Vehicle axial flow cooling fan noise prediction method based on PSO-BP neural network

By using PSO-BP neural network in the prediction of noise of automotive axial flow cooling fan, and using particle swarm optimization algorithm to optimize neural network parameters, the problem of difficulty in setting the number of hidden layer neurons and high resource consumption of traditional methods is solved, and high efficiency is achieved , accurate noise prediction.

CN119021891AActive Publication Date: 2024-11-26CHONGQING UNIV OF TECH

Patent Information

Application Number
CN202411132996.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-26
Estimated Expiration
2044-08-19

AI Technical Summary

Technical Problem

The existing BP neural network is difficult to accurately set the number of hidden layer neurons in the prediction of automotive axial cooling fan noise, resulting in a large calculation burden, low prediction accuracy and easy to fall into local optimality. Traditional experiments and numerical simulation methods are costly, Computing resources are expensive and it is difficult to achieve large-scale application.

Method used

Using a PSO-BP neural network method, the neural network parameters are optimized through particle swarm optimization algorithm, and a three-layer network structure is constructed, combining forward propagation and backpropagation training to optimize weights and thresholds to improve prediction accuracy and efficiency.

Benefits of technology

It significantly improves the accuracy and efficiency of noise prediction of automotive axle flow cooling fans, saves computing resources and manpower, is suitable for industrial applications, and provides a more accurate noise prediction method.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119021891A_ABST
    Figure CN119021891A_ABST
Patent Text Reader

Abstract

The invention discloses a vehicle axial flow cooling fan noise prediction method based on a PSO-BP neural network, and relates to the technical field of vehicle noise prediction. The method at least comprises the following steps: 1, building a vehicle axial flow cooling test fan and a simulation model, carrying out reasonability verification on the simulation model, namely confirming that the simulation model can accurately simulate the actual fan behavior, and carrying out simulation analysis through the simulation model after the reasonability verification to obtain a simulation result; and further obtaining corresponding fan noise levels under different fan parameter configurations. According to the method, the particle swarm optimization and the BP neural network are fused, the advantages of the particle swarm optimization and the BP neural network are combined to improve the prediction efficiency and precision, the advantages of the BP algorithm in the aspect of precise optimization are reserved, the optimization process is accelerated by means of the global optimization capacity of the particle swarm optimization, and local optimal traps are avoided; therefore, the accuracy and efficiency of life prediction are obviously improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of vehicle noise prediction, in particular to a vehicle axial flow cooling fan noise prediction method based on a PSO-BP neural network. Background Art

[0002] In recent years, as consumers' requirements for car ride comfort and vibration and noise control continue to improve, and the in-car configurations have become more diverse, such as the popularization of air conditioning systems and exhaust purification devices, the temperature in the engine compartment has increased significantly, which in turn has increased the workload of the cooling fan, making the noise problem more prominent. Therefore, more stringent and specific requirements have been put forward for the design and performance of the engine cooling fan, aiming to effectively reduce noise and ensure efficient heat dissipation.

[0003] Fans are mainly divided into two types: centrifugal and axial. The characteristic of centrifugal fans is that the fluid enters the fan along the axial direction and is then discharged from the side under the action of the centrifugal force generated by the rotation of the blades. In contrast, axial fans guide the fluid to flow continuously along its axial direction. Because they occupy less space and have a lower noise level, they are more common in vehicle applications. In the noise composition generated by automotive axial cooling fans, aerodynamic noise occupies a dominant position, and its order of magnitude is significantly higher than mechanical noise and electromagnetic noise, becoming the main source of noise for automotive axial cooling fans.

[0004] The existing BP neural network is well-known for its excellent nonlinear mapping ability, but it is often limited by the difficulty in accurately setting the number of hidden layer neurons. For example, too many neurons will increase the computational burden and slow down convergence; too few neurons will affect the prediction accuracy and easily fall into the local optimum and slow convergence.

[0005] Aeroacoustics is closely linked to fluid mechanics. In order to deeply understand the aerodynamic noise field, it is first necessary to accurately capture the flow field characteristics of the fluid. Traditionally, the study of fan flow field and acoustic field characteristics relies on two methods: experimental measurement and numerical simulation. However, traditional experimental methods are limited by high costs, lengthy cycles, complex operations, and the non-universality of models, making it difficult to achieve large-scale application in industry. In contrast, numerical simulation methods significantly reduce computational costs and can provide more detailed flow field information than experiments. However, in the face of models with complex geometries and high Reynolds number flow characteristics, even with numerical simulation, it is necessary to deal with huge grid volumes and complex computational challenges, consuming a lot of computing resources.

[0006] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention

[0007] The purpose of the present invention is to provide a method for predicting the noise of an automotive axial cooling fan based on a PSO-BP neural network. The noise of an automotive axial cooling fan is taken as the research object. After selecting appropriate neural network feature input and output, the obtained feature data is substituted into the neural network for training. The trained neural network shows good prediction accuracy and generalization on a test set. This method greatly reduces the time for calculating the noise of an axial cooling fan, so as to solve the technical problems raised in the background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a method for predicting noise of an axial flow cooling fan for a vehicle based on a PSO-BP neural network, comprising at least the following steps:

[0009] Step 1: Build an automotive axial cooling test fan and simulation model, verify the rationality of the simulation model, that is, confirm that it can accurately simulate the actual fan behavior, and perform simulation analysis through the simulation model after rationality verification to obtain the fan noise level corresponding to different fan parameter configurations;

[0010] Step 2: Collect parameter data and noise data of the automotive axial cooling test fan or simulation model, wherein the parameter data include rotation speed, blade tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio, and divide the parameter data into two subsets: training group data and test group data. The training group data is used for model training and learning, and the test group data is used to evaluate the performance and accuracy of the trained model.

[0011] Step 3: Based on the given neural network input and output sample sets, a topological structure of the BP neural network model is constructed, and the neural network is trained;

[0012] Step 4: Determine the particle swarm optimization algorithm. The dimension D of the position and velocity vector of each particle in the particle swarm optimization algorithm is determined by the number of parameters of the neural network. Specifically, D is equal to the number of connection weights from the input layer to the hidden layer, plus the number of connection weights from the hidden layer to the output layer, plus the threshold number of all neurons in the hidden layer, and finally the threshold number of all neurons in the output layer. These parameters together constitute the position and velocity vector of the particle in the search space, which is used to find the optimal neural network parameter configuration in the iterative process;

[0013] Step 5: Initialize the particle swarm optimization algorithm;

[0014] Step 6: Calculate the fitness of each particle based on the input and output samples;

[0015] Step 7: After calculating the fitness values ​​of all particles, by comparing the sizes of these fitness values, the individual extreme value point of each particle can be determined. At the same time, by further comparing the individual extreme value points of all particles, the global optimal extreme value point can be determined;

[0016] Step 8: According to the particle position and velocity update formula defined in the particle swarm optimization algorithm, the position and velocity of each particle are iteratively updated. When the updated value exceeds the specified range, appropriate adjustments or constraint processing are required to ensure the effectiveness and rationality of the algorithm.

[0017] Step 9: Calculate the error of the algorithm;

[0018] Step 10: Evaluate whether the current error has reached the preset accuracy requirement, or whether the number of iterations has exceeded the set maximum value; if the error meets the preset accuracy standard, the algorithm is considered to have converged, and the weight value and threshold of each dimension in the global optimal solution Pg obtained in the last iteration are the optimal parameter configurations for the problem to be solved; if the number of iterations has not reached the maximum value and the error still does not meet the accuracy requirement, the algorithm will return to step 6 and continue the iterative process to find a better solution; if the number of iterations has reached the maximum but still has not reached the preset accuracy, the algorithm will terminate execution and give an approximate solution to the problem based on the current global optimal solution;

[0019] Step 11: Optimize BP neural network parameters using particle swarm optimization algorithm: that is, use particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network to improve the prediction ability of the BP neural network for sample data and achieve the prediction application of automotive axial cooling fan noise.

[0020] Furthermore, the construction of a topological structure of a BP neural network model includes at least the following steps:

[0021] First, determine the initial structure of the BP neural network. When balancing model performance, training efficiency, and generalization ability, the initial selection of a three-layer network structure is a reasonable starting point.

[0022] The three-layer network includes an input layer, a hidden layer and an output layer;

[0023] The neurons between the input layer, hidden layer and output layer are connected to each other through weights and thresholds, which are encoded in the form of real vectors, each of which represents an individual particle in the population in the particle swarm optimization algorithm. In this way, the optimization algorithm is used to find the optimal combination of weights and thresholds to optimize the performance of the neural network;

[0024] The input quantity of the BP neural network input layer is the speed, tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio of the automotive axial cooling fan, and the output quantity of the output layer is the noise of the automotive axial cooling fan. Therefore, the number of nodes m of the input layer is 7; the number of nodes n of the output layer is 1; the number of nodes of the hidden layer is determined according to the empirical formula for calculating the number of hidden layer elements: (a is a constant from 1 to 10) to determine the range, select the appropriate number of iterations, learning rate, etc., and determine the optimal number of hidden layer neurons of the model through repeated comparisons by trial and error;

[0025] The initial connection weights between the input layer neurons and the hidden layer neurons, and between the hidden layer neurons and the output layer neurons are W i j , O j k , the threshold bias of hidden layer neurons and output layer neurons are b j , Where i = 1, 2, 3 ... m, j = 1, 2, 3 ... q, k = 1, 2, 3 ... n, the activation function is f (x) = 1 / (1 + e -x );

[0026] In order to facilitate calculation and comparison, the neuron data of the input layer and the output layer need to be normalized. The formula commonly used for normalization is:

[0027]

[0028] Where: x * is the normalized value, x i is the training data, x min 、x max They are the minimum and maximum values ​​in the data respectively. This processing method can scale the data to the range of 0 to 1, which is convenient for subsequent calculation and analysis.

