Dear Dan Ko, your request have been registered and saved in the recor......wops, dropped it. We will instead recommend you "This week with John Oliver" and "Mr Tfue" videos with a direct wormhole to both "Keeping up with the Kardashians" and "Instant regret" playlist, Kind regards - Yotbe, probably.
yeah and you will see this video recommended over and over instead of newer video's from same channel:/ i've watched this channel for a long time yet this video only now popped up in the side list after watching some other 3d print related stuff..also on my youtube recommandation page are only video's i've watched or a good 90% of them..epic fail..
Love to see proper engineering applied to design! As you have pointed out, the MAIN introduction of error in the approximation is the assumption about the boundary conditions and of course material properties, PLA under load behaves highly non-linear and, as pointed out, the FDM technology causes the resulting material to behave anisotropically. This means that these tools cannot and should not be used for analyzing absolute stress distribution or decision making due to absolute stresses. That said, comparative studies can be carried out. Your discussions about proper boundary conditions are superb! Keep up the good work! The makers world needs more high quality videos like this.
Border conditions can be assessed by destructive testing, while taking in account and adding to the database the conditions the 3d printed part was made. Although that testing is not made under any industry standard, it is valuable data though, because it considers the worst case, real life scenario.
As a previous engineer pretty familiar with FEM (ANSYS) you did an EXCELLENT job with this video. Not sure about Fusion but you could have simulated the wall with a fixed constraint to not allow movement in the -X direction on the back of the support surface. Also, you might want to include design constraints such as being able to easily mount the bracket. For instance, your design makes it very difficult to securely attach that center and bottom hole with a Phillip's screw as the bracket would impede the screwdriver. Love that fact you correlated the analysis results with actual testing.
THIS IS THE VIDEO THAT I WAS HOPING FOR!!! I had to learn about stress analysis in Fusion 360 for my 3d printing designs. It will confirm my intuition before starting a print! I will be able to compare design A vs B vs C vs D and only print the two bests ones!!!! Thank you so much Stephan Alex from Québec, Canada
@@TheGoodoftheLand The problem with commenters like yourself, is that you don't take time to listen to information in the actual video, and rather bash a comment you don't agree with. Or, in this case, defend a stupid question. Because the reason why "using a friction less surface is bad" , is given directly after he applied it in the video @9:53. Just because you don't know about a topic, doesn't mean no one does.
@@rikdol This is an informative video for the general public. People like myself watch an learn from it. The question I asked may sound stupid from the point of view of someone who is more familiar with the topic, but for me, it is totally reasonable. I watched the entire video and the reference @9:53 still didn't clear my confusion about this point so I was proactive and asked for clarification. If you know better, please go ahead and explain this point.
@haytham hakla a frictionless surface will be unable to move away from the wall. Earlier in the video ( @1:50, look closely to the bottom part) it shows that the part is bending. When a surface is set to be frictionless, it will be unable to move during a stresstest. So a test like this for stress and bending or warping of material under high load, won't be realistic. Another example of the part moving away from the wall is @20:00, circled in red..
I love playing with the stress analysis features in Fusion 360, and this was a very interesting video. But I am disappointed that the end result was "make it thicker", I was hoping you'd do some kind of automatic structural optimisation. You don't need finite state analysis to tell you that a thicker part can take higher load then the original.
Oooh, you got some o' that stronger, yellow filament! Nice! If it's possible in Fusion 360, make sure your fillets are curvature continuous, otherwise there might be imperceptible discontinuities that can be treated like cracks. Another important thing to consider is controlling the failures. If a failure at spot X would be dangerous, but a failure at spot Y would be safe, make sure there's a higher probability of failure at Y.
I'd love to see Joel do a video himself where he makes simulations himself. It's always useful to have another set of eyes from a different background for design problems like this. 3d printing has been an interesting area where you don't need engineering to make things work, but learning tools from the engineering tool box can help you so much.
There's so much more to FEM than people think. I dare to say it's one of the most underestimated parts of engineering, because people assume that "pretty colorful images" guarantee accurate and realistic results. Even if you choose the right load case and boundary conditions you might still use the wrong element type or solver and get numerical artifacts without even knowing ( if you don't know what to look for).
What a great video!! I'm so glad you included the lesson about the FEA constraints at the end! This has the greatest influence on the analysis results. I think normally, fixed supports for bolt holes is not a good assumption because they should allow free rotation with linear support at bolt head(Fixed support makes part artificially stiffer) But looks like your failure correlated well!
Just watched one on topology optimisation and all the way through was thinking 'Sort your boundary conditions out!"... Nice to see him come back to that at the end of this video.
