Method and system for rapidly forecasting noise of underwater vehicle based on long short-term memory neural network
The LSTM neural network method addresses the inefficiencies of existing prediction methods by providing rapid and precise underwater vehicle noise forecasting, leveraging data-driven learning to overcome computational and environmental challenges.
Patent Information
- Application Number
- CN202411299535.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-12-17
AI Technical Summary
Existing methods for underwater vehicle noise prediction are plagued by long time consumption and low precision, particularly in handling complex scenarios and large structures, and require extensive data preprocessing and feature engineering.
A method utilizing a long short-term memory (LSTM) neural network for rapid prediction of underwater vehicle noise, involving data collection, preprocessing, model construction, training, and validation to achieve accurate noise forecasting without complex simulation environments.
The LSTM neural network model enables quick and accurate prediction of underwater vehicle noise across various types and conditions, adapting to different environments and parameter changes, while reducing computational time and resource requirements.
Smart Images
Figure CN119149911A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of underwater vehicle noise prediction, and in particular relates to a method for rapid prediction of underwater vehicle noise based on a long short-term memory neural network. Background Art
[0002] According to the different properties of noise sources, underwater noise of underwater vehicles is generally divided into mechanical noise (or structural noise), propeller noise and hydrodynamic noise (flow noise). At present, the main research methods for noise prediction of underwater vehicles include: analytical method, finite element-boundary element method, statistical energy method and intelligent algorithm.
[0003] The analytical method is based on the theory of structural vibration and acoustic radiation. The acoustic quantity is obtained according to the boundary conditions that the shell sound pressure should meet. It can be used as a prediction tool when the frequency is low (lower than the ring frequency), but there are still problems such as complex calculation and difficulty in multi-band noise prediction. The finite element-boundary element method discretizes the underwater vehicle structure, connects the units and calculates them using the motion differential matrix equation. In principle, this method can calculate the boundary shape of various structures. It has the characteristics of strong adaptability and relatively accurate numerical calculation. It is currently widely used to solve structural vibration problems and solve sound field problems in a limited area. However, for large structures or structures vibrating at medium and high frequencies, the finite element-boundary element method requires a large model or a very fine grid division, which greatly increases the amount of numerical calculation and calculation time. The statistical energy method is based on the principle of statistical physics and uses the statistical average of time and space to reflect the average level of underwater vehicle structural vibration and acoustic radiation. It can well describe the average vibration and acoustic characteristics of each component of the system in the medium and high frequency range, but when calculating in the lower frequency band, it is difficult to obtain accurate prediction results due to the small modal density.
[0004] With the advancement of AI technology, intelligent algorithms have gradually expanded their applications to various fields. Machine learning is the core of artificial intelligence. As an important algorithm of machine learning, the neural network model has broad application prospects in the field of underwater vehicle noise prediction. Compared with the above methods, the existing prediction methods usually rely on mathematical models and statistical methods, requiring a lot of data preprocessing and feature engineering. However, neural networks do not work by simulating the way underwater vehicles generate noise, but by mining a large amount of historical data to learn noise patterns. This type of method learns the nonlinear mapping relationship between input variables and output variables from the data, so it can predict underwater vehicle noise more quickly and efficiently. In addition, neural networks have better scalability and can better adapt to the increasing amount of data and more complex scenarios. Therefore, it is of great practical significance to realize intelligent algorithms to predict underwater vehicle noise. Summary of the invention
[0005] The present invention provides a method for rapid underwater vehicle prediction based on a long short-term memory neural network, so as to overcome the technical problems of long time consumption and low precision in the existing underwater vehicle prediction methods.
[0006] To achieve the above object, the present invention provides the following solutions:
[0007] The present invention provides a method for rapid prediction of underwater vehicle noise based on a long short-term memory neural network. The method for rapid prediction of underwater vehicle noise specifically comprises the following steps:
[0008] Step S1: Collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0009] Step S2: preprocessing the underwater vehicle noise data and dividing it into a training set, a test set and a validation set;
[0010] Step S3: construct a long short-term memory neural network model and define the initial values of various weights and thresholds;
[0011] Step S4: input the training set data into the long short-term memory neural network model for model training;
[0012] Step S5: input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0013] Step S6: Compare the model prediction results with the validation set data to verify the accuracy of the model;
[0014] Step S7: Use the optimal long short-term memory neural network model to quickly predict the noise of the underwater vehicle.
[0015] Furthermore, in another preferred embodiment, the above step S1 specifically includes:
[0016] Step S11: the propulsion system data includes the propulsion system type, rotation speed, and power; the underwater vehicle structure and performance parameters include the hull material, size, and the speed and hydrodynamic parameters of the vehicle; the mechanical vibration data is the vibration data of the hull and equipment, including vibration amplitude and frequency;
[0017] Step S12: The data set is collected through channels including real-time sensor collection, underwater vehicle historical noise system records, and public data sets.
[0018] Furthermore, in another preferred embodiment, the above step S2 specifically includes:
[0019] Step S21: Considering that the propulsion system type and hull material data are discrete feature data, one-hot vector encoding is used for storage, where each propulsion system type and hull material type has its own Boolean field;
[0020] Step S22: Considering the possible impact of data dimension differences on the forecast results, all the data are normalized:
[0021]
[0022] In the formula, is the normalized value of x, x is the actual value of the original data, x min and x max are the minimum and maximum values of the original data sequence.
[0023] Step S23: defining the normalized propulsion system type, rotation speed, power, hull material, size, underwater vehicle speed, hydrodynamic parameters, vibration frequency and amplitude of hull equipment as input variables of the neural network model, and defining the normalized underwater vehicle noise data as output variables of the neural network model;
[0024] Step S24: Randomly divide the preprocessed data into three different data sets, namely, a training set (60%), a test set (20%), and a validation set (20%). The use of three different data sets (training, testing, and validation) limits the overfitting of the data when selecting hyperparameters of complex models, ensuring a fair evaluation.
[0025] Furthermore, in another preferred embodiment, the above step S3 specifically includes:
[0026] Step S31: Each unit structure of the LSTM neural network model consists of an input gate (i), a forget gate (f), an output gate (o) and a memory unit (c), and the formula is as follows:
[0027] i t =σ(w i ·[h t-1 ,x t ]+b i )
[0028] f t =σ(w f ·[h t-1 ,x t ]+b f )
[0029]
[0030] o t =σ(wo ·[h t-1 ,x t ]+b o )
[0031]
[0032] In the formula, i t 、f t , o t Represents the results of the input gate, forget gate, and output gate at the current time t; c t Represents the state of the memory cell at the current time t, which is affected by the input gate i t and forget gate f t The impact of t represents the final output of LSTM; w i and b i Represent the weight matrix and bias term of the input gate respectively. Similarly, w f and b f Represent the weight matrix and bias term of the forget gate respectively, w o and b o Represent the weight matrix and bias term of the output gate respectively, w c and b c Respectively represent the weight matrix and bias term of the unit state; [h t-1 ,x t ] The matrix is composed of the output h at the previous time t-1 t-1 and the input x at the current time t t Combination The symbol represents the Hadamard product between vector matrices; σ(·) and tanh(·) represent the sigmoid activation function and the hyperbolic tangent function, respectively;
[0033]
[0034] First, the cell state c at the current time t t The cell state c at the previous moment t-1 is controlled by the forget gate t-1 What information should be discarded and retained? Secondly, due to the cell state c t Some information is discarded in the forget gate, so new information is learned through the input gate; finally, after a series of calculations, the unit state c t The LSTM uses the output gate and the cell state c t and a tanh function to determine the final output value h t ;
[0035] Step S32: Initialize the weight thresholds of each LSTM neural network model, including the weight matrix w and bias term b of the input gate, forget gate, memory unit and output gate.
[0036] Furthermore, in another preferred embodiment, the above step S4 specifically includes:
[0037] Since the data set has been normalized, the predicted noise result data should be denormalized, that is, the predicted data should be restored to the scale range of the original data.
[0038] Furthermore, in another preferred embodiment, the above step S5 specifically includes:
[0039] Step S51: The performance evaluation indicators selected for evaluating the model are root mean square error (RMSE) and mean absolute error (MAE), and the formulas are as follows:
[0040]
[0041] In the formula, n is the total number of samples in the model test set. is the noise prediction value, y i is the actual value of noise;
[0042] Step S52: There are four groups of parameters that need to be updated in the model, namely, the weights and biases of the input gate, forget gate, memory unit, and output gate. Assume that the expected output of the model at time t is E t+1 , the actual output is h t , then the loss function of the LSTM neural network model at time t is:
[0043]
[0044] The loss function of the entire model is The model parameters are updated using the stochastic gradient descent method to input the weight matrix w of the gate i For example, the updated weights are:
[0045]
[0046] In the formula, η is the learning rate, which ranges from 0 to 1. The update process of other parameters is similar.
[0047] Step S53: Use the test set data to verify the model prediction results until the evaluation indicators meet the accuracy requirements. The evaluation criteria are: root mean square error (RMSE) ≤ 3%, mean absolute error (MAE) ≤ 3%; if the accuracy evaluation criteria are not met, repeat steps S51 and S52 to adjust and update the model parameters until the error meets the requirements or the maximum number of iterations is reached.
[0048] Furthermore, in another preferred embodiment, the above step S6 specifically includes:
[0049] Input the input variables of the validation set to obtain the underwater vehicle noise prediction results based on the long short-term memory neural network model, and compare them with the output variable data of the validation set to obtain the corresponding goodness of fit to verify the accuracy of the model:
[0050]
[0051] In the formula, is the predicted value, y is the true value, is the mean.
[0052] Furthermore, in another preferred embodiment, the above step S7 specifically includes:
[0053] Step S71: collecting various data of the underwater vehicle, processing them and inputting them into the LSTM neural network prediction model to obtain the underwater vehicle noise prediction result under the working condition;
[0054] Step S72: Pay attention to the real-time requirements of the model to ensure that the model can accurately predict the noise in a short time.
[0055] The method for rapid prediction of underwater vehicle noise based on long short-term memory 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 noise based on long short-term memory neural network, the system comprising a storage device, the storage device comprising:
[0056] Used to collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0057] Used to preprocess underwater vehicle noise data and divide it into training set, test set and validation set;
[0058] Used to build a long short-term memory neural network model and define the initial values of each weight and threshold;
[0059] Used to input training set data into the long short-term memory neural network model for model training;
[0060] It is used to input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0061] Used to compare the model prediction results with the validation set data to verify the model accuracy;
[0062] It is used to quickly predict the noise of underwater vehicles using the optimal long short-term memory neural network model.
[0063] 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 underwater vehicle noise based on long short-term memory neural network is implemented.
[0064] 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, the method for rapid prediction of underwater vehicle noise based on a long short-term memory neural network as described above is implemented.
[0065] The technical solution of the present invention has the following beneficial effects:
[0066] 1. The present invention provides a method for rapid prediction of underwater vehicle noise based on long short-term memory neural network. Unlike all other existing methods that use basic physical equations and calculations, the present invention develops and verifies a long short-term memory neural network model by utilizing large data sets of underwater vehicle propulsion system data, underwater vehicle structure and performance parameters, mechanical vibration data and noise data. The model can predict underwater vehicle noise of different types and working conditions.
[0067] Furthermore, the model can be executed in a short time while providing accurate noise forecasts. More importantly, the model can be deployed without setting up a complex simulation environment.
[0068] The method provided by the present invention can adapt to different underwater environments and changes in aircraft parameters, thereby improving the applicability and robustness of the forecast. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] Figure 1 It is a flow chart of a method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to the present invention;
[0070] Figure 2 It is a structural diagram of the long short-term memory (LSTM) neural network model established by the present invention;
[0071] Figure 3 It is a training process diagram of the long short-term memory (LSTM) neural network model described in the present invention;
[0072] Figure 4 This is the noise prediction result diagram described in the present invention. DETAILED DESCRIPTION
[0073] 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.
[0074] Implementation method 1, see Figure 1 This embodiment describes a method for rapid prediction of underwater vehicle noise based on a long short-term memory neural network, which is used to overcome the technical problems of long time consumption and low accuracy in existing underwater vehicle prediction methods.
[0075] The rapid prediction method of underwater vehicle noise specifically includes the following steps:
[0076] Step S1: Collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0077] Step S2: preprocessing the underwater vehicle noise data and dividing it into a training set, a test set and a validation set;
[0078] Step S3: construct a long short-term memory neural network model and define the initial values of various weights and thresholds;
[0079] Step S4: input the training set data into the long short-term memory neural network model for model training;
[0080] Step S5: input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0081] Step S6: Compare the model prediction results with the validation set data to verify the accuracy of the model;
[0082] Step S7: Use the optimal long short-term memory neural network model to quickly predict the noise of the underwater vehicle.
[0083] In practical application, if Figure 1As shown, by collecting underwater vehicle noise and related influencing factor data, preprocessing the data and dividing the data set, establishing a long short-term memory (LSTM) neural network model, inputting the training set to train the model in sequence, inputting the test set to tune the parameters, and inputting the verification set to verify the model accuracy, finally obtaining the optimal long short-term memory neural network model, which is applied to the rapid prediction of underwater vehicle noise. Based on the neural network method, the present invention has the ability to quickly process a large amount of data, realizes the rapid prediction of noise, and at the same time utilizes the advantages of the long short-term memory neural network to retain long-term state information, so that the underwater vehicle noise can be accurately predicted, and the accuracy of the prediction is improved.
[0084] In summary, the method for rapid prediction of underwater vehicle noise provided by this embodiment is different from all other existing methods that use basic physical equations and calculations. This embodiment develops and verifies a long short-term memory (LSTM) neural network model by utilizing a large data set of underwater vehicle propulsion system data, underwater vehicle structure and performance parameters, mechanical vibration data, and noise data. This model is capable of predicting underwater vehicle noise of different types and working conditions.
[0085] Furthermore, the model can be executed in a short time while providing accurate noise forecasts. More importantly, the model can be deployed without setting up a complex simulation environment.
[0086] Implementation Method 2: See Figures 1 to 4 This embodiment is described. This embodiment is an example of the specific steps of the method for rapid prediction of underwater vehicle noise based on long short-term memory neural network described in the first embodiment;
[0087] like Figure 1 As shown in FIG. 1 , the rapid prediction method for underwater vehicle noise specifically includes the following steps:
[0088] Step S1: Collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0089] Preferably, step S1 specifically includes:
[0090] Step S11: the propulsion system data includes the propulsion system type, rotation speed, and power; the underwater vehicle structure and performance parameters include the hull material, size, and the speed and hydrodynamic parameters of the vehicle; the mechanical vibration data is the vibration data of the hull and equipment, including vibration amplitude and frequency;
[0091] Step S12: The data set is collected through channels including real-time sensor collection, underwater vehicle historical noise system records, and public data sets.
[0092] Step S2: preprocessing the underwater vehicle noise data and dividing it into a training set, a test set and a validation set;
[0093] Preferably, step S2 specifically includes:
[0094] Step S21: Considering that the propulsion system type and hull material data are discrete feature data, one-hot vector encoding is used for storage, where each propulsion system type and hull material type has its own Boolean field;
[0095] Step S22: Considering the possible impact of data dimension differences on the forecast results, all the data are normalized:
[0096]
[0097] In the formula, is the normalized value of x, x is the actual value of the original data, x min and x max is the minimum and maximum value of the original data sequence;
[0098] Step S23: Analyze the correlation of the factors affecting the underwater vehicle noise and calculate the Person correlation coefficient as a reference for initializing the weights and thresholds:
[0099] Step S24: defining the normalized propulsion system type, rotation speed, power, hull material, size, underwater vehicle speed, hydrodynamic parameters, vibration frequency and amplitude of hull equipment as input variables of the neural network model, and defining the normalized underwater vehicle noise data as output variables of the neural network model;
[0100] Step S25: Randomly divide the preprocessed data into three different data sets, namely, a training set (60%), a test set (20%) and a validation set (20%). The use of three different data sets (training, testing and validation) limits the overfitting of the data when selecting hyperparameters of complex models and ensures a fair evaluation. The total number of data sets must not be less than 3,000 to ensure the accuracy of the model.
