/sci/ - Science & Math

Im not sure what you expect them to do as an alternative

>elegancy
stop brute forcing english, retard

>>11958614
not my fault english is retarded

>>11958614
https://www.google.com/search?q=%22elegancy%22

>>11958636
>brute forcing a search engine
ok nerd

>>11958646
Google brute forced your brain lol XDD.... KEK!

>>11958594
Finite element analysis is where I discovered the limits of what I was able to comprehend. I could use it and had a grasp of what was going on, but I would be fucked to explain the mathematics of it.

>>11958629
and it's not engineers' fault that many-body physics is retarded

>>11959253
I'm not sure what you mean, wouldn't you just explain it like
>we know how the laws of physics behave in theory, but applying these laws to complex systems causes our math to become immeasurably complicated. so we approximate these systems as discrete chunks where the physics can be solved on a computer without running out of memory.

>>11959259
Oh I meant truly understanding just what the hell was going on while applying the Galerkin Method, not what finite element analysis was.
Like a chimp with a gun, I could use the software competently and go through the steps to solve simple problems manually, but just couldn't grasp the math behind it.

>>11958594
Finite elephant analysis :D

>>11958594
but numerical methods are chad af, both from an engineering and mathematical perspective

>>11958594
More like elephantcy lol

>>11958594
because it just werks©

>virgin curve vs CHAD disjointed line

i failed numerical analysis, feels bad lads, idk why im so bad at maths

>>11958594
Why should I do something analytically if I'm working with non ideal conditions and my computer can calculated the answer to 5 decimal places in seconds

>>11960226
because sometimes the model is not precise enough under certain conditions

>>11958594
If it works, it works. We don't need our airplanes to look exactly like a bird and we don't need it to flap its wings like a bird. As long as we understand the real mechanics behind flight, we can rebuilt it anyway we want.

>>11958594
numerical methods are the only way to solve most differential equations. It's not brute force unless you're trying everything. Also finite element analysis is better than having mathematicians do the integration. They have much better things to do than integrate functions.

>>11958594

FEM isn't exactly Brute force. Brute forcing a password would require trying every single combination. With an efficient algorithm the solution is approached with every iteration, so not every combination is not tried. There is an initial random guess, but the components of that random guess that are not part of that solution space are removed

So for an analogy imagine trying to find a password, but every time you coincidentally guess one correct number that get locked in and you just have to guess the other numbers, but every time you iterate the number just goes closer to the actual solution. It's very beautiful actually. Convergence of a numerical algorithm is to me the most beautiful thing in all math and physics. Most questions in the real world are too complicated for an exact solution, so in general the solution to most problems must come from a converging numerical algorithm.

>>11958594
lol

>>11958594
The perspective on the legs is like the elephant is doing the charleston

>>11959344
Kek

I'm going to take FEM and CFD at the same time. Am I going to kill myself?

>>11958646
>Calling somone a "nerd" as if it were an insult because they know more than you and want to feel better about your pathetic uneducated brain

>>11962585
>expand nabla notation to austistic extent
>get stupidly long equation
why are we surprised?

>>11958594
Not brute forcing when it is a viable option is based but also maybe dumb. They do it because with the problems they are working on it's a good solution

>>11962900

No you're gonna be 300k starting

>>11962943

>brute forcing a greentext

>>11960226
Reality is not an approximation.
Humans will forever be locked away from the truest heights of understanding and engineering due to this fact, sadly.

>>11962585
ah yes, physics

>>11958594
When state in which you know solution that is different is called psychosis, what do you expect?

>>11959293
iktf, but what scares me much more is people with little/no knowledge of FEM running FEM models or using FEM results. I don't know how many times I've seen analyses use (unduly) averaged results to find a margin of safety, or use results from near the boundary of a model that only partially models a structure (ignoring St Venant's principle), or extract results in the wrong coordinate system, or just straight up use bad modelling methods. People come up with FEM validation checklists and procedures, but in the end, there's no substitute for common sense, a sharp eye, and good engineering judgment.

>>11962585

is this the ultimate ToE super-equation?

>>11965236
Recommend any texts to up my FEM game?

>>11966791
Personally, I keep Bathe's Finite Element Procedures on my desk. I rarely crack it open, but it's a nice safety blanket in case something really unusual comes up. I like the breadth that it covers as well as the depth, and the "basics" it covers are pretty readable, which is nice. But, there are other good FEM books out as well.

I like this Arxiv paper, it's a list of good rules of thumb that I like to skim through now and then:
https://arxiv.org/pdf/1811.05753.pdf

Most of all, I recommend texts relevant to the particular software you're using. I use MSC Nastran, so I keep their Quick Reference Guide and manuals handy. I also have a recent edition of the MSC/Nastran Primer, a really nice text that covers theory, but relates the theory directly to how it's implemented in Nastran, and also covers how Nastran performs on various test models.

