Analytics at the edge
Why the industrial control systems of the future will need intelligence everywhere
by David Robinson, Market Development Manager
Data in itself has limited value, it’s what you do with it that counts. The true value of the industrial internet of things (IIoT) doesn’t come from its ability to capture and share data from products and machines. It comes from the transformation of that data into information that can be used to make those products and machines work more efficiently and effectively.
Analytics is the key to that transformation, and its impact is already reshaping many sectors. While traditional supermarkets had rich data on customer shopping habits, they struggled for many years to turn it into real operational or commercial benefits. Today’s leading internet retailers use the power of analytics to inform everything they do, tailoring promotions, pricing and inventory locations on the basis of customer clicks.
In the industrial context, analytics is set to have a similarly revolutionary impact, but most companies are only scratching the surface today. ABI Research estimates that, in 2014, 84 per cent of the analytics associated with IoT data was of the simplest “descriptive and diagnostic” kind: companies were using their data to learn what had happened in their systems, or why things turned out as they did.
That’s useful, but not nearly as useful as the next steps in the analytics hierarchy: predictive analytics, which uses data to explain what will happen next, and prescriptive analytics, which shows how outcomes can be shaped, or even takes decisions automatically to improve system performance.
By 2020, the researchers say, those predictive and prescriptive approaches will account for more than half of all industrial analytics by value, and that’s in the context of an analytics market that will be around five times bigger than it was in the middle of the decade.
Industrial companies hoping to make the best of the analytics opportunity have a lot of questions to answer. They’ll have to think hard about what insights they need from their data, which algorithms that will reveal those insights, and the processes and business models that will make use of them.
They’ll also have to think about how the analysis is done. IoT data is profoundly different from the event-based data obtained from customer browsing histories or order records. Sensors on machines may generate hundreds or thousands of readings every second, 24 hours a day. ABI estimates that of the 233 exabytes of data recorded by IoT devices in 2014, less than 10 per cent was transmitted across networks for offsite analysis.
With IoT data volumes expected to grow by an order of magnitude over the next decade, it is clear that industrial users simply won’t have the network bandwidth available to replicate the highly centralised analytics approach adopted by the big internet companies. Instead, they will need to push analytical capabilities into the edge of their networks.
In the smart control systems of the future, responsibility for analytics will be highly devolved. Intelligent sensors, machines and local control systems will do the bulk of the work, sharing insights across the network as required. Distributing analytics in this way has advantages beyond bandwidth reduction: allowing the low-latency decision-making needed in industrial control, simplifying security considerations and maximizing availability.
Edge analytics will have profound implications for equipment selection, however. It will no longer be enough to ensure that devices are sufficiently well-connected. They’ll also need enough innate intelligence to fulfil their demanding new roles.
For more information visit www.capula.co.uk
Source: ABI Research: Competitive Edge from Edge Intelligence. IoT Analytics Today and in 2020, May 2015