He was using the term failure in an engineers sense. He was measuring how much torque it took to break the mechanism. If the answer is higher than anything it will see in actual use, then it's strong enough. The safety factor is how much higher. If the answer is too low, he has information about how to adjust the design.
As Thomas Edison said: “I have not failed. I've just found 10,000 ways that won't work.” “Many of life's failures are people who did not realize how close they were to success when they gave up.” I wonder how how two flex plates like you designed installed in opposite direction would hold up.
Same philosophy here. Despite the mechanism failing (i.e.: reaching a failure state such as slippage or breakage), the process of experimentation was a success as it yielded worthwhile information.
I honestly don't know any engineers (I'm in the US) that use imperial. The only time we do is when we're forced to by dang archaic machine shops! Metric is love, metric is life
Despite being in the united states I have fully metricized my shop, it makes collaborating online so much easier. Highly recommended. Although in my open source designs I try to use 3mm 8mm and 30cm dimensions as much as possible as there are very close imperial standard sizes so people can often use a design as-is with what they have on hand.
Whoops! Every time I said "plastic deformation" I meant "elastic deformation". I guess that's what I get for not writing scripts. Also, I filmed this video with a new camera, so I was still figuring out things like focus and white balance. Oh well :)
I guess the important word was "deformation". Didn't notice anything annoying with the camera... focus a little slow on close-ups, but better than most RUclips offerings.
Hmmm... What about making a dual-layer deforming piece, where one layer is conformal to provide springiness and the other incorporates a 'disc-and-socket' joint at either end of a 'bone' for resistance against torque without preventing deformation. Both layers would be part of a single model. There would just need to be a small gap between the disc and the socket to allow for rotation. (Three layers would be better as you can trap the rotating joint in between two conformal hinges, but that would require some thought in regards to support.) Does that make sense?
Maybe also add some radius or chamfer to the attachment point from the spring to the internal and external gear so the plastic won't break at the sharp corner.
And if you keep adding depth until it's as deep as a regular harmonic drive, you've basically created a stronger (better?) version of the regular harmonic drive.
I first saw a 3D printer in Chiang Mai a few years ago and realised how the future would unfold rapidly in the hands of young inventors. You are a great example of how mechanisms can be rapidly prototyped and improved, in ways that could never have been done last century. It's exciting to hear your thought processes, and see your working models.
Something that comes to mind, you could alternate the spring "direction" so that for every "left" spring there is a "right" spring. That way the mechanism has support in both directions. Another improvement would be to make the springs thicker vertically.
You mentioned that one torque direction could withstand more than the oher. So I thought that you could add the same pattern on the first (in the same part), like it is another 'layer', but mirrored, so it will compensate the inequality. Really loved this design, to me it's the most interesting I have seen that is 3D printed. I think you could expand the research from here to make a very good design, great work!
Another option may be to include springs in alternating directions. That would probably come at the expense of radial space, but would still allow for a low vertical profile
@@LeviJanssen I mean you could always just print two parts and assemble them together in some way. You could also print supports between the two layers of springs, but those would be hard to remove, I reckon.
@@LeviJanssen you can do that as a single print leaving a small gap between the two spring layers keeping a single hub. Also, increase the depth of the springs and gear without modifying the thickness. That should provide more strength without compromising the elasticity. Also, consider using TPE. Though, a TPU spring with PLA center would be even better. That, however, would require a printer able to print with two filaments, or a lot of fun doing it manually.
It's cool that people just put information like this out there. There's an entire different side of the spectrum which would try to find ways to patent and close off this knowledge you're willing to share. We really are coming into a new age. It makes me happy.
'Elastically' deformed, not plastically. Also, this project was AWESOME! Please keep developing. Torque is usually expensive - either in size or cost, so cheap torque will always be in demand.
Swap in some spring steel strips with the PLA as just housing? Junior hacksaw blades are a good source of thin spring steel which can (usually) be bent 90 degrees without breaking...
There's a youtube channel where all he does is test people's fan blade design. You should do the same, but with a torque transfer plate (or whatever you call this). I have an idea that I think would work.
At the very least give the name of the channel, Major Hardware, and you should also better describe the channel as he does far more than just the Fan Showdown series.
Great work Levi! I have a couple of suggestions - use fillets at each end of the compliant spring to distribute more force and reduce stress risers. You could have more of them per revolution, and you aren't limited to having them in just one plane, so you could layer them up for even greater torque transmission.
Awesome video! I'm working on my own harmonic drive myself aiming to be super compact as well, but using an inside-out HTD belt instead of 3D printing a flex spline. Initial results are promising but I still need to get a high power motor driver and do some torque tests!
well done, you are really close to a solution, both the vibration and the strength. All you have to do is print the compliant mechanism seperate from the spline, print two, put one of them up side down and glue them both into the spline. Now the two compliment each other, and the mechanism should be much stronger and there will be less vibrations
Hey if you want less "wobble" without over stressing the threads I would suggest adding some springs to the bolts. It does add "more load" to the threads at the same torque, but that also allows you to convert that load into frictional between the 3d printed part and the metal beam. This is kinda how we use springs on watercoolers for CPU's/GPU's, you don't want to crack the chip so you add spring loaded mounting brackets that distributes the load more evenly ensuring a "tighter" gap between the cooling surface and die... so in turns the force is applies more evenly. If you ruffen up the metal beam that would add more friction between the metal and plastic, kinda the opposite that we do with coolers. it's probably not going to make it "stiff" but it should prevent a couple of wibbly wobbly, timey wimey stuff to your presentation :)
ABS/ASA are finicky but they could work indeed. PETG only works up to a point, under repeat stress and certain applications of force it paradoxically crystallises and becomes brittle! But there are printable TPU/TPE varieties, and not all of them are that low Durometer, they could be of use as well, they're basically made to handle repeat stress. Polyamides (Nylons) as well after one-time conditioning.
I think that the curved part of the complient mechanism could be made thicker without compromising springiness. It may also be beneficial to use alternating left and right turning springs. Just a thought... Nice video! Thanks!
Cool Project! That's the tradeoff with Harmonic drives - no backlash, but your stiffness is compromised. Even happens on the indutrial drives. They are a lot stiffer than the plastic ones because they are metal, but not nearly as the stiffness you get out of a normal gearbox.
I designed a backdrivable 40:1 Cycloidal Drive which I am super happy with...all my attempts to get a harmonic drive going failed terribly!! This is super interesting!! Keep it up!! 🤩
What if you replaced the longest section of each of those flexures with an off the shelf metal strip or wire or something? You could beef up the plastic sections and have all of the bending be done by the metal.
Surely design is interactive with material dcience? Different materials have different properties to design with. PLA isn't a great choice for a flex element.
As long as you designed the mechanism to stay below the fatigue limit of the material, any material is a good material for the application. It all depends on what you are shooting for. If you are looking to minimize cost, but don't care about size, you could use a cheaper material and give it more room to flex without stressing the material much. If you care about size, but not cost, you could make it from a more expensive material that is better for higher stresses. And if you are really creative, you can come up with a design that optimizes cost, size, and performance using a mix of materials. Engineering should never be constrained to "this is the material for this application". That will severely limit creativity. Creativity, plus a clear, unrestrictive definition of the problem, the willingness to fail, and trusting but testing other's advice equals finding the best solution possible.
I think the best bet is to use a more flexible material. With something like ABS or PETG, you could get the same flexibility from thicker (and thus stronger) parts.
Great idea! The most obvious way to increase the capacity is to simply scale up the springy part in the Z direction. Why do you (and so many other makers) use such small bearings on the bearing bar? Seems to me that you'd want to use the largest you can fit, to produce a smoother bend in the flex spline, reducing fatigue and getting more teeth engaged and supported. Either that or use a bunch of small bearings placed along an ellipse to simulate the proper elliptical bearing of a harmonic drive, counting on the stiffness of the flex spline to produce a somewhat smooth curve across the spaces between each individual small bearing.
What if the whole bearing bar would be smaller, say only 2/3 of the diameter of the flex spline? Now you have whole outer 1/3 to try various compliant mechanisms to optimize both flexing and torque. And leading to space saving overall.
I used a harmonic drive back in the eighties, but it was different to the one you have. It had two outer gears, both same pitch diameter but one had one less tooth. The flex spline engaged in both the output would rotate one tooth for each rotation of the flex spline.
Addressing favoring direction: move the inner support to the middle and double the outer to both sides of the "spring". Angle the inner as "v" hitting the core allowing load to be transferred via compression and extension. You can do the same with the outer.
The super low parts cost for 3D printed harmonic drives makes them really favorable to cycloidal drives that rely heavily on bearings. To the point of the compliance being variable in rotation direction, you can mirror one of the flex splines and run 2 inside the same rigid spline connected to the same output. That should give you more symmetric torque, and also double the amount of material you've got in contact.
Would love to see you try this again. But this time, turn your spring. An inner core, an outer gear set, and two sidewalls with outward flex, but solid and without flex in the direction of motion. Think Goodyear tire. For bonus points, look up some of the solid tire concepts from Goodyear for the military.
I was building something similar, only a little bit more fancy. My thought: it's not ideal to make the compliant struts curved, since they would bend under tension -> reduced accuracy. I would make those struts straight, maybe even with thinned out joints at the ends. Also something I was trying out as well: there's a type two harmonic drive, where you have one flex spline that is basically just a toothed ring on an eliptic bearing, and two rings with inside teeth. One of those rings stays stationary while the other is connected to the output and they both have a tooth count difference of 2. This also makes a very compact drive, but it has high demands on the accuracy of the inside gears since any deviation of the tooth shape would result in backlash or binding.
I'm looking at similar aspects, but somewhat different conclusions.. First, at the end with it broken, notice how they seem to have 'curled up' at the ends. The opposing curves at the ends of the elements makes the longer element want deflect on a diagonal when under the 'compression' direction force. Instead of a curve and flexing at each end, they need a trapezoidal pillar as an anchor for each base connection. Then the element joined flat to the top of the trapezoids, make the joint much stronger so it can't bend out of line or curl. IOW strengthen the ends, and make all the flex of these elements happen in the longer middle area. Stronger end corners will keep the entire element from shifting out of line and make the force spread out into most of the longer element between the base pillars..
