Compliant Harmonic Drive (3D Printed)

yeah it's working that's super cool [Music] hey i'm levi and i'd like to talk about harmonic drives if you're unaware a harmonic drive or strain wave gear is a high gear reduction system and it's become a staple of robotics for years they can allow for extremely high reduction ratios little wear and are sometimes even back drivable and yet i really don't like them yeah i really don't like them at all i much prefer psychological drives except that i may have a solution which could redeem them for me but before i show you and test out my solution to one of the inherent problems to all harmonic drives i figure i should probably explain how they work this is an exploded view of an industrial harmonic drive and this is a simplified cross-section of a harmonic drive the blue circle is the circular spline and it is fixed in place the red circle is the flex spline and it is being plastically deformed by the green ellipse which is the wave generator and the wave generator isn't often an elliptical bearing the idea is pretty simple in that you have teeth in both the circular and flex splines except both of them have different numbers of teeth usually the flex spline will have two fewer teeth than the circular spline so then as the wave generator turns the gears mesh and the flex spline is forced to slowly rotate due to it having fewer teeth if you watch just a single tooth as the wave generator goes around you can see that it slowly crawls over one tooth at a time thus the output rotation which is derived from the flex spline is significantly slower than the input rotation which is on the wave generator so then what is my issue with the harmonic drive well it's this part right here a harmonic drive has to transmit rotation from the flex spline which of course is flexing to the output which is rigid so the solution is that they add a bunch of extra depth here so that this side can bend while this side remains rigid what that means is that the harmonic drive ends up with a lot of wasted space on the inside just because of this dead zone so the primary reason i have avoided harmonic drives is because they are not very compact or at least they're not as compact as they could be but anyways i may have a solution for this problem and if you looked at the title of the video you may know that it involves compliant mechanisms so then what is a compliant mechanism a compliant mechanism is a flexible mechanism that transmits forces and torques via elastic deformation instead of pivoting or sliding they bend or squish and they're pretty cool i designed and 3d printed these little plastic sticks so that i could get a better idea of the material properties and the feel of 3d printed pla i designed these in point four millimeter increments because that's the diameter of my 3d printer nozzle so essentially this is one pass two pass three four five six passes by the 3d printer i've always thought of pla as being really brittle and not wanting to bend at all but these are actually pretty bendy this first one is point four millimeter thickness and i can bend it a lot so i'm trying to stay within the range of elastic deformation that's where you can deform it and it'll snap back and i can go really far before the material is actually damaged i never thought i could see pla bend so far and even the 0.8 millimeter bends a ton and the 1.2 millimeter still bends quite a bit when i mess around with these sticks i really like the 1.2 1.6 and 2 millimeter versions so i went ahead and designed some actual springs of those thicknesses so that i could further see how they behave and these springs are almost surprisingly springy despite some of these being on the thicker end of what i tested out the springs are almost bendier than the actual sticks even this two millimeter spring deforms by up to almost a centimeter so what this confirmed for me is that not only does the thickness and the material properties matter a lot but the geometry matters almost more so this brings me to my solution how am i going to make a harmonic drive more compact my solution is to use a flat axial compliant spring to transmit rotation from the flex spline to the output all in one plane i'll show you what i mean having a better understanding for the pla and how it bends i started to try and design something that i could actually use and i ended up designing quite a few these were all failed iterations i was looking for something very specific in my compliant mechanism i needed it to be able to squish horizontally and vertically really in any rotational axis but it needed to be able to not turn under a torque the sides are being squished in and out constantly but the outside needs to rotate just as much as the inside otherwise it's inaccurate or it'll break so all of these designs either didn't squish enough for example this one barely squishes at all or they didn't properly transmit the torque this one i can rotate the inside without the outside rotating and i eventually came up with this guy this one can be squished by quite a bit which is a lot more than it needs to be it only needs to squish by about a millimeter in each axis but it does do a pretty good job of transmitting torque if the outside is rotating the inside will also want to rotate what's special about this design is the geometry of these springs you can see that they curve along the circumference of the circle so the spring is perpendicular to any squishing force so it can bend easily but to any torque it is parallel to it so it can resist it pretty significantly so this is what i'll use in my final design i took what i learned from my 3d printed compliant mechanism test to design the full actuator and then i 3d printed everything and i even got some red filament for a little bit more on camera pop which is pretty new for me all right i've printed all my pieces so now i'm ready to start assembly the parts list for this actuator is actually pretty small i have this which is the circular spline and the outer housing then i have the flex spline here in red i have a bearing to go on top i have the bearing bar which is what will be flexing the flex spline and i have this output i'm going to have to put multiple pieces together at the same time for things to cooperate so i'm going to start by putting these little screws inside the corresponding screw holes at the bottom of the base frame now we'll put my