[0029] Furthermore, the training of the neural network includes at least the following steps:

[0030] The first is the forward propagation stage, in which the input data is passed forward layer by layer through the neural network, and after calculation and activation of nodes in each layer, the output value of the network is finally obtained;

[0031] This is followed by the back propagation phase, in which the network adjusts the connection weights between layers and the bias of each neuron from back to front according to the difference between the output value and the target value, aiming to reduce this error;

[0032] This process repeats itself over and over again, and each time a combination of forward propagation and backpropagation is completed, it is considered that a training iteration has been completed;

[0033] Through multiple iterations of training, until the network error is reduced to below the preset target threshold, the network weights and bias terms reach a stable state, indicating that the model training is completed and an optimized neural network model is successfully constructed.

[0034] Furthermore, the step 5 at least includes the following steps:

[0035] First, the size of the particle swarm, i.e., the total number of particles, is set. Then, each particle is randomly assigned an initial position and initial velocity, which represent a set of potential neural network parameter configurations.

[0036] At the same time, the inertia factor w of the algorithm is determined, which is used to adjust the influence of the current velocity in the particle velocity update;

[0037] Set the learning factors c1 and c2, which control the tendency of the particle to move to its own historical optimal position and the global optimal position respectively;

[0038] Define the maximum number of iterations as the termination condition for the algorithm;

[0039] In addition, the individual extremum of each particle is initialized to the fitness value corresponding to its initial position, and the global optimal value is assumed to be the best individual extremum in the initial particle group;

[0040] As the algorithm iterates, these values ​​are updated based on the fitness of the particles.

[0041] Furthermore, step 6 at least includes the following steps:

[0042] First, for each particle in the particle swarm, perform the following operations in sequence;

[0043] For each sample in the sample set, the forward calculation method of the BP neural network is used to obtain the output value of the neural network represented by the particle for the sample, and the error of the sample is calculated based on it;

[0044] This process will traverse all samples and accumulate the errors of all samples for the current particle;

[0045] Then, the mean square error of these errors is calculated to obtain the fitness value of the particle;

[0046] After completing the fitness calculation of a particle, return to the previous step and continue to perform the same operation on the next particle in the particle group until the fitness of all particles is calculated.

[0047] Furthermore, the optimization of BP neural network parameters using particle swarm algorithm at least includes the following steps:

[0048] (3) Initialize the BP neural network structure:

[0049] The initialization of the BP neural network structure includes setting the number of neurons in the input layer, hidden layer, output layer of the network, learning rates α, β and the input and output of training samples.

[0050] (2) Initialize the particle swarm:

[0051] When applying the particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network, it is necessary to set the column parameters in detail to control the operation of the algorithm;

[0052] The parameters include the particle size N, the position vector and velocity vector of each particle, the individual extreme value and global optimal value of each particle, the iteration error accuracy ε, the constant coefficients c1 and c2, the maximum inertia weight w max , minimum inertia weight w min , maximum speed v max and the maximum number of iterations;

[0053] (4) Iteratively update position and speed;

[0054] In each iteration, the new velocity of each particle is first calculated using the following velocity update formula;

[0055] Then, check whether the updated speed exceeds the preset maximum speed v max ;

[0056] If the updated velocity is greater than v max , then the velocity of the particle is adjusted to v max to ensure that it does not exceed the allowed range; if the updated speed is less than or equal to v max , then keep the speed unchanged;

[0057] Finally, the updated velocity and position update formula are used to calculate the new position of each particle;

[0058] The specific formula is as follows:

[0059] V ij (t+1)=w·v ij (t)+c1r1(t)[p ij (t)-x ij (t)]+c2r2(t)[P gj (t)-x ij (t)]

[0060] X ij (t+1)=xij (t)+v ij (t+1)

[0061] Where: i = 1, 2...N, N is the number of particles;

[0062] j=1,2……D, D is the dimension of the particle;

[0063] v ij (t), x ij (t) are the velocity and position of particle i in dimension j at time t;

[0064] v ij (t+1), x ij (t+1) are the velocity and position of particle i in dimension j at time (t+1);

[0065] p ij is the optimal position of particle i so far;

[0066] p gj is the optimal position of all particles so far;

[0067] c1, c2 are learning factors, usually set to 2; r1, r2 are random numbers in [0,1];

[0068] (4) Use the following fitness function to calculate the fitness value of each particle:

[0069]

[0070] Where: n t is the number of training samples; O ir , T ir are the actual output and expected output of the network under the network weights and thresholds determined by the position of the i-th particle of the training sample r;

[0071] (5) Calculate the global minimum fitness value f of the particle swarm g =min{f1, f2, ..., f N};

[0072] If the current number of iterations reaches the preset maximum number of iterations, or the global minimum value f g If it is less than the given accuracy threshold ∈, the algorithm stops iterating;

[0073] Otherwise, the algorithm will continue, first updating the individual extremum P of each particle i and the global extreme value P g , and then update the particle speed and position based on these extreme values ​​to continue searching for a better solution in the solution space;

[0074] (6) When the algorithm reaches the stopping condition, the global extreme value P g The position represents a set of optimal network parameters, which include the weights and thresholds of the network. The optimal network parameters determined by the global extreme value Pg are output. The output parameters include the weights and thresholds of all layers. The optimal network parameters and output parameters will be used as the configuration of the trained neural network.

