Wednesday, May 11, 2011

Who should train new parts?

I imagine that in some factories a vision system can be implemented on a line and then left to run for years without ever again being touched by an engineer. Unfortunately I don’t work in such a plant. In my world the products change every few months, not by much but enough that the inspection programs need changing.

To give a few examples:
  • Marketing like to reposition the logo every few months, so the logo pattern match tool has to be moved.
  • Engineering add, and then later delete, a drain hole. Each time, the inspection programs have to be modified.
  • Purchasing switch coating supplier so the reflectivity changes and I have to adjust thresholds.

I like the engineers in the plant to make these changes, but it does mean that:

  • The developer has to expose all the relevant controls and provide a way to save each new part configuration file.
  • The engineer needs sufficient understanding of the way the vision system works to be able to make the changes.

The alternative is that the machine vision integrator re-teaches the system each time the product changes slightly, invariably, for a small fee. (“Just to cover our expenses.” Yeah, right, like that 42 foot boat!)

If I was running an integration business this is the approach I’d take: it’s a guaranteed income stream for ten years or more. But from a factory point of view, I think its madness. So what should a plant do about it?

It starts before buying the vision system, at the specification stage. Try to identify aspects of the product to be inspected that might change (I’ve given three examples above,) and spell out to the integrator that you want control over these aspects of the vision program.

Next, when the vendor submits his proposal, look to see how teaching of new or modified parts is handled. I like to see a configuration file that can be opened in Notepad though I’m sure there are other approaches too.

Lastly, don’t allow your engineers to buy “black box” vision systems. Push them to be actively involved in design, build and testing because that way they’ll have a reasonable grasp of how the thing works and the likely impact of product design changes.

Yes, as always, the in-house approach looks harder and more expensive, but what’s the alternative? Being held hostage by a very clever machine vision integrator? I don’t like that one bit, and neither should you.

No comments: