Faculty profile

Shiyu Zhou (23K JPG)

Years ago, manufacturers also were inventors. Many of their goods were one-of-a-kind or first-of-a-kind products made in simple steps on a production line.

But in today's global marketplace, there are abundantly more variables, says Shiyu Zhou, the department's newest assistant professor. New, high-tech goods often mean complex, lengthy fabrication processes. And profitability boils down to which companies can make the highest-quality and most-appealing product faster and cheaper than the competition.

As a result, manufacturers still have a lot to learn, particularly about what's going on in their processes. "People put a lot of sensors into and get a lot of data from the manufacturing process," says Zhou, who joined IE this fall. "However, the technology to deal with this huge amount of data and transfer them into useful knowledge is inadequate at the current stage."

His research focuses on how to analyze this data, develop models of manufacturing processes, and determine "rules" based on the processes' characteristics and outcomes. But Zhou doesn't just approach data analysis with a statistician's eye. "Normally a statistician bases it purely on data — they don't consider the underlying physical law of the process," he says. "When I do the data analysis, I try to integrate the process knowledge."

Formerly an assistant scientist in the Department of Industrial and Operations Engineering at the University of Michigan, Zhou is working with his former colleagues on a project to model the dimensional variation of a multistage machining process. "At each stage, there will be a deviation, or an error, because we cannot machine the product perfectly," he says. "So when we transfer the product to the next station in the process, this error could in some situations propagate and affect adversely at the following station."

His group is monitoring the process and then building a model of an error's propagation along the process. "That model can link the process fault, such as a machining error or a fixture error, and the product quality," Zhou says. And based on the model, his group can apply advanced statistical analysis and estimate where the process is breaking down.

Similarly, he and colleagues are analyzing cycle-based signals — "portraits" of what happens in a process over time. Typical cycle-based signals include the forming force of a forging or stamping cycle, or the welding current of a spot-welding cycle. "Normally, each curve is very similar to another curve, but with certain variations," says Zhou, who will help analyze normal and abnormal cycle-based signals to learn more about the variations in a manufacturing process.

Zhou earned a master's degree in industrial and operations engineering in 2000 from the University of Michigan and a PhD in mechanical engineering also from Michigan. A member of the Institute of Industrial Engineers, the Institute for Operations Research and the Management Sciences, the Institute of Electrical and Electronics Engineers, and the American Society of Mechanical Engineering, he is eagerly establishing himself at UW-Madison. In spring, he will teach a class in statistical data modeling and analysis with real-world applications to manufacturing. "There is a lot of support for faculty starting up in research and teaching," he says of both the university and the department.

He and his wife, Yifan, who holds a master's degree in computer science, recently bought a house in Fitchburg, adjacent to Madison's southwest side. Tennis players and avid travelers, they now devote much of their free time to house and yard work.