As an offseason robot project, we’re designing, and possibly building, a box-grabbing robot. We want to play around with some heavy lifting, and we were having some issues coming up with a strong pivot/bearing and drive mechanism for the arm and we were wondering what experience people have had with high torque mechanism.
Currently our goal is to make the maximum torque around 200+ ft-lbs.
This is in the ballpark of 2010 climbers. I’ve read through team 33’s presentation of their hanger. I’m trying to find team 254’s PDF as well as some pictures, which had some torque calculations and IIRC, material selection.
I’m looking at chains, belts, and gears.
The maximum tensile strength of 35 chain is 1,760 lb, so with a driven sprocket of PD (200/1760) = 1.36" it is AT the breaking point. It seems reasonable that a near 8" diameter sprocket would be safe (max size we can make easily). Is this right?
I’ve been told that chains can never handle this much torque, but I think that they can. For instance, my car, which maxes out at 150 ft-lbs of torque, at speeds a few hundred times faster than the robot’s arm will go transmits all of its power from the torque converter to the input shaft of the transmission with a chain (or five :yikes: ).
I’ve pretty much decided against belts. They’re about 10 times weaker with working tensions than chain, and they’ll probably slip unless I use a serious belt/massive sprocket, both of which are expensive.
I’m considering gears because I have tooling (7 pitch, 14.5* PA involute cutters) for it. It’s really slow/boring to make these gears, so I plan on avoiding this if possible.
In the simplified case you present, yes it would be safe. Missallignment or other factors will reduce the load.
I’ll say something a lot of people overlook on arms is shafts/bearings. We did 1/2" with FR8 bearings in 2011 and it worked, but in hindsight we wanted something more rigid.
In 2013 and 2014 we used 1" 1/8" wall tube turned to 25mm (barely any material removal as 1" is 25.4mm), and 25x37x7mm bearings. This is a 6805 bearing from the 6800 series and can be purchased for very cheap. We use USA bearings and belts. We used two bearings in each bearing block (as they are so thin), for 4 total. Our two bearing blocks were ~12" apart in 2013, and 18" in 2014.
We turned the shaft to 7/8" diam for long lengths that weren’t interacting with anything.
We were VERY happy with the strength and stiffness of it during both seasons.
Both years we drove the arm with a 120T, 10DP 3/4" wide 6061 gear (waterjetted tooth bolted into a hub). We’d likely go a tad beefier for games with more arm to arm interaction, but this certainly was pretty robust for the last two years.
We also used an E6 2500 count encoder from US Digital both years directly on the shaft, our sensor had zero backlash because of this.
You don’t necessarily have to rotate the joint of the arm to get the arm to rotate. You could use a lead screw/cable to pull/push somewhere on the arm. There a lot less torque involved if you do it this way.
If you’re interested in 254’s 2010 climber, they have pictures on their build blog.
It looks like the teeth on one of the sprockets are a little broken.
I’m having trouble linking to the pictures, so here’s a picture uploading to google docs.
We used a lead screw with a push rod back in '07. We also used gas shocks to offset the weight of the arm. This method was nice because it avoided a large sprocket and chain.
Just because your arm is lifting many pounds, does not mean your motor has to handle the brunt of the load.
You could balance/offset the load via a counterweight.
I’m not necessarily talking about a strictly mass-based counterweight either. Attached is a graphic representation of our 2011 robot, minus some irrelevant complexity**
While we were only lifting the arm itself and an inner tube, this setup could be beefed up with real springs and more powerful motors
The arm was a simple translating parallelogram “4-bar” arm, with surgical tubing connecting two opposing corners. When empty-handed, the arm (without motors) defaulted to straight forward (purple). With a tube, 95% of the weight* of the assembly was on the tubing, allowing just 2 window motors to easily and quickly lift the tube into the scoring position. The motors actually had to “fight” to put the arm in the lowest position, though the “brake” setting on the motor controllers was enough to combat that
If your arm was to lift, for example, 50lbs, I would set the virtual counterweight (tension) to maybe 25lbs. That way, unloaded, it takes 25lbs* to lower the arm, and loaded, takes 25lbs to lift the arm. Halving the load at the expense of doubling the load time is well worth the trade off. You will have a much easier (and lighter) time building a mechanism that is not as beefy.
