Surrogate-guided parametric inverse design of periodic labyrinthine ventilated acoustic metamaterialsa).

Abstract

Labyrinthine ventilated acoustic metamaterials (LVAMs) enable simultaneous sound insulation and air ventilation but remain difficult to design due to their geometric complexity. This study develops a surrogate-guided inverse design framework for LVAMs with target phononic bandgaps. A parametric LVAM model is constructed, and a large finite element dataset is generated to link geometry with band structure. The framework integrates a finite-element-trained forward predictor with an inverse generator, allowing feasible designs to be directly generated from target bandgaps without repeated simulations during inverse training. Numerical and experimental results show high design accuracy, with R2=0.9975 and 0.9908 for the lower and upper bandgap bounds, respectively, surpassing a baseline inverse model. The results demonstrate that the proposed framework provides an efficient surrogate-guided approach for the parametric inverse design of complex acoustic metamaterials.