Unplanned downtime costs industrial manufacturers an estimated $50 billion per year. Nowhere is the problem more urgent than in the corrugated business, where producers have been pushing equipment – both old and new – to its limits in order to meet surging demand.

That’s why Gokul Gopakumar, Senior Director of Development and Data Science at Glen Arm Maryland based Helios IIoT, is betting that their new failure prediction capabilities could save box manufacturers as much as $122,000 per machine per year.

Helios is an OEM-agnostic machine learning IIoT platform that provides insight into corrugated machine performance. This year at Corrugated Week, Gopakumar presented the company’s new failure prediction capabilities – software that allows manufacturers to not only monitor machine performance in real time, but to view an ongoing assessment of the likelihood of failure or interruption within a 30-minute window.

This new machine learning algorithm provides operators with actionable insight into the well-being of their equipment. This data can help them forecast future problems and reroute production to prioritize critical projects and avoid delays.

“By analyzing thousands of data points per second, Helios is able to see patterns in machine behavior and performance that are otherwise undetectable to even the most experienced maintenance professionals,” Gopakumar explained. “These behaviors, which may take the form of idiosyncrasies, vibrations, noises or shudders, are monitored and analyzed by machine learning to create patterns and draw conclusions about potential downtime with high accuracy and very short lead time.”

During the initial study, Helios reported that its technology successfully predicted 74 percent of machine breakdowns. But they stress that that’s only a starting point. “The platform gets smarter, faster, and more accurate every day as it takes in more data,” said Gopakumar. “These progressive solutions generate increasingly accurate downtime predictions and, in turn, improve uptime and ROI.”

Brian Kentopp, Helios’s Vice President of Sales, noted that a generation of machine technicians are retiring, carrying decades of knowledge and experience about the day-to-day operations of converting equipment out the door. Much of that knowledge has never been recorded or systematized – at best, it resided in the instincts and gut feelings of experienced machine operators.

With the labor market as tight as it is, it’s difficult for box plants to find new technicians with the qualifications to step in. It could be years before new machine operators get the experience to develop the right instincts.

“So many corrugated plants rely on human intuition and experience to drive their decisions,” added Kentopp. “But the information produced by corrugated machines is often very challenging for humans to understand with the naked eye. If humans are using this data at all, they are usually doing so retroactively to try to diagnose the root cause of a problem after it has already occurred. But machine learning has the power to turbocharge each of these same data points and to turn them into actionable intelligence that can be used proactively for decision making, or even to predict and prevent problems before they happen.”

This kind of machine-generated prediction tech used to be purely theoretical. But Helios reports these machine learning models are now fully operational and live on the platform, ready for customers who have subscribed to the company’s full suite of features. Using current sensor information along with historical downtime data, the Helios platform provides an up-to-date picture of operating equipment and actionable information about future functioning so operators can make proactive decisions.

“The algorithms are no longer simply providing diagnostic information about past machine performance,” said Gopakumar. “They’re providing a picture of ongoing and forecasted future performance.”

This gives operators the chance to make decisions in advance before an interruption occurs. If a machine is operating at high risk of failure, operators can stop it in advance to investigate the cause before a disruption. Managers can also reroute work ahead of time for high priority projects or alter production schedules accordingly to avoid a break-down.

“There is also more to machine learning than simply detecting anomalies,” said Gopakumar. “All usage data that is recorded and reported from each machine can be not only aggregated and seen in real-time, but can also be analyzed historically, allowing operators and supervisors to make data-driven decisions regarding quality and fully optimized operations.”

Typical maintenance schedules require servicing on fixed periods, regardless of actual usage, but they don’t help operators understand what’s actually happening inside the equipment. With Helios’ sensors collecting real-time information, box plants can determine how much a machine actually ran since it was last maintained: what was the load, what was the burn rate, etc. And by monitoring new equipment right from the start, operators can track degradation paths over time, monitoring how the machine begins to vary from the OEM specifications. 

In order to generate its predictive models, Helios needs two distinct sets of data. The first is the ongoing raw information provided by the sensors Helios installs on the machines to monitor ongoing operations. The second is historical downtime data – listing the start and end times of previous planned and unplanned disruptions – along with explanations and descriptions of the problems that were encountered.

Initial studies have shown that it takes about eight to twelve weeks’ worth of data collection for the machine learning mechanism to function optimally. Helios’ data scientists work with customers to input this information in order to get the predictive algorithm fully operational. Once the systems are in place, the predictive power of Helios’ machine learning gets more refined over time.

“The corrugated industry has been booming, and new machines are going online every day,” said Gopakumar. “The rise of e-commerce has brought rapid growth and increased demand means it’s more important than ever to make sure equipment is operating as efficiently as possible. It’s critical to harness the power of information. Machine learning is revolutionizing how box plants understand, predict, and maintain their corrugated converting equipment, and Helios’ new failure prediction capabilities place it at the forefront of the industry. The sooner you adopt this powerful new set of tools for understanding and maintaining your equipment, the greater the rewards you’ll reap down the road.”

Visit gohelios.us for more information.

Spread the love