Propeller noise rapid forecasting method and system based on PSO-Elman neural network

Abstract

The invention discloses a propeller noise rapid forecasting method and system based on a PSO-Elman neural network, and belongs to the field of propeller noise forecasting, and the method comprises the steps: collecting ship propeller noise data and propeller noise influence factor data, carrying out the preprocessing, combing the data correlation, constructing a sample set, and carrying out the prediction of the propeller noise; an Elman neural network model is established and optimized by adopting a particle swarm optimization algorithm, a sample set is input into the model for training and verification, and an optimal PSO-Elman neural network model is obtained and is applied to forecast the propeller noise of a certain ship. According to the method, the Elman neural network is combined with the particle swarm optimization algorithm, a large amount of data can be efficiently processed, rapid calculation of noise forecasting is achieved, meanwhile, propeller noise forecasting requirements under different environments and parameter conditions can be met, the accuracy of noise forecasting is improved, self-adaptive optimization of a noise forecasting model is achieved, and the noise forecasting accuracy is improved. And the accuracy and the stability of the forecasting result are continuously improved.

Description

Technical Field

[0001] The invention belongs to the field of marine propeller noise prediction, and in particular relates to a propeller noise rapid prediction method based on a PSO-Elman neural network. Background Art

[0002] The unstable hydrodynamic load generated by the rotation of the propeller induces a non-uniform flow field to excite the propeller and cause noise. There are currently three main means of predicting propeller noise, namely model tests, theoretical methods and numerical simulations. The model test uses a model propeller and simulates the underwater environment to analyze its acoustics, but it cannot meet the similarity criteria of the model propeller and the real propeller. In addition, based on the current experimental hardware equipment and signal processing methods, the background noise has a great impact on the test results. The theoretical method is only applicable to simple structures, and it is difficult to accurately analyze propeller noise in complex environments. With the rapid development of computer technology, numerical simulation methods have become an effective way to predict propeller noise, but there are also significant defects such as high hardware performance requirements and long calculation and analysis cycles.

[0003] With the development of artificial intelligence technology, intelligent algorithms have shown powerful capabilities in various fields. In the field of rapid prediction of propeller noise, intelligent algorithms also have broad application prospects. Neural network algorithms have certain advantages in dealing with nonlinear systems, and have the characteristics of self-learning ability, high speed, and adaptability. Neural networks can learn and understand the complex laws of propeller noise and model them as mathematical expressions. This modeling method is more flexible and efficient than traditional numerical algorithms, and can quickly generate prediction results. Therefore, it is of great engineering significance to realize the rapid prediction of propeller noise by neural networks. Summary of the invention

[0004] As a propulsion device for ships, propellers generate forward power through the interaction with water. Since the propellers are constantly cutting through water, the turbulence and cavitation generated will generate relatively large noise. The purpose of the present invention is to provide a method for rapid prediction of propeller noise based on a PSO-Elman neural network, which updates model parameters through a particle swarm algorithm, optimizes the Elman neural network model, and realizes rapid prediction of propeller noise, providing strong support for the design and construction of propellers and noise control during ship navigation.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a method for rapid prediction of propeller noise based on a PSO-Elman neural network, and the method for rapid prediction of propeller noise comprises the following steps:

[0007] Step S1: collecting ship propeller noise data and propeller noise influencing factor data, and preprocessing the propeller noise data;

[0008] Step S2: by analyzing the propeller noise characteristics and the influence of various parameters on the propeller noise results, characteristic condition data is determined, the relevant condition data is used as input data, the propeller noise is used as output data, a sample set is constructed, and it is divided into a training set, a validation set and a test set;

[0009] Step S3: Establishing an Elman neural network model, and optimizing the Elman neural network model using a particle swarm algorithm to obtain a PSO-Elman neural network model;

[0010] Step S4: input the training set samples into the constructed PSO-Elman neural network model for training. After the training is completed, the propeller noise prediction model is verified by the verification set. After the verification is qualified, the optimal PSO-Elman neural network model is obtained;

[0011] Step S5: Input the test set samples to predict the propeller noise of the ship.

[0012] Furthermore, in another preferred embodiment, the above step S1 specifically includes:

[0013] Step S11: The propeller noise influencing factor data includes propeller speed, shape and number of propeller blades, propeller rotor pitch, hull structure, ship mass and vibration; the propeller blade shape can be described by the blade bending angle, pitch, blade width and blade thickness;

[0014] Step S12: Clean the collected raw data, remove noise and outliers, smooth the outliers in the propeller noise data, and use the moving mean to fill the missing data. Specifically: output = smoothts (input, 'g', wsize, stdev), where 'g' represents the Gaussian window method, wsize represents the window width, and stdev represents the standard deviation; use the moving median with a window length of 10 to fill the missing data, specifically: F = fillmissing (A, 'movmedian', 10), where 'movmedian' represents the use of moving median filling.

[0015] Furthermore, in another preferred embodiment, the above step S2 specifically includes:

[0016] Step S21: obtaining propeller noise changes caused by changes in each propeller noise influencing factor to determine characteristic conditions, and performing standardization processing on the characteristic conditions to eliminate scale differences between various characteristics;

[0017] Step S22: performing correlation analysis on the standardized data, and calculating the Pearson correlation coefficient between the characteristic condition and the propeller noise.

[0018] Furthermore, in another preferred embodiment, the Pearson correlation coefficient between the above characteristic condition and the propeller noise is specifically:

[0019]

[0020] Where N represents the number of samples, x i ,y i Both represent sample data of characteristic conditions and noise, They all represent the sample data mean of the characteristic conditions and the noise. |Corr|<0.3 indicates a low linear correlation, |Corr|≤0.6 indicates a significant linear correlation, and |Corr|>0.6 indicates a high linear correlation. This is used as the basis for initializing the weights and thresholds of each layer of the neural network model to speed up the convergence of the model.

[0021] Furthermore, in another preferred embodiment, the above step S3 specifically includes:

[0022] Step S31: setting the Elman neural network model to include an input layer, a hidden layer, a receiving layer and an output layer;

[0023] Step S32: The weights and thresholds of the Elman neural network are optimized by using a particle swarm algorithm to form a PSO-Elman neural network model; in the particle swarm algorithm, each particle is an independent solution without the concept of volume and mass in the search space, and has its own position and speed. According to the position and speed of the particles in the search space, the population particles are randomly initialized, and each particle is continuously aggregated to its own historical best position using an adaptation function; the weights and thresholds to be optimized in the Elman neural network model form an a-dimensional vector, and the vector is initialized to form M particles, whose initialization speed and position are:

[0024] v j = {v j1 ,v j2 ,……,v ja}, x j ={x j1 ,x j2 ,……,x ja}, j = 1, 2, ..., M

[0025] The speed update and position update formulas of global particles are as follows:

[0026] v(d)=q·v(d-1)+c1r1·(pbest(d)-x(d))+c2r2·(gbest(d)-x(d))

[0027] x(d)=x(d-1)+v(d-1)

[0028] Where, v(d), x(d) represent the velocity and position of the particle when the number of iterations is d, q represents the inertia weight, which is used to maintain the velocity influence of the particle, pbest(d) is the current optimal particle position, gbest(d) is the historical optimal particle position, c1, c2 are the individual learning factor and the social learning factor, respectively, and r1 and r2 are random numbers in (0,1);

[0029] Step S33: using the minimum mean square error cost function E(X) of N training data as the fitness function of the Elman neural network model;

[0030]

[0031] Among them, y i represents the predicted value of the i-th sample, represents the true value of the i-th sample; based on this, the fitness values ​​of the M particles in step S32 can be obtained. If the current fitness value of a particle is better than the previously recorded optimal value of the particle, pbest is updated; if the current fitness value of a particle is better than the previously recorded global optimal solution, gbest is updated, that is:

[0032] gbest=min[E(pbest1),E(pbest2),……,E(pbest M )]

[0033] Furthermore, there is a preferred embodiment, the above-mentioned Elman neural network model is specifically:

