On acoustic prediction models for the intrinsic parameters of rigid-frame porous media using Bayesian optimization

Abstract

Knowledge of several intrinsic material parameters is crucial to the modeling accuracy of sound propagation in porous media. This paper studies the performance of several popular acoustic prediction models used to invert key intrinsic parameters of rigid frame porous media such as characteristic lengths, permeabilities and median pore size. These prediction models along with acoustic surface impedance experimentally measured with a standard impedance tube setup are evaluated in an optimization process which is essentially a Bayesian estimator. The effect of the number of intrinsic parameters in the model, sample thickness and frequency range on the accuracy of the Bayesian estimator is studied. The quality of model predictions is characterized using quantitative measures such as mean, variance and the interdependence of the estimated parameters. This paper also discusses uncertainties of the intrinsic parameters and effect of these uncertainties on the accuracy of the acoustic parameter prediction.