I've read over the years that infill adds little to part strength and that it's mostly down to perimeters. The experimental one with a hollowed oval inside effectually doubles the perimeters. All that said , reading things on the internet is one thing while doing experiments like you are is on another level. I'm digging this series!
What I might try is to take the oval and try different shapes instead, maybe just slight rounded vs very rounded, all of them must be the same weight to have an apples-apples contest. I could further increase the wall thickness by changing the inside corner radius less and less and redistributing that material. I'm glad you're enjoying it, I'll have lots more to come!
I routinely cut internal tubes out around through holes for bolts on molds that I design so that it adds 3 to 6 extra perimeters around those through holes so that clamping the bolts down at a higher torque doesn't crush the print. A few extra parameters can be incredible at adding strength
From physics standpoint - the closer to you get to the middle of part (or the bending axis) the less strength you gain with infill. When bending a part, majority of forces are on it's skin, and because of that, hollow parts are industry standard - hollow bike tubes, hollow car's bodies and thin metal sheets being stamped to mimick a hollow part. You also made an improvement of adding more material to the corners, which are further from the middle and thus provide more % of strength, but I'm not entirely sure that in rectangular part the inner hollow should optimally be an oval - probably something like very rounded rectangle would be better. Infills generally are wasteful - making some structural ridges just below the surface is sufficient for most of uses.
That makes them stiffer but generally doesn't raise failure. A hollow tube will deflect less than a comparable solid steel rod, but the tube will still fail sooner.
@ObservationofLimits if both tubes are of comparable weight, the hollow tube will be stronger until we reach the point where buckling becomes the main failure point
@@FierceFire14 a tube is actually a terrible crossection for bending loads exactly due to ovalization buckling failure mode, and internal reinforcement (like spars or infill) can help with this greatly, less so for box sections. From pure bending standpoint, I-beam is the best of course. A tube is unrivalled for torsion.
doesn't ikea put a honeycomb basically of cardboard inside of its tables and desks to improve rigidity? In that case the honeycomb is rigid perpendicular to the load, but I feel like it definitely does something compared to just hollow particle boards
@@Negi2468 They do, but mostly because of reasons mentioned above plus cost. IKEA wants to use as little of expensive material - outside veneers - as possible without it breaking easily. On top of that, very large flat panels need to be rigid to not only bending but also compression, and that's why cardboard infill is used. They'd also easily buckle or break at that size and thickness.
That's why Injection Molded parts achieve a greater ratio between resistance per weight (material used) - it is not just the process, it is also because there is no material being wasted in infill, the whole thing goes to the walls. When you need reinforcement, you put ribs. Things are better planned. We should see infill as a wasteful lazy way of giving some internal structure and achieve easy printability - but it is not the best usage of material. EXCELLENT VIDEO! Thanks!
Injected molded parts perform much worse in this regard in my experience because the process is not conducive to super thick walls. Printed parts with 2.5+ mm wall thickness and low infill do a lot better.
@@ry7hym It provides compressive strength and can give trellis type rigidity benefits. Interestingly, lightning infill actually provides the most added strength in my experience, if you use it with infill multiplier 2 so it's a single closed path, by auto generating "ribs".
For your torsion testing, the wrench end of the part should be supported as well so you are purely twisting the part along a single axis. Right now, you will be loading it in multiple directions while trying to apply the torque with the wrench. Loving the in-depth testing, it's great!
the density of gyroid infill can be continuously varied without any sharp transitions. today's slicers don't implement it but it'd be a neat way to redistribute material from the inside towards the surface
I'd like to see that, ideally we can get to a sort of Bone structure, which can flex well and also is resistant to breakage. We should also be able to set areas to have higher density to be more rigid etc. I think we have a few more years of FDM and then we'll be using different types of 3D printers which won't have the same limitations...
@@NeedItMakeIt exactly, for example corners even on decorative prints need extra strength because they'll experience all the impact forces when a part is dropped. adaptive cubic infill attempts to do this but the density jumps too much between levels, you can't edit the distance-to-wall density curve, and overall it's not the best load carrying structure
Why can't it be done with an algorithm that reads hand-placed weights (points in space that are placed and assigned a weight value) and then interpolates between different scales of infill structure based on the weights? And the user could also specify custom infill grid shapes that can be tiles and applied at various scales.
I think you're touching one of the main point for next gen slicer from an engineering point of view : modeling the inside of parts! That is a very interesting video!
It's an old comment, but incase you are interested: take a look at the prusa mini 3d printable parts (probably also other models, havent checked), specially inside of the models... They have included internal ribs that force the slicer to include interior perimeters inside the parts, adding hugely to the strength. I'm sure many designers do this, but it's a very good example. Would be awesome to have an option to add these directly on the slicer... a bit complicated as it doesn't know how the part will receive stress.
I've always printed my infill lines at 2x nozzle size thinking it adds strength being it should be less flexible. Be an interesting thing to actually test
Flexibility does not equal weekness. Remember that a bending branch lasts longer in the wind than a stiff one. Stiffness can make then much much weaker to "shock" forces.
Add a large nut, and a threaded bolt to the end of your lever and either add a motor or a hand crank. This way you can apply a very consistant force. If you add a stepper motor you could even calculate the number of turns it took at breaking point. I know its another upgrade but I think it would provide a better result than your hand would. Measuring with consistency is best especially with sometimes wildly different results.
I was thinking about switching to a motor and hydraulic cylinder setup, do you think that could work? I'm happy to do the upgrades, if the channel grows enough I will buy some proper testing equipment as well... I think I'm a ways off at the moment.
@@NeedItMakeItI wouldnt bother with hydraulics at this stage. Just get some adjustable construction support rods, they're dirt cheap, super rugged and usually come in trapezoidal thread variants. With that and a big wrench, or crowbar, you can get a lot of force going quite easily.
@@Argosh My idea would go with a motor like @newmon and a pulley system. It could be easily built under the wooden structure and would provide quite good measure for deflection before deform/breakage depending on pulley ratio even with regular motor, just counting revs. Stepper for super accurate. Torsion test could use its own frame to separate bending force from affecting the test and maybe use the same motor, pulley(lever), rev/pos counter to keep amount of parts lower.
Good news for you: this video got recommended to me, who's not usually 3D printing video watcher. I'm usually watching engineering or machining videos.
Most of the strength in any structural part is in the surface. The polish glass rod experiment best taught in the the engineering causes of the 60s and 70s, demonstrated this phenomenon. Glass cutting shows just how fundamental this is. Layer lines are effectively scratches from a tensile strength perspective. The strength of all plastic materials is Van de Waals forces and a consequence of polymer molecules length and shape. Infill effects compression forces that cause members to buckle. Cura provides the ability to insert members/internal walls to compensate for compression...
This is a very common and useful technique for designing mechanical parts. The further away the material is from the neutral axis the more it will contribute to the bending strength of the part (you can possibly think of it like torque), so for the best bending strength you want as much mass as far away from the neutral axis as possible. Along the neutral axis (in a simple rectangular part that would be right in the middle of its thickness) the stress and strain is zero and the further away from this axis you get the greater the stress and strain become. Steel H or I beams are a good example. When used in the right orientation they have very little material in the middle, close to the neutral axis but they have lots of material further out. They are much stronger than a solid rectangular beam of the same mass because of this more optimal material usage. They also have a very good strength to weight ratio, much higher than if you used a solid beam with the same external dimensions whilst also having most of the strength of the solid beam since in a solid beam most of the mass doesn’t add much strength. So when designing parts to be used in bending it is best to use the largest part you can (external dimensions) and distribute the mass as far from the neutral axis as possible. This is also why two stacked beams won’t be as strong as a single thicker beam, each of the stacked beams bends independently and therefore the maximum distance material is from the neutral axis of each beam is half of its thickness whereas for a single thicker beam the maximum distance material is from the neutral axis is doubled so the material contributes much more to the strength.
There's one option that I'd love to see tested. Cura has an "alternate extra perimeter" option. Every other layer gets an extra wall so that the infill is getting sandwiched between walls. In theory it would make a stronger part, but I don't know if it's actually significant.
About UA-cam algorithm: I've been watching many of your videos - also the ones about different joints... BUT, I need to tell you: I found the topic unpractical. I seriously doubt I will ever design parts that need joints like that. I prefer just having a bigger printer and print parts in one piece. This could be one of the reasons for worst performance of the two past videos, you need to investigate. Make a Community Post with a poll: "Have you watched this video? (link) Would you ever use one of these joints in your models?" Good luck my friend!
Printing direction matters a lot if you want the best strength to weight ratio. Just like fiber-glassing an aircraft. There are two type. Unidirectional, and Bidirectional fiberglass over foam core. Fill will be the foam fore that brings rigidity and transfers the load from one side to the next. Since I mostly print engineering parts (gear, pulleys, enclosures,..), I need the strongest part. Hence, I never use the parts fan and don't exceed 50mm/s speed. ex. PETG @N240ºC/B75ºC will not come apart at the layer. I found white PETG to be the best regardless of the manufacturer. I guess they use Titanium oxide pigments like any other white paint. That has been my experience I thought I share.
The web, while doing little to the overall strength, is probably what lead to the increased stiffness. In order to bend your part needs to flatten the inside oval, which is opposed in tension by horizontally webbing and compression by vertical webbing. I suspect you could leave the webbing out, though, and still have roughly the same performance, albeit a little less stiffness.
Great work on doing practical testing! Keep going! On the topic of the video, your results are not surprising at all. For both bending and torsion the more material your have away from the neutral axis, the stiffer the part will be. For bending you're increasing the moment of inertia of the part and for torsion you're increasing the polar moment of inertia. Check out the bending and torsion stress formulas. Running a FEA analysis should generate very similar increases in stiffness. While the macro shapes can easily be computed by FEA software (or even hand calcs in simple bending or torsion cases), the micro structure (interaction of intra-layer bonding) cannot easily be computed. If you continue with testing, consider going deeper in that area. Perhaps do some comparisons between FEA analysis, hand calcs and practical tests to see if you can determine a generic "correction factor" which can be applied to FEA analysis which will reflect the actual part strength. DM me if you want to chat further on this topic.
I’ve come to really appreciate your content because you’re probing in areas of 3d printing that others are not. It’s really hard to find original content anymore. These tests have been very helpful in informing my own 3d designs. Thanks!
Why is nobody talking about the fact that this man has the incredible skill of writing numbers and creating diagrams *upside down*? That hurt my brain to experience.
LOL. I think I have better printing upside down... I think this skill was developed mainly when I was teaching my son how to read and write, I was always on the opposite side of the table.
From my engineer view and experience. What structural properties you get depends on the material. Materials which are more flexible benefit from filling the volume. While more rigid (brittle) materials benefit from wall thickness. You can imagine the difference between a steel square pipe and same size laminated wood beam. I'm actually insane enough to design my infill in CAD. I have even written gcode by hand to achieve specific things. I'm annoyed that slicers don't actually allow me to have finer control over generated geometry, like directly adding planes or scripting layer behavior. Because I run flashforge, I actually use both Orca and Flashprint, because they do different things better, and if I just want to do quick test I choose the one I know to be easier to do that. However in engineering mindset your desired properties are always derived 1st from design, 2nd from material. But because my background is in welded steel products, I naturally account for rolling direction and profile geometry, as they affect things a lot. They are much like print direction and layer joining. But you will not find any sharp edges in my designs, unless it has a specific use.
i always avoid generated infill at all cost. i usually cut/hollow the vertical faces with hexagons instead of circles (to avoid overhangs). i also usually design the supports if I really need them so it prints faster with less waste and better finish. there's many tricks you can do when designing for FDM, I even automated some of them as features in my cad software.
Thanks for the extra test, I appreciate the work. I even went back to your last video to get the comparison. PETG-CF (red) Continuous with infill: 30,6 Nm PETG-CF (black) Continuous with infill: 48,4 Nm So, an Improvement of 58,17% 😮 in the torque Test, more that I expected. Also the failure mode changed to "fractured in 3 Parts" indicates that the black Filament has more Stiffness, imo. I heared before that the filament color can change the products parameters, on other filament test runs by CNC Kitchen or Makers Muse they tested on test cupons, but on a more practical part the diffrence seems quite huge. ok the Samplesize is 1/1 but the trend is plausible imo., because all the other Red Parts in the Joint Test where the weak link too.
Oh wow, that was unexpected! Thanks, this actually may be a game changer to be honest - if we are trying to make the tough parts. Great results and successful experiment! One observation from my side: at 7:25 You may see the cracks started to form before the part went flying. I am wondering if those cracks started happening after some threshold or prior to that....? Reason: when You have infill it's somewhat flexible thus the *amount* of times You can put a high load is not small but if You have smth more rigid it might be tricky - if, let's say, You go to 80% of max load and those cracks start to form this means we decrease the max value with every load put. It's not _necessarily_ the case - I am just thinking out loud. I would propose the durability test additionally to the one You've already done - and if that will be a success as well - we can basically ditch the infill (in 90+% of the cases). Thanks again - it was really interesting :)
I think it's not far from the direction we need to go, it needs some refinement of course. Yes I messed up the final test on that part because it was too hard to force the lever down, I'd need a longer lever. I like the idea of a durability test, that's a great idea actually. I could setup a part to have force applied and then removed and continue this until it fails by no longer being able to take load by deformation or breakage. I think there is some potential here for sure, I think we need to do some more research and testing. I also want to test extremely sparce infill with far increased thickness. And I have a few more ideas too and maybe there is a way to combine all of the concepts into something even better.
Great Video and testing. I have some inputs on your methods, as Ive done 4 Point Bending on Composite Sandwiches for my Bachelor Thesis: I would choose to repeat the Test 5 times per sample, as a big scattering of examples is to be expected. (But you have Said that already). The Text to distinguish the samples Looks neat, but creates a Point of failure. 4PB is sometimes used over 3 PB, because you get an area of the sample, which is just loaded with shear force, as the bending Moment in not present inbetween the cylinders for force intruduction. Your cylinders should be further apart. Thanks for your contribution to the 3d printing community
As always kudos for actually experimentally measure results, and trying new variations. Don't beat yourself too hard about the last videos not having the same performance of the previous ones, the fatigue of a topic is a real thing. Breakthrough videos of an interesting topic could have an increase in views due to high engagement in the algorithm, but people rarely stick around about continuous testing of the same topic, maybe switch around the topics a bit between each release, and then make "series" of videos by topic. At the same time wildly different content of topics also deter the core audience so it's a gamble push to far away the "main" audience, so the main takeaway is stay in your lane, but have a little more breathing room between topics continuous "main" topics.
Hexagon, strait and triangle infill should help make a part ridgid as well. Most of the other infill like gyroid are pretty but when you look at it, it's just there to fill a void not add strength.
Mathematically speaking, pound for pound your best options are either a large hollow cylinder or I beam for holding loads. For your tests, an i beam would work best with most of the length facing "up" and enough width at the top and bottom to prevent torquing, you wean the whole load straight on the support. Conceptually just thing about how much material force would need to move in order to bend it. The more material the better right? If you can imagine drawing lines through it at a perfect 90 degrees from the earth, that's how much material is holding the weight. The more lines that run through a cross section, the more material would need to be moved.
If you look into beam strength in bending, and why I-beams were created, you'll see that having more material under tension/compression at the furthest extents from the neutral plane(axis, for torsion) increases stiffness. But if stress goes too high, you can get a split down the web of an I-beam (on that neutral plane) so I think it would be worth comparing 2-3 printed I-beams of similar mass, but with different web:flange ratios. Chances are they're not worth the complexity to model/print in comparison to a shelled model, or one with minimal infill, but it may be a good demonstration. But I suspect they would show that prints have need of both the largest possible outer beam dimensions (flange on an I-beam; floor/ceiling of your experimental beam), as well as a sufficient bond between the neutral layers.
Your vertical should have a radius and feather into the inside walls. During the twist test you can see that the vertical line cracks where it enters the inside wall. A small radius would displace that shearing point.
Really liking this series! Looking forward for the optimized version. One thing that would be worth mentioning is if the new approach has any effect on the print time. Without infill I would expect the print could go even faster. Faster, stronger, same material cost, that would be a win win win.
I have always been fascinated by taking 7 equal diameter "tubes' and putting them together. One of them always gravitates toward the center, and the other 6 form a neat "structure" around the center tube. (Try this with 7 pencils to see what I mean). You can see something that happens in nature - a Hexagon! Now - as far as your infill goes - draw a straight line through the centers of all those tubes and align ONE of those lines as the vertical support for inside the square. The other two lines should go to the four corners if your object is a true square! I would bet that this type of "infill" would add even more strength!
I truly think this is valuable for specific use-cases. However, all I've learned for my personal projects is that nearly all methods are far stronger than I require for my application haha
To print mechanical parts it would be nice to have a function in the design program (or the slicer) that is able to calculate the arrangement of infill using finite elements method (FEM). Finally we would get some design similar to bones. In bones we see a full material shell and an infill that is increasingly foamy from outside to inside. Close to joints bones are also more stable. Of course, to construct with FEM we need to know our requirements, the forces acting on the part.