[0034] y(k)=g(w 2 h(k)+b y )

[0035] h(k)=f(w 1 u(k-1)+w 3 x c (k)+b h )

[0036] x c (k) = h(k-1)

[0037] Among them, y(k), h(k), x c(k), u(k) represents the output value of the output layer, the output value of the hidden layer, the output value of the receiving layer and the output value of the input layer at time k, g(·) represents the activation function of the output layer, f(·) represents the activation function of the hidden layer, and w 1 , w 2 , w 3 They represent the weight from the input layer to the hidden layer, the weight from the hidden layer to the output layer, and the weight from the receiving layer to the hidden layer, respectively. h , b y They represent the threshold of the hidden layer and the threshold of the output layer respectively.

[0038] Furthermore, in another preferred embodiment, the above step S4 specifically includes:

[0039] Step S41: input the training set into the PSO-Elman neural network model for training;

[0040] Step S42: When the maximum number of iterations is reached and the global optimal value satisfies the minimum fitness value, the global optimal weight and threshold are obtained, and the global optimal weight and threshold are substituted into the constructed Elman neural network model to obtain the optimal PSO-Elman neural network model;

[0041] Step S43: Input the verification set into the optimal PSO-Elman neural network model, compare the actual results with the predicted results, and verify the reliability of the model.

[0042] The method for rapid prediction of propeller noise based on PSO-Elman neural network described in the present invention can be fully implemented by computer software. Therefore, correspondingly, the present invention also provides a rapid prediction system for propeller noise based on PSO-Elman neural network, and the system includes a storage device, and the storage device includes:

[0043] Used to collect ship propeller noise data and propeller noise influencing factor data, and pre-process the propeller noise data;

[0044] It is used to determine characteristic condition data by analyzing propeller noise characteristics and the influence of various parameters on propeller noise results, take relevant condition data as input data and propeller noise as output data, construct a sample set, and divide it into training set, verification set and test set;

[0045] Used to establish an Elman neural network model, and use a particle swarm algorithm to optimize the Elman neural network model to obtain a PSO-Elman neural network model;

[0046] The method is used to input the training set samples into the constructed PSO-Elman neural network model for training, and after the training is completed, the propeller noise prediction model is verified by the verification set, and the optimal PSO-Elman neural network model is obtained after the verification is qualified;

[0047] Used to input test set samples and predict the propeller noise of the ship.

[0048] 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 propeller noise based on a PSO-Elman neural network is implemented.

[0049] 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 propeller noise based on a PSO-Elman neural network is implemented.

[0050] The technical solution of the present invention has the following beneficial effects:

[0051] 1. The invention provides a method for rapid prediction of propeller noise based on PSO-Elman neural network, which combines particle swarm optimization (PSO) algorithm and Elman neural network. Elman network has temporal memory capability and can capture the dynamic changes of propeller noise. Through PSO algorithm optimization, network parameters can better adapt to training data and provide high-precision noise prediction results.

[0052] 2. The propeller noise prediction task usually needs to be performed in real time or near real time. The propeller noise prediction method provided by the present invention is based on the PSO-Elman neural network model, which can adjust parameters and update the model according to the data collected in real time, perform training quickly, and after the training is completed, quickly predict the new input, which enables the network to continuously optimize and adapt to the changing characteristics of propeller noise, thereby improving the prediction accuracy.

[0053] 3. The propeller noise prediction method provided by the present invention has strong scalability, robustness and adaptability. Due to its recursive structure and the characteristics of the particle swarm optimization algorithm, it can be extended to more complex and changeable propeller systems. It can be applied to various types of propellers, including fixed-pitch propellers, variable-pitch propellers and reverse propellers, etc. Whether it is a multi-engine configuration of a submarine or a ship with multiple propellers, the method can adapt to and predict propeller noise of different numbers and configurations, and has strong robustness to abnormal conditions and changes in propeller noise, and can effectively deal with distortion or incompleteness of noise data.

