Wavelet neural network underwater vehicle mechanical noise rapid forecasting method and system based on Adam optimization

Abstract

The invention discloses a wavelet neural network underwater vehicle mechanical noise rapid forecasting method and system based on Adam optimization, belongs to the field of ship underwater radiation noise forecasting, and aims to solve the problem of rapid forecasting of mechanical noise of an underwater vehicle. According to the method, a wavelet neural network noise forecasting model based on Adam optimization is designed, and key variables including the rotating speed and power of mechanical equipment, vibration isolation measures, the form of a base, the thickness of a web and a panel, the material and the thickness of a shell and fluid medium density data are selected as characteristic conditions; an Adam optimization algorithm is introduced to optimize gradient descent of a wavelet neural network model, first-order moment estimation and second-order moment estimation are adopted to dynamically adjust the learning rate of each parameter, model learning is accelerated, the network recognition precision is improved, and oscillation is reduced during convergence. And the training set is used to carry out iterative training on the optimization model to obtain a high-precision noise prediction model, the prediction of the mechanical noise of the underwater vehicle is realized, and the method has the advantages of wide applicability, high precision and the like.

Description

Technical Field

[0001] The invention belongs to the field of ship underwater radiation noise prediction, and specifically relates to a method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network. Background Art

[0002] For underwater weapon systems such as submarines and torpedoes, their survivability and combat effectiveness depend to a large extent on their acoustic stealth and acoustic detection capabilities. Underwater vehicles are equipped with mechanical devices and equipment to achieve their set functions. When these devices and equipment are in operation, the vibration of the equipment is transmitted through the structure and radiates noise to the external waters, which is called mechanical noise.

[0003] At present, the main research methods for mechanical noise of underwater vehicles include analytical method, finite element-boundary element method and statistical energy analysis method. The analytical method is based on the classical theory of structural vibration and acoustic radiation, and establishes the motion differential equation of the prediction analysis model and the initial conditions and boundary conditions that meet the equation conditions to obtain acoustic quantities such as radiated sound power and radiation efficiency. The finite element and boundary element methods discretize the analyzed structure, that is, divide the structure into several small units, and according to the principle that the deformation and stress at the boundary of the connected units are equal, the units are connected and calculated by computer technology using the motion differential matrix equation. However, for large-sized structures or structures vibrating at medium and high frequencies, the constructed fluid is required to be large or the grid is divided very finely, so the numerical calculation amount and calculation time are very large. The statistical energy method is based on the principles of statistical physics. The physical parameters used are the statistical average values ​​of space and time, which reflect the average level of underwater structural vibration and acoustic radiation. In the medium and high frequency range, it can well describe the average vibration and acoustic characteristics of system components. However, due to the small modal density, the statistical energy method is not suitable for mechanical noise prediction of underwater structures in lower frequency bands.

[0004] As a computational model that simulates the human brain, neural networks have strong simulation and learning capabilities, and are particularly suitable for data that cannot be expressed by mathematical functions. The neural network model method can overcome the defects of the regression analysis method, unify the mechanical noise prediction into a network, take the structure of the ship and platform or the equipment on it as parameters, and automatically determine the parameters affecting the mechanical noise and their degree of influence through the inhibition and activation of network nodes. There is less human intervention, and there is no need to establish a large-scale model, which greatly improves the prediction speed. At the same time, the neural network model is trained from historical data. Although it is very similar to the empirical formula method, the quality and accuracy of its model can be continuously optimized and improved with the continuous accumulation of historical data. Therefore, it is of great engineering significance to realize the rapid prediction of mechanical noise of underwater vehicles based on artificial intelligence.

[0005] Through literature search of the prior art, it is found that in recent years, there are few patent documents similar to the method for rapid prediction of mechanical noise of underwater vehicles based on wavelet neural network optimized by Adam. The public materials similar to the present invention application include: 1. Prediction of underwater structure vibration noise based on LM-BP neural network [C] (Proceedings of the 18th Academic Symposium on Ship Underwater Noise. 2021: 836-843); 2. Prediction of ship sonar self-noise based on neural network and genetic algorithm [J] (Journal of Huazhong University of Science and Technology (Natural Science Edition), 2006 (08): 46-47+83); 3. Engineering prediction method of mechanical noise of underwater vehicles [J]. China Shipbuilding, 2001 (02): 84-91.

[0006] Reference 1 discloses a method for predicting underwater structural vibration noise based on neural network. Aiming at the phenomenon that underwater structural vibration equipment is numerous, the structure is complex, and the vibration noise presents obvious nonlinear characteristics, a LM-BP neural network is constructed, with the surface vibration of the underwater structure as the input, the sound pressure of the sound field assessment point as the output, and the error descent curve, the percentage error of the test sample detection result and the prediction qualification rate as the evaluation indicators. The influence of sample quality on the training effect of the neural network is analyzed. The results show that the sample error is large, the prediction result qualification rate is low, and the prediction effect of this method is poor; Reference 2 studies the prediction problem of the whole ship self-noise received by the sonar part when the ship is sailing, analyzes the various parameters affecting the self-noise of the sonar part, and combines the BP neural network with the genetic algorithm for the prediction of the self-noise of the sonar part of the ship. However, the number of hidden layer neuron nodes N cannot be determined, and the specific optimization method of the number of hidden layer nodes has not yet been mastered; Reference 3 establishes an engineering prediction method for the mechanical noise of underwater vehicles under known structural input sound power, and gives the estimation method of radiation efficiency and modal density, which can realize the noise prediction of simple shell structures, but the applicability for complex structures has not been effectively verified.

[0007] In summary, there are significant differences between the currently available publicly searchable literature and the present invention, which are mainly reflected in the methods and prediction accuracy. There are many types of vibration sources and noise sources on underwater vehicles, including power machinery such as the main engine, auxiliary machinery, propellers, propulsion systems, and auxiliary machinery such as pumps and fans. Each device may produce different noises, which increases the complexity of the prediction. Therefore, the present invention establishes a rapid prediction method for underwater vehicle mechanical noise based on Adam optimized wavelet neural network, which can achieve high-precision rapid prediction of underwater mechanical noise. Summary of the invention

[0008] Mechanical noise sources are the main source of radiated noise of underwater vehicles under low-speed navigation conditions. Since the mechanical noise of underwater vehicles involves multiple physical processes, such as fluid dynamics, vibration and sound propagation, there are many problems that need to be solved in noise prediction, resulting in high computational difficulty, low efficiency and accuracy, etc. In order to solve the problem of rapid prediction of mechanical noise of underwater vehicles, the present invention has built a wavelet neural network algorithm based on Adam optimization, which can realize the rapid prediction of mechanical noise of underwater vehicles. The present invention is suitable for the design, construction and navigation stages of underwater vehicles. It can predict underwater mechanical noise under the premise of only mastering certain characteristic condition data such as the base, equipment, and vibration isolators, and can provide guidance for the concealed navigation and vibration reduction and noise reduction design of underwater vehicles.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] The present invention provides a method for rapid prediction of mechanical noise of underwater vehicles based on wavelet neural network optimized by Adam. The method for rapid prediction of mechanical noise comprises the following steps:

[0011] Step S1: Determine whether the forecast object and forecast method are applicable;

[0012] Step S2: determining characteristic condition data by analyzing the mechanical noise characteristics of the underwater vehicle and the influence of various parameters on the underwater radiation noise results;

[0013] Step S3: Simulate and preprocess the characteristic condition data to construct a sample data set as training data;

[0014] Step S4: Use deep learning algorithm to build a wavelet neural network model and set the number of neurons and the number of training iterations;

[0015] Step S5: iteratively train and verify the wavelet neural network model, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0016] Step S6: Obtain the optimal prediction model to achieve rapid prediction of underwater mechanical noise.

[0017] Furthermore, there is a preferred embodiment, the above-mentioned mechanical noise rapid prediction method is applicable to underwater navigation equipment, including but not limited to underwater submersibles, unmanned underwater vehicles, UUVs, and AUVs, which have high requirements on the underwater mechanical noise level and there is a need for rapid prediction and evaluation of underwater noise.

[0018] Furthermore, in another preferred embodiment, the above step S2 specifically includes:

[0019] Step S21: By analyzing the mechanical noise characteristics and mechanical noise impact of underwater vehicles, it is found that mechanical noise is manifested as radiation noise generated by the interaction between the shell and the surrounding water medium after the shell is stimulated to vibrate, and the mechanical noise source is the main source of radiation noise of underwater vehicles under low-speed navigation conditions. After analysis, the characteristic conditions for noise prediction are determined, and simulation is performed, and the simulation data is used as the data set for noise prediction model training;

[0020] Step S22: selecting mechanical equipment speed and power, vibration isolation measures, base form and web and panel thickness, shell material and thickness, and fluid medium density data as input variable data through analysis;

[0021] Step S23: The variable data selected through analysis are highly representative and rapid forecasting can be achieved without other conditions.

[0022] Furthermore, in another preferred embodiment, the above step S3 specifically includes:

[0023] Step S31: The number of samples of ship noise data in this study is relatively small. In order to make full use of the data, cross validation is used in the training process;

[0024] Step S32: Noise is affected by multiple factors, and each influencing factor variable is a discrete variable, while the noise level target value is a continuous variable, so the Ordinal Encoder is used to numerically encode the source data;

[0025] Step S33: Use the maximum-minimum method to normalize the input data to eliminate the difference in magnitude between the data of each dimension:

[0026]

[0027] In the formula, X min is the minimum value of the data set, X max is the maximum value of the data set.

[0028] Step S34: Randomly divide the data set into three parts: a training set, a test set, and a validation set. The sample data set must not be less than 2400.

[0029] Furthermore, in another preferred embodiment, the above step S4 specifically includes:

[0030] Step S41: The wavelet neural network model comprises an input layer, a hidden layer and an output layer, wherein the input layer has m neurons, representing the number of nodes of the input data, the hidden layer has s neurons, and the output layer has n neurons, outputting the frequency domain signal of the mechanical noise of the underwater vehicle.

[0031] Step S42: The activation function of the hidden layer node of the wavelet neural network model is the wavelet basis function. The wavelet basis function is the core of the wavelet transform, and its calculation formula is:

[0032]

[0033] In the formula, ω ij is the weight of the input layer to the hidden layer, b j and a j are the offset factor and scale factor of the wavelet basis function respectively, h is the wavelet basis function, and the Morlet mother wavelet basis function is selected, and its expression is:

[0034]

[0035] In the formula, is the Gaussian distribution function, which represents the amplitude part of the wavelet basis function.

[0036] The output y of the neural network model k for:

[0037]

[0038] In the formula, ω jk is the connection weight from the hidden layer to the output layer.

[0039] Step S43: The parameters of the wavelet neural network model are ω ij ,ω jk 、b j and a j , adjust the parameters by error, the actual result And the network output y k The error between them is:

[0040]

[0041] The parameters of the wavelet neural network are adjusted as follows:

[0042]

[0043] Correction terms in the formula and Calculated by the following formula:

[0044]

[0045]

[0046] Where η is the learning rate.

[0047] Step S44: One training of the wavelet neural network means completing one forward propagation and one reverse error correction. When the neural network is continuously trained and the output parameters meet the specified requirements, the wavelet neural network will stop training.

[0048] Furthermore, in another preferred embodiment, the above step S5 specifically includes:

[0049] Step S51: iteratively train the neural network model using the training set data;

[0050] Step S52: input the test set data into the neural network model obtained by the above process, and repeatedly experiment to obtain a high-precision model;

[0051] Step S53: Based on the wavelet neural network model determined in the previous steps, the Adam optimization algorithm is used. Taking the tth iteration of gradient descent as an example, the network weight update process is:

[0052] t=t+1

[0053]

[0054] In the formula, m t and v t are the first-order momentum term and the second-order momentum term, β1 and β2 are hyperparameters with default values ​​of 0.9 and 0.999, respectively. t and v t The deviation correction value is and

[0055]

[0056] Step S54: Update the parameters of the wavelet neural network model in the previous step. The parameters of the tth iteration are for:

[0057]

[0058] In the formula, The model parameter α at the t-1th iteration is the learning rate. The default value is 0.001. ε is a very small number. Its function is to avoid the denominator being 0. Take ε as 10 -8 ;

[0059] Step S55: By solving the above equations, the updated and optimized model parameters can be obtained, thereby obtaining a wavelet neural network prediction model based on Adam optimization.

[0060] Furthermore, in another preferred embodiment, the above step S6 specifically includes:

[0061] Step S61: using the validation set data to determine the accuracy of the neural network model, and predicting the mechanical noise result data of the underwater vehicle;

[0062] Step S62: Compare the predicted result with the actual result to evaluate the performance of the underwater machinery noise prediction model.

[0063] The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network described in the present invention can be fully implemented by computer software. Therefore, correspondingly, the present invention also provides a system for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network. The system includes a storage device, and the storage device includes:

[0064] Used to determine whether the forecast object and forecast method are applicable;

[0065] It is used to determine characteristic condition data by analyzing the mechanical noise characteristics of underwater vehicles and the influence of various parameters on the underwater radiated noise results;

[0066] Used to simulate and preprocess characteristic condition data and construct sample data sets as training data;

[0067] Used to build a wavelet neural network model using a deep learning algorithm and set the number of neurons and the number of training iterations;

[0068] It is used to iteratively train and verify the wavelet neural network model, and uses the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0069] Used to obtain the optimal prediction model and realize rapid prediction of underwater mechanical noise

[0070] The present invention also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, any one of the above-mentioned methods for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network is implemented.

[0071] The present invention also provides a computer 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 computer program, any one of the above-mentioned methods for rapid prediction of mechanical noise of underwater vehicles based on Adam-optimized wavelet neural network is implemented.

[0072] The technical solution of the present invention has the following beneficial effects:

[0073] 1. The present invention provides a method for rapid prediction of underwater vehicle mechanical noise based on a wavelet neural network optimized by Adam, and builds a wavelet neural network algorithm based on Adam optimization, wherein the wavelet neural network provides the ability to model nonlinear relationships, can better handle complex nonlinear relationships and feature extraction of large amounts of data, and is suitable for complex prediction tasks such as underwater vehicle mechanical noise.

[0074] Furthermore, the rapid prediction method for underwater vehicle mechanical noise based on Adam optimized wavelet neural network proposed in the present invention has the characteristics of fast prediction speed and high accuracy, and is suitable for the entire cycle of underwater vehicle design, construction, and use. Only specific parameters are needed to quickly obtain the underwater vehicle mechanical noise level.

[0075] 2. Different parameter initializations will cause the neural network algorithm to converge to different local minima. Compared with traditional gradient descent, the Adam algorithm combines the momentum method and the adaptive learning rate characteristics, can converge to the local optimal solution faster, and is insensitive to the selection of initial parameters, and is particularly suitable for complex wavelet neural network structures. The method disclosed in the present invention optimizes and improves the wavelet neural network based on the Adam adaptive learning rate adjustment algorithm, uses first-order moment estimation and second-order moment estimation to dynamically adjust the learning rate of each parameter to accelerate model learning, improve network recognition accuracy, and reduce oscillation during convergence, thereby achieving accurate prediction of underwater radiated noise.

[0076] 3. The method disclosed in the invention is based on the simulation data of the influence conditions of the mechanical noise of underwater vehicles. It has a high prediction efficiency and can realize the rapid prediction of underwater mechanical noise. It can avoid the noise prediction delay caused by a large amount of calculation in the traditional numerical prediction method. In actual application, experiments are carried out using real underwater vehicle mechanical noise data. By comparing the experimental results with the model prediction results, it is found that the mechanical noise prediction based on neural network has high accuracy and efficiency.

[0077] 4. The method disclosed in the present invention has a fast prediction speed and is applicable to the entire cycle of underwater vehicle design, construction, and use. Only specific parameters are required to quickly obtain the mechanical noise level of the underwater vehicle.

[0078] The method disclosed in the present invention has a wide range of applications and is applicable to all underwater vehicles that require mechanical noise prediction. It has high versatility and the same series of underwater vehicles can use the same prediction model, which is economical and efficient and convenient for technology promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a flow chart of a method for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network according to the present invention;

[0080] Figure 2 It is a schematic diagram of the structure of the wavelet neural network model established by the present invention;

[0081] Figure 3 This is the underwater machinery noise prediction result of an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0083] Implementation method 1, see Figure 1 In this embodiment, the mechanical noise source is the main source of radiated noise of underwater vehicles under low-speed navigation conditions. Since the mechanical noise of underwater vehicles involves multiple physical processes, such as fluid dynamics, vibration and sound propagation, there are many problems that need to be solved in noise prediction, resulting in high computational difficulty, low efficiency and accuracy. Therefore, this embodiment provides a method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network, the method comprising the following steps:

[0084] Step S1: Determine whether the forecast object and forecast method are applicable;

[0085] Step S2: determining characteristic condition data by analyzing the mechanical noise characteristics of the underwater vehicle and the influence of various parameters on the underwater radiation noise results;

[0086] Step S3: Simulate and preprocess the characteristic condition data to construct a sample data set as training data;

[0087] Step S4: Use deep learning algorithm to build a wavelet neural network model and set the number of neurons and the number of training iterations;

[0088] Step S5: iteratively train and verify the wavelet neural network model, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0089] Step S6: Obtain the optimal prediction model to achieve rapid prediction of underwater mechanical noise.

[0090] In practical application, if Figure 1As shown, this embodiment aims at the problem of rapid prediction of mechanical noise of underwater vehicles, and builds a wavelet neural network algorithm based on Adam optimization. The Adam algorithm combines the adaptive learning rate method of Adagrad and the weighted moving average of RMSprop, and can usually converge to a better local optimal solution faster, and is less sensitive to the selection of hyperparameters, and is particularly suitable for complex wavelet neural network structures. Therefore, the Adam algorithm is used to optimize the neural network model, which can significantly improve the prediction accuracy and computational efficiency, is suitable for processing complex and dynamically changing noise signals, and can achieve high-precision rapid prediction of underwater mechanical noise. The wavelet neural network provides the ability to model nonlinear relationships, can better handle complex nonlinear relationships and feature extraction of large amounts of data, and is suitable for complex prediction tasks such as mechanical noise of underwater vehicles.

[0091] Furthermore, the method proposed in this embodiment is based on the simulation data of the influence conditions of the mechanical noise of underwater vehicles. It has a high prediction efficiency and can realize the rapid prediction of underwater mechanical noise. It can avoid the noise prediction delay caused by a large amount of calculation in the traditional numerical prediction method. In actual application, experiments are carried out using real underwater vehicle mechanical noise data. By comparing the experimental results and the model prediction results, it can be seen that the mechanical noise prediction based on neural network has high accuracy and efficiency.

[0092] Implementation Method 2: See Figure 2 This embodiment is described. This embodiment is an example of the specific steps of the method for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network described in the first embodiment.

[0093] like Figure 1 As shown in FIG. 1 , the rapid prediction method for underwater mechanical noise specifically includes the following steps:

[0094] Step S1: Determine whether the forecast object and forecast method are applicable

[0095] Preferably, step S1 specifically includes:

[0096] The rapid prediction method for underwater mechanical noise described in this embodiment is applicable to underwater navigation equipment, including but not limited to underwater submersibles, unmanned underwater vehicles, UUVs, and AUVs, which have high requirements on the underwater mechanical noise level and there is a need for rapid prediction and evaluation of underwater noise.

[0097] Step S2: determining characteristic condition data by analyzing the mechanical noise characteristics of the underwater vehicle and the influence of various parameters on the underwater radiation noise results;

[0098] Preferably, step S2 specifically includes:

[0099] Step S21: By analyzing the mechanical noise characteristics and mechanical noise impact of underwater vehicles, it is found that mechanical noise is manifested as radiation noise generated by the interaction between the shell and the surrounding water medium after the shell is stimulated to vibrate, and the mechanical noise source is the main source of radiation noise of underwater vehicles under low-speed navigation conditions. After analysis, the characteristic conditions for noise prediction are determined, and simulation is performed, and the simulation data is used as the data set for noise prediction model training;

[0100] Step S22: selecting mechanical equipment speed and power, vibration isolation measures, base form and web and panel thickness, shell material and thickness, and fluid medium density data as input variable data through analysis;

[0101] Step S23: The variable data selected through analysis are highly representative and rapid forecasting can be achieved without other conditions.

[0102] Step S3: Simulate and preprocess the characteristic condition data to construct a sample data set as training data;

[0103] Preferably, step S3 specifically includes:

[0104] Step S31: The number of samples of ship noise data in this study is relatively small. In order to make full use of the data, cross validation is used in the training process;

[0105] Step S32: Noise is affected by multiple factors, and each influencing factor variable is a discrete variable, while the noise level target value is a continuous variable, so the Ordinal Encoder is used to numerically encode the source data;

[0106] Step S33: Use the maximum-minimum method to normalize the input data to eliminate the difference in magnitude between the data of each dimension:

[0107]

[0108] Where, X min is the minimum value of the data set, X max is the maximum value of the data set.

[0109] Step S34: Randomly divide the data set into three parts: a training set, a test set, and a validation set. The sample data set must not be less than 2400.

[0110] Step S4: Use deep learning algorithm to build a wavelet neural network model and set the number of neurons and the number of training iterations;

[0111] Preferably, step S4 specifically includes:

[0112] Step S41: The wavelet neural network model comprises an input layer, a hidden layer and an output layer, wherein the input layer has m neurons, representing the number of nodes of the input data, the hidden layer has s neurons, and the output layer has n neurons, outputting the frequency domain signal of the mechanical noise of the underwater vehicle.

[0113] Step S42: The activation function of the hidden layer node of the wavelet neural network model is the wavelet basis function. The wavelet basis function is the core of the wavelet transform, and its calculation formula is:

[0114]

[0115] In the formula, ω ij is the weight of the input layer to the hidden layer, b j and a j are the offset factor and scale factor of the wavelet basis function respectively, h is the wavelet basis function, and the Morlet mother wavelet basis function is selected, and its expression is:

[0116]

[0117] In the formula, is the Gaussian distribution function, which represents the amplitude part of the wavelet basis function.

[0118] The output y of the neural network model k for:

[0119]

[0120] In the formula, ω jk is the connection weight from the hidden layer to the output layer.

[0121] Step S43: The parameters of the wavelet neural network model are ω ij ,ω jk 、b j and a j , adjust the parameters by error, the actual result And the network output y k The error between them is:

[0122]

[0123] The parameters of the wavelet neural network are adjusted as follows:

[0124]

[0125] Correction terms in the formula and Calculated by the following formula:

[0126]

[0127] In the formula, η is the learning rate.

[0128] Step S44: One training of the wavelet neural network refers to completing one forward propagation and one reverse error correction. When the neural network is continuously trained and the output parameters meet the specified requirements, the wavelet neural network will stop training.

[0129] Step S5: iteratively train and verify the wavelet neural network model, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0130] Preferably, step S5 specifically includes:

[0131] Step S51: iteratively train the neural network model using the training set data;

[0132] Step S52: input the test set data into the neural network model obtained by the above process, and repeatedly experiment to obtain a high-precision model;

[0133] Step S53: Based on the wavelet neural network model determined in the previous steps, the Adam optimization algorithm is used. Taking the tth iteration of gradient descent as an example, the network weight update process is:

[0134] t=t+1

[0135]

[0136] In the formula, m t and v t are the first-order momentum term and the second-order momentum term, β1 and β2 are hyperparameters with default values ​​of 0.9 and 0.999, respectively. t and v t The deviation correction value is and

[0137]

[0138] Step S54: Update the parameters of the wavelet neural network model in the previous step. The parameters of the tth iteration are for:

[0139]

[0140] In the formula, The model parameter α at the t-1th iteration is the learning rate. The default value is 0.001. ε is a very small number. Its function is to avoid the denominator being 0. Take ε as 10 -8 ;

[0141] Step S55: By solving the above equations, the updated and optimized model parameters can be obtained, thereby obtaining a wavelet neural network prediction model based on Adam optimization.

[0142] Step S6: Obtain the optimal prediction model to achieve rapid prediction of underwater mechanical noise.

[0143] Preferably, step S6 specifically includes:

[0144] Step S61: using the validation set data to determine the accuracy of the neural network model, and predicting the mechanical noise result data of the underwater vehicle;

[0145] Step S62: Compare the predicted result with the actual result to evaluate the performance of the underwater machinery noise prediction model.

[0146] In summary, the method for rapid prediction of underwater vehicle mechanical noise proposed in this embodiment builds a wavelet neural network algorithm based on Adam optimization, wherein the Adam algorithm combines the adaptive learning rate method of Adagrad and the weighted moving average of RMSprop, which can usually converge to a better local optimal solution faster and is less sensitive to the selection of hyperparameters, and is particularly suitable for complex wavelet neural network structures. Therefore, the Adam algorithm is used to optimize the neural network model, which can significantly improve the prediction accuracy and computational efficiency, is suitable for processing complex and dynamically changing noise signals, and can achieve high-precision rapid prediction of underwater mechanical noise. Wavelet neural networks provide the ability to model nonlinear relationships, can better handle complex nonlinear relationships and feature extraction of large amounts of data, and are suitable for complex prediction tasks such as underwater vehicle mechanical noise.

[0147] Furthermore, the method proposed in this embodiment is based on the simulation data of the influence conditions of the mechanical noise of underwater vehicles. It has a high prediction efficiency and can realize the rapid prediction of underwater mechanical noise. It can avoid the noise prediction delay caused by a large amount of calculation in the traditional numerical prediction method. In actual application, experiments are carried out using real underwater vehicle mechanical noise data, and the experimental results are compared with the model prediction results. It is found that the mechanical noise prediction based on neural network has high accuracy and efficiency.

[0148] Furthermore, the method for rapid prediction of underwater vehicle mechanical noise proposed in this embodiment is applicable to the design, construction and navigation stages of underwater vehicles. The prediction of underwater mechanical noise based on only certain characteristic condition data such as the base, equipment, and vibration isolators can provide guidance for the covert navigation and vibration and noise reduction design of underwater vehicles. It has a wide range of applications and is applicable to all underwater vehicles that require mechanical noise prediction. It has high versatility, and the same series of underwater vehicles can use the same prediction model, which is economical and efficient and facilitates technology promotion.

[0149] Implementation method three, see Figure 3 This embodiment is described, and the embodiment is to further illustrate the method for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network described in the above embodiment;

[0150] Step S1: Determine the underwater mechanical noise prediction object according to the underwater vehicle type and navigation conditions. In the implementation case, the underwater vehicle is simplified to a certain extent, and a cone-cylinder-spherical shell structure containing a base is used as the research object;

[0151] Step S2: by analyzing the selected characteristic conditions, seven variables are set, including mechanical equipment speed and power, vibration isolation measures, base form, shell material and thickness, and fluid medium density data;

[0152] Step S3: According to the characteristic conditions determined in step S2, the data is simulated and preprocessed. The characteristic condition variables are all discrete variables, while the noise level target value is a continuous variable, so the Ordinal Encoder is used to perform numerical encoding operations on the data;

[0153] Step S4: Build a 3-layer wavelet neural network model and determine the optimal number of hidden layer nodes and hidden layer numbers by trial and error. The number of nodes in the input layer is 7, the number of nodes in the hidden layer is 20, and the number of nodes in the output layer is 8. The final wavelet neural network structure is 7-20-8. The learning rate is 0.001 and the number of iterations is 2000.

[0154] Step S5: Use the training set data to iteratively train the neural network model, input the test set data into the neural network model obtained in the above process for testing, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model, and continuously update the model parameters to improve the model convergence speed and prediction accuracy, and reduce the model oscillation during the convergence process:

[0155]

[0156]

[0157] Step S6: Use the validation set data to determine the accuracy of the neural network model, and repeat the experiment to obtain a high-precision model. Finally, the underwater mechanical noise predicted by the optimized high-precision model is obtained. Figure 3 shown.

[0158] In summary, the method proposed in this embodiment is based on the simulation data of the influence conditions of the underwater vehicle mechanical noise. It has a high prediction efficiency and can realize the rapid prediction of underwater mechanical noise. It can avoid the noise prediction delay caused by a large amount of calculation in the traditional numerical prediction method. In practical applications, experiments are carried out using real underwater vehicle mechanical noise data. By comparing the experimental results with the model prediction results, it is found that the mechanical noise prediction based on neural network has high accuracy and efficiency.

[0159] Furthermore, the method for rapid prediction of underwater vehicle mechanical noise proposed in this embodiment is applicable to the design, construction and navigation stages of underwater vehicles. The prediction of underwater mechanical noise based on only certain characteristic condition data such as the base, equipment, and vibration isolators can provide guidance for the covert navigation and vibration and noise reduction design of underwater vehicles. It has a wide range of applications and is applicable to all underwater vehicles that require mechanical noise prediction. It has high versatility, and the same series of underwater vehicles can use the same prediction model, which is economical and efficient and facilitates technology promotion.

[0160] Embodiment 4: The method for rapid prediction of underwater vehicle mechanical noise based on wavelet neural network optimized by Adam described in the above embodiment can be fully implemented by computer software. Therefore, correspondingly, this embodiment provides a rapid prediction system for underwater vehicle mechanical noise based on wavelet neural network optimized by Adam, and the rapid prediction system for underwater vehicle mechanical noise includes a storage device, and the storage device includes:

[0161] Used to determine whether the forecast object and forecast method are applicable;

[0162] It is used to determine characteristic condition data by analyzing the mechanical noise characteristics of underwater vehicles and the influence of various parameters on the underwater radiated noise results;

[0163] Used to simulate and preprocess characteristic condition data and construct sample data sets as training data;

[0164] Used to build a wavelet neural network model using a deep learning algorithm and set the number of neurons and the number of training iterations;

[0165] It is used to iteratively train and verify the wavelet neural network model, and uses the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0166] Used to obtain the optimal prediction model and realize rapid prediction of underwater mechanical noise.

[0167] The underwater vehicle mechanical noise rapid prediction system proposed in this embodiment is suitable for underwater navigation equipment, including but not limited to underwater submersibles, unmanned underwater vehicles, UUVs, and AUVs, which have high requirements on the underwater mechanical noise level and there is a need for rapid prediction and evaluation of underwater noise.

[0168] The underwater vehicle mechanical noise rapid prediction system proposed in this embodiment is based on the simulation data of the underwater vehicle mechanical noise influencing conditions. It has a high prediction efficiency and can realize the rapid prediction of underwater mechanical noise. It can avoid the noise prediction delay caused by a large amount of calculation in the traditional numerical prediction method. In actual applications, experiments are carried out using real underwater vehicle mechanical noise data. By comparing the experimental results and the model prediction results, it can be seen that the mechanical noise prediction based on the neural network has high accuracy and efficiency.

[0169] Furthermore, the underwater vehicle mechanical noise rapid prediction system proposed in this embodiment has the characteristics of fast prediction speed and high accuracy, and is applicable to the entire cycle of underwater vehicle design, construction, and use. Only specific parameters are needed to quickly obtain the underwater vehicle mechanical noise level.

[0170] Embodiment 5. This embodiment provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, a method for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network proposed in the above embodiment is implemented.

[0171] Embodiment 6. This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method for rapid prediction of mechanical noise of underwater vehicles based on Adam-optimized wavelet neural network proposed in the above embodiment is implemented.

[0172] The rapid prediction method of underwater vehicle mechanical noise is:

[0173] Step S1: Determine whether the forecast object and forecast method are applicable;

[0174] Step S2: determining characteristic condition data by analyzing the mechanical noise characteristics of the underwater vehicle and the influence of various parameters on the underwater radiation noise results;

[0175] Step S3: Simulate and preprocess the characteristic condition data to construct a sample data set as training data;

[0176] Step S4: Use deep learning algorithm to build a wavelet neural network model and set the number of neurons and the number of training iterations;

[0177] Step S5: iteratively train and verify the wavelet neural network model, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model;

[0178] Step S6: Obtain the optimal prediction model to achieve rapid prediction of underwater mechanical noise.

