How Strong are 3D Printed Connections? 

A key takeaway, is that brutal changes in section area is going to create stress concentration, at which point it's going to fail in an area of minimum section, where the stress is highest. We can see this very well in the echo lock.

Thank you for the video. i appreciate that you’re taking 3d printing in a different direction than other channels. I’m also sorry for the commenters on here that sound like they’re just angry people in general. “Useless tests” - really? He used the scientific method with a control sample and all pieces were tested under the same conditions, so the tests are valuable. If you want to test different infill patterns, that’s a different test altogether.

Nice! It would be interesting to see the same test repeated with the use of an adhesive. Would those joint with larger surface areas fair better than those with smaller surface area.

thanks for taking the time to do this

This channel and others doing great testing are pushing the material science and understanding of 3d prints forward into a golden age. Thank you.

Thanks for taking the time to do this and share it all, it's been really interesting and useful to watch. If you're doing something similar in future, personally I would have liked to see the control sample tested first as this gives a benchmark to keep in mind when viewing the subsequent joints in the test rig.

For the rig, you could install the weight scale below the sample. You fix it to the ground and place the sample just above. You can then attach the sample to the liver and use a car jack to push it from below, you'll be then able to apply exactly the force you want on the liver.

If you install the heat set inserts on the far side of the connection, instead of the near side, they have to pull THROUGH the part rather than pull OUT of the part. Should be WAY stronger.

Congrats on all your recent growth, well deserved!

Nice version 1 setup 👍🏻 Here some ideas for version 2: - Attach the scale to the top so it doesn’t fall for every test - Motorise the load so you get a more precise and linear increase - If possible, record the data from the scale so you can plot a graph of the results. Not too complicated to do with an Arduino and a load cell in case commercial scales are expensive Keep it the good work and thanks for sharing 👍🏻

Great video! I'm loving this series and it's helped me so much to get more out of my smaller printer on bigger projects

Thank you for putting the time and effort into this series. Quinte interesting results indeed!

Damn that was interesting! Thanks for making this and doing it so thorough!

That was entertaining. The results really surprised me. I tried to guess beforehand on first and last.....i was wrong. lol Love this style of content...lots to learn and take away from this. Thanks Mike.

I like watching your tests. Since you're testing beam deflection load....have you considered a 3 piece joint where the lower "half" is optimized for tension and the top for compression?

Very interesting and wonderfully executed. Thank you!

Where were you with your channel when the Westfold fell? You have answers to my questions, and i could not find you for so long!

Great work, thanks for sharing your results.

Hi! Very interesting series about connections. But it seems to me that the test is not very objective because of the difference in the length of the lock, the longest ones logically won, you can make a lock almost the length of the entire part and then it will surpass parts without a lock, since it will have more solid layers. I think it is worth making a fixed length for all locks, for example 30 mm and all should fit within this limit.

Great video series man, keep up the cool work!

For the echo lock you should reprint it with rounded dove tail shapes, it looks like all the stress was on the point of the dove tail. Edit to add. Some of these joints could’ve been much better for comparing against the continuous. The bottom of most of the connections are very secure in the top half, but the bottom half of your joint designs have very little contact to fight against the rotational force being applied. It’s like putting two unequal levers side to side and overlaying their ends and calling it a table saw even tho it looks like this ( V) vs (-)

I can certainly think of reasons to make joints without any sort of adhesive or chemical - if nothing else, avoiding the PPE costs and health risks is a bonus. But I do think that if you need to join two parts made from polymer, and that joint needs to be any kind of load-bearing, then solvent-welding is a very, very, very good idea. I'd be extremely interested in a comparison where you took the best to joints from this video, and solvent-welded them using MEK or ethyl acetate. My instinct is that they've be very nearly as strong as the continuous-print version, but I'd love to see that proved out.

I'd like to see how these joints hold up with adhesive instead of with screws or friction. The drop lock looks like it would excel with adhesive. Also, what about the force being applied along the length of the joint instead of perpendicularly? That would be interesting. And finally, what if you put a collar around each joint?

I'm most interested in testing with different nozzle sizes, will be interesting to see how the same overall thickness wall will do with less interfacing to other walls for strength at the given thickness. Then also strength in just total wall count nozzle vs nozzle.

I think a couple of connections would function better if they were rotated 90°. Most of the connections show 1 weekpoint on the bottom or at the top, while turned 90°, these weekpoints wouldn't be there and the stress would be more over the rest of the joint.

Thanks!

Excellent test rig.

Awesome! Would you test these again with adhesive?

Went pretty much as expected except that the difference between the best connection and solid is rather huge. I would've expected less of a difference there, but you did also only use screws. Any chance you could test the best two with glue and/or sticking a soldering iron to the pieces to melt them together?

are you using Fuzzy skin feature? if yes please give your settings and layer height because it looks perfect

It would be good to see a continuation where you refine the designs. The under squint in particular looks like it could be massively improved since it only failed when that “tooth” at the end deformed enough to spring out and the whole thing was still intact afterwards. If you deepen that tooth and maybe make it a sharper angle then that could help its strength a lot.

The mechanics of the fails themselves is quite interesting - there are as far as I can see two main types of it - when the weak point breaks (thin "hand") and basically when the part that had to be inside the cut from the other side slips out. The first type is "fairy" easy to address by just increasing the scale of the part - however this isn't smth what will help if we don't have a room to increase the size of the part, however the second type is far more interesting - it feels like if the overlaps would be bigger, that would help to keep the joint from breaking, but on the other hand if we look at the way the weakest joint fails it is immediately visible why that wouldn't work. It's arguably counterintuitive - but really interrsting to see. That actually made me to try to inagine what can be improved in the joints - of the second type, so actually boosted creativity, which (at least for myself) I feel like a very good thing. Thus, thanks a lot for Your content - it's really interesting! Please continue with Your approach - that works really well!

I was hoping the echo lock would do better, but in hindsight it's obvious why it didn't. One way you could improve your test rig is with a system of pulleys in series; that'd increase the ratio a lot.

I think you could have a look at the “My Tech Fun” channel and his testing methods, you could probably adapt some of his tests to work for your larger test pieces, he manages to generate some pretty large forces using some kind of ratcheting mechanism.

I don't know if your conclusion is quite correct. It looked more to me like the pressure points from the pins at the top were in between the ends of the joints on the strongest ones, while they were on the outside of the weakest ones, all except the screwed one. And the screwed one had no 'fingers' at all, which most likely contributed more to the failure than anything else. I bet if you made the drop lock joint 'longer' (as long as the sine wave or undr squin) then it would perform significantly better, because you're gaining leverage from the ends of the joint which are your 'grip points'. That, or move your pressure pins to be wider than all the joint widths, which would still show a difference just due to leverage on the joint ends.

Nice work. The "Sine Wave" seem to fail from insert pull out rather than direct stress. It would be interesting to see how much better the "Sine Wave" performs with a nut & bolt fixing. Also, the "Screw Lap" went at a weak point due to less material around the double-screw fixing. Two single-screw fixings might perform better. Likely still not as good as the "Sine Wave", but a simpler design process is the trade-off.

Very interesting!

You could extend the lever out of the other side of the rig and use a hydrolic jack to lift the lever.

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.

Have you considered testing by pulling against the joint along the length og the parts? I suspect that that might be a more realistic use case for the joint.

I wonder how these would fare in the other orientation, placing load across the joint not on the join (rotate 90 degrees)

What is the chance of you testing an "undr squinted sine wave?" Both profiles are such a similar shape, had 2 different failures, and performed really well. Also, just a thought, i could imagine a short and wide dovetail shape on top and bottom of the undr squint insert, probably with a hair of extra clearance on the dovetail side that that wouldn't be pushing outward, might add some very beneficial bite from the center of the connection. Similar to how the threaded insert and bolt are grabbing from the center of the sine wave.

you could use a strap to increase load

If your parts are designed properly and printed in the right orientation, super glue does a perfect job. I've never had a glue joint failure. Inserts are good if you plan on screwing and unscrewing things many times, without destroying the threads (like a battery cover).

3rd! love your research, keep it up!

I wonder what numbers e.g. pine or spruce would get. Harder woods are probably _harder_ to break... Overall, the stress even the worst part could take was quite impressive 👍

If you take a pipe and slide it on your lever you can increase the weight by sliding it out. For more granular and easy weight adjustment.

