My place of employ is affiliated with a university; Kuka donated a robot to us a few years ago. It is a KL 15 SL with a KR C2 controller. The version (?) of the robot is prototype, as evidenced by the tag near the bottom of the robot, and the fact that the earliest version of the controller software for which machine data on the KR 15 SL is available is 5.5 and our controller is running 5.2. We have used this robot for a few projects but never required much precision from it. Now that we are asking more in this realm I see that it has problems.
Here are some symptoms. We attached a new tool to the robot flange. The tool is long, a little over one foot. It has a very sharp point on the end and I used a nice point to teach it to (using the XYZ 4-Point method). No matter how careful I am about this process the controller will not accept the teaching because it says the error (20 mm) is too large (> 5mm). Second symptom: I have taught the robot a base frame that comprises a highly precise grid attached to a table, using the 3-Point method. Once it is taught I jog the robot in that frame and it fails to track the axes of the grid. When I place the TCP at the origin of the base frame and jog in the base Y-direction the robot drifts away from the Y-axis at a rate of about 1mm per 10mm along Y. These symptoms are not really a problem on their own, but we need to be able to pull accurate position data ($POS_ACT) during operation and that appears to not be possible right now.
We have remastered the robot. I did it with a dial gauge but saw no improvement in the above symptoms so we had a service tech come this week. He remastered with an EMT and the behavior is still the same. One thing interesting is that when the robot controller moves it to {0, -90, 90, 0, 90, 0} A4 does not appear to be parallel to the ground. This is how it looks to the eye (I see it and so do colleagues). I used a digital protractor to check this and found that joint two seems to be off by 0.75 degrees. That is, when the controller says A2 = 90, plane A2-A3 is 0.75 degrees off from the vertical. Similarly, A3 seems to be off by 3.5 degrees. When the controller says A3 = 90 the angle I measured between A4 and plane A2-A3 is 93.5 degrees. The other joints appear to be good, A6 measures perpendicular to A4 when A5 = 90. Could the "reference notch" be improperly located on joints two and three? It seems unlikely, but why else would we see this error?
The service tech mentioned loading a new hard drive with version 5.5 of the software and seeing if that fixes the problem, but he had not seen my super-scientific digital-protractor measurements then. Any insight you can share would be much appreciated.