Physics-Aware Machine-Learning-Driven Inverse Design of Broadband Ultra-Open Acoustic Metamaterials.

Ventilated acoustic silencers combining strong sound attenuation with high ventilation are pivotal for advanced noise control. However, balancing attenuation, bandwidth, openness, and thickness remains a high-dimensional challenge. Here, we present a physics-aware machine-learning-driven inverse design framework for broadband ultra-open acoustic silencers (UAS). Based on Green's function parameterization, the design space is physically decoupled into spectral and radial space, ensuring physical interpretability while reducing complexity. A two-stage coarse-to-fine surrogate model captures both broadband envelopes and sharp resonant features. Coupled with a population-based, hybrid-objective parallel (PHP) inverse strategy, our framework enables rapid exploration of non-convex landscapes, identifying hundreds of optimized candidates within seconds each. Crucially, this framework reveals the intrinsic relationship among thickness, ventilation, and spectral response, and uncovers hidden linear design rules that serve as geometric proxies for impedance-matching. We experimentally validate multiple architectures: UAS-2 represents the monolithic, single-mode design with high ventilation, whereas UAS-3 exploits multi-mode interference for improved spectral performance. To circumvent the trade-off ceiling of single-unit resonators, a parallel-composite architecture (UAS-4) is introduced to enhance performance. Results confirm a broadband bandwidth exceeding 830 Hz within 1000-2000 Hz, with ultra-thin profile (0.1-0.2λ, 4-8 cm) and 80% ventilation. This work establishes a data-driven paradigm for discovering design principles in functional metamaterials.