Height Inspection....

A little background first...

We're just finishing up a corousel-style machine to assemble caps used for spice jars. Just a 2-piece assembly, with the inner "sifter" assembled to the cap. The outer ring of the sifter 'snaps' under a lip in the cap to retain it during shipping. At a later date, when the cap gets screwed onto the jar, the sifter transfers itself to the jar, and you wind up with what you buy at the store. Like this...

[attachment]

I'm inspecting the caps on the discharge conveyor to ensure they're fully seated, and rejecting any failures. The problem I'm having is that there's only a 0.010" height difference between a correctly assembled cap and poorly assembled one. And, even if one half of the sifter is 'snapped in', it could still fall out during shipping, which the end customer does not want (hence the need for 100% inspection).

I'm using a Wenglor YP05PA3 laser gaging sensor to measure the height of the sifter in 2 locations as it passes. It was the only brand I found that could actually 'see' the sifter (it's translucent). Other lasers would look right through the sifter and read the height of the caps, which wasn't what I needed to measure.

I had to use interrupts in the program to read quick enough, as the machine runs 400/min, so the caps pass by pretty quickly! Originally, I used 2 sensors on the conveyor to create a 'look window' across the cap as it passed, but edges of the holes created erroneous 'blips' from the laser sensor that I couldn't filter out without affecting detection of 'true' unassembled caps. Now I'm using the 2 sensors to query the laser sensor at 2 distinct places on the sifter.

The reason I need to check more than one location is because the assembled caps come out with the sifters in a random rotation. Therefore, depending on it's orientation, the point I inspect could be at the correct height, while it's diametrically opposite point could be too high. Looking at 2 diametrically opposite points let's me get ALL the bad parts (If either point is high, it's bad).

Here's my problem... If I set everything up to catch EVERY bad part, I reject quite a few 'good' parts. If I set it to be 'less picky', once in a while I miss a bad one. I have a feeling that because the sifter is translucent, I'm not getting a repeatable distance measurement to the top surface. Imperfections in the sifter might be be causing differences in reflectivity? I'm not really sure...

Soooo, if anyone has any ideas on a more accurate way I could inspect these parts, I'm all ears! I've looked into using a camera, but there's basically no 'visible' difference to look for. It's all in the height of the sifter. Ultrasonic sensors wouldn't care about the translucent sifter, but I don't know of any that have the resolution necessary.

Don't be afraid to give wacky suggestions... They may spark other ideas!...

beerchug

-Eric

We have done this before.

If you ping it and check the sound spectrum can you tell if the cap is screwed on right. See have used a pinger and looked at the sound spectrum to 1 MHZ and looked for harmonics. The good part is you don't care about orientation. This is pretty far out and expensive but it does work if one can see a difference in the harmonics between a good and bad part.

That's not as 'far-fetched' as it sounds ("sounds" <--no pun intended), especially in this application. The sifter is lightweight and would vibrate when not 'connected' to the cap. Although, once it's a partially assembled, I think it would be difficult to tell just 'how' assembled it is...

Just to clarify... As far as our machine is concerned, the cap is not 'screwed on' anything yet. We're only automating the cap and sifter assembly. The cap gets screwed onto the jar at another facility (our customer's customer... i.e. McCormick, Durkee, etc.)

Thanks for the idea though!...

beerchug

-Eric

Wacky #2 - I don't know the speed capabilities of camera based sensors but, if a snapshot could be made of the sifter and compared with a reference (using appropriate pattern matching software) it may be possible to 'see' when the sifter is not seated properly by detecting out-of-focus holes in the sifter. If this is possible, detecting a missing sifter would be trivial.

.02

Unsure if you have specs on devices but take a look at Keyence photoelectric sensors. www.keyence.com Now that I am out of the "Big 3" world, I have been using different types of control devices which I have had good results with Keyence. Give Ron Maul a call and he will be able to help you with this. Michigan (734-591-9922) or [email protected] I don't work with plastics but I am sure they can help you.

Here is a brain-storming idea:

I am guessing that the plastic used is reflective at a steep angle.
The problem is that the cap might sit properly in one side, but not in another side. That means that it sits at an angle and not perfectly horisontal.
Try to direct a laser onto the top of the cap from one side, and catch the laser with a narrow-looking sensor on the other side.
Because the laser light is reflected off the top, any deviation from the horisontal (in any direction) will produce a large deflection of the position where the laser spot hits the sensor.

You probably will have to combine this approach with your original design in order to catch when the cap sits horisontal, but not screwed properly in. But in that case the error should be so marked that you can decrease the sensitivity that catches too many "good" caps.

Wacky idea #3

Eric:

You say that camera's are out because there is no visable "wrongness". That would certainly be true looking DOWN on the sifter v. misaligned sifter. But what about ACROSS the sifter? You would be measuring a slope of 0.010" over a span of about 2" (judging from your picture), which is only a 0.2^o angle, and you'd have to look at it from two different angles. And the jar would have to be PERFECTLY flat on the bottom. But at least it's SOMETHING.

Unfortunately, although I'm starting to work with vision systems at work, I don't know enough about them to be able to tell you their capabilites.

Alright Eric,
You said Wacky,
How's about a "GO - NO/GO guage" Can the bottles run underneath a maximum height gauge. Kind of like when a semi tries to go under an overpass and the cargo is just a little too high. You said 400 per minute though, man would that make a mess.
Tim

whacky?

Here's a different approach to the problem:

How much vibration can a partially assembled sifter take before it's detached completely from the cap? My thought is if you can easily detach a partially assembled sifter while the fully assembled ones stayed put you would only have to inspect for the presence of the sifter. Maybe something like a vibrating conveyor or tumbling chamber would remove the sifters that are not completely attached.

Doug...

• 
  I like the "out of focus" idea! One thing I failed to mention though is that only the outer portion of the sifter changes in height as it 'snaps' into the cap. The outer rim gets 'bowed down' into sort of a convex shape. Therefore, the center portion (with the holes) really doesn't move down much...
  
  Excellent idea nonetheless!...

Steve...

• 
  After a discussion with the customer this afternoon, we've decided to use 2 Keyence LG Laser Gauges, despite my RECENT RANT about Keyence. They are currently using ONE to inpect these parts on a different machine. It looks (at about a 45° angle) at the outermost portion of the sifter, and gives a reliable inspection. The only disadvantage is that it doesn't catch EVERY bad part (see my "Reason I need to check more than one location" paragraph in my original post). They're okay with the added expense of TWO sensors if it means catching EVERY bad part, so it looks like this will be the final solution <Eric crosses his fingers>...

Jesper...

• 
  Your idea would probably work well if a 'bad' sifter sat at a flat angle. Unfortunately, it doesn't (see my reply to Doug), and since the sifter is down inside the cap, the sides of the cap would block the laser at the angle required to look at the edges of the sifter.
  
  Excellent brainstorming though! I said I was open to ANY ideas...

Allen...

• 
  Good guess!... It IS 2" diameter!... :nodi:
  
  For the same reason I gave Jesper, I can't look 'across' the sifter, as the sides of the cap kinda 'get in the way'. Wit 'til you start playing with vision systems... It's incredible what they're capable of these days! Lighting is VERY critical though.

Tim...

• 
  I wish it could be that simple... Again, the sides of the cap prevent that. 'Mechanically' checking height would require descending into the cap to get down to the sifter. We already do that to assemble the sifter to the cap, and it takes a 24 station carousel to accomplish it at speed. We'd have to almost duplicate that to do the inspection, and it would require an individual probe on each of the heads.

Tom...

• 
  Believe it or not, your approach was discussed early in this project. They actually DO that at another one of they're facilities. This would be the ideal solution, but floor space prevented adding a secondary 'cap abuse' station to the system. Preventing the sifters from falling out during transport is what we're trying to untimately prevent, so knocking unassembled ones out before they go into the box would guarantee that.

Many thanks for all the suggestions so far! Feel free to keep bouncing around ideas.

beerchug

-Eric

I thought of this later. It sounds like the cap is inspected from the top, open side up. If the cap could be inverted and placed on a pedestal sized to fit the sifter you could check for both conditions (misplaced and missing) with just a couple of PE's.

1. If the sifter is in crooked, the cap will be too high at some point. A PE or three could be set up such that a beam is broken by the elevated part of the lid.

2. If the sifter is entirely missing the lid would be riding too low and another PE, normally blocked by the lid at the correct height, would sense this by seeing its reflector.

Yes Doug, you've got it right. The cap/sifter is inspected open side up. If we had room, I'd prefer to flip it over. Then I could either do as you suggested or, better yet, just whack the bottom with a hammer to knock loose sifters out!...

