Method and system for inverting earth sound parameters by matching broadband modal phase velocity based on sparse Bayesian learning

Abstract

The invention provides a sparse Bayesian learning-based matched broadband modal phase velocity inversion earth sound parameter method and system, and the method comprises the steps: extracting a local mode through employing a receiving signal of a vertical array, and removing the narrowband limitation of block sparse Bayesian; the broadband modal phase velocity is used for matching inversion of the earth sound parameters, and decoupling of the earth sound parameters is achieved. The method has the advantages that seabed parameters do not need to be known priori, the depth range of the extraction mode is not limited by the aperture of the VLA, and the method can be applied to a non-uniform vertical array; when the earth sound parameters are inversed, the earth sound parameters are not influenced by the terrain on the propagation path, and the earth sound parameters of any terrain can be inversed; known sound source position information is not needed, and the problem of inaccurate inversion result caused by inaccurate sound source position is avoided; and meanwhile, multi-distance sound source signals are used for inversion, so that the inversion of the earth sound parameters is more robust and accurate.

Description

Technical Field

[0001] This application belongs to the field of marine acoustic technology, and particularly relates to a method and system for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning. Background Art

[0002] The inversion of seabed acoustic parameters is an important issue in marine acoustics. Inversion is a method for quickly and cost-effectively obtaining shallow-water geoacoustic parameters, and can obtain acoustic parameters such as seabed sound speed, density, and attenuation coefficient, which have a decisive impact on sound propagation in the ocean and are of great significance for the design of various intelligent ocean-related instruments. The most widely used method for seabed sediment inversion is matched-field inversion (MFI). Commonly used matching quantities include sound pressure, dispersion curves, and propagation loss, etc. Traditional matched-field inversion is affected by the propagation path, and terrain mismatch will affect the accuracy of the inversion results. Moreover, traditional matched-field inversion requires adding the sound source position as prior information to the inversion, or taking the sound source information as the quantity to be inverted, which exacerbates the multi-valuedness problem. When inverting multiple parameters, the coupling problem between geoacoustic parameters cannot be ignored. In addition, most matched-field inversions use the received data at a certain distance point in isolation for inversion, without making full use of the potential correlation between multiple segments of acoustic data to obtain more reliable results.

[0003] The modal function and horizontal wavenumber are the shape function and propagation characteristic parameters of the normal mode, which are closely related to the acoustic field parameters (such as the sound speed in water, the depth of the sea water, and the bottom sediment characteristics, etc.), determine the distribution and propagation characteristics of the acoustic field, and contain important information of the acoustic field. For modal extraction, a horizontal array can estimate the horizontal wavenumber, but the classical spatial Fourier or Hankel transform methods require a large distance aperture. For a vertical array, among many modal extraction methods, the mode filtering method requires prior environmental information, the SVD (Singular Value Decomposition) method requires the vertical array to span the entire water depth, and the warping transform method can only be used for pulse signals. In a shallow water waveguide, the acoustic field can be regarded as a weighted superposition of multi-order normal modes. Therefore, modal extraction can be regarded as a sparse signal recovery problem, which can be solved by sparse Bayesian learning (SBL). Sparse Bayesian learning does not require prior knowledge of the seabed parameters and source information when using a vertical array to extract modes. The depth range for extracting modes is not limited by the aperture of the vertical array, and it can also be applied to non-uniform vertical arrays, which is more flexible and efficient in practical applications. Block sparse Bayesian learning uses a multi-frequency modal depth function with a block structure as the dictionary matrix, but due to the use of an approximate dispersion relation, the frequency band is limited to a narrow band, and only narrow-band modes can be extracted. In addition, when constructing a multi-frequency dictionary matrix to extract modes with only one hydrophone, the distance of the sound source needs to be known. Summary of the Invention

[0004] The purpose of this application is to eliminate the dependence of traditional matched field inversion on the terrain of the propagation path and the source position information, alleviate the problem of lack of representativeness in isolated point inversion, realize the decoupling of geoacoustic parameters, and improve the robustness and accuracy of inversion.

[0005] To achieve the above purpose, this application proposes a method for inverting geoacoustic parameters based on the matched broadband modal phase velocity of sparse Bayesian learning, including:

[0006] Step 1: Perform Fourier transform on the acoustic propagation time-domain data at multiple distances received by the hydrophone vertical array to obtain the multi-distance frequency-domain signals at each frequency point within the required bandwidth;

[0007] Step 2: Collect horizontal wavenumbers within a set interval and use the shooting method to generate corresponding local modal depth functions to form a modal depth function matrix A;

[0008] Step 3: Form a matrix Y with the multi-distance time-domain signals at each frequency point, and use the matrix A and Y through the iterative method of sparse Bayesian learning to calculate the actually excited local modes and local horizontal wavenumbers;

[0009] Step 4: Perform multi-distance sparse Bayesian learning for each frequency point within the required bandwidth, and convert the obtained local horizontal wavenumber into local phase velocity;

[0010] Step 5: Conduct data cleaning and curve fitting on the phase velocity points of the wideband;

[0011] Step 6: Substitute the environmental parameters to be searched into the acoustic propagation model to obtain the phase velocity of each frequency point; by minimizing the mean square error between the frequency-phase velocity curve obtained from the acoustic propagation model and the curve fitted in Step 5, invert the environmental parameters including seabed sound speed, density, and water depth;

[0012] Step 7: Perform depth correction on the actually measured water depth with the inverted water depth;

[0013] Step 8: Invert the seabed attenuation coefficient by using the matching propagation loss method.

[0014] As an improvement to the above method, the set interval is: [2πf / c b , 2πf / c w ; where f represents the signal frequency, c b and c w respectively represent the sound speeds in the seabed and water.

[0015] As an improvement to the above method, the formula of the shooting method is:

[0016] Φ(0, k rm ) = 0,

[0017] Φ(z1, k rm ) = h,

[0018]

[0019] where ω = 2πf is the angular frequency; z j = jh, h and J are respectively the width and total number of the discrete grid; c(z j ) is the sound speed profile; Φ(z j , k rm ) is the discrete sampling value of the m-th order normal mode depth function at the j-th grid depth z j ; k rm represents the m-th order horizontal wavenumber.

[0020] As an improvement to the above method, the actually excited local mode and horizontal wavenumber are expressed as:

[0021]

[0022] where γ m is the quantity characterizing the local mode energy, represents the value of the current iteration, represents the value of the previous iteration; L represents the number of sound sources; Φ(z,k rm ) is the discrete sampling of the depth function of the m-th normal mode at the receiving end at depth z; σ 2 is the variance of the noise; Σ y is the data covariance; is a matrix composed of K excited mode functions in matrix A, + represents the pseudo-inverse calculation, and K represents the number of actually existing propagation modes; I N represents the N-dimensional identity matrix, N represents the number of array elements; the superscript H represents the conjugate calculation; ||·||2 represents the 2-norm formula; tr represents taking the trace of the matrix.

[0023] As an improvement of the above method, the obtained local horizontal wavenumber is converted into the local phase velocity, and the formula is:

[0024]

[0025] where c pm represents the local phase velocity.

[0026] As an improvement of the above method, the inversion obtains environmental parameters including the seabed sound speed, density, and water depth, and the formula is: [[ID=,28]]

[0027]

[0028] where M d is the total number of modes used; F is the total number of frequencies; represents the set of seabed environmental parameters to be inverted; is the fitted frequency-phase velocity curve; t m (f i ,u) is the copy field frequency-phase velocity curve generated by the sound propagation model when the environmental parameter set is u; f i is the frequency.

[0029] This application also provides a system for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning, which is implemented based on the above method. The system includes:

[0030] A module for obtaining multi-distance frequency-domain signals, which is used to perform Fourier transform on the sound propagation time-domain data of multiple distances received by the hydrophone vertical array to obtain multi-distance frequency-domain signals at each frequency point within the required bandwidth;

[0031] A module for obtaining the modal depth function matrix, which is used to collect horizontal wavenumbers within a set interval and generate corresponding local modal depth functions using the shooting method to form the modal depth function matrix A;

[0032] The local mode and local horizontal wavenumber calculation module is used to form a matrix Y from the multi-distance time-domain signals at each frequency point, and through an iterative method of sparse Bayesian learning with matrices A and Y, calculate the truly excited local modes and local horizontal wavenumbers;

[0033] The local phase velocity acquisition module is used to perform multi-distance sparse Bayesian learning for each frequency point within the required bandwidth and convert the obtained local horizontal wavenumbers into local phase velocities;

[0034] The curve fitting module is used to perform data cleaning and curve fitting on the phase velocity points of the broadband;

[0035] The environmental parameter inversion module is used to substitute the environmental parameters to be searched into the acoustic propagation model to obtain the phase velocity at each frequency point; by minimizing the mean square error between the frequency-phase velocity curve obtained from the acoustic propagation model and the fitted curve, invert the environmental parameters including the seabed sound speed, density, and water depth;

[0036] The depth correction module is used to perform depth correction on the actually measured water depth with the water depth obtained by inversion;

[0037] The seabed attenuation coefficient inversion module is used to invert the seabed attenuation coefficient by using the matched propagation loss method.

[0038] Compared with the prior art, the advantages of this application are as follows:

[0039] 1. When extracting modes, the bandwidth limitation is removed, so that the modes, horizontal wavenumbers, and phase velocities of broadband data can be obtained;

[0040] 2. It is not necessary to have prior knowledge of the seabed parameters. The depth range for extracting modes is not limited by the aperture of the VLA and can be applied to non-uniform vertical arrays;

[0041] 3. When inverting the geoacoustic parameters, the method of matching the local mode phase velocities is used, and the inversion is not affected by the terrain on the propagation path, and the geoacoustic parameters of any terrain can be inverted;

[0042] 4. It is not necessary to know the sound source position information, avoiding the problem of inaccurate inversion results caused by inaccurate sound source positions;

[0043] 5. The decoupling of the geoacoustic parameters is realized, avoiding the mutual influence between parameters during inversion;

[0044] 6. At the same time, the sound source signals at multiple distances are used for inversion, making the inversion of the geoacoustic parameters more robust and accurate. Brief Description of the Drawings

[0045] Figure 1 Shown is a schematic diagram of the ocean environment and the layout of the sound source - hydrophone;

[0046] Figure 2(a) shows the multi-distance sparse Bayesian learning result curve at 250 Hz and the true horizontal wave number (red ×) in the simulation environment;

[0047] Figure 2(b) shows the modal depth function estimated by multi-distance sparse Bayesian learning (blue circles) and the true modal depth function (black line) in the simulation environment, from left to right are the 1st - 5th orders;

[0048] Figure 3 Shows the scatter plot and fitting curve of broadband multi-distance sparse Bayesian learning;

[0049] Figure 4 Shows the schematic diagram of parameter optimization of the Bayesian optimization algorithm; the orange points are the estimated points of each iteration, the red stars are the true values set in the simulation, and the blue boxes are the final inversion results;

[0050] Figure 5(a) shows the frequency-phase velocity curve corresponding to the fitting curve (red line) and the matching phase velocity inversion result;

[0051] Figure 5(b) shows the comparison diagram of the matching phase velocity inversion result and the true frequency-phase velocity curve in the simulation environment;

