In the previous post we built a test station for flat surface line scanning. Such a setup is crucial — it's where you prepare the entire system architecture and the algorithms meant to run on it. But it has its limits: a simulation will always differ from production conditions, where far more variables await than we could ever imagine.
Which factors on a production line only surface in the factory?
Some variables only reveal themselves once the installation is up and running in a real paving stone plant. In our case these were:
- diverse products — a manufacturer often has more than 100 different items in their catalog,
- curvature of the production boards,
- imperfections of the belt conveyors,
- "wet" product (uncured concrete) that, combined with a dark color (e.g. anthracite), absorbs the laser far more than other products,
- dirt on the boards left by the production process,
- vibrations,
- high dust levels,
- external light sources of varying intensity — both sunlight and artificial,
- people working on the line.
Without leaving the laboratory we wouldn't even have known these factors existed and posed a real problem. And trying to reproduce them in the lab would take a very long time — and the volume of data collected still wouldn't be as large as on the line.
Why run a pilot deployment of a vision system at a client's site?
That's why piloting the machine at a prospective client's site is a huge advantage. We took this step last year, and it was one of the best decisions for the product's development. There are above all two arguments in its favor:
- Scale of data. On the test station we performed 20–30 scans a day. The machine at the client does 2 to 3 thousand.
- Real conditions. Without the pilot we wouldn't have learned about the external factors affecting the measurement that I described above.
How to choose a spot for a pilot on the production line?
It helps when the pilot client has plenty of room on the line. Then you can install a larger machine, prepared for possible modifications — repositioning components, adding further elements and other experiments. That was also the case for us. The finished installation, integrated with the production line, can be seen in Photo 1.
Photo 1: The paving stone line-scanning machine, integrated with the client's production line during the pilot deployment.