[0075] (7) Verify the accuracy of the model to ensure that its predictive ability in practical applications meets the expected standards and then output the predicted value.

[0076] Furthermore, the accuracy verification of the model to ensure that its prediction ability in practical applications meets the expected standards and outputting the prediction value includes at least the following steps:

[0077] After optimization by the particle swarm algorithm, the optimal weights and thresholds are reconfigured to the BP neural network;

[0078] Subsequently, the neural network is tested using a set of independent test samples, and the performance of the model is evaluated by calculating the error between the output of the neural network and the corresponding true results;

[0079] If the calculated error meets the preset accuracy requirement or error threshold, it indicates that the model is successfully established;

[0080] If the error requirement is not met, the training process needs to be repeated;

[0081] Finally, when the model meets the design requirements, the BP neural network is regarded as the optimal model after particle swarm optimization, and its output is the predicted value when used for subsequent noise data prediction.

[0082] Compared with the prior art, the present invention has the following beneficial effects:

[0083] 1. The present invention integrates the particle swarm algorithm with the BP neural network to combine the advantages of both to improve prediction efficiency and accuracy. The present invention not only retains the advantages of the BP algorithm in precise optimization, but also uses the global optimization capability of the particle swarm algorithm to accelerate the optimization process and avoid local optimal traps, thereby significantly improving the accuracy and efficiency of life prediction.

[0084] 2. Compared with traditional experimental measurement methods and numerical simulation methods, the model of the present invention can save a lot of manpower and material resources, as well as huge computing resources, and can be applied to different models and can be widely used in industry.

[0085] 3. The present invention further optimizes the model in terms of sample quantity, variable factors and structural algorithm, and can provide a more accurate noise prediction method for automotive axial cooling fans. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0087] Figure 1 It is the flow chart of PSO-BP algorithm of the present invention;

[0088] Figure 2 This is the structural diagram of the BP neural network of the present invention. DETAILED DESCRIPTION

[0089] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0090] See also Figure 1 - Figure 2 A method for predicting noise of an axial cooling fan for a vehicle based on a PSO-BP neural network comprises at least the following steps:

[0091] Step 1: Build an automotive axial cooling test fan and simulation model, verify the rationality of the simulation model, that is, confirm that it can accurately simulate the actual fan behavior, and perform simulation analysis through the simulation model after rationality verification to obtain the fan noise level corresponding to different fan parameter configurations;

[0092] Step 2: Collect parameter data and noise data of the automotive axial cooling test fan or simulation model, wherein the parameter data include rotation speed, blade tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio, and divide the parameter data into two subsets: training group data and test group data. The training group data is used for model training and learning, and the test group data is used to evaluate the performance and accuracy of the trained model.

[0093] Step 3: Based on the given neural network input and output sample sets, a topological structure of the BP neural network model is constructed, and the neural network is trained;

[0094] Step 4: Determine the particle swarm optimization algorithm. The dimension D of the position and velocity vector of each particle in the particle swarm optimization algorithm is determined by the number of parameters of the neural network. Specifically, D is equal to the number of connection weights from the input layer to the hidden layer, plus the number of connection weights from the hidden layer to the output layer, plus the threshold number of all neurons in the hidden layer, and finally the threshold number of all neurons in the output layer. These parameters together constitute the position and velocity vector of the particle in the search space, which is used to find the optimal neural network parameter configuration in the iterative process;

[0095] Step 5: Initialize the particle swarm optimization algorithm;

[0096] Step 6: Calculate the fitness of each particle based on the input and output samples;

[0097] Step 7: After calculating the fitness values ​​of all particles, by comparing the sizes of these fitness values, the individual extreme value point of each particle can be determined. At the same time, by further comparing the individual extreme value points of all particles, the global optimal extreme value point can be determined;

[0098] Step 8: According to the particle position and velocity update formula defined in the particle swarm optimization algorithm, the position and velocity of each particle are iteratively updated. When the updated value exceeds the specified range, appropriate adjustments or constraint processing are required to ensure the effectiveness and rationality of the algorithm.

[0099] Step 9: Calculate the error of the algorithm;

[0100] Step 10: Evaluate whether the current error has reached the preset accuracy requirement, or whether the number of iterations has exceeded the set maximum value; if the error meets the preset accuracy standard, the algorithm is considered to have converged, and the weight value and threshold of each dimension in the global optimal solution Pg obtained in the last iteration are the optimal parameter configurations for the problem to be solved; if the number of iterations has not reached the maximum value and the error still does not meet the accuracy requirement, the algorithm will return to step 6 and continue the iterative process to find a better solution; if the number of iterations has reached the maximum but still has not reached the preset accuracy, the algorithm will terminate execution and give an approximate solution to the problem based on the current global optimal solution;

[0101] Step 11: Optimize BP neural network parameters using particle swarm optimization algorithm: that is, use particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network to improve the prediction ability of the BP neural network for sample data and achieve the prediction application of automotive axial cooling fan noise.

