Stress Singularity in FEA!

[@Ans_aka_Inzi1073]

Excellent video on this infamous FEA topic! Really appreciate your tips of non-linear analysis and sub-modelling for handling such problems P. S. At the moment, I only have to deal with relatively smaller models. So I guess fillets would do just fine xD

[@Enterfea]

Thanks Mate! I'm glad that you like it :)

[@pawesiwik8802]

Great video, very helpful :) Currently finishing my master degree about bolted conections calculated in Abaqus and by hand with eurocode. I had a small problem with uderstanding why in model stresses around edges in shell models (while also being contraint) can be enormous while mises in rest of the model do not even exceed yield stress. Whish you were my teacher haha. All the best!

[@Enterfea]

Thank you Paweł! Good luck with your masters!

[@mohamedelsarag1228]

For the reentrant corner, the point lay on 2 edges each with different boundary condition (as one is vertical and the other is horizontal), so there is no way to fulfill the requirement of both edges rather than the zero stress state. Other than that a singularity will appear as a result of contradicting boundary conditions.

[@Enterfea]

Thank you for the comment Mohamed. But let me ask you a bit deeper. What do you mean by boundary condition? I assume this has little to do with how the model is supported or loaded (to me that would be boundary conditions... but I know this can mean many things to many people). Of course, in this case, neither of those two edges is loaded or supported... so I guess you mean something else. Would you mind sharing some more details about this? THANKS!

[@MrArturlax]

Dziękuję. Najlepsze lekcje o FEM na KZbin!

[@Enterfea]

Thanks :) I'm really happy that you liked the video* * That is me being "international", bo tak po prostu to dzięki! Cieszę się że Ci się podobało :)

[@Mekhify]

Point loads are not an issue on any model, unless they incorrectly affect the results in the area(s) of interest. ;) I get what you mean, but I think it is important to stress that however you apply your BC's they will be approximations, which is fine as long as the error of the approximation is considered. As a starting engineer the error should be estimated, and as you build up your experience you can work towards guesstimates, and then judgment based on... experience of similar things you have seen and tried before.

[@Enterfea]

Hey! Yea, I understand what you mean, although I tend to be very strict with my BC... there is always an error of approximation (guesstimates - I like the phrase!) - but at the same time, I kind of always knew when my approximation is "good enough" in this case. I mean, this is definitely something you need to do with FEA, but this is also where engineering judgment comes in... but as you wrote Jan, at the start of my engineering career I did a lot of tests of "how things works" and I learned how "various BC" affect outcomes. So maybe this is where the engineering judgement in this field comes from :)

[@19klemc94]

For fatigue you need to use Hot spot stress method. And so you don't need to have super small elements in the area of interest.

[@Enterfea]

Hey @19klemc94! This is an interesting point. I'm not sure what exactly do you mean by the hot spot method since I think that this is used in various contexts around the globe, but there are ways to extrapolate stress toward joints, etc. (i.e. IIW recommendations). However, I'm under the impression that if you have a "funny shape" cut from a single piece of metal and you need o to assess the stress concentration factors to do fatigue analysis, there are no reliable "simplified methods" to achieve such a thing, and small mesh would still be needed - or am I missing something? Thanks for sharing your opinion on this!

[@19klemc94]

@@Enterfea The point of HSS method (hot spot stress) is that you interpolate from a points which are some distance away from the point you are investigating. But in those points from which you are extrapolating the strees is almost independent of element size, so extrapolating value is very much the same with different mesh sizes. Of corse there must be ''smart'' number of elements (enough elements to describe deformation shape), but no super crazy small elements is required. Distance to extrapolating points is of corse recomended by standards and is dependent of shell thicknesses.

[@Enterfea]

@@19klemc94 Absolutely - this I understand easily. You could use that in a welded joint for instance, and that is a cool method. What I'm not sure about is how this would "handle" a simple (and somewhat classical) plate in tension with a small hole in it. As we all know, near the hole the stress is 3 times the average and changes pretty quickly near the opening (and quickly "dies out" to the average). I think that HSS methods work best if you don't have features like that (which makes the selection of the "path" relatively easy). Would you agree?

