Saturday, June 4, 2011

Plan for the worst (and hope for the best)

For a vision system to be successful it really helps if the environment is tightly constrained. Put the part you’re looking at in exactly the same place and with exactly the same orientation every time, and whatever you do, keep total control over the lighting. Achieve this and you have a fighting chance, even though you still have to contend with part-to-part and batch-to-batch variation, temperature fluctuations, dirt and gradual decline in light output. (I could go on, but you get the idea.)

I was reminded of this a few weeks back when a chatty salesman told me a - probably apocryphal - tale about an installation his colleague had been involved with.

The application team had thought about controlling ambient light, so they put the system in a box. Of course, the object under inspection, (let’s call it “the part,”) had to be passed in and out of the box, so it had some rather large holes cut in the sides. It didn’t seem that this would be a problem because the box was as big as the floor space provided by Engineering would permit, and overhead lighting in a plant, almost by definition, points downwards.

Well the installation team arrived on site around midday, and as they were setting up the system they noticed the generously proportioned skylights in the roof. But the system was enclosed, so they concluded it would not be a problem.

Next morning the team got in around 7 and started up the system. This was a gauging application, and everything had been positioned and calibrated off-site, so they knew what measurements it should be reporting. Only it wasn’t.

Thinking the part had changed, they checked it carefully. Identical to the samples provided by the customer. Then they ran more parts. These were measured to be the right size. The problem had gone away.

Day three came, sign-off day. The team started in first thing by running more parts before the customer came down to check the performance. The size had shifted again.

I know you’re ahead of me here, so let’s cut to the chase. The morning sun, being low in the sky, was coming through the skylight and falling on the part as it sat inside the enclosure. In a way, the team was lucky; had they been installing at a different time of year it would have been the customer who found the problem, but probably only after producing a lot of “scrap.” Or if the vision system had been oriented differently the problem wouldn’t have happened at all, or not until the plant was next rearranged.

So here’s the moral of the story. In fact, two morals. Don’t make assumptions about the environment where your system will be deployed. Go, see for yourself, and then design accordingly. And second, expect the worst, and design for that. Even if the customer’s plant doesn’t have skylights, what if they move the machine to a new building? It happens all the time.

Plan for the worst, even though it’s okay to hope for the best.

No comments: