A data-driven framework for predicting the acoustic performance of metamaterial arrays

Acoustic metamaterials, such as ventilated sound barriers or Helmholtz resonators, offer promising solutions for broadband noise reduction in architectural and environmental applications. Research has shown that arranging metamaterial units in arrays with varying geometric features can enhance acoustic performance; however, these complex configurations also introduce significant challenges for predictive modeling due to intricate coupling effects between elements. While numerical and experimental studies have shown promise, scaling them to full arrays is computationally and financially expensive. Therefore, there is a need for a method that can efficiently predict and interpret the acoustic behavior of metamaterial arrays using limited data. This study proposes a data-driven framework to estimate the acoustic performance of metamaterial panels by accounting for spatial relationships and inter-unit interactions. The approach integrates existing surrogate models for individual unit behavior with a limited set of high-fidelity simulation data from full arrays to train an efficient and scalable predictor of array-level transmission loss. The goal is to enable rapid design iteration and accurate performance estimation for complex metamaterial geometries in noise control applications.