[0102] Furthermore, the topological structure of the BP neural network model is constructed, which includes at least the following steps:

[0103] First, determine the initial structure of the BP neural network. When designing a BP neural network, a three-layer network structure with a single hidden layer is usually preferred because such a design can maintain a certain degree of computational accuracy while maintaining the relative simplicity of the network structure. Although increasing the number of hidden layers can help further reduce computational errors and improve model accuracy, it will also significantly increase network complexity, thereby extending training time and increasing the risk of model "overfitting". Therefore, when balancing model performance with training efficiency and generalization ability, initially choosing a three-layer network structure is a reasonable starting point;

[0104] The three-layer network includes an input layer, a hidden layer and an output layer;

[0105] The neurons between the input layer, hidden layer and output layer are connected to each other through weights and thresholds, which are encoded in the form of real vectors, each of which represents an individual particle in the population in the particle swarm optimization algorithm. In this way, the optimization algorithm is used to find the optimal combination of weights and thresholds to optimize the performance of the neural network;

[0106] The input quantity of the BP neural network input layer is the speed, tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio of the automotive axial cooling fan, and the output quantity of the output layer is the noise of the automotive axial cooling fan. Therefore, the number of nodes m of the input layer is 7; the number of nodes n of the output layer is 1; the number of nodes of the hidden layer is determined according to the empirical formula for calculating the number of hidden layer elements: (a is a constant from 1 to 10) to determine the range, select the appropriate number of iterations, learning rate, etc., and determine the optimal number of hidden layer neurons of the model through repeated comparisons by trial and error;

[0107] The initial connection weights between the input layer neurons and the hidden layer neurons, and between the hidden layer neurons and the output layer neurons are W ij , O jk , the threshold bias of hidden layer neurons and output layer neurons are b j , Where i = 1, 2, 3 ... m, j = 1, 2, 3 ... q, k = 1, 2, 3 ... n, and the activation function is f (x) = 1 / (1 + e -x );

[0108] In order to facilitate calculation and comparison, the neuron data of the input layer and the output layer need to be normalized. The formula commonly used for normalization is:

[0109]

[0110] Where: x *is the normalized value, x i is the training data, x min 、x max They are the minimum and maximum values ​​in the data respectively. This processing method can scale the data to the range of 0 to 1, which is convenient for subsequent calculation and analysis.

[0111] Furthermore, the training of the neural network includes at least the following steps:

[0112] The first is the forward propagation stage, in which the input data is passed forward layer by layer through the neural network, and after calculation and activation of nodes in each layer, the output value of the network is finally obtained;

[0113] This is followed by the back propagation phase, in which the network adjusts the connection weights between layers and the bias of each neuron from back to front according to the difference between the output value and the target value (i.e., the error), aiming to reduce this error;

[0114] This process repeats itself over and over again, and each time a combination of forward propagation and backpropagation is completed, it is considered that a training iteration has been completed;

[0115] Through multiple iterations of training, until the network error is reduced to below the preset target threshold, the network weights and bias terms reach a stable state, indicating that the model training is completed and an optimized neural network model is successfully constructed.

[0116] Furthermore, the step 5 at least includes the following steps:

[0117] First, the size of the particle swarm, i.e., the total number of particles, is set. Then, each particle is randomly assigned an initial position and initial velocity, which represent a set of potential neural network parameter configurations.

[0118] At the same time, the inertia factor w of the algorithm is determined, which is used to adjust the influence of the current velocity in the particle velocity update;

[0119] Set the learning factors c1 and c2, which control the tendency of the particle to move to its own historical optimal position and the global optimal position respectively;

[0120] Define the maximum number of iterations as the termination condition for the algorithm;

[0121] In addition, the individual extreme value of each particle is initialized to the fitness value corresponding to its initial position (i.e., the neural network performance evaluation index), and the global optimal value is assumed to be the best individual extreme value in the initial particle group;

[0122] As the algorithm iterates, these values ​​are updated based on the fitness of the particles.

[0123] Furthermore, step 6 at least includes the following steps:

[0124] First, for each particle in the particle swarm, perform the following operations in sequence;

[0125] For each sample in the sample set, the forward calculation method of the BP neural network is used to obtain the output value of the neural network represented by the particle for the sample, and the error of the sample is calculated based on it;

[0126] This process will traverse all samples and accumulate the errors of all samples for the current particle;

[0127] Then, the mean square error of these errors is calculated to obtain the fitness value of the particle;

[0128] After completing the fitness calculation of a particle, return to the previous step and continue to perform the same operation on the next particle in the particle group until the fitness of all particles is calculated.

[0129] Furthermore, the optimization of BP neural network parameters using particle swarm algorithm at least includes the following steps:

[0130] (5) Initialize the BP neural network structure:

[0131] The initialization of the BP neural network structure includes setting the number of neurons in the input layer, hidden layer, output layer of the network, learning rates α, β and the input and output of training samples.

