Take the "Bane" out of your Banebot P60s - Solution to Banebot P60 Weaknesses

[fox46]

We are using the 326:1 P60s, to articulate our arm, being driven by an RS395 and the other driven by a FisherPrice motor (For all you critics out there, this is not for power reasons but because we are limited to the number of BaneBot motors we can use - an RS395 will replace the FP and run the arm very happily). We have not experienced any failures yet but on both we have begun seeing signs of the gearboxes failing and although we have spares, we elected to try and "bulletproof" them rather than risk a breakdown. I am pretty confident we have mitigated the major modes of failure that we have been seeing from teams using these gearboxes.

The four "banes" of the banebot P60 include:
1) Rounding out of the output shaft and/or drive socket on the output carrier.
2) Bearing failure due to defective bearings (ungreased) and overloading the output shaft.
3) The roll pins acting against the flat sections of the ring gear that hold the rign gear torsionally captive galling, snapping and rounding out.
4) Binding of the gear train due to axial bending of the gearbox.

In order to rectify these we have:
1) Machined a new one-piece output shaft and final carrier from Stressproof steel to eliminate the interface between the output shaft and carrier.
2) Machined a new endplate with larger 1/4-20 mounting holes which houses larger (to accommodate the 1/2" output shaft) high quality NTN bearings spaced further apart to reduce loading caused by sideloads on the output shaft.
3) Bored sockets in the new endplate and stock ring gear to internally lock the ring gear to the end plate with 6 roll pins.
4) Replaced the non-graded 6-32 bolts which hold the assembly together with high strength socket head cap screws.

So far the end result is very promising. Obviously we have not been able to test it on our robot but we are hopeful this will eliminate the weak spots previously experienced with these gearboxes. All of these parts were made with a bench-top milling machine and 9x11 lathe and are well within the capabilities of even the most basic machine shops. The shafts for the planet gears on the final carrier are made from 1/8" hardened steel dowel pins. The output shaft is made of Stressproof Steel(http://www.niagaralasalle.com/product-stressproof.html) which is stronger than chromoly grades but machines like a dream (one of my favorite metals!). The bearings are 1/2" ID 1-1/8" OD, 1/4" W double sealed NTN bearings. The end plate/bearing block is just regular 6061-T6. The pins are 3/32" heavy wall roll pins.

[thefro526]

Wow, I am so thankful for this post.

Our Arm Gearboxes failed during the semi-finals of NJ for what seems like the 10th time. This seems like it'll make them bullet proof.

[AdamHeard]

Originally Posted by thefro526

Wow, I am so thankful for this post.

Our Arm Gearboxes failed during the semi-finals of NJ for what seems like the 10th time. This seems like it'll make them bullet proof.

Using such high reductions p60s is just all around a bad idea, even with these improvements. I imagine this now shifts the failure point to the gears in the final stage when the motor stalls.

[Ether]

Originally Posted by fox46

.

Do you have drawings that you would be willing to post?

[Kevin Sevcik]

Adam,

You're probably right that this will shift the failure point of the transmission elsewhere. If it's the ring gear, there's a steel version of that. I think you're wrong that the issue is stalling the motor, though. If you ever actually manage to stall one of these things, you're going the seriously damage the gearbox very quickly. We're experiencing a failure mode very similar to fox46's in that the gearbox gradually degrades over time. I think this is a combination of the weaknesses he's helpfully pointed out and some unique load conditions. We've never stalled our 256:1 RS-775 transmission, and yet we've deformed the carrier plate twice. I'm certain this is due simply to inertial loading from our arm as we lower the arm and stop it. The stop is invariably sharp and sudden thanks to the huge gear ratio, and this creates an instantaneous load on the gearbox that exceeds the capacity of the carrier plate. I expect fox46 is actually experiencing a similar problem.

His proposed solutions will definitely strengthen the gearbox and slow the degradation process, but I think they should be coupled with programming changes designed to limit the rate of change of voltage so as to lessen the sudden shocks when motion is stopped.

[fox46]

All we had were some hand drawings when we built these but all the information is there to reverse engineer it. I'll see if I can find and scan them tomorrow- I think they are still floating around the shop.

No modification to any machine will eliminate the chance of failure. Shifting the failure mode to a stronger part of the mechanism was exactly what we wanted to accomplish. Planetary gear trains are incredibly robust. I suspect the method of failure is now the interferance fit of the sungear that drives the final reduction pressed into the 3rd reduction's carrier. In any case, it is much much stronger and can now be used as it was initially marketed to us.

I have no doubt that this thread will quickly accumulate posts accusing teams of having poor designs or not using the boxes correctly but the undeniable fact of the matter is this: The Banebot P60 has a factor of safety of less than one, even for the pint-size RS395. The factor of safety decreases the larger the motor you put on it. No machine such as these gearboxes should ever be designed with a FOS of less than 1. As Kevin just mentioned, very few people are actually exceeding the ultimate strength of the unit, they are only comming very close to it so the failure mode presented seems to be more of a fatigue/degradation issue. Most people are just "marginally" exceeding the P60 capacities so even if these mods only added 10% more strength it is most likely enough to solve the problem

FYI: Banebots now have a P60 gearbox offering 627:1
Edit: correction- it's 672:1! (thank you dislexia) I guess a suitable application for this would be powering a large clock?

