A method and system for predicting seabed acoustic characteristics based on physical neural network

Abstract

The present invention discloses a method and system for predicting the acoustic characteristics of seabed sediments based on a physical neural network. The method comprises: obtaining seabed samples from a target sea area; obtaining a seabed sample data set based on seabed sample measurements; training a pre-constructed neural network architecture based on the seabed sample data set, and then optimizing the construction using a loss function to obtain a physical information neural network; wherein the constraints of the loss function include information about a wave equation; and predicting the acoustic characteristic parameters of the seabed sediment to be predicted using a physical information neural network based on the physical parameters and environmental parameters of the seabed sediment to be predicted. The present invention has the beneficial effects of constructing a method for predicting the acoustic characteristics of seabed sediments based on a physical neural network by substituting the wave equation of the acoustic characteristics of the sediment into the physical information neural network and combining it with the constraints of the seabed sediment, thereby being able to accurately predict the acoustic characteristics of the sediment and being widely applicable to the field of marine acoustic technology.

Description

Technical Field

[0001] The present invention relates to the field of marine acoustic technology, and in particular to a method and system for predicting seabed acoustic characteristics based on a physical neural network. Background Art

[0002] The seafloor serves as the lower interface for ocean sound propagation. Its acoustic properties are essential input parameters for the ocean sound field and for evaluating sonar detection effectiveness. Accurately obtaining and predicting these acoustic properties remains a key challenge in current research. Traditional methods for predicting seafloor sediments include empirical equations and theoretical models. Empirical equations are based on a single or two physical parameters, making their input parameters difficult to obtain and incapable of fully representing the acoustic properties of the seafloor. Consequently, prediction accuracy is limited. Theoretical models rely on 13 input parameters, some of which, such as tortuosity and pore size, are difficult to measure, making their application challenging and unsuitable for practical applications. Machine learning algorithms, such as random forests and support vector machines, take training data as input and then directly train the algorithm. The output is a single-frequency prediction of acoustic parameters, such as sound velocity and attenuation. These parameters are only for a single frequency and cannot be simultaneously calculated for different frequencies. This is because training data cannot be simultaneously input for multiple frequencies. Therefore, predicting acoustic properties across multiple frequencies remains a major challenge in current research. Furthermore, traditional methods and common machine learning algorithms are prone to overfitting or large errors. Summary of the Invention

[0003] The present invention aims to at least partially address the limitations of related technologies. To this end, the present invention proposes a method and system for predicting the acoustic properties of seabed sediments based on a physical neural network, which can accurately predict the acoustic properties of seabed sediments.

[0004] In one aspect, an embodiment of the present invention provides a method for predicting the acoustic characteristics of seabed sediments based on a physical neural network, comprising the following steps:

[0005] Obtain seabed samples from the target sea area;

[0006] A seabed sample dataset is obtained based on seabed sample measurements; the seabed sample dataset includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each seabed sample;

[0007] A pre-built neural network architecture is trained based on a seafloor sample dataset, and then a physical information neural network is constructed using a loss function optimization. The loss function is constrained by information about the wave equation.

[0008] Based on the physical parameters and environmental parameters of the seabed sediment to be predicted, the acoustic characteristic parameters of the seabed sediment to be predicted are predicted using a physical information neural network.

[0009] Optionally, obtaining a seabed sample from the target sea area includes the following steps:

[0010] Use seabed sampling equipment to collect multiple seabed sediment column samples from the target sea area as seabed samples.

[0011] Optionally, the seabed sample dataset includes a training set, a validation set, and a test set; obtaining the seabed sample dataset based on seabed sample measurements includes the following steps:

[0012] Based on the preset experimental parameters, acoustic measurement experiments and physical measurement experiments are carried out on all seabed samples;

[0013] Based on acoustic measurement experiments, the acoustic characteristic parameters of each seabed sample at different frequencies are obtained;

[0014] Based on physical measurement experiments, the physical and environmental parameters corresponding to each seabed sample are obtained;

[0015] The experimental data corresponding to all seabed samples are normalized and then divided into training sets, validation sets, and test sets based on preset proportions;

[0016] Among them, the experimental data include acoustic characteristic parameters, physical parameters and environmental parameters.

[0017] Optionally, the seabed sample dataset includes a training set, a validation set, and a test set; training a pre-built neural network architecture based on the seabed sample dataset, and then optimizing the physical information neural network using a loss function, includes the following steps:

[0018] The physical and environmental parameters corresponding to each seabed sample in the training set are used as input data into the neural network architecture to obtain the predicted value of the acoustic characteristics;

[0019] The neural network architecture is pre-built based on a multi-layer fully connected neural network, which includes an input layer, multiple hidden layers, and an output layer connected in sequence. The number of neurons in the input layer is determined based on the data dimensions of physical and environmental parameters, and the number of neurons in each hidden layer decreases gradually according to a preset ratio. The multiple hidden layers are also equipped with activation functions.

[0020] Based on the predicted values ​​of acoustic characteristics, the acoustic characteristic values ​​of the wave equation solution are obtained by solving the wave equation;

[0021] Constructing a physical loss based on the acoustic characteristic prediction value and its corresponding wave equation solution; constructing a data loss based on the acoustic characteristic prediction value and its corresponding acoustic characteristic value;

[0022] The loss function is constructed by weighted summing of data loss and physical loss;

[0023] According to the loss function, the model parameters of the neural network architecture are optimized and adjusted through the back-propagation algorithm and the preset optimizer;

[0024] Based on the optimized and adjusted neural network architecture, the validation set loss is obtained through the validation set;

[0025] The number of iterations is increased by 1, and the step of inputting the physical parameters and environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain the predicted value of the acoustic characteristics is returned until the validation set loss of consecutive preset rounds meets the preset conditions, or the number of iterations reaches the preset maximum training round, and the neural network architecture after the last optimization adjustment is used as the physical information neural network.

