Robotic kinematics
-
@JoergS5
Hi! how do i chat with you? I would like to understand how it is made ľ harmonic drive you bought! and complete with crossed roller bearing? as I said, replicating it in steel is madness! but not impossible! the tricky part will be hardening the flex spline and producing more than one harmonic drive in a reasonable time!
and like this?
-
@tony73 this is exactly the model I meant. It's the only one which I have opened, the other ones are closed, so it's not easily possible to show the teeth. But the differential gear are not meant for the robot. Maybe it's usable for robot, I'm not sure. The flexspline is easier to produce than the normal one, and there is no tension between the teethed part and the vertical part. I'll have to check whether I have the wave generator for this type. There is also a 2 part case around the parts of the picture.
The wave generator is normal ball bearing (unbalanced of course). Cross roller bearing was a separate idea, not harmonic drive, with something called ServoBelt Rotary, patented by Everman and a partner of him. ServoBelt Rotary would be a nice solution for Axis 1, harmonc drives more for axes 2 to 5. Axis 1 has the hole robots load vertical to the bearing, I expect the cross roller bearing better than the harmonic drive in respect to load. (But the ServoBelt has no gear inside, it may be better to use harmonic drive for all axes).
If you want to DIY the wave generator, please be aware that when buying balls, they have tolerance classes: G2/G3 best, G10 very good, G100 good quality, taking G10. Maybe it's easiest to produce a wave generator by bending a normal ball bearing a bit, with loadlifter.
The tolerance classes are inside one sample that you buy. E.g. buying 100 balls, they have tolerance G10. But buying another 100, they have only G10 to each other, not to the first sample. In other words: buy together what you need, and some more (I always loose balls when disassembling lol).
Chatting is done at the top menu, the bubble symbol. I prefer normal blog discussion, then everybody can see the information. But in this case it's an exception. To start a chat, open the profile of the other user, then the three dots ..., top entry is start a chat. I already sent you an information by chat.
-
@tony73 I found the wave generator now.
The diameter is between 60.6 and 62.0 mm, so only 1.4 mm difference. The balls are about size 6 mm (I cannot measure exactly).
The generator is built by two ball bearings, which are only connected by the axis.The flexspline teeth are total high 1.4 mm (including inner ring) for teeth high less than 0.7 mm, teeth distance 0.7 mm, so this fits good to the generator.
-
@tony73 https://www.bengtssons-maskin.se/uploads/extrafiles_file_2.pdf page 23 top is also a good idea, combining worm drive with a big wheel.
But if possible I would prefer harmonic drive because of the zero backlash.
-
@tony73 I found a scientific article about flexspline:
http://fluid.ippt.gov.pl/bulletin/(58-4)683.pdfmaybe of interest for you, what could be helpful for a longer life of the parts.
-
@JoergS5
good morning! how is the work on the robot going? -
@tony73 hello, I am currently working on gears of axis 1 and 2. I will use Nema 17 with axis 2 now, so I can use every Duet hardware with the robot (2A max for every stepper). I made improvements to belt tensioning and the weight distributions of axis 3 and 4. I will start today with axis 5 + extruder + hotend, maybe I can make the first print tomorrow.
In the meantime I made some investigations about harmonic drive. Maybe it's possible to implement without metal, there were some promising attempts, like https://www.youtube.com/watch?reload=9&v=t4hW6lZYaQk and connected usage with servo https://www.youtube.com/watch?v=tVlKTpKXCsY
It looks nice, but jerky at some movements. I think precision is key.
The attempts on youtube and thingiverse have very different quality. -
I think making cycloidal reducers yourself is preferable to harmonic drives. The flexspline is hard to make without metal and more often than not you end up with a reducer with very low durability and torsional rigidity. Cycloidals on the other hand use just rigid components and are a lot easier to make with printing. And offer a lot more torsional strength.
-
@roiki11Did you make some yourself, this would be interesting. What I know is that the gear ratio is limited in the range of 1:8 in the most examples I saw. So they are often combined with an added 1:3 belt gear between stepper and cycloid gear.
-
@JoergS5
Yes I've made some. Somewhat big with 72.5 reduction.The reduction ratio is not limited at all since you can use multiple stages. You can also get 30:1 reduction quite easily on a single stage too. Though if small size is a priority then that can limit the cycloidal ratio.
-
@roiki11 are they open source and you can share it please if possible. If not, maybe you can provide a hint where to start or whether you started development from an example. And did you make them 3D printed or with metal?
There are some thingiverse and youtube builds, but one never knows which ones are good quality and which not, unfortunately.
The robot hinges are better with bigger diameter imho, because they give added sideways stability. I think 10 cm diameter is ok for a printer for print area 40 or 50 cm each dimenstion.
-
@JoergS5 said in Robotic kinematics:
@roiki11 are they open source and you can share it. If not, maybe you can provide a hint where to start.
There are some thingiverse and youtube builds, but one never knows which ones are good quality and which not, unfortunately.The robot hinges are better with bigger diameter imho, because they give added sideways stability. I think 10 cm diameter is ok for a printer for print area 40 or 50 cm each dimenstion.
The one I'm working now is about 140mm in diameter. It's not completely finished yet since I have some parts to print and final fit. It's completely working though. I can link the fusion file sometime.
You can also check opencyre. it's a pretty good design. There's also a
Script for fusion for generating the cycloidal gears. I used that which made the design work really easy. https://github.com/mawildoer/cycloidal_generatorJust remember that the reduction ratio is "<number of pins> - 1" or the number of lobes on the rotor.
-
@roiki11 thank you for the links, I will give cycloidal a try. I agree with you that the cycloidal drive is easier to build.
Do you plan to build a robot with it also?
-
@JoergS5
good morning! were you able to try the print? -
@tony73 no, I could not finish 5th axis. It won't take much longer, but I don't want to hurry. I changed my design to a 20/60 * 20/60 * 20/60 belt gear, resulting in 1:27. The result is a 150 g lighter axis 5.
The tension of the long belts is a real problem, resulting in flexing and imprecision. When finished prototype 1, I'll start immediately with the construction of a cycloidal gear, first for axis 2. The calculations of weights and torque are valuable, because implementing cycloidal gear with low torque when rotating will be good for precision probably.
-
@JoergS5 said in Robotic kinematics:
@roiki11 thank you for the links, I will give cycloidal a try. I agree with you that the cycloidal drive is easier to build.
Do you plan to build a robot with it also?
Yes I'm building a 6 axis robot arm. Mostly for camera work and shits and giggles.
It's a bit on hold now though with covid and all.
-
@roiki11 you're invited to comment or distribute ideas on the robot constructions if you like.
-
@tony73 do you want to try cycloidal also, or proceed with harmonic drive? Both is possible and valuable (cycloidal to try to optimize, harmonic to compare both). I'll try cycloidal next, although I have the original harmonic drives as patterns. But creation of the flexspline seems difficult to me.
-
@JoergS5
Hi! can you tell me a commercial arm robot that uses cycloidal gearboxes? can they have zero play like harmonic drives? I see more complicated to build cycloidal gearboxes in (steel) and there is always the problem of the junction point between the gearbox and the two arms of the robot, the most logical solution is a crossed roller bearing and I think also the lightest one, another is to use two tapered roller bearings, the cost is very much lower, but the support must be built to contain them with a lathe or a cnc, but the weight is certainly higher! Do your harmonic drives come complete with crossed roller bearings? if yes, you have a huge advantage, that type of bearing, for what I understand as it is! allows you to merge two robot links! I believe that the robot you are building, with belts and pulleys (type R12), is an excellent solution to be able to give many people the possibility of making a robot arm! -
@tony73 I will certainly finish prototype 1 (or should I call it Model 1) with belt gears, because someone can build it without special tools. Only some boring and a saw (even a hacksaw is sufficient).
My commercial harmonic drives have ball bearings for the wave generator, not cross roller bearings.
Cycloidal is said to be zero backlash also, but I have no experience which one is easier to build. It is only that flexspline will be complicated. But with a high gear ratio (many teeth), cycloidal needs to be precisely created also, same like harmonic drive. So I am not sure.
If I am unsure of something, I generally end up buying/making both....
I think the commercial robots use harmonic drive mainly. The noncommercial ones often cycloidal (like Sixi 2).
I wanted to avoid using the harmonic drives which I bought used, because the robot then cannot be copied. 5 Axes with 200 $ minimum means 1000 $ for gears alone. Maybe I use it for Axis 5, I have a combination of Faulhaber stepper with Harmonic drive, which is beautifully small.
What I can certainly make, is compare the commercial harmonic drive with a DIY harmonic drive, cycloidal or belt based gear.
PS I'm sorry that the prototype takes so long (our first discussion was in September, so 3 months now). I underestimated the effort and I had to learn a lot. I feel more like a marathon runner than a sprinter!