[Ether]

Originally Posted by Kevin Sevcik

I think they should be coupled with programming changes designed to limit the rate of change of voltage so as to lessen the sudden shocks when motion is stopped.

Kevin, based on earlier discussions we have had regarding the switching method used by Victors, is this necessary with a Victor with jumper set to coast mode?

[fox46]

(I know you asked Kevin but-) The coast mode would definitely help with reducing the load when stopping the motor but limiting the starting power while still keeping the speed up is our challenge. Also, we're using encoders to accurately index the arm and if the Jaguars werent in brake mode I suspect it would drive the software nuts. It would be even worse if it overshot the target position and the software then reversed the motors to bring it back into position. I've been cracking the whip trying to get our programming team to come up with a ramp function to ramp the power up and down relative to proximity to the target position of the arm but dealing with our programmers is like hearding cats and lately I can't convince them to do anything. They've been using the typical post-ship "theres no point to trying because we have no way of testing it" excuse.

[Kevin Sevcik]

Brake/Coast only ever enters the picture when you're sending a 0% command to a Jaguar/Victor. Then it will short or open the motor leads. If you're sending any command outside of the deadband, then one of the lower legs is always on and you're effectively braking during the low period of the PWM cycle.

Besides which, have you tried backdriving one of these things even in coast mode? The friction losses and cogging torque of the motor make a pretty effective brake all on their own at these ratios.

On the other hand, the huge ratio means you're mostly just pushing against the friction of the gearbox itself. So the motor operates a lot closer to no-load speeds than you might think. So "slowly" ramping down the voltage will translate pretty directly into slowly ramping down the speed of the load.

[dtengineering]

Those look like some great upgrades to a BB P60. You may well be able to get more torque out of the gearbox after the modifications than before, however it is important to remember that the purpose of the very high gear reductions is to reduce speed, not to increase torque. Note the following comment prominently displayed on BaneBot's product page for their high-reduction P60s...

"Maximum torque recommendation
We recommend maximum torque not exceed 35 ft-lb for all P60 Series Gearboxes. It is possible to mount motors that will exceed this in higher gear reductions. Higher reduction gearboxes should be utilized primarily for speed reduction. Designs utilizing a P60 gearbox / motor combination that will exceed 35 ft-lb should include a method of limiting torque to prevent damage to the gearbox."

As for the "safety factor" being at or near one, I always felt that as a robot designer that safety factor calculations were my job. If the manufacturer says "don't exceed 35 ft-lb", then it was my job to ensure that I didn't. If I needed a safety factor to make sure that I didn't, then that was my job to engineer it in. Sometimes, when pushing a machine to the limit, with no risk of injury to people or property, it made sense to design a system with a very low "safety factor".... but no matter what, I always knew that when you run equipment at the extremes of its design limits that it will sometimes break.

Looks like rather than dialling back the demands, you're cranking up the specs. It looks great and I hope it works well!

Jason

[sanddrag]

Well done. Should have done this to ours.

[Kevin Sevcik]

Jason,

I don't think I specifically declared it above, but our arm design is fairly well inside BB's 35 ft-lb spec at about 8 ft-lbs, and we're still slowly failing our gearbox due to the shock loading I mentioned above. Yes, as an engineer I should've realized this was a possibility and demanded more than a 4:1 margin in the system. Still, BB's warning there doesn't really go nearly far enough. It's not that you have to limit the torque IF your motor-gearbox combo can exceed 35 ft-lbs. You have to limit torque if the load can exert more than 35 ft-lbs on the shaft. I think you could actually fail the gearboxes without a motor attached to them at all, in fact. So your only real option for using them safely as they come is to put a torque limiter on the output shaft and hope for the best.

[Ether]

Originally Posted by Kevin Sevcik

If you're sending any command outside of the deadband, then one of the lower legs is always on and you're effectively braking during the low period of the PWM cycle.

I was thinking back to this post:

If you're running the motor at high/full speed and drop to a lower PWM command, the current will decay quicker in the tan jag, but you'll never get reverse torque, so you'll "coast" during the off cycle.

... and wondering if the Victors would also behave as described above.

[dtengineering]

Originally Posted by Kevin Sevcik

Jason,

I don't think I specifically declared it above, but our arm design is fairly well inside BB's 35 ft-lb spec at about 8 ft-lbs, and we're still slowly failing our gearbox due to the shock loading I mentioned above.....

Ahh, I'm not surprised... either that you did a good design job, or that the little BB is getting beat around, depite that. I hadn't really appreciated the magnitude of the shock loading on the gear shafts until we built a mecanum drive machine and were trying to tune the PID loops for wheel speed with the robot sitting on blocks and the wheels spinning freely. As we slammed the motors from full forward to full reverse we hammered and hammered away at the gearbox. You're probably right that a warning about repeated impact loads should accompany the overall torque load.

It actually sounds like the fact that your system worked through a regional but is showing signs of wear is a sign that you designed it perfectly... after all, if it isn't overloaded, it's overbuilt!

Jason

[fox46]

Our arm is biased with gas shocks to the point that you can release it from full height with the drive chains removed and it will "float" back down to rest position and we still were seeing signs of failure.