[0052] Figure 6(a) shows the distance positioning result in the matched field positioning result. The red star is the true position of the sound source, and the blue circle is the MFP positioning result;

[0053] Figure 6(b) shows the depth positioning result in the matched field positioning result. The red star is the true position of the sound source, and the blue circle is the MFP positioning result;

[0054] Figure 7 Shows the scatter plot and fitting curve of the estimated phase velocity of broadband multi-distance sparse Bayesian learning;

[0055] Figure 8 Shows the frequency-phase velocity fitting curve (red line) estimated by broadband multi-distance sparse Bayesian learning and the frequency-phase velocity curve (black line) corresponding to the broadband matched phase velocity inversion result;

[0056] Figure 9(a) shows the distance positioning result of the matched phase velocity inversion; the red star is the sound source position recorded by GPS, and the blue circle is the MFP positioning result;

[0057] Figure 9(b) shows the depth positioning result of the matched phase velocity inversion; the red star is the sound source position recorded by GPS, and the blue circle is the MFP positioning result;

[0058] Figure 9(c) shows the distance positioning result of the traditional matched field inversion; the red star is the sound source position recorded by GPS, and the blue circle is the MFP positioning result;

[0059] Figure 9(d) shows the depth localization results of traditional matched-field inversion; the red star is the sound source position recorded by GPS, and the blue circle is the MFP localization result;

[0060] Figure 10(a) shows the cost function of traditional matched-field inversion for 59 points;

[0061] Figure 10(b) shows the results of traditional matched-field inversion for water depth for 59 points;

[0062] Figure 10(c) shows the results of traditional matched-field inversion for seabed sound speed for 59 points;

[0063] Figure 10(d) shows the results of traditional matched-field inversion for seabed density for 59 points;

[0064] Figure 10(e) shows the results of traditional matched-field inversion for seabed attenuation coefficient for 59 points;

[0065] Figure 11 The figure shows the flow chart of the method for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning. Detailed implementation mode

[0066] The technical solution of the present application will be described in detail below with reference to the accompanying drawings.

[0067] The present invention proposes a method and system for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning. This method uses the received signals of a vertical array, extracts local modes using multi-distance sparse Bayesian learning, and removes the narrowband limitation of block sparse Bayesian. The broadband modal phase velocity is used for the matching inversion of geoacoustic parameters, realizing the decoupling of geoacoustic parameters. Moreover, the inversion is not affected by the terrain on the propagation path and does not require the information of the known sound source position. During the search process of multi-dimensional parameters, the Bayesian optimization algorithm is used to improve the search efficiency. In simulation and experimental data, matched-field processing (MFP) localization is performed based on the inversion results, verifying the feasibility of this method and providing a feasible new method for geoacoustic parameter inversion.

[0068] The principle and derivation process of the method for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning are as follows:

[0069] 1. Signal model and multi-distance sparse Bayesian learning

[0070] According to the normal mode theory, the point sound source at the l-th sound source position The frequency-domain received signal y (l) (z) at depth z is composed of the weighted superposition of M modes:

[0071]

[0072] wherein, is the sound source distance at the l-th sound source position, is the sound source depth at the l-th sound source position; Φ(z, k rm ) is the discrete sampling of the m-th order normal mode depth function at the receiving end at depth z, often referred to as the local mode. k cm = k rm + ik im is the complex eigenvalue, the real part k rm represents the m-th order horizontal wavenumber, and the imaginary part k im represents the mode attenuation. Formula (1) is applicable in both distance-independent and distance-dependent scenarios. Essentially, it decomposes the sound pressure onto the local mode Φ(z, k rm ). The difference lies in that the modal amplitude has different expressions under adiabatic and coupled conditions, that is, different decomposition coefficients. When the sound speed profile is known, for a given horizontal wavenumber k rm , the modal depth function Φ(z, k rm ) can be obtained by the shooting method.

[0073] We express the received signals at different distances using the local mode depth function at the same receiving array. Let the received signal of N array elements at each distance be y (l) . At the same time, considering the received signals of L sound sources at different distances to form the received signal matrix Y = [y (1) , …, y (l) , …, y (L) , l = 1, 2, …, L. represents the modal amplitudes of each order at L distances, as follows:

[0074]

[0075] At this time, the decomposition of the signals at different sound source distances can use the same set of modal depth function matrices, written as matrix A, and the received signal set can be expressed as:

[0076] Y = AX + N (3)

[0077] A is composed of the M-order modal depth functions:

[0078]

[0079] is the noise that follows the complex Gaussian distribution .

[0080] Estimate the signal mode using the sparse Bayesian method. In the interval [2πf / c b , 2πf / cw (where \(f\) represents the signal frequency) a large number of acquisitions (\(M\) n pieces) of the horizontal wavenumber \(k\) rm and use the shooting method to generate the corresponding local modal depth function, \(c\) b and \(c\) w are the sound speeds in the seabed and water respectively. \(c\) b is unknown and can be set to a relatively large value during the execution process. At this time, the number of columns of \(A\) is also extended to \(M\) n columns, and the number of rows of \(X\) is also extended to \(M\) n rows. Among the \(M\) n column modes of \(A\), only \(K\) are real propagating modes, and the rest are virtual. Since \(M\) n is much larger than the true number of propagating modes \(K\), most of the rows in \(X\) are zero, making it \(K -\)sparse.

[0081] According to the sparse Bayesian learning theory and calculation formula, use the dictionary matrix \(A\) to perform sparse Bayesian learning on the multi - distance received dataset \(Y\) to estimate the modal coefficient set \(X\). The non - zero values in \(X\) represent the energy of the modes. The appearance indicates the existence of the corresponding propagating mode, and the positions of the non - zero values can be used to calculate the local horizontal wavenumber and the local modal depth function.

[0082] 2. Matched phase velocity inversion

[0083] The horizontal wavenumber \(k\) rm of the \(m\) - th mode and the phase velocity \(c\) pm have the following relationship:

[0084]

[0085] Therefore, the local phase velocity can be calculated from the local horizontal wavenumber estimated by multi - distance sparse Bayesian learning. Since \(k\) r (or \(c\) p ) is affected by factors such as geo - acoustic parameters, water depth, and sound speed profile, these environmental parameters can be inverted by matching the phase velocity through the sound propagation model. Because multi - distance sparse Bayesian learning involves local modes and local phase velocity, the result of the inversion is the parameters at the receiving point. Since the geo - acoustic parameters are approximately constant within a certain distance range, the inverted geo - acoustic parameters can be regarded as the values in this sea area. The inverted sea depth \(H\) at the receiving array can be used to correct the instrument sounding result. If the sounding is unknown and the sea depth is horizontal, \(H\) is the water depth; if the sounding is known, then in the case of horizontal and varying sea depths, the inverted \(H\) can correct the sounding. Theoretically speaking, the matched phase velocity inversion method is not affected by the terrain and can be used to invert the geo - acoustic parameters of any terrain, and its inversion accuracy is higher than that of the traditional matched sound pressure inversion method affected by terrain changes on the propagation path.

[0086] In the existing methods, since the block sparse Bayesian learning method uses multi-frequency signals to estimate the horizontal wavenumber and requires the use of an approximate dispersion relation, the frequency band needs to be a narrow band, and only the horizontal wavenumber and phase velocity of the narrow band can be estimated. In the multi-distance sparse Bayesian method of the present application, since multi-distance received data is used for each frequency point to independently estimate the horizontal wavenumber, the narrow-band limitation is lifted, and the phase velocity of a wider frequency band can be estimated. We use broadband data to reduce the error influence of the multi-distance sparse Bayesian learning in estimating the horizontal wavenumber of a single frequency point. The received data within a large bandwidth is selected, and the multi-distance sparse Bayesian learning is used for each frequency point to estimate the horizontal wavenumber and phase velocity. Since the phase velocity has a decreasing trend with the increase of frequency, we perform curve fitting on the estimated multi-frequency phase velocities. Then, a series of environmental parameters to be searched are brought into the acoustic propagation model (KRAKEN) to obtain the phase velocity of each frequency point. By minimizing the mean square error (MSE) between the frequency-phase velocity (f-c p ) curve obtained by the acoustic propagation model and the above fitting curve, the ocean environmental parameters are inversed. We call this method the matched-phase-velocity inversion (MPVI), and the formula is shown as follows:

[0087]

[0088] where M d is the total number of modes used; F is the total number of frequencies; represents the set of seabed environmental parameters to be inversed. is the fitted f-c p curve, t m (f i , u) is the f-c p curve of the copy field generated by KRAKEN when the environmental parameter set is u, and f i is the frequency. Since only the real part k p of the complex eigenvalue k c is used when calculating c r , and the imaginary part k i representing the modal attenuation is not included, the calculation process does not involve attenuation. Also, since the k r including the attenuation information is implicit in the modal coefficient X when using the multi-distance sparse Bayesian learning to estimate the horizontal wavenumber k i , and we only use the positions of the non-zero values of X and do not use the values of the non-zero values themselves, so k i is not involved. Therefore, the finally estimated horizontal wavenumber k rIt does not contain information on seafloor attenuation. For the above two reasons, the matching phase velocity inversion method is insensitive to the seafloor attenuation coefficient, that is, the attenuation coefficient is decoupled from other environmental parameters such as seafloor sound speed, density, and water depth.

[0089] Example 1

[0090] Based on the above principle, as Figure 11 shown, the method for inverting geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning includes:

[0091] Step 1: Perform Fourier transform on the acoustic propagation time-domain data at multiple distances received by the hydrophone vertical array to obtain the multi-distance frequency-domain signals at each frequency point within the required bandwidth.

[0092] Step 2: Abundantly collect horizontal wavenumbers within the interval [2πf / c b , 2πf / c w and generate the corresponding local modal depth functions using the shooting method to form matrix A. c b and c w are the sound speeds in the seafloor and water respectively. c b is unknown and can be set to a relatively large value during the execution process. The shooting method formula is as follows:

[0093]

[0094] where ω = 2πf is the angular frequency and f is the frequency; z j = jh, where h and J are the width and total number of discrete grids respectively, and c(z) is the sound speed profile.

[0095] Step 3: Form matrix Y from the multi-distance time-domain signals at each frequency point, and use matrix A and Y through the iterative formula of sparse Bayesian learning to calculate the truly excited local modes and local horizontal wavenumbers.

[0096]

[0097]

[0098] where γ m is the quantity characterizing modal energy, represents the value of this iteration, represents the value of the previous iteration; Φ m is the abbreviation of Φ(z, k rm ); σ 2 is the variance of the noise, Σ y is the data covariance, is the matrix composed of K excited modal functions in A, and + represents the pseudoinverse calculation; I Ndenotes the N-dimensional identity matrix; ||·||2 denotes the 2-norm formula; tr denotes the trace of a matrix.

[0099] Step 4: After performing multi-distance sparse Bayesian learning for each frequency point within the required bandwidth (the frequency range of the required local horizontal wavenumber), convert the obtained local horizontal wavenumber into local phase velocity.

[0100]

[0101] Step 5: Perform data cleaning and curve fitting on the wideband phase velocity points.

[0102] Step 6: Substitute a series of environmental parameters to be searched into the acoustic propagation model (KRAKEN) to obtain the phase velocity of each frequency point. By minimizing the mean square error between the frequency-phase velocity (f-c p ) curve obtained from the acoustic propagation model and the fitting curve in Step 5, invert the environmental parameters including the seabed sound speed, density, and water depth.

[0103]

[0104] Step 7: Perform depth correction on the actually measured water depth with the inverted water depth to reduce the error caused by instrument measurement.

[0105] Step 8: Invert the seabed attenuation coefficient by using the matched propagation loss method, and together with the environmental parameters obtained in Step 6, can be used as the inversion results for matched field construction and subsequent positioning processing. The matched propagation loss method can adopt existing mature methods.

[0106] Embodiment 2

[0107] This application also provides a system for inverting geoacoustic parameters of matched broadband modal phase velocity based on sparse Bayesian learning, which is implemented based on the above method. The system includes:

[0108] A module for obtaining multi-distance frequency-domain signals, which is used to perform Fourier transform on the acoustic propagation time-domain data of multiple distances received by the hydrophone vertical array to obtain multi-distance frequency-domain signals of each frequency point within the required bandwidth;

[0109] A module for obtaining a modal depth function matrix, which is used to collect horizontal wavenumbers within a set interval and generate corresponding local modal depth functions by the shooting method to form a modal depth function matrix A;

[0110] A module for calculating local modes and local horizontal wavenumbers, which is used to form a matrix Y with the multi-distance time-domain signals of each frequency point, and calculate the actually excited local modes and local horizontal wavenumbers by the iterative method of sparse Bayesian learning using matrix A and Y;

[0111] Obtain a local phase velocity module, which is used to perform multi-distance sparse Bayesian learning for each frequency point within the required bandwidth, and convert the obtained local horizontal wavenumber into local phase velocity;

[0112] A fitting curve module, which is used to perform data cleaning and curve fitting on the phase velocity points of the wideband;

[0113] An inversion environmental parameter module, which is used to substitute the environmental parameters to be searched into the acoustic propagation model to obtain the phase velocity of each frequency point; by minimizing the mean square error between the frequency-phase velocity curve obtained from the acoustic propagation model and the fitted curve, the environmental parameters including the seabed sound speed, density, and water depth are inversely obtained;

[0114] A depth correction module, which is used to perform depth correction on the water depth obtained by inversion with the actually measured water depth;

[0115] An inversion seabed attenuation coefficient module, which is used to inversely obtain the seabed attenuation coefficient by using the matched propagation loss method.

[0116] The problem addressed by the present invention is to inversely obtain seabed acoustic parameters and water depth using a vertical array in a shallow sea environment. In simulation and experimental data, the feasibility and accuracy of multi-distance sparse Bayesian learning for extracting local modes and horizontal wavenumbers are verified. A process for inversely obtaining environmental parameters (geophysical parameters and water depth) by matching phase velocities is given, and the inversely obtained parameters are substituted into the matched field positioning model for positioning to verify the reliability of the inversely obtained parameters. In experimental data, a comparison is also made with traditional matched field inversion, and the higher accuracy of the present invention is demonstrated by the positioning results.

[0117] Example 1, Simulation Environment

[0118] To ensure the consistency between simulation and sea trials, in the simulation, we referred to the actual situation of the sea trials (as Figure 1 shown) to set parameters. The transmitting ship conducted a range stretching experiment from the receiving point to the east, and a total of 59 signal flares with a nominal depth of 7 m were launched within 60 km ( Figure 1 right part). The experimental water depth was approximately 40 m, and the seawater sound speed was approximately 1507 m / s. At the receiving point was a 16-element vertical array, distributed in most of the water body (4.44 - 34.65 m). In the simulation, we used the Figure 1 uniform seabed model shown in the left part. Below the water layer with a water depth of H meters was a semi-infinite space, with a seabed sound speed c b , a seabed density ρ b and a seabed sound attenuation coefficient α b three seabed acoustic parameters. Among them, c b was set to 1680 m / s, ρ b was 1.9 g / cm 3 , α b According to the empirical formula αb = 0.29f 1.91 Set at dB / m, H is 40 m. SNR = 27 dB.

[0119] Select a single-frequency signal with a frequency of 250 Hz, and use multi-distance sparse Bayesian learning to calculate the γ curve representing the modal energy. As shown in Fig. 2(a). It can be seen that the peaks of the γ curve are very sharp and have high resolution. The curve has five obvious peaks, corresponding to five horizontal wavenumbers k rm , indicating that five modes are retrieved. The red "×" in the figure represents the true horizontal wavenumber of the sound field generated by KRAKEN in this simulation environment. It can be seen that the horizontal wavenumber values estimated by multi-distance sparse Bayesian learning are relatively consistent with the true values, and the calculated k rm error rates at this time are 0.0372%, 0.0643%, 0.1291%, 0.0187% and 0.0537% respectively, and the average error rate is 0.0606%. The corresponding five-mode depth functions are shown in Fig. 2(b). It can be seen that the mode functions estimated by multi-distance sparse Bayesian learning are in good agreement with the mode functions of the KRAKEN simulation field. It can be seen from the simulation that the multi-distance sparse Bayesian learning method can better estimate the horizontal wavenumber and mode function, with high accuracy.

[0120] Use a broadband signal of 50 - 500 Hz for multi-distance sparse Bayesian learning to estimate the horizontal wavenumber, and calculate the error between the estimated broadband horizontal wavenumber k rm and the true horizontal wavenumber. The frequency-averaged error rates of the fifth-order k rm are 0.1624%, 0.3279%, 0.5929%, 0.9958% and 2.7046% respectively, and the overall average error rate is 0.9567%. It can be seen that the overall error rate of the method for estimating the horizontal wavenumber by multi-distance sparse Bayesian learning is relatively low. The phase velocities at multiple frequency points are finally obtained as follows Figure 3 shown by the black scatter points in, and the red line is the fitting curve. It can be seen that the final fitting result is relatively consistent with the trend of the original data.

