Safety feature |has it been done?
Close
Login to Your Account
Page 1 of 3 123 LastLast
Results 1 to 20 of 46
  1. #1
    Join Date
    Apr 2021
    Country
    SOUTH AFRICA
    Posts
    2
    Post Thanks / Like
    Likes (Given)
    0
    Likes (Received)
    1

    Default Safety feature |has it been done?

    Good day,

    I am a cnc programmer and operator and I hate it when a machine crashes so I was thinking of making a safety feature with a camera that watches the tool and the part and checks to see if there is going to be a collision. Has this been done with something similar.

  2. #2
    Join Date
    Sep 2016
    Country
    UNITED STATES
    State/Province
    Connecticut
    Posts
    15
    Post Thanks / Like
    Likes (Given)
    13
    Likes (Received)
    12

    Default

    Well this should be a fun thread to follow...


  3. Likes CORONA VIRUS, empwoer liked this post
  4. #3
    Join Date
    Jun 2012
    Country
    UNITED STATES
    State/Province
    Michigan
    Posts
    13,362
    Post Thanks / Like
    Likes (Given)
    4530
    Likes (Received)
    4821

    Default

    Many if not most crashes happen in rapid travel so very difficult to catch in time.

    QT: A camera that watches the tool and the part and checks to see if there is going to be a collision.

    Could have a sensor and then a voice that says "Ops" after a crash.

  5. Likes steve45, 1yesca, TeachMePlease liked this post
  6. #4
    Join Date
    Feb 2005
    Location
    Hatch, NM Chile capital of the WORLD
    Posts
    9,646
    Post Thanks / Like
    Likes (Given)
    16305
    Likes (Received)
    11740

    Default

    Quote Originally Posted by michiganbuck View Post
    Many if not most crashes happen in rapid travel so very difficult to catch in time.
    Have you ever done the math on "Rapid Travel"? Its not very rapid.

    1000 inches per minute is just less than 1mph. Even if you have a rocket fast
    machine, its only going 2 or 3 miles an hour.. Less than walking speed, unless
    you are talking about the lady in front of me in the Walmart aisle, in which
    case, less than 1mph would be Rapid.

    Complexity for complexity sake. It would be cool, but guys running one offs won't pay
    for it, or need it, and places running production would need that feature once, EVER,
    and won't pay for it.

  7. #5
    Join Date
    Apr 2011
    Location
    Nebraska
    Posts
    1,421
    Post Thanks / Like
    Likes (Given)
    2354
    Likes (Received)
    474

    Default

    How exactly is a camera supposed to know before hand that there is going to be a crash? Think about it.

    -Ron

  8. Likes CarbideBob, GregSY, 1yesca, digger doug liked this post
  9. #6
    Join Date
    Jan 2005
    Location
    Houston
    Posts
    6,890
    Post Thanks / Like
    Likes (Given)
    1220
    Likes (Received)
    3900

    Default

    You'd have more nuisance trips than actual accident avoidance trips.

    As we progress (?) into the future, the more I become aware that technology is becoming a real pain in the ass. Smart doorbells! Cars that foresee accidents and avoid them! Lord, it's a miracle!

    Everyone wants their asses wiped for them nowadays. It's not a good thing.

  10. #7
    Join Date
    Jan 2007
    Location
    Flushing/Flint, Michigan
    Posts
    10,310
    Post Thanks / Like
    Likes (Given)
    596
    Likes (Received)
    8393

    Default

    Quote Originally Posted by MetalCarnage View Post
    How exactly is a camera supposed to know before hand that there is going to be a crash? Think about it.
    -Ron
    Lets do this.
    If the camera knew distance to go, had the right camera angle and a high enough resolution, fast update and imaging processing rate, the machine with a good decel, fast interconnect camera to cnc....
    Bob

  11. #8
    Join Date
    Aug 2011
    Country
    UNITED STATES
    State/Province
    Oregon
    Posts
    2,669
    Post Thanks / Like
    Likes (Given)
    632
    Likes (Received)
    809

    Default

    More appropriate maybe would be controllers that run synthetic programs so crashes happen on the screen, a simulation. One of our machines has that built into the controller. My employee is many years past needing it and never uses it.

