91����

For more than 10 years, 91���� Professor Paul R. Chiarot has worked on perfecting a low-cost manufacturing technique called electrospray deposition to make microscopically thin polymer films.

The process could have a wide range of applications, from electronics manufacturing to healthcare. Imagine a special coating that could eliminate corrosion in mobile phone components or prevent dangerous bacterial buildup on medical implants.

However, one obstacle limiting the adoption of electrospray is making sure it is consistently applied to the desired specifications, because it can be hard to control the characteristics for a film thinner than a human hair. Studying the process at a microscopic scale is also difficult.

“The role of electric charge in the process is really important, and that is not something you can physically see — you kind of infer it based on how it interacts with its neighbors or how it interacts in its environment,” said Chiarot, the chair of the Department of Mechanical Engineering at the Thomas J. Watson College of Engineering and Applied Science.

“With electrospray, the material it spits out has a high electric charge, and that charge accumulates on the surface as the material is depositing. Measuring the accumulation and decay of that charge is very difficult to do experimentally.”

A will bring together faculty from 91���� and the University at Buffalo to integrate experiments, computational modeling and artificial intelligence/machine learning methods to develop a comprehensive framework for the electrospray deposition process.

Co-investigators with Chiarot are Associate Professor Daehan Won and Professor Sangwon Yoon from Watson College’s School of Systems Science and Industrial Engineering; and Buffalo Associate Professor Xin Yong and Assistant Professor Yu “Chelsea” Jin, both former Watson faculty members.

For Won — who brings his AI skills to the project — the purpose of the research is clear.

“If we have a better understanding about the underlying physics in electrospray deposition, can we also control the parameters?” he said. “And what are the optimal parameters to get the desired level of quality we want? It’s a very complex problem, and it’s very hard to control.”

Current use of electrospray deposition involves what Chiarot jokingly calls a “shake-and-bake process” to narrow down the results until an optimal product is produced. That kind of research costs both time and money.

“Right now, we have to do some trial and error to get the ideal characteristics for the film,” he said. “We’d like to use the AI tools and modeling to know exactly how we need to operate our process to achieve those desirable characteristics.”

Because experimental observation is not easy, one challenge that the research team will need to overcome is having enough data for AI to create simulations that reflect real outcomes. If successful, though, the models could be used for more than electrospray.

“The reason why we work with AI is to minimize human labor and save time but to get a high-quality AI solution, we require a large data set,” Won said. “My main concern is how to get good outcomes with a limited amount of the data and how to leverage physics principles to get results that should be very close to experimental observations.”

As part of the project, researchers will work with the Alliance for Manufacturing and Technologies, a nonprofit based in New York’s Southern Tier that helps manufacturers to overcome the challenges of today’s competitive economy. Those goals align with national initiatives since the COVID-19 pandemic showed vulnerabilities in international supply chains.

“While we are revitalizing the U.S. manufacturing industry, one of the keys is smart manufacturing, because it will help to reduce unnecessary labor and increase efficiency,” Won said. “With labor costs here compared to other countries like China or India, that is one way we could make it work.”

Chiarot sees the electrospray research as the kind of collaboration that Watson College is known for doing, and he expects it will lead to similar projects in the future. A multidisciplinary approach allows experts in different fields to approach problems from many angles and spark ideas off each other.

“I have no experience in AI and machine learning, so Daehan gets to teach me about all of that. He’ll have to be patient with what might be one of his worst students!” Chiarot said with a laugh.