[0054] The invention is applicable to the design and construction of propellers and the noise control during the navigation of ships. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of a method for rapid prediction of propeller noise based on PSO-Elman neural network described in the present invention;

[0056] Figure 2 It is a flow chart of the neural network model algorithm of the present invention;

[0057] Figure 3 This is a diagram of the prediction result of a certain ship propeller noise according to the present invention. DETAILED DESCRIPTION

[0058] 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.

[0059] Implementation method 1, see Figure 1 In this embodiment, the propeller, as a propulsion device of the ship, generates forward power through the force of the water. Since the propeller is always cutting the water, the turbulence and cavitation generated will generate large noise. Therefore, this embodiment provides a method for rapid prediction of propeller noise based on PSO-Elman neural network, updates the model parameters through particle swarm algorithm, optimizes the Elman neural network model, realizes rapid prediction of propeller noise, and provides strong support for the design and construction of propellers and noise control during ship navigation.

[0060] The propeller noise prediction method specifically includes the following steps:

[0061] Step S1: collecting ship propeller noise data and propeller noise influencing factor data, and preprocessing the propeller noise data;

[0062] Step S2: by analyzing the propeller noise characteristics and the influence of various parameters on the propeller noise results, characteristic condition data is determined, the relevant condition data is used as input data, the propeller noise is used as output data, a sample set is constructed, and it is divided into a training set, a validation set and a test set;

[0063] Step S3: Establishing an Elman neural network model, and optimizing the Elman neural network model using a particle swarm algorithm to obtain a PSO-Elman neural network model;

[0064] Step S4: input the training set samples into the constructed PSO-Elman neural network model for training. After the training is completed, the propeller noise prediction model is verified by the verification set. After the verification is qualified, the optimal PSO-Elman neural network model is obtained;

[0065] Step S5: Input the test set samples to predict the propeller noise of the ship.

[0066] In practical application, if Figure 1 As shown in the figure, by combining the particle swarm optimization (PSO) algorithm and the Elman neural network, the dynamic changes of propeller noise can be captured. Among them, the Elman neural network is a typical local regression network. The Elman network can be regarded as a recursive neural network with local memory units and local feedback connections. The Elman network has a multi-layer structure similar to a multi-layer forward network. On the basis of the basic structure of the BP network, a receiving layer is added to the hidden layer as a one-step delay operator to achieve the purpose of memory, so that the system has the ability to adapt to time-varying characteristics and enhance the global stability of the network. It has stronger computing power than the feedforward neural network and can also be used to solve the problem of fast optimization. In addition, the neural network algorithm continuously improves the accuracy of the prediction by continuously learning and accumulating data, while having high computing efficiency and low prediction cost. Through the optimization of the PSO algorithm, the network parameters can better adapt to the training data and provide high-precision noise prediction results.

[0067] Furthermore, the propeller noise prediction method described in this embodiment can adjust parameters and update models based on data collected in real time, perform training more quickly, and quickly predict new inputs after training is completed. This allows the network to continuously optimize and adapt to the changing characteristics of propeller noise, thereby improving prediction accuracy.

[0068] Implementation Method 2: See Figures 1 to 3 This embodiment is described. This embodiment is to illustrate the specific steps of the method for rapid prediction of propeller noise based on PSO-Elman neural network described in the first embodiment by way of example;

[0069] like Figure 1 As shown, the propeller noise rapid prediction method includes the following steps:

[0070] Step S1: collecting ship propeller noise data and propeller noise influencing factor data, and preprocessing the propeller noise data;

[0071] Preferably, step S1 specifically includes:

[0072] Step S11: The factors affecting propeller noise include propeller speed, shape and number of propeller blades, propeller rotor pitch, hull structure, ship mass and vibration, etc. The shape of the propeller blades can be described by the bending angle, pitch, blade width, blade thickness, etc. of the blades;

[0073] Step S12: Clean the collected raw data to remove noise and outliers to ensure data quality. Smooth the abnormal data in the propeller noise data using the smoothts function, specifically: output = smoothts (input, 'g', wsize, stdev), where 'g' represents the Gaussian window method, wsize represents the window width, and stdev represents the standard deviation; fill the missing data using the moving median with a window length of 10, specifically: F = fillmissing (A, 'movmedian', 10), where 'movmedian' represents the use of moving median filling.

[0074] Step S2: sort out the propeller noise characteristic analysis and the influence of each parameter on the propeller noise result, determine the characteristic conditions that are strongly correlated with the propeller noise result, take the relevant conditions as input data and the propeller noise as output data, construct a sample set, and divide it into a training set, a validation set and a test set;

[0075] Preferably, step S2 specifically includes:

[0076] Step S21: obtaining the propeller noise change caused by the change of each propeller noise influencing factor through simulation and actual measurement; determining the characteristic conditions that are strongly correlated between the preprocessed data and the propeller noise result, and standardizing the characteristic condition data to eliminate the scale differences between the various features;

[0077] Step S22: performing correlation analysis on the standardized data, and calculating the Pearson correlation coefficient between the characteristic condition and propeller noise:

[0078]

[0079] Where N is the number of samples, x i ,y i Both represent the sample data of the characteristic conditions and noise, Both represent the sample data mean of the characteristic conditions and noise. |Corr|<0.3 indicates low linear correlation, |Corr|≤0.6 indicates significant linear correlation, and |Corr|>0.6 indicates high linear correlation. This is used as the basis for initializing weights and thresholds of each layer of the neural network model to accelerate the convergence of the model.

[0080] Step S3: Establishing an Elman neural network model, and optimizing the neural network model using a particle swarm algorithm to obtain a PSO-Elman neural network model;

[0081] Preferably, step S3 specifically includes:

[0082] The Elman neural network model includes an input layer, a hidden layer, a receiving layer, and an output layer. Since a receiving layer is added as a delay operator, it has a memory function. The output of the hidden layer at the previous moment and the network input data at the current moment are used as the input of the hidden layer through the receiving layer, which is equivalent to historical state feedback, and the data at the previous moment is memorized as the input of the output at the next moment, so that the output in the current state is not only related to the current input but also to the input at the previous moment. The model is as follows:

[0083] y(k)=g(w 2 h(k)+b y )

[0084] h(k)=f(w 1 u(k-1)+w 3 x c (k)+b h )

[0085] x c (k) = h(k-1)

[0086] Among them, y(k), h(k), x c (k), u(k) represent the output value of the output layer, the output value of the hidden layer, the output value of the receiving layer, and the output value of the input layer at time k respectively; g(·) represents the activation function of the output layer, f(·) represents the activation function of the hidden layer, the activation function of the hidden layer is selected as the Sigmond function, and the activation function of the output layer is selected as the Purelin function; w 1 , w 2 , w 3 They represent the weight from the input layer to the hidden layer, the weight from the hidden layer to the output layer, and the weight from the receiving layer to the hidden layer respectively; b h , b y Represents the threshold of the hidden layer and the threshold of the output layer.

[0087] Step S32: The weights and thresholds of the Elman neural network are optimized by using a particle swarm algorithm to form a PSO-Elman neural network model. In the particle swarm algorithm (PSO) algorithm, each particle is an independent solution without the concept of volume and mass in the search space, and has its own position and speed. According to the position and speed of the particles in the search space, the population particles are randomly initialized, and each particle is continuously aggregated to its own historical best position using an adaptation function. The weights and thresholds that need to be optimized in the Elman neural network model form an a-dimensional vector, which is initialized to form M particles, and their initialization speed and position are:

[0088] v j = {v j1 ,v j2 ,……,v ja}, x j ={x j1 ,x j2 ,……,x ja}, j = 1, 2, ..., M

[0089] The speed update and position update formulas of global particles are as follows:

[0090] v(d)=q·v(d-1)+c1r1·(pbest(d)-x(d))+c2r2·(gbest(d)-x(d))

[0091] x(d)=x(d-1)+v(d-1)

[0092] Where v(d) and x(d) represent the velocity and position of the particle when the number of iterations is d, q represents the inertia weight, which is used to maintain the velocity influence of the particle, pbest(d) is the current optimal particle position, gbest(d) is the historical optimal particle position, c1 and c2 are the individual learning factor and social learning factor, respectively, and r1 and r2 are random numbers in (0,1).

[0093] Step S33: In the solution space of preset dimensions, the fitness function is used as an evaluation index to seek the optimal solution of the model. The Elman neural network model uses the minimum mean square error cost function E(X) of N training data as the fitness function:

[0094]

[0095] Among them, y i represents the predicted value of the i-th sample, represents the true value of the i-th sample; based on this, the fitness values ​​of the M particles in step S32 can be obtained. If the current fitness value of a particle is better than the previously recorded optimal value of the particle, pbest is updated; if the current fitness value of a particle is better than the previously recorded global optimal solution, gbest is updated, that is:

[0096] gbest=min[E(pbest1),E(pbest2),……,E(pbest M )]

[0097] Step S4: Input the training set samples into the constructed PSO-Elman neural network model for training. After the training is completed, the propeller noise prediction model is verified by the verification set. After the verification is qualified, the optimal PSO-Elman neural network model is obtained. The specific algorithm flow of the neural network model is as follows: Figure 2 As shown;

[0098] Preferably, step S4 specifically includes:

[0099] Step S41: inputting the training set as input data into the neural network model for training;

[0100] Step S42: When the maximum number of iterations is reached and the global optimal value satisfies the minimum fitness value, the global optimal weight and threshold are obtained, and the global optimal weight and threshold are substituted into the constructed Elman neural network model to obtain the optimal PSO optimized noise prediction model;

[0101] Step S43: input the verification set into the optimal PSO-Elman neural network model, compare the actual results with the predicted results, and verify the reliability of the model.

[0102] Step S5: Input the test set samples to predict the propeller noise of the ship.

[0103] Preferably, step S5 specifically includes:

[0104] The optimal PSO-Elman neural network model finally obtained is applied to the propeller noise prediction of a certain ship. The corresponding input data is input to obtain the noise spectrum of the propeller in the frequency band of 500-2000Hz, as shown in Figure 3 shown.

[0105] The propeller noise prediction method proposed in this embodiment combines the particle swarm optimization (PSO) algorithm and the Elman neural network. The Elman network has temporal memory capability and can capture the dynamic changes of propeller noise. Through PSO algorithm optimization, the network parameters can better adapt to the training data and provide high-precision noise prediction results.

[0106] Furthermore, the propeller noise prediction method proposed in this embodiment is based on the PSO-Elman neural network model, and can adjust parameters and update models according to the data collected in real time, perform training quickly, and quickly predict new inputs after training is completed, which enables the network to continuously optimize and adapt to the changing characteristics of propeller noise, thereby improving the prediction accuracy. At the same time, it has strong scalability, robustness and adaptability. Due to its recursive structure and the characteristics of the particle swarm optimization algorithm, it can be extended to more complex and variable propeller systems. It can be applied to various types of propellers, including fixed-pitch propellers, variable-pitch propellers, and reverse propellers, etc. Whether it is a multi-engine configuration of a submarine or a ship with multiple propellers, this method can adapt to and predict propeller noise of different numbers and configurations, and has strong robustness to abnormal conditions and changes in propeller noise, and can effectively deal with distortion or incompleteness of noise data.

[0107] Embodiment 3: The method for rapid prediction of propeller noise based on PSO-Elman neural network described in embodiment 1 can be implemented entirely by computer software. Therefore, correspondingly, this embodiment provides a rapid prediction system for propeller noise based on PSO-Elman neural network, and the rapid prediction system for propeller noise includes a storage device, and the storage device includes:

[0108] Used to collect ship propeller noise data and propeller noise influencing factor data, and pre-process the propeller noise data;

[0109] It is used to determine characteristic condition data by analyzing propeller noise characteristics and the influence of various parameters on propeller noise results, take relevant condition data as input data and propeller noise as output data, construct a sample set, and divide it into training set, verification set and test set;

[0110] Used to establish an Elman neural network model, and use a particle swarm algorithm to optimize the Elman neural network model to obtain a PSO-Elman neural network model;

[0111] The method is used to input the training set samples into the constructed PSO-Elman neural network model for training, and after the training is completed, the propeller noise prediction model is verified by the verification set, and the optimal PSO-Elman neural network model is obtained after the verification is qualified;

[0112] Used to input test set samples and predict the propeller noise of the ship.

