Method and system for rapidly forecasting hydrodynamic noise of underwater vehicle based on BP neural network optimized by genetic algorithm

Abstract

The invention discloses an underwater vehicle hydrodynamic noise rapid forecasting method and system based on a BP neural network optimized by a genetic algorithm, belongs to the field of ship underwater radiation noise forecasting, and designs a BP neural network noise forecasting model optimized by the genetic algorithm. Key variables including shell thickness, molded line parameters, surface roughness, material attributes, sailing speed, fluid medium density, viscosity coefficient and underwater vehicle diving depth are selected as characteristic conditions. The initial weight and threshold value of the optimal BP network are obtained through the global optimization capacity of the genetic algorithm, the optimal initial weight and threshold value obtained through the optimization algorithm serve as the initial weight and threshold value of the BP neural network, and training is carried out to avoid falling into the local minimum value. And performing iterative training on the optimization model by using the training set to obtain a high-precision noise forecasting model, thereby realizing forecasting of the hydrodynamic noise of the underwater vehicle. The method is suitable for forecasting the hydrodynamic noise of the underwater vehicle, and has the advantages of wide applicability, high accuracy and the like.

Description

Technical Field

[0001] The invention belongs to the field of prediction of underwater radiation noise of ships, and in particular relates to a method for quickly predicting hydrodynamic noise of underwater vehicles based on genetic algorithm optimization of BP neural network. Background Art

[0002] With the transformation and upgrading of the shipping industry, the concept of green and environmental protection of ships is constantly improving. The impact of vibration and noise generated by marine platforms and ships on the health of seafarers and marine life is increasingly attracting people's attention. As one of the three major noise sources, hydrodynamic noise is caused by many factors such as the pulsating pressure in the turbulent boundary layer on the surface of the navigation body, the vibration of the wall structure under pulsating excitation, and the vortex caused by the roughness of the surface. With the increase of the speed of the underwater vehicle, the hydrodynamic noise increases rapidly and accounts for the main part of the total noise. The propagation of hydrodynamic noise is affected by many factors, including the characteristics of the fluid medium, the shape and motion state of the navigation body, etc. The complexity and uncertainty of these factors make it difficult to establish an accurate propagation model. Hydrodynamic noise has an important impact on the concealment and comfort of ships. It is necessary to make a more accurate prediction of its hydrodynamic noise in the early stage of construction of ships and marine engineering structures, so as to adopt appropriate vibration and noise reduction measures as early as possible in the preliminary design and construction stage to achieve the control of hydrodynamic noise, so as to more efficiently build ships that meet the requirements. At present, the commonly used methods for predicting hydrodynamic noise mainly rely on numerical calculation methods in addition to using semi-empirical and semi-theoretical engineering estimation methods. The engineering estimation method is mainly a semi-theoretical and semi-empirical formula formed by combining structural sound radiation theory and actual structural test data. It is easy and quick to use, but it is mainly suitable for vibration noise prediction of similar structures. The numerical calculation method mainly includes statistical energy method, finite element method, boundary element method and other methods. Its effectiveness has been generally recognized by the engineering community, but the preliminary processing is relatively cumbersome and requires a more detailed model to be established in the computer. It also has poor generalization and the prediction takes a long time.

[0003] The intelligent algorithm can directly use the structure of ships and platforms or the equipment on them as parameters, and does not require large-scale model building, so its prediction speed is greatly improved compared to the numerical method. At the same time, the model required by the intelligent algorithm 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 can be foreseen that as time goes by, the intelligent algorithm can continuously approach or even surpass the numerical algorithm in accuracy while maintaining the rapidity of prediction. Therefore, the rapid prediction of hydrodynamic noise of underwater vehicles based on artificial intelligence has important engineering significance.

[0004] 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 hydrodynamic noise of underwater vehicles based on BP neural network optimized by genetic algorithm. 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. Research on engineering estimation method of underwater structure radiated noise [J]. Journal of Harbin Engineering University, 2002(01): 91-94; 3. Numerical prediction of underwater vehicle resistance and flow noise characteristics [D]. Dalian University of Technology, 2022. DOI: 10.26991.