There's a technique I once saw used by a fairly famous entity that allows someone to hollow out a solid piece while minimizing loss of integrity. I remember very little of what it was called or what it was used by, but I do recall they primarily used it for reducing weight, so I'd guess NASA or SpaceX if I had to. The technique was basically using some sort of software to calculate where forces would be distributed inside of a part (when under load in various situations), and just cutting wherever the forces were too minimal to be worth the cost. This allowed them to remove a very significant amount of material with barely any reduction in durability. (I feel like I even remember one case where it *increased* the durability of the part compared to a full solid, but that seems too far-fetched and thus possibly a false memory) This might be a bit of an insane request, considering how little information I have to give, but I'm wondering if it could be applied here, where you can relocate the mass from the infill to increase the density of wherever the software says will get the most use.
If infill is to support top layers, it doesn't matter much, how it is printed, as long as it's fairly dense just below bridging. If infill needs to bear load, it needs to be thick - 0.6-0.8 on 0.4 nozzle - and dense - 50%+ (just changing width changes line spacing as well). In Cura it is possible to print infill with multiple lines, creating loops. Combined with alternating extra wall (Cura) it should produce pretty _solid_ infill. PrusaSlicer honeycomb is also 2 lines wide. ( Or use non-filled plain filament for max strength 😅 )
This is very interesting. I make structural parts. I am doing hydrofoil blades at the moment. I print these hollow and then fill with expanding foam. If I need extra strength, then I put a very narrow slot, 0.1 mm width in the wall, and extending in a few mm, but not all the way through. I am using a 0.6 mm nozzle, so the slot walls in practice touch, i.e. it is like a web but with an open side. This adds to the cross section of that part of the wall, i.e. stronger. I have tried all sorts of different materials. My current favourite is PLA Meta. It has good inter-layer strength, and is relatively easy to print with. And not too expensive. It is a bit stringy so there can be some surface cleaning up to do, And I use light colours - dark colours overheat in strong sunlight and distort.
Great testing! I think the optimal structure would be to have thick walls with all interior corners rounded e.g. 2 mm and inside filled with big honeycomb-like structure with a couple of layers worth of wall thickness. The idea would be to have some internal walls but not pure infill. I would guess a single hexagonal cell should have around 15 mm side length. I'm not sure if the walls of hexagonal structure should also have rounded interior corners. Having rounded corners would make the structure stronger but additional testing would be required to see if it's worth having more material. I agree with your methodology that the mass of each part should be identical.
To avoid the parts flying off in random directions from the testing station, put a box made from some sort of transparent plastic that's big enough to contain the entire testing rig. This way it can still be filmed and the broken plastic won't fly out in a random direction (one of them being your head).
Thank you for doing these! I’m working on designing stuff that needs to have strength so it’s really nice to have someone testing. I’d love to see a video about temporary joinery (like sliding rails, snap in place, etc) if you’re looking for topics. I’m working on something modular that has side pieces that need to slide on and off and I’m having trouble finding ideas that will work well.
You could try minimal Lightning infill with maximum walls. The purpose of the infill would just be to support the roof of the part while printing. And you could do a + shaped interior flange, and maybe make it thicker. Looking forward to the next set of tests.
Loose paper bags to catch/decelerate the sample fragments, like airbags/parachutes; Or lace a cord (twine?) through the ends so that most of the pieces are captive; like a floppy-rail/zip-line. -Loss of spectacle aside,- I think that having more control over where they fly to would allow for a safer (and easier to clean up) test, while also providing some illustration of how much energy each failure releases by how much they affect (or punch through) those controls.
If there's one thing I've noticed, it's that color Greatly affects the strength of some filaments. I've seen it with PLA+ for sure, but I can't say for certain about other filaments though. I would just guess that's the case. With the PLA+ the drastics I've seen are between Black and Silk Silver. The black is so much more rigid and stronger while the Silk Silver is so much more flexible and is significantly weaker (but it does print better/easier).
In cura that I use for printing my parts I am able to add custom nesting models that I then set as modifiers, I then set those embedded parts in areas of high stress, say around areas where I am embedding fixing points or connecting structures between fixing points to then add internal structure by setting these models as areas of infill that are 100% solid to allow me to bulk out areas of the print where nessesary and then use normal infill desity to fill the rest of the part as you normally would. this process then allows maximum dencity where you actualy need it and then sparse where it is needed for te printing structural support only. worth looking into.
i've been doing quadruple wall with lightning infill (7 layer top ironing enabled) since january and im satisfied with speed, strength and surface quality. working on an end table soon
Increase infill line multiplier as cura calls it, and reduce density in half. This will make the infill stiffer and much more structural. I found this while experimenting with structural properties of TPU, which makes the effects immediately a tangibly experience, no breaking fancy tools or setups necessary for crude results.
I liked your last two testing videos! I was very impressed! They reminded me of the tests that CNC Kitchen does. If I had to guess, it's either your thumbnails didn't attract which people or YT is crazy.
I have been loving these testing videos you have done lately. Please keep at it! :) I have been toying with the idea of a printed printer frame and your joints and tests are really helpful with that. Also Mr. URGENT has a point, seeing results with more common, regular filaments would be nice as well, but only if you have the resources for including them in the tests.
When optimizing you need to consider the load profile. For the bending test you want most of the material near the top and bottom, with the minimum sidewalls to enforce spacing. Which is of course not optimum for other load profiles. For torque you'd want a circular tube if the material was isotopic, which FDM prints aren't.
Torsion stress concentrate on the middle of square/rectangular profile members. You should create a cylinder in the middle leaving the corners hollow. The cylinder is the strongest shape under torsional loads. Would be interesting to know how it compares under flexion as well.
To smoothly pull on lever, OR direct pull on gauge, use a boat winch with steel wire rope (cable). can anchor it to table or test press and run through pulleys as needed for space management or "leverage" increase through multiplication. VERY CHEAP AND EASY/SIMPLE! Can get ALOT of pull from it and only 4 bolts mounts it vertically OR horizontally for ease/comfort of turning handle.... OR could use a cheap $99 electric 12vdc 1,500 lb atv winch for steady pull rate.
I'd be curious to see Stefan, CNC Kitchen, recreate these tests you're doing. His motorized screw provides a more consistent and continuous force. He also uses a three-point bar rather than a four-point; fours have different geometric loading and induce stresses differently, there's a literal reason why threes are used in material testing-- stemming from this, your previous tests with the joinery would've also had problems with the evenness of the loading from test to test, a very long diagonal dovetail would've been loaded very differently from a standard perpendicular dovetail, i.e. testing methodology was not actually consistent where one style of joint would've had unfair advantage within the testing setup. Even look at Matthias Wandel's work for how to test properly. I'm sure there's other creators within the relevant space of DIYers and makers, who either have an engineering background or at least an understanding of engineering, who also have proper testing methodology and produce good data. On the previous tests you also had print orientation issues, which should have been tested for. On this test you had design issues with perimeter count, of which should have been accounted for in the design phase; thin-wall with infill is very different from thick-wall without infill. This test also had potential print direction issues. Infill also does very little for this testing, you'd be better off testing infill with tension, compression, and buckling tests; certain design aspects fit certain tests, figure out which match and don't test mismatched sets. If you're going to be doing engineering testing, for the sake of good data and good information, use proper methodology. I want to see more actual engineering within the space, but if things can't be done properly then creators should stick to being entertainment only, we've all seen how bad information spreads like wildfire in DIY-friendly communities where the basis of the majority are undereducated on the relevant subjects when it comes to hard data and testing of materials and designs; this is also why a lot of maker channels don't have testing setups and attempt to scientifically measure aspects, because they're not qualified to speak on the matter and frankly won't invest into understanding how to do it properly. Do your due diligence to properly educate your audience. But to educate your audience, you need to educate yourself first. You need to understand why your methodology is lackluster, in both design and testing, you need to understand when testing actually matters and how to test for it. You need to label your produced testing videos as non-educational and put a disclaimer that results should not be taken as fact. If you presented your testing to a room of formally educated engineers, you'd be laughed out of that room, told to go back to physics 101, and that your testing and design method is nothing more than a science fair experiment that you didn't properly research. Good data is important, but unfortunately an undereducated community cannot recognize bad data, which is why it's your job as the content creator to do better and provide that good data outright. Also, use a blast shield. Not everything will break cleanly. Again, it's a due diligence topic, I'm surprised I haven't seen too many comments harping about personal safety on your videos. I hope you take this as constructive criticismb and not the content creator usual of being an egotistic pissbaby. Everyone has to learn at some point, however people should rather learn before they attempt to educate.
If i really need strength for something i just put something stong inside. I made internal corner joints for some aluminum profiles and put Allen keys inside them for strength (the ones you get with furniture from IKEA). I had kept them from when I bought my furniture when I moved out of my parents house, finally found a use for them 😅
If you make it solid by adding walls that can cause other weakness because you don't have the crisscross layers so I recommend using normal 2 or 3 walls and 100% lines infill. For this use case I don't think it matters much but for some parts it matters like the light I made
I could be wrong but the reason why I think the experimental part was more ridgid as well as stronger is because those inner cavities have the same walls as the outside im pretty sure I didn't take the close of a look at your actual settings but if your using more walls generally speaking parts in my experience are stronger so its likely due to that I'd imagine.
Infill is only useful for lauer adhesion and supporting top layers. This doesn't only apply to 3d printing, the outed diameter of a beam contribute nearly all of the strength. Its why steel pipes and I-beams are so common, as opposed to full pieces. Try corrugated interiors
Great video, coming from an engineering strength of materials point of view you really should have had the vertical version for the 4 point bend test as that geometry would have a greater area moment of inertia than the horizontal version, think of an I beam and how they are oriented with the vertical member of the cross section in parallel with the force acting on it. As for the torsional test I am surprised to see a difference between the prints but I think if you did multiple tests you would see a similar average strength. It makes sense that the strength of the “hallow tubes” is greater than the 30% infill as torsional stress develops, it is greatest around the neutral axis (in this case the center of the part) and theoretically zero at the surface. Would love to see more extensive testing and the possible application of stress strain curves
Interesting experiment. If you look at the bone structure of different kind of animals, you may see that land animals have hollow bones with thick walls, while birds usually have a specific bone structure with thin walls and foam like internal part, similar to the gyroid infill. I wonder in what situation would be the infill more advantageous. There might be a wall/infill ratio, where strength/weight ratio goes up at the infill end of the spectrum. Somewehere around a wall thickness (ot thinness in this case) where buckling is an actual problem before critical wall failure. Instead of gyroid, lightning infill may be more suitable in this application.
It'd be great if your application of force was always consistent, For the 4 point load test, could you rig an upside down bottle jack above the other end of the handle so you can add load by pumping the jack? That way it should be fairly consistent between the two parts, and it also won't be as reliant on your physical ability to reach the breaking points.
Interesting video! One thing you can do to make infill more useful is to increase the thickness if the infill. Cura is the only one I found that has this and you can reduce the infill really low and increase the "Infill Wall Multiplier" as you please. Plus I think this can go even further if you try changing the angle of the infill aswell. :)
I would really appreciate seeing 3-5 parts put through the test, one value is a hopeful guess, 3 is a pattern. There could very easily be a small crack in one that greatly reduces the strength of one or another. Good testing idea, and your results are as to be expected, but the arent really usable once you get into more complex topics
Great video....try a ribbed or shallow corrugated inner profile (think early Junkers aircraft construction) - will increase both stiffness and torsion. Even more so if used on outer profile too. You'd need to reduce wall thickness and optimise to keep the same weight for true comparison.
Great video! I would like to see larger sample size. Three samples per test seem resonable to me. I get that you were trying to prove the concepty. Awesome ideas in every video.
If you did a topology optimization for the bending load cases alone I would expect some triangulation inside the rod, would be the optimal solution. But when you add the torsional load case and somehow desides on a weighting between the two load cases, then I don't have a guess on what would be the optimal solution. For torsion allone the optimal solution would probably just be a hole through the center, maybe save some weight at the ends if you allow for it. You could maybe also get some compensation for the stress concentrations you get at the countersunk text, witch would be a week point from where a crack could start.
I imagine from the torsion testing has to do with how close to the center the material is. where the inner core really doesn't experience any rotation compared to the outer sides. it would be interesting to make an inverted part where the infill is on the outer side and the Center is actually solid
I think the strongest option would be to use hollow spheres for saving material/time/weight. Lets say you want to have a high % of infill, you'll just distribute some medium-sized hollow spheres as "anti-infill". If you want to have less % infill, you'll distribute larger-sized hollow spheres as "anti-infill" and anti-fill the space between the larger hollew shperes with smaller hollow spheres.
Infill has only ever been a sort of internal support structure for printing purposes that is also as uniform as possible so the part has a predictable weight distribution. It also sometimes contributes to structural support but the use of infill should only ever be an intentional design decision. You can and you should try to find potential solutions that work best for your specific use-case and printing parameters.
Thanks a bunch for this. If you do a new test, please do more tests with PETG or even PLA without CF, and if you have the time, please test 3+ of each sample to average them. I've read the CF tends to be overrated for such projects.
Very nice idea and research! What if you transferred as much mass to the central vertical web, with fillets, stealing from the walls, so it’s more an I-beam with thin outer walls?
i wonder if you built a jig for holding the round end of the torsion test if that would lead to better results as it seems you are not just twisting but also applying a pull force. if you make it so the end of the socket can't displace you might get more repeatable results. supper cool idea though, i wonder if the results change if you post process the parts by annealing them in an oven?
I think your torque test should be balanced. It’s offset because your torque wrench with applies a vertical force component. But I’m not a mechanical- so I probably don’t know.
The problem I have with these kind of strengths tests is that they only deal with instantaneous strength and not material fatigue from repeated stresses (e.g. taking into account things like crack propagation and material changes due to fatigue).
Certainly interested in this. Just to clarify, are you printing the horizontal center line Horizontally? And in what orientation is that center line in the weighted pull tests? Fantastic experiment... and results
I found 5% and 15% gyroid(or triangle infills gave better support on a wing profile..I also found going below the nozzle size for line width made the parts significantly weaker but oversizing was fine or useful.
Kind of reminded me of solid vs hollow anti-sway bars... and then you kind of almost had an I beam inside... what about recreating other structural shapes inside, I beam, Strut channel shape, T-channel shape, etc.
Enjoyed tests. Makes me curious if you tested single strands of unused filament, would their strength, tension etc. have a linear correlation with the 3-D printed parts of the same material. If so, think about how easy it would be to test new materials! Bending tests not possible, in that case print 3 or 4 layers. I feel possible correlations with fully printed pieces could be quite useful.
How much structural integrity might be lost adding holes to the hollow rod? because in theory having it hollow also allows for running hidden wires, no? I mean I don't know how useful that would be or anything, but its an interesting idea. Also I am not really knowledgeable in this field, I am assuming deflection is how much it bent before it broke, is it better for that number to be higher or lower? and how hard is it for something to go back to its original state if it does bend? that is something I think would be a good test (Though I suppose structural integrity would be compromised at that point).
I've read over the years that infill adds little to part strength and that it's mostly down to perimeters. The experimental one with a hollowed oval inside effectually doubles the perimeters. All that said , reading things on the internet is one thing while doing experiments like you are is on another level. I'm digging this series!
What I might try is to take the oval and try different shapes instead, maybe just slight rounded vs very rounded, all of them must be the same weight to have an apples-apples contest. I could further increase the wall thickness by changing the inside corner radius less and less and redistributing that material. I'm glad you're enjoying it, I'll have lots more to come!
I routinely cut internal tubes out around through holes for bolts on molds that I design so that it adds 3 to 6 extra perimeters around those through holes so that clamping the bolts down at a higher torque doesn't crush the print. A few extra parameters can be incredible at adding strength
Try cnc kitchen who has tested this and probably everything else you can think of more scientifically
@@nathanp3366 Oh, I follow Stefan too but it's nice to see different types of testing like the torque wrench.
Yeah I think even in solid metal parts, the majority of the strength comes from the outer skin
From physics standpoint - the closer to you get to the middle of part (or the bending axis) the less strength you gain with infill.
When bending a part, majority of forces are on it's skin, and because of that, hollow parts are industry standard - hollow bike tubes, hollow car's bodies and thin metal sheets being stamped to mimick a hollow part.
You also made an improvement of adding more material to the corners, which are further from the middle and thus provide more % of strength, but I'm not entirely sure that in rectangular part the inner hollow should optimally be an oval - probably something like very rounded rectangle would be better.
Infills generally are wasteful - making some structural ridges just below the surface is sufficient for most of uses.
That makes them stiffer but generally doesn't raise failure.
A hollow tube will deflect less than a comparable solid steel rod, but the tube will still fail sooner.
