A method for analyzing the statistical characteristics of extremely low frequency sound fluctuations
Through the finite element model and boundary coupling algorithm, the ups and downs of extremely low-frequency acoustic signals under the complex layered seabed are analyzed, and combined with normal distribution fitting, the problem of detection of extremely low-frequency acoustic signals in complex seabed environments is solved, and the accuracy of signal detection is improved sex.
Patent Information
- Application Number
- CN202111475583.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-12-06
AI Technical Summary
The prior art is difficult to accurately analyze the fluctuations of extremely low-frequency acoustic signals in complex stratified elastic seabed environments, resulting in insufficient signal detection results, especially under the influence of isolated sub-wave activities.
The finite element model is used to construct a water acoustic environment model and an internal wave motion model of a layered elastic seabed. The sound field is solved by combining the boundary coupling algorithm, and the statistical results are combined with the normal distribution to fit the energy fluctuation characteristics of the acoustic signal through the Gaussian function.
This method can more accurately consider the impact of complex stratified seabed on extremely low-frequency acoustic signals, provide analysis results that are closer to the real ocean conditions, and improve the accuracy and reliability of signal detection.
Smart Images

Figure CN114235139B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of signal detection, and in particular relates to a method for analyzing the statistical characteristics of frequency sound fluctuations. Background Art
[0002] Extremely low frequency sound waves have strong penetrability and can penetrate sedimentary layers. The propagation of sound signals will be affected by the bedrock below the sedimentary layer. Compared with medium and high frequencies, the propagation of extremely low frequency sound signals in shallow waters is more strongly affected by the interaction with the bottom. The bottom characteristics of the layered seabed are important factors affecting the propagation of extremely low frequency sound.
[0003] Soliton internal waves are a common hydrological phenomenon in shallow continental shelf waters, especially in areas with strong thermoclines. Soliton internal wave activity causes dramatic changes in the sound velocity level of seawater, which in turn causes energy coupling between simple normal waves, and can cause fluctuations in sound energy of more than 20dB within a propagation distance of more than ten kilometers. The fluctuations in the energy of the acoustic signal caused by internal wave motion will have a great impact on the subsequent signal detection results, so studying the analysis method of the fluctuation characteristics of the acoustic signal is of great significance to signal detection.
[0004] At present, conventional methods for studying the fluctuation of acoustic signals include theoretical formula derivation, field test measurement, and simulation analysis using sound propagation models. The propagation process of internal waves is a very complex process. The derivation of theoretical formulas is not only difficult, but also the results are not intuitive enough. Although field tests can obtain more intuitive results, the complex and changeable marine environment often has a great impact on the accuracy of experimental results, and offshore tests are very labor-intensive and financial. Simulation analysis uses actual underwater acoustic environment parameters for theoretical simulation calculations to obtain the sound field under different scenarios. Not only can a more intuitive sound field distribution be obtained, but it also saves manpower and financial resources, and can provide certain result predictions for actual offshore tests.
[0005] The fluctuation of acoustic signals is obtained by processing the calculated acoustic field information. Currently, the models used for acoustic field calculation include: simple normal wave model, ray model, parabolic equation model, beam integral model, finite element model, etc. The simple normal wave model is a traditional acoustic field calculation model with fast calculation speed and simple operation. However, for layered elastic seabed and distance-related models, the simple normal wave method often cannot give more accurate results. The ray model is not suitable for low-frequency signals. The finite element model is a general numerical solution for complex boundary problems on partial differential equations. Its basis is to divide the physical domain into a finite number of discrete units, thereby solving the exact solution or approximate solution within these units. Compared with other models, the biggest feature of the finite element model is that it considers the boundary coupling problem. For complex elastic seabed, the calculation results of the finite element method are more accurate. The application background of the finite element model method is extremely low frequency. The extremely low frequency must consider the influence of complex layered seabed on the acoustic signal, so using the finite element model to study the acoustic fluctuation characteristics caused by internal wave motion is the best choice.
[0006] Internal wave motion at extremely low frequencies causes energy fluctuations in acoustic signals. In the context of this problem, and considering the impact of complex stratified elastic seabed, it is necessary to fill the gap in this part of the research. It is necessary to study a method for analyzing the statistical characteristics of acoustic fluctuations based on the finite element acoustic field modeling method to provide a reference for subsequent acoustic signal detection. Summary of the invention
[0007] In order to overcome the shortcomings of the prior art, the present invention provides a method for analyzing the statistical characteristics of extremely low-frequency sound fluctuations. First, a hydroacoustic environment model and an internal wave motion model with a layered elastic seabed are constructed, and then the sound field is solved using a boundary coupling algorithm; next, the results of the acoustic signal energy fluctuations caused by the moving internal waves are statistically analyzed; finally, the statistical results are fitted using a function. The method of the present invention uses a finite element model to simulate the ocean environment. Compared with other models, this method can take into account more complex layered elastic seabeds and is closer to real ocean conditions. At the same time, the statistical results are combined with the normal distribution (Gaussian distribution function), and the fitting results are good, which can provide a valuable reference for subsequent signal detection methods.
[0008] The technical solution adopted by the present invention to solve the technical problem includes the following steps:
[0009] Step 1: Construct and solve the ELF sound propagation model under the background of moving internal waves;
[0010] Step 1-1: Construct a hydroacoustic environment model with a layered elastic seabed;
[0011] Construct M plane hydroacoustic environment models in the sea with a width of R0 and a depth of H0, extending from the sea surface to the seabed. The upper part of the hydroacoustic environment model is the water area, and the lower part is the seabed; the water area is a pressure acoustic model, and the seabed is a solid mechanics model;
[0012] The sea surface is defined as a soft boundary condition, the sound source frequency range is within the range of 10-50Hz, and the sound source is placed on a vertical boundary of the water area;
[0013] The seabed is layered, and the seabed is divided into 2 to 5 layers from the sedimentary layer to the bedrock according to the seabed formation parameters, and the compression wave velocity, shear wave velocity, compression wave attenuation, shear wave attenuation and density parameters of each layer are set;
[0014] A perfectly matched layer with a width of three times the wavelength of the sound wave extends outside the bottom edge of the hydroacoustic environment model and outside another vertical boundary of the water area;
[0015] Step 1-2: Construct an extremely low frequency sound propagation model under moving internal waves;
[0016] Step 1-2-1: Use the soliton internal wave sound velocity profile c(r,z) as the static internal wave model, the formula is as follows:
[0017]
[0018] Where c(r,z) represents the sound velocity profile of the internal wave, z is a variable representing the depth, and r is a variable representing the horizontal distance; z2 represents the depth of the upper interface of the thermocline, and z1 represents the depth of the lower interface of the thermocline; the sound velocities c1 and c2 represent the sound velocities within the corresponding depth range in equation (1), H represents the distance from the sea surface to the seabed, and ε represents the sound velocity gradient of the thermocline; η(r) represents the change in the depth of the thermocline caused by the internal wave, which is expressed by the KDV equation:
[0019]
[0020] Where a is the amplitude of the soliton internal wave, sech is the hyperbolic secant function, W is a physical quantity that characterizes the size in the propagation direction of the internal wave, L represents the distance range of the soliton internal wave disturbance on the waveguide, R = vT represents the distance from the center of the internal wave to the sound source, v is the internal wave propagation speed, and T is the propagation time;
[0021] Step 1-2-2: Use multiple static internal wave models to construct an extremely low frequency sound propagation model under moving internal waves:
[0022] When the internal wave center moves from R1 to R2, Establish a static internal wave model, with a total of M static internal wave models, and use M models to simulate the process of internal waves moving from R1 to R2; R1 and R2 represent different distances between the center of the internal wave and the sound source;
[0023] M static internal wave models are added to the water areas of M underwater acoustic environment models respectively to obtain the extremely low frequency sound propagation model under the moving internal wave;
[0024] Step 1-3: Solve the acoustic field using the boundary coupling algorithm;
[0025] Using the boundary coupling algorithm in the finite element method, the finite element discretization form of the fluid domain wave equation is applied to the water area, and the finite element discretization form of the elastic wave equation is applied to the seabed;
[0026] The finite element discretization form of the fluid domain wave equation is:
[0027]
[0028] The finite element discretization form of the elastic wave equation is:
[0029]
[0030] Where p is the sound pressure vector, C, D, E, and f are the global integral parameters in the finite element method calculation process, v is the node displacement vector, and C s is the discretized mass matrix of the finite element method, D s is the damping matrix, E s is the stiffness matrix, g represents the force vector;
[0031] The boundary conditions for the normal displacement continuity on the water area and seabed boundaries are:
[0032]
[0033] in, u n is the normal displacement of the boundary between the water area and the seabed, n is the interface normal, δ represents the impulse function matrix, and ρ0 represents the water area density;
[0034] Formulas (3) and (4) are coupled using boundary conditions (5) to simultaneously solve for the fluid pressure;
[0035] Step 2: Statistic the results of the energy fluctuation of the acoustic signal caused by the motion internal wave;
[0036] Step 2-1: Extract the sound energy calculation results at different distances from the sound source;
[0037] If R1>R2, then extract sample points within the distance range of R>R1; if R2>R1, then extract sample points within the distance range of R>R2; ensure that the extracted sample points are not inside the internal wave;
[0038] Step 2-2: Extract N sample points from each of the M models and construct an M×N matrix P ij , where N is the number of sample points of each model;
[0039]
[0040] Among them, the matrix P ij The jth row [p 1j p 2j p 3j … p Nj ] represents the variation of acoustic signal energy with distance at the jth depth, and the matrix P ij The i-th column [p i1 p i2 p i3 … p iM ] T represents the fluctuation of the i-th sample point over time during the internal wave motion;
[0041] Step 2-3: For the matrix P ij Find the mean of each column and get N mean values A i :
[0042]
[0043] Each element in each column is then subtracted from the mean of each column, and this difference is regarded as the magnitude of the sound energy fluctuation, recorded as the statistic C ij :
[0044]
[0045] A total of M×N sound energy fluctuation values C are obtained ij , C ij All elements in are arranged from small to large, and divided into an interval every H decibels, and a total of P intervals; count the number of sample points in each interval, and divide it by the total number of sample points M×N, and draw a histogram;
[0046] Step 3: Fit the statistical results using a function;
[0047] The median of each interval divided in step 2-3 is recorded as x, and the ratio of the number of sample points in the interval corresponding to x to the total number of sample points is recorded as y, and the corresponding relationship between x and y is obtained (x, y). The corresponding relationship of the kth interval is recorded as (x k ,y k ), (x k ,y k ) is the sample data set used for fitting; k = 1, 2, …, P;
[0048] The fitting function uses a Gaussian function with a constant term a:
[0049]
[0050] The fitting method uses the least squares method: solve a, b, c, d in equation (9) so that P points (x k ,y k) is the distance δ from the curve y = f(x) k The sum of squares is the smallest, that is, the value of formula (10) is the smallest:
[0051]
[0052] Use the gradient descent method to solve the values of a, b, c, and d. The results are recorded as a0, b0, c0, d0, and Gaussian function This is the fitting function of the data (x, y).
[0053] Preferably, the value range of M is 10-30.
[0054] Preferably, in step 2, when N sample points are extracted from M models respectively, the interval between the sample points is 1 m to 5 m.
[0055] Preferably, the value range of H is 1dB-5dB.
[0056] The beneficial effects of the present invention are as follows:
[0057] 1. The method of the present invention uses a finite element model to simulate the marine environment. Compared with other models, this method can take into account a more complex layered elastic seabed and is closer to the actual marine situation.
[0058] 2. The method of the present invention combines the statistical results with the normal distribution (Gaussian distribution function), and the fitting results are good, which can provide valuable reference for subsequent signal detection methods. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1 Schematic diagrams of different seabed topography according to embodiments of the present invention, including (a) wedge-shaped uphill seabed, (b) wedge-shaped downhill seabed, (c) shelf slope seabed, and (d) flat seabed.
[0060] Figure 2 Statistical results of acoustic energy at different depths under different terrains according to the embodiments of the present invention, where (a) the wedge-shaped upslope seabed depth is 250 m, (b) the wedge-shaped downslope seabed depth is 150 m, (c) the shelf slope terrain depth is 400 m, and (d) the flat seabed depth is 800 m.
[0061] Figure 3 These are the fitting results of different terrains and depths of the embodiments of the present invention, including (a) the fitting result of the wedge-shaped upslope seabed at a depth of 250 m, (b) the fitting result of the wedge-shaped downslope seabed at a depth of 150 m, (c) the fitting result of the shelf slope terrain at a depth of 400 m, and (d) the fitting result of the flat seabed at a depth of 800 m.
[0062] Figure 4Statistical results of the fluctuation of acoustic energy caused by internal waves under the measured bottom parameters of the South China Sea according to the embodiments of the present invention, where (a) is a depth of 100 m, (b) is a depth of 300 m, (c) is a depth of 600 m, and (d) is a depth of 900 m.