[0005] Among them, document 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 input, the sound pressure of the sound field assessment point as output, and the error descent curve, the percentage error of the test sample detection result and the prediction pass rate as 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 pass rate is low, and the prediction effect of this method is poor; document 2 proposes an estimation method based on statistical assumptions. This method is based on theoretical analysis and experimental research. After determining the radiation efficiency of structural noise, the mean square acceleration distribution of the structure surface is obtained by direct or indirect methods, so as to realize the engineering estimation method of underwater structural radiation noise. In addition, comparative experimental studies were carried out. The results showed that the estimation method is simple and feasible and can be easily applied in practical engineering. However, its applicability still needs further experimental verification. Reference 3 used the Myring equation to determine the boat types with different head shapes, tail shapes, and parallel mid-body lengths. The Realizable6)-turbulence model was used based on the CFD method to calculate the resistance of different boat types. The large eddy simulation turbulence model combined with the FW-H equation was used to calculate the flow noise of different boat types. However, its applicability to complex structures has not been effectively verified.

[0006] In summary, the currently available publicly searchable literature on rapid prediction of hydrodynamic noise still has problems such as low pass rate of prediction results, applicability in practical applications, and applicability to complex structures that have not been effectively verified. Summary of the invention

[0007] The formation mechanism of underwater vehicle hydrodynamic noise is complex, involving multiple sound source models, each of which represents different source motion, flow field characteristics, sound field directivity and sound radiation efficiency, so it is difficult to accurately and efficiently predict it. Aiming at the problem of rapid prediction of underwater vehicle hydrodynamic noise, the present invention builds a BP neural network algorithm based on genetic algorithm optimization, which can realize the rapid prediction of underwater vehicle hydrodynamic noise.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] The present invention provides a method for quickly predicting the hydrodynamic noise of an underwater vehicle based on a genetic algorithm optimized BP neural network. The prediction method comprises the following steps:

[0010] Step S1: Determine whether the forecast object and forecast method are applicable;

[0011] Step S2: determining characteristic condition data by analyzing the hydrodynamic noise characteristics of the underwater vehicle and the influence of various parameters on the hydrodynamic noise results;

[0012] Step S3: Perform finite element simulation on the underwater vehicle and its characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0013] Step S4: Using deep learning algorithm to build BP neural network model;

[0014] Step S5: iteratively train and test the BP neural network model according to the sample data set, and use a genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model;

[0015] Step S6: Use the optimal BP neural network prediction model to quickly predict hydrodynamic noise.

[0016] Furthermore, there is a preferred embodiment that the above-mentioned hydrodynamic noise rapid prediction method is suitable for underwater navigation equipment that has high requirements on hydrodynamic noise performance and has the need for rapid prediction and evaluation of underwater noise, including but not limited to underwater submersibles, unmanned submersibles, UUVs and AUVs.

[0017] Furthermore, in a preferred embodiment, the characteristic condition data for noise prediction include underwater vehicle hull thickness, profile parameters, surface roughness, material properties, navigation speed, fluid medium density, viscosity coefficient and vehicle diving depth.

[0018] Furthermore, in another preferred embodiment, the above step S3 further includes:

[0019] S31: During network training, the sample data set is randomly divided into a training set and a test set, wherein the training set accounts for 70% of the total sample data, the test set accounts for 30%, and the total sample data set is not less than 3000;

[0020] S32: Since the number of samples of underwater vehicle noise data is small, cross validation is used in the training process to make full use of the data;

[0021] S33: 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 it is necessary to use Ordinal Encoder to numerically encode the source data;

[0022] S34: Standardize the input data to ensure that the data is within a reasonable range, which is conducive to the training and convergence of the neural network model. Use the Box-Cox transformation method to convert the data into a normally distributed data to reduce the impact of outliers and produce a more robust model that is less sensitive to extreme values.

[0023]

[0024] In the formula, y represents the original data and λ represents the transformation parameter.

