Fabrics can be not only soft and strong but also influence the propagation of sound. However, selecting the right weave is difficult: even for a 10 by 10 pattern there are more possible combinations than can be checked by hand, and detailed acoustic modeling takes more than an hour.
Researchers from the Laboratory of Tribology and System Dynamics at the Ecole Centrale de Lyon/CNRS, the WPI-AIMR Institute for Advanced Materials at Tohoku University and RIKEN iTHEMS have proposed an accelerated method based on artificial intelligence. The work was published in the journal Communications Materials.
First, the scientists generated weave variants that satisfy basic manufacturing and connectivity requirements. They then grouped them by four types of repeating symmetry, including rotational and mirror symmetry. The AI system organized the patterns on a design map, placing structurally similar patterns closer to each other. The team modeled part of the structures and trained a predictive model on these data. After training, the model evaluated variants and guided the computer search in less than a minute.
The selected solutions were checked with independent detailed modeling: the predicted sound absorption matched the simulation results to within 5%.
The study also showed how symmetry affects sound absorption. Different types of symmetry distribute thread intersections differently, changing the movement of air through the gaps. Some patterns create more tortuous paths, increasing friction and heating, while others leave more direct and open channels. The size of the smallest repeating element also affects the available range of air flows and acoustic properties. Thus, symmetry can serve as a design parameter, not just a visual property.
"By combining symmetry, physical modeling and machine learning, we can explore woven structures that are practically impossible to enumerate by hand," says lead author Thanakorn Wanglomklang. "The method does not replace physics but uses it more strategically to find promising and manufacturable designs."
The team plans to fabricate the selected 3D-woven prototypes and experimentally test their sound absorption, as well as extend the method to multilayer materials, larger repeating structures and variations that arise in manufacturing.