Not a "simple" concept but a very intelligent and eloquent solution to the problem - inspired work - congratulations. May I suggest movement limitation stops in the flexing wheel that will prevent the spring element from being overtaxed? This could cause lockup if the spring element is over-stressed but that is good - don't drive the mechanism beyond its mechanical limit. I wish that I could share a doodle here to visually explain what I'm talking about...
you could make half of the springs one direction and half in the other direction so it is still springy in opposite directions and a higher strength for both directions
Can't believe this is the first time I'm finding your channel! This video alone has earned a sub! This design with a less brittle material for the compliant gear could be very good.
So cool, hope I can learn to use what I learn at uni like you do. Best of luck making a reliable 3d printed harmonic drive. It would be fun to experiment and take your compliant mechnisum and extrude it on the y axis to add more torque but keep it compliant on the x and z axis. (Extrude the y axis until it's big enough to fill the empty space from the regular design)
I wonder how hard it would be to make a design that replaces the springs with metal spring blades and the outer circle that deforms with a timing belt. The only challenge with that would be to somehow attach the metal spring blades to the timing belt. Some kind of small riveting or stapling might be able to work with that if the timing belt is wide enough to allow enough teeth area to remain intact despite having the rivets/staples there. If it all came together you could get some really solid performance out of that thing. At that point it wouldn't matter if going one way is stronger than the other way, cause event the weaker one would be insanely strong for a mostly 3d printed design.
So I just discovered this video… fascinating for a non-engineer like me. My initial thought was that the compliant spine could be duplicated with the reverse orientation and combined. Therefore, when one is being stretched by the torque, the other is being compressed, and neither direction is favored.
Thanks ever so much. Especially for your explanation and show and tell on how a harmonic drive works. I have used many HD's in steerable antenna, but never understood exactly how they work, despite trying.
If the spring is indeed stronger one way vs the other way, you could possibly add in a mirror image of the spring system and that would definitively help, that way it can be under tension both ways instead of just one.
Heck yeah there are more optimizations to come! Make that spring twice as thick. Use 2 springs, one flipped for symmetrical load. Make more springs instead of only 6 around the circle. Great design! I think you've started something great!
That's a VERY nice design mate... The compliant cage makes it actually quite feasable to do in steel and just water-jet cut the middle. Maybe with a water-laser to keep as much material as possible.
I think you should alternate the direction of the springs, so that way half are always in tension. Should make the failure load a lot higher for the device
Very interesting way to get a gear ratio like that. Although you can increase torque in these methods, You still will only get the torque that the materials can handle. And thin plastic parts don't handle torque very well. But this does show potential if you are able to get your hands on more appropriate materials. Be cool to see that done. Also the only experiment that is a failure is one where you don't learn anything. I think we learned alot from this.
great video! One comment, for those of us who are not in the know of what these sort of devices are used for, a intro of common uses or what you might be using them for would be spectacular. 👍
The broken springs are mostly broken past the bend. I don't think they need to even have a bend like that, and it doesn't need to be where they flex the most either. I'm thinking a design with concentric circles connected by short thick spokes, out of phase such that each outer spoke sits in the middle between two inner spokes and vice versa, may be worth trying out. The test showed where the weakest point for this particular design was for the rotary force, not that it had to be that weak there. We want rotational stiffness and radial flexibility for this task. We could also make approximate evaluations with FEM (built into e.g. FreeCAD). A part with these same goals is typically mounted in 3D printers: the shaft coupler. Compliant versions usually have a spiral cut in their midsection, allowing some flexing. There its main purpose is to adapt to slight differences in axis alignment. Unfortunately, many makers then skip the part about stabilizing the output shaft against linear axial force.
this made with titanium and more springs would work great. the titanium would be flexible enough to work as a spring and be strong enough to not snap due to torque. good idea!
Make the springy bits smaller and alternate their direction to prevent them from collapsing in your torque test. If they get less springy, make them narrower and higher.
You can maybe achieve an bidirectional rupture force if you mirror the design and kind of glue them together... Furthermore, you may be able to increase the number of Springs of this design if you give some angle to then and pile one above (alongside) the other...
The direction bias could probably be fixed with mirroring the springs backwards, like a modified Y shape. Maybe have to sets of rings s othere's no bias between the hub or ring having more connections to the Ys, as in invert every other one so that the connections are -1-2-1-2- on the ring and -2-1-2-1- on the hub, where the horizontal of the Ys are offset to form two rings of the concentric portion; this should also add some strength overall. Better materials would also help with the strength quality of a compliant harmonic, just as there's a reason industrial-grade harmonics always use metals and not plastic. This could easily evolve into a viable solution, and with it being open sourced it'd make industrial-grade robotics more affordable.
Try T with a small bump on the ends sort of like a Y shape, but more ability to bend inwards in your mechanism. Yet still connect to outer ring shape. Should bend but still be rigid in both directions. Allowing your reverse drive. just a thought.
About a month ago I designed and 3d printed a flex-splineless harmonic drive. I was working out shrinkage dimensions on new filament at the same time, so mine is quite clunky, but it basically negates the need for the spline altogether and worked well enough as a proof of concept in my view. Maybe it's not of interest if you're specifically looking to explore compliant mechanisms, seeing as I took the flex out of flex-spline lol.
Hey, another idea: Try putting teeth inside of the flex spline directly above the wave generator. Then put a gear with the diameter equal to the short side of the flexed spine to transfer the power. If I am not misunderstanding the concept this should be similar to a two stage harmonic drive taking less space than a single one.
Add some meat/plastic on the gear ring between each spring and design a slot through with a metal pin could be slipped in during 3D printing. The torque could then be transferred more directly from the outer ring to the center via the pins as opposed to the being transferred via the springs. with a hole in the center of the long straight section of the spring tuning the spring dimension might be required. Very nice though.
Get rid of all curves in the torque-transferring spokes. For example the outside (teeth) and inside (hub) should be 2 separate pieces with STRAIGHT spokes connecting them, hinged at the connection points (from the inside hub to the flexible outside gear). There will need to be some axial-movement play in the hinges.
I just got an idea: make bendable ring and rigid disk coupled by teeth so they can move absolutely freely in radial direction against eachother but get in traction in tangential direction. TL;DR: Make 1 : 1 gear transmission between elastic ring and rigid disk. That will introduce backlash but increase torque dramatically.
You could probably make the flat flex spline into a metal flexture using subtractive manufacturing, itd give better strength, then you could layer two mirrored gears to allow for backdrivability
I think the long parts of that compliant element should be straight, not tangential. Sure, they're transmitting a torque between the hub and the gear teeth, but each element actually "sees" a force, either in compression or in tension.
Amazing! Very well done. Just a thought... Perhaps you could have two spring mechanisms on top of each other, one biased clockwise and the other anti-clockwise. This way the overall mechanism would have the same torque capability in either direction. You might have to place some type of grease or thin spacer between the two springs to reduce friction.
It seems to me that the conventional solution is already a compliant component and there's room for it inside your housing. Also, just like the conventional solution your approach requires exotic metallurgy to make it work. I also don't think any of this diminishes your exploration of the topic, thank you.
A bit late to the party here, but what about having a 3d printed hub and ring, with U shaped metal links? The U would have to have a tab off the ends to engage the hub and ring, and the body of the U inline with the rotation. They can be in alternating directions to balance the load between directions as well.
Great video. A weird idea: While it increases the complexity (maybe too much for hobbyists), couldn't an outrunner style motor be placed inside a traditional flex spline? Mount the wave generator ellipses with a flex bearing on the housing of the motor at the mount end and a normal bearing at the other end. Fit it inside the normal flex spline. Put this inside the outer housing, with the motor wires coming out the bottom (where your motor shaft enters the gear), and screw the motor to the housing. It might need a tall housing with a bearing like you have to keep the spline in place but should still be even more compact and solve your primary issue with the harmonic drive without resorting to compliant mechanisms.
Man, compliant mechanisms and 3D printing makes such an interesting combination... I designed a compliant mechanism spider robot (you can check it out) and it's so fun! Great work and keep it coming :)
Failure??? You were creative, you designed, you build, you tested, you learned and you shared. Hence, a big success! In one word: IMPRESSIVE!
As a close friend once said to me: “We live and we learn. I am now wiser. This is not a complete loss.”
He was using the term failure in an engineers sense. He was measuring how much torque it took to break the mechanism. If the answer is higher than anything it will see in actual use, then it's strong enough. The safety factor is how much higher. If the answer is too low, he has information about how to adjust the design.
Failure might just be as useful of a result as a success.
As Thomas Edison said: “I have not failed. I've just found 10,000 ways that won't work.” “Many of life's failures are people who did not realize how close they were to success when they gave up.”
I wonder how how two flex plates like you designed installed in opposite direction would hold up.
Same philosophy here. Despite the mechanism failing (i.e.: reaching a failure state such as slippage or breakage), the process of experimentation was a success as it yielded worthwhile information.
you have no idea how much I appreciate you using metric instead of imperial
Always have!
I honestly don't know any engineers (I'm in the US) that use imperial. The only time we do is when we're forced to by dang archaic machine shops! Metric is love, metric is life
@@3DprintedLife imperial is less common in engineering, more of a general purpose measurement.
Come on... how hard is it when the software does all the heavy lifting??
Despite being in the united states I have fully metricized my shop, it makes collaborating online so much easier. Highly recommended.
Although in my open source designs I try to use 3mm 8mm and 30cm dimensions as much as possible as there are very close imperial standard sizes so people can often use a design as-is with what they have on hand.
Whoops! Every time I said "plastic deformation" I meant "elastic deformation". I guess that's what I get for not writing scripts. Also, I filmed this video with a new camera, so I was still figuring out things like focus and white balance. Oh well :)
I guess the important word was "deformation".
Didn't notice anything annoying with the camera... focus a little slow on close-ups, but better than most RUclips offerings.
It was elastic deformation of plastic, so I’ll allow it.
Hmmm... What about making a dual-layer deforming piece, where one layer is conformal to provide springiness and the other incorporates a 'disc-and-socket' joint at either end of a 'bone' for resistance against torque without preventing deformation. Both layers would be part of a single model. There would just need to be a small gap between the disc and the socket to allow for rotation. (Three layers would be better as you can trap the rotating joint in between two conformal hinges, but that would require some thought in regards to support.) Does that make sense?
perhaps if u use metal like what actual springs are made of i could handle more pressure
maybe?
I think you can give the compliant spring more depth ( not width) it will keep its elasticity and give more support against torque
Maybe also add some radius or chamfer to the attachment point from the spring to the internal and external gear so the plastic won't break at the sharp corner.