[0121] After obtaining the fitting curve, the environmental parameter inversion of the matching phase velocity can be carried out. We invert three parameters: the seabed sound speed c b , density ρ b and sea depth H, and the attenuation is considered known. In the search process of three-dimensional environmental parameters, in order to reduce the amount and improve the inversion efficiency, we use the Bayesian optimization algorithm. Figure 4 It is the process and result diagram of Bayesian optimization. Figure 4Among them, the orange dots are the evaluation points selected in each of the 1,000 iterations, and in that iteration, the expected return corresponding to this set of environmental parameters is the highest. The red star represents the true value of the environmental parameters, and the blue square represents the final parameter inversion result of Bayesian optimization, and this point corresponds to the minimum cost function. During the optimization process, the inversion result of the environmental parameters converges to the true value.

[0122] Finally, the inversion result obtained by the Bayesian optimization algorithm for the matching phase velocity inversion is H = 39.86 m, c b = 1685.0 m / s, ρ b = 1.92 g / cm 3 , which is very close to the true H = 40 m, c b = 1680 m / s, ρ b = 1.9 g / cm 3 . The f-c p curve corresponding to the fitting curve of multi-distance sparse Bayesian learning and the matching inversion result is shown in Figure 5(a) below. The two are very close, indicating that minimizing the MSE can find the curve in the matching field that is closest to the shape of the fitting curve, and thus find the environmental parameters that are closest. Comparing the f-c p curves corresponding to the environmental parameters obtained by the matching inversion and those set in the simulation, as shown in Figure 5(b) below, the two are also very close and almost overlap. It can be seen that the matching broadband phase velocity inversion of environmental parameters has a very good effect and can obtain relatively accurate environmental parameters.

[0123] Substitute the inverted water depth, seabed sound speed, density, and the known attenuation coefficient into KRAKEN to calculate the copy field for MFP positioning, and the frequency band is 100 - 300 Hz. The sound source positioning results are shown in Figures 6(a) and 6(b).

[0124] It can be seen that at this time, both the MFP positioning distance (Figure 6(a)) and depth (Figure 6(b)) are in good agreement with the true values. The average error of distance positioning is 0.23 km, and the average error rate is 0.78%; the average error of depth positioning is 0.55 m, accounting for 1.38% of the 40 m full sea depth. It shows that the inverted seabed parameter results are effective, that is, it demonstrates the feasibility of the matching broadband phase velocity inversion method based on multi-distance sparse Bayesian learning.

[0125] Example 2. Experimental environment

[0126] The experimental environment has been described in Example 1. Since the seabed attenuation coefficient is insensitive to the proposed matching phase velocity method and is decoupled from other parameters, we first use the matching phase velocity to invert the seabed sound speed and density. At this time, the seabed attenuation coefficient is not involved, and then we use the matching propagation loss to invert the attenuation coefficient for subsequent matching field positioning to verify the inversion results.

[0127] Perform multi-distance sparse Bayesian learning within 50 - 500 Hz using the received signals at 59 distances, select the first five-order horizontal wavenumbers, and obtain the corresponding phase velocities as Figure 7 shown. It can be seen that Figure 7 there are obvious fifth-order curves in Figure 7 . The fitting curve is as shown by the red line in

[0128] . Then select the signals in the range of 100 - 450 Hz to perform matching broadband phase velocity inversion for seabed parameters. During the parameter cruise, the Bayesian optimization algorithm is used. The search results are that H is 38.96 m, c b is 1608.97 m / s, and ρ b is 1.97 g / cm³. At this time, draw the f - c p fitting curve (red line) of broadband multi-distance sparse Bayesian learning and the f - c p curve (black line) obtained by substituting the seabed parameter inversion results into KRAKEN as shown in Figure 8 . It can be seen that the similarity between the two is relatively high.

[0129] Use the method of matching propagation loss to invert the attenuation coefficient. Select 59 distance points on the survey line to calculate the propagation loss. Take 8 frequency points within 100 - 350 Hz as the center frequencies of the propagation loss. Match the experimental propagation loss and the propagation loss calculated by KRAKEN to search for the attenuation coefficient (at this time, the water depth, seabed sound speed, and density obtained from the above inversion are substituted into KRAKEN), and obtain the analytical formula of the curve showing the variation of the non-linearly fitted attenuation coefficient with frequency as α b = 0.8390f 1.94 , where the unit of f is kHz.

[0130] The locally inverted water depth is 38.96 m, while the water depth at the location of the VLA measured in the experiment is 38.5 m. Therefore, there is a measurement error of 0.46 m. Add 0.46 m to the overall water depth measured on the survey line as the new water depth. Together with the inverted three seabed acoustic parameters of seabed sound speed, density, and attenuation coefficient, input them into KRAKEN to generate a copy field, and perform broadband matched field localization (distance and depth) for 59 sound source points, with the frequency used being 100 - 300 Hz. The results are shown in Figures 9(a) and 9(b), where the red stars are the true positions of the sound sources and the blue circles are the MFP localization results. It can be seen that the distance localization results basically conform to the GPS recorded results, with an average error of 0.81 km and an average error rate of 2.94%; the depth localization is also basically near the true depth of 7 m, with an average error of 1.46 m, accounting for 3.64% of the total water depth of 40 m. The above results indicate that the seabed sound speed, density, and water depth obtained by inverting the matched phase velocity and the attenuation coefficient obtained by combining the matched propagation loss are relatively reliable, which to a certain extent demonstrates the feasibility and effectiveness of the method of multi - distance sparse Bayesian learning for matching broadband phase velocity to invert environmental parameters.

