基于人工神经网络的共振吸声超材料声学性能快速预测及结构优化设计.

A sound performance prediction method based on the artificial neural network (ANN) was proposed to meet the requirements of rapid prediction and optimization design of resonant sound-absorbing metamaterials. Firstly, a theoretical model was established for multilayer perforated resonant sound-absorbing metamaterials (MPRSMs) composed of microperforated panels and Helmholtz resonators, which was then verified through simulation and experiments; subsequently, a dataset was generated with the theoretical model, and in turn an ANN model was constructed by means of the back propagation (BP) neural network to build the mapping relationship between structural parameters and acoustic performances; afterwards, the trained ANN model was combined with the genetic algorithm to optimize the acoustic performance of the MPRSMs. The results show that, the trained ANN model can accurately predict the sound absorption performance of the MPRSMs, and the prediction efficiency improves by more than 50% compared to the theoretical model; the combination of the ANN model and the optimization algorithm can not only improve the optimization efficiency, but bring good low-frequency broadband sound absorption performance of the optimized structure. The ANN provides convenience for large-scale structural performance prediction calculations and has broad application prospects in structural design and optimization of metamaterials. [ABSTRACT FROM AUTHOR]