  12. Likes Thunderjet liked this post
  13. #9
    Join Date
    Jan 2011
    Country
    UNITED STATES
    State/Province
    Michigan
    Posts
    221
    Post Thanks / Like
    Likes (Given)
    3
    Likes (Received)
    109

    Default

    Not a CNBC guy here but wouldn’t the camera have to see in 3D?

  14. #10
    Join Date
    Jan 2012
    Location
    Midland, Texas
    Posts
    1,620
    Post Thanks / Like
    Likes (Given)
    634
    Likes (Received)
    839

    Default

    If using coolant, you would you keep the lens clean? Guess it could have proximity sensors instead of a 'camera'.

  15. #11
    Join Date
    Dec 2012
    Location
    Se Ma USA
    Posts
    2,159
    Post Thanks / Like
    Likes (Given)
    176
    Likes (Received)
    1173

    Default

    You could paint Crash Test Dummy pics on different parts of the machine. Maybe operator had a hoodie that makes him look like a crash test dummy.
    No worry. When the machines become aware all will be well. They will know. There will be a man and a dog running your shop. The man will feed the dog. The dog will prevent the man from touching the machines. Harmony.

  16. Likes 1yesca, empwoer liked this post
  17. #12
    Join Date
    Oct 2009
    Location
    Oregon
    Posts
    5,943
    Post Thanks / Like
    Likes (Given)
    6433
    Likes (Received)
    3309

    Default

    My Tesla does it. I wouldn't be surprised if someone got it working in a machine tool.

    Would I pay for it?

    No.

  18. #13
    Join Date
    Sep 2010
    Location
    Oklahoma City, OK
    Posts
    5,450
    Post Thanks / Like
    Likes (Given)
    1051
    Likes (Received)
    2373

    Default

    Quote Originally Posted by GregSY View Post
    Everyone wants their asses wiped for them nowadays. It's not a good thing.
    I agree.

    Sent via CNC 88HS

  19. #14
    Join Date
    Jan 2007
    Location
    Flushing/Flint, Michigan
    Posts
    10,310
    Post Thanks / Like
    Likes (Given)
    596
    Likes (Received)
    8393

    Default

    Quote Originally Posted by Garwood View Post
    My Tesla does it. I wouldn't be surprised if someone got it working in a machine tool.
    .
    Unless of course the white semi trailer in front of you looks like a cloudy sky. (famous SDV fail)
    You do make a good point on automatic braking systems.
    The integration into a normal cnc machine tool would be so difficult.
    Rewrite the inside of the cnc core for this ? Not a trivial exercise.
    Yes I do think it will be happen but I will be very long dead.
    Kudos to those who try it now and run into all those brick walls. I see so many problems but this will be so cool when it does work.
    Bob

  20. Likes 1yesca, Garwood liked this post
  21. #15
    Join Date
    May 2017
    Country
    UNITED STATES
    State/Province
    Massachusetts
    Posts
    325
    Post Thanks / Like
    Likes (Given)
    21
    Likes (Received)
    83

    Default

    2-4 cameras mounted in the enclosure. They take a couple photos, build a 3d model of everything there. With the right cameras and lighting setup, something like .050" accuracy should be doable. Auto ID the stock from CAM. Run all the crash detection in software, not real time because of coolant.

    I imagine this whole space is a bit of a intellectual property minefield at the moment without deep pockets to guide you. There may be some new IP for machine tool specific 3d crash detection. The whole cameras to 3d model space must be pretty congested though. Hopefully there are some free and clear methods.

    Consider though, the wide variety of crash detection currently available and the economics behind it. I am fairly naïve here, but how many "modern" controls will stop you from sending a long boring bar in the turret through the headwall? That is about as simple as it gets as the machine already "knows" where the boring bar is.