@@julstr6303 ^ this
@@julstr6303 very much dis^
And if you keep adding depth until it's as deep as a regular harmonic drive, you've basically created a stronger (better?) version of the regular harmonic drive.
@@frother thats not how mafia works
I first saw a 3D printer in Chiang Mai a few years ago and realised how the future would unfold rapidly in the hands of young inventors. You are a great example of how mechanisms can be rapidly prototyped and improved, in ways that could never have been done last century.
It's exciting to hear your thought processes, and see your working models.
Something that comes to mind, you could alternate the spring "direction" so that for every "left" spring there is a "right" spring. That way the mechanism has support in both directions. Another improvement would be to make the springs thicker vertically.
Compliant mechanisms and strain wave drives are both cool. So combining them is even cooler! I'd love to see more development of this project.
You mentioned that one torque direction could withstand more than the oher. So I thought that you could add the same pattern on the first (in the same part), like it is another 'layer', but mirrored, so it will compensate the inequality.
Really loved this design, to me it's the most interesting I have seen that is 3D printed. I think you could expand the research from here to make a very good design, great work!
Another option may be to include springs in alternating directions. That would probably come at the expense of radial space, but would still allow for a low vertical profile
I was thinking the same thing, I wasn't sure how to implement it easily. I guess layering it wouldn't be too difficult.
ah, you had the same thought
@@LeviJanssen I mean you could always just print two parts and assemble them together in some way. You could also print supports between the two layers of springs, but those would be hard to remove, I reckon.
@@LeviJanssen you can do that as a single print leaving a small gap between the two spring layers keeping a single hub. Also, increase the depth of the springs and gear without modifying the thickness. That should provide more strength without compromising the elasticity. Also, consider using TPE. Though, a TPU spring with PLA center would be even better. That, however, would require a printer able to print with two filaments, or a lot of fun doing it manually.
I think this is now my fav new channel for 3d printing.
Original ideas, competitive, first principles thinking, proper testing; keep it up dude.
It's cool that people just put information like this out there.
There's an entire different side of the spectrum which would try to find ways to patent and close off this knowledge you're willing to share.
We really are coming into a new age. It makes me happy.
'Elastically' deformed, not plastically. Also, this project was AWESOME! Please keep developing. Torque is usually expensive - either in size or cost, so cheap torque will always be in demand.
Yeah, I had to correct myself at least once, I probably missed other occasions.
...Like Ford Mustangs.
I agree. Keep up the good work! What you're doing is awesome. And interesting to boot!
Swap in some spring steel strips with the PLA as just housing?
Junior hacksaw blades are a good source of thin spring steel which can (usually) be bent 90 degrees without breaking...
There's a youtube channel where all he does is test people's fan blade design. You should do the same, but with a torque transfer plate (or whatever you call this). I have an idea that I think would work.
that's a good idea ^^
Not sure what i mean - not sketched - 'radial slots'
At the very least give the name of the channel, Major Hardware, and you should also better describe the channel as he does far more than just the Fan Showdown series.
@@xaytana I came here to say this. He's a smart guy over at Major Hardware, moreso than he lets on in his videos
Great work Levi!
I have a couple of suggestions - use fillets at each end of the compliant spring to distribute more force and reduce stress risers. You could have more of them per revolution, and you aren't limited to having them in just one plane, so you could layer them up for even greater torque transmission.
Awesome video! I'm working on my own harmonic drive myself aiming to be super compact as well, but using an inside-out HTD belt instead of 3D printing a flex spline. Initial results are promising but I still need to get a high power motor driver and do some torque tests!
Please continue with this project. Its really interesting.
well done, you are really close to a solution, both the vibration and the strength. All you have to do is print the compliant mechanism seperate from the spline, print two, put one of them up side down and glue them both into the spline. Now the two compliment each other, and the mechanism should be much stronger and there will be less vibrations
Very interesting! I like your style of going through the design history, so we understand the reasoning behind it. Keep up the good work!
Appreciate the use of engineering units ;)
Hey if you want less "wobble" without over stressing the threads I would suggest adding some springs to the bolts.
It does add "more load" to the threads at the same torque, but that also allows you to convert that load into frictional between the 3d printed part and the metal beam.
This is kinda how we use springs on watercoolers for CPU's/GPU's, you don't want to crack the chip so you add spring loaded mounting brackets that distributes the load more evenly ensuring a "tighter" gap between the cooling surface and die... so in turns the force is applies more evenly.
If you ruffen up the metal beam that would add more friction between the metal and plastic, kinda the opposite that we do with coolers.
it's probably not going to make it "stiff" but it should prevent a couple of wibbly wobbly, timey wimey stuff to your presentation :)
Awesome! More compact, less parts and no imbalance.
Thats amazing, I can see PETG or ABS making a huge difference in strength. I would go at it again, good video! Thanks for sharing
ABS/ASA are finicky but they could work indeed. PETG only works up to a point, under repeat stress and certain applications of force it paradoxically crystallises and becomes brittle! But there are printable TPU/TPE varieties, and not all of them are that low Durometer, they could be of use as well, they're basically made to handle repeat stress. Polyamides (Nylons) as well after one-time conditioning.
