A neural network model modeling method for predicting front side window flow field excitation of an automobile by frequency segments

Abstract

The application discloses a neural network model frequency-division modeling method for predicting automobile front side window flow field excitation, and steps include: 1) obtaining frequency domain flow field excitation data of the automobile front side window surface; 2) dividing the prediction frequency band according to the contribution analysis result; 3) screening out the top-ranking parameters in each frequency band as key geometric parameters of the frequency band, and performing segmented processing on the flow field excitation data according to the frequency band; 4) taking the key geometric parameters in each frequency band as the input of a neural network model, taking the excitation data of the corresponding frequency band as the output of the neural network model, and training the neural network model; and 5) combining the prediction results of the neural network models of the frequency bands to establish a neural network model for rapidly predicting the front side window flow field excitation in the whole prediction frequency band. The application divides the frequency band based on the contribution analysis result, and realizes accurate prediction of the front side window excitation by training the network model in the frequency-division manner.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerodynamics and machine learning, in particular to a neural network model frequency-division modeling method for predicting the flow field excitation of the front side window of a vehicle. BACKGROUND

[0002] In the field of research on wind noise in the vehicle cabin, computational fluid dynamics (CFD) is currently the most commonly used research method. However, the CFD method has problems such as long calculation time, large consumption of computing resources, and complex model grid processing. Often a vehicle has multiple sets of shape optimization schemes, and each set of scheme requires several days of time for CFD simulation. This seriously hinders the progress of the work in the stages of vehicle shape optimization design and in-cabin wind noise optimization and noise reduction.

[0003] The unique black box property of neural networks enables them to accurately establish the relationship between input parameters and output values of a nonlinear system, and thus realize the ability to predict flow field excitation through key geometric parameters. This seems to be able to serve as an efficient prediction method to replace CFD simulation. However, there are many geometric factors that affect the flow field state, which means that the neural network model may select too many input parameters, thereby increasing the demand for the number of training samples. In addition, since the influence of different geometric parameters on the flow field excitation varies at different frequencies, considering all input parameters will repeatedly consider parameters with low correlation at each frequency. These low-correlation parameters will interfere with the model's identification of the main parameters, increase the degree of nonlinearity of the system, and affect the model training effect. SUMMARY

[0004] The purpose of the present application is to provide a neural network model frequency-division modeling method for predicting the flow field excitation of the front side window of a vehicle, comprising the following steps:

[0005] 1) Calculate the time-domain flow field data of the front side window surface of the vehicle using computational fluid dynamics simulation software;

[0006] 2) Process the time-domain flow field data using wave number frequency spectrum decomposition technology to obtain the frequency-domain flow field excitation data of the front side window surface of the vehicle;

[0007] 3) Measure the geometric parameter data of the vehicle and form a mapping with the frequency-domain flow field excitation data of the front side window surface of the vehicle;

[0008] 4) Calculate the contribution of the geometric parameters of the vehicle to each flow field excitation using a contribution analysis method, and divide the prediction frequency range according to the contribution size to obtain multiple sub-frequency ranges;

[0009] 5) Take the geometric parameters of the vehicle whose contribution is in the top n% in each sub-frequency range as the key geometric parameters;

[0010] According to the sub-frequency band, the frequency domain flow field excitation data is segmented and processed to obtain sub-frequency band frequency domain flow field excitation data;

[0011] A sub-frequency band sample data set is constructed with the key geometric parameter data as input and the sub-frequency band frequency domain flow field excitation data as output;

[0012] 6) A neural network model is established by using machine learning technology;

[0013] 7) The neural network model is trained by using the sub-frequency band sample data set to obtain the front side window flow field excitation prediction model corresponding to each sub-frequency band;

[0014] 8) The front side window flow field excitation prediction model is constructed by combining the front side window flow field excitation prediction model corresponding to each sub-frequency band;

[0015] 9) The geometric parameter data of the vehicle to be measured is measured and input into the front side window flow field excitation frequency band prediction model to obtain the front side window flow field excitation prediction value of the vehicle.

