Part of the promise of massive databases is a transformation of process analytics, first to predictive modeling and ultimately enabling adaptive change. That transformation likely is years away, however. It will require many more data scientists than are currently available, says Light, and higher education is just beginning to create the data science curricula needed to fill the skills gap.
A recent report by Harvard Business Review (HBR) Analytic Services decries the lack of business investment in Big Data development. Most Big Data projects are done on an ad hoc basis, the report notes, with fewer than one in five organizations pursuing them as part of a comprehensive strategy.
That may be more a reflection of most companies’ inability to invest millions in infrastructure development with long-term paybacks than a disinterest in the technology itself. Automation vendors’ poster child for IIoT investment in the food industry is Sugar Creek Packing Co., a pork processor that christened a new facility in Cambridge City, Ind., in 2015. Sugar Creek invested $6 million in networking hardware and software. The technology was a necessary foundation for a high-performance work team system that executive management introduced at the new facility. Eliminating workflow bottlenecks provided the return on investment.
Machine assets, not human assets, are the focus of most IIoT initiatives, however. For the process engineer, real-time data exchange between automated machines is the issue, with remote access through the internet to performance reports a residual benefit. The Big Data proxy for Google is the cloud or, possibly, OEMs in the role of Pretty Big Data.
Analyzing controls data from multiple machines that are identical or very similar can yield much more powerful information on machine condition and process performance than data from a single machine. If the database included hundreds of machines, regardless of ownership, in dozens of processing environments, it would be even more powerful.
The likelihood of that scenario, however, hovers around nil: machine performance data is jealously guarded. For years, OEMs and skid builders have provided remote diagnostic capabilities in advanced machine controls. If those diagnostics ran continuously in the background, the IIoT could be the conduit to vendor-supported alarms and analytics.
But the very idea of an internet portal to a plant’s Ethernet communications sends a shudder down the spine of IT. Instead of easier access from outside the plant, food manufacturers are placing more restrictions. “One word: firewalls,” summarizes Ola Wesstrom, senior industry manager-food and beverage for instrument supplier Endress+Hauser (www.us-endress.com), Greenwood, Ind. Expect more restrictions, not less, to allowing third parties to listen in on machine performance.
The Shadow knows
Data security is the obsession of IT and, to some extent, executive management. Engineers and operations personnel, on the other hand, are more receptive to IIoT collaboration.
In the Food Processing-ABB automation survey, plant operations professionals were twice as likely to view remote access to machine controls favorably as C-suite executives. They also had a much more favorable view of vendor access to controls data and the connection of field devices to a wireless network.
A possible workaround to security concerns is creation of a parallel controls network, a system of “shadow sensors,” in the words of Rob McGreevy, vice president-operations, information & asset management at Schneider Electric (software.schneider-electric.com/), Andover, Mass.
“Low-cost sensing and other technologies allow engineers to enhance performance monitoring in a fraction of the time and cost,” he explains.
“Shadow sensors costing $200 or $300 each could sit on top of the high-fidelity, deterministic controls needed for high-speed machines.”
A handful of these wireless devices would communicate via Bluetooth to the cloud and monitor machine condition, product quality and other factors. “It’s not that complicated and doesn’t require pulling wires and costly systems integration,” he adds.
Another avenue to better process control and improvement in yield and product quality passes through in-line inspection systems. The performance of upstream machinery is inferred in the products being inspected. While inspection equipment is only required to render a pass/fail decision, it often has the computing power to do much more.
An example is the software suite that Key Technology Inc. (www.key.net), Walla Walla, Wash., began embedding in high-speed optical sorters two years ago. The immediate benefits relate to the quality of raw materials, but there also is potential for improved process control and machine performance.
“Optical sorters could be looked at as digital information centers,” observes Marco Azzaretti, who oversees Key’s advanced inspection systems. Images of each item moving down the line are captured, and real-time processing of the captured data can be used to adjust the process.