Here are some tricks posted by Audrey Boucher-Genesse at that might help in your automation journey:

  1. Use Feedback. Using a feedback system to simplify the path teaching process. You could use a force sensor, a camera, etc. The principle: teach the path approximately, plug-in feedback device, use feedback from the device to compute the variation from the programmed trajectory and adjust the path to reduce the error.
  2. Keep the operator in the loop!  Operators know the real deal. They know the process better than anyone else. Better still, they know the variations of the process. The normal variations, the doesn’t-happen-much-variations, and the I’ve-seen-this-only-once-variation, the this-could-maybe-happen-if-variations. None of the above is probably stated in the official specs, so if the person that tries to automate the process bases its choices only on the theoretical specs, problems will arise one day or another…
  3. Define numerical boundaries. There’s nothing more annoying for a programmer than not having a clear understanding of what needs to be accomplished. If the spec is unclear, the result will be unreliable. In order to make the purpose clear, sit down with the people that know the process and start asking questions. Here are a couple of examples of questions that might trigger some questions of your own:
    • For an inspection process: What size of defect is acceptable? Take a bunch of parts, from various batches, inspected by your experts. Make them take notes of everything they notice, then their classification (bad / acceptable / borderline acceptable, and why). Examine each one of these notes: “This scratch is considered OK – but is it because it’s not deep enough, long enough, wide enough?”
    • For a painting process: What is the speed of application, the quantity of paint to be dispensed? Should the trajectory go beyond the part to make sure that the paint covers evenly, even in the borders? If not, how far from the border should it stop?
    • For machine tending: After putting the part in the machine, how many seconds should the robot wait before opening the machine again? Or should it wait for a signal?
    • For a pick and place application: What Gripper stroke is required to grasp the object correctly? What robot acceleration will provide a secure grip? What is the limit friction coefficient for a part with cutting fluid on it?
    • Drilling: What is the exact location of the holes to be drilled? Are they relative to a reference plane? Should the hole be drilled at a 90° angle or at some other angle?
  4. Use the CAD model. Using the CAD to teach paths that require high-precision or that have to follow a complex shape. A CAD model should not be used in every automation.  If you can teach the path using Robotiq’s Kinetic teaching, or using your traditional way of teaching, and this works out well for you, then go ahead.  If, however, the use of your traditional technique means using the robot’s precious time to program a complex path, then consider going offline with your robot’s simulator and the part’s CAD model. You will be able to:
    • Improve the path’s precision
    • Use less of the robot’s online time
    • Simulate the path (I don’t know about you, but I so much prefer crashing a simulated robot in a CAD model than all of this in the real world).
    • Program a path much more efficiently
  5. Don’t overlook the importance of calibration. If you use a CAD model, as stated above, don’t ASSUME that the part’s reference frame stands exactly where your CAD model was. Of course, you might have designed a bracket that holds the object at this specific XYZ location, but every setup has some imprecisions, some slack in a joint, etc. The robot itself has some variations from one robot to another, and is more precise in some areas of its work envelope. So what should you do? Calibrate. That stands for the use of CAD models, but also for everything that uses a reference frame that is used for the path of the robot. For example, a camera’s focus point should also be calibrated: if it isn’t, don’t be surprised to get blurred pictures. It is good practice to run the calibration routine regularly (it could be every day, every month, etc. depending on your need). If a change has occurred, the path will automatically be adjusted.
    • Pro tip #1: Since the path you have taught will be adjusted with calibration, make sure your path is not close to a robot limit. If you’re close to the robot reach, or close to a joint limit (e.g. 89 degrees), consider changing this point or changing the robot configuration for this point, so you’ll have more buffer for changes when the path gets adjusted.

Robitiq is offering a free book entitled, “How to Simplify a Complex Task.”

> Read more by Audrey Boucher-Genesse, robotiq, August 8, 2016

5 Ways to Ease the Automation of a Process