Automation of hand chamfering prisms

    Automation of hand chamfering prism


    HiIm new to this forum and would like to ask for some opinion and discussion regarding the Automation of hand chamfering prism solution. I includes the photos and dimensions with some explanation of the set up. So far my understanding to approach the problem with classical robotics methods consists of the robot uses a vision system to scan the chamfering tool diameter and the prism to determine just how round or out-of-round the tool and prism are . After gathering this information, the robot arm traces the exact contour of the camfering tool using a tooling to hold the sponge. Another robot arm grasp the prism and a laser displacement sensor does a 360-degree check to verify the position of the prism. This is sort of closed-loop concept but would appreciate if any discussion or help to improve the system or add additional components. Another option would be using machine learning/reinforcement learning to control the robot arm and for example can be use also a Stochastic Gradient Descent that affect the learning rate taking into account the minimizing the distance to the target as a function of the angle. Any other idea or methodology how to approach the problem in a better way and how can be done more efficiently? Thanks

    Quote from astronaut

    So far my understanding to approach the problem with classical robotics methods consists of the robot uses a vision system to scan the chamfering tool diameter and the prism to determine just how round or out-of-round the tool and prism are . After gathering this information, the robot arm traces the exact contour of the camfering tool using a tooling to hold the sponge.

    The diagram and photo look more like the sponge is being used to flatten the circular face of the prism, not correct diametrical out-of-roundness.


    Vision systems typically have serious issues dealing with glassy surfaces, with either refraction/translucence or extreme reflectivity.


    What degree of accuracy is required for this application? Industrial robots are typically not high-accuracy devices.

    Yes, the sponge is being used to flatten the circular face of the prism, not correct diametrical out-of-roundness.I have more photos of the set up with exact explanation and dimensions but i can not upload. Haw can I attach 6 more photos? Sponge is just to drop bit water

    Let's back up a bit. What is wrong with the current process that you want to improve?


    What needs to be measured? What are the go/no-go criteria?


    There probably exist specialized optical profilometers for glasses/prisms, but I doubt most generic off-the-shelf vision systems are capable.


    A typical articulated robot arm probably doesn't have the motion accuracy to move a tool to precise positions on the glass. OTOH, if the robot is holding a simple abrasive too and just needs to hold it steady while a turntable spins the glass under the tool (the robot acting as the lathe toolpost to the turntable's lathe spindle, so to speak), it might be possible. Especially if the end effector is designed for compliance, or has some sort of high-precision small-throw motion stage to handle the last sub-mm accuracy motions.


    If precision is the main requirement, then something like a SCARA arm may be a better choice.


    I've attached the accuracy pamphlet for a fairly typical "high-accuracy" model robot as an example.

    So, the only thing you want to automate is the chamfering of the outer, upper circumference?


    Is this positionally controlled or force-controlled?


    What degree of precision is required?


    What are the irregularities in the process? Robots are good at doing something the same way every time. If every prism to be chamfered varies in thickness or diameter, or has to be chamfered to a different depth/angle, that makes automating the process difficult.


    On first look, this seems more like a special-machine application than a robot application. A precision motion jig that can step the chamfering tool to precise increments relative to the center of the spindle, with a cut/measure/cut integration with whatever profilometer is used.


    I would first attack this problem thusly:

    1. Place uncut prism on the turntable

    2. Spin the prims with the turntable while scanning the uncut corner with the profilometer, looking for the highest/widest point

    3. Step the chamfering tool (preset to the correct angle, riding a precision linear axis along the turntable's radius) in to that highest/widest point, minus some number of microns (whatever the optimal "cutting depth" of the chamfering tool is

    4. Once no part of the prism circumference is exceeding the height/width being aimed for in Step 3, goto Step 2.


    Repeat the process, "sneaking up" on the final diameter, much like cutting high-precision diameters on a lathe.

    Its more about automation. How to automate the whole process instead doing it manually . Understand. The manual work is described in the documents photos I attached. So instead of doing it manually Im seeking for automated solution

    You don't want two PIDs in series like that. And a PID control should be fed by the error between the setpoint and the actual. PID_Input=Desire-Position, to use your diagram titles.


    Force-Torque Control on the robot is possible. It would require a very precise calibration. Most robots that support FTC have their own programming structure to build a PID algorithm to closely couple the FTC feedback into the robot's motion.


    I'm not sure how the vision would tie in, since the surface of the prism you want to use the vision on will be pressed against the grinding wheel. You might need the robot to periodically remove the prism from the wheel and present it to the vision system, and use that to generate an offset. Something like:


    1. Robot presents unchamfered edge to the vision system for initial measurement

    2. Robot moves prism down onto grinding wheel using FTC to apply a gradual force until a certain height limit is reached

    3. Robot presents chamfered edge to vision system for measurement

    4. Vision system tells robot to repeat Step 2, but going 0.? mm deeper this time

    5. Once the chamfered edge is within tolerance, set the prism aside.

Advertising from our partners