Kolmogorov-Arnold Network-Aided Design of Conch Shell-Inspired Acoustic Metamaterial for Rail Noise Mitigation

Abstract

Urban transportation noise has become a critical challenge for modern cities striving to maintain an acceptable acoustic environment and ensure public well- being. Traditional noise mitigation approaches, such as bulky barriers and thick soundproofing panels, often struggle to balance broadband and high-efficiency sound absorption with lightweight, thin-layer configurations. Acoustic metamaterials (AMMs) offer a promising alternative, enabling high-efficiency sound absorption across broad frequency ranges with reduced weight and thickness. In addition, incorporating biomimicry into the design of AMMs allows us to draw inspiration from the natural structures like conch shells with remarkable sound-dampening properties due to their unique geometries and material compositions. By mimicking these natural designs, we can create AMMs that not only enhance sound absorption but also maintain lightweight and compact configurations. However, the bio-mimic AMM design is often complex, computationally expensive, and requires intricate geometric layouts and multi- parameter optimization. In this study, we propose a novel design framework to address the design challenges associated with conch shell-inspired AMMs by combining machine learning and optimization algorithms. Our framework uses Kolmogorov-Arnold Networks (KANs) to model and predict the acoustic absorption performance based on various geometric and material parameters, specifically targeting the 500–2000 Hz frequency range, which encompasses many of the most disruptive frequencies in rail transport. Particle swarm optimization (PSO) then guides the design towards configurations that achieve average absorption coefficients above 0.80, while minimizing simulation complexity and computational cost. Compared to conventional noise control solutions, this framework significantly reduces volume and complexity without compromising acoustic absorption, making it more viable and cost-effective for train noise abatement. Moreover, by shortening design cycles and reducing iteration costs, our approach can accelerate the development of next-generation noise mitigation acoustic materials, holding a significant promise for creating quieter, more sustainable, and more efficient rail systems.