DEWBOT VI Drive Train Tread Wear
One of the issues we had at Drexel is that one wheel had extreme tread wear. This page is a discussion on what the problem is, what possible reasons are and how to test for the reason before we do a repair.
- 1 Problem
- 2 Thesis: Brake Jumper on Jaguar
- 3 Thesis: Pivot Plate Dimension
- 4 Thesis: Loose Axle Bearing
- 5 Thesis: Position Sensor Mounted Differently
- 6 Thesis: Different Drive Code
- 7 Thesis: Bent Axle
- 8 Follow-On Issue: Losing Traction
- 9 Information
- 10 Clem McKown email 11-April
- 11 Gary Deaver 11-April
- 12 Gary Deaver April 15
- 13 Gary Deaver April 21
The wheel on pivot #4 (front right) exhibits extreme tread wear when compared to the other wheels.
All wheels all were retreaded on 22-Feb, just prior to shipping. At that time, the tread wear on Pivot 4 was noticably greater than for the other wheels. Wear was noticed again sfter the 7th match at Drexel and this tread replaced after the last qualifying match. At that time, the wheel had 19 qualifying matches + 7x2 practice matches + practice field time worth of wear (about 90 minutes of drive time). The other three treads were not replaced at Drexel.
Wear primarily effects the tread edges.
Other Observed Inconsistencies
- Wheel #4 does not calibrate correctly zz% of the time in the aa version of the calibration code. It is the only wheel with this problem.
- It can however, be aligned manually in the pit pre-match. When this is done and the calibration code is not run in autonomous, there the wheel doesn't experience a calibration problem (true?)
- Simultaneously with noticing the tread wear, the pit crew discovered the wheel's locking collar was completely unsecured. It appeared to have loosened rather than sheared (i.e. it could be repositioned and tighten again).
- This wheel appears to have a binding spot in the drive mechanism as well (is this confirmed?)
- Gary> When pivot #4 was serviced before ship. there was not noticeable wear on the tread worse than the others. The chain was much looser than the others from the time it was installed. This is most likely from improperly drilled steering motor mounting holes and play in the 1' pivot tube and bearings. This can be addressed by removing a link and adding a 1/2 link and/or oblonging the mounting holes. The chain tensioner may be adding friction that can cause problems with PID. (see PID comments later)
- Gary>One side effect from the method of 360 wheel drive is that there is a torque opposite the rotation of the wheel drive gear rotation. This load is felt by the steering motor and the motor must over come this force to move against it. Turning the other way aides the wheel steering. If from wear pivot #4 can not turn the wheel as freely as the others the steering motor will have to over come this higher torque. Is the pivot drive train damaged? ( PID constants affect the motors ability to deal with this added load)
Even wear across an entire Plaction Wheel tread that is inconsistent with the other wheels can be caused by
- that wheel spinning (consistently) at a different rpm than the others
- that wheel being driven while the others are stationary and vice versa (causing it to slip)
- a tread defect (this seems unlikely given the other issues, but is easily checked by continued use of the new tread)
Thesis: Brake Jumper on Jaguar
The brake jumper is installed on the drive cRio. This means every time the robot changes direction the wheel will stop, dragging the tread. The other wheels would continue to spin along the carpet.
Tests and Results
Look at the drive Jaguar for that pivot.
Thesis: Pivot Plate Dimension
The holes that the steering chain run between on the #4 pivot plate are slightly off, making the chain loose. Though it was subsequently tensioned, the drag of the tensioner or the looseness of the chain delays the wheel's steering. This makes it point in an incorrect direction (at least initially), causing it to slip.
Tests and Results
- Attempt to tighten chain (sans tensioner). Switch pivot plate? (seems drastic)
- The first step to this problem is to dismantle pivot #4 and assess the mechanical condition of the pivot, steering and sensor. If it is not sound mechanically, no further analysis of the programming issues can be done. We have spares of all parts.h
Thesis: Loose Axle Bearing
The axle bearing collar was loose, this would allow the wheel to slide back and forth across the axle. This would lead to wear because of ???????
Tests and Results
Thesis: Position Sensor Mounted Differently
The magnet for the Cherry Position Sensor is mounted differently than the other three. This would cause wear because ?????
- Gary> The Cherry sensor mounting plate on #4 was slightly more off center than the other. This was supposedly corrected during mounting. Maybe it is not. I do not believe this would give the error that we see. Off center acceptable alignment is up to 2 MM. I'm almost certain it is within that amount. Offset error Would show up as a +- sinusoidal offset error around 4 points on the linear output line and be very repeatable. The magnet on pivot #4 was too close to the Cherry sensor on pivot # 4. This was corrected at install time by adding 2 washers instead of 1 washer on top off the 1" standoffs. This should give a 2 - 2.5 MM magnet to sensor distance. Is it?
