Underwater sound propagation loss prediction method and system based on feedforward neural network

Abstract

The invention provides an underwater sound propagation loss prediction method and system based on a feedforward neural network, and belongs to the technical field of underwater sound waves, and the method comprises the steps: carrying out the dimensionality reduction and classification of historical marine environment data through employing a principal component analysis method and a K-means algorithm, constructing training data, carrying out the offline training of an FNN model, and deploying the trained FNN model into a UUV system. And the propagation loss of the UUV and the underwater platform can be quickly calculated when a small amount of in-situ measurement data is received. Compared with a Bellhop model, the method has the advantages that the calculation precision is ensured, the occupation of calculation resources is remarkably reduced, the calculation rate is greatly improved, and an efficient and accurate prediction technology is provided for the underwater communication of the UUV.

Description

Technical Field

[0001] The present invention relates to the field of underwater sound wave technology, and in particular to a method and system for predicting underwater sound propagation loss based on a feedforward neural network. Background Art

[0002] Unmanned underwater vehicles (UUVs) are widely used in tasks such as marine environment detection, underwater reconnaissance and search, and underwater communication due to their convenient operation, low price, and high flexibility. Due to the particularity of the underwater environment, the energy attenuation of electromagnetic waves and light waves is large when propagating in water. Hydroacoustic sound is a common means of communication and data transmission between UUV and underwater platforms. During hydroacoustic communication, the sound propagation loss is affected by factors such as the sound velocity profile of the underwater environment, resulting in different propagation losses of UUV at different locations, thereby affecting the communication quality between UUV and underwater platforms. In addition, due to the limitations of UUV size, load, and cost, the supporting hydroacoustic communication system is required to be small in size and light in weight. Therefore, it is crucial to use limited computing resources to quickly and accurately calculate the propagation loss of communication between the sea area where the UUV is located and the underwater platform based on in-situ measurement data, and select a propagation loss that is less than the communication propagation loss threshold to determine the optimal communication position of the UUV.

[0003] At present, the main models for calculating sound propagation loss are simplified underwater sound propagation models and ocean waveguide sound propagation models. Simplified underwater sound propagation models include spherical wave models and cylindrical wave models. Although these models are simple to calculate, they do not take into account the changes in the ocean environment and cannot accurately describe the impact of the complex ocean environment on sound wave propagation. The propagation path and attenuation characteristics of sound waves are difficult to accurately predict, and the calculation accuracy is low. Ocean waveguide sound propagation models include ray models, simple normal wave models, and parabolic equation models. The corresponding acoustic software includes Bellhop, Kraken, and RAM. Although they take into account the complex ocean environment, these models usually take a long time to calculate and require high computing resources. Summary of the invention

[0004] The present invention provides a method and system for predicting underwater sound propagation loss based on a feedforward neural network, which are used to solve the defects of the underwater sound propagation detection model used in the prior art, such as low accuracy in obtaining environmental characteristics, low calculation accuracy, and complex calculation. By using fewer computing resources to achieve high-precision and low-complexity calculation of underwater sound propagation loss, the optimal communication position between an unmanned underwater vehicle and an underwater platform can be quickly predicted.

[0005] In a first aspect, the present invention provides a method for predicting underwater sound propagation loss based on a feedforward neural network, comprising: Acquire historical underwater acoustic environment data, convert the historical underwater acoustic environment data into a sound velocity profile, and use principal component analysis and K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; Constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; Deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; The in-situ measurement data of the unmanned underwater vehicle is obtained, the in-situ measurement data is input into the underwater sound propagation loss prediction model after dimension reduction, and the underwater sound propagation loss prediction result is output.

[0006] According to a method for predicting underwater sound propagation loss based on a feedforward neural network provided by the present invention, the historical underwater acoustic environment data is converted into a sound velocity profile, and the sound velocity profile is reduced in dimension and classified using a principal component analysis method and a K-means clustering algorithm K-means to obtain model training data, including: Determine a sound velocity matrix corresponding to the historical underwater acoustic environment data, and obtain an eigenvector matrix of the sound velocity matrix; Calculating a reduced-dimensional sound velocity profile according to the eigenvector matrix; Clustering the sound velocity profile after dimensionality reduction using a K-means algorithm to obtain a sound velocity profile category; Calculating the sound velocity profile category through the Bellhop model to obtain the propagation loss; Binarizing the propagation loss according to the propagation loss threshold value to obtain final propagation loss data; The reduced-dimensional sound velocity profile, the distance between the unmanned underwater vehicle and the underwater platform, the underwater platform depth, the underwater unmanned vehicle depth and the propagation loss threshold value are used as model input data, and the final propagation loss data are used as model output data.

