A method for testing equivalent parameters of underwater acoustic materials

Abstract

This application relates to the field of acoustic performance testing of underwater acoustic materials, specifically disclosing a method for testing the equivalent parameters of underwater acoustic materials. Utilizing the relationship between the target, environment, and sound field, it aims to obtain the equivalent acoustic parameters of the material under test by accurately constructing a finite-space sound field environment model. First, a machine learning model is used to establish a mapping relationship between the finite-space environment parameters and the sound field. The finite-space environment parameters are then inverted through measured sound fields, achieving accurate construction of the acoustic environment model. Next, the material under test is placed in this finite-space environment. At this point, the equivalent acoustic parameters of the material are the only unknowns. Again, a machine learning model is used to establish a mapping relationship between the sound field and the material's equivalent parameters. The material's equivalent parameters are then inverted through measured sound fields. Finally, the acoustic performance data of the material at each incident angle direction are calculated using the equivalent acoustic parameters.

Description

Technical Field

[0001] This application relates to the field of acoustic performance testing of underwater acoustic materials, and in particular to a method for testing equivalent parameters of underwater acoustic materials. Background Technology

[0002] Underwater acoustic materials are key functional materials for controlling sound wave transmission, and accurate evaluation of their acoustic performance is crucial for accelerating the iteration of sound-absorbing material performance and improving the acoustic concealment of materials. The equivalent acoustic parameters of underwater acoustic materials (such as equivalent density, equivalent sound velocity, and equivalent thickness) are core indicators for measuring their acoustic performance and important input parameters for acoustic structure design, simulation calculations, and performance evaluation. Therefore, accurately obtaining the equivalent acoustic parameters of underwater acoustic materials under service conditions has always been a fundamental issue in the field of underwater acoustic materials research.

[0003] Currently, the testing methods for acoustic parameters of underwater acoustic materials are mainly divided into three categories: the acoustic tube method, the reverberation method, and the free field testing method.

[0004] The acoustic tube method is one of the most widely used testing methods. Its basic principle is to place the material sample at the end of an acoustic tube and obtain the normal acoustic performance parameters of the material by measuring the standing wave ratio or transfer function within the tube. The acoustic tube method has advantages such as relatively simple operation and strong applicability to low frequencies. In a laboratory environment, it can accurately obtain the acoustic impedance and absorption coefficient of a material under normal incident conditions. However, the fundamental limitation of this method is that it can only obtain the acoustic performance under perpendicular (normal) incident conditions and cannot reflect the anisotropic acoustic behavior of the material under oblique incident conditions. In actual underwater service environments, sound waves often incident on the material surface at multiple angles; therefore, the test results of the acoustic tube method cannot comprehensively characterize the actual acoustic performance of the material.

[0005] The reverberation method, based on reverberant sound field testing theory, places the material under test in a reverberant water tank and obtains the average sound absorption coefficient of the material by measuring the change in reverberation time. The advantage of the reverberation method is its ability to simulate a relatively realistic diffuse sound field environment. However, the sound absorption coefficient it measures is a statistical average value for each incident angle, lacking phase information. Therefore, it cannot reflect the differences in the anisotropic acoustic performance of the material under different incident angles, and it is difficult to use for the refined verification of material acoustic models.

[0006] Free-field testing methods typically rely on anechoic water tanks. By measuring the reflected and transmitted signals of a material sample, acoustic performance parameters of the material at different incident angles can be obtained. Theoretically, free-field testing can acquire the most comprehensive acoustic performance data of materials. However, this method is limited by the multipath interference effect caused by reflections from the anechoic water tank walls in the low-frequency range and the diffraction effect at the sample edges. In actual testing, it is difficult to effectively separate the incident and reflected waves, resulting in the need for sufficiently high test frequencies to obtain reliable results. Furthermore, the testing process is complex and requires stringent experimental conditions, limiting its widespread application in the low-frequency range.