@ObservationofLimits if both tubes are of comparable weight, the hollow tube will be stronger until we reach the point where buckling becomes the main failure point
@@FierceFire14 a tube is actually a terrible crossection for bending loads exactly due to ovalization buckling failure mode, and internal reinforcement (like spars or infill) can help with this greatly, less so for box sections. From pure bending standpoint, I-beam is the best of course. A tube is unrivalled for torsion.
doesn't ikea put a honeycomb basically of cardboard inside of its tables and desks to improve rigidity? In that case the honeycomb is rigid perpendicular to the load, but I feel like it definitely does something compared to just hollow particle boards
@@Negi2468 They do, but mostly because of reasons mentioned above plus cost.
IKEA wants to use as little of expensive material - outside veneers - as possible without it breaking easily. On top of that, very large flat panels need to be rigid to not only bending but also compression, and that's why cardboard infill is used.
They'd also easily buckle or break at that size and thickness.
That's why Injection Molded parts achieve a greater ratio between resistance per weight (material used) - it is not just the process, it is also because there is no material being wasted in infill, the whole thing goes to the walls. When you need reinforcement, you put ribs. Things are better planned. We should see infill as a wasteful lazy way of giving some internal structure and achieve easy printability - but it is not the best usage of material. EXCELLENT VIDEO! Thanks!
Injected molded parts perform much worse in this regard in my experience because the process is not conducive to super thick walls. Printed parts with 2.5+ mm wall thickness and low infill do a lot better.
infill is basically support material for the insides , not much more😅
@@ry7hym It provides compressive strength and can give trellis type rigidity benefits. Interestingly, lightning infill actually provides the most added strength in my experience, if you use it with infill multiplier 2 so it's a single closed path, by auto generating "ribs".
Yeah injection moulds better be planned out well. Because a fuckup will cost 20k of machining instead of 20euro worth of filament
Also like the other comment said, good luck injecting any thick sections
For your torsion testing, the wrench end of the part should be supported as well so you are purely twisting the part along a single axis. Right now, you will be loading it in multiple directions while trying to apply the torque with the wrench.
Loving the in-depth testing, it's great!
Glad I was not the only one that thought this.
I agree. The torque sensor must be supported by a ball bearing.
the density of gyroid infill can be continuously varied without any sharp transitions. today's slicers don't implement it but it'd be a neat way to redistribute material from the inside towards the surface
I'd like to see that, ideally we can get to a sort of Bone structure, which can flex well and also is resistant to breakage. We should also be able to set areas to have higher density to be more rigid etc. I think we have a few more years of FDM and then we'll be using different types of 3D printers which won't have the same limitations...
@@NeedItMakeIt exactly, for example corners even on decorative prints need extra strength because they'll experience all the impact forces when a part is dropped.
adaptive cubic infill attempts to do this but the density jumps too much between levels, you can't edit the distance-to-wall density curve, and overall it's not the best load carrying structure
I'd like to see a real example of progressive gyroid. I don't think it is possible
Why can't it be done with an algorithm that reads hand-placed weights (points in space that are placed and assigned a weight value) and then interpolates between different scales of infill structure based on the weights? And the user could also specify custom infill grid shapes that can be tiles and applied at various scales.
@@NeedItMakeIt FDM is unlikely to go anywhere, it's the most affordable and user-friendly 3D printer tech.
I think you're touching one of the main point for next gen slicer from an engineering point of view : modeling the inside of parts! That is a very interesting video!
It's an old comment, but incase you are interested: take a look at the prusa mini 3d printable parts (probably also other models, havent checked), specially inside of the models... They have included internal ribs that force the slicer to include interior perimeters inside the parts, adding hugely to the strength. I'm sure many designers do this, but it's a very good example. Would be awesome to have an option to add these directly on the slicer... a bit complicated as it doesn't know how the part will receive stress.
2:49 wow, seeing how flexible the infill is really emphasizes the idea that walls are more important for strength
I've always printed my infill lines at 2x nozzle size thinking it adds strength being it should be less flexible. Be an interesting thing to actually test
Flexibility does not equal weekness. Remember that a bending branch lasts longer in the wind than a stiff one. Stiffness can make then much much weaker to "shock" forces.
@@Stevieboy7 you make a good point
@@Stevieboy7very true but everything is still governed by the wall flexibility.
inside the side-rails of your destruction tower, you should add some netting to catch the projectiles you produce :)
That's for sure! I think I'll do a re-deign of the setup too so we can test woods and metals as well down the line.
@@NeedItMakeIt a good source for nets is 5kg potato bags
Add a large nut, and a threaded bolt to the end of your lever and either add a motor or a hand crank. This way you can apply a very consistant force. If you add a stepper motor you could even calculate the number of turns it took at breaking point. I know its another upgrade but I think it would provide a better result than your hand would. Measuring with consistency is best especially with sometimes wildly different results.
I was thinking about switching to a motor and hydraulic cylinder setup, do you think that could work? I'm happy to do the upgrades, if the channel grows enough I will buy some proper testing equipment as well... I think I'm a ways off at the moment.
@@NeedItMakeItI wouldnt bother with hydraulics at this stage. Just get some adjustable construction support rods, they're dirt cheap, super rugged and usually come in trapezoidal thread variants.
With that and a big wrench, or crowbar, you can get a lot of force going quite easily.
@@Argosh My idea would go with a motor like @newmon and a pulley system. It could be easily built under the wooden structure and would provide quite good measure for deflection before deform/breakage depending on pulley ratio even with regular motor, just counting revs. Stepper for super accurate. Torsion test could use its own frame to separate bending force from affecting the test and maybe use the same motor, pulley(lever), rev/pos counter to keep amount of parts lower.
Probably should put some basic cage around the outward ends of the structure as well, unless you're aiming for a nice swiss cheese drywall theme
@@NeedItMakeIt you could get by with a simple car jack if you wanted to. They are relatively smooth to operate and maintain pressure the entire way.
Good news for you: this video got recommended to me, who's not usually 3D printing video watcher. I'm usually watching engineering or machining videos.
Yep, same.
Most of the strength in any structural part is in the surface. The polish glass rod experiment best taught in the the engineering causes of the 60s and 70s, demonstrated this phenomenon. Glass cutting shows just how fundamental this is. Layer lines are effectively scratches from a tensile strength perspective.
The strength of all plastic materials is Van de Waals forces and a consequence of polymer molecules length and shape.
Infill effects compression forces that cause members to buckle. Cura provides the ability to insert members/internal walls to compensate for compression...
This is a very common and useful technique for designing mechanical parts. The further away the material is from the neutral axis the more it will contribute to the bending strength of the part (you can possibly think of it like torque), so for the best bending strength you want as much mass as far away from the neutral axis as possible. Along the neutral axis (in a simple rectangular part that would be right in the middle of its thickness) the stress and strain is zero and the further away from this axis you get the greater the stress and strain become.
Steel H or I beams are a good example. When used in the right orientation they have very little material in the middle, close to the neutral axis but they have lots of material further out. They are much stronger than a solid rectangular beam of the same mass because of this more optimal material usage. They also have a very good strength to weight ratio, much higher than if you used a solid beam with the same external dimensions whilst also having most of the strength of the solid beam since in a solid beam most of the mass doesn’t add much strength.
So when designing parts to be used in bending it is best to use the largest part you can (external dimensions) and distribute the mass as far from the neutral axis as possible.
This is also why two stacked beams won’t be as strong as a single thicker beam, each of the stacked beams bends independently and therefore the maximum distance material is from the neutral axis of each beam is half of its thickness whereas for a single thicker beam the maximum distance material is from the neutral axis is doubled so the material contributes much more to the strength.
There's one option that I'd love to see tested. Cura has an "alternate extra perimeter" option. Every other layer gets an extra wall so that the infill is getting sandwiched between walls. In theory it would make a stronger part, but I don't know if it's actually significant.
It works, but the difference isn't that major.
About UA-cam algorithm: I've been watching many of your videos - also the ones about different joints... BUT, I need to tell you: I found the topic unpractical. I seriously doubt I will ever design parts that need joints like that. I prefer just having a bigger printer and print parts in one piece. This could be one of the reasons for worst performance of the two past videos, you need to investigate. Make a Community Post with a poll: "Have you watched this video? (link) Would you ever use one of these joints in your models?" Good luck my friend!
Printing direction matters a lot if you want the best strength to weight ratio. Just like fiber-glassing an aircraft. There are two type. Unidirectional, and Bidirectional fiberglass over foam core. Fill will be the foam fore that brings rigidity and transfers the load from one side to the next.
Since I mostly print engineering parts (gear, pulleys, enclosures,..), I need the strongest part.