[0025] Furthermore, in another preferred embodiment, the above step S4 specifically includes:

[0026] S41: The hidden layer of the BP neural network adopts a single hidden layer or a multi-hidden layer structure to complete the approximation and the mapping from n-dimension to m-dimension, and determine the number of BP neural network layers. For example, a 3-layer BP neural network is selected for modeling:

[0027] a [1](i) =g(θ [1] X (i) +b [1] )

[0028]

[0029] In the formula, a [1](i) represents the output of the i-th sample in the hidden layer, θ [1] represents the network weight between the input layer and the hidden layer, X (i) represents the i-th input sample, b [1] represents the hidden layer neuron threshold, represents the i-th sample output, θ [2] represents the weight between the hidden layer and the output layer, b [2] Represents the output layer neuron threshold, and the above parameters all represent matrix vectors;

[0030] S42: A deeper structure will bring great challenges to the training convergence process of the neural network, so residual connections are introduced into the network structure to suppress the vanishing gradient;

[0031] S43: In order to make the BP neural network approach any nonlinear function and map the input value to the vicinity of 0-1, reduce the data discrimination and standardize the output of the neuron, Sigmoid is selected as the hidden layer activation function, and Softmax is selected as the output layer activation function to enable the neural network to perform multi-task classification;

[0032] Sigmoid function:

[0033] g(x)=1 / (1+e -x )

[0034] Softmax function:

[0035]

[0036] S44: Import the cost function into the BP neural network structure to measure the error between the predicted value of the intelligent algorithm and the actual value:

[0037]

[0038] S45: According to the properties of BP neural network, back propagation and gradient descent algorithm are used to update the weight θ of BP neural network [l] and threshold b [l] :

[0039]

[0040] In the formula, l = 1, 2, α is the learning rate, and the default value is 0.001.

[0041] Furthermore, in another preferred embodiment, the above step S5 specifically includes:

[0042] S51: Use the training set data to continuously iteratively train the BP neural network model until a high-precision forecasting effect is achieved;

[0043] S52: Using genetic algorithm, the weights and thresholds of the BP neural network are updated and optimized by fixing the topology of the network:

[0044]

[0045] In the formula, θ and b represent weights and thresholds respectively, Ω represents the value space of weights and thresholds, and a i is the output of the i-th sample hidden layer, y i is the output of the network;

[0046] S53: Obtaining the optimal BP neural network prediction model according to the optimized weights and thresholds.

[0047] Furthermore, in another preferred embodiment, the above step S6 specifically includes:

[0048] The underwater radiated noise is quickly predicted based on the optimal BP neural network prediction model, and the underwater vehicle hydrodynamic noise prediction model is evaluated based on the prediction results and prediction conditions.

[0049] The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP 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 hydrodynamic noise based on genetic algorithm optimized BP neural network, wherein the system storage device comprises:

[0050] Used to determine whether the forecast object and forecast method are applicable;

[0051] It is used to determine characteristic condition data by analyzing the hydrodynamic noise characteristics of underwater vehicles and the influence of various parameters on the hydrodynamic noise results of underwater vehicles;

[0052] It is used to perform finite element simulation on underwater vehicles and their characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0053] Used to build BP neural network model using deep learning algorithm;

[0054] It is used to iteratively train and test the BP neural network model according to the sample data set, and use the genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model;

[0055] It is used to quickly predict hydrodynamic noise using the optimal BP neural network prediction model.

[0056] 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 hydrodynamic noise of underwater vehicles based on genetic algorithm optimized BP neural network is implemented.

[0057] 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 hydrodynamic noise of an underwater vehicle based on a genetic algorithm-optimized BP neural network is implemented.

[0058] The technical solution of the present invention has the following beneficial effects:

[0059] 1. The present invention provides a method for rapid prediction of hydrodynamic noise of underwater vehicles based on genetic algorithm optimized BP neural network. By adopting genetic algorithm to optimize BP neural network, problem solving and global optimization are realized. The optimized BP neural network model is more efficient in the training and prediction process, can more accurately reflect the changes in noise characteristics in complex underwater environments, and can dynamically adapt to changes in the underwater environment to maintain the real-time and accuracy of the prediction.

[0060] Furthermore, compared with the prior art, the present invention proposes a new method for rapid prediction of hydrodynamic noise of underwater vehicles by optimizing BP neural network with genetic algorithm, and can significantly improve the prediction accuracy and efficiency of hydrodynamic noise of underwater vehicles.

[0061] 2. Traditional BP neural networks may encounter problems such as long training time and slow convergence speed. The method for rapid prediction of hydrodynamic noise proposed in the present invention is based on the adaptive learning rate adjustment algorithm of genetic algorithm to optimize and improve the traditional BP neural network. The genetic algorithm has global optimization capability and can automatically search and optimize the structural parameters and learning parameters of the BP neural network, etc. It can improve the network recognition accuracy and reduce oscillation during convergence, thereby realizing the rapid prediction of hydrodynamic noise of underwater vehicles.

[0062] 3. The rapid prediction method for hydrodynamic noise proposed in the present invention is based on the simulation data of the hydrodynamic noise influencing conditions of underwater vehicles. It can effectively process complex input data and output noise signals. It has a high prediction efficiency and can realize the rapid prediction of hydrodynamic noise. It can avoid the noise prediction delay caused by a large amount of calculation in traditional numerical prediction methods and can provide more sophisticated numerical simulation and prediction.

[0063] 4. The rapid prediction method for hydrodynamic noise proposed in the present invention has a fast prediction speed and is suitable for the design, construction and navigation stages of underwater vehicles. It can predict hydrodynamic noise only on the premise of mastering certain characteristic condition data such as shell thickness, material properties, navigation speed, etc., and can provide guidance for the stealth navigation and vibration and noise reduction design of underwater vehicles.

[0064] 5. The method disclosed in the present invention has a wide range of applications and can better adapt to the hydrodynamic noise prediction problems of different types of underwater vehicles. It has high generalization ability, and the same series of underwater vehicles can use the same prediction model, which is economical and efficient and convenient for technology promotion.

[0065] The present invention is applicable to the entire cycle of underwater vehicle design, construction and use, and only requires specific parameters to quickly obtain the hydrodynamic noise level of the underwater vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flow chart of a method for rapid prediction of hydrodynamic noise of underwater vehicles based on BP neural network optimized by genetic algorithm according to the present invention;

[0067] Figure 2 It is the prediction result of the hydrodynamic noise of the underwater vehicle described in the present invention. DETAILED DESCRIPTION

[0068] 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.

[0069] Implementation method 1, see Figure 1 This embodiment describes a method for rapidly predicting the hydrodynamic noise of an underwater vehicle based on a genetic algorithm to optimize a BP neural network, comprising the following steps:

[0070] Step S1: Determine whether the forecast object and forecast method are applicable;

[0071] Step S2: determining characteristic condition data by analyzing the hydrodynamic noise characteristics of the underwater vehicle and the influence of various parameters on the hydrodynamic noise results;

[0072] Step S3: Perform finite element simulation on the underwater vehicle and its characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0073] Step S4: Using deep learning algorithm to build BP neural network model;

[0074] Step S5: iteratively training and testing the BP neural network model according to the sample data set, and optimizing the gradient descent of the BP neural network model using a genetic algorithm to obtain the optimal BP neural network prediction model;

[0075] Step S6: Use the optimal BP neural network prediction model to quickly predict hydrodynamic noise.

[0076] In practical application, if Figure 1As shown in the figure, by analyzing the characteristics of the hydrodynamic noise of underwater vehicles and the influence of the hydrodynamic noise results of underwater vehicles, the characteristic condition data used for noise prediction is determined, so that the prediction of hydrodynamic noise can be achieved under the premise of mastering certain characteristic condition data such as shell thickness, material properties, navigation speed, etc.; by optimizing and improving the traditional BP neural network with an adaptive learning rate adjustment algorithm based on genetic algorithm, the network recognition accuracy can be improved, and the oscillation can be reduced during convergence, thereby realizing the rapid prediction of hydrodynamic noise of underwater vehicles.

[0077] The prediction method described in this embodiment is based on the simulation data of the influence conditions of hydrodynamic noise of underwater vehicles. It can effectively process complex input data and output noise signals. It has a high prediction efficiency and can achieve rapid prediction of hydrodynamic noise. It can avoid the noise prediction delay caused by a large amount of calculation in traditional numerical prediction methods and can provide more sophisticated numerical simulation and prediction.

[0078] Implementation method 2: This implementation method is to illustrate the specific steps of the method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimization of BP neural network described in implementation method 1 by way of example;

[0079] Step S1: Determine whether the forecast object and forecast method are applicable;

[0080] Preferably, the step S1 specifically includes:

[0081] The method for rapid prediction of hydrodynamic noise of underwater vehicles described in this embodiment is suitable for underwater navigation equipment that has high requirements on hydrodynamic noise performance and has the need for rapid prediction and evaluation of underwater noise, including but not limited to underwater submersibles, unmanned submersibles, UUVs (unmanned underwater vehicles) and AUVs (autonomous underwater vehicles).

[0082] Step S2: determining characteristic condition data for noise prediction by analyzing the hydrodynamic noise characteristics of the underwater vehicle and the influence of various parameters on the hydrodynamic noise results;

[0083] Preferably, the step S2 specifically includes:

[0084] Step S21: Analyze the characteristics and impact of the hydrodynamic noise of the underwater vehicle, and determine the simulation data of the characteristic conditions for noise prediction as a data set for noise prediction model training;

[0085] Step S22: selecting typical variables such as underwater vehicle hull thickness, profile parameters, surface roughness, material properties, navigation speed, fluid medium density, viscosity coefficient, and vehicle diving depth as input parameters;

[0086] Step S23: The variable data selected through analysis are highly representative and rapid forecasting can be achieved without other conditions.

[0087] Step S3: Perform finite element simulation on the underwater vehicle and its characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0088] Preferably, the step S3 further comprises:

[0089] Step S31: during network training, the data is randomly divided into a training set and a test set, wherein the training set accounts for 70% of the total sample data and the test set accounts for 30%. To fully train the model, the total sample data set should be no less than 3,000;

[0090] Step S32: Since the number of samples of underwater vehicle hydrodynamic noise data is relatively small, in order to make full use of the data, this embodiment adopts cross-validation in the training process, that is, repeatedly using data, dividing the obtained sample data, and combining them into different training sets and test sets.

[0091] Step S33: 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 Ordinal Encoder is used to numerically encode the source data;

[0092] Step S34: Neural networks can learn from various types of data distributions. Training data is not required to follow a normal distribution, but converting data to a normal distribution is beneficial to model training. Therefore, the input data is standardized to ensure that the data is within a reasonable range, which is beneficial to the training and convergence of the neural network model. The Box-Cox transformation method is used to convert the data into a normal distribution, reduce the impact of outliers, and produce a more robust model that is less sensitive to extreme values.

[0093]

[0094] In the formula, y represents the original data and λ represents the transformation parameter.

[0095] Step S4: Using deep learning algorithm to build BP neural network model;

[0096] Preferably, the step S4 specifically includes:

[0097] Step S41: using a single or multi-hidden layer BP neural network to complete the approximation and the mapping from n-dimension to m-dimension, and selecting a suitable number of BP neural network layers for modeling;

[0098] For example, a single hidden layer BP neural network can be used to complete the approximation, and a 3-layer BP neural network can complete the mapping from n-dimension to m-dimension. This forecast uses a 3-layer BP neural network for modeling.

[0099] a [1](i) =g(θ [1] X (i) +b [1] )

[0100]

[0101] In the formula, a [1](i) represents the output of the i-th sample in the hidden layer, θ [1] represents the network weight between the input layer and the hidden layer, X (i) represents the i-th input sample, b [1] represents the hidden layer neuron threshold, represents the i-th sample output, θ [2] represents the weight between the hidden layer and the output layer, b [2] Represents the output layer neuron threshold, and the above parameters all represent matrix vectors;

[0102] Step S42: A deeper structure will bring great challenges to the training convergence process of the neural network, so a residual connection is introduced into the network structure to suppress the vanishing gradient;

