Multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation joint data

Abstract

The multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation joint data includes: first, according to the environmental prior information and transceiver configuration of the experimental sea area, a forward model consisting of an acoustic propagation model and a reverberation model is selected to establish a simulation environment; then, the objective function E is selected. PC (m) = E Pro (m)+E Rev (m), where E Pro (m) is the array element coherence objective function; the sensitivity of each parameter at each frequency is then calculated to characterize the difficulty of parameter inversion at the selected frequency; finally, a multi-step inversion strategy is formulated based on the parameter sensitivity, and the final inversion result is obtained through multiple rounds of execution until the objective function converges. The present invention transforms traditional multi-frequency joint data inversion into a multi-step inversion determined by parameter sensitivity, narrowing the parameter search interval after each round, reducing the inversion time. At the same time, after the highly sensitive parameters are determined, the less sensitive parameters are inverted, improving the reliability of the inversion results.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater acoustic environment perception and parameter inversion, and in particular to a multi-step inversion method for seabed acoustic parameters and scattering characteristics based on multi-frequency propagation-reverberation joint data. Background Art

[0002] In an ocean waveguide environment, some sound rays emitted by underwater sound sources are reflected at the interface between the water column and the seabed due to multipath effects. The uneven seabed causes scattering, so the received signal contains seabed information. Acoustic inversion methods, which use signal processing to extract seabed information from received signals, employ inversion methods based on measurement data and forward models of sound propagation and scattering to estimate seabed acoustic parameters (including sediment velocity, density, attenuation coefficient, etc.) and scattering coefficients. Seabed acoustic parameters play a crucial role in predicting sound propagation loss and evaluating sonar performance. The seabed scattering coefficient is crucial for sonar design, sonar performance analysis, and sonar range. It is also a crucial information carrier in seabed acoustic imaging and seabed topography mapping.

[0003] Matched field inversion is the most common acoustic inversion method. Its basic principle is to match data generated by a forward model with measured data and then use optimization algorithms to search for the seafloor acoustic parameters and scattering coefficients with the highest degree of matching in a high-dimensional parameter space. The forward model includes an acoustic propagation model and a reverberation model. Using joint propagation-reverberation data can reduce the impact of inter-parameter correlation on the estimation results. Parameter sensitivity is used to measure the difficulty of parameter inversion; the more sensitive the parameter, the easier the inversion. Different parameters have different sensitivities at the same frequency, resulting in different inversion results. The sensitivity of the same parameter also varies at different frequencies, with some parameters exhibiting higher sensitivity at certain frequencies. Furthermore, the parameter search space has local optima, which vary at different frequencies. Therefore, using multi-frequency joint propagation-reverberation data as the inversion input can avoid the influence of local optima and improve the sensitivity of individual parameters at a single frequency. However, when directly inverting multi-frequency data, parameter sensitivity is averaged over the number of frequencies, which can sometimes result in lower sensitivity of individual parameters when inverting multi-frequency joint data than when inverting at a single frequency. Summary of the Invention

[0004] The present invention aims to address the shortcoming that traditional multi-frequency propagation-reverberation joint data inversion does not fully utilize the high sensitivity of parameters at specific frequencies, and provides a multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation joint data.

[0005] The present invention adopts a multi-step inversion strategy, fully utilizes the high sensitivity characteristics of parameters at specific frequencies and reduces the calculation time of inversion.

[0006] The multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation combined data of the present invention comprises the following steps:

[0007] S1. Establish a forward model;

[0008] According to the prior information of the experimental sea environment and the transceiver configuration, the acoustic propagation model and reverberation model are selected to establish the forward model;

[0009] S2. Select the objective function;

[0010] An objective function is constructed to measure the matching degree between the model data and the measured data.

[0011] For propagation data, the objective function usually selects the array element coherent objective function or the frequency coherent objective function to measure the matching degree between the propagation data calculated by the model and the measured propagation data. When the spectrum of the transmitted signal is unknown, the array element coherent objective function is usually used:

[0012]

[0013] Alternatively, when the uncertainty in the array element positions is large, the objective function for propagating data often uses a frequency-coherent objective function that is insensitive to array shape mismatch:

[0014]

[0015] Where m=[m1 m2...m K ] T represents a parameter vector consisting of K seabed acoustic parameters, (·) T and(·) H Represents transpose and conjugate transpose respectively. N F and N H are the frequency of the received sound field and the number of receiving array elements respectively. Let d n (f i ) indicates that the frequency at the nth array element is f i The measured sound pressure value, Yes N H The array element at frequency f i The measured sound pressure vector at , The nth array element is in N F The measured sound pressure vector at the frequency. Let g n (f i ,m) means that given the parameter vector m, the frequency at the nth array element is f i The copy sound pressure value, is a given parameter vector m, N H The array element at frequency fi The copy sound pressure vector at Given a parameter vector m, the nth element in N F The copied sound pressure vector at the frequency.