[0063] Figure 5 These are the fitting results of the acoustic energy fluctuations at different depths according to the embodiments of the present invention, including: (a) the fitting result at a depth of 100 m, (b) the fitting result at a depth of 300 m, (c) the fitting result at a depth of 600 m, and (d) the fitting result at a depth of 900 m. DETAILED DESCRIPTION
[0064] The present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0065] This invention provides an analysis method for the statistical characteristics of acoustic fluctuations caused by internal wave motion under a complex stratified elastic seabed, which is used to study the laws of acoustic fluctuations caused by internal wave motion and provide a reference for subsequent signal detection.
[0066] A method for analyzing extremely low frequency sound fluctuation statistical characteristics comprises the following steps:
[0067] Step 1: Construct and solve the ELF sound propagation model under the background of moving internal waves;
[0068] The basic idea of the finite element method is to discretize the solution domain into a finite number of units and obtain the approximate solution within the unit through the variable values of the unit nodes. The variable value of each unit node is calculated based on the input parameters of the physical domain.
[0069] Step 1-1: Construct a hydroacoustic environment model with a layered elastic seabed;
[0070] Construct M plane hydroacoustic environment models with a width of R0 and a depth of H0, extending from the sea surface to the seabed. The upper part of the hydroacoustic environment model is the water area, and the lower part is the seabed. The water area is a pressure acoustic model, and the seabed is a solid mechanics model. The value range of M is 10-30.
[0071] The sea surface is defined as a soft boundary condition, the sound source frequency range is within the range of 10-50Hz, and the sound source is placed on a vertical boundary of the water area;
[0072] The seabed is layered, and the seabed is divided into 2 to 5 layers from the sedimentary layer to the bedrock according to the seabed formation parameters, and the compression wave velocity, shear wave velocity, compression wave attenuation, shear wave attenuation and density parameters of each layer are set;
[0073] A perfectly matched layer with a width of three times the wavelength of the sound wave extends outside the bottom edge of the hydroacoustic environment model and outside another vertical boundary of the water area;
[0074] Step 1-2: Construct an extremely low frequency sound propagation model under moving internal waves;
[0075] Step 1-2-1: Use the soliton internal wave sound velocity profile c(r,z) as the static internal wave model, the formula is as follows:
[0076]
[0077] Where η(r) represents the change in thermocline depth caused by internal waves, which is expressed by the KDV equation:
[0078]
[0079] Where a is the amplitude of the soliton internal wave, W is a physical quantity that characterizes the size in the propagation direction of the internal wave, L represents the distance range of the soliton internal wave disturbance on the waveguide, R=vT represents the distance from the center of the internal wave to the sound source, v is the propagation speed of the internal wave, and T is the propagation time;
[0080] Step 1-2-2: Use multiple static internal wave models to construct an extremely low frequency sound propagation model under moving internal waves:
[0081] When the internal wave center moves from R1 to R2, A static internal wave model is established. There are M static internal wave models in total. M models are used to simulate the process of internal waves moving from R1 (unit: m) to R2 (unit: m); R1 and R2 represent different distances from the center of the internal wave to the sound source.
[0082] M static internal wave models are added to the water areas of M underwater acoustic environment models respectively to obtain the extremely low frequency sound propagation model under the moving internal wave;
[0083] Step 1-3: Solve the acoustic field using the boundary coupling algorithm;
[0084] Using the boundary coupling algorithm in the finite element method, the finite element discretization form of the fluid domain wave equation is applied to the water area, and the finite element discretization form of the elastic wave equation is applied to the seabed;
[0085] The finite element discretization form of the fluid domain wave equation is:
[0086]
[0087] The finite element discretization form of the elastic wave equation is:
[0088]
[0089] Where p is the sound pressure vector, C, D, E, and f are the global integral parameters in the finite element method calculation process, v is the node displacement vector, and C s is the discretized mass matrix of the finite element method, Ds is the damping matrix, E s is the stiffness matrix, g represents the force vector;
[0090] The boundary conditions for the normal displacement continuity on the water area and seabed boundaries are:
[0091]
[0092] in, u n is the normal displacement of the boundary between the water area and the seabed, n is the interface normal, δ represents the impulse function matrix, and ρ0 represents the water area density;