Tests and Results
Look at the sensor and see how many washers are on it. Should be 2 - 2.5 MM magnet to sensor distance.
Thesis: Different Drive Code
There is different drive code for that pivot. This would cause wear because ?????
Gary> There are several things about the program that concern me. The Cherry sensors should have very good repeatability. In other words the pivot drive motors and sensors once assembled should always have the same calibration coefficient's. Since day one after calibration this value for each pivot should stay the same. Only if we take apart the chain on the steering motors would this change. If it varies some thing is very wrong.
The potentiometers should only be used for moving the steering motors to obtain the calibration. Why do we have to keep calibrating? What bothers me is that the programmers talk about the calibration routine at auto start up. Are they using the pots at all then? They shouldn't.
The calibration values should be able to be hard coded into the least distance routine as a constant. The PID drive VI should be able to be given a position to drive to at the start of Auto (straight forward) and the steering motors should reach that position in less than 1 second. Why does it take so long to get the wheels pointing forward?
Could part of the lag be attributed to the time to read the offset values from flash? Are the values from flash assigned to a variable by value? Could Labview be using a pointer to a flash memory location? This would be very bad. Flash reads are much slower than ram. A whole sector has to be read to get to one byte. Question? If the offset values are written to flash and stored after the compiled code, could they be over written if the code grew considerably and they are not re-written?
- The original (Foster) idea was to read the positions and store them into Flash. The team opted to create a board with 4-10K pots on it. The pots are adjusted to align the wheels. During startup, the pot values are read into memory (ADC conversion of ~10ms) and kept for that run of the program. The pot values are only read during alignment.
- On of the changes made on Sunday before Drexel was to "force" the wheels to go to the zero position. This had the effect of moving all the pivots to the same orientation. When doing this we saw "pivot palsy" where the pivot moved back and forth from 355-0-355 degrees due to the PID trying hard to get exactly to zero.
Could there be something about auto mode that is different than telop that affects the loop rate on the PID of auto? Does something in the code affect the loop rate?
- Carly noted that the robot drifts to one side during auto mode. Could this be a code issue or mechanical drag?
Gary> A comment by Paul bothers me. He said that at Drexel he had to change the PID values back to what they were and the lag went away. This does not make sense to me. Sunday before Drexel the test bed set up showed very sluggish response. When the PID coefficient where adjusted the response was very good. This should have nothing to do with lag. Remember the motors had no load on them then. When the robot is on the floor there is more friction. If anything they may need to be higher under load. Additional friction from a damaged pivot #4 would make the PID lack of responsiveness much more apparent. The 2 together could be the problem.
Foster>What are the PID values from Drexel?
Tests and Results
- Talk about how the calibration routines work with the entire team so we understand how it works.
- Why does it take so long to get the wheels pointing forward?
- Examine pivot #4 for additional friction issues
Thesis: Bent Axle
There is a bent axle at xxx location in the pivot mechanism, this would lead to wear because of ?????
Follow-On Issue: Losing Traction
As of April 15, wheel 4's calibration/alignment issue appears to be resolved (See Gary Deaver April 15). The only remaining issue is the wheel's tendency to break traction and spin, again affecting tread wear. This isn't necessarily a major concern, especially if tread replacement isn't an issue and we can bulwark against possible worsening in later-generation assemblies.
- Thesis: 3 Points to a Plane
- The contact area on wheel 4 is not in the plane defined by the other 3. We've seen this before (to a much worse extent) on DEWBOT IV.
- Tests and Results
- If tread hasn't been replaced since latest upgrade and shows signs of wear, replace it and see if the issue reappears. (Even if so, it's still possible to continue fixing the symptom, preferably with faster replacement method.)
- Examine wheels for misalignment. Force plates can be nice if available. If cause and possible bulwarks remain unidentified, consider suspension for next generation.
- Work in R4. May present logistical difficulties.
Gary Deaver and Clem McKown inspected the drive-train and especially Pivot 4 on 10-April.
Mechanically, there was nothing which set Pivot 4 apart from the other pivots except for the fact that Pivot 4's steering chain was fitted with a chain tensioner (at Finger Lakes). Pivot 4's backlash with this tensioner in place was not obviously the greatest of the 4 pivots (but it obviously did not have the least backlash of the 4). The PVC chain tensioner was removed and examined. There was some light chain wear on the tensioner, which was to be expected. There is nowhere near enough slack in Pivot 4's steering chain to replace a link pair with a half-link.
All pivots can be shifted side-to-side in their mounts. This appears to be due to cumulative slop from the 1" flanged bearings, clearance between the bearings and the pivot plate holes (we went overboard in enlarging some of these holes) and clearance between the pivot tubes and the bearing ID (it looks as if Wamac turned down the pivot tube OD a smidge). In future pivots, this shift/slop can be reduced by:
- Using sealed bearings (adds ~$3.50 per bearing x 8);
- Providing tighter fit at the bearing plates; and
- Avoid turning down Pivot Tube diameter at the bearing contact points.
Not clear that Pivot 4's side-to-side "slop" is worse than the other wheels'.
There were no bent axles. No bent pivot tubes. All pivots turn with low frictional resistance. In short, other than the tensioner, Pivot 4 is mechanically consistent with all of the other pivots. The tensioner installed at Finger Lakes was not set tightly and did not significantly increase pivot frictional resistance.
Nylon tensioners were installed on (3) of the pivot steering chains (4th not installed due to perceived lack of time). These tensioners were installed on pivots 1, 3 & 4. The tensioners were set tight, basically eliminating all chain backlash (there is still a little gearbox backlash). The installed tensioners do increase pivot frictional resistance. It is possible that these need to be loosened a bit.
During driving practice following this inspection and tensioner installation, we observed that:
- Pivot 4 still tends to go out of steering alignment with the other three
- Pivot 4 wheel tread was torn/worn up and needs to be changed again (other 3 are still okay)
- The steering motors on the wheels with the new tensioners get hot
During driving practice, tread 4 probably had more interaction with the carpet edge than the other (3).
Plan to loosen the tensioners a bit.
Clem McKown email 11-April
I suggest we log data from wheel 4 and at least one other wheel (all wheels if possible). Can we do this? I think labview would be a natural for this (because this is what people use labview for in the real world).
I would log:
- Joystick x & y
- cherry sensor value by wheel
- steering output signal by wheel
We should log in crab mode. Driving on carpet and stationary on cart/stand (2 tests). Best to log until we "break" wheel 4 (but a shorter log could be very revealing).
Once we have this log, we can analyze in excel and look to see how Pivot 4 differs from its siblings.
Gary Deaver 11-April
Adding the chain tensioners Reduced virtually all lash in the chain. There is some gear lash in the window motor. Now the Cherry sensor will see all movement in the system. We can expect the PID to make allot more adjustment and put more current into the motors. Friction from the chain tensioners will also increase current draw. The window motors are sealed and not designed For continuous run. Question? Is the heat build up OK for a normal match time? How long does it take for them to cool down?
I saw the wheels all tracking properly in the beginning of the drive session on Saturday. The pivot 4 problem must occur after some drive time. I suspect a thermal or intermittent electrical problem. The window motor has thermal protection. It could be defective. The Jag has thermal protection. If the fan is not working the jag would intermittently shut down. All 4 Cherry sensors have cracked housings by the connector part. The pivot 4 Cherry sensor could have a intermittent electrical connection. This damage was probably caused by ball intrusion.
There is always the possibility of the infamous bad PWM cable.
At this time I do not believe this is a programming issue. However, I reserve the right to blame them later.
After the last chain tensioner is added the programmers can work on drive straight and kick Auto.
Further from Gary Deaver (same date):
One more thought. We are using worm gears (integral to the Nisso-Denko steering motors). I have seen worn worm gears that want to lock up. Read the Chief Delphi post. It goes off track until the end.
Maybe a window motor replacement is the next step. Or try swapping jag for a victor?
Gary Deaver April 15
At last night's drive secession the pivots were observed. I did not see the behavior in pivot #4 that has been occurring. Over all the pivots have been behaving as expected. occasionally all pivots show a little out of alignment but do recover. Paul did re-write the code and the response does seam crisper. For some reason pivot # 4 wheel does break traction and spin allot. Would affect tread wear. The possessor sucks so whats new? The kicker hung up twice and the zip ties were intact. Kicker bar is loose. At this point I don't see a problem with pivot 4 steering. Tread wear is an issue. 2 days to think about the processor.
Gary Deaver April 21
At tonight's meeting we replaced all the treads. This took most of the meeting. Two screws were stripped and had to be cut off. After pivot #3 (right rear) was reassembled it was noticed that there was substantial binding in the drive train. After disassembling the pivot there was wear noted on the vertical dive shaft. The decision was made to swap out the pivot. After the new pivot was replaced the dive chain to the cim was too tight and the pivot was stiff. We loosened the chain and the pivot now is free running like the rest. In the future a case hardened shaft should be considered. The issue of cutting key ways needs to be looked into. Also The lithium grease in not performing. It seams to dry out and become very thick. Some better bearing grease should be purchased. The last chain tensioner still needs to be installed. The possessor bar needs to be changed. The robot needs to be calibrated and driven to check tonight's repairs. The damage to pivot #3 must have happened since the Saturday when the chain tensioners were installed because at that time all wheels were checked and turning free. Considering the total # of hours the robot has been driven the mechanical wear is good.