bearing bar inside of my flex spline you can see it's squishing it out a little bit it's not quite a circle it's more of an oval now i'll put the flex blind inside the circular spline and i'll have to make sure that the gaps on both sides are about the same the flex spline is in and you might be able to see that we have those equal gaps on either side so now i can put this on the motor and stick the shaft through my d-hole in my bearing bar there we go now i can tighten down those screws before i attach the output to here i want to make sure that everything's working as expected i'm using a drv8825 stepper motor driver here and i have a little potentiometer that i can use to control the speed in other direction so here's a first test yeah it's working look at that you can see that that little bit of flexing and of course that overall output motion that's super cool such a good looking motion now i can press this bearing into the top here and then my output will go inside the bearing i also have some bolts that will go through the bearing and thread into this flex spline [Music] the output is now on you and see if it actually works oh look at that it's totally working i can't stop it with my finger so it's a good sign to begin with very cool i should also mention before we move on the gear reduction of this the gear reduction of a harmonic drive is modeled by the difference between the teeth of the circular and flex splines divided by the teeth of the flex spline i have 92 teeth on my circular and 90 on my flex spline so that means that this whole thing has a reduction of 45 to 1 which is pretty significant for this compact size but now that i've got this thing together it's time for me to do some real torque tests okay so i now have the actuator clamped kind of crudely to the edge of my bench and i've got this aluminum bar attached to it then my motor driver is behind it here let's see if we can't turn it that's pretty good quick disclaimer you probably did see a whole lot of vibration here um that's not because of the actuator the actuator is actually pretty much zero backlash but i don't have any actual nuts in here i'm screwing into just 3d printed plastic so i can't go very hard and because of that this is this is pretty wobbly here so take that with a grain of salt it's not actually this bad it's it's actually really good i'm running a max of 600 milliamps i'm gonna increase the current on my motor driver [Applause] [Music] yeah it definitely turns my next step is going to be to do some empirical torque testing so i got a full bottle of water and almost empty bottle of water and then i've got that cap tied to the end of my arm i'm going to screw this bottle into the cap and then i can slowly add more water until the mechanism gives what i'm looking to measure here is not going to be the stall torque i'm very confident that this mechanism is not back drivable and that the first thing that's going to happen here is the compliant mechanism is going to give it's going to snap in some way so that's the data point that i'm looking for the direction i have it right now so clockwise is going to be the weaker side if i go the other way the failure method is going to be by stretching those springs but in this case it's going to be collapsing those springs which should be a lot easier i've got maybe a quarter of the bottle full right now i'll put it on maybe just about to break there see if i can turn it that was bad yeah that sounds like a broken mechanism i honestly thought i'd be able to do more than that thing so that right there is the limitation of this design it's not surprising but confirmation is always good this actuator was almost able to handle that weight but not quite and that was 392 grams at 26 centimeters so once you do the math that's almost exactly one newton meter it actually came out to be 0.9998 newton meters so i thought that was pretty cool so there you have it that was my solution developed and tested and granted i would have liked to see the actuator handle a little more torque but i'm not going to call it a failure this is by no means the ultimate design for a compliant harmonic drive this is just what i was able to come up with within a couple days i'm sure there are plenty of designs and optimizations that can be made to allow this to withstand significantly higher torques while still maintaining the ability to be squished in all axes and if you'll remember my goal here was just to make something more compact and i certainly did that this actuator could have been a lot larger so i do think that if you're looking for something compact and that is relatively precise this could be a very viable option it's not going to handle a ton of torque but it does shine in other areas and that's why i'm not going to call this a failure that's why i'm pretty happy with it but a lot of improvements can be made namely i'd like a compliant mechanism that doesn't favor one direction like this one did if you are interested in the continuation of this project any further development of this kind of compliant harmonic drive actuator let me know in the comments this was something very different to what i usually work with and it was pretty refreshing but if you enjoyed this project and seeing me make it you can like the video subscribe to the channel you know do all the youtube things and if you'd like to even further support me you can do so via patreon there will be a link for that in the description and there you can get some of the cad files to the stuff i work on including this project so you can make one of these for yourselves or even improve upon the design i also wanted to mention that my discord server which was previously patrons only is now open to everybody there's still a patron only channel but for the most part everything is open to everybody it's place for people to talk about ideas and share their projects i would love for you to join and show me what you're working on i absolutely love seeing what people are doing it's been really fun being able to do that there will be a link for my discord in the description so my conclusion is that i'm pretty happy with this as a proof of concept design it worked pretty well anyways that's all i have for now i have been levi janssen and i'll see you later bye you

Channel: Levi Janssen

Views: 368,349

Keywords:

Id: fGI-4dHYz9U

Length: 13min 38sec (818 seconds)

Published: Mon Mar 29 2021