[0026] Optionally, the method comprises the following steps:

[0027] Determining constraint boundary conditions based on preset statistical range values ​​of physical parameters and environmental parameters;

[0028] The data in the seabed sample dataset is screened and processed based on the constraint boundary conditions.

[0029] Optionally, the method further comprises the following steps:

[0030] Determine the pore fluid pressure involved in the constitutive relationship based on the constitutive equation of the seabed sediment;

[0031] Based on the pore fluid pressure, the wave equation is obtained by coupling the solid wave equation and the fluid wave equation.

[0032] Optionally, the method further comprises the following steps:

[0033] The test set is used to perform model evaluation operations on the physical information neural network to obtain the model evaluation results.

[0034] On the other hand, an embodiment of the present invention provides a system for predicting seabed acoustic characteristics based on a physical neural network, comprising:

[0035] The first module is used to obtain seabed samples in the target sea area;

[0036] The second module is used to obtain a seabed sample dataset based on seabed sample measurements; the seabed sample dataset includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each seabed sample;

[0037] The third module is used to train the pre-built neural network architecture based on the seabed sample dataset, and then use the loss function to optimize the physical information neural network. The constraints of the loss function include information about the wave equation.

[0038] The fourth module is used to predict the acoustic characteristic parameters of the seabed sediment to be predicted based on the physical parameters and environmental parameters of the seabed sediment to be predicted using the physical information neural network.

[0039] Optionally, the system further includes:

[0040] A fifth module is used to determine constraint boundary conditions based on preset statistical range values ​​of physical parameters and environmental parameters;

[0041] The sixth module is used to filter and process the data in the seabed sample dataset based on constraint boundary conditions.

[0042] Optionally, the system further includes:

[0043] The seventh module is used to determine the pore fluid pressure involved in the constitutive relationship based on the constitutive equation of the seabed sediment;

[0044] The eighth module is used to obtain the wave equation based on the pore fluid pressure by coupling the solid wave equation and the fluid wave equation.

[0045] Optionally, the system further includes:

[0046] The ninth module is used to perform model evaluation operations on the physical information neural network using the test set to obtain the model evaluation results.

[0047] On the other hand, an embodiment of the present invention provides an electronic device, including: a processor and a memory; the memory is used to store programs; the processor executes the program to implement the above-mentioned method for predicting the acoustic characteristics of the seabed based on the physical neural network.

[0048] On the other hand, an embodiment of the present invention provides a computer storage medium storing a program executable by a processor. When the program executable by the processor is executed, it is used to implement the above-mentioned method for predicting the acoustic characteristics of the seabed sediment based on the physical neural network.

[0049] The embodiment of the present invention obtains seabed samples from the target sea area; obtains a seabed sample data set based on the seabed sample measurements; the seabed sample data set includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each seabed sample; trains a pre-constructed neural network architecture based on the seabed sample data set, and then optimizes the construction using a loss function to obtain a physical information neural network; wherein the constraints of the loss function include information about the wave equation; based on the physical parameters and environmental parameters of the seabed sediment to be predicted, the acoustic characteristic parameters of the seabed sediment to be predicted are predicted using the physical information neural network. The present invention includes the following beneficial effects: the present invention constructs a seabed sediment acoustic characteristic prediction method based on a physical neural network by substituting the wave equation of the sediment acoustic characteristic into the physical information neural network, combined with the constraints of the seabed sediment, thereby accurately realizing the prediction of the sediment acoustic characteristic. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation to the technical solution of the present invention.

[0051] Figure 1 This is a schematic diagram of an implementation environment for predicting seabed acoustic characteristics based on a physical neural network, as provided in an embodiment of the present invention;

[0052] Figure 2 This is a flow chart of a method for predicting seabed acoustic characteristics based on a physical neural network provided by an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of an example process of a method for predicting seabed acoustic characteristics based on a physical neural network provided in an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of the structure of a seabed sediment acoustic characteristics prediction system based on a physical neural network provided in an embodiment of the present invention;

[0055] Figure 5 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0057] It should be noted that although the system diagrams illustrate functional module divisions and the flowcharts illustrate a logical sequence, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the system or the sequence in the flowcharts. The terms "first / S100," "second / S200," and the like in the specification, claims, and drawings are used to distinguish similar objects and are not necessarily intended to describe a specific sequence or precedence.

[0058] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0059] It is understandable that the method for predicting the acoustic characteristics of the seabed sediment based on the physical neural network provided in the embodiment of the present invention can be applied to any computer device with data processing and computing capabilities, and this computer device can be various terminals or servers. When the computer device in the embodiment is a server, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network, content distribution network), and big data and artificial intelligence platforms. Optionally, the terminal is a smart phone, tablet computer, laptop computer, desktop computer, etc., but is not limited to this.

[0060] like Figure 1 FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention. Figure 1 , the implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected to the network in a wireless or wired manner to complete data transmission and exchange.

[0061] Server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0062] In addition, server 101 can also be a node server in a blockchain network. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0063] The terminal 102 may be a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited thereto. The terminal 102 and the server 101 may be connected directly or indirectly via wired or wireless communication, which is not limited in this embodiment of the present invention.

[0064] Based on the example Figure 1 In the implementation environment shown, an embodiment of the present invention provides a method for predicting the acoustic characteristics of the seabed sediment based on a physical neural network. The following is explained using the example of applying the method for predicting the acoustic characteristics of the seabed sediment based on a physical neural network to the server 101. It can be understood that the method for predicting the acoustic characteristics of the seabed sediment based on a physical neural network can also be applied to the terminal 102.

[0065] Reference Figure 2 , Figure 2 The flowchart of the method for predicting the acoustic characteristics of the seabed sediment based on a physical neural network applied to a server provided in an embodiment of the present invention is provided. The execution subject of the method for predicting the acoustic characteristics of the seabed sediment based on a physical neural network can be any of the aforementioned computer devices (including servers or terminals). Figure 2 , the method comprises the following steps:

[0066] S100, obtaining seabed samples from the target sea area;

[0067] It should be noted that, in some embodiments, step S100 may include the following steps: using a seabed sampling device to collect multiple seabed sediment columnar samples from the target sea area as seabed samples.

[0068] For example, in some specific embodiments, multiple columnar samples of seabed sediments can be collected from the target sea area using seabed sampling equipment such as gravity samplers and box samplers.

[0069] S200, obtaining a seabed sample dataset based on seabed sample measurements;

[0070] The seabed sample dataset includes the acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each seabed sample; the seabed sample dataset includes a training set, a validation set, and a test set;

[0071] It should be noted that, in some embodiments, step S200 may include the following steps: based on preset experimental parameters, performing acoustic measurement experiments and physical measurement experiments on all seabed samples; obtaining the acoustic characteristic parameters of each seabed sample at different frequencies based on the acoustic measurement experiment; obtaining the physical parameters and environmental parameters corresponding to each seabed sample based on the physical measurement experiment; normalizing the experimental data corresponding to all seabed samples, and then dividing them into training sets, validation sets and test sets based on preset ratios; wherein the experimental data include acoustic characteristic parameters, physical parameters and environmental parameters.

[0072] For example, in some specific implementations, acoustic measurement experiments are carried out on samples in a laboratory environment to obtain acoustic characteristic data such as sound velocity and sound attenuation coefficient of different samples at different frequencies; at the same time, the physical parameters of the samples are measured, such as porosity, density, particle size distribution, etc. In addition, environmental parameters such as offshore distance, sea depth, water temperature, and primary productivity data are recorded. The various types of collected data are normalized and mapped to the [0,1] interval to facilitate subsequent neural network training. The data set is randomly divided into training set, validation set, and test set according to the ratio of 70%, 15%, and 15% (the ratio can be adjusted accordingly according to the actual application conditions).

[0073] S300: Training a pre-built neural network architecture based on a seabed sample dataset, and then optimizing the physical information neural network using a loss function;

[0074] The constraints of the loss function include information about the wave equation; the seabed sample dataset includes a training set, a validation set, and a test set;

[0075] It should be noted that, in some embodiments, step S300 may include the following steps: inputting the physical parameters and environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain the acoustic characteristic prediction value; wherein, the neural network architecture is pre-built based on a multi-layer fully connected neural network, and the multi-layer fully connected neural network includes an input layer, a multi-layer hidden layer and an output layer connected in sequence; the number of neurons in the input layer is determined based on the data dimension of the physical parameters and the environmental parameters, and the number of neurons in each hidden layer in the multi-layer hidden layer gradually decreases according to a preset ratio, and the multi-layer hidden layer is also provided with an activation function; based on the acoustic characteristic prediction value, the acoustic characteristic value of the wave equation solution is obtained by solving the wave equation; according to the acoustic characteristic prediction value and the corresponding wave equation solution, the acoustic characteristic is obtained. The physical loss is constructed based on the acoustic characteristic prediction value and its corresponding acoustic characteristic value; the data loss is constructed based on the weighted sum of the data loss and the physical loss to construct a loss function; based on the loss function, the model parameters of the neural network architecture are optimized and adjusted through the backpropagation algorithm and a preset optimizer; based on the optimized and adjusted neural network architecture, the validation set loss is obtained through the validation set; the number of iterations is increased by 1, and the execution returns to the step of inputting the physical parameters and environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture to obtain the acoustic characteristic prediction value, until the validation set loss of consecutive preset rounds meets the preset conditions, or the number of iterations reaches the preset maximum training round, and the neural network architecture after the last optimization adjustment is used as the physical information neural network. Specifically, the number of iterations is initially 0.

[0076] For example, in some specific implementations, given the nonlinear correlation between the acoustic characteristics of the seabed and numerous complex factors, a multi-layer fully connected neural network is constructed. The number of neurons in the input layer is determined based on the input physical parameters and environmental parameter dimensions, incorporating parameters such as porosity, density, key characteristic values ​​of particle size distribution, offshore distance, water depth, water temperature, salinity, frequency, and primary productivity data. After statistical analysis, the number of neurons in the input layer is determined to be n. 3-4 hidden layers are set in the middle, and the number of neurons in each layer gradually decreases according to a certain ratio, and the ReLU activation function is enabled; the output layer has 2 neurons, which correspond to the prediction of sound speed and sound attenuation coefficient at different frequencies.

[0077] The key idea behind PINN (Physics-Informed Neural Networks) is to use constitutive and wave equation information as constraints, encoding them into a neural network loss function for training. This loss function consists of two components: a data loss and a physical loss.

[0078] Data loss is the error between the model algorithm's prediction and the actual measured acoustic characteristic values. Physical loss is the error between the model algorithm's prediction and the acoustic characteristic values ​​solved by the wave equation. Therefore, the total loss function is a weighted sum of data loss and physical information loss. This total loss function guides model training, ensuring that the model fits real data while also complying with the laws of physics.

[0079] Specifically, during training, the total loss function is minimized using the backpropagation algorithm and the Adam+L-BFGS, SGD, and RMSprop optimizers. At each iteration, the model parameters are updated based on the gradient of the loss function. It is important to note that since the acoustic characteristic values ​​include both sound velocity and sound attenuation, they must simultaneously minimize the total sound velocity loss function and the total sound attenuation loss function at different frequencies during the iteration process. These two mutually constrain each other, thereby improving prediction accuracy.

[0080] Ultimately, through multiple iterations and feedback, physical and data losses are reduced, minimizing the total loss while satisfying the laws of physics, thus ensuring that the acoustic property predictions given by the model are suitable for practical applications.

[0081] In some embodiments, the method may further include the following steps: determining constraint boundary conditions based on preset statistical range values ​​of physical parameters and environmental parameters; and screening and processing data in the seabed sample data set based on the constraint boundary conditions.

[0082] For example, in some specific embodiments, the statistical range values ​​of the porosity, density, median particle size, and average particle size parameters corresponding to different sediment types can be given according to different sediment types. Letter functions are used here to replace them, f_range(porosity, type), f_range(Mz, type), f_range(Md, type), and f_range(density, type). The above four functions are boundary constraints; the frequency variation range is given, f_range(f).

[0083] Specifically, when collecting training data, we ensure that parameters such as porosity, particle size, density, and frequency values ​​within the data fall within the ranges defined by their corresponding functions. This ensures that the data received by the input layer naturally conforms to physical laws, and the acoustic properties predicted by the output layer are more realistic, ensuring that the neural network architecture fundamentally adheres to physical boundaries.

[0084] In some embodiments, the method may further include the following steps: determining the pore fluid pressure involved in the constitutive relationship based on the constitutive equation of the seabed sediment; and obtaining the wave equation by coupling the solid wave equation and the fluid wave equation based on the pore fluid pressure.

[0085] For example, in some specific embodiments, the seabed substrate is regarded as a sediment composed of a solid skeleton and pore fluid, which has both skeleton elastic properties and fluid properties. For an isotropic elastic medium, its constitutive equation only requires two elastic moduli, while for an isotropic pore elastic medium, its constitutive equation relationship expression is as follows:

[0086]

[0087] Where,

[0088] w=β(uU)

[0089] The second constitutive relation involves the pore fluid pressure P f , which is expressed as follows:

[0090]

[0091] The frequency-dependent complex elastic modulus M appears in the above formula.

[0092] When sound waves propagate in the seabed, their wave equation is formed by the coupling of the solid wave equation and the fluid wave equation, so their wave equation is:

[0093]

[0094] The parameters G, β, and ρ represent the shear modulus of the solid framework, the porosity of the bottom sediment, and the overall density of the sediment, respectively. The parameter w is the displacement vector of the interstitial fluid relative to the solid framework. Other parameters include permeability κ, dynamic viscosity η, and the frequency-dependent viscous repair factor F.

[0095] In some embodiments, the method may further include the following steps: performing a model evaluation operation on the physical information neural network using a test set to obtain a model evaluation result.

[0096] For example, in some specific implementations, after stopping training, the test set can be used to evaluate the final model's ability to predict the acoustic properties of the seabed sediment, output the prediction results, and conduct in-depth analysis of the model's performance in different samples and different frequency bands.

[0097] S400: Based on the physical parameters and environmental parameters of the seabed sediment to be predicted, the acoustic characteristic parameters of the seabed sediment to be predicted are predicted using a physical information neural network.

[0098] For example, in some specific implementations, processed data can be sequentially input into a trained PINN model during actual use, ensuring that all parameters are entered in the order and format required by the model. The model prediction program is then activated. Drawing on the complex physical relationships and data patterns learned during training, the model rapidly computes the input data and outputs predicted values ​​for sound velocity and attenuation at different frequencies. Because the PINN model incorporates the constraints of physical equations, the predictions are not simply data fits but also physically plausible.

[0099] In order to explain the principle of the technical solution of the present invention in detail, the overall process of the present invention is described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and cannot be regarded as a limitation of the present invention.

[0100] First of all, it should be explained that, in response to the relevant problems of the existing technology, the present invention proposes a dispersion physical mechanism for analyzing the acoustic characteristics of the seabed sediment, combines the physical neural network PINN algorithm, substitutes the wave equation and constitutive equation of the seabed sediment acoustic characteristics into the PINN algorithm, and combines the boundary conditions of the seabed sediment to construct a seabed sediment acoustic characteristics prediction method based on physical neural network, thereby realizing the prediction of multi-band sediment acoustic characteristics.

[0101] In some specific embodiments, the technical principle of the method for predicting seabed acoustic characteristics based on a physical neural network provided by the present invention can be implemented through the following process steps:

[0102] S1. Analyze the constitutive equation and wave equation of the acoustic characteristics of the seabed:

[0103] The S1-1 seabed sediment is considered to be composed of a solid skeleton and pore fluid. It has both skeletal elastic properties and fluid properties. For an isotropic elastic medium, its constitutive equation only requires two elastic moduli. For an isotropic pore elastic medium, its constitutive equation relationship is expressed as follows:

[0104]

[0105] Where,

[0106] w=β(uU)

[0107] The second constitutive relation involves the pore fluid pressure P f , which is expressed as follows:

[0108]

[0109] The frequency-dependent complex elastic modulus M appears in the above formula.

