Lowering the sound transmission loss of impedance-matching structures: Data-driven optimization assisted with a priori knowledge

Abstract

Impedance-matching structures (IMSs) with low sound transmission loss (TL) are of great importance in engineering applications. Due to the limitations of the traditional research methods, the designs of the IMSs were usually empirically guided, e.g., following impedance-gradient. In this work, we propose an approach introducing a priori knowledge into data-driven structural optimization of IMSs at low frequencies (0–10 kHz), reaching a 31.78% decrease of average TL compared to conventional impedance-gradient designs. Contradicting the long-standing wisdom (impedance-gradient), an impedance-jumping layer appears in the optimized structure, which is revealed to lower TL at frequencies 0–5 kHz but at a cost of TL increasing at higher frequencies (5–10 kHz). Then the introduction of inhomogeneity alleviates this degeneration and enhances the overall performance of the IMSs. The methodology and results in this work motivate the novel designs of high-performance acoustic metamaterials for waterborne applications.

Keywords

• Impedance jumping
• Sound transmission loss
• Acoustic metamaterials
• Digital materials
• Machine learning