You said you were also releasing the STEP files and/or Fusion 360 files, but I only see STL and 3MF on the linked page. Am I missing something, or have you not released them? I'm dying to see exactly how they're all constructed. Your explanation in prior videos didn't quite do it for me.

Am I misunderstanding how these measurements are being taken? The scale clearly reads 41kg at 7:40 but you state the maximum force is 25kg? I assumed this would be like Matthias Wandel's experiments where he measures the maximum force it can withstand before the force decreases.

If possible, put the heat set threads on the outside. My thinking is that more material between the stress points, i.e., head and nut.

Would reducing the amount of "fingers" in the drop and echo lock improve it's performance?

Few points of criticism. Hopefully you'll be level-headed and take this as a learning opportunity, which is fairly rare in the content creator space. There's a reason why most bend loading tests use three points, other than convenience there's also advantages with strain rate and loading geometry. In the case of testing joinery like this, it would also put the loading in the center of the joint rather than in two points mirrored from center, the main issue here is that none of your joints are equalized in width making for bad comparisons, a short dovetail will have different properties than a long butt or scarf joint, especially with the distance of your central two points. You should also be posting results in terms of stress, not outright force, the equation for bending being the bending moment divided by the section modulus; Stefan (CNC Kitchen) has explained this in at least their 3Dp v wood video. Speaking of Stefan, you may heavily benefit from understanding their testing methodology and understanding why properly scientific testing is more productive. You should also invest into a motor-based setup that can apply force gradually and consistently over time, adding bags of sand, while minimal (for now, imagine if you got into the scale of adding 10 pound weights as a metric), will introduce shock forces if the load is not eased into position, and this is also demonstrated in your final test of snapping your solid bending bar given you had to apply multiple attempts to break it, which also introduces steps of applied force that are not gradual nor consistent which will remove any elastic deformation between each applied step before applying more force to the bar; it's just bad science, and that leads to bad numbers, hitting a wall with a wrecking ball multiple times does not show total breaking force needed in one gradual shot rather it shows how many times you need to hit with a consistent force, same with driving a nail with wood where it's about number of strikes of consistent shock force rather than a one-shot of consistent gradual force, etc. And while on this topic, I do also think that shock forces are another important metric, though the testing methodology is entirely different, mostly because of realistic use of the tested items, no realistic force within normal use of an item is going to have gradual and consistent application of force on it, realistic use is almost always shock forces or near-instantaneous loading; take for example setting down a heavy object, how many times have you put a gradual and consistent force from the instant surfaces touch to where the full weight of the object is down, the answer is never, because the force curve is a very rapid ramp from no weight to full weight even with the gentlest of setting a heavy object down; another great example is walking and running, you always have a heel-strike, and it's called a strike for a reason. Material properties are one thing, but realistic use is another thing, this is why doing both a loading test and a shock test are important. If you're looking to become a channel based in engineering, for the love of integrity, and the love of good numbers and good science, do things properly. Honestly, just reach out to Stefan, or even Matthias Wandel, and they could probably give you a crash course on how to design a proper testing setup for both tension and bending testing of material, given both of them have designed their own test setups and I believe both of them may have tested shock forces as well in the past. I would also include more test samples. Average of 3 is a starting point, most engineers prefer at least an average of five to discover early outliers within a set. You should also test different print orientations, as there's three cartesian axes to work with, then preferred orientation if a part requires it. Testing in more common materials is also more helpful for the community, carbon-filled plastic will always be a specialty filament and there'll always be debate around this specific filament. In joinery like this you should also be testing joint orientation, not just print orientation. Etc. This is essentially R&D, and R&D is not cheap, that's just part of engineering. You should also be separating attachment method into separate groups, you should not be mixing fastener-based joints with friction-based joints, similarly you should only include adhesion-based joints within their own discrete set of results; of which the latter is where butt joints actually belong, a fastener set between two surface does not make a true butt/lap joint. If you really want to get scientific with specific joints, such as the dovetail in this series, you should explore varieties of the dovetail, start with the classic perpendicular then expand into dovetails on an angle, then other varieties such as a curved dovetail, etc.; again this also gets into why print orientation testing and joint orientation testing are important variables, you're exploring what makes a joint stronger and explaining why by extrapolating the data from proper testing. I'd also do an intermittent loading test on an extra sample just to separate out elastic and plastic deformation and the point of transition; I would also reconsider where you mark your failure point, a lot of people will consider the transition into plastic deformation the failure point as it's all downhill from that point forward, some people will consider damage (cracking, in most cases) to be the failure point, then there's people who consider full separation as the failure point of which this should be defined as catastrophic failure, where earlier material failure transitions into catastrophic failure being a better way to look at the situation- moment of plastic deformation should be the initial point, damage should be the point of no return and in need of repair, and catastrophic being the end of the spectrum and need of replacement. Also, if you're testing pieces to failure, especially on pieces that will loosely fall away, ALWAYS have personal safety set up. Again, I hope you take this as a learning opportunity to better your content creation. I want to see your channel grow, but if you're going to be producing lackluster numbers with bad science, then you should be for entertainment purposes only and not dabble in anything within the sphere of engineering. Why? Because you'd be laughed out of a room of actual engineers and anyone who understands good science and decent numbers. The maker community already has enough bad information floating around by people who don't quite understand what they're doing or talking about, which is why proper engineers like Stefan are a blessing for the community, because good science, decent numbers, and good communication abilities can fix a lot of misconceptions that are bred from bad information. Anyone who dabbles into the education side of things, of which engineering pieces are a part of, has a due diligence of doing things properly for the purposes of good information; else anyone who spreads less than good information, out of ignorance or malice, is a fraud. Be on the side of good and accurate information with proper methodology, use proven proper scientific methodologies and don't fall for the fallacy of 'I think this is good enough therefore it is good enough' that many people inevitably fall for then believe is the factually correct method.

