This was a follow-on case study with Schneider Electric where we used product data to drive configuration management. Configuring industrial electrical devices is complex, and dangerous. When you are dealing with thousands of volts of electricity, getting it wrong results in buildings burning down, equipment being ruined, and people dying.
The previous approach was to wait until a new line of products was to be rolled out in a particular country. At that point, a group of skilled engineers would review the specifications and adjust the rules that determined compatibility.
This compatibility information was reduced to a series of quite complex spreadsheets. Each spreadsheet had tens of thousands of rows in many tabs. These were designed in a way that made it possible to drive the configuration rule engines in their ERP system, SAP.
During this time, we had been evolving the knowledge graph representation and beginning to think about the deeper nature of electrical devices. One thing led to another. Our sponsor suggested going to the source (the designers of these parts) and we were able to uncover the simplicity behind the complexity.
It turned out, we were able to distill all the information contained with spreadsheet into about two dozen rules that established whether two parts were “intrinsically compatible” and a few dozen more exceptions to these rules. We were able to implement these rules in SPARQL. Our implementation partner, mPhasis, was then able to take this new information and build a more sophisticated configuration management platform.
One interesting thing about simplifying this product model representation: as soon as the product was designed, we knew immediately what it was compatible with. We merely needed to wait until the product was offered for sale, say in Austria, to know all the other Austrian parts it was compatible with. The model allowed re-use and extensibility