Having a small set of intelligent devices is pretty reasonable to understand and manage. Many intelligent devices for manufacturing allow parameter adjustment (or control) and most have status or diagnostic information, again pretty easy to manage with a limited number of devices. So perhaps the first challenge we face is that while having autonomous devices is pretty interesting, the reality is that it would be impracticable or even undesirable to have them operate autonomously. Typically a device is part of a larger system that manages a process, and with most processes there needs to be coordination. After all, wasn’t this the reason that manufacturers recognized the need for Control Systems so many years ago? The concerns over security/safety and process optimization/profitability don’t go away because we have intelligent devices. What does happen, though is that we start to look at new ways to incorporate devices. We have been talking about “Islands of Automation” for the past 20 years, and they still exist today, and it is pretty normal to find that a manufacturer may only have 30% of his field devices connected to a Distributed Control System (DCS). One of the reasons for this has been the cost to connect them. With intelligent devices, the promise is that connectivity is built in, reducing the interface cost, but we are still faced with the often overlooked integration issue: unless there is context to the data, then it is just data!
Metcalfe’s Law, formulated circa 1980 postulated that the value of a (telecommunications) network increased exponentially with the number of connections. If you had 2 subscribers with telephones, then you had one possible connection. With 100 subscribers it would be possible to have 10,000 different connections. While that is true, the flaw in the argument is that all have to be speaking the same language. If 50% of the subscribers speak only English while the other 50% speak only Swahili, the number of connections (value) goes down significantly. Even if we agree on a common language, we still have to factor in experience. My Swahili speaking friend may talk about “Spar” and I would be wondering what the discussion has to do with ships, while they would be talking about a retail store. Only if we have a common context (language and experience) do we get the full value of the network.
We have the same issue with intelligent devices. Unless we have some type of model that brings everything to the same context level then the value of making decisions not only decreases, but can become dangerous. Think about what would happen if one device returned a status of “1” that meant “I’m going to fail soon” while your program was expecting that “1” meant “everything is OK”. In order to get value out of these intelligent devices we need to ensure that data is transformed into actionable information, which means that some type of contextual model needs to be applied to all of these intelligent devices.