this way (completely with trial and error) you can easily do it too!

Search for help!

This is something that did not occur to me. I admit that in 2008 blogging was not a common thing, and even searching help on forums wasn’t that easy. Today, it is so much easier. If you can’t do something – just ask. You can send me an email, or find a forum where you will ask. There are a lot of online courses and training materials available.

You are not alone in this! A lot of people learned, and are still learning FEA. Search them out and hang out with them! This alone is a great advise I would say. Learning alone is rather sad and requires a lot of mental energy (trust me I know), doing that with others is so much better!

Just be aware that if someone starts using complex math to explain something… things might not be perfect. Practitioners usually use much simpler language! If you feel you need to learn to solve some math (beyond maybe a square root) in order to do nonlinear FEA… something went wrong along the way!

Focus on meanting not equations!

If you learn some more advanced stuff (like nonlinear approach), focus on what parameters do. Don’t search for equations, just ask what this parameter change and what values are proper. Rarely the answer is “default is ok” – usually parameter does something, and it’s good to understand what it does : )

In the worst case scenario… run some tests! I have solved hundreds of identical models changing one parameter after another… just to verify what they do!

Make it fun!

Our lives are full of distractions. Try to make the process as enjoyable as possible. Even give yourself a cookie if you really need to 😛 Learning FEA will take you some time. This means you will have to do it more or less regularly for weeks (if not months). Making the effort pleasurable will greatly increase the chance of success.

Try to think of some really stupid household test. Just so you can tell your friends that you are using numerical computing (sounds scientific right?) to asses the breaking moment for a pencil or something. Just have fun, and you will get there!

Last few lines

I hope you have found this useful. If you like it, share this with your friends and colleagues. Maybe they will find something interesting here as well. It also gives you a chance of finding others that are willing to learn FEA!

And if you start to learn and reach the wall – let me know! Maybe I will be able to help you out!

All the best

Ł

Leave A Comment