*in the straight out position
In the real photo there is also a wrist mechanism, which explains much of the difference between the concept drawing and real deal
You’ll still need to worry about bearings and loads from hitting the arm sideways. If you can find any good pictures from 2005, when we had to lift those heavy tetras, you’ll see some pretty cool double-tetra arms. There were also plenty of bad arms too.
A few pieces of advice:
I agree with Adam about getting a good fit. You don’t want a sloppy fit anywhere in the big pivot. If your pivot is huge, you can use SilverThin bearings, which are pretty neat.
How the bearing is retained by the frame is just as important as the bearing itself. You should try to get the entire length of the bearing supported. It puts a lot less load on races. That being said, most of the bearings you’ll look at are much stronger than they need to be, so you might be able to get away with only partial support. If your arm is very long, you’ll have to worry about people pushing on it from the side. There can be an incredible amount of force down at the pivot. You may want to use thrust bearings here. This was a killer in 2005, when teams got tangled in the towers.
A common mistake seems to be making pivots that are well supported for carrying weight, but are very weak when pushed on from the side. They often “parallelogram”, or just twist and deform due to inadequate support.
Check out CAD models from 1114 and 118, who both had great claws.
Over the years have built several arms. We have always used fiberglass pultrusions for the arm material. Never had a failure. One advantage to the pultrusions is that as the load builds they flex along the entire length as compared to AL where they tend to concentrate the stress at the joint. Our driver in 2011 did, while illegal, manage to fork lift a bot. In arms like air plane wings sometimes a little flex can be a good thing especially for absorbing shock loads.
Okay, I have a few more questions.
I think we’re going to be using some gearbox with a single chain reduction.
We are going to drive the arm pivot because we want the arm to be able to rotate 180 degrees.
This transmission can backdrive, so I’m looking for ways to prevent it.
I have three ideas:
Worm gearbox (worm reduction + chain):
I need a single start worm. The goal is (at stall) 180 ft-lbs at the joint, or 43 ft-lbs at the drive sprocket. In in-lbs, that’s 518 in-lbs on the worm gear.
At half speed, I have 7.2 rpm at the arm, 30 rpm at the worm gear, and 900 rpm at the worm.
Both of these meet operating specs for 12 pitch ground/hardened worms running with 30 tooth bronze worm gears.
The disadvantage is that it is expensive, and large shock loads could damage it.
Locking gearbox:
I’d use a dog clutch somewhere. My concern is backlash if it’s early in the reduction, and force to engage/disengage it if it’s near the end.
Also, I could go with something like the robonaut’s claw locking thing where they have preset positions it locks in. I’m not sure if I like limiting myself to only a handful of presets though.
I’ll point you to one of the robots from 2005, namely 330’s. Large sprocket driving a single pivot, dead-axle (sprocket directly attached to the base of the arm). To get to the sprocket from the FP motors (x2), the power first went through 1 FP gearbox/motor to a common shaft, to a small sprocket, to a medium sprocket on an intermediate shaft via chain, to a small sprocket on the same intermediate shaft, to the big sprocket. No backdriving without a bit of force (>8 lb@5’), and it could lift us under power). Also no lock…
There are a number of methods for reducing force on the chain; try running 2 chain runs in parallel, which will cut the amount of force each chain has to take in half. Also note the counterweight/counterforce suggestions. The counterweights will help stop backdriving.
For purposes of locking an arm, a small pneumatic cylinder fired into holes will do the trick. Or a small servo into a slot. Something like that.