[0093] Formulas (3) and (4) are coupled using boundary conditions (5) to simultaneously solve for the fluid pressure;
[0094] Step 2: Statistic the results of the energy fluctuation of the acoustic signal caused by the motion internal wave;
[0095] Step 2-1: Extract the sound energy calculation results at different distances from the sound source (unit: dB);
[0096] If R1>R2, then extract sample points within the distance range of R>R1; if R2>R1, then extract sample points within the distance range of R>R2; ensure that the extracted sample points are not inside the internal wave;
[0097] Step 2-2: Extract N sample points for each of the M models, with the interval between the sample points ranging from 1m to 5m, and construct an M×N matrix P ij , where N is the number of sample points of each model;
[0098]
[0099] Among them, the matrix P ij The jth row [p 1j p 2j p 3j … p Nj ] represents the variation of acoustic signal energy with distance at the jth depth, and the matrix P ij The i-th column [p i1 p i2 p i3 … p iM ] T represents the fluctuation of the i-th sample point over time during the internal wave motion;
[0100] Step 2-3: For the matrix P ij Find the mean of each column and get N mean values A i :
[0101]
[0102] Each element in each column is then subtracted from the mean of each column, and this difference is regarded as the magnitude of the sound energy fluctuation, recorded as the statistic C ij :
[0103]
[0104] A total of M×N sound energy fluctuation values C are obtained ij , C ij All elements in are arranged from small to large, and each interval is divided into an interval of 1dB-5dB decibels, and a total of P intervals; the number of sample points in each interval is counted, and divided by the total number of sample points M×N, and a histogram is drawn;
[0105] Step 3: Fit the statistical results using a function;
[0106] The median of each interval divided in step 2-3 is recorded as x, and the ratio of the number of sample points in the interval corresponding to x to the total number of sample points is recorded as y, and the corresponding relationship between x and y is obtained (x, y). The corresponding relationship of the kth interval is recorded as (x k ,y k ), (x k ,y k ) is the sample data set used for fitting; k = 1, 2, …, P;
[0107] The fitting function uses a Gaussian function with a constant term a:
[0108]
[0109] The fitting method uses the least squares method: solve a, b, c, d in equation (9) so that P points (x k ,y k ) is the distance δ from the curve y = f(x) k The sum of squares is the smallest, that is, the value of formula (10) is the smallest:
[0110]
[0111] Use the gradient descent method to solve the values of a, b, c, and d. The results are recorded as a0, b0, c0, d0, and Gaussian function This is the fitting function of the data (x, y). Specific embodiment:
[0113] The method of the present invention is further explained in combination with internal wave scenarios under various terrains.
[0114] Firstly, the finite element method is used to construct the internal wave model under four types of terrains, namely, wedge-shaped upslope seabed, wedge-shaped downslope seabed, shelf slope terrain and flat seabed.
[0115] Schematic diagram of the wedge-shaped upslope seabed Figure 1 As shown in (a) (not a real-scale model), the sea depth changes from 900m to 100m, with a slope of about 4.57°. The internal wave moves from 2km to 4km at a speed of 4km / s. There is a 20m thick sediment layer between the water area and the hard substrate.
[0116] Schematic diagram of the wedge-shaped downslope seabed Figure 1 As shown in (b) (not a real scale model), the sea depth changes from 200m to 900m, with a slope of about 4°. The internal wave moves from 6km to 4km at a speed of 4km / s. There is a 20m thick sediment layer between the water area and the hard substrate.
[0117] Schematic diagram of the shelf slope seafloor Figure 1 As shown in (c) (not a real scale model), the sea depth is constant at 900m from 0km to 5km, there is a steep slope with a slope of 11.3° from 5km to 8km, and the depth is constant at 300m from 8km to 10km. The internal wave moves from 2km to 4km at a speed of 4km / s. There is a 20m thick sediment layer between the water area and the hard substrate.
[0118] Schematic diagram of flat seabed Figure 1 As shown in (d) (not a real-scale model), the sea depth is constant at 900 m. The internal wave moves from 4 km to 2 km at a speed of 4 km / s. There is a 20 m thick sediment layer between the water area and the hard substrate.
[0119] Table 1 Formation parameters
[0120] Compression wave speed Shear wave speed Compression wave attenuation Shear wave attenuation density Sedimentary layer (sand) 1600m / s 170m / s 0.8dB / λ 2.5dB / λ <![CDATA[1650kg / m 3 ]]> Base (limestone) 5300m / s 3074m / s 0.1dB / λ 0.2dB / λ <![CDATA[2600kg / m 3 ]]>
[0121] Figure 1 Table 1 shows the topographic diagram and bottom parameters. The sound velocity profile is a negative sound velocity gradient, and the internal wave speed changes from 1530m / s to 1490m / s. The internal wave amplitude is 200m, the horizontal scale is 300m, the movement speed is 4km / h, the sound source frequency is 20Hz, and the sound source depth is 50m.
