Banebot 56mm gearbox - double D - V3 - It's Show Time.

I am sorry to keep closing old threads and starting new ones, but I am doing it again.

This is the thread that started the discussion: Banebot 56mm gearbox - double D related

This is the tread that continued things: Banebot 56mm gearbox - double D - RELOADED

A new chapter is about to begin and I thought it was time to start from scratch.

Bottom line first: Banebots has come up with a plan to get harder plates to teams that need them. Expect details in an update soon (today perhaps).

It is not ideal in that we don’t have a genie to make the parts with a blinking her eyes nor do the parts magically install themselves, but given the constraints associated with living in the real world, I think that Banebots has really done a great job pulling this off.

Now for my two cents. In order have any chance of meeting timing constraints, Banebots has had to commit something like $30,000 to solving a problem that has not yet been shown to cause a robot to stop driving.

Think about that. The only total failure of the DD joint that I can document is the destructive torque to failure test that I did on a bench not on a real robot. We have had a handful of reports of Bowtied carrier plates, which I admit is not good, but we have had no reports of transmissions that have stopped working due to that time.

In short, to date no testing by FIRST or testing by teams we have recruited for this purpose have been able to cause a DD joint failure to the point that the robot will not still drive (and drive quite well).

Test results to date:
Data point #1: FIRST
FIRST tested these transmissions (1-CIM and 2-CIM) last summer and found no failures – not even bowtied carrier plates.

Data point #2: Team Testing
1-CIM, 16:1 ratio
We have done 500 hard full forward, full revevse cycles with a 6WD with roughtop conveyor wheels on a 115lbs robot. Result: "noticeable wear, however I think it is nowhere near the point of failure."

Data point #3: Team Testing
2-CIM, 12:1 ratio
We have done 350 hard full forward, full revevse cycles (test is still in progress – they ran out of batteries) with a 6WD with roughtop conveyor wheels on a 115lbs robot. Result: "noticeable backlash"

I expect pictures and a completed 2-CIM test results by the end of the day. The other team I have testing should be able to get us results today or at latest tomorrow. But that is all we have for now.

So, here is what I would like to see. I want every team that is using the 56mm gearbox and can drive their chassis to load it up to competition weight and do some evaluation for us.

If possible, help us with as much data as you can give us

  1. approximate weight of the robot
  2. the type of wheels you have (size and surface type)
  3. the ratio from 56mm shaft to the wheels,
  4. the number of CIMs (1 or 2),
  5. the ratio of the Banebots Transmission (12:1 or 16:1)
  6. tell us if you are using support on the output shaft or if not, how far from the bearing plate is the sprocket
  7. if possilbe the angular backlash at your wheels before and after testing/driving.
  8. If the backlash increases signficantly from start to finish, give us a picture of the carrier plate
  9. if possible, give us pictures of your robot chassis so we can judge your traction etc.
  10. tell us what kind of driving you were doing (climbing stairs, crashing down brick walls, starting and stopping, pushing around old robots, etc).
    I know that Banebots is committed to a addressing this issue regardless of whether it is a real or theoretical problem. Even so, I would like to know that the money they are spending is going for more than just public relations.

So, It’s Show Time!!! Show me some robots that actually stops driving because of this type of failure. Heck I would even settle for seeing a bunch of bowtied carrier plates. It is Show Time… …so show me.

Joe J.

Update: Had to split pictures into two files because they were so large. JJ

Carrier Pictures from Alan Fallone A.zip (2.37 MB)
Carrier Pictures from Alan Fallone B.zip (962 KB)


Carrier Pictures from Alan Fallone A.zip (2.37 MB)
Carrier Pictures from Alan Fallone B.zip (962 KB)

I would like to express my thanks to Banebots for stepping up
and producing hardened carrier plates just in case this becomes
a problem during FIRST competitions. I would like to encourage
everyone who meets a Banebots representative to personally thank
them for what they have done, and continue to do, for FIRST.

Eugene

Dr’s Brooks and Johnson,
Your input during this process has been invaluable!
Joe, thank you for staying on top of this and seeing it through to completion.

I would like to encourage
everyone who meets a Banebots representative to personally thank
them for what they have done, and continue to do, for FIRST.

Dr. Brooks, your suggestion is one that should be heralded from the rooftops. Banebots has definitely earned my respect and future business.

TWO BIG THUMBS UP TO BANEBOTS!!!

I, too, would like to express my appreciation to the BaneBots crew and to all the teams who put their drivetrain on the line to do testing. With the help of the largest and most knowledgeable consulting firm I know of (the FIRST community), this problem has been thoroughly analyzed and a solution worked in just a couple weeks…that is truly amazing to me.

In my real job, this kind of situation comes up more often than anyone involved would like, or would like to admit. Sometimes I have been the customer’s engineer trying to get the bottom of a problem with the quality of material that comes in from a supplier. And sometimes I have been the supplier’s engineer, trying to correct such a problem before it causes a costly schedule delay for the customer. Either way the situation is not fun. But the character and skills of the participants always show clearly in the ways that they respond.

By the yardstick of my experience, it is clear that Banebots, FIRST Staff, and many people here on CD have responded professionally and with grace under pressure.

Okay, the test results are in from one of my test teams.

Special thanks are due to Alan Fallone and the rest of Team 1279. They put in a lot of work for us. Thank you thank you.

The Test:
500 Hard Cycles: Full Forward, Allow the robot to get to full speed, Full Reverse, Allow the robot to get to full reverse speed, repeat.

115lbs robot
6WD
Roughtop Conveyor wheel tread

Test 1: 16:1 1-CIM
Test 2: 12:1 2-CIM

Note that Alan reports that with the 2 CIM set up, they only could go 100 cycles before needing to replace the batteries. This is not a low energy jog around the park.

See the attached pictures, but given the severity of the test, I expected much worse.

There needs to be more data gathered, but I am beginning to believe that this problem would not have been a significant factor for almost all teams using the Banebot 56mm transmission.

Stay tuned.

Joe J.

Carrier Pictures from Alan Fallone.zip (143 KB)


Carrier Pictures from Alan Fallone.zip (143 KB)

When trying to open your attachment, WinRAR says:

! C:\Documents and Settings\Tim\My Documents\My Downloads\Carrier Pictures from Alan Fallone.zip: Unexpected end of archive

Same error with WinZip.
Joe, can you re-upload the file or get a new copy loaded?

Please see this link for details on ordering hardened replacement carrier plates.

http://www.banebots.com/p/GP-56012-RCP

Thanks to everyone in the FIRST community for your help and support

BaneBots

I just placed our order.
There is a little funny behavior when attempting to “continue” on the second page. You must enter “none” in the “Order Number” field if you didn’t have any additional orders to request plates for.

Thanks again to everyone for recognizing this problem quickly and dealing with it. While I haven’t counted the number of forward/backward transistions, I have posted earlier reporting one of our DD joints was bowtied after less than 30 minutes of testing. We have since driven around for another 90-120 minutes, some of it quite aggressively. The transmissions have not failed… yet… but the backlash is increasing. Here is some data that I hope will help:

  1. Approximate weight: 75 pounds (plus 15 for battery)… bumpers, body and actuators yet to be added.

  2. Wheels: Andy Mark Mecanum. Testing conducted mostly on carpet (similar to FRC carpet) and some on linoleum.

  3. Ratio from tranny to wheel: 1:1 (wheel mounted directly to shaft)

  4. #Cims: One CIM and one BB tranny per wheel

  5. BB Tranny Ratio: 12:1

  6. Outboard end of shaft IS supported. See http://www.chiefdelphi.com/forums/showthread.php?t=53033&highlight=1346 for pictures and a link to an older video.

  7. I measure the backlash one one transmission to be between 15 and 20 degrees and on the other three to be roughly 22-27 degrees. It may be that the one tranny with the low backlash was the one on which we destroyed the encoder (see below) and it was not “worked” as hard while the programmers had the drive base up on blocks adjusting their PID code.

I do not have “initial backlash” readings, other than to say they were “pretty tight when new”. I am surprised (and pleased) that the trannies have held out as long as they have after my earlier posting reporting a bow tie… but the metal around the double D is definitely “mushing”. Perhaps work hardening is ocurring in the deformed area. If I get a chance to take one apart and snap a picture, I will share it here (wish I had taken a picture when I had one apart last night)… but we are a bit behind in the rest of our build at the moment.

