Processing math: 16%

EI / SCOPUS / CSCD 收录

中文核心期刊

基于长短期记忆网络与射线声学的浅海声速剖面反演方法

吴隆昊, 刘松, 吴照志, 潘才能, 袁飞

吴隆昊, 刘松, 吴照志, 潘才能, 袁飞. 基于长短期记忆网络与射线声学的浅海声速剖面反演方法[J]. 声学学报, 2025, 50(1): 12-22. DOI: 10.12395/0371-0025.2023215
引用本文: 吴隆昊, 刘松, 吴照志, 潘才能, 袁飞. 基于长短期记忆网络与射线声学的浅海声速剖面反演方法[J]. 声学学报, 2025, 50(1): 12-22. DOI: 10.12395/0371-0025.2023215
WU Longhao, LIU Song, WU Zhaozhi, PAN Caineng, YUAN Fei. Inversion for sound speed profile in shallow water based on long short-term memory networks and ray theory[J]. ACTA ACUSTICA, 2025, 50(1): 12-22. DOI: 10.12395/0371-0025.2023215
Citation: WU Longhao, LIU Song, WU Zhaozhi, PAN Caineng, YUAN Fei. Inversion for sound speed profile in shallow water based on long short-term memory networks and ray theory[J]. ACTA ACUSTICA, 2025, 50(1): 12-22. DOI: 10.12395/0371-0025.2023215
吴隆昊, 刘松, 吴照志, 潘才能, 袁飞. 基于长短期记忆网络与射线声学的浅海声速剖面反演方法[J]. 声学学报, 2025, 50(1): 12-22. CSTR: 32049.14.11-2065.2023215
引用本文: 吴隆昊, 刘松, 吴照志, 潘才能, 袁飞. 基于长短期记忆网络与射线声学的浅海声速剖面反演方法[J]. 声学学报, 2025, 50(1): 12-22. CSTR: 32049.14.11-2065.2023215
WU Longhao, LIU Song, WU Zhaozhi, PAN Caineng, YUAN Fei. Inversion for sound speed profile in shallow water based on long short-term memory networks and ray theory[J]. ACTA ACUSTICA, 2025, 50(1): 12-22. CSTR: 32049.14.11-2065.2023215
Citation: WU Longhao, LIU Song, WU Zhaozhi, PAN Caineng, YUAN Fei. Inversion for sound speed profile in shallow water based on long short-term memory networks and ray theory[J]. ACTA ACUSTICA, 2025, 50(1): 12-22. CSTR: 32049.14.11-2065.2023215

基于长短期记忆网络与射线声学的浅海声速剖面反演方法

基金项目: 国家自然科学基金(62371404, 62271425, 62071401)资助
详细信息
    通讯作者:

    袁飞, yuanfei@xmu.edu.cn

  • 中图分类号: 43.30, 43.60

  • PACS: 
      43.30,43.60

Inversion for sound speed profile in shallow water based on long short-term memory networks and ray theory

  • 摘要:

    针对水声多途信道环境下的水下声速剖面反演问题, 将深度学习与射线声学理论结合, 提出了一种基于长短期记忆网络(LSTM)的反演方法。该方法根据有序线阵等距的特点, 将到达时间差、到达角度等多模态数据融合组成的感知矩阵作为输入, 利用LSTM网络处理时序数据的能力挖掘空间上有序分布的接收阵元之间的关联信息, 从而实现声速剖面反演。在此基础上, 还提出了基于互相关函数的硬阈值估计方法, 通过降低感知矩阵的测量误差提高模型的抗多途性能。通过数值仿真验证了该反演方法的可行性和准确性, 与传统优化算法相比, 所提算法能够更好地捕捉声速剖面的非线性特征, 具有更高的反演精度和较强的抗噪能力。

    Abstract:

    To address the problem of underwater sound speed profile (SSP) inversion in underwater acoustic multipath channels, this paper combines deep learning and ray theory to propose an inversion method using a long short-term memory network (LSTM). Based on the equidistant characteristics of the horizontal line array, the proposed method takes the perceptual matrix composed of multi-modal data, such as time difference of arrival and angle of arrival, as input, and utilizes the ability of LSTM network to process time-series data to mine the correlations between spatially ordered receiving array elements for sound speed profile inversion. On this basis, a time delay estimation method based on hard threshold estimation method and cross-correlation function is proposed to reduce the measurement errors of the perceptual matrix and improve the anti-multipath performance. The feasibility and accuracy of the proposed method are verified through numerical simulations. Compared with the traditional optimization algorithm, the proposed algorithm better captures the nonlinear characteristics of SSP, with higher inversion accuracy and stronger noise resistance.

  • 图  1   场景模型

    图  2   整体框架图

    图  3   数据集中部分声速剖面图(最大深度200 m)

    图  4   基于LSTM的声速剖面反演网络结构图

    图  5   发送信号时频特性与自相关特性 (a) 时频特性; (b) 自相关特性

    图  6   接收信号时频特性与自相关特性(SNR = −5 dB) (a) 时频特性; (b) 自相关特性

    图  7   时延估计算法性能比较 (a) 不同SNR下的TDE误差对比; (b) 不同距离下的TDE误差对比

    图  8   LSTM单元数量、隐藏层神经元数量和训练步数对声速剖面反演的影响 (a) 不同组合(LSTM单位数、隐层神经元数); (b) 训练步数

    图  9   不同输入参数对模型训练的影响对比

    图  10   不同信噪比下声速剖面反演结果比较

    图  11   部分反演结果对比 (a) 样本1; (b) 样本2

    图  12   声速剖面反演误差分布 (a) 整体误差分布直方图; (b) 样本1反演误差; (c) 样本2反演误差

    表  1   CCF-HTP算法流程

    01 :iter, l,u, err, u // iter: l// uerr:  // u
    02 {\boldsymbol{S}} = \dfrac{1}{M}{\boldsymbol{S}}_{_u}^{\text{H}} \cdot {{\boldsymbol{S}}_{{u^{'}}}} // 计算互功率谱
    03 {\boldsymbol{\varPhi }} = \left[ {\boldsymbol{a}\left( {{\tau_0}} \right), \boldsymbol{a}\left( {{\tau_1}} \right), \cdots , \boldsymbol{a}\left( {{\tau_{N - 1}}} \right)} \right] // 计算观测矩阵
    04 {{\boldsymbol{r}}^0} = {\boldsymbol{S}}, \, {{\boldsymbol{\xi }}^0} = 0, \, {\mu ^0} = 1 // 初始化参数
    05 {\text{while}}(i < {\text{iter}})\,{\text{and}}\,\left( {\left\| {{{\boldsymbol{r}}^n} - {{\boldsymbol{r}}^{n - 1}}} \right\|_{\text{2}}^{\text{2}} > {\text{err}}} \right)
    06   \begin{gathered} {{\varGamma }^n} = {\text{\{ indices of }}k{\text{ largest entries of }} \\\qquad {\text{ }}{{\boldsymbol{x}}^n} + {\mu ^n}{{\boldsymbol{\varPhi }}^T}{{\boldsymbol{r}}^{n - 1}}{\text{\} }} \\ \end{gathered} // 取备选更新信号的l
    最大值索引构建索引
    集合
    07   {\boldsymbol{v}} = \arg \min \{ ||{\boldsymbol{S}} - {\boldsymbol{\varPhi v}}||,{\text{supp}} ({\boldsymbol{v}}) \subseteq {{\varGamma }^n}\} // 选择支撑集包含于索引集合的有效信号
    08 {{\boldsymbol{\xi }}^n} = {\boldsymbol{v}} // 更新信号
    09 {\mu ^n} = \dfrac{{{\text{||(}}{{\boldsymbol{\varPhi }}^{\text{*}}}{\text{(}}{\boldsymbol{S}} - {\boldsymbol{\varPhi }}{{\boldsymbol{x}}^n}{\text{)}}{{\text{)}}_{{{\varGamma }^n}}}{\text{||}}_{\text{2}}^{\text{2}}}}{{{\text{||}}{\boldsymbol{\varPhi }}{{{\text{(}}{{\boldsymbol{\varPhi }}^{\text{*}}}{\text{(}}{\boldsymbol{S}} - {\boldsymbol{\varPhi }}{{\boldsymbol{x}}^n}{\text{))}}}_{{{\varGamma }^n}}}{\text{||}}_{\text{2}}^{\text{2}}}} // 更新步长
    10 {{\mathrm{end}}}
    11 输出: \; {\boldsymbol\xi }^{n}
    下载: 导出CSV

    表  2   参数设置

    参数 仿真取值
    采样率 (kHz) 100
    Chirp信号频率范围 (kHz) 20~30
    Chirp信号周期 (ms) 40
    水声信道多途数量 10
    最大时延拓展 (ms) 30
    输入层神经元个数 4
    输出层神经元个数 25
    接收浮标个数 20
    PSO粒子数 50
    PSO迭代次数 50
    学习率 0.0001
    声速剖面训练集数量 8800
    声速剖面测试集数量 2200
    声速剖面深度采样间隔 (m) 8
    训练迭代次数 10000
    下载: 导出CSV
  • [1]

    Fuda J L, Millot C, Taupier-Letage I, et al. XBT monitoring of a meridian section across the western Mediterranean Sea. Deep Sea Res. Part I, 2000; 47(11): 2191−2218 DOI: 10.1016/S0967-0637(00)00018-2

    [2]

    Gettelman A, Geer A J, Forbes R M, et al. The future of Earth system prediction: advances in model-data fusion. Sci. Adv., 2022; 8(14): eabn3488 DOI: 10.1126/sciadv.abn3488

    [3] 张忠兵, 马远良, 杨坤德, 等. 浅海声速剖面的匹配波束反演方法. 声学学报, 2005; 30(2): 103−107 DOI: 10.3321/j.issn:0371-0025.2005.02.002
    [4]

    Tolstoy A, Sotirin B. Acoustic tomography via matched field processing. J. Acoust. Soc. Am, 1995; 89(5): 393−406 DOI: 10.1121/1.411711

    [5]

    Shang E C. Ocean acoustic tomography based on adiabatic mode theory. J. Acoust. Soc. Am, 1989; 85(4): 1531−1537 DOI: 10.1121/1.397355

    [6]

    Roux P, Iturbe I, Nicolas B, et al. Travel-time tomography in shallow water: Experimental demonstration at an ultrasonic scale. J. Acoust. Soc. Am, 2011; 130(3): 1232−1241 DOI: 10.1121/1.3621271

    [7]

    Li F H, Zhang R H. Inversion for sound speed profile by using a bottom mounted horizontal line array in shallow eater. Chin. Phys. Lett., 2010; 27(8): 084303 DOI: 10.1088/0256-307X/27/8/084303

    [8]

    Voronovich A G, Shang E C. Numerical simulations with horizontal-refraction-modal tomography. J. Acoust. Soc. Am, 1997; 101(5): 2636−2643 DOI: 10.1121/1.418504

    [9]

    Skarsoulis E K, Athanassoulis G A, Send U. Ocean acoustic tomography based on peak arrivals. J. Acoust. Soc. Am, 1996; 100(2): 797−813 DOI: 10.1121/1.416212

    [10]

    Yang T C, Yates T. Matched-beam processing: Application to a horizontal line array in shallow water. J. Acoust. Soc. Am, 1998; 104(3): 1316−1330 DOI: 10.1121/1.424341

    [11] 廖光洪, 朱小华, 林巨, 等. 海洋声层析观测技术和方法. 海洋学报, 2010; 32(3): 14−22
    [12] 唐俊峰, 杨士莪. 由传播时间反演海水中的声速剖面. 哈尔滨工程大学学报, 2006; 27(5): 733−736 DOI: 10.3969/j.issn.1006-7043.2006.05.022
    [13] 李鹏程, 冯海泓, 李记龙. 基于参考点之间水声传播的声速剖面反演. 声学技术, 2023; 42(4): 446−451 DOI: 10.16300/j.cnki.1000-3630.2023.04.006
    [14] 张忠兵, 马远良, 倪晋平. 基于声线到达时差的浅海声速剖面反演. 西北工业大学学报, 2002; 20(1): 36−39 DOI: 10.3969/j.issn.1000-2758.2002.01.009
    [15]

    Liu C, Qu K. Wide-area sound speed profile estimation based on a pre-classification scheme for sound speed perturbation modes. Front. Marine Sci., 2023; 10: 1130061 DOI: 10.3389/fmars.2023.1130061

    [16]

    Chen Z, Wang P, Bao S, et al. Rapid reconstruction of temperature and salinity fields based on machine learning and the assimilation application. Front Marine Sci., 2022; 9: 985048 DOI: 10.3389/fmars.2022.985048

    [17]

    Huang W, Li D, Jiang P. Underwater sound speed inversion by joint artificial neural network and ray theory. Proceedings of the 13th International Conference on Underwater Networks & Systems, ACM, Shenzhen, China, 2018: 1−8

    [18] 吴碧, 陈长安, 林龙. 多种声速经验公式的比较分析研究. 全国水声学学术会议, 中国声学学会, 广东湛江, 2013: 300−302
    [19] 王子蘅, 王振杰, 聂志喜, 等. 声速剖面EOF重构的实测数据采样深度研究. 海洋科学, 2021; 45(6): 126−134 DOI: 10.11759/hykx20200711001
    [20]

    Xia H, Yang K, Duan R, et al. Analysis and estimation of critical depth based on SODA datasets in the Philippine Sea. IEEE/OES China Ocean Acoustics, IEEE, Harbin, China, 2016: 1−7

    [21] 刘伯胜. 水声学原理. 北京: 科学出版社, 2019
图(12)  /  表(2)
计量
  • 文章访问数:  220
  • HTML全文浏览量:  31
  • PDF下载量:  98
  • 被引次数: 0
出版历程
  • 收稿日期:  2023-09-04
  • 修回日期:  2023-11-10
  • 刊出日期:  2025-01-10

目录

    /

    返回文章
    返回