Other Voices: Cognitive manufacturing and thinking inside the box

Editor’s note: The following column by Patrick Murphy, partner and practice leader, cognitive manufacturing, IBM, is part of Modern’s Other Voices column, a series featuring ideas, opinions and insights from end-users, analysts, systems integrators and OEMs. Click here to learn about submitting a column for consideration.

————-

The average factory produces more than a terabyte’s worth of information every day, but 99 percent of this data isn’t being analyzed. The result is a closed loop — a manufacturing box that contains information but can’t effectively leverage it to streamline production. As the shift to Industry 4.0 picks up speed, enterprises are looking for ways to empower process visualization and lay the foundation for cognitive manufacturing. How do they think inside the box?

Can You Hear Me Now?
While companies recognize the critical role of cognitive-driven manufacturing, most of them don’t have the necessary infrastructure and oversight to leverage cognitive potential.

Picture a line of people passing something from hand to hand. They naturally form a kind of cognitive unit, able to communicate relevant information about the size or weight of the object they’re holding. Put them in a dark room and ask them to wear earplugs, though, and the unit falls apart.

This is the current state of the Internet of Things (IoT) in the manufacturing sector: Devices are capable of moving information, but they’re unable to make decisions based on what they know.

The Quiet Revolution
IoT isn’t a new concept in manufacturing. Firms have been leveraging connected devices for decades. Many companies still use programmable logic controllers and distributed control systems across Ethernet or dedicated protocols to streamline production operations. But devices have quietly been getting smarter. They’re now able to communicate across networks, process and store more data, run local algorithms, and send relevant data to specific end-points for processing or further analysis.

First-generation connected light bulbs operated much like these original IoT networks: They were capable of following instructions from central network hubs but were effectively dumb unless directly controlled. Now it’s possible to offload critical workloads and create conditional lighting schedules linked to geographical locations, time of day, ingested schedule model and now AI learning.
But while reducing network overhead sets the stage for cognitive manufacturing, it alone isn’t enough to enable it.

Smartphones at Scale
Graphics processing units have emerged as front-runners of industry efforts to deliver computing power at scale. The most obvious example? Smartphones — powerful, flexible devices significantly smaller than their desktop and laptop predecessors. Remove the traditional telephone concept while leveraging all of the connective capabilities of the device, and you have the foundation of IoT solutions for manufacturing. Devices that use cameras can observe production lines in real time; microphone-enabled devices can detect sound changes that may indicate potential asset failure points. And, by creating custom models in an “app” accessed through the device, you now have a highly-integrated and powerful edge device. As part of a larger system, such devices can trigger alerts and generate work orders with detailed problem descriptions for human technicians.

This is cognitive manufacturing: the combination of low network overhead and on-site action. At scale, it empowers companies to be proactive by eliminating the need for redundant systems.

Takt-ical Advantage
Industry 4.0 — the manufacturing side of digital transformation — focuses on five key goals: connection, collection, visualization, analysis and optimization. The first two are possible with existing IoT networks; the next pair are feasible thanks to GPU-driven smart devices. But what about optimization? What does this look like in practice for manufacturing firms?

Consider takt time. The phrase, borrowed from German by way of Japanese, denotes the average time between the start of production of one unit to the next, determined by existing customer demand. For many organizations, existing paper processes force takt time to be assessed at the end of stage or product line, hindering organizations’ adaptability. By introducing IoT devices capable of edge computing and onboard processing, it’s possible to measure quality parameters and time-to-completion at any workstation. From this data, organizations can glean actionable insight, effectively allowing inside- and outside-the-box thinking.

Of Pairs and Platforms
To maximize the impact of IoT solutions for manufacturing, companies need more than smart devices capable of acting autonomously. They need platforms capable of creating governance and management frameworks for IoT networks and combining data from IoT devices to optimize production processes at scale. Solutions such IBM’s Watson IoT Platform are on the leading edge of this platform development, helping companies combine the abilities of generic IoT devices to deliver business-specific processes that optimize production end to end.

What’s next? Digital twins: virtual versions of products and processes created using data that allow companies to manipulate and test process improvements before moving them to live production lines. While this technology shows promise, it remains in adolescence — and it’s predicated on a foundation of comprehensive cognitive IoT.

Manufacturing processes no longer exist in isolation. New cognitive solutions and IoT platforms can help enterprises think inside the box.