[0007] According to a method for predicting underwater sound propagation loss based on a feedforward neural network provided by the present invention, a sound velocity matrix corresponding to the historical underwater acoustic environment data is determined, and a eigenvector matrix of the sound velocity matrix is ​​obtained, including: Determining that the sound velocity matrix includes different interpolations of a sound velocity profile and sound velocities at different times at the same depth; A disturbance matrix is ​​obtained from the sound speed matrix and the sound speed matrix mean; Calculating the covariance matrix of the disturbance matrix, performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue matrix and an original eigenvector matrix; The eigenvalues ​​in the eigenvalue matrix are sorted, the cumulative variance is calculated from the sorted eigenvalues, and the eigenvectors corresponding to the eigenvalues ​​whose cumulative variance is greater than a preset percentage are determined to form the eigenvector matrix.

[0008] According to a method for predicting underwater sound propagation loss based on a feedforward neural network provided by the present invention, the K-means algorithm is used to cluster the reduced-dimensional sound velocity profile to obtain sound velocity profile categories, including: Step 1, dividing the sound velocity profile after dimensionality reduction into several categories, obtaining the center of any category, and randomly determining several category centers as cluster center point sets; Step 2, calculate the minimum Euclidean distance of all sound velocity profiles from the center of each category, and divide each number of sound velocity profiles into the category closest to the center point; Step 3, obtaining the number of any type of sound speed profiles, and calculating the average value of each type of sound speed profile as the new center point; Step 4: Calculate the objective function according to each type of sound speed profile, the new center point and the number of sound speed profile samples, and calculate the difference between the two adjacent objective functions in any number of divisions. If it is determined that the objective function difference is less than the difference threshold, clustering is completed and the sound speed profile category is obtained. Otherwise, iterative calculation is started again from step 2.

[0009] According to a method for predicting underwater sound propagation loss based on a feedforward neural network provided by the present invention, a feedforward neural network is constructed, and model training is performed on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model, including: Determine that the feedforward neural network consists of an input layer, a hidden layer and an output layer, the input layer includes 7 nodes, the hidden layer includes two layers, and the first hidden layer includes nodes, and the second hidden layer includes nodes, the output layer includes 1 node; Determining that the two hidden layers include a sigmoid activation function, and the output layer includes a softmax activation function; Determine the preset training accuracy and the preset training times; The training data is forward propagated according to the network structure of the feedforward neural network, the loss function is calculated using the minimum mean square error according to the output result and the target value, and the network weights and biases are adjusted using the gradient descent method until the preset training accuracy or the preset number of training times is reached, thereby obtaining the underwater sound propagation loss prediction model.

[0010] According to a method for predicting underwater sound propagation loss based on a feedforward neural network provided by the present invention, in-situ measurement data of the unmanned underwater vehicle is obtained, the in-situ measurement data is input into the underwater sound propagation loss prediction model after dimension reduction, and the underwater sound propagation loss prediction result is output, including: Acquire a sound source level by transmitting a sound source from an underwater platform, and acquire a sonar detection threshold of the unmanned underwater vehicle, a sonar receiving array directivity index, and an environmental noise calculated by a sensor; Calculate the sound source level, the sonar detection threshold, the sonar receiving array directivity index and the ambient noise based on the passive sonar equation to obtain a propagation loss threshold; The sensor of the unmanned underwater vehicle obtains the current position of the unmanned underwater vehicle, the depth of the unmanned underwater vehicle and the distance between the unmanned underwater vehicle and the underwater platform; Acquiring a current sound speed profile through a sensor of the unmanned underwater vehicle, and calculating the current sound speed profile after dimensionality reduction; The current sound speed profile after dimensionality reduction, the depth of the unmanned underwater vehicle, the distance between the unmanned underwater vehicle and the underwater platform, and the depth of the underwater platform are input into the underwater sound propagation loss prediction model, and the sound propagation loss of the area where the unmanned underwater vehicle and the underwater platform are located is calculated, and the communicative area is determined by the sound propagation loss.

