Vision System Camera Adjustment - Change Verification?

Just looking for information or any advice you industry leaders can offer on a vision project I'm working on.

Currently we depend on a lot of cameras (mostly DVT) for process control poka yoke - operator cycles the machine with the wrong component used, the camera detects this and the part is locked in place until a team leader can investigate. At times we have false fails; sometimes ambient or stray light, sometimes variation in the part color, other times dust collection, etc. This generally leads to a camera adjustment (exposure modified, a threshold increased or decreased, etc.). Technicians make adjustments with good intent but sometimes these changes expose us to risk.

What do you guys do for vision changes in your industry? Just run a few bad parts post-change to verify implementation? Do you re-run the entire capability study? Do you communicate changes with your peers - if so, how?

Any information you can provide is appreciated!!

kb103 said:

...Technicians make adjustments with good intent but sometimes these changes expose us to risk...

Click to expand...

I use the equipment for measurement rather than P/Y, but I think the concept is the same...your statement above is the root...

My technicians do not have the ability to make adjustments...Equipment lockout, software password protected.

Stray lighting is controlled to the degree required rather than compensating for it in the machine. This way I don't have the risk that you describe.
Controlling lighting is pretty cheap and easy. So is dust. Controlling lighting can also account for color variation to a degree (shades of gray). We use Red light which washes out some of the color variation.

(Don't use Red light if color is one of your P/Y acceptance criteria!)

I use a challege to test, like a mini Operational Qualification using 'bad' parts and 'good parts'. Infact, the challenge is procedurualized and is done at the beginning of each shift, and if there are any problems with the production line.