Worm gears are great. They’re nonbackdrivable, and lend themselves to both manual and closed loop control. Basically, where you put them they stay, which means that you can have a much simpler and less finicky control loop. I think we just had a P loop on our worm powered claw tilt this year, and had it not been for the precise angle requirements, open loop control would have worked just fine.
12 DP seems like a reasonable pitch based on my experience, but you could probably get away with as high as 20 DP.
Realistically, whenever I’ve seen large shock loads running through a worm gearbox, either the spacers around the worm or the thrust bearings on the worm are what get damaged. When the worm gear teeth are 3/16" of hardened steel, and your washers on your thrust bearing are 3/64" of unevenly supported stainless, it’s not hard to see what will fail first. Of course, you can bump up the bearings to angular contact bearings, and make all your spacers steel, but at that point, you’re likely going to see other things break before the worm gear. We ran a worm gear setup as our claw tilt this year, and shock loads (like stopping the claw suddenly) caused the screw holes holding the sprocket to the claw and the hex on the worm gearbox sprocket to round out before the worm and worm gear were damaged.
If you do go worm, you’re going to want to throw a lot of power at it. I’d guess (without having run any numbers) that you’d need at the very least a CIM to power whatever you’re describing. We ran a 2 CIM, 1 lead worm + 2 spur reductions gearbox this year, and observed efficiencies in the 6-7% range. That gearbox was the result of a lot of institutional knowledge too, I doubt our earlier versions were even that efficient. Worm gears are great, but they do dump a lot of power.
We’ve been doing worm gearboxes for about four years now. Let me know if you’d like to see some of our gearbox CAD. I’d be happy to share designs and experience.
Worm gears are pretty good for this type of application.
We used a two start worm on our climber with decent success. We used a 2 start 16 pitch worm and a 20 tooth worm gear. We must have climbed our practice tower hundreds of times with the setup (2 CIMs direct driving the worm) and we had no issues.
Worm gears aren’t indestructible though. We broke two, one bronze, one steel, once we increased our climber to 3 CIMs and doubled the speed.
I think you can do better on efficiency than the previous poster has mentioned.
Our climber, which was two belts, the worm reduction described above, two #25 chains, and a 5 start lead screw, and we saw over 50% efficiency (measured by the amount of time it took the robot to lift itself).
Making sure the worm/worm gear are in the exact right spot is very important. We added .002" extra to the center to center like we do with spur gears. Also, make sure you put plenty of grease on the gears.
Chain’s do not “handle torque”, they handle tension as your rating called out. Chains are excellent for doing what you are looking to do.
With a big enough sprocket, you could make thousands of foot pounds of torque with 25 pitch chain. It just might be a a very big sprocket. As Archimedes said, with a large enough lever, I could move the world.
As far as pivots go, FRC 33 used large bushing for the 2005 and 2007 robot. The joints were really nice to work with and gave smooth operation, though the arm construction could be difficult.
With high torque though can come high forces, and you do need to be mindful of the loads. If you would like help doing calculations and layouts, let me know and we can figure these out as a an open case study.
We would start with calculating the amount of load and how far/fast we wanted to move it, and thus would dictate a minimum motor power. Based off of that, we can pick a motor, and then figure what sort of methods we could use to get the right torque.
If you’re interested in counterbalancing your arm, this year we figured out a way of using surgical tubing to perfectly balance our collector arms throughout their range of motion, which could possibly be upscaled to deal with heavier arms.
I’m hoping that the student who worked on that system will chime in here (I can only tell you that it worked like a charm and integrated into our design quietly, inside a frame tube where nobody ever noticed it without us pointing it out). The value of counterbalancing the arm was that it allowed us to use a much smaller pneumatic cylinder than we otherwise would have, reducing air usage while giving us very quick actuation in both directions. Similar benefits would have been in store if we’d used a motor to actuate the arms.
The links show a mechanism (and components) for counterbalancing a lift gate that works on the same principles. The first link also has an entertaining video that shows it in action:
