Wednesday, May 25, 2011

Process signature monitoring

The main reasons for using machine vision are to find defective or faulty product before shipping, and to ensure that no bad parts are made. For both these classes of application we aim our cameras at the parts, but what if we turned them on the process instead?

That’s the idea behind process signature monitoring, as discussed in a very thorough article in IVD Technology magazine, “Process monitoring: reducing costs and improving quality.” (May 11th, 2011.)

This lengthy piece, written by Ron Pawulski of Sciemetric, compares process monitoring to destructive testing. Unsurprisingly, since Sciemetric make process monitoring equipment, he comes down on the side of monitoring, though to me his arguments make perfect sense.

From a Lean Manufacturing perspective, destructive testing is extremely wasteful. Not only do you throw out good product, but holding inventory while testing is performed delays shipment, takes up floorspace and so on.

The point Pawulski fails to make though is that destructive testing will only catch shifts in the process. What about those random or freak events that mean just one part didn’t get welded? The odds of finding those depend on the size of the sample: improving the odds of detection also increases the costs.

I know you’re nodding in agreement, but I suspect you’re also asking what this has to do with machine vision. Well the IVD Technology piece skimps over the “how” to monitor processes, but I’m suggesting that machine vision is one weapon in the armory. As a smart sensor it should have many uses: go out into you plant and find them!

No comments: