Failing equipment, whether a dead car battery or paper jam, can ruin your day and alter your plans. Or a manufacturer’s production line fails for an unknown reason, causing delays in production, shipping and sales.
Currently, there is very little to no warning given to address these problematic issues before they happen. However, Kangwon Seo, an assistant professor in industrial and manufacturing systems engineering with a joint appointment in statistics, is researching how machines can use predictive maintenance themselves to detect when they are going to fail.
“Predictive maintenance is a big application research topic of the fourth industrial revolution and machine learning,” Seo said. “Basically, instead of making a maintenance action every month, people try to use machine learning to predict future failure so that they can make a maintenance action as needed.”
Seo and his doctoral student, Wonjae Lee, recently published this article detailing how a data-driven predictive maintenance exercise led to their proposed algorithms showing better performance for detecting manufacturing equipment failure than other state-of-the-art techniques.
“This research started from a data challenge in 2019 from the IISE (Institute of Industrial and Systems Engineers),” Seo said. “We were one of the three finalists in that competition.”
The data provided for the competition was from an unnamed paper manufacturer.
If a paper manufacturer used the developed data algorithm for predictive maintenance, consumers may find cheaper prices on their products.
“Predictive maintenance helps reduce the consumers’ costs in terms of maintenance activity for a manufacturer,” Seo said. “This will reduce the costs that flow through the price pipeline.”
Predictive maintenance will also help reduce overall costs for manufacturers as well.
“A manufacturer will be able to reduce their effort for maintenance activity,” Seo said. “Traditional maintenance activity is centered on period-based or even reactive maintenance, but that can have a lot of costs and take a lot of time to investigate what is wrong.”
Seo thinks research into predictive maintenance is a perfect match with his interests in being an industrial engineer as well as applied statistics.
“When I was a PhD student, my research was about reliability,” Seo said. “But it was more of a traditional statistical model-based research. I think predictive maintenance research is a progression for my career.”
Seo earned his master’s and doctoral degrees from Arizona State University. He received his bachelor’s degree from Hongik University in South Korea. Applied statistics, reliability engineering, and design of experiments are his technical focuses.
Learn more about the Department of Industrial and Manufacturing Systems Engineering.