Shell high-frequency sound radiation forecasting method based on local optimization and deep learning

Abstract

The invention relates to the technical field of noise prediction, and discloses a shell high-frequency sound radiation forecasting method based on local optimization and deep learning, comprising: acquiring historical operation data of a shell of an underwater vehicle, the historical operation data comprising speed data, vibration data and sound radiation data; processing the historical operation data, and extracting key features; constructing a deep learning network model, and training the deep learning network model according to the key features; optimizing the deep learning network model by adopting a local optimization algorithm to obtain an optimal solution of parameters of the deep learning network model; the trained deep learning network model is endowed with the optimal solution of the parameters, and a shell high-frequency sound radiation forecasting model is obtained; and acquiring real-time operation data of the shell of the underwater vehicle, and determining a high-frequency sound radiation forecast value of the shell according to the real-time operation data based on the shell high-frequency sound radiation forecast model. According to the invention, the accuracy and efficiency of sound radiation prediction are improved, and online monitoring of shell noise is realized.

Description

Technical Field

[0001] The present invention relates to the field of noise prediction technology, and in particular to a method for predicting high-frequency sound radiation of a shell based on local optimization and deep learning. Background Art

[0002] With the improvement of industrial automation and information technology, the requirements for noise control of equipment are getting higher and higher. In elastic media, as long as the frequency of the longitudinal wave excited by the wave source is between 20-20000Hz, it can cause human hearing. The vibration within this frequency range is called acoustic vibration, and the longitudinal wave excited by acoustic vibration is called sound wave. Sound waves propagate in all directions with the help of various media. Sound waves are usually longitudinal waves, but there are also transverse waves. The particles where the sound waves reach vibrate near the equilibrium position along the propagation direction. The propagation of sound waves is essentially the transfer of energy in the medium. Especially in the fields of ships, underwater vehicles, etc., underwater sound is still the most effective means of underwater information transmission. Accurate prediction of high-frequency sound radiation is of great significance for improving product performance and reducing environmental noise pollution.

[0003] However, traditional acoustic radiation prediction methods often rely on complex numerical simulations and experimental tests, which are time-consuming and labor-intensive and difficult to meet real-time requirements. Therefore, there is an urgent need for a shell high-frequency acoustic radiation prediction method based on local optimization and deep learning to achieve high-frequency acoustic radiation prediction for underwater shells. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting high-frequency sound radiation of a shell based on local optimization and deep learning, so as to solve the problem that traditional sound radiation prediction methods often rely on complex numerical simulation and experimental testing, which is time-consuming and labor-intensive and difficult to meet real-time requirements.

[0005] The present invention provides a method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning, the method comprising:

[0006] Acquiring historical operation data of a hull of the underwater vehicle, the historical operation data including speed data, vibration data, and acoustic radiation data;

[0007] Performing data cleaning and standardization on the historical operation data, and extracting key features;

[0008] Constructing a deep learning network model, and training the deep learning network model according to the key features;

[0009] And using a local optimization algorithm to optimize the deep learning network model to obtain the optimal solution of the deep learning network model parameters;

[0010] Assigning the optimal solution of the parameters to the trained deep learning network model to obtain a shell high-frequency sound radiation prediction model;

[0011] Real-time operation data of the hull of the underwater vehicle is acquired, and based on a high-frequency acoustic radiation prediction model of the hull, a high-frequency acoustic radiation prediction value of the hull is determined according to the real-time operation data.

[0012] In some embodiments of the present application, the speed data includes the hull travel speed of the underwater vehicle, the water flow direction and the water flow speed;

[0013] The vibration data includes vibration frequency and vibration amplitude;

[0014] The sound radiation data includes sound frequency and sound pressure level.

[0015] In some embodiments of the present application, the data cleaning includes: missing value processing, outlier value processing and duplicate value processing.

[0016] In some embodiments of the present application, extracting key features includes:

[0017] Performing Fourier transform on the time series image of the historical operation data to obtain a spectrum image of the historical operation data;

[0018] Key features in the historical operation data are extracted according to the spectrum image.

[0019] In some embodiments of the present application, a local optimization algorithm is used to optimize the deep learning network model to obtain an optimal solution for the parameters of the deep learning network model, including:

[0020] Using a gradient descent algorithm to optimize the deep learning network model;

[0021] Setting the optimization target and learning rate of the deep learning network model;

[0022] Initializing the parameters of the deep learning network model;

[0023] Calculate the gradient of the parameter and take the parameter with the smallest gradient as the current optimal solution;

[0024] Iterative optimization begins until the maximum number of iterations is reached or the preset conditions are met, and the final optimal solution is obtained.

[0025] In some embodiments of the present application, the calculation formula of the gradient descent algorithm is:

[0026]

[0027] Among them, θ is the parameter of the deep learning network model, J(θ) is the loss function, α is the learning rate, is the gradient.

[0028] In some embodiments of the present application, after obtaining the shell high-frequency sound radiation prediction model, the method further includes:

[0029] Randomly extracting a number of data from the historical operation data to form a test set;

[0030] Testing the shell high-frequency sound radiation prediction model according to the test set, and obtaining the shell high-frequency sound radiation prediction model after the test according to the test result;

[0031] If the deviation of the test result is within an acceptable range, the shell high-frequency sound radiation prediction model is retained;

[0032] If the deviation of the test result is not within the acceptable range, the parameters of the shell high-frequency sound radiation prediction model are adjusted and the test is performed again.

[0033] In some embodiments of the present application, the deep learning network model includes a hidden layer and a fully connected layer.

[0034] The present application also discloses a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The instruction, program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning.

[0035] The present application also discloses a device, which includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by the processor to implement the above-mentioned method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning.

[0036] Compared with the prior art, the beneficial effect of the present invention lies in that the historical operation data of the underwater vehicle shell is analyzed and processed to extract key features, and a deep learning network model is established and trained based on the key features. At the same time, the model is locally optimized to ensure that the prediction effect of the model is optimal, improve the accuracy of the sound radiation prediction, and better reflect the sound radiation characteristics of the shell under actual working conditions. Automatic feature extraction through deep learning reduces human intervention and improves prediction efficiency. The application of local optimization algorithms enables the model to fit actual data more accurately and improves the reliability of prediction. It can achieve rapid prediction, meet real-time requirements, and is suitable for online monitoring and control systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0038] Figure 1 It is a flow chart of a method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0040] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0041] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0042] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0043] like Figure 1 As shown, the present invention provides a method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning, the method comprising:

[0044] S1, obtaining historical operation data of the hull of the underwater vehicle, wherein the historical operation data includes speed data, vibration data and sound radiation data.

[0045] S2, performing data cleaning and standardization processing on the historical operation data, and extracting key features.

[0046] S3, constructing a deep learning network model, and training the deep learning network model according to the key features.

[0047] S4, and using a local optimization algorithm to optimize the deep learning network model to obtain the optimal solution for the parameters of the deep learning network model.

[0048] S5, assigning the optimal solution of the parameters to the trained deep learning network model to obtain a shell high-frequency sound radiation prediction model.

[0049] S6, acquiring real-time operation data of the hull of the underwater vehicle, and determining a high-frequency sound radiation prediction value of the hull based on the real-time operation data based on a high-frequency sound radiation prediction model of the hull.

[0050] In some embodiments of the present application, the speed data includes the hull travel speed of the underwater vehicle, the water flow direction and the water flow speed.

[0051] The vibration data includes vibration frequency and vibration amplitude.

[0052] The sound radiation data includes sound frequency and sound pressure level.

[0053] In this embodiment, the noise generated by the underwater vehicle during operation is not only related to the vibration generated by the underwater vehicle itself, but also to the driving speed of the underwater vehicle. Therefore, when collecting the historical operation data of the underwater vehicle shell, it is also necessary to collect the driving speed of the underwater vehicle shell, the water flow direction and the water flow speed.

[0054] In some embodiments of the present application, the data cleaning includes: missing value processing, outlier value processing and duplicate value processing.

[0055] In this embodiment, missing value processing is an important part of data analysis and machine learning, which involves how to deal with blanks or incomplete parts in the data. The following are some methods for processing missing values: deletion, constant filling, interpolation filling, KNN filling, random forest filling, multiple interpolation, model prediction, expert knowledge filling, and this application uses one or several of these methods to process missing values. Outliers, also known as outliers, refer to values ​​that deviate significantly from other data points in a data set. Processing outliers is an important step in data analysis because outliers may have an adverse effect on the analysis results. The following are several commonly used outlier processing methods: deleting outliers, changing outliers, interpolation, grouping, using statistical methods, and retaining outliers. It is worth noting that not all outliers are invalid or wrong. Sometimes outliers may represent some important information or trends, so the causes and background of outliers need to be carefully analyzed before processing. Processing duplicate values ​​is a key step to ensure data quality and accuracy. Before data analysis and modeling, it is usually necessary to perform deduplication on the data set. The following are some methods for processing duplicate values: detecting duplicate values, deleting duplicate values, replacing duplicate values, merging duplicate values, and analyzing duplicate values. In this application, if the deletion method is used to process historical operation data, all data at the same time corresponding to the value to be deleted need to be deleted.

[0056] In some embodiments of the present application, extracting key features includes: performing Fourier transform on a time series image of the historical operation data to obtain a spectrum image of the historical operation data; and extracting key features in the historical operation data based on the spectrum image.

[0057] In this embodiment, Fourier transform is a widely used tool in signal processing, which can decompose a complex time series signal into a combination of sine waves and cosine waves of different frequencies. The purpose of doing so is to better understand the frequency characteristics of the signal. The time series image of the historical operation data can be transformed into a spectrum image of the historical operation data by Fourier transform, and then the key features can be determined by the spectrum image.

[0058] In some embodiments of the present application, a local optimization algorithm is used to optimize the deep learning network model to obtain the optimal solution of the parameters of the deep learning network model, including: optimizing the deep learning network model using a gradient descent algorithm; setting the optimization target and learning rate of the deep learning network model; initializing the parameters of the deep learning network model; calculating the gradient of the parameters, and taking the parameter with the smallest gradient as the current optimal solution; starting iterative optimization until the maximum number of iterations is reached or the preset conditions are met to obtain the final optimal solution.

[0059] In some embodiments of the present application, the calculation formula of the gradient descent algorithm is:

[0060]

[0061] Among them, θ is the parameter of the deep learning network model, J(θ ) is the loss function, α is the learning rate, is the gradient.

[0062] In this embodiment, the gradient descent algorithm is an optimization algorithm used to find parameter values ​​that minimize the loss function.

[0063] In machine learning and deep learning, the loss function measures the difference between the model's predicted value and the true value. The purpose of gradient descent is to reduce this difference by updating the model's parameters. It works by calculating the gradient of the loss function with respect to the model parameters, and then updating the parameters in the opposite direction of the gradient. In this way, each iteration will reduce the loss function value, gradually approaching the minimum value of the loss function.

[0064] The specific steps of the gradient descent algorithm usually include the following steps: Initialize parameters: Randomly initialize the parameters of the model. Calculate gradients: Use the current parameters to calculate the gradient of the loss function with respect to these parameters. Update parameters: Update the parameters based on the calculated gradient and learning rate. Repeat iterations: Repeat the above process until the stopping condition is met or the preset number of iterations is reached.

[0065] In some embodiments of the present application, after obtaining the shell high-frequency sound radiation prediction model, it also includes: randomly extracting a number of data from the historical operation data to form a test set; testing the shell high-frequency sound radiation prediction model according to the test set, and obtaining the tested shell high-frequency sound radiation prediction model according to the test result; if the deviation of the test result is within an acceptable range, retaining the shell high-frequency sound radiation prediction model; if the deviation of the test result is not within an acceptable range, adjusting the parameters of the shell high-frequency sound radiation prediction model, and retesting.

[0066] In this embodiment, the test set accounts for 1:9-10 of all historical operation data. Regarding the acceptance range, the acceptance range can be set to 95% or 98%. When the deviation of the test result is within the acceptance range, that is, the ratio of the test value to the actual value is greater than 95% or 98%, it means that the obtained shell high-frequency sound radiation prediction model can accurately predict the shell high-frequency sound radiation prediction value. Therefore, it is necessary to retain the shell high-frequency sound radiation prediction model for application; if the deviation of the test result is not within the acceptance range, that is, the test result is less than 95% or 98%, it means that the obtained shell high-frequency sound radiation prediction model cannot accurately predict the shell high-frequency sound radiation prediction value. Therefore, it is necessary to adjust the parameters of the shell high-frequency sound radiation prediction model and retest until the deviation of the test result is within the acceptable range.

[0067] In some embodiments of the present application, the deep learning network model includes a hidden layer and a fully connected layer.

[0068] In this embodiment, the deep learning network model generally includes a hidden layer and a fully connected layer. The hidden layer is a layer in a neural network other than the input layer and the output layer, and is mainly responsible for extracting features from the input data. There can be multiple hidden layers, each of which is composed of multiple neurons, which perform nonlinear transformations on the input data through activation functions, so that complex data representations can be learned. In a deep neural network, multiple hidden layers can extract increasingly abstract high-level features layer by layer. A fully connected layer is a layer in a neural network in which each neuron is connected to all neurons in the previous layer. This means that each neuron in the fully connected layer is able to receive signals from all neurons in the previous layer. The fully connected layer is usually located at the end of the network and is used to combine the features extracted by the previous layers to produce the final output. In classification tasks, the output of the fully connected layer is usually combined with an activation function (such as softmax) to generate a probability distribution representing the predicted probabilities of different categories.

[0069] The present application also discloses a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The instruction, program, code set or instruction set is loaded and executed by a processor to implement the above-mentioned method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning.

[0070] Preferably, the computer-readable storage medium may be an internal storage unit, such as a hard disk or a memory. The computer-readable storage medium may also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc.

[0071] As is known to those of ordinary skill in the art, the term computer-readable storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0072] The present application also discloses a device, which includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by the processor to implement the above-mentioned method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning.

[0073] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

[0075] The system provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be decomposed or combined. For example, the modules in the above embodiment can be combined into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps, and are not regarded as improper limitations of the present invention.

[0076] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

Claims

1. A shell high-frequency acoustic radiation prediction method based on local optimization and deep learning, characterized in that: The method comprises: Acquiring historical operation data of a hull of the underwater vehicle, the historical operation data including speed data, vibration data, and acoustic radiation data; Performing data cleaning and standardization on the historical operation data, and extracting key features; Constructing a deep learning network model, and training the deep learning network model according to the key features; And using a local optimization algorithm to optimize the deep learning network model to obtain the optimal solution of the deep learning network model parameters; Assigning the optimal solution of the parameters to the trained deep learning network model to obtain a shell high-frequency sound radiation prediction model; Real-time operation data of the hull of the underwater vehicle is acquired, and based on a high-frequency acoustic radiation prediction model of the hull, a high-frequency acoustic radiation prediction value of the hull is determined according to the real-time operation data.

2. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 1 is characterized in that: The speed data includes the hull travel speed of the underwater vehicle, the water flow direction and the water flow speed; The vibration data includes vibration frequency and vibration amplitude; The sound radiation data includes sound frequency and sound pressure level.

3. The shell high-frequency acoustic radiation prediction method based on local optimization and deep learning according to claim 1 is characterized in that: The data cleaning includes: missing value processing, outlier value processing and duplicate value processing.

4. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 1, characterized in that: Extract key features, including: Performing Fourier transform on the time series image of the historical operation data to obtain a spectrum image of the historical operation data; Key features in the historical operation data are extracted according to the spectrum image.

5. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 1, characterized in that: The deep learning network model is optimized using a local optimization algorithm to obtain the optimal solution for the parameters of the deep learning network model, including: Using a gradient descent algorithm to optimize the deep learning network model; Setting the optimization target and learning rate of the deep learning network model; Initializing the parameters of the deep learning network model; Calculate the gradient of the parameter and take the parameter with the smallest gradient as the current optimal solution; Iterative optimization begins until the maximum number of iterations is reached or the preset conditions are met, and the final optimal solution is obtained.

6. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 5, characterized in that: The calculation formula of the gradient descent algorithm is: Among them, θ is the parameter of the deep learning network model, J ( θ ) is the loss function, α is the learning rate, is the gradient.

7. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 1, characterized in that: After obtaining the shell high-frequency sound radiation prediction model, it also includes: Randomly extracting a number of data from the historical operation data to form a test set; Testing the shell high-frequency sound radiation prediction model according to the test set, and obtaining the shell high-frequency sound radiation prediction model after the test according to the test result; If the deviation of the test result is within an acceptable range, the shell high-frequency sound radiation prediction model is retained; If the deviation of the test result is not within the acceptable range, the parameters of the shell high-frequency sound radiation prediction model are adjusted and the test is performed again.

8. The method for predicting high-frequency acoustic radiation of a shell based on local optimization and deep learning according to claim 1, characterized in that: The deep learning network model includes a hidden layer and a fully connected layer.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by a processor to implement the shell high-frequency acoustic radiation prediction method based on local optimization and deep learning as described in any one of claims 1 to 8.

10. A device, characterized in that: The device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the instruction, program, code set or instruction set is loaded and executed by the processor to implement the shell high-frequency acoustic radiation prediction method based on local optimization and deep learning as described in any one of claims 1 to 8.

Images