Kaibo Liu, an associate professor of industrial and systems engineering at the University of Wisconsin-Madison, will use two new grants to apply a machine learning technique called transfer learning to nuclear reactor safety and maintenance and industrial manufacturing production systems.
Transfer learning involves analyzing a given system or sample and then leveraging those insights to better model and analyze similar systems or samples, even when facing shortages in data or an abundance of variables.
In one project, Liu will use a $400,000 grant from the United States Department of Energy to apply the technique to better model void swelling in nuclear reactor steels, a key maintenance and safety issue in the industry.
The other grant, for $135,463 from industry partner 3M, funds a second year of the collaboration between Liu’s lab and the company, with a shift in focus to transfer learning. Liu notes that applying the technique to process control in production systems is a relatively unexplored area.