[0016] Further, in step 1), the step of calculating the time domain flow field data of the automobile front side window surface comprises:

[0017] 1.1) A virtual numerical wind tunnel is established in the fluid dynamics simulation software, and the virtual numerical wind tunnel is a cuboid, wherein two wall surfaces are used as velocity inlet wall surfaces and pressure outlet wall surfaces;

[0018] 1.2) A closed vehicle model to be measured is constructed;

[0019] 1.3) The vehicle model to be measured is placed in the virtual numerical wind tunnel and close to the velocity inlet wall surface;

[0020] 1.4) A grid dense area is set in the front side window area of the vehicle model to be measured;

[0021] 1.5) The velocity inlet wall surface is set to have a coming flow wind speed, the pressure of the pressure outlet wall surface is set to 0 Pa, and the remaining four wall surfaces and the vehicle body wall surface are set to be no-slip wall surfaces;

[0022] 1.6) The simulation software is run to obtain the time domain flow field data of the automobile front side window surface.

[0023] Further, the length of the virtual numerical wind tunnel is 15 times the length of the vehicle body of the vehicle to be measured, the width is 13 times the width of the vehicle body, and the height is 7 times the height of the vehicle body;

[0024] The distance between the vehicle model to be measured and the velocity inlet wall surface is 3 times the length of the vehicle body, and the distance between the two sides of the vehicle body and the two side wall surfaces is 6 times the width of the vehicle body;

[0025] Further, the step of processing the time-domain flow field data of the front side window surface of the automobile in step 2) comprises:

[0026] 2.1) performing Fourier transform on the time-domain flow field data of the front side window surface of the automobile to obtain:

[0027]

[0028] wherein ω is the circular frequency, x and y are spatial variables, t is time, e is the natural constant, i is the imaginary number; p(x, y, t) is the time-domain flow field data; p * (x, y, ω) is the time-domain flow field data after Fourier transform;

[0029] 2.2) performing second Fourier transform on the pressure fluctuation in the x and y directions in the frequency domain to obtain the pressure data P * (k x ,k y , ω) in the wave number domain, i.e.:

[0030]

[0031] wherein k x is the wave number in the x direction, and k y is the wave number in the y direction;

[0032] 2.3) in the two-dimensional wave number domain (k x ,k y ), separating the acoustic energy and the hydrodynamic energy according to the difference in the propagation speed of the acoustic energy and the hydrodynamic energy;

[0033] wherein the distribution of the acoustic energy is as follows:

[0034]

[0035] wherein c is the sound speed; f is the frequency;

[0036] The distribution of the hydrodynamic energy is as follows:

[0037]

[0038] wherein u is the air flow speed.

[0039] 2.4) taking the separated acoustic energy as the sound pressure excitation of the front side window surface of the automobile, and taking the separated hydrodynamic energy as the hydrodynamic pressure excitation of the front side window surface of the automobile; taking the sound pressure excitation of the front side window surface of the automobile and the hydrodynamic pressure excitation of the front side window surface of the automobile as the frequency-domain flow field excitation data of the front side window surface of the automobile.

[0040] Further, the automobile geometry parameter data is measured by a geometry measurement means;

[0041] The automobile geometry parameters include, but are not limited to, the following parameters that can describe the geometry of A-pillar and rearview mirror: vehicle width, front spoiler angle, A-pillar width, A-pillar deflection, A-pillar side step height, A-pillar side inclination angle, A-pillar included angle, rearview mirror handle thickness, rearview mirror inner side included angle, rearview mirror outer side included angle, rearview mirror upper deflection, rearview mirror bottom deflection, rearview mirror distance to side window, rearview mirror height, rearview mirror width, etc.