Besides books, I like these MIT OCW video series by Dr. Bathe himself, covering Linear and Non-Linear FEM, respectively

https://www.youtube.com/playlist?list=PLD4017FC423EC3EB5
https://www.youtube.com/playlist?list=PL75C727EA9F6A0E8B

Have fun my dude

>>11965236
How can you promote "good engineering judgement" over the results of a model? How would you suggest an engineer interpret FEM model results (ideally conservative for the setup) if not at face value?

You respond with a list of FEM resources that shows you are competent. How do you reconcile we solve inaccurate FEM models and the inexperienced engineers who utilize them to make decisions?

t. Aero eng

>>11967848
I'll answer this in the scope of linear static FEM of complex structures for aerospace applications, since that's my field. For very easy FEM models, a lack of experience is less of a problem, since you can validate the model by some procedure and run it with confidence. For very complex models, empirical testing is a must. In my field, however, it's not unusual to create models of significant complexity, but extensive testing may not be performed.

I'd say the best solution is to not let inexperienced engineers mess with FEM models without close supervision. Thankfully, at least in commercial aerospace, we're still at a point where it's cheaper to analyze many structures by some sort of hand analysis than by FEM, since if there's ever a design iteration, you can often update the hand analysis by just changing some geometrical parameters, while FEM would require you to create, or significantly modify, a mesh. So, we can put young engineers on these sorts of problems, which helps them develop some intuition on how certain structures behave. If it's necessary to have someone work on FEM for the first time, then we need to make sure that they're given proper guidance and that they receive feedback on their work.

>>11968892
Unfortunately, this is all easier said than done. "Easy" problems that are ideal for young engineers are also ideal for outsourcing to low-cost engineering centers. In these cases, young engineers can spend almost all of their time devising statements of work - figuring out what work has to be done - only to have the actual engineering work itself be outsourced. As for gaining FEM experience, a lot of FEM work is inherently "deep thinking" work, and some managers are biased against giving young people this sort of work. While I don't have any respect for a manager that doesn't judge someone based on their individual merits, the general stereotype that young people are better at multitasking smaller problems while older people are better at deep and creative thinking probably holds some weight.

Another "solution" is to "appify" FEM. That is, to have a team of experienced engineers and programmers develop specialized programs that accept geometrical parameters, materials, and loads and boundary conditions, run a FEM model in the background, and provide you with a result (a maximum stress or even a margin of safety) as an output. The benefits are obvious - within an allowed range of inputs, the FEM model won't be wrong, and there are huge time/cost savings in generating a FEM model automatically vs manually, especially if it's a very common type of problem. However, this does nothing for developing engineers. Despite that, since the main goal of an engineering organization is to release drawings on time and on budget, such "apps" are already in use in aerospace and their scope will only continue to grow. Maybe someday, a practical, generalized FEM app will be developed, in which case I'll need to find another job. I don't think we'll have anything like that any time soon, though.

>>11968894
Maybe the future isn't bright, but here's the advice I'd give:
1. If you're a young engineer, get a Master's degree and learn how to program. Today's companies have nearly limitless access to problem-solvers, whether human or computerized, for easy problems, but difficult problems remain difficult. Being able to program will show that you can do deep, creative thinking, and will make you much more efficient than your peers. Plus, FEM and programming go hand-in-hand; you don't need to be able to program if you're running a single load condition, but if you're applying 1000 load conditions to a model, you need to be able to program.
2. If you're a manager, consider the cost of not developing your workforce. If you outsource, what contractual obligations does your outsourcing company have? Are they required to provide "correct" analyses, or only to correct errors found by your engineers? Are they immune from the effects of natural disasters or national politics? What certainty do you have that they'll be able and willing to provide services to you in the future? On top of that, computational power isn't growing as fast as it used to, and there hasn't been a "quantum leap" in the efficiency of the FEM method itself in decades. Most potential improvements in the design of structures seem to be in new manufacturing methods and in architected materials, which means new kinds of geometry and material properties. Is your engineering team ready to tackle these new challenges?

Anyway, thanks for coming to my TED talk.

>>11966715
I think it's the full standard model

>>11968892
>linear static FEM of complex structures
>linear
How is it even possible to fuck up something so simple?

>>11958594
There is no elegant solution to real world problems.

>>11959253
I think you must have developed a FEA software by your own to grasp it, at least that's what someone who did, told me.

We spend years in school, stressed ourselves the fuck out, slave away in a STEM career just to make a measly 6 figure salary. Meanwhile fucking social media "influencers" do jack for society and mankind, but make more than we ever will in our lifetimes. It's not fair.

>>11970547
That's what you'd think, but I keep finding new and exotic ways of fucking up a model, and I promise it's not just me. Once a model is big enough, it becomes almost as much about configuration management as it is about FEM.