[0011] In a second aspect, the present invention further provides an underwater sound propagation loss prediction system based on a feedforward neural network, comprising: A construction module is used to obtain historical underwater acoustic environment data, convert the historical underwater acoustic environment data into a sound velocity profile, and use a principal component analysis method and a K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; A training module, used for constructing a feedforward neural network, performing model training on the feedforward neural network based on the model training data, and obtaining an underwater sound propagation loss prediction model; A deployment module, used for deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; The prediction module is used to obtain the in-situ measurement data of the unmanned underwater vehicle, reduce the dimension of the in-situ measurement data and input it into the underwater sound propagation loss prediction model, and output the underwater sound propagation loss prediction result.

[0012] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements any of the above-described methods for predicting underwater sound propagation loss based on a feedforward neural network.

[0013] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for predicting underwater sound propagation loss based on a feedforward neural network.

[0014] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for predicting underwater sound propagation loss based on a feedforward neural network.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The underwater sound propagation loss prediction method and system based on feedforward neural network provided by the present invention reduce the dimension and classify the historical underwater acoustic environment data by using principal component analysis and K-means algorithm, construct training data, train the feedforward neural network model offline, and deploy the trained feedforward neural network model to the UUV system, so that the propagation loss of the area where the UUV and the underwater platform are located can be quickly calculated after receiving the in-situ measurement data. Compared with the existing calculation model, while ensuring the calculation accuracy, the occupation of calculation resources is significantly reduced, and the calculation speed is greatly improved, providing a new efficient and accurate prediction technology for ensuring the underwater communication accuracy of the UUV. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the underwater sound propagation loss prediction method based on the feedforward neural network provided by the present invention; Figure 2 It is a flow chart of feedforward neural network model training provided by the present invention; Figure 3 It is a flow chart of deployment of the feedforward neural network model provided by the present invention; Figure 4 It is a comparison chart of underwater sound propagation loss prediction provided by the present invention; Figure 5 It is a structural schematic diagram of the underwater sound propagation loss prediction system based on the feedforward neural network provided by the present invention; Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] In view of the defects of the prior art, the present invention proposes a method for predicting underwater sound propagation loss based on a feedforward neural network (FNN) to improve the reliability of UUV underwater communication.

[0020] Figure 1 FIG. 1 is a flow chart of a method for predicting underwater sound propagation loss based on a feedforward neural network provided in an embodiment of the present invention. Figure 1 As shown, including: Step 100: Acquire historical underwater acoustic environment data, convert the historical underwater acoustic environment data into a sound velocity profile, and use principal component analysis and K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; Step 200: constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; Step 300: deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; Step 400: Acquire in-situ measurement data of the unmanned underwater vehicle, reduce the dimension of the in-situ measurement data and input it into the underwater sound propagation loss prediction model, and output the underwater sound propagation loss prediction result.

[0021] Specifically, the underwater sound propagation loss prediction method based on a feedforward neural network proposed in the embodiment of the present invention is implemented in the following steps: Figure 2 The training process shown in Figure 3 The deployment process shown.

[0022] The first step is to obtain historical underwater acoustic environment data, usually taking the marine environment as the research object to obtain corresponding data, and processing to obtain historical sound speed profile data. The sound speed profile is reduced in dimension and classified by principal component analysis (PCA) and K-means algorithm to obtain the sound speed profile data after dimensionality reduction and clustering. The model training data is combined with the UUV depth, the horizontal distance from the underwater platform, the underwater platform depth and the calculated propagation loss threshold. Then, the constructed FNN model is trained with the model training data to obtain an underwater sound propagation loss prediction model. After the model training is completed, the model is deployed on the UUV, and the sound speed profile is reconstructed and reduced in dimension according to the acquired sound speed information, and then input into the underwater sound propagation loss prediction model to obtain the underwater sound propagation loss prediction result, and determine the optimal communication position between the UUV and the underwater platform.

[0023] By using principal component analysis and K-means algorithm to reduce the dimension and classify historical underwater acoustic environment data, construct training data, and train the feedforward neural network model offline, the trained feedforward neural network model is deployed to the UUV system, so that it can quickly calculate the propagation loss in the area where the UUV and the underwater platform are located after receiving a small amount of in-situ measurement data. Compared with the existing measurement model, while ensuring the calculation accuracy, it significantly reduces the occupation of computing resources and greatly improves the calculation speed, providing a new, efficient and accurate prediction technology for the underwater communication accuracy of UUV.

[0024] Based on the above embodiment, step 100 includes: Determine a sound velocity matrix corresponding to the historical underwater acoustic environment data, and obtain an eigenvector matrix of the sound velocity matrix; Calculating a reduced-dimensional sound velocity profile according to the eigenvector matrix; Clustering the sound velocity profile after dimensionality reduction using a K-means algorithm to obtain a sound velocity profile category; Performing sound propagation calculation on the sound velocity profile category by using the Bellhop model to obtain the propagation loss; Binarizing the propagation loss according to the propagation loss threshold value to obtain final propagation loss data; The reduced-dimensional sound velocity profile, the distance between the unmanned underwater vehicle and the underwater platform, the underwater platform depth, the underwater unmanned vehicle depth and the propagation loss threshold value are used as model input data, and the final propagation loss data are used as model output data.

[0025] Specifically, the embodiment of the present invention constructs FNN model training data including: The first step is to calculate the reduced-dimensional sound speed profile of the mission sea area.

[0026] According to the sound speed profile data matrix of the UUV mission area over the years , PCA is used to process the sound velocity profile data to obtain the previous Eigenvector , the calculation method is as follows: The sound speed matrix of the UUV mission ocean history is defined as:

[0027] Each column represents M interpolations of a sound velocity profile, and each row represents the sound velocity at different times at the same depth. A sample of the sound velocity profile.

[0028] According to the sound speed matrix and the sound speed mean, the disturbance matrix is ​​calculated :

[0029] in, is the sound velocity matrix, is the mean speed of sound.

[0030] Compute the covariance matrix of the perturbation matrix :

[0031] Pair Matrix Perform eigenvalue decomposition to obtain the eigenvalue matrix and eigenvector matrix:

[0032] in, is the eigenvalue matrix, is the eigenvector matrix.

[0033] Sort by the size of the eigenvalues , select the cumulative variance greater than 95% The eigenvectors corresponding to the eigenvalues ​​form an eigenvector matrix , where the cumulative variance expectation calculation formula is as follows:

[0034] The second step is to calculate the reduced-dimensional sound velocity profile based on the eigenvector matrix. Calculate the sound velocity profile after dimensionality reduction , the calculation formula is as follows:

[0035] in, is the eigenvector matrix, is the perturbation matrix.

[0036] The third step is to use the K-means algorithm to cluster the sound speed profiles after dimensionality reduction.

[0037] Step 1: Divide the sound velocity profile after dimensionality reduction into Class, The center of the class is , randomly selected Group as the center point of the cluster { ; Step 2: Calculate the minimum value of the Euclidean distance of all sound velocity profiles from the center of each type, and The sound velocity profile is divided into the nearest center point category, and the calculation formula is as follows:

[0038] Step 3, The number of sound velocity profiles of the class is , calculate the average value of each class as the new center point , the calculation formula is as follows:

[0039] Step 4: Calculate the objective function , The sound velocity profile samples are calculated as follows:

[0040] Calculate the difference between the two objective functions. If the difference is less than the threshold , the clustering ends. If it is not satisfied, it will start from step 2 and iterate again.

[0041] The fourth step is to calculate the propagation loss through the Bellhop model.

[0042] From the sound velocity profile categories calculated in the third step, randomly select 5 groups of sound velocity profiles for each category and , Distance between UUV and underwater platform , underwater platform depth , UUV depth , using the Bellhop model in the Acoustic Toolbox developed by Michael B. Porter to calculate the sound propagation loss :

[0043] The fifth step is to calculate the binary propagation loss.

[0044] According to different propagation loss thresholds The Bellhop propagation loss is binarized to obtain the final sound propagation loss data. :

[0045] The sixth step is to construct the FNN neural network input data and output data.

[0046] The reduced-dimensional sound velocity profile , Distance between UUV and underwater platform , underwater platform depth 、UUV Depth Harmonic propagation loss threshold As the data input of FNN network , the transmission loss binarization result As the output data of the FNN model.

[0047] Based on the above embodiment, step 200 includes: The first step is to determine the FNN network structure. The FNN network model consists of an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is 7; the hidden layer contains 2 layers, the number of nodes in the first hidden layer is 30, and the number of nodes in the second hidden layer is 50; the number of nodes in the output layer is 1, which represents the binarization result of the sound propagation loss.

[0048] The second step is to determine the FNN network activation function and training parameters. The activation function of the two hidden layers of the FNN is the sigmoid function, and the activation function of the output layer is the softmax function. Set the training accuracy and number of training times of the FNN model, and use the gradient descent method for training.

[0049] The third step is to train the FNN model. The data in the first step is forward propagated through the network, and the loss function is calculated using the minimum mean square error based on the output results and the target value. The network weights and biases are adjusted using the gradient descent method. Repeat the first to third steps until the training accuracy or the upper limit of the number of training times is reached.

[0050] Based on the above embodiment, step 400 includes: Acquire a sound source level by transmitting a sound source from an underwater platform, and acquire a sonar detection threshold of the unmanned underwater vehicle, a sonar receiving array directivity index, and an environmental noise calculated by a sensor; Calculate the sound source level, the sonar detection threshold, the sonar receiving array directivity index and the ambient noise based on the passive sonar equation to obtain a propagation loss threshold; The sensor of the unmanned underwater vehicle obtains the current position of the unmanned underwater vehicle, the depth of the unmanned underwater vehicle and the distance between the unmanned underwater vehicle and the underwater platform; Acquiring a current sound speed profile through a sensor of the unmanned underwater vehicle, and calculating the current sound speed profile after dimensionality reduction; The current sound speed profile after dimensionality reduction, the depth of the unmanned underwater vehicle, the distance between the unmanned underwater vehicle and the underwater platform, and the depth of the underwater platform are input into the underwater sound propagation loss prediction model, and the sound propagation loss of the area where the unmanned underwater vehicle and the underwater platform are located is calculated, and the communicative area is determined by the sound propagation loss.

[0051] Specifically, first, the UUV calculates the UUV propagation loss threshold based on the ocean environment noise and equipment information obtained through sensors. The sound source level is obtained based on the sound source emitted by the underwater platform. , read the UUV sonar detection threshold Sonar receiving array directivity index , read the ambient noise calculated by the UUV sensor , the propagation loss threshold is calculated according to the passive sonar equation, the formula is as follows:

[0052] in, is the sound source level, is the environmental noise, is the sonar detection threshold, It is the directivity index of the sonar receiving array; Then, the UUV obtains the current position through the sensor, and obtains the depth of the UUV and the distance from the underwater platform.

[0053] Next, the UUV acquires the sound velocity profile through the sensor, and calculates the reduced-dimensional sound velocity profile according to the above-mentioned embodiment.

[0054] Finally, the reduced-dimensional sound velocity profile, UUV depth, distance between UUV and underwater platform, and depth of underwater platform are input into the underwater sound propagation loss prediction model to calculate the sound propagation loss in the area where UUV and underwater platform are located, and the communicative area is obtained.

[0055] It should be noted that the method of the present invention is used to conduct simulation experiments, select the sound velocity profile data of a certain sea area in the South China Sea for each month from 2015 to 2023, calculate the reduced-dimensional sound velocity profile, and classify the reduced-dimensional sound velocity profile. The seawater depth is 100m, the underwater platform depth is 10m, the transmitting sound source frequency is 10kHz, the maximum horizontal distance between the UUV and the underwater platform is 10km, the position of the UUV changes at intervals of 10m in the horizontal direction and 0.5m in the vertical direction, input Bellhop to calculate the propagation loss, the propagation loss threshold is 45dB, and the calculated propagation loss is as follows: Figure 4 As shown, they are the Bellhop propagation loss value, the Bellhop propagation loss binarization value and the propagation loss prediction value of the feedforward neural network model.

[0056] The underwater sound propagation loss prediction system based on the feedforward neural network provided by the present invention is described below. The underwater sound propagation loss prediction system based on the feedforward neural network described below and the underwater sound propagation loss prediction method based on the feedforward neural network described above can be referenced to each other.

[0057] Figure 5 is a schematic diagram of the structure of an underwater sound propagation loss prediction system based on a feedforward neural network provided by an embodiment of the present invention. Figure 5 As shown, it includes: a construction module 51, a training module 52, a deployment module 53 and a prediction module 54, wherein: The construction module 51 is used to obtain historical underwater acoustic environment data, convert the historical underwater acoustic environment data into sound velocity profiles, use principal component analysis and K-means algorithm to reduce the dimension and classify the sound velocity profiles, and obtain model training data; the training module 52 is used to construct a feedforward neural network, and perform model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; the deployment module 53 is used to deploy the underwater sound propagation loss prediction model on an unmanned underwater vehicle; the prediction module 54 is used to obtain the in-situ measurement data of the unmanned underwater vehicle, reduce the dimension of the in-situ measurement data, and input it into the underwater sound propagation loss prediction model, and output the underwater sound propagation loss prediction result.

