摘要
多孔材料作为吸声材料可降低噪音污染对人们生活、工作、学习的危害,如何设计吸声性能更加优异的多孔吸声材料成为研究者们日益关注的问题。而仿真模拟技术可以在理论上得到性能最优的多孔吸声材料的结构参数,有效地为设计性能优异的多孔吸声材料提供理论依据和减少不必要的实验过程。由于仿真模拟技术存在多样化,选择合适的仿真模拟技术预测多孔吸声材料吸声系数极其重要。近年来,各种仿真模拟技术被广泛应用于多孔吸声材料吸声系数的预测,但尚无各种仿真模拟技术预测效果优劣性的系统比较。本文主要探讨了四种非线性神经网络预测多孔吸声材料吸声系数的方法。就预测过程及结果来说,径向基函数神经网络所需的样本数量更少,预测精度也较高。优化算法在理论上可以解决其他神经网络所需样本数量大的问题,具体的做法是在建立神经网络模型之前,利用优化算法对少量的样本集进行处理得到最佳的训练样本和测试样本,然后代入神经网络模型中进行模拟预测。另外,探讨了三种经典理论模型经验公式计算多孔吸声材料吸声系数的方法,其中Johnson-Champoux-Allard模型误差较小。虽然Johnson-Champoux-Allard模型是一种五参数模型经验公式,但影响多孔吸声材料吸声系数的影响因素众多,不止五种。量纲分析是建立数学模型的重要方法之一,通过量纲化可以将复杂的数学物理问题转化为精确的数学模型公式。因此,可以通过量纲分析的方法建立所有影响因素与多孔吸声材料吸声系数之间的关系式,从而使通过模型公式计算得到的吸声系数误差更小。本文还探讨了二维和三维等效模型模拟仿真多孔吸声材料各项性能的方法。相比于二维模型,三维模型虽然建模过程复杂,但能够更加直观准确地对材料的各项性能进行模拟仿真,并且材料的三维建模数据可为后续的生产制备提供依据。本文简要介绍了吸声材料的分类及其原理,分别对三种方法中的九种仿真模拟技术进行了介绍,并总结了每种方法中性能最优的技术,以期为模拟仿真技术在多孔吸声材料吸声系数的预测提供参考。
Porous materials have been used as sound-absorbing materials to reduce the impact of noise pollution on people’s lives, work and studies,which raises a major research concern about how to design porous sound-absorbing materials with better sound absorption performance.Simulation technology can theoretically determine the structural parameters of porous sound-absorbing materials with optimal performance, effectively providing a theoretical foundation for designing porous sound-absorbing materials with excellent performance while reducing unnecessary experimental processes.Due to the variety of simulation techniques available, it is crucial to choose the appropriate one for predicting the sound absorption coefficient of porous acoustic materials. Recently, various simulation techniques have been used for predicting the sound absorption coefficient of porous acoustic materials, but no systematic evaluation of the benefits and drawbacks of each simulation technique’s prediction effect exists.In this review, four nonlinear neural networks to predict the sound absorption coefficient of porous acoustic materials are discussed. The radial basis function neural network uses fewer samples and offers a higher prediction accuracy in terms of the prediction process and results. Further, an optimization algorithm can theoretically solve the problem of other neural networks requiring several samples;the optimization algorithm is used to process a small set of samples to obtain the best training and test samples before building the neural network model and then substituting them into the neural network model for simulation prediction. What’s more, three classical theoretical models of empirical formulations for estimating the absorption coefficient of porous acoustic materials are discussed, among which the Johnson-Champoux-Allard model exhibits little error.Although the Johnson-Champoux-Allard model is a five-parameter model empirical formula, several factors(more than five) influence the absorption coefficient of porous sound-absorbing materials. Dimensional analysis, one of the important methods for establishing mathematical models,can transform complex mathematical physical problems into accurate mathematical model formulas using quantiles. Therefore, an equation relating all the influencing factors and the absorption coefficient of porous sound-absorbing materials can be established using the method of dimensional analysis, which reduces the absorption coefficient error calculated from the model equation. Two-and three-dimensional equivalent models are proposed to simulate the properties of porous acoustic materials. Although the modeling process is complicated, the three-dimensional model can simulate the properties of the material more intuitively and accurately than the two-dimensional model, and it can provide the basis for subsequent production and preparation with the three-dimensional modeling data of the material.In this review, the classification of sound-absorbing materials and their principles, as well as nine simulation techniques in three methods, are briefly introduced. The technique with the best performance in each method is presented to provide a reference for simulation-based absorption coefficient prediction of porous sound-absorbing materials.
作者
梁李斯
郭文龙
马洪月
弥晗
张宇
米嘉毓
李林波
LIANG Lisi;GUO Wenlong;MA Hongyue;MI Han;ZHANG Yu;MI Jiayu;LI Linbo(College of Metallurgical Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China;Shaanxi Metallurgical Engineering Technology Research Center,Xi’an 710055,China;Shaanxi Provincial Key Laboratory of Goldand Resources,Xi’an 710055,China)
出处
《材料导报》
EI
CAS
CSCD
北大核心
2022年第23期85-92,共8页
Materials Reports
基金
陕西省教育厅重点实验室项目(Z20200151)
国家自然科学基金青年项目(51404187)。
关键词
多孔吸声材料
非线性神经网络
等效模型
经验公式
吸声性能
porous sound-absorbing material
nonlinear neural network
equivalent model
empirical formula
sound absorption property