[0042] Further, in step 4), the step of calculating the contribution amount of the automobile geometry parameters to each flow field excitation includes:

[0043] 4.1) normalizing the automobile geometry parameters;

[0044] 4.2) taking the normalized geometry parameters as input variables and the automobile front side window flow field excitation as output, an expression in the form of regression analysis is established, i.e.:

[0045]

[0046] wherein, X j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable, β0 is a constant; Y is the output;

[0047] 4.3) using a regression problem optimization algorithm to calculate the regression coefficient corresponding to each input variable;

[0048] 4.4) calculating the contribution amount of each input variable to the output, i.e.:

[0049]

[0050] wherein, is the contribution amount.

[0051] Further, in step 5), the step of taking the automobile geometry parameters with the contribution amount in the top n% in each sub-frequency band as the key geometry parameters includes:

[0052] In each sub-frequency band, the average value of the contribution amount of each geometry parameter is calculated, and the geometry parameter with the average value of the contribution amount in the top n% is taken as the key geometry parameter of the sub-frequency band.

[0053] Further, in step 6), the step of establishing a neural network model using machine learning technology includes:

[0054] 6.1) determining the basic structure of the neural network, including an input layer, a single or multiple hidden layers, and an output layer;

[0055] 6.2) determining the number of neurons in each layer of the neural network.

[0056] Further, in step 7), the step of training the neural network model by using the sub-frequency band sample data set comprises:

[0057] 7.1) normalizing the data of the sample data set, and dividing the normalized sample data set into a test set and a training set;

[0058] 7.2) selecting a training function of the neural network model and an activation function of each hidden layer;

[0059] 7.3) dividing each sub-frequency band corresponding sample set into a training set and a test set in units of the divided plurality of sub-frequency bands;

[0060] 7.4) training the neural network model by using the training set;

[0061] 7.5) testing the neural network model by using the test set, and if the error is less than a preset value, outputting a front side window flow field excitation prediction model, otherwise, re-constructing the sample data set;

[0062] 7.6) repeating steps 7.3)-7.5) to obtain a front side window flow field excitation prediction model corresponding to each sub-frequency band.

[0063] Further, the neural network model comprises a back propagation neural network model.

[0064] The technical effect of the present application is self-evident. The present application proposes a neural network model frequency band division modeling method for predicting the flow field excitation of the front side window of a vehicle. The method divides the prediction frequency band based on the parameter contribution analysis result, selects the key geometric parameters in units of the divided frequency band, and divides the excitation data. Through the frequency band division modeling method, the input dimension of the neural network model in each frequency band is reduced, and the prediction accuracy of the model is improved. The method has the advantages of high prediction efficiency and accurate results. It has great application value and broad application prospect in the field of vehicle aerodynamic shape optimization design and the field of vehicle wind noise prediction. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 It is a flowchart of a neural network model frequency band division modeling method for predicting the flow field excitation of the front side window of a vehicle;

[0066] Figure 2 It is a schematic diagram of the shape of a virtual numerical wind tunnel;

[0067] Figure 3 It is a schematic diagram of the position of the front side window grid encryption area;

[0068] Figure 4 It is a diagram of the sound pressure excitation and hydrodynamic pressure excitation data obtained after the time domain flow field data is decomposed by the WFS method;

[0069] Figure 5 The schematic diagram of the measurement position of the geometric parameters of the automobile;

[0070] Figure 6 The schematic diagram of the neural network structure;

[0071] Figure 7 The curve diagram of the contribution amount of each geometric parameter to the sound pressure excitation under different frequencies;

[0072] Figure 8 The curve diagram of the contribution amount of each geometric parameter to the hydrodynamic pressure excitation under different frequencies;

[0073] Figure 9 The schematic diagram of the BPNN neural network structure used in Example 2;

[0074] Figure 10 The comparison of the predicted sound pressure excitation and the expected sound pressure excitation obtained by the method in Test Set 1;

[0075] Figure 11 The comparison of the predicted sound pressure excitation and the expected sound pressure excitation obtained by the method in Test Set 2;

[0076] Figure 12 The comparison of the predicted hydrodynamic pressure excitation and the expected sound pressure excitation obtained by the method in Test Set 1;

[0077] Figure 13 The comparison of the predicted hydrodynamic pressure excitation and the expected sound pressure excitation obtained by the method in Test Set 2. DETAILED DESCRIPTION

[0078] The application will be further described in conjunction with the examples below, but should not be understood as limiting the above-mentioned subject matter of the application only to the following examples. Various substitutions and modifications can be made according to ordinary technical knowledge and conventional means in the art without departing from the above-mentioned technical idea of the application, and all of them should be included in the protection scope of the application.

