【摘要】 声学超材料作为一类新型人工复合结构材料，拥有天然材料所不具备的众多新颖和反常物理特性，为声波的有效控制和精准调控提供了崭新的研究途径和应用机遇。然而，为获得声学超材料的特定结构功能响应，传统设计方法需要在理论推导、数值模拟和试验验证过程中不断反复调整材料参数或结构形态，大幅增加了研究计算成本。机器学习具有强大的非线性拟合能力，能够通过优化算法绕过物理建模过程，直接在参数空间内构建恰当映射关系以达到满足目标功能的目的，为突破传统物理设计策略的高度限制提供了可能性。综述了近几年来机器学习在声学超材料的应用进展。首先，简要概述了声学超材料的基本发展情况和主流机器学习算法的基本原理，然后详细介绍了机器学习在声子晶体、声学超材料和声学超材料拓扑设计中的最新应用研究成果，最后就该领域的研究现状开展了相应讨论和展望。更多还原

【Abstract】 Acoustic metamaterials, as a new type of artificial composite structural materials possess numerous novel and anomalous physical properties which natural materials do not possess to provide new study avenues and application opportunities for effective control and precise regulation of sound waves. However, in order to obtain specific structural functional responses of acoustic metamaterials, traditional design methods require repeated adjustments of material parameters or structural forms in processes of theoretical derivation, numerical simulation and experimental verification to significantly increase study and computational costs. Machine learning has powerful nonlinear fitting capabilities, it can bypass physical modeling process with optimization algorithms and directly construct appropriate mapping relations in parametric space to reach the goal of meeting functional requirements, and provide the possibility to break through height limitations of traditional physical design strategies. Here, an overview of recent application progress of machine learning in acoustic metamaterials was summarized. Firstly, a brief overview of basic development of acoustic metamaterials and fundamental principles of mainstream machine learning algorithms was presented. Then, the latest application study results of machine learning in phononic crystals, acoustic metamaterials and acoustic metamaterial topology design were introduced in detail. Finally, the current study status and prospects in this field were discussed and outlooked accordingly.更多还原