[0103] Step S43: In order to make the BP neural network approach any nonlinear function and map the input value to the vicinity of 0-1, reduce the data discrimination and standardize the output of the neuron, Sigmoid is selected as the hidden layer activation function, and Softmax is selected as the output layer activation function to make the neural network perform multi-task classification;

[0104] The Sigmoid function is:

[0105] g(x)=1 / (1+e -x )

[0106] The Softmax function is:

[0107]

[0108] Step S44: Importing a cost function into the BP neural network structure; the imported cost function is used to measure the error between the predicted value of the intelligent algorithm and the true value:

[0109]

[0110] Step S45: According to the properties of the BP neural network and the correlation between the input of the neural network model and the overall noise level, the relationship between the characteristic variables and the target variables and the characteristic variables is analyzed to update the weight θ of the BP neural network by back propagation and gradient descent algorithm. (J) and threshold b (J) .

[0111]

[0112] In the formula, l = 1, 2, α is the learning rate, and the default value is 0.001.

[0113] Step S5: iteratively train and test the neural network model, and use a genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model;

[0114] Preferably, the step S5 specifically includes:

[0115] Step S51: using the training set data to continuously iterate and train the neural network model until a high-precision model is achieved;

[0116] Step S52: input the test set data into the neural network model obtained in the above process, and predict the result data of the underwater vehicle hydrodynamic noise;

[0117] Step S53: using a genetic algorithm to update and optimize the weights and thresholds of the BP neural network by fixing the topological structure of the network;

[0118]

[0119] In the formula, θ and b represent weights and thresholds respectively, Ω represents the value space of weights and thresholds, and a i is the output of the i-th sample hidden layer, y i is the output of the network;

[0120] Step S54: By solving the above equations, the updated optimized weights and thresholds can be obtained, and the ultimate optimal BP neural network prediction model based on genetic algorithm optimization can be obtained.

[0121] Step S6: Use the optimal BP neural network prediction model to quickly predict underwater radiation noise.

[0122] Preferably, the step S6 specifically includes:

[0123] According to the optimal BP neural network model optimized by genetic algorithm determined in the above steps, the prediction results are compared with the actual results to evaluate the performance of the hydrodynamic noise prediction model.

[0124] Compared with the existing underwater radiation noise prediction method, the method for rapid prediction of underwater vehicle hydrodynamic noise based on BP neural network optimized by genetic algorithm described in this embodiment has global optimization capability, can automatically search and optimize the structural parameters and learning parameters of BP neural network, can improve the network recognition accuracy, and reduce oscillation during convergence, thereby realizing rapid prediction of underwater vehicle hydrodynamic noise;

[0125] Furthermore, the rapid prediction method for hydrodynamic noise proposed in this embodiment can effectively process complex input data and output noise signals, has a high prediction efficiency, can achieve rapid prediction of hydrodynamic noise, avoid noise prediction delays caused by a large amount of calculation in traditional numerical prediction methods, and provide more sophisticated numerical simulation and prediction.

[0126] Implementation method three, see Figure 2 This embodiment is described as an example of a method for rapid prediction of hydrodynamic noise of underwater vehicles based on genetic algorithm optimization of BP neural network described in the above embodiment;

[0127] Step S1: Determine the hydrodynamic noise prediction object according to the underwater vehicle type and navigation conditions. This embodiment simplifies the underwater vehicle to a certain extent and adopts a cone-column-spherical shell structure containing a base as the research object;

[0128] Step S2: by analyzing and selecting characteristic conditions, five variables, namely, base form, equipment type, vibration isolation measures, web thickness, and panel thickness, are set as input variables;

[0129] 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;

[0130] Step S4: Build a 3-layer BP neural network model, import the loss function and Sigmoid hidden layer activation function, use the chain derivation rule to calculate the gradient, and minimize the cost function J(θ,b) along the reverse gradient direction, calculate layer by layer from the output layer to the input layer, and iteratively optimize the neural network through reverse propagation of the error;

[0131]

[0132] g(x)=1 / (1+e -x )

[0133] Step S5: Use the training set data to iteratively train the neural network model, repeatedly experiment to obtain a high-precision model, input the test set data into the neural network model obtained in the above process, and use a genetic algorithm with exponential weighted average and deviation correction to optimize the gradient descent of the BP neural network model obtained in step S4, and continuously update the weights and thresholds to improve the model convergence speed and prediction accuracy, and reduce model oscillation during the convergence process.