These joints seem to be tested on Their weak sides orient the part 90* and test again. I believe they will do better.

What tolerances did you use with the 0.4mm nozzle to make the fit so snug?

You need a sign wave locking ying and yang bar.

Thanks 👍

The lap joint didn't fail the printed cf filament failed. Also you should have done a continuous black one as well because it was the weaker one.

Very nice comparison but the max load that you report are not always consistent to the max value that we see in the video of the test.. (40kg on the video vs 24kg reported). But all other aspects of your test methodology looks really good.

Perhaps a better way of weighting the scale is instead of adding weights to it, secure the entire apparatus to the floor or secured workbench, then fashion a hand crank and spool with some steel cable. The more you wind the cable the more force gets put on the scale. It would also be easier on your body 😂. Perhaps a quicker solution would be to buy a very heavy duty Ratchet strap fastened to the floor.

The joint should be rotated to the direction of the load, so they dont deflect apart. Perhaps even a design that offered locking in each plane.

Your test rig clearly needs some netting on each side to catch the launched parts out the side :-).

The screw lap joint looks like it would have been higher up in the rankings if the screws were in a straight or zig zag line along the length. The two screws side by side weakened the joint too much.

good video!

The two contact point block on the test rig skew the test results because the length of the joints varies.

Why gyroid instead of Cubic infill? Shouldn't the Cubic infill be the best for overall strength in all directions? Edit: For the drop lock you recorded 24.9kg yet then scale went to 43kg before it mostly failed and touched the foam. Your weight is from testing an already weakened piece.

Awesome video, please please try them with 5 top layers and 100 bottom layers or however many is needed to complete fill the model with zero infill. The print wont take that much longer and will be alot harder to break

start @14:36 your welcome.

Hand wound crank and pulley maybe?

you could use a hydraulic jack instead of weights. Than you have tons of weight:D A small 1 ton cylindrical jack is quite accessible both as price and as size

In sinewave replace heat insert with a nut

The tests should be done with the same filament for each parts and the different filament have different strengths. From the video the black parts break first.

Drop lock got robbed in this testing. It hit way higher numbers than got recorded

👍👍😎👍👍

Why do the continuous bar by hand when you can do it by seat? 🍑 Keepo

Gyroid made this whole test useless

sine wave seems promising.. change the heat insert with print in place nut :)