As readers of these fora know, the 56mm Banebots transmission has a failure mode that involves the double D joint between the last carrier stage and the output shaft.
The 42mm gearbox shares this same failure mode.
I have done some testing and I have the following recommendation with regard to the use of these gearboxes:
My analysis is below. I think the numbers are somewhat conservative but I don’t think they are way out of line with what teams will see.
[FONT=Arial][/FONT]
This will be my advice to teams in summary:
42mm w/FP: If you plan to stall the motor, use ratio less than 145:1
42mm w/BB: If you plan to stall the motor, use ratio less than 214:1
[FONT=Arial][/FONT]
Note that even though you cannot BUY a 144:1 gearbox it is possible to build one if you buy a 256:1 and combine it with parts from a 36:1 BUT… BE WARNED doing so is probably marginal if used with a FP motor
[FONT=Arial][/FONT]
Finally, I do not include dynamic effects in these calculations. If you are expecting the mechanism to have significant impact loads, I recommend even lower ratios.
[FONT=Arial][/FONT]
As always, your mileage may vary.
[FONT=Arial][/FONT]
Joe J.
[FONT=Arial][/FONT]
[FONT=Arial][/FONT]
Rockwell C to Tensile Yield:
RC 22 – 115 Ksi
RC 23 – 117 Ksi
RC 24 – 119 Ksi
RC 25 – 123 Ksi
RC 26 – 125 Ksi
[FONT=Arial][/FONT]
42mm
Carrier hardness:
RC 23.6, 23.8, 24.3
Gear Brass:
RA 37.7, 35.4, 35.8 <<suspect due to small surface to test
Shaft:
RC 43.1, 40.6, 43.7
[FONT=Arial][/FONT]
Given the analysis of the D on the 56mm gearbox, due to scaling, the 42mm D should take 54% of 350in-lbs failure for the 56mm gearbox. But the yield of the carrier on the 42mm gearbox is harder (119Ksi rather than 64Ksi) so it should 186% stronger due to better material. The net effect should be that the joint should fail at almost exactly the same value as the 56mm gearbox (350in-lbs).
[FONT=Arial][/FONT]
If that is true, then I predict that the 42mm gearbox with a FP motor (with 12V stall = .42N-m = 3.7in-lbs) will fail if the effective ratio (the ratio including losses due to efficiency) is 95:1 or higher.
[FONT=Arial][/FONT]
In order to get 95:1 it will take 4 stages. I usually use 85% per planetary stage as my efficiency when I design a gearbox output, but in this case I will use 90% per stage in order to be safe (more torque getting through means more stress on the output). Given that, I predict that the FP motor on the 42mm gearbox with a ratio of 145:1 or greater will fail.
[FONT=Arial][/FONT]
Similarly for the BB motor in the kit (12V stall = .28N-m = 2.5in-lbs) will fail if the effective ratio (the ratio including losses due to efficiency) is higher than 140:1.
[FONT=Arial][/FONT]
In order to get 140:1 it will take 4 stages. Again using 90% per, predict that the BB motor on the 42mm gearbox with a ratio of 214:1 or greater will fail.
[FONT=Arial][/FONT]
Note that these are not one time failure predictions but a failure that will fail upon repeated cycling back and forth.
[FONT=Arial][/FONT]
Now to the actual torque to failure test: I used a 256:1 gearbox. The input torque required to fail the gearbox with the output shaft locked was .6N-m (5.3in-lbs). The peak (after failure, the torque grows as the shaft plows through the carrier) was 1.0N-m
[FONT=Arial][/FONT]
This is scarily close to the stall torque of the FP motor and not too far away from the BB motor. Also, this does not include any dynamic loading of the gearbox.
[FONT=Arial][/FONT]
But, based on the 5.3in-lbs failure, I get that the D joint fails at 5.3in-lbs * (4 * .85)^4 = 710in-lbs. Note while this is higher than the 350in-lbs predicted above, that number was not 1 time failure load, but a load that if cycled caused failure, so I am not too worried at this difference.
[FONT=Arial][/FONT]
If we have our arm geared so that we have sufficiently more torque than we need, then we shouldn’t be stalling the motors and shouldn’t be tearing these out, right?
Thanks so much for your efforts to analyze and solve this problem over the past few days, Dr. Joe. My team is in the “fortunate” situation of having a 12:1 single CIM drive system, but I am worried about our arm which is driven by the 256:1 42mm transmission with FP motor. While I don’t anticipate the motor ever being driven to anywhere near stall, I tend to think that impulsive loading from the arm itself is the thing to watch out for.
I can easily imagine a large arm changing directions in a matter of milliseconds, hitting a dead stop, or gaining momentum before a chain engages and suddenly yanking on the drive shaft with much more torque the the FP could supply. And at 256:1, the gearbox would have very little give from backdrive at those time scales. A good test might be to short the motor terminals (worst-case scenario with Victor e-brakes on?) and drop a weight from an arm attached to the shaft. I bet it could do a lot of damage.
If this is the more likely way to get failure, then there are a bunch of ways for teams to alleviate the problem, many of which were mentioned in previous threads (gas springs, limit switches, etc.). Do you think this could be the more dangerous culprit, or am I missing something important?
You are probably right to be concerned about shock loading in your arm possibly causing damage. There are many ways shock loading can occur such as the arm hitting hard stops, the arm hitting another machine or field element, or a rapid change in direction caused by the operator or a control program.
You may be able to use software to alleviate shocks associated with rapid changes in direction by using an accelleration and decelleration curve. This will dampen the responsiveness of your arm, but will also reduce the shock loads that the system experiences.
Another option could be to install a clutch in the system somewhere between the arm and the BB gearbox. Check out page 993 of the Mcmaster catalog for a start at some off the shelf type options. The clutch would simply slip if torques that are too high are applied (like shock loads). Keep in mind however that this could wreak havoc on feedback systems that are attached to the arm.
Another possible solution similar to the clutch would be to link the drive system to the arm through a spring which would be strong enough to transfer “normal” loads but would absorb strong shock loads (I think team 571 did something like this on their 2004 machine to absorb shock loads through the arm).
350 in-lbs is nothing to sneeze at. And remember that it takes 100’s of such cycles to fail the joint.
As others have mentioned, you potentially have some control via software that can help you keep from exceeding this output from the motor side.
Clutches and springs can help you from the arm side.
One way to drive through a spring. Think of a spring in general terms. One Year to save weight, we replaced a section of our arm drive chain with Spectra cable. This cable is very strong (nearly as strong as steel) but it was much much springier. Thing of that.
You can also think of perhaps using #25 chain to drive your arm vs. #35. If my memory serves me well, #25 chain breaks at 800 lbs or so rather than 3500 lbs or more for #35 chain, if you use a half inch sprocket radius for the output of the gearbox, the chain will break before the gearbox gets damaged – not fun but at least you’d be alive again quickly.
Finally, you can back calculate what force on your arm will cause the joint to see 350 in-lbs or more. Is it reasonable that such a force would occur? If so, would it be enough to cause an instant failure or just a slight widenning of the bowtie. The second you can live with if it doesn’t happen too often, the first one is something to worry about.
Maybe a small bit of good news. We’ve been using the 42mm 256:1 w/ FP setup for our arm. After the output shaft, there is an aditional 72:10 chain reduction to the arm itself. The arm is just a light 1x1 Al box extrusion, no more than 2 lbs and 4 feet long. It hasn’t been put through anything rigorous or a ton of cycles yet, but we have used it to right the whole robot a few times. We tore it down today to check out the internals and the carrier looks pretty good to me (picture attached). There is slight deformation at the corners, but it’s only at the surface (maybe 10% of the total depth).
Obviously more demanding and cyclic loads need to be tested before this proves anything, but maybe with TLC the 42mm can be used effectively for light arms.
I tried a possible quick fix: pinning the shaft to the carrier, but after a couple broken bits, I can confirm that the shaft is, to use technical terms, really, really hard. The carrier hardness falls somewhere between a filing drawer hanger and a screwdriver. Note to self: leave the real engineering to Dr. Joe. But 119 ksi is pretty hard, am I right? Can we do a lot better with a different steel? Heat treatment?
What about shims to take out the small bit of play between the shaft and the carrier? This might help since the corners won’t dig in so much. It would approach Dr. Joe’s linearly-distributed loading condition more. I’m thinking of what happens when you use a metric hex wrench in an English hole. It might work, but will be a bit too small and as a result will strip the hole more easily.
A reminder to teams (and I am going to shout, so plug your ears): I AM NOT SAYING THAT THESE GEARBOXES ARE FRAGILE LITTLE WIMPS THAT WILL BREAK IF YOU LOOK AT THEM CROSSWISE!!!
They have limitations. It is possible to use them over their limits and they are likely to have issues if you do.
Note that my recommendation of 350in-lbs max is the load that I predict they will fail at with 100’s of cycles! It will not take a single or even a 100 such events to make the gearbox fail in the mode.
Now the good news/bad news. First the bad news: Banebots tells me they are out of the 256:1 42mm gearboxes. now the good news: with 64:1 being the highest ratio they sell, these gearboxes can not only take full stall loads of the FP or the Banebots motor, they have dynamic loading safety margins of 2.5 for the FP and 3.5:1 for the BB.
Here is the bottom line (and I am going to shout again so you can hear me): IF I WERE BUILDING A ROBOT THIS YEAR I WOULD ALMOST CERTAINLY HAVE FOUR (YES FOUR) OF THESE 42MM GEARBOXES ON MY ROBOT – TWO FOR THE FP MOTOR AND TWO FOR THE BB MOTORS.
Is there any chance the BaneBots will make available for purchase the individual carriers? Two reasons: I would like some to test with that I’m not afraid to break. Also, I think teams would like to have spares if their carriers do fail due to some unforseen stall condition in a match.
I am going to take my 42mm carrier to a materials lab today and see if I can at least confirm the hardness values. Perhaps they’ll also let me try some heat treatment. (Anneal -> Quench -> Temper would be best?)
Joe I see where you are going with this, the lower gear ratio gearbox’s are going to be much more robust because the fewer number of stages. But there is a good reason the 256:1 gearbox’s have been selling like hotcakes, those are the only ones which are highly applicable to this year’s game. Teams are looking to use the FP motors for high torque, low speed situations this year. Not like last year where some teams used them to power shooter wheels or run belting systems, in a higher speed less torque intensive situation. Making the 64:1 gearbox far less useful to teams than the 256:1 gearbox this year. Many teams looked at the 256:1 gearbox and thought “Wow that’s a perfect ratio to use a FP motor to power an arm system”. My team received our 256:1 gearbox’s yesterday, and promptly dissected and inspected them, comparing the D design the design from last year. Lets say we are disappointed and concerned about the life time this system will work. We have designed our system to share the load across two of the gearbox’s and sprocketed them down to a very reasonable level where we will will be using only 1/5 of the FP’s stall torque. It now worries me that these gearbox’s are no longer available, if we choose to use the ones we have will we be able to purchase replacements in the future?? Why are they no longer available?? simply out of stock, or quality problems?
Using 1/5 of the stall torque of the FP would mean that you are pretty safe with respect to torque from the motor side. You have a pretty good shot at being able to limit the torque input by the motor.
If you can figure a way to limit the backdrive input from the arm to the gearbox, you’ll be fine.
Now to your other points.
If you damage the gearbox, you can guy another ratio gearbox and rob parts to rebuild your 256:1 gearbox.
As to why they are no longer available, I didn’t ask specifically, but I was under the impression that whatever stock they had in house has been sold. I don’t know what or when or whether they will order more. My impression is that Banebots was not even planning on marketing these to FIRST folks. For our own reasons, we find them useful so we buy them but that was not their intent when they decided to add them to their catalog.
Here are the results of some testing done at a materials lab today:
Before Heat Treatment:
HRC 21.4, 23.7, 23.8 - right on target with Dr. Joe’s tests
Annealed at 900C for 15 minutes, then quenched in water:
HRC 59.9, 59.8
According to the lab technical instructor, who knows more about materials than I could ever imagine, this is tool steel. It comes annealed, which makes it easy to machine, but can be hardened quite a bit with heat treatment. (My chart says that as-quenched, the tensile stength has gone from ~119 ksi to almost 300 ksi.)