[0110] When sound waves propagate in the seabed, their wave equation is formed by the coupling of the solid wave equation and the fluid wave equation, so their wave equation is:

[0111]

[0112] The parameters G, β, and ρ represent the shear modulus of the solid framework, the porosity of the bottom sediment, and the overall density of the sediment, respectively. The parameter w is the displacement vector of the interstitial fluid relative to the solid framework. Other parameters include permeability κ, dynamic viscosity η, and the frequency-dependent viscous repair factor F.

[0113] S1-2 gives the constraint boundary conditions:

[0114] According to different sediment types, the statistical range values ​​of porosity, density, median particle size and average particle size parameters corresponding to the sediment type are given. Letter functions are used here to replace them, f_range(porosity, type), f_range(Mz, type), f_range(Md, type), f_range(density, type). The above four functions are boundary constraints; the frequency variation range is given, f_range(f).

[0115] S2. Constructing the physical neural network PINN algorithm model:

[0116] The main idea of ​​S2-1PINN is to use the constitutive equation and wave equation information as constraints and encode them into the neural network loss function for training. The loss function here includes two types: one is data loss and the other is physical loss.

[0117] Data loss is the error between the model algorithm's prediction and the actual measured acoustic characteristic values. Physical loss is the error between the model algorithm's prediction and the acoustic characteristic values ​​solved by the wave equation. Therefore, the total loss function is a weighted sum of data loss and physical information loss. This total loss function guides model training, ensuring that the model fits real data while also complying with the laws of physics.

[0118] When constructing the S2-2 PINN algorithm model, boundary constraints need to be incorporated into the physical loss. Based on the boundary condition input range, the range of acoustic characteristic values ​​of the wave equation solution is given, and then subtracted from the acoustic characteristic values ​​predicted by the physical model to obtain the physical loss. Since the wave equation contains a frequency-dependent viscous repair factor F and a frequency range, f_range(f), the acoustic characteristic parameter values ​​given by the wave equation solution here are frequency-varying values, such as the sound velocity and sound attenuation values. Therefore, during model training, the input label data should also be the measured sound velocity and sound attenuation values ​​at different frequencies. The input training data is digitized data on different factors affecting acoustic characteristics, including offshore distance, multiple particle size parameters, physical parameters, water depth, frequency, and primary productivity data. The predicted values ​​given by the model training are also multi-frequency data that varies with frequency.

[0119] During the training process, S2-3 uses the backpropagation algorithm and the Adam+L-BFGS, SGD, and RMSprop optimizers to minimize the total loss function. At each iteration, the model parameters are updated based on the gradient of the loss function. It is important to note that since the acoustic characteristic values ​​include both sound velocity and sound attenuation, they must simultaneously minimize the total sound velocity loss function and the total sound attenuation loss function at different frequencies during the iteration process. These two constraints constrain each other, thereby improving prediction accuracy.

[0120] Through multiple iterations and feedback, physical loss and data loss are reduced, so that the total loss is minimized while satisfying the physical laws, thus ensuring that the acoustic characteristic prediction values ​​given by the model are suitable for practical applications.

[0121] S3. Model Validation:

[0122] S3-1 uses measured data that did not participate in model training, including sound velocity and sound attenuation data at different frequencies, to verify the trained model. If the verification data cannot meet the prediction accuracy, repeat the above steps to adjust the parameters again and iterate the model training until the model can meet the prediction accuracy when making predictions on data that has never been verified.

[0123] In some specific application scenarios, such as Figure 3 As shown, the method for predicting the acoustic characteristics of seabed sediments based on physical neural networks provided by the present invention can be implemented by the following process:

[0124] Step 1: Seabed sample data collection:

[0125] Using seafloor sampling equipment such as gravity samplers and box samplers, multiple core samples of seafloor sediment were collected from the target sea area. Acoustic measurements were conducted on the samples in a laboratory setting to obtain acoustic properties such as sound velocity and attenuation coefficient at different frequencies. Physical parameters such as porosity, density, and particle size distribution were also measured. Environmental parameters such as offshore distance, sea depth, water temperature, and primary productivity were also recorded. The collected data was normalized and mapped to the [0, 1] interval to facilitate subsequent neural network training. The dataset was randomly divided into training, validation, and test sets, with a ratio of 70%, 15%, and 15%.

[0126] Step 2: Neural network architecture construction:

[0127] Given the nonlinear correlation between the acoustic properties of the seafloor and numerous complex factors, a multi-layer, fully connected neural network was constructed. The number of neurons in the input layer was determined based on a comprehensive set of physical and environmental parameters, including porosity, density, key characteristic values ​​of particle size distribution, offshore distance, water depth, water temperature, salinity, frequency, and primary productivity data. Statistical analysis determined the number of neurons in the input layer to be n. Three to four hidden layers were constructed in the middle, with the number of neurons in each layer decreasing proportionally. The ReLU activation function was used. The output layer consisted of two neurons, one for predicting the speed of sound and the other for predicting the acoustic attenuation coefficient at different frequencies.

[0128] Statistical ranges for porosity, density, median particle size, and average particle size are given for each sediment type. f_range(porosity, type) gives the statistical range of porosity for each sediment type, f_range(Mz, type) and f_range(Md, type) define the median and average particle size ranges, respectively, and f_range(density, type) specifies the density range. Additionally, f_range(f) determines the frequency range. When collecting training data, ensure that parameters such as porosity, particle size, density, and frequency values ​​within the data fall within the ranges defined by their corresponding functions. This ensures that the data received by the input layer naturally conforms to physical laws, and the acoustic properties predicted by the output layer are more realistic, ensuring that the neural network architecture fundamentally adheres to physical boundaries.

[0129] When sound waves propagate in the seabed, their wave equation is formed by the coupling of the solid wave equation and the fluid wave equation, so their wave equation is:

[0130]

[0131] The parameters G, β, and ρ represent the shear modulus of the solid framework, the porosity of the bottom sediment, and the overall density of the sediment, respectively. The parameter w is the displacement vector of the interstitial fluid relative to the solid framework. Other parameters include permeability κ, dynamic viscosity η, and the frequency-dependent viscous repair factor F.

[0132] Using the staggered grid finite difference method, the partial derivatives of the wave field variables are discretized in space, and a second-order precision time integration format is used in the time dimension to rewrite the continuous wave equation as an iterative relationship at discrete spacetime points. The wave field variable values ​​predicted by the neural network are input, and the difference between the two sides of the discretized wave equation is calculated based on the discretized wave equation to obtain the wave equation residual vector. The elements of the wave equation residual vector are squared and accumulated to construct the wave equation loss term L wave ,Right now Where N is the number of discrete space-time points involved in the calculation, i is the residual of the wave equation at the ith discrete point.

[0133] For each frequency point and each set of data, the difference between the acoustic characteristic value predicted by the model and the actual measured acoustic characteristic value is calculated one by one. The differences corresponding to all frequency points and all sample data are squared and then summed to construct the data loss. The commonly used calculation method is the mean square error (MSE), which is expressed as Where N is the total number of training samples, F is the number of frequency points considered, and predicted ij is the predicted acoustic characteristic value of the i-th sample at the j-th frequency point, and the actual ij The corresponding actual measured value.

[0134] Integrate data-driven mean square error loss L data and wave equation loss L wave , construct the final loss function L = ω1L data +ω2L wave . Set appropriate weight values ​​through preliminary experimental adjustments.

[0135] Step 3: Neural network training optimization:

[0136] Neural network training uses optimizers such as Adam+L-BFGS, SGD, and RMSprop. Initially, the Adam optimizer uses its adaptive learning rate to accelerate convergence, while later, L-BFGS is switched to use second-order derivatives to find the global optimum. SGD and RMSprop assist in comparing performance. Each iteration updates model parameters based on the gradient of the loss function.

[0137] After each round of training, the model's performance is comprehensively evaluated on the validation set, closely monitoring the loss function value and prediction accuracy indicators. Particular attention is paid to the prediction errors of sound velocity and sound attenuation. Given that the two are mutually constrained, if either exhibits abnormal fluctuations or persistent non-convergence, the model hyperparameters must be immediately backtested. Training is terminated when the validation set loss does not decrease for five consecutive rounds, or when the preset maximum number of training rounds is reached. The final model's ability to predict the acoustic properties of the seafloor sediment is evaluated on the test set. The prediction results are output and the model's performance is further analyzed for different samples and frequency bands.

[0138] Step 4: Sound velocity and sound attenuation prediction:

[0139] In actual use, the processed data is sequentially fed into the trained PINN model, ensuring that all parameters are entered in the order and format required by the model. The model prediction program is then activated. Drawing on the complex physical relationships and data patterns learned during training, the model rapidly calculates the input data and outputs predicted values ​​for sound velocity and attenuation at different frequencies. Because the PINN model incorporates the constraints of physical equations, the predictions are not simply data fits but also physically plausible.

[0140] Step 5: Model application:

[0141] This model can output the predicted values ​​of sound velocity and sound attenuation at different frequencies. It can provide sound velocity and sound attenuation values ​​at different frequencies according to the application scenario. For example, in ocean acoustic detection, the sound wave frequency of active sonar is around 5kHz. At this time, the sound velocity and sound attenuation values ​​at 5kHz can be provided for sonar detection according to demand and input into the sonar detection model to improve the sonar detection efficiency.

[0142] Other application scenarios, such as shallow layer profiling, multi-beam detection, and scanning sonar, can provide the sound velocity and sound attenuation values ​​of the required frequency band according to the required frequency band and detection accuracy requirements, thereby improving the detection accuracy and range.

[0143] In some feasible application scenarios, the operating frequency band of the shallow marine layer profile is generally in the range of several hundred hertz to several thousand hertz. The specific frequency band may vary depending on different geological exploration purposes and technical means. Common frequency bands include the following:

[0144] 1. Low frequency band (tens of Hz to hundreds of Hz): Used to detect deeper strata and suitable for large-scale geological surveys. It can penetrate thicker sedimentary layers, but has lower resolution.

[0145] 2. Medium frequency band (several hundred hertz to several thousand hertz): suitable for medium-depth stratum detection, can provide clearer profiles, and is often used in shallow geological exploration and resource exploration.

[0146] 3. High frequency band (several thousand hertz to more than ten kilohertz): suitable for detailed detection of shallow strata, and can provide high-resolution geological structure information, but has weak penetration ability. It is usually used to study thinner sedimentary layers or detect specific objects, such as buried objects, pipelines, etc.

[0147] The choice of different frequencies should be determined according to specific survey requirements, geological conditions, and target depth. These detection frequency bands all require the input of the sound velocity and sound attenuation values ​​of the seabed sediments for data analysis. Therefore, the prediction model of the present invention can provide the sound velocity and sound attenuation values ​​of sediments in different frequency bands, thereby improving the quality of data analysis and reducing data analysis work time.