[0058] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the underwater sound propagation loss prediction method based on the feedforward neural network, the method comprising: obtaining historical underwater acoustic environment data, converting the historical underwater acoustic environment data into a sound velocity profile, reducing the dimension and classifying the sound velocity profile by using the principal component analysis method and the K-means algorithm, and obtaining model training data; constructing a feedforward neural network, and training the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; obtaining in-situ measurement data of the unmanned underwater vehicle, reducing the dimension of the in-situ measurement data and inputting it into the underwater sound propagation loss prediction model, and outputting the underwater sound propagation loss prediction result.

[0059] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, 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, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in 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, etc. Various media that can store program codes.

[0060] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program 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 underwater sound propagation loss prediction method based on a feedforward neural network provided by the above-mentioned methods. The method includes: obtaining historical water acoustic environment data, converting the historical water acoustic environment data into a sound velocity profile, and using principal component analysis and K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; obtaining in-situ measurement data of the unmanned underwater vehicle, reducing the dimension of the in-situ measurement data and inputting it into the underwater sound propagation loss prediction model, and outputting the underwater sound propagation loss prediction result.

[0061] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the underwater sound propagation loss prediction method based on a feedforward neural network provided by the above-mentioned methods, the method comprising: obtaining historical water acoustic environment data, converting the historical water acoustic environment data into a sound velocity profile, and using principal component analysis and K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; obtaining in-situ measurement data of the unmanned underwater vehicle, reducing the dimension of the in-situ measurement data and inputting it into the underwater sound propagation loss prediction model, and outputting an underwater sound propagation loss prediction result.

[0062] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0063] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting underwater sound propagation loss based on a feedforward neural network, characterized in that: include: Acquire historical underwater acoustic environment data, convert the historical underwater acoustic environment data into a sound velocity profile, and use principal component analysis and K-means clustering K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; Constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model; Deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; The in-situ measurement data of the unmanned underwater vehicle is obtained, the in-situ measurement data is input into the underwater sound propagation loss prediction model after dimension reduction, and the underwater sound propagation loss prediction result is output.

2. The underwater sound propagation loss prediction method based on feedforward neural network according to claim 1 is characterized in that: The historical underwater acoustic environment data is converted into a sound velocity profile, and the sound velocity profile is reduced in dimension and classified using a principal component analysis method and a K-means clustering algorithm to obtain model training data, including: Determine a sound velocity matrix corresponding to the historical underwater acoustic environment data, and obtain an eigenvector matrix of the sound velocity matrix; Calculating a reduced-dimensional sound velocity profile according to the eigenvector matrix; Clustering the sound velocity profile after dimensionality reduction using a K-means algorithm to obtain a sound velocity profile category; Performing sound propagation calculation on the sound velocity profile category by using the Bellhop model to obtain the propagation loss; Binarizing the propagation loss according to the propagation loss threshold value to obtain final propagation loss data; The reduced-dimensional sound velocity profile, the distance between the unmanned underwater vehicle and the underwater platform, the underwater platform depth, the underwater unmanned vehicle depth and the propagation loss threshold value are used as model input data, and the final propagation loss data are used as model output data.

3. The underwater sound propagation loss prediction method based on feedforward neural network according to claim 2 is characterized in that: Determining the sound velocity matrix corresponding to the historical underwater acoustic environment data and obtaining the eigenvector matrix of the sound velocity matrix includes: Determining that the sound velocity matrix includes different interpolations of a sound velocity profile and sound velocities at different times at the same depth; A disturbance matrix is ​​obtained from the sound speed matrix and the sound speed matrix mean; Calculating the covariance matrix of the disturbance matrix, performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue matrix and an original eigenvector matrix; The eigenvalues ​​in the eigenvalue matrix are sorted, the cumulative variance is calculated from the sorted eigenvalues, and the eigenvectors corresponding to the eigenvalues ​​whose cumulative variance is greater than a preset percentage are determined to form the eigenvector matrix.