Hence, I never use the parts fan and don't exceed 50mm/s speed. ex. PETG @N240ºC/B75ºC will not come apart at the layer.
I found white PETG to be the best regardless of the manufacturer. I guess they use Titanium oxide pigments like any other white paint.
That has been my experience I thought I share.
The web, while doing little to the overall strength, is probably what lead to the increased stiffness.
In order to bend your part needs to flatten the inside oval, which is opposed in tension by horizontally webbing and compression by vertical webbing.
I suspect you could leave the webbing out, though, and still have roughly the same performance, albeit a little less stiffness.
Great work on doing practical testing!
Keep going!
On the topic of the video, your results are not surprising at all. For both bending and torsion the more material your have away from the neutral axis, the stiffer the part will be.
For bending you're increasing the moment of inertia of the part and for torsion you're increasing the polar moment of inertia.
Check out the bending and torsion stress formulas.
Running a FEA analysis should generate very similar increases in stiffness.
While the macro shapes can easily be computed by FEA software (or even hand calcs in simple bending or torsion cases), the micro structure (interaction of intra-layer bonding) cannot easily be computed.
If you continue with testing, consider going deeper in that area.
Perhaps do some comparisons between FEA analysis, hand calcs and practical tests to see if you can determine a generic "correction factor" which can be applied to FEA analysis which will reflect the actual part strength.
DM me if you want to chat further on this topic.
I’ve come to really appreciate your content because you’re probing in areas of 3d printing that others are not. It’s really hard to find original content anymore. These tests have been very helpful in informing my own 3d designs. Thanks!
Why is nobody talking about the fact that this man has the incredible skill of writing numbers and creating diagrams *upside down*? That hurt my brain to experience.
LOL. I think I have better printing upside down... I think this skill was developed mainly when I was teaching my son how to read and write, I was always on the opposite side of the table.
From my engineer view and experience. What structural properties you get depends on the material.
Materials which are more flexible benefit from filling the volume. While more rigid (brittle) materials benefit from wall thickness.
You can imagine the difference between a steel square pipe and same size laminated wood beam.
I'm actually insane enough to design my infill in CAD. I have even written gcode by hand to achieve specific things. I'm annoyed that slicers don't actually allow me to have finer control over generated geometry, like directly adding planes or scripting layer behavior. Because I run flashforge, I actually use both Orca and Flashprint, because they do different things better, and if I just want to do quick test I choose the one I know to be easier to do that.
However in engineering mindset your desired properties are always derived 1st from design, 2nd from material.
But because my background is in welded steel products, I naturally account for rolling direction and profile geometry, as they affect things a lot. They are much like print direction and layer joining.
But you will not find any sharp edges in my designs, unless it has a specific use.
i always avoid generated infill at all cost. i usually cut/hollow the vertical faces with hexagons instead of circles (to avoid overhangs). i also usually design the supports if I really need them so it prints faster with less waste and better finish. there's many tricks you can do when designing for FDM, I even automated some of them as features in my cad software.
Thanks for the extra test, I appreciate the work.
I even went back to your last video to get the comparison.
PETG-CF (red) Continuous with infill: 30,6 Nm
PETG-CF (black) Continuous with infill: 48,4 Nm
So, an Improvement of 58,17% 😮 in the torque Test, more that I expected. Also the failure mode changed to "fractured in 3 Parts" indicates that the black Filament has more Stiffness, imo.
I heared before that the filament color can change the products parameters, on other filament test runs by CNC Kitchen or Makers Muse they tested on test cupons, but on a more practical part the diffrence seems quite huge.
ok the Samplesize is 1/1 but the trend is plausible imo., because all the other Red Parts in the Joint Test where the weak link too.
Oh wow, that was unexpected! Thanks, this actually may be a game changer to be honest - if we are trying to make the tough parts. Great results and successful experiment!
One observation from my side: at 7:25 You may see the cracks started to form before the part went flying. I am wondering if those cracks started happening after some threshold or prior to that....? Reason: when You have infill it's somewhat flexible thus the *amount* of times You can put a high load is not small but if You have smth more rigid it might be tricky - if, let's say, You go to 80% of max load and those cracks start to form this means we decrease the max value with every load put. It's not _necessarily_ the case - I am just thinking out loud. I would propose the durability test additionally to the one You've already done - and if that will be a success as well - we can basically ditch the infill (in 90+% of the cases). Thanks again - it was really interesting :)
I think it's not far from the direction we need to go, it needs some refinement of course. Yes I messed up the final test on that part because it was too hard to force the lever down, I'd need a longer lever. I like the idea of a durability test, that's a great idea actually. I could setup a part to have force applied and then removed and continue this until it fails by no longer being able to take load by deformation or breakage. I think there is some potential here for sure, I think we need to do some more research and testing. I also want to test extremely sparce infill with far increased thickness. And I have a few more ideas too and maybe there is a way to combine all of the concepts into something even better.
Great Video and testing. I have some inputs on your methods, as Ive done 4 Point Bending on Composite Sandwiches for my Bachelor Thesis:
I would choose to repeat the Test 5 times per sample, as a big scattering of examples is to be expected. (But you have Said that already).
The Text to distinguish the samples Looks neat, but creates a Point of failure.
4PB is sometimes used over 3 PB, because you get an area of the sample, which is just loaded with shear force, as the bending Moment in not present inbetween the cylinders for force intruduction. Your cylinders should be further apart.
Thanks for your contribution to the 3d printing community
This is fascinating and I love seeing all your results and tests! Thanks for sharing; I can’t wait to see more tests!
As always kudos for actually experimentally measure results, and trying new variations.
Don't beat yourself too hard about the last videos not having the same performance of the previous ones, the fatigue of a topic is a real thing.
Breakthrough videos of an interesting topic could have an increase in views due to high engagement in the algorithm, but people rarely stick around about continuous testing of the same topic, maybe switch around the topics a bit between each release, and then make "series" of videos by topic.
At the same time wildly different content of topics also deter the core audience so it's a gamble push to far away the "main" audience, so the main takeaway is stay in your lane, but have a little more breathing room between topics continuous "main" topics.
Hexagon, strait and triangle infill should help make a part ridgid as well. Most of the other infill like gyroid are pretty but when you look at it, it's just there to fill a void not add strength.
3D Honeycomb was found to be strongest by a few channels I believe
This kind of testing is what drew me in to the channel, I hope you keep it up!
Mathematically speaking, pound for pound your best options are either a large hollow cylinder or I beam for holding loads. For your tests, an i beam would work best with most of the length facing "up" and enough width at the top and bottom to prevent torquing, you wean the whole load straight on the support. Conceptually just thing about how much material force would need to move in order to bend it. The more material the better right? If you can imagine drawing lines through it at a perfect 90 degrees from the earth, that's how much material is holding the weight. The more lines that run through a cross section, the more material would need to be moved.
It's mad that the last two videos haven't done well, I'm loving this series!
If you look into beam strength in bending, and why I-beams were created, you'll see that having more material under tension/compression at the furthest extents from the neutral plane(axis, for torsion) increases stiffness. But if stress goes too high, you can get a split down the web of an I-beam (on that neutral plane) so I think it would be worth comparing 2-3 printed I-beams of similar mass, but with different web:flange ratios.
Chances are they're not worth the complexity to model/print in comparison to a shelled model, or one with minimal infill, but it may be a good demonstration.
But I suspect they would show that prints have need of both the largest possible outer beam dimensions (flange on an I-beam; floor/ceiling of your experimental beam), as well as a sufficient bond between the neutral layers.
Your vertical should have a radius and feather into the inside walls. During the twist test you can see that the vertical line cracks where it enters the inside wall. A small radius would displace that shearing point.
Really liking this series! Looking forward for the optimized version.
One thing that would be worth mentioning is if the new approach has any effect on the print time. Without infill I would expect the print could go even faster. Faster, stronger, same material cost, that would be a win win win.
I have always been fascinated by taking 7 equal diameter "tubes' and putting them together. One of them always gravitates toward the center, and the other 6 form a neat "structure" around the center tube. (Try this with 7 pencils to see what I mean). You can see something that happens in nature - a Hexagon! Now - as far as your infill goes - draw a straight line through the centers of all those tubes and align ONE of those lines as the vertical support for inside the square. The other two lines should go to the four corners if your object is a true square! I would bet that this type of "infill" would add even more strength!
i love this kind of stuff! i like seeing new ideas of making prints stronger with less material
I truly think this is valuable for specific use-cases.