[0122] After the model calculates the results, it extracts sample points for different scenarios. Figure 1 (a) Extract sample points 4 km away, Figure 1 (b) Extract sample points 6 km away, Figure 1 (c) Extract the sample point 4 km away, Figure 1(d) Extract sample points after 4 km. For the extracted sample point data, calculate the acoustic energy fluctuation of each sample point during the internal wave motion according to the method in step 2, count the number of sample points in each divided interval and divide it by the total number of sample points. The following are some examples of statistical results. Figure 2 shown.
[0123] The above results are fitted with Gaussian function (normal distribution) according to step 3, and the results are as follows Figure 3 shown.
[0124] It can be seen that the statistical results can be well fitted to the normal distribution. Finally, a set of measured terrain data is used to increase the universality of the method and the credibility of the results.
[0125] The seabed geology adopts the measured bottom parameters of the South China Sea, which divide the bottom into four layers with different sound speed and density, as shown in Table 2.
[0126] Table 2 Measured bottom parameters in the South China Sea
[0127] Thickness of the formation density Compression wave speed Shear wave speed Compression wave attenuation Shear wave attenuation 0-10m <![CDATA[1420kg / m 3 ]]> 1500m / s 80m / s 0.2dB / λ 1.0dB / λ 10-70m <![CDATA[1650kg / m 3 ]]> 1550m / s 120m / s 1.0dB / λ 1.5dB / λ 70-100m <![CDATA[1750kg / m 3 ]]> 1600m / s 170m / s 0.8dB / λ 2.5dB / λ 100-1000m <![CDATA[2600kg / m 3 ]]> 5300m / s 3074m / s 0.1dB / λ 0.2dB / λ
[0128] The first three layers are sediments with different sound speeds and densities, and the fourth layer is hard bedrock. The flat seabed is 1000m deep, the sound source frequency is 20Hz, and the sound source depth is 100m. The internal wave moves from 9km to 2km, and the sound speed ranges from 1530m / s to 1490m / s. Calculate the results within a range of 30km, and count the information of all sample points at different depths from 10km to 30km. The statistical results are as follows: Figure 4 shown.
[0129] The above results are fitted with a Gaussian function (normal distribution), and the results are as follows: Figure 5 .
[0130] All the above results can be well fitted to the normal distribution curve, indicating that the analysis method is feasible.
Claims
1. A method for analyzing the statistical characteristics of extremely low frequency sound fluctuations, characterized in that: The steps include: Step 1: Construct and solve the ELF sound propagation model under the background of moving internal waves; Step 1-1: Construct a hydroacoustic environment model with a layered elastic seabed; Construct M plane hydroacoustic environment models in the sea with a width of R0 and a depth of H0, extending from the sea surface to the seabed. The upper part of the hydroacoustic environment model is the water area, and the lower part is the seabed; the water area is a pressure acoustic model, and the seabed is a solid mechanics model; The sea surface is defined as a soft boundary condition, the sound source frequency range is within the range of 10-50Hz, and the sound source is placed on a vertical boundary of the water area; The seabed is layered, and the seabed is divided into 2 to 5 layers from the sedimentary layer to the bedrock according to the seabed formation parameters, and the compression wave velocity, shear wave velocity, compression wave attenuation, shear wave attenuation and density parameters of each layer are set; A perfectly matched layer with a width of three times the wavelength of the sound wave extends outside the bottom edge of the hydroacoustic environment model and outside another vertical boundary of the water area; Step 1-2: Construct an extremely low frequency sound propagation model under moving internal waves; Step 1-2-1: Use the soliton internal wave velocity profile c(r, z) as the static internal wave model, the formula is as follows: Where c(r, z) represents the sound velocity profile of the internal wave, z is a variable representing the depth, and r is a variable representing the horizontal distance; z2 represents the depth of the upper interface of the thermocline, and z1 represents the depth of the lower interface of the thermocline; the sound velocities c1 and c2 represent the sound velocities within the corresponding depth range in equation (1), H represents the distance from the sea surface to the seabed, and ε represents the sound velocity gradient of the thermocline; η(r) represents the change in the depth of the thermocline caused by the internal wave, which is expressed by the KDV equation: Where a is the amplitude of the soliton internal wave, sech is the hyperbolic secant function, W is a physical quantity that characterizes the size in the propagation direction of the internal wave, L represents the distance range of the soliton internal wave disturbance on the waveguide, R = vT represents the distance from the center of the internal