[@19klemc94]

​@@Enterfea HSS method works even in this case. You can try it. Of corse you will not get the stress of 3 times as nominal, but something smaller. So this result is not actual state in material but some fictional value of stress. However for such an approach in standards you have recommendations how large this fictional stress could be for detail to survive the desired number of cycles. It turns out, that for steel (welds) this value is always around 100MPa, regardless of shape of the detail.

[@Enterfea]

@@19klemc94 I guess that I'm in favor of a slightly more accurate approach in "geometry cases" then - but I understand where you are coming from. Next time I will do some fatigue stuff on details I will try to remember to do the HSS method as well (in places where I would go with the mesh refinement instead) - just to see how much it "misses" the mark. Because I'm still not sure if the accuracy in certain details would be decent - but this is a "hunch" only - I have absolutely no grounds not to believe you... it just needs testing on my end to make challenge my own point of view on this :)

[@danielecostantini8722]

Outstanding video. Dzienkuje 💪 If you are in Warsaw I’d dig to hit a beer with you man.

[@Enterfea]

Hey Daniele! Thank you so much! I really appreciate your kind comment :) I'm in Wrocław - this is like 3-4h drive from Warsaw (not sure, haven't driven there in some time). Sadly, I'm not going to the capital anytime soon (to be honest, I'm simply swamped with work!) but if you will ever go to Wrocław definitely let me know! I even know a few awesome coffee shops (I'm not a beer guy I'm afraid... which is SUPER WEIRD for Poland, I know :P)

[@peterszalai79]

Thank you! 👍☺️

[@Enterfea]

I'm really glad that you like it Peter!

[@GenaEnSamIAm]

Thank you 😊

[@Enterfea]

I'm happy that you like it Eugene :)

[@adarshnataraj7904]

Awesome.. Thanks a lot 😊

[@Enterfea]

I'm really glad that you found this useful Adarsh! All the best!

[@russreadsbooks]

Great video!

[@Enterfea]

Thank you Russell! I'm glad that you found it useful :)

[@hemantbhoosnurmath4007]

Brilliant video! Thank you :) Would you know how can I calculate the loads I would apply on a smaller detailed model of my actual model, given I know the load for the global model?

[@Enterfea]

Hey Hemant! Thank you for your kind words :) The loads are only a part of the problem - the boundary conditions may be even more "tricky" - there is a whole "range of topics" on this called submodeling - I think this is what you want to "search for" :)

[@kuba4077]

Dzięki :).

[@Enterfea]

Proszę bardzo :)

[@GenaEnSamIAm]

Perfect 👌

[@Enterfea]

Thanks :)

[@ayomasukaja]

when i modeled. i was wish i had an experiment to confirm my model, so i can ignore the singularity. Like it is happened or what. Is there any site that or books providing fea checks to the real strain gauge monitoring experiments.

[@Enterfea]

Well I'm making my own lab tests for benchmarking in my Nonlinear FEA Masterclass, and I would guess that many scientific articles use tests to benchmark analysis. But I doubt there are easily accessible online resources containing the databases of such things.

[@ayomasukaja]

could you compare pin lug analysis like this vs petersen book peak stress vs ASME / Eurocode /AISC lug pin analysis.those code allow, but the stress is beyond yield@@Enterfea

[@Enterfea]

@@ayomasukaja to be honest with you, I just don't have the time to do it... I recently did a PIN case-study with some FEA and Eurocode comparisons, but that's about it. But maybe one day I will have the time to sit on this... right now, there is just to much to do. All the best! Ł

[@supreethrkoundinya2872]

Great video! Can you do a video/tell us on how we can do hand calculations to verify the FEA results? Can hand calcs be used as a means of verification all the time? Is it possible?

[@Enterfea]

Hey! Thank you for reaching out! Hand calculations are a great way to check stuff... when they are ways to calculate (or even estimate) stuff you wish to analyze with FEA. This is of course not always a case, but usually, there is something you can check by hand (and hope that "since this one worked, the rest should be fine as well"!). If this topic interests you, you may want to read this: enterfea.com/5-steps-fea-results-verification/ and this: enterfea.com/fea-results-verification-with-hand-calculations/ All the best!

[@supreethrkoundinya2872]

@@Enterfea thanks a ton. But what about a complex geometry, like a steering system that I'm doing an FEA on, there will be multiple forces on multiple locations. Can hand calcs be done in such a complex case of geometry?

