Skip to main content
March 14, 2017

“Smart” equipment will manage its own maintenance

Written By: Silke Schmidt

The number of devices connected to the internet—ranging from “smart” home appliances to health and fitness sensors—is predicted to increase from about 25 billion today to one trillion by 2025. But making sense of this highly interconnected data-rich world, which is enabled by “Internet of Things” (IoT) technology, requires much more than simply installing a few sensors.

“The IoT technology has little value by itself,” says Kaibo Liu, an assistant professor of industrial and systems engineering at the University of Wisconsin-Madison. “You create value by analyzing and interpreting the data in the context in which they are generated.”

In order to advance these analytic capabilities, the U.S. Department of Defense Office of Naval Research recently funded Liu’s research on IoT-enabled condition-based maintenance, diagnosis and prognostics for Navy equipment.

The U.S. Navy hopes that a growing number of sensors—measuring conditions like temperature, pressure and turbine vibration—will better predict the degradation of ships and other equipment over time. With continuous data collection, analysis and visualization, the Navy can design better maintenance schedules to reduce costs and prevent catastrophic failures, increasing the safety of the sailors and staff who operate or rely on the equipment.

But since the research is fundamental, the applications for the statistical methods Liu will develop don’t end with the Navy. The broader goal of the four-year project, which is supported by a total of $275,000, is to devise novel methods for analyzing any continuously generated data from multiple sensors in real time. This will address shortcomings of existing methods that have focused on a single sensor, or on data collected at a limited number of discrete time points.

“Think of your car maintenance schedule,” Liu says. “The common recommendation to get a check-up and oil change every 3,000 miles or every three months could become much more precise if road conditions, driving style and other information about the specific car, driver and external environment were taken into account. This can be accomplished by analyzing the data from a variety of sensors installed on the car.”