The role of software and math in FEA
I used to think about the role of math and software in FEA as about two different phenomenons. But recently…21 November 2022
I’ve seen dozens of FEA reports done by others. Mostly because Customers asked me to verify them or provide an opinion if the calculations are any good. More often than not, I have to inform the person that asked, that the report they have has serious flaws…
The ability to briefly check an FEA report is really useful for managers and business owners. After all, you really want to make sure that the calculations you’ve ordered are good! But the same knowledge will help FEA specialist to make their reports even better!
There are some key things you want to pay attention to. And while some, will require at least a bit of knowledge and critical thinking, many can be verified by folks without deep engineering training.
Of course, as always, I will try to keep this as simple, and as easy to understand, as possible!
I’m absolutely in love with FEA aesthetics. I love animations of outcomes and good snapshots of interesting parts of the model. I mean, doesn’t this just look great?
Judging by the reports I verified, I’m not alone in this love! Usually, FEA reports are full of such images.
Don’t get me wrong, I have nothing against images in reports… not only do they look great, but they also “prove” that you actually did the FEA calculations of the model (!).
But the value of the report is not in the images (sadly). Just think about it, you would not be able to tell if the above part is good or not based on that image.
Experience shows, that FEA reports often look like this:
Sadly, the above is just wrong! I know quite a few design codes that allow using FEA… non of those allows us to judge the capacity of the model “based on the image”. To be honest, I can’t really imagine what that procedure would look like!
I think that the above, approach is completely wrong. But it is still the most popular way of preparing FEA reports. And I even understand why is that. The idea is quite simple: you do the model, load it somehow, show the images with the “outcomes”, and “by the eye” judge that this “looks good”. But this is not how you should do this…
Remember: FEA images look great, but they do not convey enough information to judge what is the outcome of the analysis!
If the report contains mostly images without descriptions and outcome analysis for them… there is a very high chance, that this is not what you are searching for!
Engineering is pretty much “standardized”. This means, that if you want to “design something”, there are official documents (most of the time on a “national level”!) that dictate how you should do the design!
This is a huge asset, as the standards require you to make certain activities, and give rules on how you should interpret the outcomes. Those requirements are there, even if the Customer is unaware of them! So when someone orders a design “by the ABC standard”, it’s the engineers’ responsibility to follow that code!
This means that ordering design done by a decent code is actually beneficent. It gives the Customer a framework, that will be checkable later on (even if the Customer does not know that framework… it’s just there!).
The standards I like the most:
All those codes seem overwhelming?
Don’t worry! If you just want to check the FEA report provided to you, you don’t really need to know the standard. All you need is to be aware that design codes exist, and that the report should follow one of those!
Ok, so let’s get back to the report itself!
I’m a firm believer, that the FEA report should follow a selected standard/code. Usually, it’s the Customer’s choice, but some national rules may “force” a choice of a regional code on some occasions. It’s always good to specify it at the start of the project to avoid trouble later on!
Of course, the Customer/Manager doesn’t really have to know the code. But it is definitely good to know that the codes exist and that you need to choose one for the design!
Practically I’m always more than happy to help my Customers select a decent design code that “fits” what they need. But I have to say that nowadays, most often they just know what they need themselves.
Even if you haven’t specified anything, and you are not sure what to choose – don’t worry! A good FEA specialist will ask you about it and guide you toward selecting a proper standard for your task.
Good FEA report will specify, which design code was followed in the design.
But be careful! It’s not enough to slap on the cover that design is done in accordance with a given code… it actually has to follow it, for it to be worth something!
This is where the technical knowledge comes in!
As you already know, there are many codes, and realistically, it’s impossible to know them all! But I feel it’s also important to be sure, that the design was following the code procedure. How can you check this?
Yea, it should be easy as that! I mean, codes are usually pretty thick (200-800 pages) with multiple chapters and rules. Almost always, they include more than a single check that the designer has to prove (stress, stability, fatigue, you name it!).
This is why good reports always state the procedures used in a given part of the report. They should also show, what is the outcome of that procedure.
So expect quite a few code references like “in accordance with EN 1993-1-6 chapter X.Y” etc. in the document. Even if you don’t know what those chapters are about, the fact that someone mentions a code is a good sign! And if you are making the report, adding the code reference is just a small fraction of the work, but it really helps your report to look professional and “stand out” – so I would advise you to do it if you are not doing this already.
If you don’t know the code of the design, but you feel that something may be “off” with the references – don’t worry. You can always ask someone for help! Or just randomly look up a few chapter references to see if they match the report content (if equations look the same, etc.)
The engineering world likes numbers!
Most code procedures end with the “code check”. It’s usually a simple equation where *something* is smaller than *something else*.
One of those things comes from FEA analysis. It will be some sort of an outcome, like stress, deflection of a beam, plastic strain, etc. The other comes from the code, but the designer may have to calculate that value “by the code” to adjust it for a given case.
So expect that each code check ends with “A<B which means that the requirement of chapter 14.3 is fulfilled”.
There will be no vague descriptions like “as you can see, all is good”. Although this is a bit tricky if the outcome is “obvious” for an experienced engineer…
Imagine a model where some check would be 1 < 1000! In such a case, it’s quite obvious that the check is ok… so the author of the report could writhe something about “as obviously seen”. Still, such situations don’t happen all that often. And writing the equation is good practice anyway.
Sadly, this is not all black and white. I.e. ASME VIII div. 2 has a few checks that succeed when the analysis converges for a given “state” (I really dislike those!). Verification of such a check is a bit more tricky. It will simply end with a funny sentence like “since the analysis converged for the load case C, verification according to chapter 5.3 positive”. There will be no equation or numeric outcome. So clearly, this may sound “unspecific”. Still, those things happen rarely (I can only think about ASME from the top of my head).