I think that the curved part of the complient mechanism could be made thicker without compromising springiness. It may also be beneficial to use alternating left and right turning springs. Just a thought... Nice video! Thanks!
Cool Project! That's the tradeoff with Harmonic drives - no backlash, but your stiffness is compromised. Even happens on the indutrial drives. They are a lot stiffer than the plastic ones because they are metal, but not nearly as the stiffness you get out of a normal gearbox.
I designed a backdrivable 40:1 Cycloidal Drive which I am super happy with...all my attempts to get a harmonic drive going failed terribly!!
This is super interesting!! Keep it up!! 🤩
What if you replaced the longest section of each of those flexures with an off the shelf metal strip or wire or something? You could beef up the plastic sections and have all of the bending be done by the metal.
the solution should be from the design first before looking for the material science...
Surely design is interactive with material dcience? Different materials have different properties to design with.
PLA isn't a great choice for a flex element.
@@anoirbentanfous shouldn't you keep material properties in mind when designing? And designing for one material might not work for another
As long as you designed the mechanism to stay below the fatigue limit of the material, any material is a good material for the application. It all depends on what you are shooting for. If you are looking to minimize cost, but don't care about size, you could use a cheaper material and give it more room to flex without stressing the material much. If you care about size, but not cost, you could make it from a more expensive material that is better for higher stresses. And if you are really creative, you can come up with a design that optimizes cost, size, and performance using a mix of materials. Engineering should never be constrained to "this is the material for this application". That will severely limit creativity. Creativity, plus a clear, unrestrictive definition of the problem, the willingness to fail, and trusting but testing other's advice equals finding the best solution possible.
Pretty elegant design. Curious to see what geometry, layering and other stuff you will create to put more springs and material to handle the torque.
I think the best bet is to use a more flexible material. With something like ABS or PETG, you could get the same flexibility from thicker (and thus stronger) parts.
1:04 I think you meant elastically deformed. If that piece was being plastically deformed repeatedly, it would fail due to fatigue very quickly.
Great idea! The most obvious way to increase the capacity is to simply scale up the springy part in the Z direction.
Why do you (and so many other makers) use such small bearings on the bearing bar? Seems to me that you'd want to use the largest you can fit, to produce a smoother bend in the flex spline, reducing fatigue and getting more teeth engaged and supported. Either that or use a bunch of small bearings placed along an ellipse to simulate the proper elliptical bearing of a harmonic drive, counting on the stiffness of the flex spline to produce a somewhat smooth curve across the spaces between each individual small bearing.
Mostly cost I bet. Smaller bearings are cheaper both in material cost and in shipping cost.
What if the whole bearing bar would be smaller, say only 2/3 of the diameter of the flex spline? Now you have whole outer 1/3 to try various compliant mechanisms to optimize both flexing and torque. And leading to space saving overall.
Very specific, very cool. Loved the style!
Brilliant start of something bigger. I'm calling it.
I used a harmonic drive back in the eighties, but it was different to the one you have. It had two outer gears, both same pitch diameter but one had one less tooth. The flex spline engaged in both the output would rotate one tooth for each rotation of the flex spline.
Addressing favoring direction: move the inner support to the middle and double the outer to both sides of the "spring". Angle the inner as "v" hitting the core allowing load to be transferred via compression and extension. You can do the same with the outer.
I'm just impressed all around. Well done and nice display of your knowledge.
Man, congratulations! That's impressive. It broke on load, but it's just a trial and error process. Please keep on!
This is freakin awesome!! I'd been thinking about harmonic drives for like the last year and a half; this never even occurred to me!
The RUclips algorithm brought me here and I'm not complaining :) Good work!
The super low parts cost for 3D printed harmonic drives makes them really favorable to cycloidal drives that rely heavily on bearings. To the point of the compliance being variable in rotation direction, you can mirror one of the flex splines and run 2 inside the same rigid spline connected to the same output. That should give you more symmetric torque, and also double the amount of material you've got in contact.
Impressive! Harmonic Drives do fill a gap where hole bunch of alternative solutions should be possible..good approach!
Would love to see you try this again. But this time, turn your spring. An inner core, an outer gear set, and two sidewalls with outward flex, but solid and without flex in the direction of motion. Think Goodyear tire. For bonus points, look up some of the solid tire concepts from Goodyear for the military.
I was building something similar, only a little bit more fancy. My thought: it's not ideal to make the compliant struts curved, since they would bend under tension -> reduced accuracy. I would make those struts straight, maybe even with thinned out joints at the ends.
Also something I was trying out as well: there's a type two harmonic drive, where you have one flex spline that is basically just a toothed ring on an eliptic bearing, and two rings with inside teeth. One of those rings stays stationary while the other is connected to the output and they both have a tooth count difference of 2. This also makes a very compact drive, but it has high demands on the accuracy of the inside gears since any deviation of the tooth shape would result in backlash or binding.
I'm looking at similar aspects, but somewhat different conclusions..
First, at the end with it broken, notice how they seem to have 'curled up' at the ends.
The opposing curves at the ends of the elements makes the longer element want deflect on a diagonal when under the 'compression' direction force.
