Heuristic algorithm-enhanced inverse design of deep-subwavelength hybrid metamaterials for ultra-broadband underwater sound absorption

This study proposes a hybrid acoustic metamaterial (HAM) composed of four quasi-Helmholtz resonators (QHRs, with rubber liners and embedded necks) and one boundary-constrained viscoelastic material (BVM) arranged in parallel. It achieves ultra-broadband absorption through dual physical dissipation mechanisms: monopole resonance of QHRs (dominant in low frequencies) and compressional-to-shear wave conversion of the BVM (dominant in mid-to-high frequencies). A theoretical model is established to calculate its acoustic impedance and sound absorption coefficient, validated via finite element simulations with high consistency. To address the inefficiency of traditional parameter scanning and the multi-solution ambiguity in inverse design, an inverse design methodology based on a heuristic algorithm-enhanced multi-layer perceptron neural network is proposed. This method efficiently predicts optimal structural parameters from a target sound absorption spectrum. Our theoretical and numerical predictions show that the optimized HAM achieves a sound absorption coefficient above 0.8 across the entire 0.4–10 kHz frequency band. Remarkably, its thickness is only 1/71 of the wavelength at the lowest frequency (400 Hz), showcasing deep-subwavelength performance. The metamaterial also maintains effective broadband absorption under 45° oblique incidence and 2.5 MPa hydrostatic pressure, demonstrating significant acoustic robustness. This work confirms the advantage of combining multiple dissipation mechanisms in metamaterials for efficient broadband absorption and presents a viable inverse design strategy for optimizing multi-parameter acoustic systems.