Inverse design of acoustic coatings using Bayesian inference

Abstract

A Bayesian approach for the inverse design of acoustic coatings for underwater noise control is presented. The coating model comprises a periodic arrangement of voids embedded in a soft viscoelastic material. The viscoelastic material is attached to a steel backing with water on its incidence side and air on its transmission side. The acoustic performance of the coating is strongly dependent on the geometric properties of the voided inclusions as well as the number of layers of voids in the direction of sound propagation. The geometric optimization process using Bayesian inference proceeds inversely from a target absorption coefficient spectrum to the required number of voided layers, the geometric design parameters in each layer, and the distance between each layer. The Bayesian design process demonstrates that broadband sound absorption can be achieved using a multilayered coating with a gradient change in both void diameter and distance between each layer. Optimized designs using the Bayesian approach are validated against results obtained numerically as well as experimental results from the literature.