[0016] For reverberation data, the objective function is:

[0017]

[0018] Among them, let I n (f i ) represents the frequency f i The measured reverberation intensity at the nth distance point (unit: dB), The frequency is f i Time N R The measured reverberation intensity vector at the distance point. Let s n (f i ,m) represents a given parameter vector m based on the reverberation model, with a frequency of f i The copy reverberation intensity value at the nth distance point (unit: dB), is a given parameter vector m, with a frequency of f i Time N R The vector of copied reverberation strengths at distance points.

[0019] For the joint propagation-reverberation data, the objective function is:

[0020] E PC (m) = E Pro (m)+E Rev (m) (4)

[0021] The objective function of propagation data is selected from array element coherence objective function or frequency coherence objective function according to experimental conditions.

[0022] S3, calculate parameter sensitivity;

[0023] Parameter sensitivity can characterize the difficulty of parameter inversion at a selected frequency and is calculated by the following formula:

[0024]

[0025] Where m represents one of the parameters, m0 is the true value, and Δm represents the parameter offset. A larger SI(m) indicates a faster decrease in the objective function around the true value, a higher sensitivity, and an easier parameter inversion. The parameter sensitivity of each parameter to be inverted at each frequency included in the multi-frequency data is calculated using Equation (5).

[0026] S4, formulate and execute a multi-step inversion strategy;

[0027] Based on the parameter sensitivities calculated in step S3, a multi-step inversion strategy is formulated and executed. Multi-step inversion begins by randomly selecting initial parameter values. The parameters to be inverted are then divided into three steps, sequentially inverted according to their sensitivity, starting with high, medium, and low. The first step uses high-frequency data, the second step uses a mixture of high- and low-frequency data, and the third step uses low-frequency data. After each inversion step, the initial parameter values ​​are updated. These three steps constitute a round, and after each round, the parameter search interval is narrowed based on the inversion results. This process is repeated until the objective function in step S2 converges.

[0028] The present invention addresses the shortcomings of traditional multi-frequency propagation-reverberation joint data inversion, utilizes a multi-step inversion strategy, fully utilizes the high sensitivity characteristics of parameters at specific frequencies, and greatly reduces the calculation time required for inversion without affecting the accuracy of the inversion results.

[0029] Compared with the traditional multi-frequency propagation-reverberation joint data inversion method, the characteristics of the present invention are:

[0030] 1) Make full use of the high sensitivity of parameters at specific frequencies. Parameter sensitivity is used to measure the difficulty of parameter inversion. The more sensitive the parameter, the easier it is to invert. At some specific frequencies, the parameters will show higher sensitivity. The traditional inversion method of multi-frequency propagation-reverberation joint data takes all data as input. The parameter sensitivity calculated according to formula (5) is the average of the parameter sensitivity at all frequencies, that is, during the inversion process, the sensitivity of the parameter is lower than that at a certain frequency in the data set, which may lead to inaccurate estimation results for parameters with low sensitivity. The present invention groups the parameters with higher sensitivity at a specific frequency separately and inverts them step by step, making full use of the high sensitivity of the parameters at a specific frequency. After the high and second-highest sensitive parameters are determined, the low-sensitivity parameters are inverted, and the inversion results are more reliable.

[0031] 2) Reduce computing time by executing multiple steps in a small number of steps. The traditional inversion method for multi-frequency propagation-reverberation joint data uses the inversion results of all parameters as the output of a single program execution. The parameter search space has a high dimensionality, the search interval is large, and the program execution takes a long time. The present invention performs parameter inversion in steps, inverting a few parameters in each step, reducing the dimensionality of the parameter search space. At the same time, the parameter search interval is narrowed after each round of inversion, reducing the interval length, simplifying the optimization algorithm search process, and greatly reducing the computing time required for inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is an overall flow chart of the method of the present invention.

[0033] Figure 2 This is a schematic diagram of the experimental environment.

[0034] Figure 3It is a multi-step inversion strategy based on multi-frequency propagation-reverberation joint data.

[0035] Figure 4 It is a scatter plot of the results of each round of multi-step inversion.