Great video but I have a suggestion. This is basically "euler bernoulli beam theory" By increasing the area of inertia the structural strength increases. You can test this by flexing a ruler. The side with the greater inertia is harder to flex. My suggest is that the middle support in the triangle should be flipped 90 degrees to increase its inertia. But anyways this was a great solution and a great video
This is exactly the kind of video I was looking for, thanks! I am new to mechanical design, but had heard of FEA in the past, and wondered how difficult it would be to use for designing my functional parts.
Stefan, Very nice video. You certainly improved on the original design and taught us all a few things about the finite analysis package. However I would like to point out that the original testing concept was flawed in order to simplify the testing. The final loading of the brackets was to occur with weights more or less equally placed on the front and rear rails. Consequently this would place greater shear forces on the top inner corner of the bracket. I suggest that under these conditions the only point to fail in the bracket will be that same top left corner. and I also presume that the ultimate load would be less than half of the loads that you measured (probably a lot less). This mode of loading could be readily tested by placing a light weight cylinder on top of two wooden blocks in the place of the rails and then loading the cylinder using the same strap tensioner. The finite element simulation should also be changed to involve the two weights (and don't forget the small twisting force due to the filament spools forcing the wooden rails apart).
OMG! Noch einmal! The German engineering excellence is not a myth! Rigorous demo. Love it. Thanks a lot Stefan! I used to work in Lahr, Germany, 50 km away from Strasbourg where I live. I did appreciate the seriousness of German engineers (IT sectors) but never understood their humor - still wondering whether it exists ;) - my German language knowledge being so wanting...
I'm glad you did this video. It's very helpful to learn how to make your parts stronger. When Joel first did his brackets, it was pretty clear that they weren't the best design. Simply looking at rafters for roof construction demonstrates an optimal design of a truss. Regardless, I did appreciate the basic idea of making shelf brackets for my filament rolls. They work great.
+CNC Kitchen I would like to see you try adding a thin section in the middle (like an I beam) with strengthened "cutouts" for the screws instead of thickening the material. If you print from the base and up this should avoid having to add support material.
Adding a very minor relief cut on the back surface of the bracket near the bottom screw would allow you to create a small constraint surface just at the bottom corner. This would prevent unrealistic results due to constraining the entire surface.
@@CNCKitchen love your channel by the way, especially your methodical but garage-doable testing regiments! Also your video on heat resistance testing of various materials led me to be able to powder coat PLA! More on that another time
Many of us pointed Joel to your video on this subject when he was making those brackets :-) On his second video on it he mentioned you and that video. I'm so happy you are doing a follow up to it. Thank you.
Ich weis nicht was besser ist - der Inhalt, oder die Machart und die Umsetzung der Videos. Ich denke, weil hier alles in jeder Hinsicht perfekt ist :) i love it, you get a Hug & a Kiss :D
Cool video. This is the difference between brute force engineering vs mathematical. They both work. One is typically quicker to market, and the other just works.
Fascinating video! I have already identified ways I can incorporate this in my workflow. The real danger is getting to the point where I’m over-engineering my shop furniture!
I love your videos, we all learn something new from them. Correct me if i am wrong but your test scenario here is more suited for hanger and not shelf bracket. Shelf bracket should have load concentrated on the middle of horizontal part (if you simplify it with one force) or distributed evenly over entire length of horizontal part of the bracket. You would get other results then. In any case awesome video and a lot learned about fusion 360
It would be awesome if (assuming they haven't done this already) they could simulate the FDM part using their own slicer. Design part -> Slice -> generate GCode -> import theoretical FDM part with infill properties -> do FEA analysis. Bonus if this is automatic somehow!
iirc, fusion will also do generative design, where you tell it what geometry you want, and where forces will be, and it will generate an optimized structure to support it.
Thank you for sharing this tutorial! I can now get even further over my head resulting in even more fun :) I hope you get feeling better soon. Worst case, punt: extra spicy doner kebap, chips with extra paprika, and a whole bottle of rot trocken... unless you have chicken soup over there.
This is really awesome. What would be the effect of adding, say 2mm thick orthogonal strips which protrude from the cross member girders. This would make them more of a cross shape in cross section, and I think (in my special custom buckling simulation, in my brain) that configuration would spread the load stress more evenly and perhaps improve buckling load resistance. Great video thanks!
Three quick observations: How did you get a scew driver to the middle fixing hole? (Every bracket designer forgets this) Would changing the bottom member to a triangular cross section affect the result? The bottom member has a high slenderness with respect to the plane of stress; and narrower, deeper section may have performed better. Great video, I really appreciate your work.
Optimizing for 3D printing is going to be very big in the future, as we start to make more and more parts using 3D printing, the ability to get a stronger part at the same weight is going to be crucial, and we will need clever tools to iterate the designs. The organic optimizing tools i have seen often leave fragile brace members and thus are not usable in real life. The optimization has to take into account fragility. Buckminster Fuller theorized that using his octet truss to make octet trusses would yield an almost infinite strength to weight ratio, but each support member of the braces gets thinner and more fragile, and thus the structure would collapse like a house of cards.
