Page 173 - 《应用声学》2023年第3期

第 42 卷 第 3 期 Vol. 42, No. 3
2023 年 5 月 Journal of Applied Acoustics May, 2023

⋄ 研究报告 ⋄

基于深度学习的低频宽带隔声器件设计 ∗

孙雪聪 1,2 贾 晗 1,2† 杨玉真 1 杨 军 1,2†

(1 中国科学院声学研究所 噪声与振动重点实验室 北京 100190)
(2 中国科学院大学 北京 100049)
摘要：在实际应用中，通常需要将多个声人工结构单元进行组合来实现低频宽带的隔声降噪。这种组合结构
往往参数较多，传统的设计方法很难对其进行高效的自动化设计。该文在集总参数模型的基础上，提出了一种
基于深度学习的低频宽带隔声器件设计方法，并基于该方法完成了由 9 个二阶亥姆霍兹共鸣器单元组合而成
的低频宽带隔声装置的设计。仿真结果表明，该隔声装置在 158∼522 Hz 范围内均具有良好的隔声效果，从而
验证了所提出方法的有效性。与传统方法相比，该文所提出的设计方法不仅减少了对设计者专业知识和设计
经验的依赖，而且具有更高的设计效率、更强的通用性，未来有望进一步推广至其他声人工结构的设计领域。
关键词：低频宽带隔声；亥姆霍兹共鸣器；深度学习；集总参数模型
中图法分类号: O424 文献标识码: A 文章编号: 1000-310X(2023)03-0611-09
DOI: 10.11684/j.issn.1000-310X.2023.03.020

Low-frequency broadband sound insulation device design method based on
deep learning

SUN Xuecong 1,2 JIA Han 1,2 YANG Yuzhen 1 YANG Jun 1,2
(1 Key Laboratory of Noise and Vibration Research, Institute of Acoustics, Chinese Academy of Sciences,
Beijing 100190, China)
(2 University of Chinese Academy of Sciences, Beijing 100049, China)

Abstract: In practical applications, it is necessary to combine several acoustic structures to realize broadband
sound insulation in the low frequency range. Considering that the composite structures often have multiple
parameters, it is difficult to design them efficiently using the traditional design methods. In this paper, we
proposed a design method of the low-frequency broadband sound insulation device using deep learning model
based on the lumped-parameter technique. Moreover, we designed a composite structure with 9 two-order
Helmholtz resonators using the proposed method. The simulation results show that the composite structure
has good sound insulation effect in the range of 158–522 Hz, which demonstrates the effectiveness of the
proposed method. Compared with the traditional methods, the proposed method can not only reduce the
dependency on the designer’s skills and experiences, but also improve design efficiency. The proposed model
has a strong versatility and scalability, which can be further extended to other acoustic structures.
Keywords: Low-frequency broadband sound insulation; Helmholtz resonator; Deep learning; Lumped-
parameter technique

2021-12-29 收稿; 2022-02-16 定稿
广东省重点领域研发计划项目 (2020B010190002), 国家自然科学基金项目 (11874383, 12104480), 中国科学院声学研究所前沿探索项
∗
目 (QYTS202110)
作者简介: 孙雪聪 (1995– ), 女, 天津人, 博士研究生, 研究方向: 声人工结构的智能化设计与应用。
† 通信作者 E-mail: hjia@mail.ioa.ac.cn; jyang@mail.ioa.ac.cn