However, all I've learned for my personal projects is that nearly all methods are far stronger than I require for my application haha
To print mechanical parts it would be nice to have a function in the design program (or the slicer) that is able to calculate the arrangement of infill using finite elements method (FEM). Finally we would get some design similar to bones. In bones we see a full material shell and an infill that is increasingly foamy from outside to inside. Close to joints bones are also more stable. Of course, to construct with FEM we need to know our requirements, the forces acting on the part.
There's a technique I once saw used by a fairly famous entity that allows someone to hollow out a solid piece while minimizing loss of integrity. I remember very little of what it was called or what it was used by, but I do recall they primarily used it for reducing weight, so I'd guess NASA or SpaceX if I had to.
The technique was basically using some sort of software to calculate where forces would be distributed inside of a part (when under load in various situations), and just cutting wherever the forces were too minimal to be worth the cost. This allowed them to remove a very significant amount of material with barely any reduction in durability. (I feel like I even remember one case where it *increased* the durability of the part compared to a full solid, but that seems too far-fetched and thus possibly a false memory)
This might be a bit of an insane request, considering how little information I have to give, but I'm wondering if it could be applied here, where you can relocate the mass from the infill to increase the density of wherever the software says will get the most use.
If infill is to support top layers, it doesn't matter much, how it is printed, as long as it's fairly dense just below bridging.
If infill needs to bear load, it needs to be thick - 0.6-0.8 on 0.4 nozzle - and dense - 50%+ (just changing width changes line spacing as well).
In Cura it is possible to print infill with multiple lines, creating loops. Combined with alternating extra wall (Cura) it should produce pretty _solid_ infill. PrusaSlicer honeycomb is also 2 lines wide.
( Or use non-filled plain filament for max strength 😅 )
This is very interesting. I make structural parts. I am doing hydrofoil blades at the moment. I print these hollow and then fill with expanding foam. If I need extra strength, then I put a very narrow slot, 0.1 mm width in the wall, and extending in a few mm, but not all the way through. I am using a 0.6 mm nozzle, so the slot walls in practice touch, i.e. it is like a web but with an open side. This adds to the cross section of that part of the wall, i.e. stronger.
I have tried all sorts of different materials. My current favourite is PLA Meta. It has good inter-layer strength, and is relatively easy to print with. And not too expensive. It is a bit stringy so there can be some surface cleaning up to do, And I use light colours - dark colours overheat in strong sunlight and distort.
Great testing! I think the optimal structure would be to have thick walls with all interior corners rounded e.g. 2 mm and inside filled with big honeycomb-like structure with a couple of layers worth of wall thickness. The idea would be to have some internal walls but not pure infill. I would guess a single hexagonal cell should have around 15 mm side length. I'm not sure if the walls of hexagonal structure should also have rounded interior corners. Having rounded corners would make the structure stronger but additional testing would be required to see if it's worth having more material.
I agree with your methodology that the mass of each part should be identical.
To avoid the parts flying off in random directions from the testing station, put a box made from some sort of transparent plastic that's big enough to contain the entire testing rig. This way it can still be filmed and the broken plastic won't fly out in a random direction (one of them being your head).
Great video. No need to reinvent the wheel. Design the interior like an H-beam with rounded corners.
Thank you for doing these! I’m working on designing stuff that needs to have strength so it’s really nice to have someone testing.
I’d love to see a video about temporary joinery (like sliding rails, snap in place, etc) if you’re looking for topics.
I’m working on something modular that has side pieces that need to slide on and off and I’m having trouble finding ideas that will work well.
You could try minimal Lightning infill with maximum walls. The purpose of the infill would just be to support the roof of the part while printing. And you could do a + shaped interior flange, and maybe make it thicker. Looking forward to the next set of tests.
Loose paper bags to catch/decelerate the sample fragments, like airbags/parachutes;
Or lace a cord (twine?) through the ends so that most of the pieces are captive; like a floppy-rail/zip-line.
-Loss of spectacle aside,- I think that having more control over where they fly to would allow for a safer (and easier to clean up) test, while also providing some illustration of how much energy each failure releases by how much they affect (or punch through) those controls.
Great video! Appreciate the testing approach used.
Thanks, I'll continue to improve my test setups for better more consistent results as well.
If there's one thing I've noticed, it's that color Greatly affects the strength of some filaments. I've seen it with PLA+ for sure, but I can't say for certain about other filaments though. I would just guess that's the case. With the PLA+ the drastics I've seen are between Black and Silk Silver. The black is so much more rigid and stronger while the Silk Silver is so much more flexible and is significantly weaker (but it does print better/easier).
In cura that I use for printing my parts I am able to add custom nesting models that I then set as modifiers, I then set those embedded parts in areas of high stress, say around areas where I am embedding fixing points or connecting structures between fixing points to then add internal structure by setting these models as areas of infill that are 100% solid to allow me to bulk out areas of the print where nessesary and then use normal infill desity to fill the rest of the part as you normally would. this process then allows maximum dencity where you actualy need it and then sparse where it is needed for te printing structural support only. worth looking into.
i've been doing quadruple wall with lightning infill (7 layer top ironing enabled) since january and im satisfied with speed, strength and surface quality. working on an end table soon
post it somewhere i was looking to do a similar project!
Increase infill line multiplier as cura calls it, and reduce density in half. This will make the infill stiffer and much more structural.
I found this while experimenting with structural properties of TPU, which makes the effects immediately a tangibly experience, no breaking fancy tools or setups necessary for crude results.
I liked your last two testing videos! I was very impressed! They reminded me of the tests that CNC Kitchen does.
If I had to guess, it's either your thumbnails didn't attract which people or YT is crazy.
I have been loving these testing videos you have done lately. Please keep at it! :) I have been toying with the idea of a printed printer frame and your joints and tests are really helpful with that.
Also Mr. URGENT has a point, seeing results with more common, regular filaments would be nice as well, but only if you have the resources for including them in the tests.
When optimizing you need to consider the load profile. For the bending test you want most of the material near the top and bottom, with the minimum sidewalls to enforce spacing. Which is of course not optimum for other load profiles. For torque you'd want a circular tube if the material was isotopic, which FDM prints aren't.
Torsion stress concentrate on the middle of square/rectangular profile members. You should create a cylinder in the middle leaving the corners hollow. The cylinder is the strongest shape under torsional loads. Would be interesting to know how it compares under flexion as well.
To smoothly pull on lever, OR direct pull on gauge, use a boat winch with steel wire rope (cable). can anchor it to table or test press and run through pulleys as needed for space management or "leverage" increase through multiplication. VERY CHEAP AND EASY/SIMPLE! Can get ALOT of pull from it and only 4 bolts mounts it vertically OR horizontally for ease/comfort of turning handle.... OR could use a cheap $99 electric 12vdc 1,500 lb atv winch for steady pull rate.
i always love when people test limits:D
You should do experimental with diagonals -- I expect that to be much better over all.
Please continue to do these testing videos - its majorly interesting
I'd be curious to see Stefan, CNC Kitchen, recreate these tests you're doing. His motorized screw provides a more consistent and continuous force. He also uses a three-point bar rather than a four-point; fours have different geometric loading and induce stresses differently, there's a literal reason why threes are used in material testing-- stemming from this, your previous tests with the joinery would've also had problems with the evenness of the loading from test to test, a very long diagonal dovetail would've been loaded very differently from a standard perpendicular dovetail, i.e. testing methodology was not actually consistent where one style of joint would've had unfair advantage within the testing setup. Even look at Matthias Wandel's work for how to test properly. I'm sure there's other creators within the relevant space of DIYers and makers, who either have an engineering background or at least an understanding of engineering, who also have proper testing methodology and produce good data. On the previous tests you also had print orientation issues, which should have been tested for. On this test you had design issues with perimeter count, of which should have been accounted for in the design phase; thin-wall with infill is very different from thick-wall without infill. This test also had potential print direction issues. Infill also does very little for this testing, you'd be better off testing infill with tension, compression, and buckling tests; certain design aspects fit certain tests, figure out which match and don't test mismatched sets.
If you're going to be doing engineering testing, for the sake of good data and good information, use proper methodology. I want to see more actual engineering within the space, but if things can't be done properly then creators should stick to being entertainment only, we've all seen how bad information spreads like wildfire in DIY-friendly communities where the basis of the majority are undereducated on the relevant subjects when it comes to hard data and testing of materials and designs; this is also why a lot of maker channels don't have testing setups and attempt to scientifically measure aspects, because they're not qualified to speak on the matter and frankly won't invest into understanding how to do it properly. Do your due diligence to properly educate your audience. But to educate your audience, you need to educate yourself first. You need to understand why your methodology is lackluster, in both design and testing, you need to understand when testing actually matters and how to test for it. You need to label your produced testing videos as non-educational and put a disclaimer that results should not be taken as fact. If you presented your testing to a room of formally educated engineers, you'd be laughed out of that room, told to go back to physics 101, and that your testing and design method is nothing more than a science fair experiment that you didn't properly research. Good data is important, but unfortunately an undereducated community cannot recognize bad data, which is why it's your job as the content creator to do better and provide that good data outright.