[@Enterfea]

@@supreethrkoundinya2872 I don't know the problem that you are solving, so I also don't know the question to this answer. My experience is, that usually, you are able to verify at least some things with hand calculations, but of course, this is not a general rule, and I cannot guarantee that it will be so in your case as well! Without a doubt, the better understanding of engineering you have (and better skill in hand calculations), the easier it is to find something you can verify this way. All the best! Ł

[@araglar]

I keep noticing very high concentrations in very small areas in the FEAs of various types of industrial threads. These stresses are in the thread root and contact areas. Typically it’s in the first three threads that the strain is highest. Do you have any advice on how one could mitigate or defend the results of an FEA with that type of result?

[@Enterfea]

Hey Conner! I assume that by "thread" you mean the "coil" that is around the bolt that you "thread in" into the nut - right? I'm asking since the thread may be rather "rich in meaning" :) If this is what you are asking, my usual reply is - I don't calculate thread in FEA. I would rather calculate loads in the bolt, and then check the thread manually (well, the bolt really, as it is done in accordance to EN 1993-1-8). There are many problems in FEA regarding the bolts (computing is insane, convergence, amount of needed elements, but also problems with analyzing outcomes) that make me unwilling to do the thread FEA... I won't be able to help you beyond that, maybe I will only ask if you are using nonlinear material - as the first thread may yield to "redistribute" the load to the "deeper ones", although you are mentioning strains not stresses, so i guess you already did that. Allt he best! Ł

[@araglar]

@@Enterfea, thank you very much! I love your videos.

[@Enterfea]

@@araglar Thanks Mate!

[@Iboxx]

My key concern with singularities is as following: if the stress result provided by my FEA in an inside sharp corner is wrong due to a singularity and this is confirmed by the results diverging as the mesh is refined, how do I find out what the real stress is?

[@Enterfea]

Well... this is a tricky thing, because a perfectly sharp inside corner is not a good idea in the first place. I mean, if the region is not that stresses, you don't have fatigue and all that, this may work, but usually you should make some sort of rounded corner, simply not to concentrate stresses (this is a real thing after all). So if you worry about a given spot, then doing a rounded corner makes sense.

[@eartheartbaratheon791]

Hi Łukasz, I know it's an offtopic but I'm having a hard time sarching for S-N curves of structural steels. Do you know some sources where I could find those? (s235, 355, 460 and so on). I'd really appreciate suggestions!

[@Enterfea]

Hey, try in Eurocode EN 1993-1-9

[@dmytrosky]

Thanks😄

[@Enterfea]

Sure, I'm so glad that you like it!

[@vinothvino5870]

Sir I am Doing Static Structural Analysis. By using more refinement the stress values increasing why and how to reduce the stress in small areas. Please help me sir

[@Enterfea]

Ach Mate, I would love to help you, but I've learned that vague questions don't really have an answer. Let's be realistic here - I have no idea what you are doing, how the model is loaded nor supported, where the problem is etc. So whatever I would write here would be a blind guess... and since you could believe that I know what I'm doing, you could believe in my blind guess regardless if it would be a good one or a bad one... and I could get you in trouble this way, while full of good intentions. It's really hard to help solve technical engineering problems, and doing that in the comment section seems literally impossible I'm afraid :(

[@AwestrikeFearofGods]

48:13 What a cruel act, to use such vocabulary with a Pole. I've heard "spurious" fewer than 5 times in my life.

[@Enterfea]

I definitely learned something there :)

[@ΜΙΧΑΗΛΚΑΤΤΗΣ]

if the material yields there will be no singularity

[@Enterfea]

Hey! I never pondered on this theoretically, but without a doubt, yielding solves the "crazily high stresses problem". All the best!

[@ΜΙΧΑΗΛΚΑΤΤΗΣ]

@@Enterfea Thank you for the reply! . I'v found that perfect elasto-plastic material will not have singularities. But if some hardening is involved then there is still some problem. Also logarithmic scale in the stress colormap is much better visualisation when singularities exist

[@Enterfea]

@@ΜΙΧΑΗΛΚΑΤΤΗΣ Hey! Well, if you have a perfectly plastic material the stress simply can't increase above yield, so obviously this solves the problem of "higher stresses' :) Also, I rarely use log scales to anything, I feel they make things harder to understand and interpret... especially in color scales. But perhaps we simply do different models, and we have different experiences :)