High-precision modeling method and system of pool sound field
By inverting the wall parameters of the water tank using a machine learning model, the problem of obtaining wall parameters in water tank acoustic field modeling was solved, achieving high-precision water tank acoustic field modeling, expanding the range of low-frequency acoustic experiments, and supporting the construction of a water tank digital twin system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-12
- Publication Date
- 2026-04-07
AI Technical Summary
In the existing technology, the acoustic environment modeling method of water tank cannot effectively obtain the acoustic parameters of the water tank wall, which limits the testing of low-frequency acoustic characteristics and makes it difficult to achieve high-precision modeling of the sound field of the water tank.
By combining machine learning models with equivalent modeling methods, a numerical model is constructed by directly measuring parameters, and the indirect measured parameters are retrieved using a BP neural network, thus achieving equivalent modeling of the pool wall parameters.
It achieves high-precision modeling of the acoustic field of the water tank, expands the low-frequency range of the acoustic test of the water tank, and provides technical support for the construction of the digital twin system of the water tank.
Smart Images
Figure CN121809233A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of underwater sound field technology, and in particular to a high-precision modeling method and system for sound fields in a water tank. Background Technology
[0002] Acoustic test tanks are important experimental facilities in the field of underwater acoustics, enabling various experiments to verify underwater acoustic mechanisms. However, due to insufficient understanding of the acoustic environment of the test tanks, the frequency range of acoustic experiments based on test tanks is greatly limited.
[0003] For anechoic water tanks, the effective sound absorption frequency range of their walls is above 500Hz, allowing for acoustic testing within this range. However, below 500Hz, experimental results are poor due to boundary reflections. Reverberation water tanks based on reverberation theory offer advantages such as low construction costs and simple experimental operation, but their testing frequencies are limited to below 300Hz due to tank size constraints. Therefore, new testing methods are needed to obtain the low-frequency acoustic characteristics of the target.
[0004] Based on the relationship between the target, environment, and sound field, the characteristics of the sound source can theoretically be obtained by accurately modeling the environment and combining it with the measured sound field. However, the key to modeling the sound field environment of a water tank lies in obtaining the acoustic parameters of the tank wall, and currently, no effective modeling method for the sound field of a water tank has been proposed. Summary of the Invention
[0005] The purpose of this invention is to provide a modeling method and system for making the acoustic environment of a water tank transparent, providing technical support for low-frequency extension based on water tank acoustic experiments and the construction of a digital twin system for water tanks.
[0006] In this invention, parameters affecting the modeling accuracy of the water tank are categorized into directly measured parameters and indirectly measured parameters. The values of directly measured parameters can be directly obtained through measuring equipment and input directly into the numerical model of the water tank. However, the values of indirectly measured parameters (water tank wall parameters) cannot be obtained through conventional methods. They require equivalent modeling to create an equivalent model, and then the equivalent values are obtained through machine learning model inversion. Based on the acquisition of these two types of parameter values, high-precision modeling of the water tank can be achieved.
[0007] In a first aspect, the present invention provides a high-precision modeling method for the sound field of a water tank, comprising the following steps: S1: Acoustic parameter acquisition: Acquire all acoustic parameters of the target water tank and divide them into direct measurement parameters and indirect measurement parameters; S2: Numerical Model Construction: Construct a numerical model of the water tank, input the direct measurement parameter values of the target water tank, and perform equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown; S3: Machine learning model training: Set up a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power being constant; continuously change the equivalent values of the indirect measurement parameters to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; build an initial machine learning model and train the machine learning model based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. S4: Obtaining Equivalent Parameters: Place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water tank. Obtain the measured sound pressure field data in the target water tank. Substitute this sound pressure field data into the trained machine learning model in S3 to obtain the equivalent values of the indirectly measured parameters in the target water tank. Input this equivalent value into the numerical model in S2 to obtain the accurate numerical model of the target water tank.
[0008] Specifically, the direct measurement parameters mentioned in step S1 above include: the shape and size of the pool, the sound velocity and density of the water medium, and the water-air interface parameters. These parameter values can be directly obtained using measuring equipment. The indirect measurement parameters include the pool wall parameters, which cannot be directly measured and need to be obtained through inversion.