[0007] Therefore, it is evident that all three existing methods have their own technical limitations: the acoustic tube method can only obtain normal performance, the reverberation method can only provide angle-averaged results and lacks phase information, and the free-field method is limited by the difficulty of low-frequency testing and the complex testing process. None of these methods can easily and accurately obtain the equivalent acoustic parameters of underwater acoustic materials at multiple incident angles in a confined space environment (such as a typical water tank). Summary of the Invention

[0008] To address the limitations of existing underwater acoustic material parameter testing techniques, which are restricted by single incident direction, angle averaging results, or ideal testing environments, this application provides a method for testing equivalent parameters of underwater acoustic materials.

[0009] The method for testing equivalent parameters of underwater acoustic materials provided in this application adopts the following technical solution: A method for testing equivalent parameters of underwater acoustic materials includes the following steps: Step S1: Establish an initial numerical model of the finite space, and obtain the equivalent environment model of the finite space by inverting the boundary parameters of the finite space through machine learning; Step S2: Place the underwater acoustic material to be tested in the finite space, construct an equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, use the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknown, use a machine learning model to establish the mapping relationship between the equivalent acoustic parameters of the underwater acoustic material to be tested and the sound field, and invert the equivalent acoustic parameters of the underwater acoustic material to be tested through measured sound field data.

[0010] Furthermore, in step S1, the boundary parameters are the wall equivalent parameters of the finite space, including at least the equivalent density, equivalent sound velocity, and equivalent thickness; the geometric parameters and medium parameters of the finite space are obtained through direct measurement.

[0011] Further, step S1 includes: S11: Establish an initial numerical model based on the geometric and medium parameters of the finite space, and use the unknown boundary parameters of the finite space as variables to be inverted; S12: Randomly generate multiple sets of parameter samples within the preset value range of the unknown boundary parameters, substitute each set of parameter samples into the initial numerical model, and generate corresponding multiple sets of sound field simulation data through forward modeling. S13: Using the multiple sets of parameter samples and their corresponding sound field simulation data, train the first machine learning model and establish a nonlinear mapping relationship between boundary parameters and sound field; S14: Obtain measured sound field data in a limited space, input the measured sound field data into the trained first machine learning model, and invert to obtain the boundary parameters; S15: Substitute the boundary parameters obtained from the inversion into the initial numerical model to obtain an equivalent environment model for the finite space.

[0012] Furthermore, the forward modeling calculation employs any one of the finite element method, boundary element method, or finite difference method; the measured sound field data includes sound pressure amplitude data and / or sound pressure phase data at multiple locations within a finite space.

[0013] Further, step S2 includes: S21: Add an equivalent model of the underwater acoustic material to be tested to the numerical model of the equivalent environment model obtained in S1, construct an equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and take the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknown. S22: Randomly generate multiple sets of parameter samples within the preset range of the equivalent acoustic parameters, substitute each set of parameter samples into the equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and generate corresponding multiple sets of sound field simulation data through forward modeling. S23: Using the multiple sets of parameter samples and their corresponding sound field simulation data, train a second machine learning model to establish a nonlinear mapping relationship between the equivalent acoustic parameters and the sound field; S24: Obtain the measured sound field data of the limited space, input the measured sound field data into the trained second machine learning model, and invert to obtain the equivalent acoustic parameters of the underwater acoustic material to be tested.

[0014] Furthermore, the equivalent acoustic parameters of the underwater acoustic material to be tested include at least equivalent density, equivalent sound velocity, and equivalent thickness.

[0015] Furthermore, it also includes step S3: based on the equivalent acoustic parameters of the underwater acoustic material to be tested, the acoustic performance data of the underwater acoustic material to be tested at each incident angle are obtained through simulation calculation.

[0016] Furthermore, in step S3, the simulation calculation adopts the transfer matrix method or the finite element method. By setting periodic boundary conditions and / or a perfectly matched layer at the model boundary, the plane wave excitation at different incident angles is simulated, and the acoustic performance data at each incident angle is calculated based on the sound pressure distribution on the material surface and the incident sound pressure.

[0017] Furthermore, the first machine learning model and / or the second machine learning model is any one of a neural network model, a random forest model, a support vector regression model, or a Gaussian process regression model; the first machine learning model and the second machine learning model adopt the same type or different types of machine learning models.

[0018] Furthermore, the confined space is an underwater acoustic test pool; the equivalent environment model is a two-dimensional model or a three-dimensional model; and the sound field data is acquired by a hydrophone array deployed within the confined space.

[0019] In summary, this application includes the following beneficial technical effects: This application utilizes machine learning to construct a two-layer mapping framework of "environmental parameters-sound field" and "material parameters-sound field." Through a two-step operation of "calibrating the environment first, then inverting the material," the interference of environmental boundary reflections on material parameter inversion is effectively eliminated. In the first round of inversion, all environmental parameters are fixed, and the equivalent acoustic parameters of the underwater acoustic material under test become the only unknowns. Then, a second round of machine learning achieves high-precision inversion of material parameters. If material parameter inversion is performed directly in an unknown or uncalibrated finite space environment, the interference sound field caused by boundary reflections will significantly reduce the inversion accuracy. However, the "fixed environmental variables" prerequisite created by the two-step method in this application achieves synergistic decoupling between environmental and material factors, improving measurement accuracy.

[0020] The two-layer mapping model is both independent (modeling and training independently for environmental parameters and material parameters respectively) and logically related (the equivalent environmental model output by the first layer serves as the fixed input condition for the second layer's inversion). This cascaded machine learning architecture allows this application to flexibly adapt to different testing scenarios: when changing the test pool, only the first round of inversion needs to be re-executed to complete the new environment modeling; when changing the material under test, only the second round of inversion needs to be executed based on the existing environmental model to obtain the material parameters, without the need for repeated environmental calibration. Compared to the existing technology where environmental modeling and material testing are separated and independent processes, this application has stronger adaptability and repeatability.

[0021] Compared to existing technologies such as the acoustic tube method (which only obtains normal acoustic performance), the reverberation method (which only provides angle-averaged results without phase information), and the free-field method (which relies on an ideal anechoic environment and is difficult to test at low frequencies), this application can comprehensively obtain parameters such as equivalent density, equivalent sound velocity, and equivalent thickness of underwater acoustic materials in a confined space environment. Based on the obtained equivalent parameters, the acoustic performance data of the material at various incident angles are further obtained through simulation calculations. Under different sound source locations and different verification paths, the equivalent model and the sound pressure field distribution of the target water tank are highly consistent, and the sound absorption coefficients of the equivalent material and the material under test change with the incident angle in the same way, verifying the inversion accuracy and reliability of this application. This application does not rely on expensive facilities such as anechoic tanks and can be implemented in ordinary water tanks. The testing process is simple and efficient, which can reduce the cost of testing the acoustic performance of underwater acoustic materials.

[0022] The method for equivalent modeling of acoustic environment in confined space and cascade inversion of material parameters proposed in this application is not only applicable to laboratory testing of equivalent parameters of underwater acoustic materials, but can also be extended to the fields of acoustic parameter inversion of fluid media such as marine sediment, analysis of acoustic characteristics of underwater targets, and digital twin modeling of underwater acoustic environment, providing technical support for the study of sound absorption mechanism of underwater acoustic materials and the construction of digital twin system of underwater acoustic environment. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for testing equivalent parameters of underwater acoustic materials according to an embodiment of this application; Figure 2 This is a numerical model diagram of the target water tank in an embodiment of this application; Figure 3 This is an equivalent numerical model diagram of the target water tank in an embodiment of this application; Figure 4 This is a comparison diagram of the sound field distribution along L1(a) and L2(b) of the target water tank (real) and the equivalent simulated water tank (simulation) in the embodiments of this application; Figure 5 This is an equivalent simulated water tank diagram (a) after changing the sound source position in the embodiments of this application, and a comparison diagram of the sound field distribution along L3 (b) between the target water tank (real) and the equivalent simulated water tank (simulation); Figure 6 This is a numerical model diagram of the target water tank containing the material to be tested, according to an embodiment of this application. Figure 7 This is an equivalent numerical model diagram of the target water tank including the equivalent model of the material to be tested in the embodiments of this application; Figure 8 These are sound field distribution diagrams in the water tank of this application embodiment, containing the material to be tested (a) and the equivalent material (b), respectively; Figure 9This is a comparison diagram of the sound pressure distribution along L1(a) and L2(b) under the conditions of the material under test and the equivalent material in the embodiments of this application; Figure 10 This is an equivalent model diagram of the embodiment of this application after changing the position of the sound source; Figure 11 This is a comparison diagram of the sound pressure distribution along L3(a) and L4(b) under the conditions of the material under test and the equivalent material in the embodiments of this application; Figure 12 This is an equivalent model diagram of the embodiment of this application after the sound source position is changed again; Figure 13 This is a comparison diagram of the sound pressure distribution along L5(a) and L6(b) under the conditions of the material under test and the equivalent material in the embodiments of this application; Figure 14 This is a numerical model diagram of the reflection coefficient calculation of the equivalent material (a) and the material to be tested (b) in the embodiments of this application; Figure 15 This is a comparison chart of the reflection coefficients of the equivalent material and the material under test in the embodiments of this application under different incident angles. Detailed Implementation

[0024] The following is in conjunction with the appendix Figure 1-15 This application will be described in further detail.

[0025] First, it should be noted that in real-world application scenarios, the sound field of the target water tank and the sound field of the target water tank containing the hydroacoustic material to be tested are obtained through actual measurements using a hydrophone array deployed within the water tank. To verify the accuracy and effectiveness of the method in this application, this embodiment uses numerical simulation to simulate the above-mentioned measurement process. That is, the forward modeling results of all known parameters of the target water tank and the actual parameters of the material to be tested are used as the measured sound field data for inversion verification. The following uses a two-dimensional fluid model as an example to provide a detailed description of the technical solution of this embodiment.

[0026] This application discloses a method for testing the equivalent parameters of underwater acoustic materials, such as... Figure 1 As shown, it includes the following steps: Step S1: Establish an initial numerical model of the finite space, and obtain an equivalent environment model of the finite space by inverting the boundary parameters of the finite space through machine learning. This specifically includes the following steps: S11: Establish an initial numerical model based on the geometric and medium parameters of the finite space, and use the unknown boundary parameters of the finite space as variables to be inverted.

[0027] In this embodiment, the limited space is a water tank for underwater acoustic testing, so as to... Figure 2 The water tank shown is used as the target water tank. Direct measurement parameters of the water tank are obtained directly using measuring equipment, including: tank width W = 40 m, height H = 40 m, and water density ρ.w =1000 kg / m 3 The speed of sound in water medium c w =1482 m / s. Considering the huge difference in acoustic impedance between the water and air interfaces, an absolutely soft boundary condition is used to model the water-air interface.

[0028] The wall information of the target pool cannot be directly measured by sensors and is therefore an indirect measurement parameter. Figure 3 The equivalent model in the paper presents an equivalent model of the wall of the target pool. The equivalent wall structure consists of two layers: the inner layer (equivalent layer 1) is a finite structure with an equivalent density ρ. e1 Equivalent speed of sound c e1 Equivalent thickness d e1 Three parameters to be inverted; the outer layer (equivalent layer 2) is an infinite structure with an equivalent density ρ. e2 Equivalent speed of sound c e2 Two parameters to be inverted. The specific values ​​of the above five wall equivalent parameters are unknown and need to be obtained by inverting the sound pressure field data using machine learning methods. A perfectly matched layer (PML) is used outside the equivalent wall structure to simulate the non-reflective boundary conditions.

[0029] An initial numerical model is established based on the direct measurement parameters of the target pool, and the above five wall equivalent parameters are used as boundary parameters to be inverted.

[0030] S12: Randomly generate multiple sets of parameter samples within the preset range of unknown boundary parameters, substitute each set of parameter samples into the initial numerical model, and generate corresponding sets of sound field simulation data through forward modeling.

[0031] Specifically, based on prior knowledge of the target pool wall material, the five equivalent parameters (boundary parameters to be inverted) are controlled within the range of variation shown in Table 1.

[0032] Table 1. Variation range of equivalent parameters

[0033] 7000 sets of equivalent wall parameters were randomly generated within the range shown in Table 1, and each set of parameter samples was substituted into the initial numerical model. A cylindrical wave radiation sound source with a sound pressure amplitude of 1 Pa and an operating frequency of 300 Hz was set at the center of the numerical model. Forward modeling was performed using the finite element method to generate 7000 sets of sound field simulation data. Extraction Figure 3 medium shaded area (i.e.) Figure 3 The sound pressure amplitude data of the sound pressure acquisition area (11×11=121 uniformly distributed measurement points) is used as the sound field data.

[0034] S13: Using multiple sets of parameter samples and their corresponding sound field simulation data, train the first machine learning model and establish a nonlinear mapping relationship between boundary parameters and sound field.

[0035] Specifically, an initial neural network model was constructed as the first machine learning model. The number of neurons in its input layer was the same as the number of sound pressure sampling points (121), and the number of neurons in its output layer was the same as the number of wall equivalent parameters to be inverted (5). The number of hidden layers and neurons were optimized and adjusted based on the training results. The neural network was trained using the aforementioned 7000 sets of wall equivalent parameters (labels) and the corresponding sound field data (input) to establish a nonlinear mapping relationship between the wall equivalent parameters and the sound pressure field.

[0036] S14: Obtain measured sound field data in a limited space, input the measured sound field data into the trained first machine learning model, and invert to obtain the boundary parameters.

[0037] It is important to note that in a real water tank environment, the sound field of the target water tank is obtained through actual measurements using a hydrophone array deployed within the tank. In this embodiment, to verify the effectiveness of the method, the actual wall parameters of the target water tank need to be known for numerical simulation: the actual wall has a three-layer structure: Layer 1 (finite structure): density ρ1 = 2400 kg / m³ 3 Sound velocity c1 = 3000 m / s, thickness d1 = 5 m; Layer 2 (finite structure): density ρ2 = 2200 kg / m³ 3 Sound velocity c2 = 3200 m / s, thickness d2 = 5 m; Layer 3 (infinite structure): density ρ3 = 2000 kg / m³ 3 The speed of sound, c3 = 1800 m / s. Place sound sources of the same size and type in the target water tank, such as... Figure 2 At the same locations shown, the sound field of the target water tank was calculated using the finite element method, and sound pressure amplitude data at the same locations (121 measuring points) in the equivalent model were extracted. This set of measured sound pressure data was then input into the first machine learning model, which had already been trained, to invert and obtain the boundary parameters: ρ e1 =2456 kg / m 3 c e1 =3067 m / s, d e1 =9.57 m, ρ e2 =1985kg / m 3 c e2 =1788 m / s.

[0038] S15: Substitute the boundary parameters obtained from the inversion into the initial numerical model to obtain an equivalent environment model in a finite space.

[0039] To verify the accuracy of the equivalent environment model, the sound pressure amplitude distribution along line segments L1 and L2 of the target water tank and the equivalent environment model were extracted and compared. The comparison curves are shown below. Figure 4 (a) and Figure 4 As shown in (b), the two match very well. Furthermore, the sound source position was moved to... Figure 5 (a) shows the location; comparing the sound pressure amplitude distribution along line segment L3, from... Figure 5 (b) shows that even if the location of the sound source is changed, the sound field calculated by the equivalent environment model is still consistent with the sound field of the target water tank. Therefore, the equivalent environment model can simulate the acoustic environment of the target water tank well.

[0040] Step S2: Place the underwater acoustic material to be tested in a limited space, construct an equivalent environment model containing an equivalent model of the underwater acoustic material to be tested, and use the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknowns. Use a machine learning model to establish the mapping relationship between the equivalent acoustic parameters of the underwater acoustic material to be tested and the sound field. Invert the equivalent acoustic parameters of the underwater acoustic material to be tested through measured sound field data. Specifically, this includes the following steps: S21: Add an equivalent model of the underwater acoustic material to be tested to the numerical model of the equivalent environment model obtained in S1, construct an equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and take the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknown.

[0041] Specifically, the underwater acoustic material to be tested is placed in the target water tank, such as... Figure 6 As shown, an equivalent environmental model is constructed that includes an equivalent model of the underwater acoustic material to be tested, such as... Figure 7 As shown. In this embodiment, the equivalent acoustic parameters of the underwater acoustic material to be tested include three unknowns: equivalent density ρ e Equivalent speed of sound c e With equivalent thickness d e In the equivalent environmental model containing the equivalent model of the underwater acoustic material to be tested, the equivalent acoustic parameters of the underwater acoustic material to be tested are the only unknowns, and the environmental parameters (including the equivalent parameters of the pool wall) have all been obtained and fixed by S1 inversion.