[0148] In summary, the present invention, through the physical neural network PINN algorithm, measures acoustic characteristics at different frequencies, including sound velocity and sound attenuation, on the basis of data, combines the seabed sediment acoustic wave equation and boundary conditions, constructs a total loss function including data loss and physical loss, and minimizes the total loss function through the back propagation algorithm and Adam+L-BFGS, SGD, RMSprop optimizer. At the same time, due to the mutual constraints of the sound velocity and sound attenuation values ​​at different frequencies, the prediction model can meet the measured data fitting on the basis of conforming to the laws of physics, improve the prediction accuracy requirements, so that the constructed prediction model can be put into practical application. The constructed prediction model can simultaneously give the sound velocity and sound attenuation prediction values ​​at different frequencies, thereby meeting the frequency requirements for the acoustic characteristics of the sediment under different application scenarios, expanding the applicability of the method model, and the present invention is widely used in technical fields such as marine science, marine acoustics, and marine resource exploration.

[0149] On the other hand, Figure 4 As shown, an embodiment of the present invention provides a seabed sediment acoustic characteristics prediction system 900 based on a physical neural network, which may include:

[0150] The first module 901 is used to obtain seabed samples in the target sea area;

[0151] The second module 902 is configured to obtain a seabed sample dataset based on seabed sample measurements; the seabed sample dataset includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each seabed sample;

[0152] The third module 903 is used to train the pre-built neural network architecture based on the seabed sample dataset, and then optimize the physical information neural network using a loss function; wherein the constraints of the loss function include information about the wave equation;

[0153] The fourth module 904 is used to predict the acoustic characteristic parameters of the seabed sediment to be predicted using a physical information neural network based on the physical parameters and environmental parameters of the seabed sediment to be predicted.

[0154] In some embodiments, the system may further include:

[0155] A fifth module is used to determine constraint boundary conditions based on preset statistical range values ​​of physical parameters and environmental parameters;

[0156] The sixth module is used to filter and process the data in the seabed sample dataset based on constraint boundary conditions.

[0157] In some embodiments, the system may further include:

[0158] The seventh module is used to determine the pore fluid pressure involved in the constitutive relationship based on the constitutive equation of the seabed sediment;

[0159] The eighth module is used to obtain the wave equation based on the pore fluid pressure by coupling the solid wave equation and the fluid wave equation.

[0160] In some embodiments, the system may further include:

[0161] The ninth module is used to perform model evaluation operations on the physical information neural network using the test set to obtain the model evaluation results.

[0162] The contents of the method embodiments of the present invention are all applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0163] In another aspect, an embodiment of the present invention further provides an electronic device comprising a memory and a processor. The memory stores a computer program, and the processor, when executing the computer program, implements the aforementioned method for predicting seafloor acoustic properties based on a physical neural network. The electronic device can be any intelligent terminal, including a tablet computer and an in-vehicle computer.

[0164] It can be understood that the contents of the above method embodiments are applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0165] like Figure 5 As shown, Figure 5 The hardware structure of an electronic device 1000 according to another embodiment is shown. The electronic device 1000 includes:

[0166] The processor 1001 may be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute relevant programs to implement the technical solutions provided by the embodiments of the present invention.

[0167] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1002 and is called by the processor 1001 to execute the network node population optimization method of the embodiment of the present invention.

[0168] Input / output interface 1003, used to implement information input and output;

[0169] Communication interface 1004, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0170] Bus 1005 , which transmits information between various components of the device (e.g., processor 1001 , memory 1002 , input / output interface 1003 , and communication interface 1004 );

[0171] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0172] The electronic device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one location or distributed across multiple network units. Some or all of these modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0173] The contents of the method embodiments of the present invention are all applicable to the electronic device embodiments. The functions specifically implemented by the electronic device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0174] Another aspect of an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the above method.

[0175] It should be noted that the computer-readable medium shown in the embodiments of the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or component. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0176] The contents of the method embodiments of the present invention are all applicable to the computer-readable storage medium embodiments. The functions specifically implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0177] The present invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the above method.

[0178] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0179] It should be noted that although several modules of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0180] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD to ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present invention.

[0181] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0182] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It will also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skill of an engineer. Therefore, a person skilled in the art will be able to implement the present invention as set forth in the claims using ordinary skill without undue experimentation. It will also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0183] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it 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 the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0184] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution apparatus, device, or apparatus (e.g., a computer-based apparatus, a device including a processor, or other apparatus that can fetch instructions from and execute instructions on an instruction execution apparatus, device, or apparatus). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution apparatus, device, or apparatus.

[0185] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0186] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution device. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0187] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0188] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0189] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for predicting the acoustic characteristics of seabed sediments based on physical neural networks, characterized in that: The following steps are involved: Obtain seabed samples from the target sea area; Obtaining a seabed sample dataset based on the seabed sample measurements; the seabed sample dataset includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each of the seabed samples; Training a pre-built neural network architecture based on the seabed sample dataset, and then optimizing the construction of a physical information neural network using a loss function; wherein the constraints of the loss function include information about the wave equation; The seabed sample dataset includes a training set, a validation set, and a test set; and the training of a pre-built neural network architecture based on the seabed sample dataset and the use of a loss function to optimize the construction of a physical information neural network include the following steps: Inputting the physical parameters and the environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain acoustic characteristic prediction values; The neural network architecture is pre-built based on a multi-layer fully connected neural network, and the multi-layer fully connected neural network includes an input layer, multiple hidden layers, and an output layer connected in sequence; the number of neurons in the input layer is determined based on the data dimensions of the physical parameters and the environmental parameters, and the number of neurons in each hidden layer in the multi-layer hidden layer gradually decreases according to a preset ratio, and the multi-layer hidden layer is also provided with an activation function; Based on the predicted acoustic characteristic value, obtaining an acoustic characteristic value of a wave equation solution by solving the wave equation; Constructing a physical loss based on the acoustic characteristic prediction value and the corresponding acoustic characteristic value by solving the wave equation; constructing a data loss based on the acoustic characteristic prediction value and the corresponding acoustic characteristic value; Performing a weighted summation on the data loss and the physical loss to construct the loss function; According to the loss function, the model parameters of the neural network architecture are optimized and adjusted through a back-propagation algorithm and a preset optimizer; Based on the optimized and adjusted neural network architecture, obtaining a validation set loss through the validation set; Increment the number of iterations by 1, and return to the step of inputting the physical parameters and environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain acoustic characteristic prediction values, until the validation set loss of consecutive preset rounds meets a preset condition, or the number of iterations reaches a preset maximum training round, and the neural network architecture after the last optimization adjustment is used as the physical information neural network; Based on the physical parameters and environmental parameters of the seabed sediment to be predicted, the physical information neural network is used to predict and obtain the acoustic characteristic parameters of the seabed sediment to be predicted.