Some FEA software like ABAQUS can model shell elements e.g. only modelling elements on the surface of the part, and then you define the thickness with a separate parameter. That and a way to approximate part infill would be really nice in Fusion.
Having specific shell type mesh and simulations is partly for ease of computation, and partly for accuracy. A shell mesh generally just covers the outside surface of the part, so the mesh is less complex to model the same part, or you can have a finer mesh for the same computational cost. It generalises it a bit using just a surface mesh e.g. it may assume the variation in stress between the inside and outside of the shell is negligible, which it is if the shell thickness is low compared to the size of the overall model. In FEA, it's important to make the mesh variation smooth e.g. for a given element (smallest 3D shapes in the mesh formed between mesh nodes) you don't want the difference in length of the edges to be too much, something like 1:3 (ideally it should be as close to 1 as possible) ratio between any two edges on an element, otherwise the results become quite innacurate. If you modelled the entire 3-dimensional shell, you may have small edges between the inside and outside of the model, but much longer ones on the surfaces of the objects. So in order to get an accurate simulation, all the edges would have to be similar in length to the ones passing through the shell, and that could make the mesh way more complex again, for not much gain in simulation accuracy. Kind of like in the video where Stefan mentions you should have no less than two elements over the thickness of the part in order to accurately model it with this type of mesh. I'm not sure if one element thickness is good enough for a shell assumption (I'm not an expert in FEA), but even so a very fine mesh would be required. This is basically what a shell type mesh property is for, greatly reducing the computational effort (and user effort) to make an accurate model of this type. In terms of the infill, I was only really thinking of having an internal mesh with different material properties to the surface, modelling the infill exactly would be very difficult unless it could be done in a slicer or with the G-code or something. I'm just thinking the internal mesh could have material properties which better models infill e.g. with 10% infill the Youngs modulus is only 10% for the infill (I'm not exactly sure that example is correct as again I'm not an expert in FEA but you get the idea). I suppose overall the mesh may be just as complex as these ones but I think it would be far more representative. Rather than modelling a solid part, modelling the surface to have diferent properties than the infill is essentially what I'm advocating.
I'm assuming the elements in Fusion are solid continuum elements, since there's no way to change the element type from what I can see. The modelling feature in Fusion evidently is simplified for basic simulations and ease of use. Shell elements are solved differently to solid continuum elements and are better suited to that type of structure. Since practically all 3D prints are shell type structures, something built into Fusion which could model that type of structure accurately would be nice. I'm not an expert but have used FEA to an in depth level for a specific application. Perhaps the standard elements are suitable for that type of structure, I don't know. I do know that in ABAQUS which I have used before, there are standard continuum elements and shell continuum elements, so evidently there is a need for them. Reading into it a bit more, the shell continuum elements have a plane stress assumption, so only the stresses along the surface are measured and stresses normal to the surface are assumed to be neglible. The elements can distinguish the stress variation through the thickness of the part however, with more shell element layers giving a more accurate picture, although that's probably not very important since the shells are generally quite thin on 3D prints. I'm not sure if it would be suitable to all types of model actually. For the infill, I don't see how it would see it as an incompressible solid if the infill was represented as a mesh. If you gave the infill mesh different bulk material properties which reflected it's behaviour e.g. such as Stefans video on infill strength. With low infill settings, the lattice structure is large compared to the overall object and could be significant, but low infill doesn't affect part strength much as we all know. Also at higher infill settings, the lattice is more dense and begins to act as a continuum, so a solid mesh should be a fairly accurate model.
In retrospect, I think I was proposing something (using a shell type mesh to possibly model a 3d printed part more accurately) which seemed to make sense at first but actually might not be correct since I'm not an expert in FEA like I've mentioned, but I have used FEA to understand how it works. Ultimately the proposal is that Autodesk adds some feature, to either fusion or another piece of software, which more accurately and efficiently models 3D printed objects. If they could make a tool which generates a mesh for the shell of the object, then generates a mesh on the inside of the part for the infill with one click and simple data e.g. shell thickness, Youngs Modulus etc. A step further would be to model the anisotropic structure of a 3D printed part i.e a 3D printed part has different strength as well as other properties depending on the direction it is printed. If they gave you a little arrow and allowed you to point it in the direction the part will be printed, then it could apply the properties to the model based on that, perhaps even structure the mesh to be in line with the layer lines of the print. You'll never be able to model a 3D printed part perfectly, but each of the things above would bring it one step closer to being accurate. You could defnitely model a 3D printed part using existing FEA software, but it would be time consuming and laborious. If someone could make something that could automate the above, that would be really useful for the 3D printing community. I guess there hasn't really been any incentive for that type of modelling until recently, since 3D printing has mostly been used for prototypes and not fully functioning parts. Maybe this software exists already although I hope it's not behind a high paywall.