[0042] S22: Randomly generate multiple sets of parameter samples within the preset range of equivalent acoustic parameters, substitute each set of parameter samples into the equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and generate corresponding multiple sets of sound field simulation data through forward modeling.

[0043] Specifically, based on prior knowledge of the underwater acoustic material to be tested, the three unknown equivalent acoustic parameters of the underwater acoustic material to be tested are controlled within the range of variation shown in Table 2.

[0044] Table 2. Variation range of material equivalent parameters

[0045] 6000 sets of material equivalent acoustic parameter samples were randomly generated within the range shown in Table 2. Each set of parameter samples was then substituted into... Figure 7 In the equivalent environmental model containing the equivalent model of the underwater acoustic material to be tested, a cylindrical wave radiation sound source is set at the center of the model, with a sound pressure amplitude of 1 Pa and an operating frequency of 300 Hz. Forward modeling is performed using the finite element method, generating 6000 sets of sound field simulation data. The data is then extracted from the shaded area (i.e.,...). Figure 7 The sound pressure amplitude data of 121 measuring points in the sound pressure acquisition area were used as sound field data.

[0046] S23: Using multiple sets of parameter samples and their corresponding sound field simulation data, train a second machine learning model to establish a nonlinear mapping relationship between equivalent acoustic parameters and sound field.

[0047] Specifically, a random forest model is constructed as the second machine learning model. It is trained using the aforementioned 6,000 sets of material equivalent acoustic parameters (labels) and corresponding sound field data (inputs) to establish a nonlinear mapping relationship between the material equivalent acoustic parameters and the sound pressure field.

[0048] S24: Obtain the measured sound field data of the limited space corresponding to the target water tank, input the measured sound field data into the trained second machine learning model, and invert to obtain the equivalent acoustic parameters of the underwater acoustic material to be tested.

[0049] In real-world scenarios, the sound field of the target water tank containing the underwater acoustic material under test is obtained through actual measurements using a hydrophone array. This embodiment, to verify the effectiveness of the method, requires known parameters of the underwater acoustic material under test for numerical simulation: the material's parameters are density ρ = 2300 kg / m³. 3 Sound velocity c = 2800 m / s, thickness d = 5 m. Sound sources of the same size and type are placed in a target water tank containing the underwater acoustic material to be tested. Figure 6 At the same locations shown, the sound field was calculated using the finite element method, and sound pressure amplitude data at the same locations (121 measuring points) in the equivalent environment model were extracted. This set of measured sound pressure data was then input into a pre-trained second machine learning model to invert and obtain the equivalent acoustic parameters of the underwater acoustic material under test: ρ e =2464 kg / m 3 c e =2878 m / s, d e =5 m.

[0050] The sound pressure field distributions of the target water tank (the material under test) and the equivalent environmental model (equivalent material) are extracted, as shown below. Figure 8 (a) and Figure 8As shown in (b), the two have a high degree of similarity. Further extraction of the sound pressure amplitude distribution along line segments L1 and L2 under the conditions of the tested material and the equivalent material is shown in the figures below. Figure 9 (a) and Figure 9 As shown in (b), the sound pressure distribution is basically consistent. Except for the larger error near the valley point, the error of the other points is less than 3 dB.

[0051] To further verify the generalization ability of the inverted material equivalent parameters at different sound source locations, the sound source locations were changed as follows: Figure 10 As shown, when the sound source frequency is 300 Hz, the distribution of the sound pressure amplitude along line segments L3 and L4 under the conditions of the material under test and the equivalent material are as follows: Figure 11 (a) and Figure 11 As shown in (b), the sound pressure distribution is basically consistent, except for a larger error near the valley point, the error at other points is less than 3 dB. The sound source location is changed again as follows... Figure 12 As shown, the distributions of sound pressure amplitude along line segments L5 and L6 under the conditions of the tested material and the equivalent material are respectively as follows: Figure 13 (a) and Figure 13 As shown in (b), the sound pressure distribution is also basically consistent. The above verification results show that the material equivalent acoustic parameters obtained by inversion in this embodiment can accurately characterize the acoustic behavior of the underwater acoustic material under test in a confined space environment.

[0052] Step S3: Based on the equivalent acoustic parameters of the underwater acoustic material to be tested, obtain the acoustic performance data of the underwater acoustic material under each incident angle direction through simulation calculation.

[0053] Finite element calculation model such as Figure 14 As shown in (a), the upper medium of the material is water, the lower medium is concrete, and the top and bottom layers use a perfectly matched layer (PML) to simulate non-reflective boundary conditions. The parameters of the water medium are: density ρ w =1000kg / m 3 Speed ​​of sound c w =1482 m / s; Concrete parameters: density ρ c =2500 kg / m 3 Speed ​​of sound c c =3200 m / s. The left and right boundaries are simulated using Floquet periodic boundary conditions to represent infinitely large periodic boundary conditions.

[0054] A background pressure sound field is set in an aqueous medium to simulate plane waves incident from different directions. The incident sound field satisfies the following formula: P in =e -i(k x ·x+k y ·y) In the formula: i is the imaginary unit, k x k is the wavenumber component in the horizontal direction (along the material surface). x =2πf / c w ·sin(θ0), where x is the horizontal position coordinate; k y k is the wavenumber component in the vertical direction (normal direction of the material surface). y =2πf / c w ·cos(θ0); θ0 is the incident angle, y is the position coordinate in the vertical direction, and f is the operating frequency of the sound source.

[0055] The material's reflectance coefficient R is obtained using the following formula: R=∫ L (P-Pin)dx / ∫ L 1dx In the formula: L is the length of the upper surface of the material; P is the measured sound pressure.

[0056] The calculated curve of the material reflection coefficient as a function of the incident angle is shown below. Figure 15 As shown in the figure, the reflection coefficients of the tested material and the equivalent material exhibit the same variation with the incident angle. Therefore, the equivalent acoustic parameters of the material obtained from the inversion in the equivalent environment model and their simulation results can be used as the true acoustic performance parameters of the tested underwater acoustic material.

[0057] The above verification results show that this method can be implemented in confined spaces such as ordinary water tanks, without relying on ideal environments such as anechoic pools, and without separating incident and reflected waves, and can accurately obtain the equivalent acoustic parameters of underwater acoustic materials and their acoustic performance at various incident angles.

[0058] The above embodiments illustrate the technical solution of this application using a two-dimensional fluid model as an example, but the implementation of this application is not limited thereto. Those skilled in the art should understand that this application is also applicable to three-dimensional models. The setting methods for wall equivalent parameters and material equivalent parameters in a three-dimensional model are similar to those in a two-dimensional model; only the geometric dimensions and measurement point layout of the third dimension need to be added during modeling. In this application, the first machine learning model is not limited to a neural network model, and the second machine learning model is not limited to a random forest model; other machine learning models such as support vector regression and Gaussian process regression can also be used. The first and second machine learning models can be of the same or different types, and the specific choice can be made according to the amount of training data and accuracy requirements. In this application, forward modeling is not limited to the finite element method; numerical calculation methods such as the boundary element method or the finite difference method can also be used. The equivalent environment model constructed in step S1 of this application is not limited to a two-layer wall equivalent structure. Depending on the actual wall structure of the target pool, a single-layer or three-layer equivalent structure can be used. Each equivalent layer independently sets its equivalent density, equivalent sound velocity, and equivalent thickness (as well as optional attenuation parameters). Correspondingly, the number of boundary parameters to be inverted increases with the number of equivalent layers. The simulation calculation in step S3 of this application is not limited to the finite element method; other numerical methods such as the transfer matrix method can also be used to calculate the acoustic performance data of the material at various incident angles. This application is not limited to underwater acoustic testing pool environments; it can also be extended to fields such as acoustic parameter inversion of fluid media such as marine sediments, analysis of underwater target acoustic characteristics, and construction of digital twin systems for underwater acoustic environments.