[0009] The equivalent modeling method for the indirect measurement parameters (pool wall parameters) mentioned in step S2 above is as follows: the pool wall is modeled using two layers of material, with the inner layer being a finite structure having the following equivalent parameters: density ( ), longitudinal wave velocity ( ), P-wave attenuation ( ), transverse wave velocity ( ), shear wave attenuation ( ) and thickness ( The outer layer is an infinite structure with the following equivalent parameters: density ( ), longitudinal wave velocity ( ), P-wave attenuation ( ), transverse wave velocity ( ), shear wave attenuation ( ), and its outer region uses an infinite domain as a low-reflection boundary condition.
[0010] The machine learning model described in step S3 above is illustrated using a BP neural network as an example. It includes an input layer, an output layer, and a hidden layer. The number of neurons in the input layer is consistent with the number of selected sound pressure measurement points, the number of neurons in the output layer is consistent with the number of equivalent parameters, and the number of layers and neurons in the hidden layer can be adjusted according to different situations.
[0011] The equivalent values of the indirect measurement parameters in the dataset mentioned in step S3 above are obtained by randomly generating a large number of parameter combinations by giving the range of each parameter; the sound pressure data in the dataset are sound pressure amplitude values.
[0012] The measurement point locations of the measured sound pressure field data in step S4 above must be the same as the sound pressure field measurement point locations in the dataset in S3.
[0013] Secondly, the present invention also provides a high-precision modeling system for the sound field of a water tank, comprising: The first acquisition module is used to acquire all acoustic parameters of the target water tank and divide all acoustic parameters into direct measurement parameters and indirect measurement parameters; The construction module is used to build a numerical model of the water tank. It takes the direct measurement parameter values of the target water tank as input and performs equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown. The training module is used to set up a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power being constant; continuously change the equivalent values of the indirect measurement parameters to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; build an initial machine learning model, and train the machine learning model based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. The second acquisition module is used to place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water pool, acquire the measured sound pressure field data in the target water pool, substitute the sound pressure field data into the trained machine learning model to obtain the equivalent value of the indirect measurement parameter in the target water pool, and input the equivalent value into the numerical model to obtain the accurate numerical model of the target water pool.
[0014] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a high-precision modeling method for the sound field of a water tank as described above.
[0015] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the high-precision modeling method for the sound field of a water tank as described above.
[0016] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements a high-precision modeling method for the sound field of a water tank as described above.
[0017] The high-precision modeling method and system for the sound field of a water tank provided by this invention achieves high-precision modeling of the sound field environment of the water tank by treating the water tank wall as a medium combining materials of finite and infinite thickness, and inverting its equivalent parameters through a machine learning model. This can provide technical guidance for the construction of a digital twin system for water tanks. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating the water tank acoustic field modeling method provided by the present invention; Figure 2 This is a schematic diagram of the structure of the target water tank provided by the present invention; Figure 3 This is a schematic diagram of the equivalent structure of the target water tank provided by the present invention; Figure 4 This is a schematic diagram of the structure of the neural network model provided by the present invention; Figure 5 This is a graph showing the relationship between the error between the predicted value and the label value of the neural network model provided by this invention and the number of iterations; Figure 6 This is a comparison diagram of the sound field distribution along L1 and L2 at 300Hz for the target water tank (real) and the equivalent simulated water tank (simulation) provided by this invention. Figure 7 This is a comparison diagram of the sound field distribution along L3 at 300Hz between the target water tank (real) and the equivalent simulated water tank (simulated) after changing the sound source position, provided by the present invention. Figure 8 This is a comparison diagram of the sound field distribution along L4 at 300Hz between the target water tank (real) and the equivalent simulated water tank (simulated) after the sound source position has been changed again, provided by the present invention. Figure 9 This is a comparison diagram of the sound field distribution along L1 of the target water tank (real) and the equivalent simulated water tank (simulation) after changing the sound source frequency to 100Hz, provided by the present invention. Figure 10 This is a comparison diagram of the sound field distribution along L1 of the target water pool (real) and the equivalent simulated water pool (simulation) after changing the sound source frequency to 500Hz, provided by the present invention. Figure 11 This is a comparison diagram of the sound field distribution along L1 of the target water tank (real) and the equivalent simulated water tank (simulation) after changing the sound source frequency to 800Hz, provided by the present invention. Figure 12 This is a schematic diagram of the water tank acoustic field modeling system provided by the present invention; Figure 13 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0021] To address the problems existing in the prior art, this invention has found that the key to modeling the acoustic field environment of a water tank lies in obtaining the acoustic parameters of the water tank wall. Based on machine learning technology, this invention proposes a high-precision modeling method for the acoustic field of a water tank, laying a theoretical foundation for underwater target characteristic experiments based on water tanks.
[0022] Figure 1 This is a flowchart illustrating the water tank acoustic field modeling method provided in an embodiment of the present invention, as shown below. Figure 1 As shown, it includes: Step 100: Acoustic parameter acquisition: Acquire all acoustic parameters of the target water tank and divide them into direct measurement parameters and indirect measurement parameters; Step 200: Numerical Model Construction: Construct a numerical model of the water tank, input the direct measurement parameter values of the target water tank, and perform equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown; Step 300: Machine learning model training: Set up a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power being constant; continuously change the equivalent values of the indirect measurement parameters to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; construct an initial machine learning model, and train the machine learning model based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. Step 400: Obtaining Equivalent Parameters: Place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water tank, and obtain the measured sound pressure field data in the target water tank. Substitute this sound pressure field data into the machine learning model that has been trained in Step 300 to obtain the equivalent values of the indirectly measured parameters in the target water tank. Input this equivalent value into the numerical model in Step 200 to obtain the accurate numerical model of the target water tank.
[0023] Specifically, this embodiment of the invention uses a two-dimensional fluid model as an example, and the structural schematic diagram of the target water pool is as follows. Figure 2 As shown, the purpose of this embodiment is to establish an equivalent numerical model of the target water tank.
[0024] First, acoustic parameters are acquired. The acoustic parameters of the target water tank can be divided into direct and indirect measurement parameters. Direct measurement parameters include: the shape and dimensions of the water tank, the sound velocity and density of the water medium, and the water-air interface parameters. For this target water tank, its shape is as follows: A rectangle; the density of the water medium. =1000kg / m3; the velocity of sound in water is =1482m / s; the water-air interface can be modeled using soft boundary conditions (due to the large impedance difference between water and air).
[0025] The indirect measurement parameters are the wall parameters of the pool. For this pool, the wall consists of three layers: the first two are finite, and the last layer is an infinite structure (the Perfectly Matched Layer (PML) can simulate an infinitely large area). The material parameters of each layer are as follows: Layer 1: Density =2400 kg / m³, speed of sound =3000 m / s, thickness =5m; Layer 2: Density =2200 kg / m³, speed of sound =3200m / s, thickness =5m; Layer 3: Density =2000 kg / m³, speed of sound =1800m / s. These indirect measurement parameters are unavailable when modeling a real water tank, and will not be used in this example when performing an equivalent model of the target water tank.
[0026] The second step is the construction of the numerical model. An equivalent numerical model is built based on the directly measured parameters of the target water tank, such as... Figure 3 As shown, since the wall parameters of the target pool are unknown, an equivalent model needs to be created. The equivalent wall structure consists of two layers: the inner layer is a finite structure with equivalent density. Equivalent speed of sound equivalent thickness Three parameters; the outer layer has an infinite structure and an equivalent density. Equivalent speed of sound Two parameters are required. The specific values of these five parameters are unknown and need to be obtained through inversion of sound pressure field data based on machine learning models.
[0027] Next is the training of the machine learning model. Taking a neural network model as an example, a neural network model can serve as a powerful nonlinear fitting tool to establish the mapping relationship between sound pressure field data and equivalent parameters. In this example, the steps for training the neural network model are as follows: (1) Generation of training dataset The training dataset contains input data and labels. In this example, the input data is the collected sound pressure field data, and the labels are the equivalent parameter data.
[0028] 1) Generation of equivalent parameter dataset The range of equivalent parameters can be roughly determined by prior knowledge. In this example, the range of equivalent parameter variation is set as shown in Table 1. Based on the range of equivalent parameter variation in Table 1, 7000 sets of equivalent parameter data combinations are randomly generated.
[0029] Table 1. Variation range of equivalent parameters
[0030] 2) Generation of sound pressure field dataset The combination of 7000 sets of equivalent parameter data is input sequentially into... Figure 3 In the equivalent numerical model, a cylindrical wave radiation sound source is set at the center of the numerical water tank. The radiated sound pressure is 1 Pa and the frequency is 300 Hz. The radiated sound field is calculated, and the sound pressure field data of the shaded area in the figure are collected (11 The sound pressure amplitudes at 11 uniformly distributed measurement points are calculated. After the calculation is completed, 7000 sets of sound pressure field data and their corresponding equivalent parameter data (labels) can be obtained as the training dataset for the neural network.