As for the destroyed encoder, one of our trannies had sufficient end play that the banebots encoder disc would rub against the optical sensors. While we were able to solve this problem by putting a thin spacer between the two tranny shaft bearings, the encoder on this wheel had already ceased to function. We have since had to add a spacer to another transmission as it developed end play and are monitoring the other transmissions. This is off-topic, perhaps, for this thread and may be a unique problem to mecanum drives using the banebots encoders, but it needs to be noted somewhere and perhaps dealt with in any revisions of the 56mm gearbox – a gearbox which I quite like, despite the teething pains we’re experiencing with it this year.

I hope this information is helpful in producing a solution. We need our transmissions to be reliable not just for drive testing, but for two regionals as well as a couple years of demonstrations and experimentation afterwards.

Jason

dtengineering

I hope this information is helpful in producing a solution. We need our transmissions to be reliable not just for drive testing, but for two regionals as well as a couple years of demonstrations and experimentation afterwards.

Jason

This is just the kind of data we need. I am really surprised by the load you must be putting on the transmissions. It was your initial posting more than any other that really lit a fire under me. At the time, I thought, how much load can those crumby mecanum wheels be putting on the transmission? But now I see that in fact, it must be a lot because you showed more wear than the 2-CIM test we ran.

bilbo911

Joe, can you re-upload the file or get a new copy loaded?

Done.

Joe J.

I believe that the reason that the mechanum transmissions bowtie more readily than in a standard drive train is the dynamics of their control. Where as in a regular drive, in order to put large forces on the transmissions, you must actually change the direction of the robot. Each shift is noticeable and easily documented as a fatigue cycle.

With the mechanum setup, you are likely not putting the transmissions under the massive load of decelerating/accelerating the mass of the entire robot, but with every sort of direction change, and to a lesser extent, slight adjustments course, you are fatiguing that connection.

So, the actual loading on the transmission may be half or less that of a traditional setup, but the number of cycles experienced could be a magnitude higher or more. So long as that yield is reached, it’ll go, and a greater number of cycles would account for the greater amount of bow tie. Maybe anyway! Side loading could possibly account for something too. An axial misalignment between the shaft and the carrier plate created by an improper mounting setup could potentially increase the bow tie effect.

Regardless, I am relatively surprised at the amount of strength that the connection retains, even increases, as the bowtie widens. Work hardening is the only thing that I can imagine is occurring, though it is still quite astounding. I was quite sure that teams would be reporting near or actual failures by now. Something’s happening in there…

Following up on my earlier post, I have had a chance to pull one of the carrier plates, clean it up and get a photo to post.

This is similar to what I saw in the transmission I opened up last night… and is from one of the transmissions exhibiting about 25 degrees of backlash.

A couple of things to note:

The hardness test indentations: Although my hardness tester is old and uncalibrated, the two tests in the deformed area showed “Rc 40” (uncalibrated… not likely the real value) while the tests in the base metal showed closer to “Rc 30”, again uncalibrated, sorry… these should not be taken as absolute values, but rather the difference between them should be sufficient to back up the concept of work hardening.

The crack formation: unfortunately this does not show up particularly well, however if you follow the radius of the double d you can see a faint line beginning to trace out a circle on three of the corners. I suspect this will be the ultimate failure mode.

The mecanum may subject the drivetrain to extra cycles, and certainly when the robot was up on blocks and the programmers were testing their drive code we would do many, many, many forward and reverse max to max cycles. With the AM mecanums, that is a fair bit of impact… while they might be lightweight for mecanum wheels, they sure pack more rotational momentum than a standard wheel.

In any case, we’re really good at ripping the trannies apart and throwing them back together, and will be packing a 2.5mm hex key with us to Portland.

Jason

P.S. The deformation extends about 80%-90% of the way through the carrier plate, so it is pretty obvious that the shaft does not fully engage the plate… a condition noted elsewhere already that may or may not contribute to the potential failure.





