My 3D Printed Harmonic Drive that Performs Surprisingly Well!

this is my harmonic drive gearboxes like this are critical in achieving incredibly high torque in compact packages and why would i want to make this well i was hanging out my discord server one day and had a thought wouldn't it be awesome if i had my own robot arm to do complex coordinated motion holding my camera to capture some crazy b-roll obviously yes that would be so cool but being that i'm me the only option here was designing and building my own arm from scratch i wanted to make an arm powerful enough to hold my camera with a reasonable range capable of smooth somewhat accurate motion but here's the catch i wanted to make it as cheap as reasonably possible a lot of diy designs use bldc motors with o drive controllers which are really nice but this alone costs a bare minimum of 100 bucks per joint no thanks i wanted to use steppers instead a not nearly as powerful but way cheaper alternative but i need more torque a lot more so i'll need to gear it down there's a few different ways that we can do this one is to use belts with different size drive and driven pulleys but unfortunately if i wanted to get 10 newton meters from my steppers it would look something like this so that's not gonna work the other option is a gearbox these aren't nearly as efficient but they're a lot more compact there are some off-the-shelf options but as you may have guessed they're not cheap so in order to keep to my budget and try to keep this design accessible to all i decided to 3d print my own and luckily the youtube algorithm delivers i quickly began discovering tons of videos about stream weave gearboxes also known by one of their brand names harmonic drives these gearboxes use a semi-flexible tooth profile which is distorted by another component called the wave generator as the wave generator spins around it will slowly but surely rotate the flex spline one problem though these gearboxes are pretty chunky and quite difficult to design for high torques when using 3d printing the reason is this pretty long flexible region which has to be long to allow it to flex or does it spoiler no not really so with all the videos i watched as inspiration i began work on my own design but how exactly is this design different well rather than directly tying the flex spline to the output instead it uses the flex blind to drive another half of the gearbox this allows you to use a way shorter smaller and simpler flex spline since it doesn't have any cross-sectional bracing it can flex more easily of course this is not without its downsides but that's a problem for future me for this first iteration i decided to use an inside out hd belt as the flex spline which as you may have guessed is nice and flexy so time to test at first i tried using one of my osr control boards but one amp output just is not enough for these nema 23s i'm working on designing my own motherboard which will be for the entire robot arm and capable of driving these nema 23s at maximum power but all my components are out of stock so for now i just bought this cheap stepper driver on amazon to get going with my testing and since i haven't written any code yet i'm going to generate these pulses myself and sweet it works so i guess i'll actually write that code now and look at it go it's pretty smooth quiet seems like a really good first attempt with just one flaw it doesn't have much torque oh and it has a lot of backlash and it's way too big to use in my robot arm and this stepper is stalling way too easily um one sec the amazon page lied the steppers i'm using can handle up to three amps i can't really fix this for now so i guess i'll just fix all the mechanical issues and come back to this once i have my own motherboard so back to cad and beholds the smaller version which should be capable of similar output performance while both costing less and fitting in a smaller main bearing i also tighten up the tooth profile a bit to cut down on backlash and i printed a test stand so i can start actually measuring performance and compare versions so let it rip uh that's some something's wrong see this part that's that's supposed to be spinning okay so yeah this part broke so i went ahead made some changes and printed a new wave generator and we're back so time for another torque test [Music] the length of the moment arm i'm using for this is 200 millimeters to point of contact so using the force of the scale we've recorded six newton meters before maxing it out not bad but it's clear i need a better way to test these so i bought a 20 kilogram load cell and threw together a fancy little jig here and then i wrote way too much code to allow me to autonomously load and cycle test these gearboxes was that part necessary no but i did it anyways okay so rather than testing that last gearbox again let's just jump straight to the next design and the script i'm reading says that it worked perfectly and definitely didn't stall under no load so you might notice that this design is now using a 3d printed strain wave gear in addition to using a new tooth profile tooth tooth and you'd be right as it turns out the nice and flexy felt was a little too flexy and was introducing a lot of slop and compliance into the output of the final gearbox so i went ahead and modeled a new flex spline with a proper invuluted gear profile is this