[0131] For comparison, perform traditional matched field inversion on the received data. Traditional matched field inversion can invert multiple parameters such as water depth, seabed sound speed, density, attenuation, etc. simultaneously, but it is more complex to apply in horizontally varying scenarios. Here, it is roughly considered that the water depth is horizontally invariant, and the inverted water depth is the average depth of the entire area. Perform broadband matched field inversion on 59 distance points respectively. Different from the method proposed in the present invention, traditional matched field inversion requires accurate sound source position information. Therefore, we input the 59 sound source distances recorded by GPS and the nominal depth of 7 m into the inversion. The cost functions of the 59 points are shown in Figure 10(a). It can be seen that when taking the 3rd distance point, the cost function of the inversion is the smallest, reaching 0.2143. At the same time, draw the inversion results of water depth (Figure 10(b)), seabed sound speed (Figure 10(c)), density (Figure 10(d)), and attenuation coefficient (Figure 10(e)) at different distance points respectively. It can be seen that the four parameters are very scattered within the search range. This is because: the solution of matched field inversion has the characteristic of multiple values (that is, the copy fields calculated using different combinations of multi - dimensional parameters will have similar acoustic field characteristics to the measured acoustic field), and matched field inversion is performed separately for each single distance. Therefore, even for adjacent distance points, the inversion results may differ greatly. However, in the same sea area, the changes in the inversion results using received signals at different distances should not be large. Therefore, it is necessary to consider which distance points' inversion results are used as representative final results, and this is usually a problem faced when using matched field inversion. Here, since the inversion results of the parameters at each distance point are relatively scattered, take the seabed parameters corresponding to the minimum cost function as the inversion results. The corresponding seabed parameters are H = 39.90 m, c b= 1601.60 m / s, ρ b = 2.05 g / cm3, α b = 0.13 dB / λ. Similarly, substituting into the KRAKEN model to generate the copy field, the positioning results in Figures 9(c) and 9(d) are obtained. It can be seen that the ranging positioning effect deteriorates at long distances. The average ranging error is 2.28 km, and the average error rate is 6.35%. The depth positioning is better only at short distances and very poor at long distances. The average error is 13.54 m, accounting for 33.86% of the entire 40 m water depth. Traditional matched-field inversion is generally used in an environment with very little horizontal variation. When inverting with a horizontally invariant model, the inversion result is approximately an average result. Therefore, when this model is still used in the case where horizontal variation cannot be ignored, errors will occur. When the result of multi-distance sparse Bayesian learning matching broadband phase velocity inversion is used for matched-field positioning, the depth positioning result is better than that of matched-field inversion, improving the problem of difficult depth determination often faced in MFP. At the same time, the ranging result is also better than the traditional method. Moreover, we need to note that when traditional matched-field inversion of seabed parameters, relatively accurate sound source information is utilized, while the matched phase velocity inversion method does not use sound source information and still achieves good results in subsequent positioning.

[0132] In summary, in the present invention, when multi-distance sparse Bayesian learning extracts normal modes, it does not require geoacoustic parameters or sound source information, making it more simple and efficient. Using the information of multi-distance points at each frequency point to estimate the horizontal wavenumber lifts the bandwidth limitation and allows the use of broadband signals to improve the accuracy. During inversion, the present invention decouples the seabed attenuation coefficient, enabling independent inversion and reducing the interaction between parameters. Matched phase velocity inversion does not require the sound source position information and does not have the inversion accuracy affected by the accuracy of the sound source position as in traditional matched-field inversion. The proposed method simultaneously uses the received data of multi-distance points, not only reducing the large computational amount of individually inverting different distance signals one by one in matched-field inversion but also overcoming the problem that the inversion result of isolated points lacks representativeness. Matched phase velocity inversion matches the local phase velocity and is not affected by the terrain changes on the propagation path, enabling matched phase velocity inversion to perform geoacoustic inversion and depth correction regardless of whether the terrain is horizontally invariant or horizontally varying. On the contrary, matched-field inversion matches the sound pressure and is affected by the propagation path, being sensitive to terrain mismatch, especially when a horizontally invariant model is applied to a horizontally varying seabed environment. These advantages make the matched phase velocity inversion method more robust and accurate than the traditional matched-field inversion method.

[0133] The present application may also provide a computer device, including: at least one processor, a memory, at least one network interface, and a user interface. Each component in the device is coupled together through a bus system. It can be understood that the bus system is used to implement the connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0134] Among them, the user interface may include a display, a keyboard, or a pointing device. For example, a mouse, a trackball, a touchpad, or a touch screen, etc.

[0135] It can be understood that the memory in the disclosed embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM). The memory described herein is intended to include but not be limited to these and any other suitable types of memory.

[0136] In some embodiments, the memory stores the following elements, executable modules, or data structures, or subsets thereof, or extended sets thereof: an operating system and applications.

[0137] Among them, the operating system includes various system programs, such as the framework layer, the core library layer, the driver layer, etc., which are used to implement various basic services and handle hardware-based tasks. The application programs include various application programs, such as the Media Player, the Browser, etc., which are used to implement various application services. The program for implementing the method of the embodiments of the present disclosure may be included in the application programs.

[0138] In the above-mentioned embodiments, the program or instruction stored in the memory may also be called. Specifically, it may be the program or instruction stored in the application program. The processor is used for:

[0139] Executing the steps of the above method.

[0140] The above method may be applied to the processor or implemented by the processor. The processor may be an integrated circuit chip with signal processing capabilities. During the implementation process, each step of the above method may be completed by the integrated logic circuit in the hardware of the processor or the instruction in the form of software. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed above. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. Combining the steps of the above-disclosed method may be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0141] It can be understood that these embodiments described in the present application can be implemented by hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application, or a combination thereof.