I think my animation might have some poeple confused. The jar is NOT part of our machine. Here's another picture to clarify exactly what I'm inspecting. This is what the cap/sifter looks like as it's moving down the conveyor (at 95 feet/min). I put white dots at the places I'll be measuring with the laser sensors. The circumference that crosses these white dots is the portion that gets 'snapped down' into the cap, amd is the only portion that has a measurable difference in height from 'good' to 'bad'.

[attachment]

Oh, and by the way, there are also 8 different color caps, so I had to use a color mark sensor to do the initial "sifter present" inspection. Think about it... The laser sensor will accept the cap if there's no sifter, since nothing is "too high". To make it simpler for operators to set up, I didn't bother to have the laser ALSO check for "too low". It's easy enough to do the "missing sifter" check with a simple color mark sensor. At least THAT part of the inspection works 100%...

beerchug

-Eric

Standby - Brainstorming in progress:
Is the plastic sufficiently reflective so that a very bright object can be seen when using the sifter as a "mirror" ?

If yes, then I think this approach can be used:
Let a camera look down onto the sifter, and let a grid-shaped light source mirror it self in the sifter. Any flexing or bending of the sifter will cause a large distortion in the mirrored image. You know this effect allready from when you look in a normal mirror from a distance; distortions grow proportionally with the distance to the mirror. It will be much easier to catch the distortions of the mirrored image than catching any differences in the normal camera image of the sifter.

Hey Eric,

Believe it or not, I used to do this with parmesan cheese cover inspections, but we were inspecting the foil, your project is quite a bit more difficult... we were able to just check for reflectivity of the foil to see if it was pressed in properly...

Vision system, not a bad idea, but they tend to measure pixelation, etc.. and as you said, each cap is rotated, so it would be a bugger to setup all the rotations of the holes in the sifter.. have used those in the past, but usually with some form of reflective addition to help the vision system, ie, the part is dark, but the add in is refelctive, then compared within a range of total part vs. pixels. Although, this may work with your missing sifters, this will not check for angled inserts.

The idea you are using now for 'snap in' is great, and the one that I have found to work the best. I'm sure that most of the sifters are clear, translucent or white, so your sensors should work just fine. If you were to try and measure the cap color itself, well, that could be an issue, check your QA department on the range of colors they will allow for 'go/no go'. They are usually pretty liberal, and with injection molding colors, it gets varied by the second, depending on regrind content, virgin, etc...

As for the missing sifters, what about some form of the previously mentioned 'ping' to see if it's present? We use a 'taptone' system in the bottling industry to check for bottles with product, and they are pretty accurate to see what the liquid level is in a bottle. It's accurate within .1 ounces, and I'm sure you could 'retrofit' to your needs. Plus, these lines will run up to 800 inspections per minute (speed of the filler) and not sure what they max out at (taptone), but I feel that they would be sufficient for your speeds. In other words, 'taptone' them first, reject what does not pass, then do your sensor check for 'flatness'. I'm guessing that these basically run down a conveyor then dumped into a box for shipping, making it difficult for the operator to inspect every one that comes down the line... I know I'd be cross eyed by the end of the day trying to inspect 400 of the buggers in a minute...lol.

Just some thoughts with my limited experience...

Hoot

Thanks for the reply, Hoot. Sounds like you've 'been around' similar stuff. I spent 13 years at an injection molding plant, so I know all about the variations you can get. Different colors have different shrinkage rates as well, so you can't count on a red part coming out the same dimensions as a black one without changing the setup. Regrind? You'd be amazed how much regrind you can stuff into a black VHS storage case without it being visible from the outside...

I borrowed a Keyence laser gauge from the customer so I can continue while waiting for the new ones to arrive. It looks like I won't need to have a separate 'missing sifter' check, as these units have adjustable upper AND lower limits. I had the machine shop make up all the brackets, etc. needed to mount them on the conveyor, with X/Y slides to allow fine tuning of their position.

Positioning them at a 30° angle to look at the points I indicated in my last picture seems to give a very reliable reading. Part positioning is also much less critical thanks to the angling. Looking straight down required very precise positioning to catch the correct point on the radius.

You mentioned foil liners... They also use foil liners on some of their products, but thankfully we don't have to be concerned with them on this system. Foil isn't so bad though. Last year we built an inspection system for lined caps to check for all kinds of defects. Using 2 DVT cameras (one with a ringlight and one backlit), we were able to pick out defective liners at a line speed of over 800 caps/min!

Great info on that 'taptone' system. I have a different application where something like that would be ideal. Do you have any links to manufacturers?

beerchug

-Eric