[0132] (2) Initialize the particle swarm:

[0133] When applying the particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network, it is necessary to set the column parameters in detail to control the operation of the algorithm;

[0134] The parameters include the particle size N, the position vector and velocity vector of each particle, the individual extreme value and global optimal value of each particle, the iteration error accuracy ε, the constant coefficients c1 and c2, the maximum inertia weight w max , minimum inertia weight w min , maximum speed v max and the maximum number of iterations;

[0135] (6) Iteratively update position and speed;

[0136] In each iteration, the new velocity of each particle is first calculated using the following velocity update formula;

[0137] The formula generally includes parameters such as current speed, difference between individual optimal position and current position, difference between global optimal position and current position, inertia weight, individual learning factor and social learning factor;

[0138] Then, check whether the updated speed exceeds the preset maximum speed v max ;

[0139] If the updated velocity is greater than v max , then the velocity of the particle is adjusted to v max to ensure that it does not exceed the allowed range; if the updated speed is less than or equal to v max , then keep the speed unchanged;

[0140] Finally, the updated velocity and position update formula are used to calculate the new position of each particle;

[0141] The specific formula is as follows:

[0142] V ij (t+1)=w·v ij (t)+c1r1(t)[p ij (t)-x ij (t)]+c2r2(t)[P gj (t)-x ij (t)]

[0143] X ij (t+1)=x ij (t)+v ij (t+1)

[0144] Where: i = 1, 2...N, N is the number of particles;

[0145] j=1,2……D, D is the dimension of the particle;

[0146] v ij (t), x ij (t) are the velocity and position of particle i in dimension j at time t;

[0147] v ij (t+1), x ij (t+1) are the velocity and position of particle i in dimension j at time (t+1);

[0148] p ij is the optimal position (individual extreme value) of particle i so far;

[0149] p gj is the optimal position (group extremum) of all particles so far;

[0150] c1, c2 are learning factors, usually set to 2; r1, r2 are random numbers in [0,1];

[0151] (4) Use the following fitness function to calculate the fitness value of each particle:

[0152]

[0153] Where: n t is the number of training samples; O ir , T ir are the actual output and expected output of the network under the network weights and thresholds determined by the position of the i-th particle of the training sample r;

[0154] (5) Calculate the global minimum fitness value f of the particle swarm g =min{f1, f2, ..., f N};

[0155] If the current number of iterations reaches the preset maximum number of iterations, or the global minimum value f g If it is less than the given accuracy threshold ∈ (which means that the training error of the network has reached the required accuracy requirement), the algorithm stops iterating;

[0156] Otherwise, the algorithm will continue, first updating the individual extremum P of each particle i (i.e., the best position each particle has ever experienced) and the global extreme value P g (i.e. the best position experienced by the entire particle swarm), and then update the particle speed and position based on these extreme values ​​to continue searching for a better solution in the solution space;

[0157] (6) When the algorithm reaches the stopping condition, the global extreme value P g The position represents a set of optimal network parameters, which include the weights and thresholds of the network. The optimal network parameters determined by the global extreme value Pg are output. The output parameters include the weights and thresholds of all layers. The optimal network parameters and output parameters will be used as the configuration of the trained neural network.

[0158] (7) Verify the accuracy of the model to ensure that its predictive ability in practical applications meets the expected standards and then output the predicted value.

[0159] Furthermore, the accuracy verification of the model to ensure that its prediction ability in practical applications meets the expected standards and outputting the prediction value includes at least the following steps:

[0160] After optimization by the particle swarm algorithm, the optimal weights and thresholds are reconfigured to the BP neural network;

[0161] Subsequently, the neural network is tested using a set of independent test samples, and the performance of the model is evaluated by calculating the error between the output of the neural network and the corresponding true results;

[0162] If the calculated error meets the preset accuracy requirement or error threshold, it indicates that the model is successfully established;

[0163] If the error requirement is not met, the training process needs to be repeated, including the use of a particle swarm algorithm to optimize the weights and thresholds of the neural network until a satisfactory prediction accuracy is achieved;

[0164] Finally, when the model meets the design requirements, the BP neural network is regarded as the optimal model after particle swarm optimization, and its output is the predicted value when used for subsequent noise data prediction.