[0134] m t =β1*m t-1 +(1-β1)*dθ t-1 [l]

[0135] v t =β2·v t-1 +(1-β2)·(dθ t-1 [l] ) 2

[0136]

[0137] Step S6: Obtain a high-precision prediction model to achieve rapid prediction of underwater vehicle hydrodynamic noise, and finally obtain the predicted hydrodynamic noise result as follows: Figure 2 shown.

[0138] Embodiment 4: The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimization of BP neural network described in embodiment 1 can be fully implemented by computer software. Therefore, correspondingly, this embodiment provides a rapid prediction system for underwater vehicle hydrodynamic noise based on genetic algorithm optimization of BP neural network, and the rapid prediction system for hydrodynamic noise includes a storage device, and the storage device includes:

[0139] Used to determine whether the forecast object and forecast method are applicable;

[0140] It is used to determine characteristic condition data by analyzing the hydrodynamic noise characteristics of underwater vehicles and the influence of various parameters on the hydrodynamic noise results of underwater vehicles;

[0141] It is used to perform finite element simulation on underwater vehicles and their characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0142] Used to build BP neural network model using deep learning algorithm;

[0143] It is used to iteratively train and test the BP neural network model according to the sample data set, and use the genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model;

[0144] It is used to quickly predict hydrodynamic noise using the optimal BP neural network prediction model.

[0145] The underwater vehicle hydrodynamic noise rapid prediction system proposed in this embodiment is suitable for underwater navigation equipment that has high requirements on hydrodynamic noise performance and has the need for rapid prediction and evaluation of underwater noise, including but not limited to underwater submersibles, unmanned submersibles, UUVs (unmanned underwater vehicles) and AUVs (autonomous underwater vehicles).

[0146] The rapid prediction system for hydrodynamic noise proposed in this embodiment is based on the simulation data of the hydrodynamic noise influencing conditions of underwater vehicles. It can effectively process complex input data and output noise signals, has high prediction efficiency, can realize rapid prediction of hydrodynamic noise, can avoid the noise prediction delay caused by a large amount of calculation in traditional numerical prediction methods, can provide more precise numerical simulation and prediction, and has a wide range of applications. It can better adapt to the hydrodynamic noise prediction problems of different types of underwater vehicles, has high generalization ability, and the same series of underwater vehicles can use the same prediction model, which is economical and efficient and convenient for technology promotion.

[0147] 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 hydrodynamic noise of an underwater vehicle based on a genetic algorithm optimized BP neural network proposed in the above embodiment is implemented.

[0148] 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 hydrodynamic noise of an underwater vehicle based on a genetic algorithm-optimized BP neural network proposed in the above embodiment is implemented.

[0149] The rapid prediction method of underwater vehicle hydrodynamic noise is:

[0150] Step S1: Determine whether the forecast object and forecast method are applicable;

[0151] Step S2: determining characteristic condition data by analyzing the hydrodynamic noise characteristics of the underwater vehicle and the influence of various parameters on the hydrodynamic noise results;

[0152] Step S3: Perform finite element simulation on the underwater vehicle and its characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data;

[0153] Step S4: Using deep learning algorithm to build BP neural network model;

[0154] Step S5: iteratively training and testing the BP neural network model according to the sample data set, and optimizing the gradient descent of the BP neural network model using a genetic algorithm to obtain the optimal BP neural network prediction model;

[0155] Step S6: Use the optimal BP neural network prediction model to quickly predict hydrodynamic noise.

[0156] 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 hydrodynamic noise of underwater vehicles based on genetic algorithm optimization of BP neural network in the above method embodiment.