[0036] Figure 5 This is a comparison chart between the traditional multi-frequency joint data inversion results and the multi-frequency joint data step-by-step inversion results. DETAILED DESCRIPTION

[0037] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0038] Reference Figure 1 The specific implementation steps of the multi-step inversion method for seabed acoustic parameters and scattering coefficients from multi-frequency propagation-reverberation joint data are as follows:

[0039] S1. Establish a forward model;

[0040] Assume that the experimental environment is Figure 2 As shown in the figure, before the experiment began, an acoustic propagation model and a reverberation model were selected to establish the simulation environment based on prior information about the water and seabed in the experimental area, as well as the geometric relationship between the sound source and the horizontal linear array (HLA). The parameters to be estimated included seabed acoustic parameters: sediment thickness h, sound velocity c1, attenuation coefficient α1, density ρ1; substrate sound velocity c2, attenuation coefficient α2, density ρ2, and scattering coefficient μ. The experimental transmission frequencies were 50 Hz, 100 Hz, 1000 Hz, and 2000 Hz.

[0041] S2. Select the objective function;

[0042] For propagation data, the objective function usually selects the array element coherent objective function or the frequency coherent objective function to measure the matching degree between the propagation data calculated by the model and the measured propagation data. When the spectrum of the transmitted signal is unknown, the array element coherent objective function is usually used:

[0043]

[0044] Alternatively, when the uncertainty in the array element positions is large, the objective function for propagating data often uses a frequency-coherent objective function that is insensitive to array shape mismatch:

[0045]

[0046] Where m=[m1 m2...m K ] T represents a parameter vector consisting of K seabed acoustic parameters, (·) Tand(·) H Represents transpose and conjugate transpose respectively. N F and N H are the frequency of the received sound field and the number of receiving array elements respectively. Let d n (f i ) indicates that the frequency at the nth array element is f i The measured sound pressure value, Yes N H The array element at frequency f i The measured sound pressure vector at , The nth array element is in N F The measured sound pressure vector at the frequency. Let g n (f i ,m) means that given the parameter vector m, the frequency at the nth array element is f i The copy sound pressure value, is a given parameter vector m, N H The array element at frequency f i The copy sound pressure vector at Given a parameter vector m, the nth element in N F The copied sound pressure vector at the frequency.

[0047] For reverberation data, the objective function is:

[0048]

[0049] Among them, let I n (f i ) represents the frequency f i The measured reverberation intensity at the nth distance point (unit: dB), The frequency is f i Time N R The measured reverberation intensity vector at the distance point. Let s n (f i ,m) represents a given parameter vector m based on the reverberation model, with a frequency of f i The copy reverberation intensity value at the nth distance point (unit: dB), is a given parameter vector m, with a frequency of f i Time N R The vector of copied reverberation strengths at distance points.

[0050] For the joint propagation-reverberation data, the objective function is:

[0051] E PC (m) = E Pro (m)+E Rev (m) (4)

[0052] The objective function of propagation data is selected based on experimental conditions: array element coherence objective function or frequency coherence objective function. Figure 2 In the experimental environment, the transmitted signal spectrum is unknown but the array position information is known, so the array element coherence objective function is selected.

[0053] S3, calculate parameter sensitivity;

[0054] Parameter sensitivity can characterize the difficulty of parameter inversion at a selected frequency and is calculated by the following formula:

[0055]

[0056] Where m represents one of the parameters, m0 is the true value, and Δm is the parameter offset. A larger SI(m) indicates a faster decrease in the objective function around the true value, a higher sensitivity, and easier parameter inversion. Calculate the sensitivity of each parameter to be inverted at each frequency in the multi-frequency data.

[0057] Table 1 Parameter sensitivity of the inverted parameters at various frequencies

[0058]

[0059] S4, formulate and execute a multi-step inversion strategy;

[0060] The multi-step inversion strategy formulated by the parameter sensitivity of each parameter at each frequency in step S3 is as follows: Figure 3 As shown. Perform three rounds until the objective function in step S2 converges. Figure 4 Plot the results of each round in the graph. Figure 5 A comparison of the results of a conventional multi-frequency propagation-reverberation joint data inversion (designated EDS-JI in the figure) and the multi-frequency propagation-reverberation joint data step-by-step inversion (designated MS-JI in the figure) proposed in this invention is presented. While the two estimate similar results, the latter requires 62.5% less time than the former. Therefore, compared to conventional multi-frequency propagation-reverberation joint data inversion, the present invention achieves similar inversion accuracy, yet offers higher reliability and significantly reduced inversion time.