Instead of a curve and flexing at each end, they need a trapezoidal pillar as an anchor for each base connection. Then the element joined flat to the top of the trapezoids, make the joint much stronger so it can't bend out of line or curl. IOW strengthen the ends, and make all the flex of these elements happen in the longer middle area. Stronger end corners will keep the entire element from shifting out of line and make the force spread out into most of the longer element between the base pillars..
"It's not actually this bad, it's actually really good." Best disclaimer ever. Will keep that one in my back pocket. 😂
you can introduce some sliding slots to transmit the rotation force, and then use smaller deeper springs for radius changes.
Great engineering video!
Not a "simple" concept but a very intelligent and eloquent solution to the problem - inspired work - congratulations.
May I suggest movement limitation stops in the flexing wheel that will prevent the spring element from being overtaxed?
This could cause lockup if the spring element is over-stressed but that is good - don't drive the mechanism beyond its mechanical limit.
I wish that I could share a doodle here to visually explain what I'm talking about...
Rather DeVinci of you , Well Done !
Made of traditional spring steel, your flexure design is unique & novel
Super cool would love to see it scaled up and tested again
you could make half of the springs one direction and half in the other direction so it is still springy in opposite directions and a higher strength for both directions
Can't believe this is the first time I'm finding your channel! This video alone has earned a sub! This design with a less brittle material for the compliant gear could be very good.
So cool, hope I can learn to use what I learn at uni like you do. Best of luck making a reliable 3d printed harmonic drive. It would be fun to experiment and take your compliant mechnisum and extrude it on the y axis to add more torque but keep it compliant on the x and z axis. (Extrude the y axis until it's big enough to fill the empty space from the regular design)
Behold, a whole compliant CVT! I can't wait to put it in my Compliant Robin
I wonder how hard it would be to make a design that replaces the springs with metal spring blades and the outer circle that deforms with a timing belt.
The only challenge with that would be to somehow attach the metal spring blades to the timing belt.
Some kind of small riveting or stapling might be able to work with that if the timing belt is wide enough to allow enough teeth area to remain intact despite having the rivets/staples there.
If it all came together you could get some really solid performance out of that thing.
At that point it wouldn't matter if going one way is stronger than the other way, cause event the weaker one would be insanely strong for a mostly 3d printed design.
So I just discovered this video… fascinating for a non-engineer like me. My initial thought was that the compliant spine could be duplicated with the reverse orientation and combined. Therefore, when one is being stretched by the torque, the other is being compressed, and neither direction is favored.
Super cool video dude! First I’ve heard of this particular mechanism
Thanks ever so much. Especially for your explanation and show and tell on how a harmonic drive works. I have used many HD's in steerable antenna, but never understood exactly how they work, despite trying.
This is an excellent design! Those springs solves a big problem with printed harmonic drives.
If the spring is indeed stronger one way vs the other way, you could possibly add in a mirror image of the spring system and that would definitively help, that way it can be under tension both ways instead of just one.
Very clever, and clear explanation of the results. Appreciated.
Very educative thank you.
Good work with your design, too! It seems like a natural fit (harmonic drives being almost compliant mechanisms themselves).
Add some spring layers in opposing directions. Slightly thicker, but should work. Other than that, a better material choice than PLA would help.
Heck yeah there are more optimizations to come!
Make that spring twice as thick. Use 2 springs, one flipped for symmetrical load.
Make more springs instead of only 6 around the circle.
Great design! I think you've started something great!
4:36 in the top left...What went wrong with that one? It looks the same as the final spring, but had 8 springs around the circle. Was it too stiff?
It had eight springs and was significantly stiffer, yes.
That's a VERY nice design mate... The compliant cage makes it actually quite feasable to do in steel and just water-jet cut the middle. Maybe with a water-laser to keep as much material as possible.
I think you should alternate the direction of the springs, so that way half are always in tension. Should make the failure load a lot higher for the device
Wow, your videos have become much better since 2019. Great work!
Very interesting way to get a gear ratio like that. Although you can increase torque in these methods, You still will only get the torque that the materials can handle. And thin plastic parts don't handle torque very well. But this does show potential if you are able to get your hands on more appropriate materials. Be cool to see that done.
Also the only experiment that is a failure is one where you don't learn anything. I think we learned alot from this.
this is absolutely genius design. very well done
Amazing work! Well done.
great video! One comment, for those of us who are not in the know of what these sort of devices are used for, a intro of common uses or what you might be using them for would be spectacular. 👍
Levi: I need a compliant mechanism that doesn't favor one direction
The Compliant Mechanism: 🎶You don't know you're beautiful!🎶
The broken springs are mostly broken past the bend. I don't think they need to even have a bend like that, and it doesn't need to be where they flex the most either. I'm thinking a design with concentric circles connected by short thick spokes, out of phase such that each outer spoke sits in the middle between two inner spokes and vice versa, may be worth trying out. The test showed where the weakest point for this particular design was for the rotary force, not that it had to be that weak there. We want rotational stiffness and radial flexibility for this task. We could also make approximate evaluations with FEM (built into e.g. FreeCAD).
A part with these same goals is typically mounted in 3D printers: the shaft coupler. Compliant versions usually have a spiral cut in their midsection, allowing some flexing. There its main purpose is to adapt to slight differences in axis alignment. Unfortunately, many makers then skip the part about stabilizing the output shaft against linear axial force.