>>11970701
>does it for money
That's your mistake, money is a byproduct shouldn't be your endgame, as long as you can cover all of your needs including non basic nones like travel and videogames you are set dude, no need to rack up excessive amounts of money by basically whoring out your privacy and dignity.

>>11958594
you mean ELEPHANCY?

>>11968894
Hi, an aside not related to FEM.

I am someone different who wants to ask about you saying the notion that younger engineers are better at multitasking holds some weight. Certainly in my day to day life I practically cannot easily tackle the big problems as fluidly as someone more experienced (it's possible but not pragmatic, since for example I'll take longer learning the basics whereas the experienced person has already done that, not that it feels it takes long to get the experience, it's just that initial barrier to entry). Now I don't have much experience in work, but I read this funny article from an embedded electronics engineer about doing the reverse - intentionally assigning inexperienced engineers to the deep problems and utilising experienced personnel for the small problems as an ideal for the manager to minimize labor expenses (well, the guy mentions more reliable software and firmware).

Uh... I can't find the original article, which showed a graph produced by some consultancy where a "more reliable" workforce crawls "up" a j-curve. I've drawn this mystical picture that I can't find in pic related.

I remember I got to the article from here... skip to the heading "Other Tidbits" and especially the link "Superprogrammers":
http://www.ganssle.com/item/guide-to-managing-firmware-teams.htm

Specifically, in this link, skip to the paragraph starting with, "Caper Jones, in an unpublished 1977 study for IBM, found...".
http://www.ganssle.com/rants/powerprogrammers.htm

I don't know how much CFD you like. I don't do engineering work like you, but I like raytracing, and I recall an article someone wrote that contrasted the difficulty of working with efficient real-valued but common-sense raytracing algorithms with the ease of working with initially unintuitive but practically more elegant geometric algebra (hypercomplex arithmetic) less efficient algorithms. I'm thinking of the difference between CFD from "first principal" vs CFD via lattice-boltzman. I know the latter is a lot easier for me to grasp, even though it emerges from unintuitive foundations - BUT I program with Scheme, and I don't have the pressures of real engineering work. I figure if back in college I'd had different most-basic tools like I do now that I'd have had a much easier time of things. So, the chart I badly drew feels like the more experienced engineers' role becomes "build superior fundamental tools for the interior engineers so they can more easily build themselves up into better engineers/so they can build the fundamental tools themselves". Ah, and pic related, I forgot the drawing in the earlier post. I swear there is a drawing like this on Ganssle's site.

>>11958594
Same.
That's why i love airplanes, cars and other weight critical vehicles.
Been working at a steel manufacturing facility this summer and goddamnit all the machines here are just ridiculously huge chunks of steel which is somewhat annoying, guess most of the blame can be put on manufacturing methods but I'd say that a large chunk of the blame can be out on laziness and the lack of generative design programs some 20 years ago.

>>11959253
Isn't FEA just dividing the part into really small triangles and then it's just statics/strength calculations.

>>11960308
>implying a bird is in any way perfected flying machine
baka

>>11959253
You literally have matrix templates that do most of the work for you, your only job is to plug the numbers and simplify it? What is there not to get? It is just beams and ties columns.

>>11973779
Pretty cool articles, thanks for posting, it surprised me to see that, for very large projects, productivity between the worst and best approach each other. Still, when this guy talks about small projects, he means one man-month projects, whereas a small project in my world would be more like half a day. I very much agree that experienced engineers on large projects should spend their time breaking down the problem into chunks and providing the tools and processes to help others be productive. I'm not experienced, but I've been spearheading a large FEM project, but have done almost no meshing at all! I spend all of my time developing trackers, reviewing others' part meshes, creating tools that simplify certain repetitive tasks, coordinating with design engineers when the design they provide doesn't work, etc. The sad truth is, every good engineer eventually becomes at least a part-time manager.

>>11970701
Yeah but we get a rather clear cut meaning in life compared to social media influencers. Also our career is way more stable and is I think not as difficult honestly.

>>11976472
retard

>>11958594
If you think engineers brute force things wait til you see what technicians are capable of.

>>11977886
They somehow get the damn perfection those engineers draw to actually work, in three space, holy fuck. Just when they thought they made an untorqueable screw those wrench monkeys fucking omit it. Yes, that is the root cause of machine failure. I don't know why they do it, this thing ain't ever going to break. Just get it on the one time and it is good for life partner. We all know that one oversight brings on the whole sloppy hole of shit and piss and problems to bear. Better cut their budget. That'll teach em for breaking machines. Management will love it.

>>11977927
This guy engineers

>>11958594
eat shit analyticfags

>>11960242
>not dumping years of r&d into unfucking your simulations
not going to make it

>>11976472
>complaining about simplified geometry on heavy machinery
i bet you fell for the "german engineering" meme too

bump

>>11978378
Oh you mean making everything super complex for little to no discernable benefit?

Is Finite Elephant Analysis just Calc of Variations?

>>11979841
god i hate krauts