[0061] The present invention discloses a multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation joint data. First, according to the environmental prior information and transceiver configuration of the experimental sea area, a forward model consisting of an acoustic propagation model and a reverberation model is selected to establish a simulation environment; then, according to the actual experimental situation, an objective function is selected to measure the degree of matching between the model calculated data and the measured data; then, the sensitivity of each parameter at each frequency is calculated to characterize the difficulty of inverting the parameter at the selected frequency; finally, a multi-step inversion strategy is formulated based on the parameter sensitivity, and the final inversion result is obtained after multiple rounds of execution until the objective function converges. The present invention converts the traditional multi-frequency joint data inversion into a multi-step inversion determined by parameter sensitivity, and narrows the parameter search interval after each round, thereby reducing the parameter search space and reducing the inversion time. At the same time, after the highly sensitive parameters are determined, the less sensitive parameters are inverted, thereby improving the reliability of the inversion result.

[0062] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The scope of protection of the present invention should not be regarded as limited to the specific forms described in the embodiments. The scope of protection of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A multi-step inversion method for seafloor acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation joint data includes the following steps: S1. Establish a forward model: Based on the prior information of the experimental sea area environment and the transceiver configuration, select the acoustic propagation model and reverberation model to establish the forward model; S2. Select the objective function and construct the objective function to measure the matching degree between the model data and the measured data; S3, calculate parameter sensitivity; S4, formulate and execute a multi-step inversion strategy; Based on the parameter sensitivity calculated in step S3, a multi-step inversion strategy is formulated and executed. In the multi-step inversion, the initial values ​​of the parameters are randomly selected, and the parameters to be inverted are divided into three steps in the order of high, medium and low parameter sensitivity, and inverted in sequence. The first step uses high-frequency data, the second step uses mixed high- and low-frequency data, and the third step uses low-frequency data. The initial values ​​of the parameters are updated after each inversion step. The three steps constitute a round. After each round of inversion, the parameter search range is narrowed according to the inversion results, and the process is repeated until the objective function in step S2 converges. Step S2 specifically includes: For propagation data, the objective function usually selects the array element coherence objective function or the frequency coherence objective function to measure the matching degree between the propagation data calculated by the model and the measured propagation data. When the spectrum of the transmitted signal is unknown, the array element coherence objective function is used: Alternatively, when the uncertainty in the array element positions is large, the objective function for propagating data uses a frequency-coherent objective function that is insensitive to array shape mismatch: Where m=[m1 m2...m K ] T represents a parameter vector consisting of K seabed acoustic parameters, (·) T and(·) H represent transpose and conjugate transpose respectively; N F and N H are the frequency of the received sound field and the number of receiving array elements respectively; let d n (f i ) indicates that the frequency at the nth array element is f i The measured sound pressure value, Yes N H The array element at frequency f i The measured sound pressure vector at , The nth array element is in N F The measured sound pressure vector at the frequency; let g n (f i ,m) means that given the parameter vector m, the frequency at the nth array element is f i The copy sound pressure value, is a given parameter vector m, N H The array element at frequency f i The copy sound pressure vector at Given a parameter vector m, the nth element in N F The copy sound pressure vector at the frequency; For reverberation data, the objective function is: Among them, let I n (f i ) represents the frequency f i The measured reverberation intensity at the nth distance point at time, unit: dB, The frequency is f i Time N R The measured reverberation intensity vector at the distance points; let s n (f i ,m) represents a given parameter vector m based on the reverberation model, with a frequency of f i The copy reverberation intensity value at the nth distance point, unit: dB, is a given parameter vector m, with a frequency of f i Time N R The copy reverberation intensity vector of the distance point; For the joint propagation-reverberation data, the objective function is: HAVE BEEN PC (m)=E Pro (m)+E Rev (m) (4) The objective function of propagation data is selected based on experimental conditions: array element coherence objective function or frequency coherence objective function; Step S3 specifically includes: parameter sensitivity can characterize the difficulty of parameter inversion at the selected frequency, which is calculated by the following formula: Where m represents one of the parameters, m0 is the true value, and △m represents the parameter offset. The larger the SI(m), the faster the objective function decreases near the true value, the higher the sensitivity, and the easier the parameter inversion. The parameter sensitivity of each parameter to be inverted at each frequency included in the multi-frequency data is calculated according to formula (5).

2. The multi-step inversion method for seabed acoustic parameters and scattering coefficients based on multi-frequency propagation-reverberation combined data according to claim 1, characterized in that: Step S4 specifically includes: multi-step inversion first randomly selects the initial value of the parameter, and divides the parameters to be inverted into three steps in the order of parameter sensitivity from high to low, and performs inversion in sequence. The first step uses high-frequency data, the second step uses high- and low-frequency mixed data, and the third step uses low-frequency data. After each step of inversion, the initial value of the parameter is updated; the step is a round, and after each round of inversion, the parameter search range is narrowed according to the inversion result, and the execution is repeated until the objective function in step S2 converges.

Images