Theoretical Analysis, Neural Network-Based Inverse Design, and Experimental Verification of Multilayer Thin-Plate...

Area of Science:

• Acoustic metamaterials
• Vibrational mechanics
• Computational materials science

Background:

• Acoustic metamaterials leverage subwavelength structures for advanced sound insulation.
• Thin-plate acoustic metamaterials, formed by resonant mass blocks on thin plates, excel in low-to-mid frequencies.
• Multilayer designs enhance performance but introduce significant design complexity.

Purpose of the Study:

• To develop a systematic, neural network-assisted inverse design framework for multilayer composite thin-plate acoustic metamaterials.
• To overcome the inefficiency of traditional design methods for complex acoustic metamaterials.
• To provide an efficient solution for low-frequency sound insulation.

Main Methods:

• Established an analytical model for thin-plate metamaterials with multiple attached masses using point matching and modal superposition.
• Constructed multilayer composite unit cells and generated a large dataset (30,000 samples) via numerical simulations.
• Developed and trained forward prediction and inverse design neural networks.

Main Results:

• The analytical model was validated by finite element simulations.
• The forward prediction network achieved a test error of 1.06%.
• The inverse design network converged to an error of 2.27%, with structures validated by experiments.

Conclusions:

• A novel theoretical and computational framework for inverse design of thin-plate acoustic metamaterials was successfully established.
• The neural network approach efficiently addresses design complexity and parameter non-uniqueness.
• The proposed method offers a practical solution for designing high-performance low-frequency sound insulation.