[0157] 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 method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP 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 hydrodynamic noise characteristics of the underwater vehicle and the influence of various parameters on the hydrodynamic noise results; Step S3: Perform finite element simulation on the underwater vehicle and its characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data; Step S4: Using deep learning algorithm to build BP neural network model; Step S5: iteratively train and test the BP neural network model according to the sample data set, and use a genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model; Step S6: Use the optimal BP neural network prediction model to quickly predict hydrodynamic noise.

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

3. The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP neural network according to claim 1 is characterized in that: The characteristic condition data include the underwater vehicle hull thickness, profile parameters, surface roughness, material properties, navigation speed, fluid medium density, viscosity coefficient and vehicle diving depth.

4. The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP neural network according to claim 1 is characterized in that: Step S3 also includes: S31: The sample data set is randomly divided into a training set and a test set, with the training set accounting for 70% of the total sample data and the test set accounting for 30%. The total sample data set is no less than 3,000; S32: Cross validation is used during training; S33: Ordinal Encoder is used to numerically encode the source data; S34: The input data are transformed into normal distribution data using the Box-Cox transformation method.

5. The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP neural network according to claim 4 is characterized in that: Step S4 is specifically as follows: S41: The hidden layer of the BP neural network adopts a single hidden layer or a multi-hidden layer structure to complete the approximation and the mapping from n-dimension to m-dimension, and determine the number of BP neural network layers. For example, a 3-layer BP neural network is selected for modeling: a [1](i) =g(θ [1] X (i) +b [1] ) In the formula, a [1](i) represents the output of the i-th sample in the hidden layer, θ [1] represents the network weight between the input layer and the hidden layer, X (i) represents the i-th input sample, b [1] represents the hidden layer neuron threshold, represents the i-th sample output, θ [2] represents the weight between the hidden layer and the output layer, b[ 2 ] represents the output layer neuron threshold, and the above parameters all represent matrix vectors; S42: Introducing residual connection into BP neural network structure; S43: Sigmoid is selected as the activation function of the hidden layer of the BP neural network, and Softmax is selected as the activation function of the output layer; S44: Import the cost function into the BP neural network structure; S45: Update the weight θ of the BP neural network using back propagation and gradient descent algorithm (J) and threshold b (J) ; In the formula, l = 1, 2, α is the learning rate, the default value is 0.

001.

6. The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP neural network according to claim 5 is characterized in that: Step S5 is specifically as follows: S51: Use the training set data to continuously iteratively train the BP neural network model until a high-precision forecasting effect is achieved; S52: Using genetic algorithm, the weights and thresholds of the BP neural network are updated and optimized by fixing the topology of the network: In the formula, θ, b represent the weight and threshold respectively, Ω represents the value space of weight and threshold, a i is the output of the hidden layer of the i-th sample, and yi is the output result of the network; S53: Obtaining the optimal BP neural network prediction model according to the optimized weights and thresholds.

7. The method for rapid prediction of underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP neural network according to claim 1 is characterized in that: Step S6 is specifically as follows: The hydrodynamic noise is quickly predicted based on the optimal BP neural network prediction model, and the underwater vehicle hydrodynamic noise prediction model is evaluated based on the prediction results and prediction conditions.

8. A rapid prediction system for underwater vehicle hydrodynamic noise based on genetic algorithm optimized BP 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 hydrodynamic noise characteristics of underwater vehicles and the influence of various parameters on the hydrodynamic noise results of underwater vehicles; It is used to perform finite element simulation on underwater vehicles and their characteristic condition data to obtain hydrodynamic noise result data, and perform preprocessing to construct a sample data set as training data; Used to build BP neural network model using deep learning algorithm; It is used to iteratively train and test the BP neural network model according to the sample data set, and use the genetic algorithm to optimize the gradient descent of the BP neural network model to obtain the optimal BP neural network prediction model; It is used to quickly predict hydrodynamic noise using the optimal BP neural network prediction model.

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 quickly predicting the hydrodynamic noise of an underwater vehicle based on a genetic algorithm optimized BP neural network as described in any one of claims 1 to 7 is implemented.

10. A computer device, characterized in that: The method comprises 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 hydrodynamic noise of an underwater vehicle based on a genetic algorithm-optimized BP neural network as described in any one of claims 1 to 7 is implemented.

Images