2. The method for predicting the acoustic characteristics of seabed sediments based on a physical neural network according to claim 1, wherein: The method of obtaining seabed samples from the target sea area includes the following steps: A plurality of columnar samples of seabed sediments are collected from the target sea area using seabed sampling equipment as the seabed samples.

3. The method for predicting the acoustic characteristics of seabed sediments based on a physical neural network according to claim 1, wherein: The seabed sample dataset includes a training set, a validation set, and a test set; and obtaining the seabed sample dataset based on the seabed sample measurements includes the following steps: Based on preset experimental parameters, acoustic measurement experiments and physical measurement experiments are performed on all the seabed samples; Obtaining the acoustic characteristic parameters of each seabed sample at different frequencies based on the acoustic measurement experiment; Obtaining the physical parameters and the environmental parameters corresponding to each of the seabed samples based on the physical measurement experiment; Normalizing the experimental data corresponding to all the seabed samples, and then dividing them into a training set, a validation set, and a test set based on a preset ratio; The experimental data includes the acoustic characteristic parameters, the physical parameters and the environmental parameters.

4. The method for predicting the acoustic characteristics of seabed sediments based on a physical neural network according to claim 1, wherein: The method comprises the following steps: Determining constraint boundary conditions based on preset statistical range values ​​of the physical parameter and the environmental parameter; The data in the seabed sample data set is screened based on the constraint boundary conditions.

5. The method for predicting the acoustic characteristics of seabed sediments based on a physical neural network according to claim 1, wherein: The method further comprises the following steps: Determine the pore fluid pressure involved in the constitutive relationship based on the constitutive equation of the seabed sediment; Based on the pore fluid pressure, the wave equation is obtained by coupling a solid wave equation and a fluid wave equation.

6. The method for predicting the acoustic characteristics of seabed sediments based on a physical neural network according to claim 1, wherein: The method further comprises the following steps: A model evaluation operation is performed on the physical information neural network using the test set to obtain a model evaluation result.

7. A system for predicting seabed acoustic characteristics based on physical neural networks, characterized in that: include: The first module is used to obtain seabed samples in the target sea area; The second module is configured to obtain a seabed sample dataset based on the seabed sample measurements; the seabed sample dataset includes acoustic characteristic parameters, physical parameters, and environmental parameters corresponding to each of the seabed samples; A third module is configured to train a pre-built neural network architecture based on the seabed sample dataset, and then optimize the construction of a physical information neural network using a loss function; wherein the constraints of the loss function include information about the wave equation; The seabed sample dataset includes a training set, a validation set, and a test set; and the training of a pre-built neural network architecture based on the seabed sample dataset and the use of a loss function to optimize the construction of a physical information neural network include the following steps: Inputting the physical parameters and the environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain acoustic characteristic prediction values; The neural network architecture is pre-built based on a multi-layer fully connected neural network, and the multi-layer fully connected neural network includes an input layer, multiple hidden layers, and an output layer connected in sequence; the number of neurons in the input layer is determined based on the data dimensions of the physical parameters and the environmental parameters, and the number of neurons in each hidden layer in the multi-layer hidden layer gradually decreases according to a preset ratio, and the multi-layer hidden layer is also provided with an activation function; Based on the predicted acoustic characteristic value, obtaining an acoustic characteristic value of a wave equation solution by solving the wave equation; Constructing a physical loss based on the acoustic characteristic prediction value and the corresponding acoustic characteristic value by solving the wave equation; constructing a data loss based on the acoustic characteristic prediction value and the corresponding acoustic characteristic value; Performing a weighted summation on the data loss and the physical loss to construct the loss function; According to the loss function, the model parameters of the neural network architecture are optimized and adjusted through a back-propagation algorithm and a preset optimizer; Based on the optimized and adjusted neural network architecture, obtaining a validation set loss through the validation set; Increment the number of iterations by 1, and return to the step of inputting the physical parameters and environmental parameters corresponding to each seabed sample in the training set as input data into the neural network architecture for processing to obtain acoustic characteristic prediction values, until the validation set loss of consecutive preset rounds meets a preset condition, or the number of iterations reaches a preset maximum training round, and the neural network architecture after the last optimization adjustment is used as the physical information neural network; The fourth module is used to predict the acoustic characteristic parameters of the seabed sediment to be predicted using the physical information neural network based on the physical parameters and environmental parameters of the seabed sediment to be predicted.

8. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 6.

9. A computer storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 6 when executed by the processor.

Images