Robot makes a small "bump" movement at the beginning of a new motion command or direction change

     

    We have a R-2000iB-210F robot with a R-30iA controller, Handling Tool v7.70p/53. I had no experience with robots prior to messing about with this one. We use it as a router to trim vac formed parts. The cut quality is not what I would expect and consider it unacceptable for most of our parts. Whenever the robot moves there is a little bobble--a small skip up (and it does seem to be upwards not down)-- immediately preceding a move but none at the end of a move. It does this when in teach mode or auto, when cutting or not, when moving slow or fast, using FINE or CNT motions, across various cutting planes, etc.

    I attached photos of a couple cut outs showing the wonky corners to illustrate the issue. This is just an example as cutting other materials and shapes shows similar behavior, as does any motion without physically cutting anything. In this example the geometry being cut is a rectangle with sharp corners and the points were inputted mathematically using the teach pendant with all axes the same except for the X and Y, with FINE motion for all points. The part being cut is black ABS plastic 5mm thick, the fixture is tooling board foam with vac hold down (previous cut paths are visible), bit is a 1/4" dia, rectangle measures about 71mm x 108mm.

    Is this behavior typical of these large payload material handling robots like something inherent in how they are geared? Or is it a mechanical or programming issue? We bought this machine used from a broker last year, the OG integrator (Robotworx) who refurbished it and designed the cell is no longer in business, and we have no support from Fanuc as it isn't licensed.


    thank you for any insight you can provide!

    It is not typical, the R-2000iB/210F should be very repeatable. Make sure the robot base and plate bolts are tight. If it were electrical, I would expect a difference when the controller is first turned on (cool) or warm.

    Based on your description, the tooling may be loose or flexing and/or one or more robot axes has excessive play. If you don't have the knowledge or experience to check robot repeatability, you should reach out to companies that specialize in Fanuc robot repair/rebuild and have them assess the robot. This should include repeatability and sampling of all the grease cavities to check for ferrous dust/particles that indicate excessive mechanical wear.

    RobotWorx had some good folks working for them (Tonya) and a great website (http://www.robots.com) but struggled sometimes with mechanical issues that likely contributed to their demise. Often their competitors (myself included) followed up and fixed those issues. I know the folks that bought many of the assets & website at auction honored some of those warranty issues before finally putting the RobotWorx name to rest.

    It sounds like backlash in the gear reducers. All articulated arm robots have backlash, the longer the arm the worse it is. Also you have an old used robot, so probably worn. Backlash would occur when changing directions.

    A CNC would have been a much better choice for this application.

    IMHO your best off cutting your losses and getting a proper CNC mill or router.

    Quote from HawkME

    It sounds like backlash in the gear reducers. All articulated arm robots have backlash, the longer the arm the worse it is. Also you have an old used robot, so probably worn. Backlash would occur when changing directions.

    A CNC would have been a much better choice for this application.

    IMHO your best off cutting your losses and getting a proper CNC mill or router.

    We have 3 5-axis and 6 3-axis CNCs. My company bought the robot mostly just to experiment with using a robot for trimming complex vac-formed parts that the 5-axis struggles with. Parts that can't be trimmed on a 5-axis are hand-trimmed using trim ledges built into the vac tools but there's a lot of potential for rejected parts requiring a precise fit due to a trimmer's approach angle, etc.

    The parts that I have been cutting holes and other cut-outs in have many curves and crevices and also get fully perimeter trimmed with the robot. Perimeter trims usually turn out much better since any of these gear moves kinda "hide" in the more complex perimeter trim.

    I had already concluded that this particular robot isn't a very good match to our needs--it has about the same reach and clearance as our 5-axis but with a worse cut quality--however we are still interested in robots for production.

    Quote from Nation

    Is your payload setup correctly?

    How many hours are on the arm?

    Payload setup? No idea

    Timers:

    F Number: F84644
    VERSION : HandlingTool
    $VERSION: V7.70107 10/3/2018
    DATE: 06-DEC-23 16:14
    TIMER INFORMATION::

    SYSTEM TIMERS:
    Timer type Total(h) Lap(m)
    -On Power time: 377.5 0.0[ OFF ]
    -Servo on time: 103.4 0.0[ OFF ]
    -Running time: 32.3 0.0[ OFF ]
    -Waiting time: 2.5 0.0[ OFF ]

    PROGRAM TIMERS:
    count comment
    [ 1] 1266028.63(s)[WARMUP]
    [ 2] 0.00(s)[]
    [ 3] 0.00(s)[]
    [ 4] 0.00(s)[]
    [ 5] 0.00(s)[]
    [ 6] 0.00(s)[]
    [ 7] 0.00(s)[]
    [ 8] 0.00(s)[]
    [ 9] 0.00(s)[]
    [10] 0.00(s)[]

    CNCs are built for high accuracy and rigidity, at the cost of $$$ and working envelope. Robots are far less $$$ for similar working envelopes and greater flexibility, but sacrifice accuracy and rigidity to do so. Robots are usually quite good at repeatability, and to achieve better accuracy, one needs to find a way to leverage that repeatability.

    Robot backlash is generally much worse than CNC backlash, even assuming the backlash of each axis is the same. If you look at the serial kinematics of the robot, as compared to the parallel kinematics of a CNC, it's obvious why. Backlash effects on a robot are amplified over a longer distance, and in unpredictable ways -- where some backlash in a single CNC axis has predictable 1-dimensional effects and can be largely corrected for, the same backlash in the shoulder or elbow of a robot can have very different effects for different joint poses. And so on, and so on. The older and/or larger the robot, generally the worse the effects.

    The worst effects are cause by direction changes, especially if any axis reverses direction. Typically, in CNC-like operations using robots, the best approach is to never do corners or change directions while in contact with the workpiece. You want to get a "running start" from out in the air so that the backlash effects are fully settled before making contact, and "run out" past the workpiece edge before changing direction.

    Even then, there can be problems. For example, imagine the robot cutting a line parallel to its Y axis, from left to right. As J1 approaches 0, J2 and J3 go to zero speed, then reverse direction. J4-J5 might as well. Thus, some degree of backlash on multiple axes is injected directly into your TCP positioning.

    The robot also tries to compensate for gravity effects using the load data, which is why having accurate load data is critical.

    Mitigations for backlash effects include writing path programs to play to the robot's strengths and avoid its weaknesses, "sneaking up" on a cut (especialy with some measurement between cuts), tracking heads, force sensing, fine-tuning the robot with something like NRK's SARCA, and other techniques. Some robots are built for tighter tolerances, but compared to a CNC, a robot should always be treated like a clapped-out mill with lots of backlash and slop everywhere. Good results are possible, but require a very different approach.

    Quote from eklug

    Payload setup? No idea

    Get that done, as it will help. I'm not sure it will eliminate all your issues, but if the robot thinks it has 210kg on the EOAT, and it has significantly less, it will overshoot.

    Quote from eklug

    -Servo on time: 103.4 0.0[ OFF ]

    Hmm. 103hours for a V7.7 era robot is not very believable. Unless it was never commissioned. I wonder if it had a new image dropped in it. V7.7 is 2007 era, and I would not be surprised if the arm has 40k-80k hours on it. Depends on its previous life. You may just have some worn out reducers. My old ABB with 80k hours on it would produce similar corners.

    Check out the Fanuc position converter I wrote here! Now open source!

    Check out my example Fanuc Ethernet/IP Explicit Messaging program here!

Advertising from our partners