[0031] (2) Initial neural network model construction A schematic diagram of the neural network model is shown below. Figure 4 As shown, it includes an input layer, hidden layers, and an output layer. The number of neurons in the input layer is 121, which is the same as the number of sound pressure sampling points. The number of neurons in the output layer is 5, which is the same as the number of equivalent parameters. The number of layers and neurons in the hidden layer can be optimized and adjusted according to the training results.
[0032] (3) Training of neural network models When a labeled dataset is input into an initial neural network, the relationship between the error between the neural network model's predicted values and the labeled values and the number of iterations is as follows: Figure 5 As shown, the error stabilizes when the number of iterations exceeds 1000. At this point, the neural network model has three hidden layers: the first layer has 128 neurons, the second layer has 64 neurons, and the third layer has 16 neurons.
[0033] Finally, to obtain the equivalent parameters, a cylindrical wave radiation sound source is set at the center of the target water tank, with a radiation sound pressure of 1 Pa (the location of the sound source and the radiation sound pressure information must be consistent with those used when the dataset was generated), such as... Figure 2 As shown. And extract 11 [units] at the same location in the target pool. The sound pressure amplitude at a uniformly distributed measuring point is used to input the sound pressure amplitude data into a trained neural network model to obtain the equivalent parameters: =2456 kg / m3; =3.67 m / s; =9.57 m; =1985 kg / m3; =1788 m / s.
[0034] Substitute this equivalent parameter into the equivalent numerical model and calculate its sound field distribution. The sound field distribution along line segment L1 is as follows: Figure 6 As shown in (a) above, the sound field distribution along line segment L1 is as follows: Figure 6 As shown in (b) of the figure, it can be seen from the figure that the sound pressure calculated by the target water pool (real) is quite consistent with the sound pressure field distribution obtained by the equivalent simulated water pool (simulation). The error between the two is larger near the valley value, and the error at other positions does not exceed 2dB.
[0035] To further verify the simulation effect of the equivalent simulated water tank on the target water tank, the position of the sound source was changed as follows: Figure 7 As shown in (a), the distribution of sound pressure amplitude along L3 calculated from the target pool and the equivalent simulated pool was extracted, as shown in (a). Figure 7 As shown in (b) in the figure, the error between the two is still very small.
[0036] Further move the sound source to the lower left corner, such as... Figure 8 As shown in (a), the distribution of the sound pressure amplitude calculated from the target pool and the equivalent simulated pool along L4 was extracted, as shown in (a). Figure 8 As shown in (b) of the figure, the error between the two is still very small. Figures 6-8 The calculation results show that the equivalent simulated water tank can effectively represent the target water tank at 300 Hz.
[0037] To further verify the simulation effect of the equivalent simulated water tank on the target water tank at other frequencies, the sound source frequency was changed to 100Hz, and the calculated sound pressure amplitudes of the target water tank and the equivalent simulated water tank were extracted along [the curve / timeline]. Figure 3 The distribution of L1, such as Figure 9 As shown in the figure, the error between the two is relatively small.
[0038] Furthermore, the sound source frequency was changed to 500Hz, and the sound pressure amplitude calculated from the target water tank and the equivalent simulated water tank was extracted along the […]. Figure 3 The distribution of L1, such as Figure 10 As shown in the figure, the error between the two is relatively small.
[0039] Furthermore, the sound source frequency was changed to 800Hz, and the sound pressure amplitude calculated from the target water tank and the equivalent simulated water tank was extracted along the […]. Figure 3 The distribution of L1, such as Figure 11 As shown in the figure, the overall error of both is still relatively small, but the error is larger near the trough value.
[0040] In summary, the modeling method proposed in this invention can perform high-precision modeling of target water tanks over a wide frequency range.
[0041] The water tank sound field modeling system provided by the present invention is described below. The water tank sound field modeling system described below can be referred to in correspondence with the water tank sound field modeling method described above.
[0042] Figure 12 This is a schematic diagram of the water tank acoustic field modeling system provided by the present invention, as shown below. Figure 12 As shown, it includes: a first acquisition module 1201, a construction module 1202, a training module 1203, and a second acquisition module 1204, wherein: The first acquisition module 1201 is used to acquire all acoustic parameters of the target water tank and divide all acoustic parameters into direct measurement parameters and indirect measurement parameters; the construction module 1202 is used to construct a numerical model of the water tank, input the direct measurement parameter values of the target water tank, and perform equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown; the training module 1203 is used to set a spherical wave sound source in the numerical model of the water tank, the position of which and the radiated sound power are constant; continuously change the equivalent values of the indirect measurement parameters, and acquire a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; An initial machine learning model is built and trained based on a dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. The second acquisition module 1204 is used to place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water pool, acquire the measured sound pressure field data in the target water pool, substitute the sound pressure field data into the trained machine learning model to obtain the equivalent values of the indirect measurement parameters in the target water pool, and input the equivalent values into the numerical model to obtain the accurate numerical model of the target water pool.
[0043] Figure 13 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 13As shown, the electronic device may include: a processor 1310, a communication interface 1320, a memory 1330, and a communication bus 1340. The processor 1310, communication interface 1320, and memory 1330 communicate with each other via the communication bus 1340. The processor 1310 can call logical instructions in the memory 1330 to execute a high-precision modeling method for the acoustic field of a water tank. This method includes: S1: Acoustic parameter acquisition: acquiring all acoustic parameters of the target water tank and classifying them into direct measurement parameters and indirect measurement parameters; S2: Numerical model construction: constructing a numerical model of the water tank, inputting the direct measurement parameter values of the target water tank, and performing equivalent modeling on the indirect measurement parameters, the equivalent parameter values of which are unknown; S3: Machine learning model training: setting a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power remaining constant; continuously changing the equivalent values of the indirect measurement parameters to acquire... A dataset containing equivalent values of indirect measurement parameters and sound pressure field is used. An initial machine learning model is constructed and trained based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of indirect measurement parameters. S4: Equivalent parameter acquisition: A spherical wave sound source with the same radiated sound power as in the numerical model is placed at the same location in the target water pool. The measured sound pressure field data in the target water pool is obtained. This sound pressure field data is substituted into the machine learning model that has been trained in S3 to obtain the equivalent values of the indirect measurement parameters in the target water pool. The equivalent values are then input into the numerical model in S2 to obtain the accurate numerical model of the target water pool.
[0044] Furthermore, the logical instructions in the aforementioned memory 1330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0045] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the high-precision modeling method for the acoustic field of the water tank provided by the above methods. The method includes: S1: Acoustic parameter acquisition: acquiring all acoustic parameters of the target water tank and dividing all acoustic parameters into direct measurement parameters and indirect measurement parameters; S2: Numerical model construction: constructing a numerical model of the water tank, inputting the direct measurement parameter values of the target water tank, and performing equivalent modeling on the indirect measurement parameters, the equivalent parameter values of which are unknown; S3: Machine learning model training: setting a spherical wave in the numerical model of the water tank. The sound source has a constant location and radiated sound power; the equivalent values of the indirect measurement parameters are continuously varied to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; an initial machine learning model is constructed and trained based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters; S4: Equivalent parameter acquisition: A spherical wave sound source with the same radiated sound power as in the numerical model is placed at the same location in the target water pool, and the measured sound pressure field data in the target water pool is obtained. This sound pressure field data is substituted into the machine learning model that has been trained in S3 to obtain the equivalent values of the indirect measurement parameters in the target water pool. The equivalent values are then input into the numerical model in S2 to obtain the accurate numerical model of the target water pool.
[0046] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a high-precision modeling method for the acoustic field of a water tank provided by the above methods. The method includes: S1: Acoustic parameter acquisition: acquiring all acoustic parameters of the target water tank and dividing all acoustic parameters into direct measurement parameters and indirect measurement parameters; S2: Numerical model construction: constructing a numerical model of the water tank, inputting the direct measurement parameter values of the target water tank, and performing equivalent modeling on the indirect measurement parameters, the equivalent parameter values of which are unknown; S3: Machine learning model training: setting a spherical wave sound source in the numerical model of the water tank, the position of which and the radiated sound power are constant. By continuously varying the equivalent values of the indirect measurement parameters, a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field is obtained. An initial machine learning model is constructed, and the machine learning model is trained based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. S4: Equivalent parameter acquisition: A spherical wave sound source with the same radiated sound power as in the numerical model is placed at the same location in the target water pool. The measured sound pressure field data in the target water pool is obtained. This sound pressure field data is substituted into the machine learning model that has been trained in S3 to obtain the equivalent values of the indirect measurement parameters in the target water pool. The equivalent values are then input into the numerical model in S2 to obtain the accurate numerical model of the target water pool.
[0047] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A high-precision modeling method for the sound field of a water tank, characterized in that, Includes the following steps: S1: Acoustic parameter acquisition: Acquire all acoustic parameters of the target water tank and divide all acoustic parameters into direct measurement parameters and indirect measurement parameters; S2: Numerical Model Construction: Construct a numerical model of the water tank, input the direct measurement parameter values of the target water tank, and perform equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown; S3: Machine Learning Model Training: Set up a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power being constant; continuously change the equivalent values of the indirect measurement parameters to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; construct an initial machine learning model, and train the machine learning model based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. S4: Equivalent Parameter Acquisition: Place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water pool, acquire the measured sound pressure field data in the target water pool, substitute the sound pressure field data into the machine learning model that has been trained in S3, obtain the equivalent values of the indirect measurement parameters in the target water pool, input the equivalent values into the numerical model in S2, and obtain the accurate numerical model of the target water pool.
2. The high-precision modeling method for the sound field of a water tank according to claim 1, characterized in that, The direct measurement parameters mentioned in S1 include: the shape and size of the pool, the sound velocity and density of the water medium, and the water-air interface parameters, which are obtained directly using measuring equipment; the indirect measurement parameters include the pool wall parameters, which are obtained through inversion to obtain the pool wall parameters that cannot be directly measured.
3. The high-precision modeling method for the sound field of a water tank according to claim 1, characterized in that, The equivalent modeling method for the indirect measurement parameters described in S2 is as follows: The pool wall is modeled using two layers of material. The inner layer is a finite structure with the following equivalent parameters: density, P-wave velocity, P-wave attenuation, S-wave velocity, S-wave attenuation, and thickness. The outer layer is an infinite structure with the following equivalent parameters: density, P-wave velocity, P-wave attenuation, S-wave velocity, and S-wave attenuation. An infinite domain is used outside the outer layer as a low-reflection boundary condition.
4. The high-precision modeling method for the sound field of a water tank according to claim 1, characterized in that, The machine learning models described in S3 include neural network models, decision trees, random forests, support vector regression, linear regression, K-nearest neighbors, and Naive Bayes.
5. The high-precision modeling method for the sound field of a water tank according to claim 1, characterized in that, The equivalent values of the indirect measurement parameters in the dataset described in S3 are obtained by randomly generating a large number of parameter combinations with a given range for each parameter; the sound pressure data in the dataset are sound pressure amplitude values.
6. The high-precision modeling method for the sound field of a water tank according to claim 1, characterized in that, The measurement points for the measured sound pressure field data described in S4 are the same as those for the sound pressure field measurement points in the dataset in S3.
7. A high-precision modeling system for the sound field of a water tank, characterized in that, Includes the following steps: The first acquisition module is used to acquire all acoustic parameters of the target water tank and divide all acoustic parameters into direct measurement parameters and indirect measurement parameters; The construction module is used to build a numerical model of the water tank. It takes the direct measurement parameter values of the target water tank as input and performs equivalent modeling of the indirect measurement parameters, the equivalent parameter values of which are unknown. The training module is used to set up a spherical wave sound source in the numerical model of the water tank, with its position and radiated sound power being constant; continuously change the equivalent values of the indirect measurement parameters to obtain a dataset containing the equivalent values of the indirect measurement parameters and the sound pressure field; build an initial machine learning model, and train the machine learning model based on the dataset to obtain the mapping relationship between the sound pressure field and the equivalent values of the indirect measurement parameters. The second acquisition module is used to place a spherical wave sound source with the same radiated sound power as in the numerical model at the same location in the target water pool, acquire the measured sound pressure field data in the target water pool, substitute the sound pressure field data into the trained machine learning model to obtain the equivalent value of the indirect measurement parameter in the target water pool, and input the equivalent value into the numerical model to obtain the accurate numerical model of the target water pool.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the high-precision modeling method for the sound field of the water tank as described in any one of claims 1 to 6.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the high-precision modeling method for the sound field of the water tank as described in any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the high-precision modeling method for the sound field of the water tank as described in any one of claims 1 to 6.