This is simple as that. If someone provided you with a report, and you are not sure about the code, or the checks look “vague” and “strangely unspecific”… ask about it!
Again, you don’t need to know the code at all. Just ask which procedures designers used in the design (ask for chapter numbers or equation numbers from the code). The answer to this question usually suffices to know if someone did a good job. You will either get the chapter numbers that the designers used… or not!
This is a bit more tricky! While you see what is in the report, you really don’t see what is not there (and maybe should be there).
Usually, design codes divide the verification into a few “categories”. And you can ask about each one of those:
What is cool, is that the above nicely fits into the previous point. Each of the above checks has a “code procedure”. This means that you can do them “correctly”, but also “wrongly”. Always ask, which chapter was followed in verification in any of the above!
Imagine a situation like this:
You ask the report provider, how they did the “stress design”, and they reply that the images (the nice FEA ones!) in the report are the “stress design”… but you already know that the images are not enough after riding this article, so you ask: “great, which chapter in the code describes how you analyzed those outcomes and found the answer satisfactory?”. Check-mate!*
*BTW, if anybody will tell you a chapter, and that code really describes how to derive capacity from nice images… please let me know!
But this is even more effective with stability! It’s absolutely commonplace for people to use LBA (Linear Buckling Analysis) of about anything, to prove that something has sufficient buckling capacity. The vast majority of codes do not allow such checks! They are too primitive and too inaccurate! In such cases, always ask which code chapter specifies such verification… and you will see if the job was done the right way.
How to check if the report follows the selected code?
It should be obvious from reading the code. Expect quite a bit code references, and checks that usually look like A<B in the report. Those are the outcomes… not the colorful pictures!
If you are not sure – ask about it! The report provider should answer which code procedures were used, but remember that you are not hiring them to train you on how to do such analysis. So be ready to read that section of the code yourself if you will still have doubts.
It’s also good to remember that usually following things should be covered in report:
- Stress design
- Stability design
- Fatigue analysis
Of course assuming that the scope was not limited or extended beyond what code requires (but you will know about that for sure).
I’m fully aware that there are industries where codes are not so dominant, or may not exist at all.
Sadly, this makes things much more complicated!
The absence of rules doesn’t mean that everything is acceptable (you wouldn’t need the FEA report in that case, right?). This means, that someone has to decide what is acceptable and what is not acceptable in some way.
I think, that in such a case, you should select a “generic code” that you can use in many areas. EN 1993-1-5 and EN 1993-1-6 come to mind as good ideas.
Otherwise, be ready for a discussion on the limits you are willing to accept, safety factors and load combinations needed in design, etc.
Sadly, I don’t have good news: this is a technical thing!
While codes describe a lot, they will not inform the engineer how to properly support their model. And the most outrageous mistakes in analysis that I’ve seen were with model supports.
There is no simple way you can check that if you don’t have sufficient engineering knowledge I’m afraid.
But at least check, if the way the model is supported in the analysis is described somehow in the report. If you feel like it, you can also wonder if the model supports “make sense”.
You can always ask about it too.
The way the model is suported may be the most important part of the analysis!
Design codes give no guidance on this – it is assumed that the engineer performing the analysis knows what to do!
Check if this is described in the report, and if in doubt ask someone to help you understand if the supporting of that model “makes sense”!
Here, we have to be realistic. You most likely won’t be able to check everything with loads… since you would literally have to write half of the report yourself again!
Codes usually nicely describe how to calculate loads from earthquakes, wind, or snow. Those calculations can have several pages of weird equations.
Funny enough, my experience is, that those are usually quite correct! I mean sure, there may be errors in following the code procedures, but you will have a very small chance of finding those anyway (unless of course, you have to validate the report… then check each and every equation and digit!).
But check if all the loads that should be in the analysis are really there (i.e. someone did not forget the earthquake load completely!).
Codes usually describe how to add loads up in so-called “load combinations”. Expect a list of those in the report, and take a look if those make sense. Usually, loads in those use “safety factors” (so are bigger than you would expect). Safety factor values vary between 1.1 to even 3.5 in some extreme cases. The most common values would be around 1.5 to 1.7 I would say.
Checking loads will most likely be an unneded nightmare!
To the point, that I’m not sure if this is what you want to focus on, unless you are charged with actually “veryfying” the report.
But do check, if all various loads were consiedered, and if load combinations look reasonable (each code should describe those, so a code reference there is expected as well).
I wonder if you were waiting for this part. I feel that many folks obsess about this. OK, I’ll admit that poor mesh quality can really make the outcomes “bad”. I would obsess far more about model supports, and about the code used in design.
It’s hard to get a “feel” of how small mesh is “small enough”. You could of course do the mesh convergence study (more on it here), but if you just want to quickly check the report, that would be way too much work!
It’s impossible to make a quick guide here (just like with supports really) but a few pointers:
As for the mesh “size”, it may be a bit tricky, but I think I know how to help. Take a look at this image, and wonder which mesh size in your opinion looks “good enough”:
Remember the number under the model you selected (doesn’t have to be accurate!). When you are ready, check on the horizontal axis where your number is. The chart will show you how big an error such mesh produced in nonlinear analysis (there is a separate curve for TRI and QUAD elements):
I know that this is just a “rough guideline”, but at least it will give you a visual hint!
We covered a lot of ground here, so I feel that a small summary is in order!
I hope that this article will help you to better analyze FEA reports, and also to write good reports yourself!
A few key points to remember:
And in the end, if you liked this post, you can learn way more by joining my FREE FEA course below!
Join my FEA Newsletter
See you on June 13th.
Join the discussion