[0179] A computer device provided in this embodiment, the hardware device of this part is a general model and is not shown in the form of a diagram. The system includes a processor and a memory, wherein the processor and the memory can be connected via a bus or other means. The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, and corresponding program instructions / modules. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, so as to realize the method and steps for rapid prediction of mechanical noise of underwater vehicles based on Adam optimized wavelet neural network in the above method embodiment.

[0180] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0181] The above description is only the implementation mode of the present invention and is not limited to the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A rapid prediction method for underwater vehicle mechanical noise based on Adam optimized wavelet neural network, characterized in that: The method is: Step S1: Determine whether the forecast object and forecast method are applicable; Step S2: determining characteristic condition data by analyzing the mechanical noise characteristics of the underwater vehicle and the influence of various parameters on the underwater radiation noise results; Step S3: Simulate and preprocess the characteristic condition data to construct a sample data set as training data; Step S4: Use deep learning algorithm to build a wavelet neural network model and set the number of neurons and the number of training iterations; Step S5: iteratively train and verify the wavelet neural network model, and use the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model; Step S6: Obtain the optimal prediction model to achieve rapid prediction of underwater mechanical noise.

2. According to claim 1, a method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network is characterized in that: The rapid prediction method for mechanical noise is applicable to underwater equipment that has the need for rapid prediction and evaluation of underwater mechanical noise, including but not limited to underwater submersibles, unmanned underwater vehicles, UUVs, and AUVs.

3. The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network according to claim 1 is characterized in that: Characteristic condition data include mechanical equipment speed and power, vibration isolation measures, base form and web and panel thickness, shell material and thickness, and fluid medium density.

4. The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: cross validation is used during the training process to divide the sample data into a training set, a test set and a validation set; Step S32: numerically encode the characteristic condition data and perform normalization preprocessing on the input variables; Step S33: Convert the data into a normal distribution.

5. The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network according to claim 4 is characterized in that: Step S4 is specifically as follows: Step S41: The wavelet neural network model comprises an input layer, a hidden layer and an output layer, wherein the input layer has m neurons, representing the number of nodes of the input data, the hidden layer has s neurons, and the output layer has n neurons, outputting the frequency domain signal of the mechanical noise of the underwater vehicle; Step S42: the activation function of the hidden layer node of the wavelet neural network model is the wavelet basis function; Step S43: adjusting the wavelet neural network model parameters according to the error between the actual result and the network output result; Step S44: One training of the wavelet neural network means completing one forward propagation and one reverse error correction. When the neural network is continuously trained and the output parameters meet the specified requirements, the wavelet neural network will stop training.

6. The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network according to claim 5 is characterized in that: Step S5 is specifically as follows: Step S51: using the training set data to continuously iterate and train the wavelet neural network model until a high-precision forecasting effect is achieved; Step S52: According to the determined wavelet neural network model, the Adam optimization algorithm is used to update the network weights: Taking the tth iteration of gradient descent as an example, update the network weights; t=t+1 In the formula, m t and v t are the first-order momentum term and the second-order momentum term, β1 and β2 are hyperparameters with default values ​​of 0.9 and 0.999, respectively. t and v t The deviation correction value is and Step S53: Update the parameters of the determined wavelet neural network model again to obtain a wavelet neural network prediction model based on Adam optimization: Parameters for the tth iteration for: In the formula, The model parameter α at the t-1th iteration is the learning rate. The default value is 0.

001. ε is a very small number. Its function is to avoid the denominator being 0. Take ε as 10 -8 .

7. The method for rapid prediction of underwater vehicle mechanical noise based on Adam optimized wavelet neural network according to claim 1 is characterized in that: Step S6 is specifically as follows: The underwater vehicle mechanical noise prediction model is evaluated according to the prediction results and prediction conditions to obtain the optimal prediction model and realize the rapid prediction of underwater radiated noise.

8. A rapid prediction system for underwater vehicle mechanical noise based on Adam optimized wavelet neural network, characterized in that: The system includes a storage device, the storage device including: Used to determine whether the forecast object and forecast method are applicable; It is used to determine characteristic condition data by analyzing the mechanical noise characteristics of underwater vehicles and the influence of various parameters on the underwater radiated noise results; Used to simulate and preprocess characteristic condition data and construct sample data sets as training data; Used to build a wavelet neural network model using a deep learning algorithm and set the number of neurons and the number of training iterations; It is used to iteratively train and verify the wavelet neural network model, and uses the Adam optimization algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the wavelet neural network model; Used to obtain the optimal prediction model and realize rapid prediction of underwater mechanical noise.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for rapid prediction of mechanical noise of underwater vehicles based on Adam-optimized wavelet neural network as described in any one of claims 1 to 7 is implemented.

10. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, a method for rapid prediction of mechanical noise of underwater vehicles based on an Adam-optimized wavelet neural network is implemented as described in any one of claims 1 to 7.

Images