Hello Stefan! Nicely explained! Maybe you could add a fixed contraint in a small area just at the back of the 3 bolts also without the frictionless support.
Really cool. I think I may like 360 over Creo. I think a visit to home depot would have been useful. Concave ridges are super nice and let you take material from one side and put it on the side under compression (all the trusses are rounded). WaaaY more time to draft, though. Filets tend to break.
2:23 "We are working with mostly that are hollow and Fusion360 is just not made for that." YET. I think they will add a most precise simulation for FDM print in the future. Great video!
This finite element analysis simulation assumes the material properties are isotropic and structurally homogeneous. You'd need to model layer alignment, layer delamination, and shell and infill to be accurate and that would require huge processing power.
Thanks for creating a very informative instructional video. Personally, I think that true optimisation would mean creating a design of equivalent mass to the original design but stronger, or the same strength with less mass but perhaps I'm just being very picky.
Congratulations for your work! Have you considered increase the area moment of inertia of the lower member instead of just increase its thickness. I mean, you could try designing a "T" cross section, for example. That would reduce the normal stress at the lower member while keeping a lower mass. Again, nice work!
Very useful video! I'm always curious about exceeding material strength in compression for malleable materials. Our PhD analyst is always asking me to beef up areas exceeding yield in compression, but it has been awhile since I've seen material compressed to failure. Thank you!
I tried making a wheel segment with spokes support, and it did amazingly with tiny amount of material and just three 2mm wide spokes. Most of the load was in pure tension on the spokes which 3d printed plastics handle perfectly well, and the "rim" was in modest compressive load so even 3mm thickness was plenty to only result in minimal buckling. The flat surface is effectively suspended between two points, at the wall and at the wheel rim, a small arc could be added to reinforce that as well. A little geometry goes a long way.
You should have left the middle screw further up where it was before anyway. It would help a lot with the strength just the screwing might be a little harder. Auf Wiedersehen!
Since there are different printing orientations in 3D printing process, every printing orientation would result in different strength, it is necessary to take this factor into consideration.
After seeing the sudden and extreme failure of one of the parts, it occurs to me that there would be significant value in designing an impending failure signal into the part. The classic engineering concept of the failure signal was based upon the materials used, such as steel or wood that bends, groans, or otherwise signals when it is close to failure. However, certain types of optimization may inadvertently take away these signals. Understanding the consequences of the optimization steps often involves careful inspection of the performance in many of the design goal dimensions.
Hold that thought! Autodesk says Starting on September 6, 2022: The ability to run simulation studies on a local computer will no longer be an option for simulation solving. All simulation study types within the simulation workspace of Fusion 360 will only have the cloud-solving option. And it will probably be a lot slower than an i9.
In my limited knowledge, I have learnt that different printing parameters change the properties of the final part. Does the simulation take all of that into account?
Given your failure mode, what if you changed the upper arm, such that you used a significantly longer screw, which went into that upper arm, with the head of the screw ending up 25mm or more away from the wall? Having a steel rod present there would help with any buckling and having the support directly in line with the tension force would make pulling it away from the wall much harder.
That was great content please do more simulations tutorial while kipping the tutorials at a level that a beginner to intermediate users can understand... there are not enough tutorials explaining those powerful tools from fusion 360
The added thickness to the bottom strut had the most impact, followed by moving the the web to create two triangles. Now if you would have added 5mm to the entire outer perimeter the web and wall support, and then print it solid...... This is a shelf bracket so I didn't really understand why skimp on the thickness in the first place, or even try to save material. Sure it will print faster but, oh well.. What I would have done is create a T-slot bracket for the wall, print that completely solid. Create a full triangle, no webbing, with a single hole in the center of the bracket to hang stuff on (or slide a pole through for more options), to fit into the T-slot but print that with infill 25% or less. The T-slot would be stronger/thicker (not to mention printed solid) eliminating stress points on the bracket and there would be added functionality with a T-slot bracket on the wall. If you decide you don't want a shelf there later on, well there are tons of accessories you can make and stick into a T-slot. Great video BTW,.You are one of the few youtubers that I've never regretted subscribing to.
Great video. Kind of hard to use the middle screw hole in the last design. Not much room to fit either the screw or screwdriver without hitting the stabilising part.
I am just wondering about the material and process data due to the 3D printing process. It seems that the FEM simulations are performed with full material, but printer settings such as filling, wall thickness ,or layer height are not considered. In this case, why do the simulation results roughly match the test results? Thank you very much for your help!
