What's new
What's new

Safety feature |has it been done?

Tyrel

Plastic
Joined
Apr 7, 2021
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.
 
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.
 
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.
 
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.
 
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
 
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.
 
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.
 
My Tesla does it. I wouldn't be surprised if someone got it working in a machine tool.

Would I pay for it?

No.
 
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
 
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.
 
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.
 
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.



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

-Ron
 
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.
 
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.
 
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 ?
 








 
Back
Top