[0113] The propeller noise prediction system proposed in this embodiment has strong scalability, robustness and adaptability. Due to its recursive structure and the characteristics of the particle swarm optimization algorithm, it can be extended to more complex and variable propeller systems. It can be applied to various types of propellers, including fixed-pitch propellers, variable-pitch propellers, and reverse propellers. Whether it is a multi-engine configuration of a submarine or a ship with multiple propellers, this method can adapt to and predict propeller noise of different numbers and configurations, and has strong robustness to abnormal conditions and changes in propeller noise, and can effectively deal with distortion or incompleteness of noise data.

[0114] 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 propeller noise based on a PSO-Elman neural network proposed in the above embodiment is implemented.

[0115] Implementation method 5: This implementation method 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 propeller noise based on a PSO-Elman neural network proposed in the above implementation method is implemented.

[0116] The propeller noise prediction method is:

[0117] Step S1: collecting ship propeller noise data and propeller noise influencing factor data, and preprocessing the propeller noise data;

[0118] Step S2: by analyzing the propeller noise characteristics and the influence of various parameters on the propeller noise results, characteristic condition data is determined, the relevant condition data is used as input data, the propeller noise is used as output data, a sample set is constructed, and it is divided into a training set, a validation set and a test set;

[0119] Step S3: Establishing an Elman neural network model, and optimizing the Elman neural network model using a particle swarm algorithm to obtain a PSO-Elman neural network model;

[0120] Step S4: input the training set samples into the constructed PSO-Elman neural network model for training. After the training is completed, the propeller noise prediction model is verified by the verification set. After the verification is qualified, the optimal PSO-Elman neural network model is obtained;

[0121] Step S5: Input the test set samples to predict the propeller noise of the ship.

[0122] 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 propeller noise rapid prediction method and steps based on the PSO-Elman neural network in the above method embodiment.

[0123] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may be subject to 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 propeller noise based on PSO-Elman neural network, characterized in that: The method is: Step S1: collecting ship propeller noise data and propeller noise influencing factor data, and preprocessing the propeller noise data; Step S2: by analyzing the propeller noise characteristics and the influence of various parameters on the propeller noise results, characteristic condition data is determined, the relevant condition data is used as input data, the propeller noise is used as output data, a sample set is constructed, and it is divided into a training set, a validation set and a test set; Step S3: Establishing an Elman neural network model, and optimizing the Elman neural network model using a particle swarm algorithm to obtain a PSO-Elman neural network model; Step S4: input the training set samples into the constructed PSO-Elman neural network model for training. After the training is completed, the propeller noise prediction model is verified by the verification set. After the verification is qualified, the optimal PSO-Elman neural network model is obtained; Step S5: Input the test set samples to predict the propeller noise of the ship.

2. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: the propeller noise influencing factor data includes propeller speed, shape and number of propeller blades, propeller rotor pitch, hull structure, ship mass and vibration; The propeller blade shape can be described by the blade’s bend angle, pitch, blade width, and blade thickness; Step S12: Clean the collected raw data to remove noise and outliers, smooth the abnormal data in the propeller noise data, and use the moving mean to fill in the missing data.

3. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 1 is characterized in that: Step S2 is specifically as follows: Step S21: obtaining propeller noise changes caused by changes in each propeller noise influencing factor, determining characteristic condition data, and performing standardization processing on the characteristic condition data to eliminate scale differences between various features; Step S22: performing correlation analysis on the standardized data, and calculating the Pearson correlation coefficient between the characteristic condition and the propeller noise.

4. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 3 is characterized in that: The Pearson correlation coefficient between the characteristic condition and propeller noise is as follows: Where N represents the number of samples, x i ,y i Both represent sample data of characteristic conditions and noise, Both represent the mean of sample data of characteristic conditions and noise.

5. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: setting the Elman neural network model to include an input layer, a hidden layer, a receiving layer and an output layer; Step S32: The weights and thresholds of the Elman neural network are optimized by using a particle swarm algorithm to form a PSO-Elman neural network model. In the particle swarm algorithm, each particle is an independent solution without the concept of volume and mass in the search space, and has its own position and speed. According to the position and speed of the particles in the search space, the population particles are randomly initialized, and each particle is continuously aggregated to its own historical best position using an adaptation function; the weights and thresholds that need to be optimized in the Elman neural network model form an a-dimensional vector, and the vector is initialized to form M particles, whose initialization speed and position are: v j ={v j1 ,v j2 ,……,v ja },x j ={x j1 ,x j2 ,……,x ja },j=1,2,……,M The speed update and position update formulas of global particles are as follows: v(d)=q·v(d-1)+c1r1·(pbest(d)-x(d))+c2r2·(gbest(d)-x(d)) x(d)=x(d-1)+v(d-1) Where, v(d), x(d) represent the velocity and position of the particle when the number of iterations is d, q represents the inertia weight, which is used to maintain the velocity influence of the particle, pbest(d) is the current optimal particle position, gbest(d) is the historical optimal particle position, c1, c2 are the individual learning factor and the social learning factor, respectively, and r1 and r2 are random numbers in (0,1); Step S33: using the minimum mean square error cost function E(X) of N training data as the fitness function of the Elman neural network model.

6. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 5 is characterized in that: The Elman neural network model is: y(k)=g(w 2 h(k)+b y ) h(k)=f(w 1 u(k-1)+w 3 x c (k)+b h ) x c (k)=h(k-1) Among them, y(k), h(k), x c (k), u(k) represents the output value of the output layer, the output value of the hidden layer, the output value of the receiving layer and the output value of the input layer at time k, g(·) represents the activation function of the output layer, f(·) represents the activation function of the hidden layer, and w 1 , w 2 , w 3 They represent the weight from the input layer to the hidden layer, the weight from the hidden layer to the output layer, and the weight from the receiving layer to the hidden layer, respectively. h , b y They represent the threshold of the hidden layer and the threshold of the output layer respectively.

7. The method for rapid prediction of propeller noise based on PSO-Elman neural network according to claim 1 is characterized in that: Step S4 is specifically as follows: Step S41: input the training set into the PSO-Elman neural network model for training; Step S42: When the maximum number of iterations is reached and the global optimal value satisfies the minimum fitness value, the global optimal weight and threshold are obtained, and the global optimal weight and threshold are substituted into the constructed Elman neural network model to obtain the optimal PSO-Elman neural network model; Step S43: Input the verification set into the optimal PSO-Elman neural network model, compare the actual results with the predicted results, and verify the reliability of the model.

8. A propeller noise rapid prediction system based on PSO-Elman neural network, characterized in that: The system includes a storage device, the storage device including: Used to collect ship propeller noise data and propeller noise influencing factor data, and pre-process the propeller noise data; It is used to determine characteristic condition data by analyzing propeller noise characteristics and the influence of various parameters on propeller noise results, take relevant condition data as input data and propeller noise as output data, construct a sample set, and divide it into training set, verification set and test set; Used to establish an Elman neural network model, and use a particle swarm algorithm to optimize the Elman neural network model to obtain a PSO-Elman neural network model; The method is used to input the training set samples into the constructed PSO-Elman neural network model for training, and after the training is completed, the propeller noise prediction model is verified by the verification set, and the optimal PSO-Elman neural network model is obtained after the verification is qualified; Used to input test set samples and predict the propeller noise of the ship.

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 a processor, the method for rapid prediction of propeller noise based on a PSO-Elman neural network according to 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. When the processor executes the computer program, a method for rapid prediction of propeller noise based on a PSO-Elman neural network as described in any one of claims 1 to 7 is implemented.

Images