Deep-Learning-Based Impedance Eduction Method for Acoustic Liners with Flow

Acoustic liners are specialized structures or surface materials designed to attenuate and control sound propagation. Accurate simulation of impedance measurement under different grazing flow conditions is essential for optimizing their performance. This paper presents an efficient impedance eduction method for a grazing incidence tube by using a mixed neural network model, which enables rapid and effective impedance eductions over a range of parametric values and working conditions. In a procedure based on the NASA grazing incidence tube benchmark experiment, the finite element method is applied to generate the dataset required for the model training process. The proposed mixed neural network model is demonstrated on the analytical results of several acoustic liners, including a single-degree-of-freedom liner, a CT57 liner, a porous material, and NASA experiment data. All the demonstrations highlight its potential and capability as a practical and accurate tool for impedance eduction and liner optimization.