[0165] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network, characterized in that: At least the following steps are included: Step 1: Build an automotive axial cooling test fan and simulation model, verify the rationality of the simulation model, that is, confirm that it can accurately simulate the actual fan behavior, and perform simulation analysis through the simulation model after rationality verification to obtain the fan noise level corresponding to different fan parameter configurations; Step 2: Collect parameter data and noise data of the automotive axial cooling test fan or simulation model, wherein the parameter data include rotation speed, blade tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio, and divide the parameter data into two subsets: training group data and test group data. The training group data is used for model training and learning, and the test group data is used to evaluate the performance and accuracy of the trained model. Step 3: Based on the given neural network input and output sample sets, a topological structure of the BP neural network model is constructed, and the neural network is trained; Step 4: Determine the particle swarm optimization algorithm. The dimension D of the position and velocity vector of each particle in the particle swarm optimization algorithm is determined by the number of parameters of the neural network. Specifically, D is equal to the number of connection weights from the input layer to the hidden layer, plus the number of connection weights from the hidden layer to the output layer, plus the threshold number of all neurons in the hidden layer, and finally the threshold number of all neurons in the output layer. These parameters together constitute the position and velocity vector of the particle in the search space, which is used to find the optimal neural network parameter configuration in the iterative process; Step 5: Initialize the particle swarm optimization algorithm; Step 6: Calculate the fitness of each particle based on the input and output samples; Step 7: After calculating the fitness values ​​of all particles, by comparing the sizes of these fitness values, the individual extreme value point of each particle can be determined. At the same time, by further comparing the individual extreme value points of all particles, the global optimal extreme value point can be determined; Step 8: According to the particle position and velocity update formula defined in the particle swarm optimization algorithm, the position and velocity of each particle are iteratively updated. When the updated value exceeds the specified range, appropriate adjustments or constraint processing are required to ensure the effectiveness and rationality of the algorithm. Step 9: Calculate the error of the algorithm; Step 10: Evaluate whether the current error has reached the preset accuracy requirement, or whether the number of iterations has exceeded the set maximum value; if the error meets the preset accuracy standard, the algorithm is considered to have converged, and the weight value and threshold of each dimension in the global optimal solution Pg obtained in the last iteration are the optimal parameter configurations for the problem to be solved; if the number of iterations has not reached the maximum value and the error still does not meet the accuracy requirement, the algorithm will return to step 6 and continue the iterative process to find a better solution; if the number of iterations has reached the maximum but still has not reached the preset accuracy, the algorithm will terminate execution and give an approximate solution to the problem based on the current global optimal solution; Step 11: Optimize BP neural network parameters using particle swarm optimization algorithm: that is, use particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network to improve the prediction ability of the BP neural network for sample data and achieve the prediction application of automotive axial cooling fan noise.

2. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 1, characterized in that: The topological structure of the BP neural network model is constructed, which at least comprises the following steps: First, determine the initial structure of the BP neural network. When balancing model performance, training efficiency, and generalization ability, the initial selection of a three-layer network structure is a reasonable starting point. The three-layer network includes an input layer, a hidden layer and an output layer; The neurons between the input layer, hidden layer and output layer are connected to each other through weights and thresholds, which are encoded in the form of real vectors, each of which represents an individual particle in the population in the particle swarm optimization algorithm. In this way, the optimization algorithm is used to find the optimal combination of weights and thresholds to optimize the performance of the neural network; The input quantity of the BP neural network input layer is the speed, tip clearance, blade installation angle, blade width, curvature radius, number of blades and hub ratio of the automotive axial cooling fan, and the output quantity of the output layer is the noise of the automotive axial cooling fan. Therefore, the number of nodes m of the input layer is 7; the number of nodes n of the output layer is 1; the number of nodes of the hidden layer is determined according to the empirical formula for calculating the number of hidden layer elements: (a is a constant from 1 to 10) to determine the range, select the appropriate number of iterations, learning rate, etc., and determine the optimal number of hidden layer neurons of the model through repeated comparisons by trial and error; The initial connection weights between the input layer neurons and the hidden layer neurons, and between the hidden layer neurons and the output layer neurons are W ij , O jk , the threshold bias of hidden layer neurons and output layer neurons are b j , Where i = 1, 2, 3...m, j = 1, 2, 3...q, k = 1, 2, 3...n, and the activation function is f (x) = 1 / (1+e -x ); In order to facilitate calculation and comparison, the neuron data of the input layer and the output layer need to be normalized. The formula commonly used for normalization is: Where: x * is the normalized value, x i is the training data, x min 、x max They are the minimum and maximum values ​​in the data respectively. This processing method can scale the data to the range of 0 to 1, which is convenient for subsequent calculation and analysis.

3. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 1, characterized in that: The training of the neural network at least comprises the following steps: The first is the forward propagation stage, in which the input data is passed forward layer by layer through the neural network, and after calculation and activation of nodes in each layer, the output value of the network is finally obtained; This is followed by the back propagation phase, in which the network adjusts the connection weights between layers and the bias of each neuron from back to front according to the difference between the output value and the target value, aiming to reduce this error; This process repeats itself over and over again, and each time a combination of forward propagation and backpropagation is completed, it is considered that a training iteration has been completed; Through multiple iterations of training, until the network error is reduced to below the preset target threshold, the network weights and bias terms reach a stable state, indicating that the model training is completed and an optimized neural network model is successfully constructed.

4. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 1, characterized in that: The step 5 at least comprises the following steps: First, the size of the particle swarm, i.e., the total number of particles, is set. Then, each particle is randomly assigned an initial position and initial velocity, which represent a set of potential neural network parameter configurations. At the same time, the inertia factor w of the algorithm is determined, which is used to adjust the influence of the current velocity in the particle velocity update; Set the learning factors c1 and c2, which control the tendency of the particle to move to its own historical optimal position and the global optimal position respectively; Define the maximum number of iterations as the termination condition for the algorithm; In addition, the individual extremum of each particle is initialized to the fitness value corresponding to its initial position, and the global optimal value is assumed to be the best individual extremum in the initial particle group; As the algorithm iterates, these values ​​are updated based on the fitness of the particles.

5. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 1, characterized in that: The step 6 at least comprises the following steps: First, for each particle in the particle swarm, perform the following operations in sequence; For each sample in the sample set, the forward calculation method of the BP neural network is used to obtain the output value of the neural network represented by the particle for the sample, and the error of the sample is calculated based on it; This process will traverse all samples and accumulate the errors of all samples for the current particle; Then, the mean square error of these errors is calculated to obtain the fitness value of the particle; After completing the fitness calculation of a particle, return to the previous step and continue to perform the same operation on the next particle in the particle group until the fitness of all particles is calculated.

6. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 1, characterized in that: The optimization of BP neural network parameters by using particle swarm algorithm at least comprises the following steps: (1) Initialize the BP neural network structure: The initialization of the BP neural network structure includes setting the number of neurons in the input layer, hidden layer, output layer of the network, learning rates α, β and the input and output of training samples. (2) Initialize the particle swarm: When applying the particle swarm optimization algorithm to optimize the initial weights and thresholds of the BP neural network, it is necessary to set the column parameters in detail to control the operation of the algorithm; The parameters include the particle size N, the position vector and velocity vector of each particle, the individual extreme value and global optimal value of each particle, the iteration error accuracy ε, the constant coefficients c1 and c2, the maximum inertia weight w max , minimum inertia weight w min , maximum speed v max and the maximum number of iterations; (2) Iteratively update position and velocity; In each iteration, the new velocity of each particle is first calculated using the following velocity update formula; Then, check whether the updated speed exceeds the preset maximum speed v max ; If the updated velocity is greater than v max , then the velocity of the particle is adjusted to v max to ensure that it does not exceed the allowed range; if the updated speed is less than or equal to v max , then keep the speed unchanged; Finally, the updated velocity and position update formula are used to calculate the new position of each particle; The specific formula is as follows: V ij (t+1)=w·v ij (t)+c1 r1(t)[p ij (t)-x ij (t)]+c2 r2(t)[P gj (t)-x ij (t)] X ij (t+1)=x ij (t)+v ij (t+1) Where: i = 1, 2...N, N is the number of particles; j=1,2……D, D is the dimension of the particle; v ij (t), x ij (t) are the velocity and position of particle i in dimension j at time t; v ij (t+1), x ij (t+1) are the velocity and position of particle i in dimension j at time (t+1); p ij is the optimal position of particle i so far; p gj is the optimal position of all particles so far; c1, c2 are learning factors, usually set to 2; r1, r2 are random numbers in [0,1]; (4) Use the following fitness function to calculate the fitness value of each particle: Where: n t is the number of training samples; O ir 、T ir are the actual output and expected output of the network under the network weights and thresholds determined by the position of the i-th particle of the training sample r; (5) Calculate the global minimum fitness value f of the particle swarm g =min{f1, f2, ..., f N }; If the current number of iterations reaches the preset maximum number of iterations, or the global minimum value f g If it is less than the given accuracy threshold ∈, the algorithm stops iterating; Otherwise, the algorithm will continue, first updating the individual extremum P of each particle i and the global extreme value P g , and then update the particle speed and position based on these extreme values ​​to continue searching for a better solution in the solution space; (6) When the algorithm reaches the stopping condition, the global extreme value P g The position represents a set of optimal network parameters, which include the weights and thresholds of the network. The optimal network parameters determined by the global extreme value Pg are output. The output parameters include the weights and thresholds of all layers. The optimal network parameters and output parameters will be used as the configuration of the trained neural network. (7) Verify the accuracy of the model to ensure that its predictive ability in practical applications meets the expected standards and then output the predicted value.

7. The method for predicting noise of an automotive axial cooling fan based on a PSO-BP neural network according to claim 6, characterized in that: Verifying the accuracy of the model to ensure that its prediction ability in practical applications meets the expected standards and outputting the prediction value at least includes the following steps: After optimization by the particle swarm algorithm, the optimal weights and thresholds are reconfigured to the BP neural network; Subsequently, the neural network is tested using a set of independent test samples, and the performance of the model is evaluated by calculating the error between the output of the neural network and the corresponding true results; If the calculated error meets the preset accuracy requirement or error threshold, it indicates that the model is successfully established; If the error requirement is not met, the training process needs to be repeated; Finally, when the model meets the design requirements, the BP neural network is regarded as the optimal model after particle swarm optimization, and its output is the predicted value when used for subsequent noise data prediction.

Citation Information

Patent Citations

  • Equipment lifetime prediction method and system based on PSO-BP neural network

    CN113743011A

  • Natural gas moisture pipeline liquid holdup prediction model method based on PSO-BP neural network

    CN114818487A

  • Cooling tower noise optimization analysis method based on noise rule and test

    CN114861560A

  • Train traction fan noise small sample prediction method based on regression-neural network

    CN115544859A

  • Fan anomaly detection method, device and equipment and storage medium

    CN117345678A

Cited By

  • Ship block manufacturing time prediction method under small sample data

    CN119443464A

  • Fan area flow field prediction method and electronic equipment

    CN120893358A

  • Fan area flow field prediction method and electronic device

    CN120893358B

  • Fan tapered roller bearing retainer structure optimization method

    CN121256993A

  • A method for optimizing a fan conical roller bearing cage structure

    CN121256993B