Scanning Laser (obstruction detection). / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Scanning Laser (obstruction detection).

1 2 next

Move Thread

LAN_403

MScooling

Thu Jul 17 2008, 02:36PM

Registered Member #1016 Joined: Fri Sept 21 2007, 08:46PM
Location:
Posts: 19

Hey guys,

I work with a machine that has occasional issues with small pieces of scrap steel (about the size of a 10mm crescent wrench) damaging our product as it goes through a water quench. The scrap comes with the product from an upstream robotic plasma cutting process and the control engineers haven't been able to fix the problem.

My initial idea was to use CCD cameras along with NI Vision software to detect scrap in the bed, but the visual variances of the system are making detection difficult. The other idea was to fire a laser down the length of the bed and near the surface. Scrap would break the beam and activate an alarm. The beam would have to be very close to the bottom surface since the scrap material can be as thin as 4.8mm.

Here is a video of the process:
Snapshot

http://filebox.vt.edu/users/sparkman/quench.avi (3.3Meg, Microsoft MPEG4)

This is the target area:
http://filebox.vt.edu/users/sparkman/target_area.jpg

The quench bed is around 50 feet long and is decently flat.

And here are two GIF animations I created to demonstrate the concept.

http://filebox.vt.edu/users/sparkman/laser2.GIF
http://filebox.vt.edu/users/sparkman/laser3.gif (shows beam location and minimum scrap thickness)

Iâ€™m looking to see if anyone knows of scanning laser system similar to the concept.

Thanks

Dave47

Thu Jul 17 2008, 04:45PM

Registered Member #84 Joined: Thu Feb 09 2006, 01:06PM
Location: Dallas, TX
Posts: 47

The DARPA urban Grand Challenge guys used one of these...

It says its distance accuracy is <2cm, though. I wouldn't know if its sensitivity could be modified for reliable 0.5cm accuracy.

David

Carbon_Rod

Thu Jul 17 2008, 09:13PM

Registered Member #65 Joined: Thu Feb 09 2006, 06:43AM
Location:
Posts: 1155

As the above post stated... most Lidars/Ladars do not have sub cm accuracy.

You may find painting the bed or work-material with a fluorescent paint for a vision system easier. In this case pairs of cameras (at 45' and 315') on each side would be needed to cover the blind spots.

OpenCV is often used for high speed detection of edges, shapes, and color clusters.

Nik

Thu Jul 17 2008, 09:28PM

Registered Member #53 Joined: Thu Feb 09 2006, 04:31AM
Location: Ontario, Canada
Posts: 638

Being an electrician (apprentice anyway) and having worked with touchy sensors Im going to say that it would probably be more trouble caused by the laser picking up things other then the debris you are looking for. Would it not be possible to have a brush/sweeper run over the area during part of the machines cycle when it would not be interrupting anything? It would cut down the complexity but it would be larger in size but, unlike a laser it would rarely (if ever) have to be cleaned.

mikeselectricstuff

Thu Jul 17 2008, 10:20PM

Registered Member #311 Joined: Sun Mar 12 2006, 08:28PM
Location:
Posts: 253

I can't play the vid on this PC - does the scrap arrive along with the bar being quenched, or do you want to check for a clean bed first?
If the former, you may have problems with the hot air around the bar causing optical shimmering, messing up any optical sensing.

Might a simple solution be to run a brush or something along the bed? Or some kind of mechanical probe that would detect any obstruction?

As it's metal you're looking for, might an inductive sensor scheme be practical - a number of rectangular coils in the bed ( far enough away from the surrouding metal) - obviously environmental protection of the coils may be an issue.

Is there somewhere a powerful magnet or magnets could be placed to catch any small bits on their way in without affecting the larger product bar?

MScooling

Fri Jul 18 2008, 02:03PM

Registered Member #1016 Joined: Fri Sept 21 2007, 08:46PM
Location:
Posts: 19

Here is a bit more information. Our product is essentially a c-channel rail the ranges from 10-40 feet in length. Before the rail enters the furnace, characteristic shapes are cut into the rail with a robotic plasma cutter. Sometimes the the scrap material hangs on to the rail and breaks off when it reaches the quench. It would be ideal to detect this scrap material before the rail entered the furnace, but I don't believe I will see the funding for a system that complex.

You may find painting the bed or work-material with a fluorescent paint for a vision system easier.

Our product enters the quench bed at 1600F. Paint is not going to survive. The bed is made from hardened tool steel, and we replace the bed plates regulary due to wear.

I can't play the vid on this PC - does the scrap arrive along with the bar being quenched, or do you want to check for a clean bed first?
If the former, you may have problems with the hot air around the bar causing optical shimmering, messing up any optical sensing.

We would only scan periodically, perhaps after 10 or 20 rails go through the quench or whenever there is a gap in production. The rail would be out of the bed when we use the laser to scan for scrap.

Im going to say that it would probably be more trouble caused by the laser picking up things other then the debris you are looking for.

If we do purchase a laser/sensor, it will need to have a sensitivity adjustment so it doesn't pick up insignificant objects.

I played around with NI Vision to detect scrap in the bed. Here is what I came up with:

http://filebox.vt.edu/users/sparkman/compare1.gif

http://filebox.vt.edu/users/sparkman/shape_detection.bmp

I was looking at this "laser fence": http://www.mcelasers.com.au/product.php?code=LR.100

If I could mount the LR100 on a frame that moved back and forth, it might get the job done.

Thanks for the ideas.

aonomus

Fri Jul 18 2008, 02:49PM

Registered Member #1497 Joined: Thu May 22 2008, 05:24AM
Location: Toronto, Ontario, Canada
Posts: 801

Why not reverse the detection scheme? Instead of looking for the scrap metal in the trough, scan the rail on its way in to check for a gap where the scrap metal should not be. It might be hard with a camera since the metal is so hot it'l give off a ton of IR, but a laser could check for however many 'holes' there should be?

GeordieBoy

Fri Jul 18 2008, 03:50PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Have you investigated capacitive proximity sensors like those used on plasma cutting heads to keep the torch the correct distance from the workpiece? Something like that may be able to detect pieces of scrap metal sticking up on the bar as it passes under a sensor.

I can't remember who makes then now, but they work for detecting proximity to conducting material and they work in extremely harsh enviroments. Hence their choice for distance control in plasma apps, where heat, light, acoustic noise, electrical noise and dust contamination are all present.

-Richie,

MScooling

Thu Jul 31 2008, 03:56PM

Registered Member #1016 Joined: Fri Sept 21 2007, 08:46PM
Location:
Posts: 19

I got around to mounting a laser level on the quench bed to determine if the concept was worth pursuing.

Here are a few pictures I took:

Setup:
Laser Setup

LINK (1.0meg)

Initial Beam
LINK (1.0meg)

Midpoint down the quench bed:
LINK (1.0meg)

End of quench bed:
LINK (1.0meg)

Past quench bed:
LINK(1.0meg)

The notebook paper is scattering quite a bit of the beam, so the beam appears more diffused than it really is.

Halfway down the bed the beam quality is decent, but by the end it has diffused too much. I'm sure I can find a laser/lense setup that would have a greater distance than the inexpensive test laser.

Anyways, just figured I would update.

mikeselectricstuff

Fri Aug 01 2008, 08:46AM

Registered Member #311 Joined: Sun Mar 12 2006, 08:28PM
Location:
Posts: 253

A helium neon laser will give a well-collimated beam over a decent distance - often available cheap on ebay.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob