Measuring the position repeatability of an industrial robot

This video introduces our TriCal 3D measurement device, designed for automatic calibration, and precision and accuracy evaluation of industrial robots. The TriCal consists of three bluetooth digital indicators (from Sylvac) mounted orthogonal to each other on a special fixture. The fixture incorporates a magnetic nest, which coupled with a 1.5" precision steel ball, allows the mastering (zeroing) of the device. In other words, while the orientation of the stems of the three digital indicators is not known with high accuracy, the position of the TCP with respect to the robot flange can be measured very accurately (on a CMM). In addition, thanks to the magnetic nest, a 1.5" SMR can be attached at the exact same TCP position, so the device can be used with a laser tracker too (for validation purposes).
The idea behind the TriCal is to use the device to bring in automated mode, and with high accuracy, the robot's TCP to given positions: the centers of 1" precision balls fixed in the robot's base. These positions can be taught manually or even calculated offline. When the robot positions the TriCal on on one of these balls, a piece of software that runs on a PC reads the position errors measured with the three indicators, then sends a command to the robot to offset the position of its end-effector. The process is repeated a couple of times until each measurement is no more than 0.001 mm.
This self alignment process can be used for calibrating the robot, evaluating its position accuracy, or simply for measuring its position repeatability. Indeed, note that the repeatability of industrial robots is typically measured with a laser tracker, but a laser tracker in not accurate enough to measure position errors in the range of 0.020 mm . Our TriCal is accurate to a few micrometers when it comes to measuring errors of up to 1 mm.
The TriCal is therefore not only better than a laser tracker at evaluating the stationary positioning performance of an industrial robot, it is also at least ten times cheaper, portable, and extremely easy to use. If the device is used for robot calibration, then the positions of the precision balls fixed on the robot's base must be known with high accuracy. Ideally, these balls as well as three tooling balls used for constraining the robot's base must be fixed on a large plate and inspected on a CMM. (The robot will then be installed on that plate).
In this first video, we show the process of evaluating the position repeatability of a FANUC LR Mate 200iD/7L. We've followed the
ISO 9283 norm as much as possible, although we do not agree with parts of it. Indeed, because of the nature of our device, we must always approach each measurement position from the above. The only way to fully respect the norm is by using a laser tracker (which is not accurate enough) or a CMM (which is not practical at all).
The results are difficult to summarize, but we can provide the raw data upon request. The mean values for each of the five tested positions, in the order seen in the video (Step 4), as calculated from 300 measurements (after 3 hours of robot warm-up), are: 0.032 mm (for the central position), 0.057 mm and 0.051 mm (for the two upper positions), 0.028 mm and 0.025 mm (for the two lower positions).

[@markemetcalf]

Unfortunately, this is not the correct way to test a robot for repeatability. When you test in this fashion you are preloading all axis gear trains in the same direction, every time. I proved this to my Director at FANUC Robotics by testing similarly, receiving readouts of +/- a half mm, and then (just when everyone thought the robot was perfect) I physically moved the wrist 3 1/2" in the opposite direction and +Z. The ONLY TRUE WAY to test robot repeatability (per Axis) is to invert the arm and counterbalance it for neutrality and then record the measurements when on the gauges and watching the Encoder Counts as you come off the gauges. You then calculate the Delta (slop) and this gives you a true indication of the GOOD the BAD and the REALLY UGLY.

[@magran17]

I love comments like this that not only show what’s wrong, but offer improvement found by someone else’s discovery. 👍

[@impactodelsurenterprise2440]

Robot inherent inaccuracy is one thing sellers and integrators never tell you, unless you ask.

[@ilianbonev5171]

Hello Mark, Sorry, I just saw your comment on my lab's channel. I'm afraid, however, that I don't quite understand it. I agree that the Trical device that we use to measure the robot's repeatability does not respect ISO 9283, because we approach and depart from each position along the same (vertical) direction. Only a laser tracker can be used to measured robot repeatability as prescribed in ISO 9283. The problem with laser trackers, of course, is that they are extremely expensive. The Trical device that we developed in my lab is much more affordable.
That said, the way you measure robot repeatability depends on your actual needs. From a user point of view, the best way to measure robot repeatability is to do exactly what a user would do in a specific application. Since in most applications where repeatability is needed, a robot would go repeatedly to a desired end-effector pose following the same approach path, I don't see why our approach would be wrong. The main problem with our device is that it involves contact between the three digital indicators and a precision ball. However, the spring forces and the friction involved are negligible for most industrial robots. We've done similar tests with a non-contact 3D probe (Trinity), and the results are the same.
The main problem with repeatability is not how you measure it, but how users interpret it. When you teach a pose, your move to it from different directions (jog a bit to the left, jog a bit to the right), at extremely slow speeds, and often while the robot is cold (not warmed up enough). Then, in the actual application, you approach the desired pose from a different direction, at a higher speed, and while the robot is warmed up. We've done tests and showed that there could be a deviation of more than 0.300 mm from the pose that was taught, even though the repeatability as per ISO 9283 is about 0.030 mm.
As for the accuracy, our Trical device was actually developed for calibrating robots. I'll soon post another video where we see the calibration procedure.

[@impactodelsurenterprise2440]

Is the trical for sale and if so how much?

[@ilianbonev5171]

@@impactodelsurenterprise2440 Yes, we have several units of a newer version that are for sale. These do not include the digital indicators and the ATI tool changer, however, which are off-the-shelf. Please contact me at [e-mail] for more details.

[@VectorCrafty]

Maybe while you’re waiting for the robot you can quantize that beat - it’s out of timing calibration. Amazing to see your robotic calibration process though…

[@williamhuang5329]

Hanzhen harmonic gear , over 30 years experience , robot arm gear