Also, use a blast shield. Not everything will break cleanly. Again, it's a due diligence topic, I'm surprised I haven't seen too many comments harping about personal safety on your videos.
I hope you take this as constructive criticismb and not the content creator usual of being an egotistic pissbaby. Everyone has to learn at some point, however people should rather learn before they attempt to educate.
If i really need strength for something i just put something stong inside. I made internal corner joints for some aluminum profiles and put Allen keys inside them for strength (the ones you get with furniture from IKEA). I had kept them from when I bought my furniture when I moved out of my parents house, finally found a use for them 😅
If you make it solid by adding walls that can cause other weakness because you don't have the crisscross layers so I recommend using normal 2 or 3 walls and 100% lines infill. For this use case I don't think it matters much but for some parts it matters like the light I made
I could be wrong but the reason why I think the experimental part was more ridgid as well as stronger is because those inner cavities have the same walls as the outside im pretty sure I didn't take the close of a look at your actual settings but if your using more walls generally speaking parts in my experience are stronger so its likely due to that I'd imagine.
Infill is only useful for lauer adhesion and supporting top layers. This doesn't only apply to 3d printing, the outed diameter of a beam contribute nearly all of the strength. Its why steel pipes and I-beams are so common, as opposed to full pieces. Try corrugated interiors
This makes total sense, material on the edges of the part always takes more force than on the inside.
Great video, coming from an engineering strength of materials point of view you really should have had the vertical version for the 4 point bend test as that geometry would have a greater area moment of inertia than the horizontal version, think of an I beam and how they are oriented with the vertical member of the cross section in parallel with the force acting on it.
As for the torsional test I am surprised to see a difference between the prints but I think if you did multiple tests you would see a similar average strength. It makes sense that the strength of the “hallow tubes” is greater than the 30% infill as torsional stress develops, it is greatest around the neutral axis (in this case the center of the part) and theoretically zero at the surface. Would love to see more extensive testing and the possible application of stress strain curves
Interesting experiment. If you look at the bone structure of different kind of animals, you may see that land animals have hollow bones with thick walls, while birds usually have a specific bone structure with thin walls and foam like internal part, similar to the gyroid infill. I wonder in what situation would be the infill more advantageous. There might be a wall/infill ratio, where strength/weight ratio goes up at the infill end of the spectrum. Somewehere around a wall thickness (ot thinness in this case) where buckling is an actual problem before critical wall failure. Instead of gyroid, lightning infill may be more suitable in this application.
It'd be great if your application of force was always consistent, For the 4 point load test, could you rig an upside down bottle jack above the other end of the handle so you can add load by pumping the jack? That way it should be fairly consistent between the two parts, and it also won't be as reliant on your physical ability to reach the breaking points.
Interesting video! One thing you can do to make infill more useful is to increase the thickness if the infill. Cura is the only one I found that has this and you can reduce the infill really low and increase the "Infill Wall Multiplier" as you please. Plus I think this can go even further if you try changing the angle of the infill aswell. :)
Great test. Can you also publish how much filament and print time for each. It helps determine if it is "worth it".
I would really appreciate seeing 3-5 parts put through the test, one value is a hopeful guess, 3 is a pattern.
There could very easily be a small crack in one that greatly reduces the strength of one or another.
Good testing idea, and your results are as to be expected, but the arent really usable once you get into more complex topics
Great video....try a ribbed or shallow corrugated inner profile (think early Junkers aircraft construction) - will increase both stiffness and torsion. Even more so if used on outer profile too. You'd need to reduce wall thickness and optimise to keep the same weight for true comparison.
Great video! I would like to see larger sample size. Three samples per test seem resonable to me. I get that you were trying to prove the concepty. Awesome ideas in every video.
If you did a topology optimization for the bending load cases alone I would expect some triangulation inside the rod, would be the optimal solution.
But when you add the torsional load case and somehow desides on a weighting between the two load cases, then I don't have a guess on what would be the optimal solution.
For torsion allone the optimal solution would probably just be a hole through the center, maybe save some weight at the ends if you allow for it. You could maybe also get some compensation for the stress concentrations you get at the countersunk text, witch would be a week point from where a crack could start.
PLEASE DO MORE TESTING!! i would love to see them!
I imagine from the torsion testing has to do with how close to the center the material is. where the inner core really doesn't experience any rotation compared to the outer sides. it would be interesting to make an inverted part where the infill is on the outer side and the Center is actually solid
If you want to focus on the bending and leave a little bit the torsion aside, strengthen the up and down sides instead of the corners
I think the strongest option would be to use hollow spheres for saving material/time/weight.
Lets say you want to have a high % of infill, you'll just distribute some medium-sized hollow spheres as "anti-infill".
If you want to have less % infill, you'll distribute larger-sized hollow spheres as "anti-infill" and anti-fill the space between the larger hollew shperes with smaller hollow spheres.
You could keep adding smaller and smaller spheres to fill the space between the spheres until you reached the % of non-infill you wanted.
I'm liking the new music you're using
I'm enjoying these. Similar content to others, but somehow you've been unique (In a good way)
Infill has only ever been a sort of internal support structure for printing purposes that is also as uniform as possible so the part has a predictable weight distribution. It also sometimes contributes to structural support but the use of infill should only ever be an intentional design decision. You can and you should try to find potential solutions that work best for your specific use-case and printing parameters.
I wish you had stress/strain plots so we could compare the points of plastic deformation, that is what will matter for most use cases
Thanks a bunch for this. If you do a new test, please do more tests with PETG or even PLA without CF, and if you have the time, please test 3+ of each sample to average them. I've read the CF tends to be overrated for such projects.
Very nice idea and research! What if you transferred as much mass to the central vertical web, with fillets, stealing from the walls, so it’s more an I-beam with thin outer walls?
i wonder if you built a jig for holding the round end of the torsion test if that would lead to better results as it seems you are not just twisting but also applying a pull force. if you make it so the end of the socket can't displace you might get more repeatable results. supper cool idea though, i wonder if the results change if you post process the parts by annealing them in an oven?
I think your torque test should be balanced. It’s offset because your torque wrench with applies a vertical force component. But I’m not a mechanical- so I probably don’t know.
i love all your testing vids, keep up the great work!
These property testing videos are great
Thanks for the testing!
The problem I have with these kind of strengths tests is that they only deal with instantaneous strength and not material fatigue from repeated stresses (e.g. taking into account things like crack propagation and material changes due to fatigue).
nice test. But one interesting thing was left out. Did the time changed you needed for the print?
I'm assuming the higher second moment of area is probably one of the main factors helping this. Makes sense, good to see testing ey
What are the print-times like? Does the experimental benefit (time-wise) from having a more 'continuous' path for the nozzle to travel?
Is there a reason you didn't design one with BOTH vertical and horizontal supports. It might be a good idea to add that to the mix.
Great vid, thanks.
Printing mechanical objects since 2015. The more experience I gain, the more I "hand jam" necessary infill into certain kinds of parts.
Love the videos lately keep up the good work!
Certainly interested in this. Just to clarify, are you printing the horizontal center line Horizontally? And in what orientation is that center line in the weighted pull tests? Fantastic experiment... and results
I found 5% and 15% gyroid(or triangle infills gave better support on a wing profile..I also found going below the nozzle size for line width made the parts significantly weaker but oversizing was fine or useful.
Kind of reminded me of solid vs hollow anti-sway bars... and then you kind of almost had an I beam inside... what about recreating other structural shapes inside, I beam, Strut channel shape, T-channel shape, etc.
Enjoyed tests. Makes me curious if you tested single strands of unused filament, would their strength, tension etc. have a linear correlation with the 3-D printed parts of the same material. If so, think about how easy it would be to test new materials! Bending tests not possible, in that case print 3 or 4 layers. I feel possible correlations with fully printed pieces could be quite useful.
How much structural integrity might be lost adding holes to the hollow rod? because in theory having it hollow also allows for running hidden wires, no? I mean I don't know how useful that would be or anything, but its an interesting idea.
Also I am not really knowledgeable in this field, I am assuming deflection is how much it bent before it broke, is it better for that number to be higher or lower? and how hard is it for something to go back to its original state if it does bend? that is something I think would be a good test (Though I suppose structural integrity would be compromised at that point).