I have the carrier tempering now at 300C for a half hour. This will bring the hardness down a bit, but make it more resistant to fracture. Very good news for the 42mm gearboxes. I’m not sure how it applies to the 56mm, but hopefully those can be hardened too.
The planet carrier plate in the 56mm gearbox appears to be made of something other than tool steel, as it starts out about half as hard as the plate in the 42mm gearbox. It can probably be treated to result in a similar doubling of hardness.
I’m curious to know what you’re doing with the planet gear pins, as I expect their strength characteristics, as well as the tightness of their press fit, will be affected by the hardening/annealing process.
Update: After tempering at 300C for 30 minutes, the new hardness numbers are:
HRC 49.5, 49.5 (~230 ksi tensile yield)
So it’s still about twice the original hardness. I’m putting it back in the gearbox tonight and we’ll test it a bunch. I’d still like to get some spares since this one has now been drilled/poked/prodded/baked/cooled/etc. If anyone from BaneBots is reading - selling these individually would be very helpful for the time being until a more permanent solution can be found.
Regarding the pins, all I can tell you is that they didn’t fall out. :rolleyes: I’ll check them again after testing.
In summary:
annealed at 900C for 15 minutes
quenched in water
tempered at 300C for 30 minutes
quenched in water
roughly doubles the hardness for the 42mm carrier plate. How this applies to the 56mm plate is still uncertain because it is apparently made of a different steel.
Bad news: our tempered (HRC 49.5) 42mm carrier failed in the gearbox by fast fracture. More specifically, three of the pins fractured just below the plane of the plate. This occurred at almost no load.
To confirm that it was not just the pins (and to let out frustration) we subjected the carrier itself to the “hit it hard with a hammer” test and sure enough it fractured as well (picture attached).
Conclusion: It was too brittle, by far, at this tempering. I can try it again with a higher temperature temper and longer time in the oven to get it a little softer and less brittle, but I will need to get my hands on more and I don’t think I want to buy whole gearboxes. I’m now less optimistic about the tempering solution since it seemed to fracture so easily, but there may still be hope.
You are beautiful. I am sorry about your failure but you give such great details that I love to read about it…
…I think you should dial down your targets a bit.
#1. Are you sure you need the extra strength? If you can make due with 350in-lbs I would highly recommend that you just leave them alone. Note that the 350in-lbs that I am predicting is not a single load failure it will take 100’s of these cycles before the gearbox will fail to function.
#2 If you need that extra bit of safety margin, I recommend that the hardness of the shaft is a good point to start. Unless you were going to harden both, you will just push your failure to the shaft. So, I would say, push out the pins, harden and then temper back to RC 40-41. I think you will find that value a good compromise between brittle and ductile failure.
I will try to get my broken sample to HRC 40-41 tomorrow with a different tempering. I won’t be able to test it in the gearbox, obviously, but I’ll break it (I’m good at that) and see what it looks like.
Our team is using two of these 256:1 gearboxes for our arm as well, and I’ve been thinking about heat treating the carrier plate and/or the output shaft. I talked to my materials professor today, and he said it could be done if I knew exactly what kind of tool steel the output shaft and the carrier plate were made of…does anyone know the material specifications? Without these specifications, I imagine we may experience the same sort of fast fracture that others are talking about if I can’t figure out the exact type of tool steel.
I can’t tell you the exact specs (I don’t think anyone knows at this point), but I’ve been doing testing with the 42mm plate for the last two days and I have the following data:
Hardness Before Heat-Treatment: ~C24
900C, 15 min, water quench: ~C60
300C temper, 30 min: ~C50
This resulted in brittle fracture almost immediately in the gearbox (very low load).
The lab technician said his best guess is either O1 (oil cooled) or W1 (water cooled) tool steel. Definitely not A-anything (air cooled).
Many have said that oil quenching with a longer/hotter temper to get to ~C40 might work well. I will test this when I get more samples. I’m also getting some A2 tool steel to try.
Note to self: leave the real engineering to Dr. Joe. But 119 ksi is pretty hard, am I right? Can we do a lot better with a different steel? Heat treatment?