[0101] Step S3: construct a long short-term memory neural network model and define the initial values of various weights and thresholds;
[0102] Preferably, step S3 specifically includes:
[0103] Step S31: Figure 2 As shown, each unit structure of the LSTM neural network model consists of an input gate (i), a forget gate (f), an output gate (o) and a memory unit (c), and the formula is as follows:
[0104] i t =σ(w i ·[h t-1 ,x t ]+b i )
[0105] f t =σ(w f ·[h t-1 ,x t ]+b f )
[0106]
[0107] o t =σ(w o ·[h t-1 ,x t ]+b o )
[0108]
[0109] In the formula, i t 、f t , o t Represents the results of the input gate, forget gate, and output gate at the current time t; c t Represents the state of the memory cell at the current time t, which is affected by the input gate i t and forget gate f t The impact of t represents the final output of LSTM; w i and b i Represent the weight matrix and bias term of the input gate respectively. Similarly, w f and b f Represent the weight matrix and bias term of the forget gate respectively, w o and b o Represent the weight matrix and bias term of the output gate respectively, w c and b c Respectively represent the weight matrix and bias term of the unit state; [h t-1 ,x t ] The matrix is composed of the output h at the previous time t-1 t-1 and the input x at the current time t t Combination The symbol represents the Hadamard product between vector matrices; σ(·) and tanh(·) represent the sigmoid activation function and the hyperbolic tangent function, respectively:
[0110]
[0111] First, the cell state c at the current time t tThe cell state c at the previous moment t-1 is controlled by the forget gate t-1 What information should be discarded and retained? Secondly, due to the cell state c t Some information is discarded in the forget gate, so new information is learned through the input gate; finally, after a series of calculations, the unit state c t The LSTM uses the output gate and the cell state c t and a tanh function to determine the final output value h t .
[0112] Step S32: Initialize the weight thresholds of each LSTM neural network model, including the weight matrix w and bias term b of the input gate, forget gate, memory unit and output gate.
[0113] Step S4: Figure 3 As shown, the training set data is input into the long short-term memory neural network model for model training;
[0114] Preferably, step S4 specifically includes:
[0115] Since the data set has been normalized, the predicted noise result data should be denormalized, that is, the predicted data should be restored to the scale range of the original data.
[0116] Step S5: input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0117] Preferably, step S5 specifically includes:
[0118] Step S51: The performance evaluation indicators selected for evaluating the model are root mean square error (RMSE) and mean absolute error (MAE), and the formulas are as follows:
[0119]
[0120] In the formula, n is the total number of samples in the model test set, is the noise prediction value, y i is the actual value of noise;
[0121] Step S52: There are four groups of parameters that need to be updated in the model, namely, the weights and biases of the input gate, forget gate, memory unit, and output gate. Assume that the expected output of the model at time t is E t+1 , the actual output is h t , then the loss function of the LSTM neural network model at time t is
[0122]
[0123] The loss function of the entire model is The model parameters are updated using the stochastic gradient descent method to input the weight matrix w of the gate i For example, the updated weights are:
[0124]
[0125] In the formula, η is the learning rate, which ranges from 0 to 1. The update process of other parameters is similar;
[0126] Step S53: Use the test set data to verify the model prediction results until the evaluation indicators meet the accuracy requirements. The evaluation criteria are: root mean square error (RMSE) ≤ 3%, mean absolute error (MAE) ≤ 3%; if the accuracy evaluation criteria are not met, repeat steps S51 and S52 to adjust and update the model parameters until the error meets the requirements or the maximum number of iterations is reached.
[0127] Step S6: Compare the model prediction results with the validation set data to verify the accuracy of the model;
[0128] Preferably, step S6 specifically includes:
[0129] Input the input variables of the validation set to obtain the underwater vehicle noise prediction results based on the long short-term memory neural network model, and compare them with the output variable data of the validation set to obtain the corresponding goodness of fit to verify the accuracy of the model:
[0130]
[0131] In the formula, is the predicted value, y is the true value, is the mean.
[0132] Step S7: Use the optimal long short-term memory neural network model to quickly predict the noise of the underwater vehicle.
[0133] Preferably, step S7 specifically includes:
[0134] Step S71: collecting various underwater vehicle data, processing them and inputting them into the LSTM neural network prediction model to obtain the underwater vehicle noise prediction result under the working condition;
[0135] Step S72: Pay attention to the real-time requirements of the model to ensure that the model can accurately predict the noise in a short time. Figure 4 shown.
[0136] In summary, the method for rapid prediction of underwater vehicle noise provided by this embodiment is different from all other existing methods that use basic physical equations and calculations. This embodiment develops and verifies a long short-term memory (LSTM) neural network model by utilizing a large data set of underwater vehicle propulsion system data, underwater vehicle structure and performance parameters, mechanical vibration data, and noise data. This model is capable of predicting underwater vehicle noise of different types and working conditions.
[0137] Furthermore, the model can be executed in a short time while providing accurate noise forecasts. More importantly, the model can be deployed without setting up a complex simulation environment.
[0138] Embodiment 3: The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network described in the above embodiment can be fully implemented by computer software. Therefore, correspondingly, this embodiment provides a rapid prediction system for underwater vehicle noise based on long short-term memory neural network, and the rapid prediction system for underwater vehicle noise includes a storage device, and the storage device includes:
[0139] Used to collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0140] Used to preprocess underwater vehicle noise data and divide it into training set, test set and validation set;
[0141] Used to build a long short-term memory neural network model and define the initial values of each weight and threshold;
[0142] Used to input training set data into the long short-term memory neural network model for model training;
[0143] It is used to input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0144] Used to compare the model prediction results with the validation set data to verify the model accuracy;
[0145] It is used to quickly predict the noise of underwater vehicles using the optimal long short-term memory neural network model.
[0146] The underwater vehicle noise rapid prediction system proposed in this embodiment collects underwater vehicle noise and related influencing factor data, pre-processes the data and divides the data set, establishes a long short-term memory (LSTM) neural network model, sequentially inputs the training set to train the model, inputs the test set to tune the parameters, and inputs the verification set to verify the model accuracy, and finally obtains the optimal long short-term memory neural network model, which is applied to the rapid prediction of underwater vehicle noise. Based on the neural network method, the present invention has the ability to quickly process a large amount of data, realizes the rapid prediction of noise, and at the same time utilizes the advantages of the long short-term memory neural network to retain long-term state information, so that the underwater vehicle noise can be accurately predicted, and the accuracy of the prediction is improved.
[0147] Furthermore, the underwater vehicle noise rapid prediction system proposed in this embodiment can be executed in a short time while providing accurate noise prediction. More importantly, the system can be deployed without setting up a complex simulation environment.
[0148] Embodiment 4: 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 underwater vehicle noise based on a long short-term memory neural network proposed in the above embodiment is implemented.
[0149] Embodiment 5. 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 underwater vehicle noise based on a long short-term memory neural network proposed in the above embodiment is implemented.
[0150] The rapid prediction method for underwater vehicle noise is:
[0151] Step S1: Collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data;
[0152] Step S2: preprocessing the underwater vehicle noise data and dividing it into a training set, a test set and a validation set;
[0153] Step S3: construct a long short-term memory neural network model and define the initial values of various weights and thresholds;
[0154] Step S4: input the training set data into the long short-term memory neural network model for model training;
[0155] Step S5: input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model;
[0156] Step S6: Compare the model prediction results with the validation set data to verify the accuracy of the model;
[0157] Step S7: Use the optimal long short-term memory neural network model to quickly predict the noise of the underwater vehicle.
[0158] 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 underwater vehicle noise based on long short-term memory neural network in the above method embodiment.
[0159] 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 noise based on long short-term memory neural network, characterized in that: The method is: Step S1: Collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data; Step S2: preprocessing the underwater vehicle noise data and dividing it into a training set, a test set and a validation set; Step S3: construct a long short-term memory neural network model and define the initial values of various weights and thresholds; Step S4: input the training set data into the long short-term memory neural network model for model training; Step S5: input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model; Step S6: Compare the model prediction results with the validation set data to verify the accuracy of the model; Step S7: Use the optimal long short-term memory neural network model to quickly predict the noise of the underwater vehicle.
2. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: The propulsion system data includes the propulsion system type, rotation speed, and power; the underwater vehicle structure and performance parameters include the hull material, size, and the speed and hydrodynamic parameters of the vehicle; the mechanical vibration data is the vibration data of the hull and equipment, including vibration amplitude and frequency; Step S12: The data set is obtained through sensors, underwater vehicle historical noise system records and public data sets.
3. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 2 is characterized in that: Step S2 is specifically as follows: Step S21: The propulsion system type and hull material data are stored using one-hot vector encoding, where each propulsion system type and hull material type has its own Boolean field; Step S22: normalize all data; Step S23: defining the normalized propulsion system type, rotation speed, power, hull material, size, underwater vehicle speed, hydrodynamic parameters, vibration frequency and amplitude of hull equipment as input variables of the neural network model, and defining the normalized underwater vehicle noise data as output variables of the neural network model; Step S24: The proportions of the training set, test set, and validation set are 60%, 20%, and 20%, and the total number of data sets is not less than 3,000.
4. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: Each unit structure of the long short-term memory neural network model consists of an input gate, a forget gate, an output gate and a memory unit, specifically: i t =σ(w i ·[h t-1 ,x t ]+b i ) f t =σ(w f ·[h t-1 ,x t ]+b f ) the t =σ(w o ·[h t-1 ,x t ]+b o ) In the formula, i t 、f t , o t Represents the results of the input gate, forget gate, and output gate at the current time t; c t Represents the state of the memory cell at the current time t, which is affected by the input gate i t and forget gate f t The impact of t represents the final output of LSTM; w i and b i Represent the weight matrix and bias term of the input gate respectively. Similarly, w f and b f Represent the weight matrix and bias term of the forget gate respectively, w o and b o Represent the weight matrix and bias term of the output gate respectively, w c and b c Respectively represent the weight matrix and bias term of the unit state; [h t-1 ,x t ] The matrix is composed of the output h at the previous time t-1 t-1 and the input x at the current time t t Combination The symbol represents the Hadamard product between vector matrices; σ(·) and tanh(·) represent the sigmoid activation function and the hyperbolic tangent function, respectively; Step S32: Initialize the weight thresholds of the long short-term memory neural network model, including the weight matrix and bias items of the input gate, forget gate, memory unit and output gate.
5. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 1 is characterized in that: Step S4 is specifically as follows: The predicted noise result data is denormalized.
6. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 1 is characterized in that: Step S5 is specifically as follows: Step S51: The performance evaluation indicators selected for evaluating the model are root mean square error and mean absolute error; Step S52: The parameters that need to be updated in the model are the weight items and bias items of the input gate, forget gate, memory unit and output gate; Step S53: Use the test set data to verify the model prediction results until the evaluation indicators meet the accuracy requirements. The evaluation criteria are: root mean square error ≤3%, mean absolute error ≤3%; if the accuracy evaluation criteria are not met, repeat steps S51 and S52 to adjust and update the model parameters until the error meets the requirements or the maximum number of iterations is reached.
7. The method for rapid prediction of underwater vehicle noise based on long short-term memory neural network according to claim 1 is characterized in that: Step S6 is specifically as follows: The input variables of the validation set are input to obtain the underwater vehicle noise prediction results based on the long short-term memory neural network model, and compared with the output variable data of the validation set to obtain the corresponding goodness of fit to verify the accuracy of the model.
8. A rapid prediction system for underwater vehicle noise based on long short-term memory neural network, characterized in that: The system includes a storage device, the storage device including: Used to collect underwater vehicle noise data and noise influencing factor data, including propulsion system data, underwater vehicle structure and performance parameters, and mechanical vibration data; Used to preprocess underwater vehicle noise data and divide it into training set, test set and validation set; Used to build a long short-term memory neural network model and define the initial values of each weight and threshold; Used to input training set data into the long short-term memory neural network model for model training; It is used to input the test set data into the long short-term memory neural network model for testing, output various evaluation indicators, judge whether the model is qualified, adjust the parameters, and obtain the optimal model; Used to compare the model prediction results with the validation set data to verify the model accuracy; It is used to quickly predict the noise of underwater vehicles using the optimal long short-term memory neural network 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 rapid prediction of underwater vehicle noise based on long short-term memory neural network 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 underwater vehicle noise based on a long short-term memory neural network as described in any one of claims 1 to 7 is implemented.