Highly efficient design of sound attenuation and ventilated metamaterial via machine learning and genetic algorithm

• A novel ventilated acoustic metamaterial with high parametric tunability was developed. • The multi-unit cell configuration achieves exceptional customizable performance. • Integrated machine learning with GA enables highly efficient design optimization for target-specific configurations. • Two distinct ventilated metamaterials were designed and experimentally validated. Ventilated acoustic metamaterials possess dual functionalities of ventilation and noise suppressing, which meets the needs of many scenarios. This work proposes a unit cell with high parametric tunability that allows flexible control of transmission loss peak quantities, facilitating customized ventilated metamaterials through cascaded unit cell configurations. To achieve the desired acoustic performance within specific frequency ranges, a partitioned optimization strategy was employed, targeting individual unit cells for distinct frequency sub-bands. By integrating machine learning and genetic algorithms, the geometrical parameters of the unit cell can be rapidly optimized to meet the acoustic performance target. In this work, we designed dual-band ventilated metamaterials and broadband ventilated metamaterials to demonstrated the framework’s effectiveness. Both ventilated metamaterials were investigated via finite element method and experiments. The transmission loss performances of experiments are perfect agreement with simulations. Machine learning model surrogate approach bypasses the repetitive process of modeling and finite element analysis, addressing the time-consuming and labor-intensive limitations of traditional trial–error and exhaustive methods, thereby establishing an accelerated pathway for ventilated metamaterial design.