iRVision - Accuracy changes with Part Rotation

    Hello dear colleagues! I was wondering if someone ever experienced the following situation: I am dealing with a high-precision pick and place application and I noticed that the accuracy on the robot is changing according to the rotation of my part. If the part is sitting "straight", close to 0º in reference to the UF, my accuracy is lower than compared when the part is rotated at around 40º.

    My setup is:

    LR-Mate 200 iD 14L with 2 GigE cameras, one mounted on the robot's EoA and another one fixed to get the part offset after picked-up.

    Since this is a high-precision application I do the picking using FOUND POSITION type of offset and I use WINDOW SHIFT TOOL + COMBINATION TOOL to get the exact coordinates of the part's feature and then I move the robot to those coordinates with the camera as the TCP so I can image the part once more but now having the feature aligned with the center-point of the camera's FoV, only after doing this I move the robot with the gripper to perform the picking. After I pick up I move over to the Fixed camera and I get the Tool Offset of that same feature. I noticed that if the part is picked when is sitting at close to 0º the offsets are around 300 microns but if the part is picked a little bit rotated, around 30º the offset is 50 microns, sometimes even less.

    The UF was created using AGFS (Automatic Grid Frame Set) and I set the camera TCP as well and the camera was calibrated using Perspective 2 Plane Calibration.

    To keep things even more accurate I am keeping the the standoff distance the same throughout the process, so if the part is 1.5mm higher than the calibration plate, I move the robot up that same 1.5mm to keep the distance consistent.

    The part Z Height was also entered, but seems to have little to no effect on the feature coordinates.

    I am struggling with it for quite a while, so any tips would help! Next Monday I will be able to post pictures or the program...

    Thanks a lot guys!

    Hello guys, as I said earlier...here are some pictures to show what I am dealing with:

    The above picture shows the part's position and Tool offset. I could not see any drifting when gripping the part with the vacuum, so I believe that this Tool offset is coming from the Robot's going to a wrong position.

    Here is some of the code for picking the part....

    Hello Hawk! Thanks for your reply! Yeah, 250 microns is good, but not enough for what I need... Since the robot is way more accurate when the part is slightly rotated I wonder is there is something that could be improved in order to have the same accuracy no matter the part's orientation.

    If you look at the picture in the bottom right you will see that the part drifted only about 70 microns in X and 50 in Y and that was fairly repeatable, and sometimes I was able to get even lower offsets, so that tells me that the robot is definitely capable of this application.

    I am about to get the J992 Vision Mastering option from FANUC hoping to improve the accuracy by remastering the robot, but honestly I am a little bit skeptical about it because it is a fairly new robot that we bought a little over a year ago and it is being used only for R&D purposes, so yes, we did crashed the robot in the past but in slow speeds, so I think it would not be enough to mess up the mastering, but who knows right?

    Vision mastering will make a robot more accurate for vision processes than factory mastering. Do your vision mastering as close to the same location as the vision application as possible.

    However, you are getting beyond the level of the arms mechanical accuracy. If your tolerance really must be that tight then an articulated arm is not the best solution.

Advertising from our partners