[0142] For software implementation, the techniques of the present application can be implemented by executing the functional modules of the present application (such as procedures, functions, etc.). The software code can be stored in a memory and executed by a processor. The memory can be implemented inside or outside the processor.

[0143] The present application can also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, each step in the above method embodiments can be implemented.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the embodiments, those of ordinary skill in the art should understand that any modification or equivalent replacement of the technical solutions of the present application does not depart from the spirit and scope of the technical solutions of the present application, and they should all be covered within the scope of the claims of the present application.

Claims

1. A method for inversing geoacoustic parameters of matching broadband modal phase velocity based on sparse Bayesian learning, comprising: Step 1: Perform Fourier transform on the acoustic propagation time-domain data at multiple distances received by the hydrophone vertical array to obtain multi-distance frequency-domain signals at each frequency point within the required bandwidth; Step 2: Collect horizontal wave numbers within a set interval and use the shooting method to generate corresponding local modal depth functions, forming a modal depth function matrix A; Step 3: Combine the multi-distance time-domain signals at each frequency point into a matrix Y, and use the matrix A and Y through the iterative method of sparse Bayesian learning to calculate the actually excited local modes and local horizontal wave numbers; Step 4: Perform multi-distance sparse Bayesian learning on each frequency point within the required bandwidth, and convert the obtained local horizontal wave numbers into local phase velocities; Step 5: Perform data cleaning and curve fitting on the wide-frequency phase velocity points; Step 6: Substitute the environmental parameters to be searched into the acoustic propagation model to obtain the phase velocity at each frequency point; by minimizing the mean square error between the frequency-phase velocity curve obtained from the acoustic propagation model and the curve fitted in Step 5, inverse the environmental parameters including seabed sound speed, density, and water depth; Step 7: Perform depth correction on the actually measured water depth using the inversed water depth; Step 8: Inverse the seabed attenuation coefficient using the matching propagation loss method.

2. The method for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning according to claim 1, characterized in that The set interval is: [2πf / c b , 2πf / c w ; where f represents the signal frequency, and c b and c w respectively represent the sound speeds in the seabed and in water.

3. The method for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning according to claim 2, wherein The formula of the shooting method is: Φ(0,k rm ) = 0, Φ(z1,k rm ) = h, where ω = 2πf is the angular frequency; z j = jh, where h and J are the width and total number of discrete grids respectively; c(z j ) is the sound speed profile; Φ(z j , k rm ) is the discrete sampling value of the depth function of the m-th normal mode at the depth z j of the j-th grid; k rm represents the horizontal wave number of the m-th order.

4. The method for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning according to claim 3, wherein, The actually excited local modes and horizontal wave numbers are expressed as: where γ m is a quantity representing the local modal energy, represents the value of the current iteration, represents the value of the previous iteration; L represents the number of sound sources; Φ(z,k rm ) is the discrete sampling of the m-th normal mode depth function at depth z at the receiving end; σ 2 is the variance of the noise; Σ y is the data covariance; is a matrix composed of K excited modal functions in matrix A, + represents the pseudo-inverse calculation, and K represents the number of actually existing propagation modes; I N represents the N-dimensional identity matrix, N represents the number of array elements; the superscript H represents the conjugate calculation; ||·||2 represents the 2-norm formula; tr represents taking the trace of the matrix.

5. The method for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning according to claim 4, wherein The formula for converting the obtained local horizontal wave numbers into local phase velocities is: Among them, c pm represents the local phase velocity.

6. The method for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning according to claim 5, characterized in that The formula for inversing the environmental parameters including seabed sound speed, density, and water depth is: Among them, M d is the total number of modes used; F is the total number of frequencies; represents the set of seabed environmental parameters to be inverted; is the fitted frequency-phase velocity curve; t m (f i , u) is the copy field frequency-phase velocity curve generated by the acoustic propagation model when the environmental parameter set is u; f i is the frequency.

7. A system for inversing geoacoustic parameters by matching broadband modal phase velocity based on sparse Bayesian learning, which is implemented based on any one of the methods described in claims 1-6, and is characterized in that The system includes: A module for obtaining multi-distance frequency-domain signals, which is used to perform Fourier transform on the acoustic propagation time-domain data at multiple distances received by the hydrophone vertical array to obtain multi-distance frequency-domain signals at each frequency point within the required bandwidth; A module for obtaining the modal depth function matrix, which is used to collect horizontal wave numbers within a set interval and use the shooting method to generate corresponding local modal depth functions, forming a modal depth function matrix A; A module for calculating local modes and local horizontal wave numbers, which is used to combine the multi-distance time-domain signals at each frequency point into a matrix Y, and use the matrix A and Y through the iterative method of sparse Bayesian learning to calculate the actually excited local modes and local horizontal wave numbers; A module for obtaining local phase velocities, which is used to perform multi-distance sparse Bayesian learning on each frequency point within the required bandwidth and convert the obtained local horizontal wave numbers into local phase velocities; A curve fitting module, which is used to perform data cleaning and curve fitting on the wide-frequency phase velocity points; An environmental parameter inversion module, which is used to substitute the environmental parameters to be searched into the acoustic propagation model to obtain the phase velocity at each frequency point; by minimizing the mean square error between the frequency-phase velocity curve obtained from the acoustic propagation model and the fitted curve, inverse the environmental parameters including seabed sound speed, density, and water depth; A depth correction module, which is used to perform depth correction on the actually measured water depth using the inversed water depth; and A seabed attenuation coefficient inversion module, which is used to inverse the seabed attenuation coefficient using the matching propagation loss method.

Images