4. The underwater sound propagation loss prediction method based on feedforward neural network according to claim 2 is characterized in that: The K-means algorithm is used to cluster the sound velocity profile after dimensionality reduction to obtain the sound velocity profile categories, including: Step 1, dividing the sound velocity profile after dimensionality reduction into several categories, obtaining the center of any category, and randomly determining several category centers as cluster center point sets; Step 2, calculate the minimum Euclidean distance of all sound velocity profiles from the center of each category, and divide each number of sound velocity profiles into the category closest to the center point; Step 3, obtaining the number of any type of sound speed profiles, and calculating the average value of each type of sound speed profile as the new center point; Step 4: Calculate the objective function according to each type of sound speed profile, the new center point and the number of sound speed profile samples, and calculate the difference between the two adjacent objective functions in any number of divisions. If it is determined that the objective function difference is less than the difference threshold, clustering is completed and the sound speed profile category is obtained. Otherwise, iterative calculation is started again from step 2.

5. The underwater sound propagation loss prediction method based on feedforward neural network according to claim 1 is characterized in that: Constructing a feedforward neural network, and performing model training on the feedforward neural network based on the model training data to obtain an underwater sound propagation loss prediction model, including: Determine that the feedforward neural network consists of an input layer, a hidden layer and an output layer, the input layer includes 7 nodes, the hidden layer includes two layers, and the first hidden layer includes nodes, and the second hidden layer includes nodes, the output layer includes 1 node; Determining that the two hidden layers include a sigmoid activation function, and the output layer includes a softmax activation function; Determine the preset training accuracy and the preset training times; The training data is forward propagated according to the network structure of the feedforward neural network, the loss function is calculated using the minimum mean square error according to the output result and the target value, and the network weights and biases are adjusted using the gradient descent method until the preset training accuracy or the preset number of training times is reached, thereby obtaining the underwater sound propagation loss prediction model.

6. The underwater sound propagation loss prediction method based on feedforward neural network according to claim 1 is characterized in that: Acquiring in-situ measurement data of the unmanned underwater vehicle, reducing the dimension of the in-situ measurement data and inputting it into the underwater sound propagation loss prediction model, and outputting the underwater sound propagation loss prediction result, including: Acquire a sound source level by transmitting a sound source from an underwater platform, and acquire a sonar detection threshold of the unmanned underwater vehicle, a sonar receiving array directivity index, and an environmental noise calculated by a sensor; Calculate the sound source level, the sonar detection threshold, the sonar receiving array directivity index and the ambient noise based on the passive sonar equation to obtain a propagation loss threshold; The sensor of the unmanned underwater vehicle obtains the current position of the unmanned underwater vehicle, the depth of the unmanned underwater vehicle and the distance between the unmanned underwater vehicle and the underwater platform; Acquiring a current sound speed profile through a sensor of the unmanned underwater vehicle, and calculating the current sound speed profile after dimensionality reduction; The current sound speed profile after dimensionality reduction, the depth of the unmanned underwater vehicle, the distance between the unmanned underwater vehicle and the underwater platform, and the depth of the underwater platform are input into the underwater sound propagation loss prediction model, and the sound propagation loss of the area where the unmanned underwater vehicle and the underwater platform are located is calculated, and the communicative area is determined by the sound propagation loss.

7. An underwater sound propagation loss prediction system based on a feedforward neural network, characterized in that: include: A construction module is used to obtain historical underwater acoustic environment data, convert the historical underwater acoustic environment data into a sound velocity profile, and use a principal component analysis method and a K-means algorithm to reduce the dimension and classify the sound velocity profile to obtain model training data; A training module, used for constructing a feedforward neural network, performing model training on the feedforward neural network based on the model training data, and obtaining an underwater sound propagation loss prediction model; A deployment module, used for deploying the underwater sound propagation loss prediction model on an unmanned underwater vehicle; The prediction module is used to obtain the in-situ measurement data of the unmanned underwater vehicle, reduce the dimension of the in-situ measurement data and input it into the underwater sound propagation loss prediction model, and output the underwater sound propagation loss prediction result.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the underwater sound propagation loss prediction method based on the feedforward neural network as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting underwater sound propagation loss based on a feedforward neural network as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting underwater sound propagation loss based on a feedforward neural network as claimed in any one of claims 1 to 6 is implemented.

Images