Very well done! I'm keeping my eye on this. I'm hoping to start working on a small robot arm soon, and this is definitely has potential.
this made with titanium and more springs would work great. the titanium would be flexible enough to work as a spring and be strong enough to not snap due to torque. good idea!
Make the springy bits smaller and alternate their direction to prevent them from collapsing in your torque test. If they get less springy, make them narrower and higher.
You can maybe achieve an bidirectional rupture force if you mirror the design and kind of glue them together... Furthermore, you may be able to increase the number of Springs of this design if you give some angle to then and pile one above (alongside) the other...
looking forward to seeing another iteration of the design
To reduce wear on it, put petroleum jelly in it as a grease.
The direction bias could probably be fixed with mirroring the springs backwards, like a modified Y shape. Maybe have to sets of rings s othere's no bias between the hub or ring having more connections to the Ys, as in invert every other one so that the connections are -1-2-1-2- on the ring and -2-1-2-1- on the hub, where the horizontal of the Ys are offset to form two rings of the concentric portion; this should also add some strength overall. Better materials would also help with the strength quality of a compliant harmonic, just as there's a reason industrial-grade harmonics always use metals and not plastic. This could easily evolve into a viable solution, and with it being open sourced it'd make industrial-grade robotics more affordable.
Try T with a small bump on the ends sort of like a Y shape, but more ability to bend inwards in your mechanism. Yet still connect to outer ring shape. Should bend but still be rigid in both directions. Allowing your reverse drive. just a thought.
The red plastic inner gear reminds me of the adapters used to play 45 RPM records on 33 RPM turntables.
About a month ago I designed and 3d printed a flex-splineless harmonic drive. I was working out shrinkage dimensions on new filament at the same time, so mine is quite clunky, but it basically negates the need for the spline altogether and worked well enough as a proof of concept in my view. Maybe it's not of interest if you're specifically looking to explore compliant mechanisms, seeing as I took the flex out of flex-spline lol.
Once again an original idea and a great quality video. Keep it up mate! Great stuff
I think you can add a lot of overlapping springs, seems like a great idea and great base for further research.
Hey, another idea: Try putting teeth inside of the flex spline directly above the wave generator. Then put a gear with the diameter equal to the short side of the flexed spine to transfer the power. If I am not misunderstanding the concept this should be similar to a two stage harmonic drive taking less space than a single one.
Add some meat/plastic on the gear ring between each spring and design a slot through with a metal pin could be slipped in during 3D printing. The torque could then be transferred more directly from the outer ring to the center via the pins as opposed to the being transferred via the springs. with a hole in the center of the long straight section of the spring tuning the spring dimension might be required. Very nice though.
Get rid of all curves in the torque-transferring spokes. For example the outside (teeth) and inside (hub) should be 2 separate pieces with STRAIGHT spokes connecting them, hinged at the connection points (from the inside hub to the flexible outside gear). There will need to be some axial-movement play in the hinges.
There are strain wave drives without the cup. One outer gear simply has the same number of teeth as the... "belt", while it still flexes
You have a bright future ahead of you my friend.
Very educational. Well done Levi.
This was awesome I love to see an optimized version of it👌
I just got an idea: make bendable ring and rigid disk coupled by teeth so they can move absolutely freely in radial direction against eachother but get in traction in tangential direction.
TL;DR: Make 1 : 1 gear transmission between elastic ring and rigid disk. That will introduce backlash but increase torque dramatically.
You could probably make the flat flex spline into a metal flexture using subtractive manufacturing, itd give better strength, then you could layer two mirrored gears to allow for backdrivability
This is AWESOME!!
Great job! Keep developing.
I think the long parts of that compliant element should be straight, not tangential. Sure, they're transmitting a torque between the hub and the gear teeth, but each element actually "sees" a force, either in compression or in tension.
that is super cool, this is indeed how they have to be made. Very cool, thanks!
Amazing! Very well done. Just a thought... Perhaps you could have two spring mechanisms on top of each other, one biased clockwise and the other anti-clockwise. This way the overall mechanism would have the same torque capability in either direction. You might have to place some type of grease or thin spacer between the two springs to reduce friction.
It seems to me that the conventional solution is already a compliant component and there's room for it inside your housing. Also, just like the conventional solution your approach requires exotic metallurgy to make it work. I also don't think any of this diminishes your exploration of the topic, thank you.
A bit late to the party here, but what about having a 3d printed hub and ring, with U shaped metal links? The U would have to have a tab off the ends to engage the hub and ring, and the body of the U inline with the rotation. They can be in alternating directions to balance the load between directions as well.
Great video.
A weird idea: While it increases the complexity (maybe too much for hobbyists), couldn't an outrunner style motor be placed inside a traditional flex spline?
Mount the wave generator ellipses with a flex bearing on the housing of the motor at the mount end and a normal bearing at the other end.
Fit it inside the normal flex spline.
Put this inside the outer housing, with the motor wires coming out the bottom (where your motor shaft enters the gear), and screw the motor to the housing.
It might need a tall housing with a bearing like you have to keep the spline in place but should still be even more compact and solve your primary issue with the harmonic drive without resorting to compliant mechanisms.
Man, compliant mechanisms and 3D printing makes such an interesting combination... I designed a compliant mechanism spider robot (you can check it out) and it's so fun! Great work and keep it coming :)