After having a week of off build, something nobody on our team wanted but weather in our area isn’t too predictable and all the roads in the area were closed for 8 days and no school for a week, we obviously were very eager to get some testing on our attachments. After running the attachment for no longer then 15-20 sec (and those first 15 seconds ran great) the thing stopped dead in its tracks. I thought at first it may have been a bad wire or battery, but this wasn’t the case… what we had was possibly the quickest *failure *(and from what FIRST claims in the Updates, one of the first failures) of the banebots 12:1 gearbox. All this under designed conditions. This happened today around 7:00, quickly putting an end to what work we needed to do. Below are pics of how it turned out.

We had taken all the precautions to avoid this from happening including a programmed smoothing function as to avoid a jerk back a forth. we had decided early on not to use them as drive gear boxes due to the quality of material, which seems now to be a profitable decision. Below are two pictures one, a cleaned up version (no lubricant) of how it looked after pulling the gearbox apart, and two the backside of it and the damage done.

We are currently drafting an email for FIRST and BaneBots as to what happened.

1.JPG
0.JPG


1.JPG
0.JPG

Just curious…what kind of attatchment was it that you had on that gearbox? assuming it was an arm or something, how much weight how far out with what kind of gear reduction?

inquring minds want to know :slight_smile:

(also…how deep is the failed area into the carrier plate? looks like there was way too much end play in that trans!)

These plates look exactly like the torque to failure tests that I did on the bench set up.

There may have been some extra end play in the gearbox but it is hard to tell because the carrier plate sort of screws off if you know what I mean.

I am very interested to know what your setup was. Did you hit end stops? Was there other significant dynamic loading?

Please share more info.

Thanks,

Joe J.

Here’s an interesting thought for everyone… ((please bear with me, I don’t know many technical terms for this))

We are using a banebot transmition on our arm. the arm weighs about 7 lbs, 2 of which are used as a counterweight. After spending a day with the programmers, we noticed at least 25/30 degrees of backlash. When we took our transmition apart, BOTH of the plates were bowtie shaped.

Was this the 56mm kit transmission? What, if any, additional reduction?

And when you say both plates, do you mean the final stage plate on two separate gearboxes or both stages of a single gearbox. (The latter case would be very interesting and as of yet undocumented.)

In light of the last few posts, I’d have to agree with everyone who has said that direction change from the load side is what is killing these plates more than the actual amount of torque applied from the motor side. A 7 lb, well-counterweighted arm may sound like a less threatening load than a 120 lb robot, but it may all be in the dynamics.

Some other data to consider: I made 42mm plates out of tool steel, which I was going to harden. Dr. Brooks suggested that they would become too brittle, so for now I am using one as it comes (annealed), which is actually slightly softer than the stock 42mm plates, which have been shown to deform the same way as the 56mm in some conditions (the original post by Karthik). (Note that the 42mm plates and my plates are both harder than the 56mm plates.)

Our setup is a 256:1 gearbox (supposedly the worst case, although I don’t think the actual reduction matters much if the loading is coming from the arm side) with a 72:10 sprocket reduction to a light-medium weight arm (1"x1"x1/16" Al box extrusion, 3.5’ length, globe motor mounted about 1/3 the way out, 1.5 lb end effector, no counterweight yet). We’ve been running the arm for three days, with and without motor braking, and we’ve been using it to deploy a 26-lb ramp and platform. THERE IS NO NOTICEABLE BACKLASH at this point.

Besides being slightly softer, the only difference between the plate we are using and the stock 42mm plate is that the double-D flats are a closer fit to the shaft. (I cut them undersized and filed out until it slipped on…not a press fit though). This leads me to believe that if you can get full engagement of the shaft on the plate (42mm or 56mm), you will be a lot better off. Also, even when deploying the ramp and platform, our motor never draws more than 10-12A…much much less than its stall torque. The extra 72:10 reduction makes a huge difference and the arm is still plenty fast, ~30 deg/sec.

Sorry for the long post. I’ve been following this and the other threads from the beginning and I still can’t wrap my head around the problem, so I occassionally need to just dump my random thoughts out so others can maybe see a pattern that will lead to a way to protect these plates.

No end stops (well there are, but there are some magnetic limit switches set a little piece from both that would slow it down gradually to a stop). Smoothing in the programming for speed, bearing opposite to keep shaft straight. Everything as suggested in other forums. Our set-up had a spool which wound in cables to lift extrusion up along sliders. And as we let the cable out slowly, it would drop it with gravity.