actually the right way to do a strain wave gear i'm really not too sure so if anybody here is an expert in designing gears and gearboxes please let me know in the comments because i'm pretty curious but at the very least this tooth profile is a lot deeper so it should transfer more torque before slipping should point is i got the dimensions wrong on that last one so the wave generator was pushing way too hard into the flex spline and distorting the housing significantly so after a few more revisions [Music] we arrive here the finalish form factor but still far from the final iteration this design has almost no backlash pretty good compliance considering it's 3d printed and can hit 10 newton meters of torque with this motor and so much more not that much more but it works and it works pretty well [Music] okay now it works pretty well so let's get testing with it i set this first test to stop at one newton meter so that i can figure out how much the gearbox is deflecting under minimal load as it turns out not too much then i set it to go as hard as it can yeah not bad not really not bad at all being able to apply 10 newton meters is only good if it can keep doing that time and time again so let's get life testing i wrote this script to cycle the arm at no load and periodically test to the maximum torque just to make sure it can still hit it and it'll keep doing this until it either fails or i get bored and stop it unfortunately in this case about 600 cycles later it suddenly started slipping like crazy so i took it apart and it turns out the wave generator was once again slipping on the motor shaft so i made a few tweaks printed a new part put it back in and off again but now it's only hitting 6 newton meters which is pretty weird i'm not sure exactly what happened but i must have accidentally modified some parameters before i reprinted the wave generator but i'm just going to go with it for now and see how it does and 1500 cycles later the torque output only slightly dropped which is really not bad and taking it apart to inspect i discovered something shocking yeah it broke a little bit it looks like mostly this is just layer delamination so i'm gonna reprint this part of the housing at a higher temperature and hopefully that will help improve durability and this new print got back up to eight newton meters which is good we're getting back to where we were so just a few more tweaks left before this design is ready to go but at this point it's probably pretty obvious my motor isn't gonna have enough torque to actually test this gearbox to failure which is annoying for the video but i guess that's a good problem to have so i'm just gonna put in a little elbow grease at first i tried using this load cell to apply force by hand to the arm while it's holding position but i was only able to exert 17 newton meters this way so i resorted to using my personally calibrated load cells and after exerting exactly 40 pounds the gearbox survived and it's still spinning smoothly and has barely any backlash not what i was expecting so one more try this time i'm not stopping until it stops and finally after at least 50 pounds of force not too surprising the flex spline just sheared down the middle and that part failure likely led to damaging the tooth profiles on both halves of the housing unfortunate but luckily my robot arm is never going to see this much force so i'm not concerned but if for some reason this is an issue i did buy some nylon filament to try to make some stronger flex blinds i've just heard that printing nylon is scary so i'm going to avoid that until i absolutely need to and now that i've destroyed my last gearbox i decided to print all the parts for a new one which as it turns out is a great excuse for me to show you exactly everything that goes into this gearbox so here's all the parts it takes about four hours to print then it just takes a few m3 bolts square nuts a little bit of glue and and just for my own sanity let's put this through one final torque test and see where we're at and thankfully we are back to 10 newton meters so what's next with this project well i'm gonna be building up a few more of these gear drives then i'm gonna be putting a lot more time into cad working on the arm it still has a long way to go but i'm really excited to see just where it's going i think i should be able to make a very capable arm and hopefully do it all for under 500. but anyway all that is for a future video as always a huge thanks to all my patreons if it weren't for all their support videos like this just would not be possible don't forget to subscribe so you don't miss any future projects and let me know in the comments below if you have any suggestions or feedback for me until next time

Channel: 3DprintedLife

Views: 253,557

Rating: 4.9418874 out of 5

Keywords: 3d printed harmonic drive, harmonic drive, strain wave gearbox, stepper motor harmonic drive, strain wave gear, robotic actuator, 3d print, 3d printed robot, 3d printed gearbox, 3d printed strain wave, 3d printed actuator, harmonic drive actuator, high torque 3d print, 3d printed robot actuator, robot actuator, stepper motor actuator, stepper motor gearbox, 3d printed project, 3d printed, robotics project, raspberry pi robot, Engineering project, Strain wave gear design

Id: Emvo3bLT-Z4

Length: 11min 48sec (708 seconds)

Published: Sun May 02 2021