March 23, 2019, 09:04:12 AM
Robotforum | Industrial Robots Community

 Gymnastics for robot

Author Topic:  Gymnastics for robot  (Read 449 times)

0 Members and 1 Guest are viewing this topic.

December 04, 2018, 11:50:13 AM
Read 449 times
Offline

TadeuFerraz


Good morning people,

Have you heard of gymnastics for robots?
The idea is to warm the robot's axes, gradually increasing the speed for 40 minutes in a movement routine until reaching the speed of execution of the program.

According to the project creator, in days that the ambient temperature is equal to or below 10ºC, the robot presents execution failures related to speeding in the axes (possibly by the viscosity of the lubricant), and this axis heating routine solved the problem (in an ABB robot).

I am in a hot country, but my boss asked to evaluate such a procedure.

Did you ever see something like that?

Sorry if there is an error in the English translation.

Thanks in advance.

Today at 09:04:12 AM
Reply #1

Advertisement

Guest

December 04, 2018, 12:53:31 PM
Reply #1
Offline

SkyeFire

Global Moderator
Well, KUKA robots actually have something like this built in (as an option).  However, rather than "gymnastics," the robot speed is limited based on the motor temperature.  So, if the robot is cold, it performs normal production operations, but at reduced speed.  The speed is ramped up gradually as the motors and gearing warm up.

I've only seen this used in very cold conditions -- in this particular case, the robots were special "Arctic" models being operated in an environment of about -40C.  The lubricant inside the robots had been replaced with a low-temperature lubricant that was so thin the robot would actually be damaged if it were operated in temperatures above +10C.  And even then, the lubricant would become gelatinous and cause the motors to suffer torque faults if the robot stopped moving for more than 5-10 minutes.  So we used this warmup option whenever the robot had been shut off and re-started.  But since upstream interruptions could cause unpredictable periods where the robot would be on, but inactive, we eventually added a small "dance" or "gymnastic" routine that the robot would run whenever it was left idle for more than 1-2 minutes, merely to keep the motors and lubricant warm.

So, if you need to add something like this to an existing production robot, I see two options:
1.  Temporarily reduce the robot's speed for normal production operation
2.  Add a special function to warm up the robot separate from production

For either of these options, you will need to have some sort of trigger condition: either temperature sensing, motor torque sensing, or a simple timer measuring how long the robot has been stationary.  Possibly a combination of these.

If the ABB(?) controller can monitor the motor torque values, and adjust the robot speed, it might be as simple as starting the robot at low speed, and increase speed until the motor torques become dangerously high.  Then run the robot at that speed for a short time, and try increasing the speed again, until eventually it reaches 100%.

If the robot is being shut off, then you would want to set a condition that would trigger the warmup for the first several robot cycles after reboot.  On the other hand, if the issue arises when the robot is powered on but idle, a simple timer to "exercise" the axes whenever the robot has been stationary for too long might suffice.  Obviously, shutting the robot off overnight or for a holiday would also count as being stationary and trigger the exercise routine.

December 05, 2018, 01:51:50 AM
Reply #2
Offline

TygerDawg


I have implemented a warm up routine before.  In a previous life I was using robots to perform extremely precise 3-Dimensional path applications.  I tested and documented that the robot's TCP location in space was more consistent for a sufficiently-warmed up robot than a cold robot.  After tests proved it with hard data, a 20-minute all axes / all configurations warmup routine was implemented.
TygerDawg
Blue Technik
Virtuoso Robotics Engineering
www.bluetechnik.com

December 05, 2018, 10:54:23 AM
Reply #3
Offline

humphrey


TygerDawg, what was the ambient temperature during your tests?

December 05, 2018, 12:06:30 PM
Reply #4
Offline

TadeuFerraz


I have doubts about the temperature too, I live in a hot region of Brazil, I do not think it would be necessary this procedure, but if there are indications that even in higher temperatures the procedure is beneficial I will carry out.

December 05, 2018, 01:29:42 PM
Reply #5
Offline

TygerDawg


Quote
what was the ambient temperature during your tests?

"Normal" shop temperatures 72F - 78F, variable and sometimes warmer.  Variable humidity.  This was not a controlled environment.

The mechanical engineer in me theorized that hydrodynamic effects of lubricating oil at different temperatures, expansion / contraction effects of gears & links & other mechanical elements in the overall kinematic chain, effect of more or less motor torque required to move elements, more or less aggregated stiction of all elements might have all contributed in some way to the influence the performance that was demonstrated.

The electrical engineer in me theorized that more or less torque required more or less electrons flowing which imparted heat and caused temperature gradients in the components, links, & elements.  The exercise routine imparted to the manipulator joules of energy and time to warm up and stabilize the temperature through the majority of the manipulator elements.

Keep in mind:  the magnitude of my dimensional process variation under investigation was approximately +/- 0.010 mm on a cylindrical tolerance zone around a 3D path, well below a typical robot's repeatability specification.  My measurements indicated a tighter statistical distribution of tolerance after warm up compared to cold state.  I do not claim that there was some magical miraculous improvement, just a tighter tolerance distribution which served my purposes at the time.  For more consistent improvement I would put the arm in a controlled environment and keep it running continuously to maintain temperatures.

December 05, 2018, 03:38:17 PM
Reply #6
Offline

SkyeFire

Global Moderator
I know from some testing I did in the past that the thermal expansion of the robot's link sections, just from the robot's own waste heat, would cause the robot's accuracy to change as the robot warmed up.  I wasn't looking at whether the robot was more or less accurate at certain temperatures -- rather, I was just examining the degree of position change over temperature. 

We tested by shutting down the robot for 12hrs overnight, then powering up and performing a detailed metrology test pattern of the robot, followed by 10 minutes of very heavy "gymnastics" designed to stress the servos close to maximum, followed by another metrology test, over and over again, all day.  What we found was that the robot would "drift" slightly over time, until the internal temperatures reached a stable level.  The time lag in the data suggested that the waste heat from the servos was "creeping" through the arm, causing thermal expansion, until eventually an equilibrium was reached between the waste heat emission rate and the ambient air cooling of the arm.

  So despite the fact that these robots were operating in a controlled environment, I had to come up with a way to ensure that the robots' internal temperatures were stable during production.  Which lead to a warm-up cycle used whenever the robots had been idle too long.  But that was a pretty exotic case.

Today at 09:04:12 AM
Reply #7

Advertisement

Guest


Share via facebook Share via linkedin Share via pinterest Share via reddit Share via twitter