wave to the sound source, v is the internal wave propagation speed, and T is the propagation time; Step 1-2-2: Use multiple static internal wave models to construct an extremely low frequency sound propagation model under moving internal waves: When the internal wave center moves from R1 to R2, Establish a static internal wave model, with a total of M static internal wave models, and use M models to simulate the process of internal waves moving from R1 to R2; R1 and R2 represent different distances between the center of the internal wave and the sound source; M static internal wave models are added to the water areas of M underwater acoustic environment models respectively to obtain the extremely low frequency sound propagation model under the moving internal wave; Step 1-3: Solve the acoustic field using the boundary coupling algorithm; Using the boundary coupling algorithm in the finite element method, the finite element discretization form of the fluid domain wave equation is applied to the water area, and the finite element discretization form of the elastic wave equation is applied to the seabed; The finite element discretization form of the fluid domain wave equation is: The finite element discretization form of the elastic wave equation is: Where p is the sound pressure vector, C, D, E, and f are the global integral parameters in the finite element method calculation process, v is the node displacement vector, and C s is the finite element method discretized mass matrix, D s is the damping matrix, E s is the stiffness matrix, g represents the force vector; The boundary conditions for the normal displacement continuity on the water area and seabed boundaries are: in, u n is the normal displacement of the boundary between the water area and the seabed, n is the interface normal, δ represents the impulse function matrix, and ρ0 represents the water area density; Formulas (3) and (4) are coupled using boundary conditions (5) to simultaneously solve for the fluid pressure; Step 2: Statistic the results of the energy fluctuation of the acoustic signal caused by the motion internal wave; Step 2-1: Extract the sound energy calculation results at different distances from the sound source; If R1>R2, then extract sample points within the distance range of R>R1; if R2>R1, then extract sample points within the distance range of R>R2; ensure that the extracted sample points are not inside the internal wave; Step 2-2: Extract N sample points from each of the M models and construct an M×N matrix P ii , where N is the number of sample points of each model; Among them, the matrix P ij The jth row [p 1j p 2j p 3j …p Nj ] represents the variation of acoustic signal energy with distance at the jth depth, and the matrix P ij The i-th column [p i1 p i2 p i3 …p iM ] T represents the fluctuation of the i-th sample point over time during the internal wave motion; Step 2-3: For the matrix P ij Find the mean of each column and get N mean values A i : Each element in each column is then subtracted from the mean of each column, and this difference is regarded as the magnitude of the sound energy fluctuation, recorded as the statistic C ij : A total of M×N sound energy fluctuation values C are obtained ij , C ij All elements in are arranged from small to large, and divided into an interval every H decibels, and a total of P intervals; count the number of sample points in each interval, and divide it by the total number of sample points M×N, and draw a histogram; Step 3: Fit the statistical results using a function; The median of each interval divided in step 2-3 is recorded as x, and the ratio of the number of sample points in the interval corresponding to x to the total number of sample points is recorded as y, and the corresponding relationship between x and y is obtained as (x, y). The corresponding relationship of the kth interval is recorded as (x k ,y k ), (x k ,y k ) is the sample data set used for fitting; k = 1, 2, ..., P; The fitting function uses a Gaussian function with a constant term a: The fitting method uses the least squares method: solve a, b, c, d in equation (9) so that P points (x k ,y k ) is the distance δ from the curve y = f(x) k The sum of squares is the smallest, that is, the value of formula (10) is the smallest: Use the gradient descent method to solve the values of a, b, c, and d, and the results are recorded as a0, b0, c0, d0, Gaussian function That is the fitting function of the data (x, y).
2. The method for analyzing the statistical characteristics of extremely low frequency sound fluctuations according to claim 1, characterized in that: The value range of M is 10-30.
3. The method for analyzing the statistical characteristics of extremely low frequency sound fluctuations according to claim 1, characterized in that: In the step 2, when N sample points are extracted for each of the M models, the interval between the sample points is 1 m to 5 m.
4. The method for analyzing the statistical characteristics of extremely low frequency sound fluctuations according to claim 1, characterized in that: The value range of H is 1dB-5dB.
