Machine learning model automates wearable technology design

Thanks. Listen to this article using the player above. ✖

Defying engineering challenges in record time, researchers at the University of Maryland have developed a machine learning model that eliminates difficulties in material design to produce green technologies used in wearable heaters.

Resistance associated with trial-and-error experimental processes often delays innovations. To solve this challenge, Po-Yen Chen, assistant professor in the Department of Chemical and Biomolecular Engineering, proposed an accelerated method for creating materials used in wearable heating applications. His model, published Saturday in Nature Communicationscould automate design processes by leveraging machine learning and collaborative robotics.

Similar to water-based gels but made with air, aerogels are lightweight, porous materials used in thermal insulation and wearable technologies for their mechanical strength and flexibility. But despite its seemingly simplistic nature, its assembly line is quite complex. Researchers rely on endless experiments and experience-based approaches to explore the vast design space and engineer these materials.

To solve these problems, Chen’s team combined robotics, machine learning algorithms, and materials science expertise to enable the design of airgel with programmable mechanical and electrical properties – breaking scientific barriers at full speed. Its prediction model is built to generate sustainable products with a 95% accuracy rate.

“Materials scientists often struggle to adopt machine learning design due to a scarcity of high-quality experimental data. Our workflow, which combines robotics and machine learning, not only improves data quality and collection rates, but also helps researchers navigate the complex design space,” said Chen.

The resulting strong and flexible aerogels were made using conductive titanium nanosheets as well as natural components such as cellulose; an organic compound found in plant cells and gelatin; a protein derived from collagen found in animal tissues and bones.

But that’s not where it ends. This tool can be expanded to address other applications in airgel design. Green technologies used for oil spill cleanup, sustainable energy storage and thermal energy products – such as insulated windows – could become more affordable sooner than expected with this fast-paced assembly process.

“The combination of these approaches is putting us at the frontier of designing materials with complex and customizable properties. We anticipate leveraging this new production-scale platform to design aerogels with unique mechanical, thermal and electrical properties for harsh work environments,” said Eleonora Tubaldi, assistant professor of mechanical engineering and study collaborator.

Looking ahead, Chen’s group will conduct studies to understand the microstructures responsible for airgel’s flexibility and strength properties. Their work was supported by the institutional Grand Challenges Team Grant for the Programmable Design of Natural Plastic Substitutes, jointly awarded to mechanical engineering professor Teng Li.

Reference: Shrestha S, Barvenik KJ, Chen T, et al. Machine intelligence-accelerated design of conductive MXene aerogels with programmable properties. National Communication. 2024;15(1):4685. doi: 10.1038/s41467-024-49011-8

This article has been republished from the following materials. Note: Material may have been edited for length and content. For more information, contact the source cited. Our press release publishing policy can be accessed here.