Saw blade calibration on robot arm

    Hello everyone,

    Have a robot arm with attach spindle motor and saw blade (see attached image). Having issues with tool angles. Because saw blade has radius of 26 cm, small angle error (~1 degree) in TCP (center of saw blade), results in significant position, angle errors on saw blade edges where the actual cutting is happening.

    Most tool calibration techniques only calculates TCP position and disregards orientation error. Maybe someone has experience with similar tools and can share ideas/techniques how tool orientation can be calibrated?

    Thanks in advance!

    Well, there's a few ways. I think in your case, grabbing a dial indicator and doing it iteratively might be the best way to go.

    I would lock the saw blade so it doesn't turn (if possible), and put some sharpie lines on it parallel with where you want the X and Y axes to be.

    Then create a program that touches one end of the X axis mark to the dial indicator, then moves to touch the other end of the X axis mark to the dial indicator. Zero the dial indicator at the first point, run the program. Based on the distance between the touch points, do something like $TOOL_DATA[1] = TOOL_DATA[1] : {X 0,Y 0, Z 0,A 0, B 0.1, C 0}, where the B value is how much you want to adjust the rotation of the TCP around its own Y axis. Repeat until you have the dial indicator showing the same value at both points.

    Then repeat the process along the Y axis, making changes to C to adjust the rotation around the X axis.

    The Z axis of the tool (going by your illustration) will be mostly irrelevant, except that you'll want to make sure your "touch points" are programmed along the Tool X and Y axes when you start.

    Note: this process works, but there are limits. Eventually you hit a point where improving X disimproves Y, and vice versa, so you'll need to find the best "happy medium" you can live with. Also, this technique depends on the saw blade being flat, and not rotating while you do this process. If the saw blade has any imperfections, you might need to program the motion to drag the dial indicator along the surface rather than just touching at two points, to ensure you get good data.

    Quote from Lemster68

    Why is the TCP in the center of the blade? This is the reason that the cutting edge can be off the intended path. I suggest to make the TCP the actual point on the blade that does the cutting.

    Excellent point. Ideally, the TCP should be at the point of contact. Of course, depending on how deep the blade is cutting, that could be a bit tricky -- should the TCP be at the center of the chord, or at the leading end of the chord?

    There are various cuts based on material design. And robot might need to cut with different edge of the blade depending on cut position and robot reachability.
    Besides, I don't think moving TCP to the edge, would solve this. Right now main issues is that when cutting along a line, saw blade moves sideways. And on longer and/or deeper cuts that increases friction between the side of a blade and material, resulting in stopping the blade completely.

    How it should move along a line. Saw blade moves from right to left in this image (top view, yellow part is the material):

    How it actually moves (exaggerated):


    One idea we are thinking of is to put Inclinometer (or multiple for different axis) sensor (for example) on spindle motor to see angle errors and make corrections. Haven't used them before so wondering what is your opinion about that.

    If the view really is top view, and you can't change the cutting position, I think an inclinometer won't help. As it can't measure the direction of deviation.

    A procedure skyefire described could help, how did you already try to calibrate orientation of tool? There exist at least two standard procedures to calibrate tool orientation.

    Till now we did standard ABB robot arm joint calibration and TCP position calibration as described in RoboDK docs. TCP orientation is just rotated 90 degrees relative to robot flange.

    Seems like angle errors comes from joint inaccuracies, as angle error changes when moving tool along single axis.

    We also did some custom calibration where we used laser distance sensor and collected real distances to the blade surface and theoretical distances using virtual environment. Did some optimization on that and got orientation offset which improved angles but not as much as we would like. Initially angle error could go up to 1.3 degree, now getting up to 0.7 degree error depending on joint configuration and where tool is pointing to. Preferred error is up to 0.2 degree (based on saw blade diameter).

    Accidentally missed SkyeFire first comment. Thanks for the idea, will give that a try.

    Quote from SkyeFire

    Well, there's a few ways. I think in your case, grabbing a dial indicator and doing it iteratively might be the best way to go.

    that works in good hands but even then it can take long time. i like to think of any cutting tools as consumables that get replaced frequently and therefore could/should be measured automatically.

    one can write program that brings robot to 3 points and trigger touch probe. then it is simple to calculate two vectors from those three points. and doing cross product of those vectors gives you the vector that is perpendicular to the blade and you are done. this would be not just convenient but also VERY accurate.

    1) read pinned topic: READ FIRST...

    2) if you have an issue with robot, post question in the correct forum section... do NOT contact me directly

    3) read 1 and 2

    Quote from Darius Grigaliunas

    There are various cuts based on material design. And robot might need to cut with different edge of the blade depending on cut position and robot reachability.

    You are free to use as many TCP's as are necessary.

    It sounds to me like much of your issue (other than using the cutting edge of the blade as the tcp) is due to backlash in the joints. I am going to guess that the speeds used while cutting are slow as well. With that being said, you might be able to use some servo tuning to get netter results.

    You might want to change the Motion Process Mode to Low Speed Stiff Mode If I made the correct assumption that your cutting speeds are slow.

    You can read up on these things in the System Parameter reference manual.

Advertising from our partners