An unexpected shift of a trajectory during 90 degrees angle

Mateusz_ABB · July 10

When trying to make a 90 degrees turn with IRB 6700 robot, it makes weird jumps as if it begun to take a different route. We tried tracking paths with a pen and a plasma cutter and during both test the trajectory looked wrong in a similar way - with the wrong turns happening on 2 opposite sides out of 4 in a square (see a photo below). When MoveL and WaitTime are used the issue is much smaller but still noticable. The robot was serviced by ABB and the 5th axis was replaced, it helped with the smoothness of movement but it didn't fix the problem. Additionally the circles aren't perfect either with jerking along the way in some parts.

What may be the issue and how to solve this problem? Thank you in advance.

lemster68 · July 10

I thought that the entire wrist was supposed to be replaced as an assembly whenever there was an issue with those axes. Anyway, if you can safely activate the brake releases for axes 5 and six, move them by pushing them through their range of motions. What you will be looking for is any feeling of crunching, grinding or spots where it becomes more difficult to move the axes.
Next, are these motions very slow? It might be necessary to change the motion mode in the motion planner.
By the way, there is no picture.

Mateusz_ABB · July 11

EDIT: I've learn that during the replacement of the wrist this check was done and the ABB provided servise couldn't solve my problem that way. However, as I said, it helped with the smoothness (there wasn't any difference of resistance when moving it by pushing) but it didn't affect the bug which was similar before and after.

Thank you I will check that as soon as I'll be able to but in time before it, I'd like to add that when auxiliary point are added before and after the corner the problem is basically resolved but it isn't the solution that would be suitable for a normal use. Does it change anything?

Additionaly, when moving on a z-axis down, stopping and then up, the robot also makes this shift on the way up (it's the most noticable)

And yes I meant that the entire wrist was replaced but I meant that it was because of 5th axis.

Sorry, here's the image:
(the upper right corner is expected (the starting point) and the lower left corner is also quite pretty but the upper left and lower right are problematic)

lemster68 · July 11

Again, what are the programmed speeds?

Mateusz_ABB · July 11

I've tried both v200 and v1000 and some inbetween with overall 25% speed.

lemster68 · July 11

Oh, I assumed it would be much slower since you mentioned the plasma cutter. Probably what you are seeing is backlash in the gearbox. Observe the joint rotation direction when it does these motions. What I want you to look for is a joint (or joints) reversing direction. I am going to guess that they do not in the good corners and do reverse where you are seeing the jump.

Mateusz_ABB · November 12

Thank you for your help and sorry for not answering but only now I'm to try to solve the issue again.

We did tests of axis 1, 4, 5 and 6 with released brakes and they moved smoothly without any suspicious noise. We couldn't do 2 and 3 for obvious reasons.

When looking for reversing direction we couldn't spot any weird movement (both in real life and from signal analysis), but maybe you could describe more what to look for?

For the further tests:
- when jogging we tried a few different position with one on a 45 degrees slope and the problem was still there. When moving in a typical y-axis the line drawn by a pen had quite big overshoot while the one drawn on x-axis didn't. I think, on a z-axis the issue was present as well. However, changing the workspace (typical y became x) the issue was still on the original axis.
- when trying Tune Master with Df parameter on the first try I was told that the cut was better but on the next day the issue was still present. Next tuning didn't change anything and the third made drawing with the pen better, but not good enough to use the robot as intended. I will check the robot again tomorrow.
- the floor it's mounted on isn't the best of the best but the pedestal is rather firm and solid, when Mounting Stiffness Tuning was done, the original values of 1 were unchanged.

- the movement during routines or when jogging can be sometime wobbly,
- we are now trying to do something with signal analysis feature. I don't think it shows any obvious flaws but I'm worried about changing speed when the power is directed to the bleeder. I would appreciate suggestion what to look at. When the robot is making circles with 2 MoveC, it make obvious straight lines before each target point but on the analysis movements on x and y are perfect sines.

Some additional information:
- our robot is the version with the one with tighter gearboxes,
- it's about 2.5 years old with only 2 years of minimal use (because of the problem),
- it's usually working without any "warming up", is it a problem?

lemster68 · November 13

OK. Let's get back to some of the installation basics. Is the correct robot type installed for the system? What is the version of Robotware? Have you verified the calibration offsets? When moving to the zero position, do all the alignment marks look OK? Also, when the wrist was replaced, were those axes fine calibrated? You might inquire about getting the Calpendulum calibration procedure getting done. Does this robot have Absolute accuracy and, if so, is it activated. Are the armloads and payloads set correctly? Have you taught a good tool center point? If not, try teaching one to see if you can get good accuracy.
Finally, what do you mean by "tighter gearboxes? What is the robot model and variant? (I know that you said IRB 6700, but there is more to it.)

Mateusz_ABB · November 13

1. Control system was bought with the robot: IRC5 Single, serial no with 6700

2. Robotware version 6.14.00.01

3. The 5th and 6th calibration offsets were changed during raplacement 5.0891 to 5.7138 and 5.4338 to 0.8374

4. The fine tuning was done during replacement, btw they did change wrist from black steel one for stainless steel one.

5. All alignments seem ok.

6. It doesn't have absolute accuracy

7. There is one small load on 3rd axis (abt. 1kg), and bigger load on a tool (10 kg). We tried to set it today but it didn't change much (with LoadID for the tool and Arm Load for the load, with reset). The problem was before and after adding those weights in the past, the weights were put down during the replacement.

8. We tried both manually entering values and TCP with 4 points, it looks good in both configurations

9. The robot is IRB 670-175/3.05. When the robot was bought, I was told, the salesman asked if we would like the faster version with less accuracy or a bit slower one with 0.5 mm accuracy, the later was chosen.

Could it be something with the regulation? Maybe it's possible to calibrate it? The robot makes slight overshoots in different axis (in x-axis when starting to move in z, in y-axis when starting to move in x, but usually one axis is ok) when jogging. When doing squares with z0 in one corner it undershoots. On graphs it can be seen that there are slight changes to speed when the power goes to bleeder and not to other motors, can this be important? or influenced?

Thank you

lemster68 · November 13

To accurately define these overshoots you are mentioning, make sure that the joystick is locked for the x, y or z which you attempting to check (lock out the ones you are not trying to jog). You mentioned this bleeder thing before, I don't know what that is.
"The fine tuning was done during replacement" Calpendulum?

Mateusz_ABB · November 14

By bleeder I (based on manual) meant the moment when some Motors decelerate but no other motor needs the power so it goes to braking resistor or back to the power grid. I understand some transient state but it's weird that it has influence on speed (about ±5%, normal 60 changes for to 300 ms to 57 and then to 62), isn't it?

Yes, it was official abb service, they used pendulums

Edit: Tests with locking axis was done but the results were the same

lemster68 · November 14

9. The robot is IRB 670-175/3.05. When the robot was bought, I was told, the salesman asked if we would like the faster version with less accuracy or a bit slower one with 0.5 mm accuracy, the later was chosen.

OK, but does the system in the controller match that robot type exactly?

Mateusz_ABB · November 14

Yes, they have exactly the same serial number, and were bought together. The name, current system and id are the same in controller properties

lemster68 · November 14

Did I ask yet if there are any orientation changes or especially large reorientation?

lemster68 · November 14

A couple more questions. What size is that box you are having the robot draw? You mentioned a plasma cutter, what would be the cutting speeds when in use? You might need the option Advanced shape tuning. I have a link here for the Product manual for the IRB 6700 series, one thing that caught my eye was wrist optimization, but I have not yet found a detailed description or how to perform it yet.

https://search.abb.com/library/Download.aspx?DocumentID=3HAC044265-001&LanguageCode=en&DocumentPartId=&Action=Launch

Mateusz_ABB · November 19

There is no orientation changes.
The cut one from the photo is about 3x3 cm but the weird thing is that the problem looks similar in different sizes, I think the biggest we tried is about 17x17cm. Plasma cutter speed is v80. We have advanced robot motion and the wrist optimization was done during the replacement with no luck but we will look into it.

lemster68 · November 19

I didn't bring this up before because you mentioned higher programmed speeds on some earlier posts. But, you might look into changing the system parameter for Motion process mode. You could try Low speed accuracy mode or Low speed stiff mode.
With the option that you just mentioned it has you could try advanced shape tuning and maybe even wristmove.

Mateusz_ABB · November 21

I did more test with those advanced options and they did have some results. Changing to low speed stiff mode made everything horrendous but low speed accuracy mode did help quite a bit.

However, tuneServo options seems a little confusing, for example changing DH to 95 ( the basic value is 100 and, supposedly, it can be changed safely in range 100-30 safely) made the robot make very weird (and noisy) jump at the point when it make one of two overshoots - I assume it is connected because one of the overshoots seems to be in a point when 2nd and 3rd axis begin to change directions and with lower gain it didn't have enough power at that moment.

Overall, combining low speed accuracy with FricIdEvaluate had the best results. After calibraton, lowering yhe speed from v300 to v80 did help as well. I wonder, how after making multiple tests FricId "decides" which path is the correct one. Do you know where I can find information about it? It helped but the problem still remains and we would like to find a way to identify and get rid of the problem. Additionally, when calibrating to make circles, the straight lines begin to look wobbly.

Btw, if somebody else is trying to use FricId and the results get worse with time I recommend to start again from default values.

lemster68 · November 21

Well, it is good to hear that you are making some progress. I have not actually used FricIdEvaluate myself, but maybe I can find some documentation.

An unexpected shift of a trajectory during 90 degrees angle

Best Answer

Answers

Categories

In this Discussion