Everyday items like car tires, plastic bags, and foam cushions come from materials called polymers that take years to develop and test. Researchers at Carnegie Mellon University and the University of North Carolina at Chapel Hill have developed a new approach to creating rubber-like materials more quickly and more quickly by combining artificial intelligence with human expertise.
Typically, when researchers use a material more, the flexibility decreases, but flexible materials tend to be weaker. To fix this issue, the team created a machine learning model that works in conjunction with human chemists. Machine Learning – a subset of AI research – involves teaching artificial intelligence to perform specific tasks. In one experiment, researchers worked with AI tools to create powerful, flexible polymers.
“There are so many applications in polymers: construction, car parts, footwear, molded parts, coatings.” olexandr (oles) isayev,(Opens in a new window) Professors Carl and Amy Jones of Interdisciplinary Science. “Whenever you create one for a particular application, certain properties are required and they are usually not able to withstand force and expand simultaneously. These new materials have excellent properties. They can do both.”
Groups enter the properties required for the polymer into the design tool. The model then proposed a series of experiments carried out by UNC-Chapel Hill chemists using automated science tools.
The researchers tested the produced materials and provided feedback to the model, allowing adjustments to be made.
“AI systems propose experiments, and after the experiment is done, they measure the properties and iterate,” says Isayev. “It dynamically adjusts to help the machine navigate and find materials with the desired properties.”
Frank Leibfarth, professor of chemistry at UNC-Chapel Hill, said working this new way is a breath of fresh air.
“Our human-focused approach was not only taking instructions, but also interacting with models,” says Leibfarth. “This allowed us to combine the best aspects of the human and machine guided process to lead us to the most optimal solution.”
Leibfarth also said he was excited by the potential use of the polymer.
“These materials can be used in running shoes, medical devices such as 3D printed dental implants, and durable parts of your car,” says Leibfarth.
“They're a great way to get to know you,” said Dylan Ansistine, a former postdoctoral researcher at Carnegie Mellons. Chemical Bureau(Opens in a new window)Assistant Professor of Chemical Engineering and Materials Science at Michigan State University. “It's clear that expert experimental chemists and expert computational chemists will be involved using the best data science tools possible. We really teased what that relationship looked like.”
Machine learning models also saved researchers a considerable amount of time and money by eliminating non-working methods and chemicals. Researchers have made the program open source, so labs have access to this tool. If adopted by other labs, this tool reduces the cost and time required for other discoveries.
This approach could accelerate the development of advanced materials in medical devices, footwear and electronic devices. By combining AI prediction with human expertise, researchers hope to be able to more effectively solve complex material challenges.
Anstine, Isayev and Leibfarth have been published Angewandte Chemie's “Design of Tough 3D Printable Elastomers with In-Loop Reinforcement Learning”(Opens in a new window) with Carnegie Mellon graduate students Philip Gusev and Philip Nikitin, as well as researchers Johann Rapp, Kelly Yun, Meredith Borden and Bitol Batt. Their work was funded by the Air Force Research Institute and the National Science Foundation.