  22. #16
    Join Date
    May 2017
    Country
    UNITED STATES
    State/Province
    Massachusetts
    Posts
    325
    Post Thanks / Like
    Likes (Given)
    21
    Likes (Received)
    83

    Default

    Quote Originally Posted by GregSY View Post
    Everyone wants their asses wiped for them nowadays. It's not a good thing.
    You know the French solved this issue MANY years ago with the Japanese making significant improvements starting 40 years ago.

    If your preference is to walk around with occasionally chapped cheeks though, I would suggest keeping that to yourself on a professional forum such as this.

  23. #17
    Join Date
    Nov 2003
    Country
    UNITED STATES
    State/Province
    Texas
    Posts
    5,866
    Post Thanks / Like
    Likes (Given)
    208
    Likes (Received)
    2011

    Default

    A camera has only one lens. So do the classic demonstration. Take two pencils, the old fashioned type with erasers. Hold them, one in each hand and at arm's distance. But then bend your elbows a bit.

    Now, starting with the eraser ends about one foot apart, slowly bring them together to make the eraser ends into contact. You did it, didn't you.

    But you have two eyes which gives you DEPTH PERCEPTION.

    So do it again, but with one eye closed. I bet you missed three out of four tries. No depth perception with one eye. And no depth perception with just one camera lens.

    A collision avoidance system would need a 3D view of the machining area and that would take at least two cameras. And a bunch of computing power to develop a 3D image of what is going on.

    I am not saying it can not be done. I am saying it will take more than just one camera.

    One thought that comes to me is perhaps a three camera system, one camera looking on each of the three axis (X, Y, and Z) and if two out of the three say there would be a collision, then stop.



    Quote Originally Posted by MetalCarnage View Post
    How exactly is a camera supposed to know before hand that there is going to be a crash? Think about it.

    -Ron

  24. #18
    Join Date
    Apr 2018
    Country
    UNITED STATES MINOR OUTLYING ISLANDS
    Posts
    7,536
    Post Thanks / Like
    Likes (Given)
    0
    Likes (Received)
    3674

    Default

    Adaptive control already slams the brakes on hard if the tool exceeds its programmed load. Not sure that a vision system would be much better, because it is typical to rapid up to .1" away before changing to feedrate, and feedrate alone these days can be pretty fast, so a vision system would only have .1" of movement to stop everything. What's .1" at 800 ipm ? Not very long, I think.

    Not very practical.

  25. #19
    Join Date
    Jun 2013
    Location
    Northern Il
    Posts
    1,647
    Post Thanks / Like
    Likes (Given)
    843
    Likes (Received)
    1507

    Default

    Quote Originally Posted by EPAIII View Post
    A camera has only one lens. So do the classic demonstration. Take two pencils, the old fashioned type with erasers. Hold them, one in each hand and at arm's distance. But then bend your elbows a bit.

    Now, starting with the eraser ends about one foot apart, slowly bring them together to make the eraser ends into contact. You did it, didn't you.

    But you have two eyes which gives you DEPTH PERCEPTION.

    So do it again, but with one eye closed. I bet you missed three out of four tries. No depth perception with one eye. And no depth perception with just one camera lens.

    A collision avoidance system would need a 3D view of the machining area and that would take at least two cameras. And a bunch of computing power to develop a 3D image of what is going on.

    I am not saying it can not be done. I am saying it will take more than just one camera.

    One thought that comes to me is perhaps a three camera system, one camera looking on each of the three axis (X, Y, and Z) and if two out of the three say there would be a collision, then stop.
    Why do you even need a camera? The controller already has all of the information it needs to predict collision issues. If you were to take an approach of a parallel processor comparing future tool position to collision limits, you would have what you are after.

    Don't see a real economic need for this as it would add to controller costs, require mapping the work holding boundaries and the part boundaries no matter which approach you take and all to solve a situation that is caused by human laziness to not double check your work or prove the program path.

  26. #20
    Join Date
    Oct 2005
    Country
    UNITED STATES
    State/Province
    Pennsylvania
    Posts
    21,381
    Post Thanks / Like

    Default

    If your programming is so bad you crash so much, doo you really think your going to also program in the crash detection properly ?

    What class are we helping you with homework ?

  27. Likes Bobw, Newman109, empwoer, Garwood liked this post

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •