Research On Classification Of Sound Speed Profiles In Water Column And Its Effects On Sound Propagation

Most of the surrounding seas in our country are shallow seas,and the marine environment is complex.Studying the sound propagation characteristics in complex hydrological environments is an important basis for achieving efficient underwater detection,communication,identification,and other applications.Sound speed is one of the key factors that affect the structure of underwater sound propagation.The types of shallow water sound speed profiles(SSP)are diverse,and different structures may appear in different seasons and different geographical locations.Due to the influence of sound speed profiles and boundary conditions,the acoustic characteristics of shallow water waveguides are relatively complex,therefore,it is of great significance to understand the complex underwater sound speed environment,classify it,and master the sound propagation laws under different sound speed classification conditions for engineering applications research based on acoustic field characteristics in complex hydrological environments.In this paper,we propose a new fitting strategy to extract the general trend of the profile using measured shallow water underwater sound speed profile data.Then,we apply a deep learning model combined with clustering algorithm to achieve a more reasonable classification of large-scale shallow water sound speed profiles.Finally,we study the sound propagation characteristics under different sound speed profiles,and summarize the propagation laws from the optimal receiving depth and optimal frequency.The specific work in this paper is as follows.To address the problem of fluctuations and outliers in actual sound speed profiles due to instruments,measurement methods,and other factors,in this paper,we first use the least squares method and global optimization algorithm to linearly fit the historical sound speed profile samples of the research area,and extract the profile trend through line segments.As only using a fixed threshold set artificially for fitting will cause a large number of overfitting and underfitting phenomena,we propose a sound speed profile fitting algorithm based on the threshold and the sum of gradient deviation,which can effectively reduce overfitting and underfitting phenomena and extract the basic trend of the profile.To address the problem of a small underwater sound speed profile classification dataset leading to a single category and insufficiently refined corresponding sound field feature classification,we propose a sound speed profile classification method based on deep clustering.We apply a deep clustering model based on an autoencoder,and use the powerful nonlinear mapping ability of neural networks to fit the clustering laws in the sound speed profile internally to achieve automatic classification on large-scale sound speed profiles.The effectiveness and feasibility of the deep clustering algorithm applied to large-scale sound speed profiles classification tasks are verified from multiple perspectives,providing rich sound speed environmental conditions for subsequent sound propagation laws studies.Based on the study of sound speed profile classification,we carry out the research on sound field feature laws.In order to compare the effects of different sound speed profiles on sound propagation structures,we analysis the optimal sound propagation receiving depth and optimal frequency in different sound speed classification environments by changing the depth and frequency of the sound source.Based on the above phenomena,we further summarize the sound propagation laws shown in the current study of the sound speed profile classification environment,and propose the appropriate deployment depth of the receiving array and the area suitable for target concealment under different sound speed profile classification environment.It provides some guidance for the engineering application of the sound field characteristics in the hydrological environment.