AI and Symmetry Accelerate Design of Sound-Absorbing Fabrics

An international research team has developed an AI-assisted method to accelerate the design of woven fabrics with improved sound-absorbing performance. The work, published in Communications Materials, involved researchers from the Laboratory of Tribology and Dynamics of Systems at École Centrale de Lyon/CNRS, the Advanced Institute for Materials Research at Tohoku University, and RIKEN iTHEMS. The study addresses a major challenge in textile engineering: identifying over-and-under yarn patterns that optimize acoustic behavior. Even modest weave sizes, such as 10 by 10, generate more possible arrangements than can realistically be tested individually, while detailed acoustic simulations can take more than an hour for each design. To overcome this bottleneck, the researchers first generated weave patterns that met basic manufacturing and connectivity requirements, then grouped them using four types of repeating symmetry, including rotational and mirror symmetry. An AI system placed structurally similar patterns close together on a design map. After simulating a selected set of designs, the team trained a prediction model capable of evaluating candidate patterns and guiding further searches in less than a minute. Independent detailed simulations confirmed the model’s predictions within 5 percent for selected designs. The research also demonstrated that symmetry is not merely a visual property but a functional design parameter. Different symmetries distribute yarn crossings in distinct ways, altering airflow through microscopic gaps between yarns. Some patterns create longer, more winding airflow routes that increase the conversion of sound energy into heat through friction, while others produce straighter, more open paths. The size of the smallest repeating unit also influences the range of acoustic behaviors available. Lead author Thanasak Wanglomklang said the approach combines symmetry, physics-based simulation, and machine learning to explore woven structures that would be impractical to evaluate through trial and error. He emphasized that the method does not replace physical simulation but uses it more strategically to identify promising, manufacturable designs. The team plans to manufacture selected 3D-woven prototypes and measure their sound absorption experimentally. Future work will extend the method to multilayer woven materials, larger repeating structures, and variations introduced by real-world manufacturing processes.