[0059] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for testing equivalent parameters of underwater acoustic materials, characterized in that: Includes the following steps: Step S1: Establish an initial numerical model of the finite space, and obtain the equivalent environment model of the finite space by inverting the boundary parameters of the finite space through machine learning; Step S2: Place the underwater acoustic material to be tested in the finite space, construct an equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, use the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknown, use a machine learning model to establish the mapping relationship between the equivalent acoustic parameters of the underwater acoustic material to be tested and the sound field, and invert the equivalent acoustic parameters of the underwater acoustic material to be tested through measured sound field data.

2. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: In step S1, the boundary parameters are the wall equivalent parameters of the finite space, including at least the equivalent density, equivalent sound velocity, and equivalent thickness; the geometric parameters and medium parameters of the finite space are obtained through direct measurement.

3. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: Step S1 includes: S11: Establish an initial numerical model based on the geometric and medium parameters of the finite space, and use the unknown boundary parameters of the finite space as variables to be inverted; S12: Randomly generate multiple sets of parameter samples within the preset value range of the unknown boundary parameters, substitute each set of parameter samples into the initial numerical model, and generate corresponding multiple sets of sound field simulation data through forward modeling. S13: Using the multiple sets of parameter samples and their corresponding sound field simulation data, train the first machine learning model and establish a nonlinear mapping relationship between boundary parameters and sound field; S14: Obtain measured sound field data in a limited space, input the measured sound field data into the trained first machine learning model, and invert to obtain the boundary parameters; S15: Substitute the boundary parameters obtained from the inversion into the initial numerical model to obtain an equivalent environment model for the finite space.

4. The method for testing equivalent parameters of underwater acoustic materials according to claim 3, characterized in that: The forward modeling calculation employs any one of the finite element method, boundary element method, or finite difference method; the measured sound field data includes sound pressure amplitude data and / or sound pressure phase data at multiple locations within a finite space.

5. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: Step S2 includes: S21: Add an equivalent model of the underwater acoustic material to be tested to the numerical model of the equivalent environment model obtained in step S1, construct an equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and use the equivalent acoustic parameters of the underwater acoustic material to be tested as the only unknown quantity. S22: Randomly generate multiple sets of parameter samples within the preset range of the equivalent acoustic parameters, substitute each set of parameter samples into the equivalent environment model containing the equivalent model of the underwater acoustic material to be tested, and generate corresponding multiple sets of sound field simulation data through forward modeling. S23: Using the multiple sets of parameter samples and their corresponding sound field simulation data, train a second machine learning model to establish a nonlinear mapping relationship between the equivalent acoustic parameters and the sound field; S24: Obtain the measured sound field data of the limited space, input the measured sound field data into the trained second machine learning model, and invert to obtain the equivalent acoustic parameters of the underwater acoustic material to be tested.

6. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: The equivalent acoustic parameters of the underwater acoustic material to be tested include at least equivalent density, equivalent sound velocity, and equivalent thickness.

7. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: It also includes step S3: based on the equivalent acoustic parameters of the underwater acoustic material to be tested, the acoustic performance data of the underwater acoustic material to be tested under each incident angle direction are obtained through simulation calculation.

8. The method for testing equivalent parameters of underwater acoustic materials according to claim 7, characterized in that: In step S3, the simulation calculation adopts the transfer matrix method or the finite element method. By setting periodic boundary conditions and / or a perfectly matched layer at the model boundary, the plane wave excitation at different incident angles is simulated, and the acoustic performance data at each incident angle is calculated based on the sound pressure distribution on the material surface and the incident sound pressure.

9. A method for testing equivalent parameters of underwater acoustic materials according to claim 3 or 5, characterized in that: The first machine learning model and / or the second machine learning model is any one of a neural network model, a random forest model, a support vector regression model, or a Gaussian process regression model; the first machine learning model and the second machine learning model use the same type or different types of machine learning models.

10. The method for testing equivalent parameters of underwater acoustic materials according to claim 1, characterized in that: The confined space is an underwater acoustic test pool; the equivalent environment model is a two-dimensional or three-dimensional model; the sound field data is acquired by a hydrophone array deployed in the confined space.

Images