[0079] Example 1:

[0080] Referring to Figures 1 to 13 A frequency segment modeling method of a neural network model for predicting the flow field excitation of the front side window of an automobile, comprising the following steps:

[0081] 1) Calculating the time-domain flow field data of the surface of the front side window of the automobile by using a computational fluid dynamics simulation software;

[0082] 2) Processing the time-domain flow field data by using a wave number frequency spectrum decomposition technique to obtain the frequency-domain flow field excitation data of the surface of the front side window of the automobile;

[0083] 3) measuring the geometric parameter data of the automobile and mapping with the frequency domain flow field excitation data of the front side window surface of the automobile;

[0084] 4) calculating the contribution amount of the geometric parameter of the automobile to each flow field excitation by using the contribution amount analysis method, and dividing the prediction frequency band according to the contribution amount to obtain a plurality of sub-frequency bands;

[0085] 5) taking the geometric parameters of the automobile with the contribution amount located in the top n% in each sub-frequency band as the key geometric parameters; n is a positive integer.

[0086] According to the sub-frequency band, the frequency domain flow field excitation data is segmented and processed to obtain sub-frequency band frequency domain flow field excitation data;

[0087] A sub-frequency band sample data set is constructed with the key geometric parameter data as input and the sub-frequency band frequency domain flow field excitation data as output;

[0088] 6) using machine learning technology to establish a neural network model;

[0089] 7) using the sub-frequency band sample data set to train the neural network model to obtain the front side window flow field excitation prediction model corresponding to each sub-frequency band;

[0090] 8) combining the front side window flow field excitation prediction model corresponding to each sub-frequency band to construct a front side window flow field excitation sub-frequency band prediction model;

[0091] 9) measuring the geometric parameter data of the vehicle to be measured and inputting into the front side window flow field excitation sub-frequency band prediction model to obtain the front side window flow field excitation prediction value of the vehicle.

[0092] In step 1), the step of calculating the time domain flow field data of the front side window surface of the automobile includes:

[0093] 1.1) establishing a virtual numerical wind tunnel in a fluid dynamics simulation software, the virtual numerical wind tunnel being a cuboid, wherein two wall surfaces are used as velocity inlet wall surfaces and pressure outlet wall surfaces;

[0094] 1.2) constructing a closed vehicle model to be measured;

[0095] 1.3) placing the vehicle model to be measured in the virtual numerical wind tunnel and close to the velocity inlet wall surface;

[0096] 1.4) setting a grid dense area in the front side window area of the vehicle model to be measured;

[0097] 1.5) setting the incoming flow velocity on the velocity inlet wall surface, setting the pressure on the pressure outlet wall surface to 0 Pa, and setting the remaining four wall surfaces and the vehicle body wall surface as no-slip wall surfaces;

[0098] 1.6) running simulation software to obtain time-domain flow field data of the front side window surface of the vehicle;

[0099] The virtual numerical wind tunnel has a length of 15 times the length of the vehicle body to be measured, a width of 13 times the width of the vehicle body, and a height of 7 times the height of the vehicle body;

[0100] The distance between the vehicle model to be measured and the speed inlet wall surface is 3 times the length of the vehicle body, and the distance between the two sides of the vehicle body and the two side walls is 6 times the width of the vehicle body;

[0101] In step 2), the step of processing the time-domain flow field data of the front side window surface of the vehicle includes:

[0102] 2.1) Fourier transform the time-domain flow field data of the front side window surface of the vehicle to obtain:

[0103]

[0104] In the formula, ω is the circular frequency, x and y are spatial quantities, t is time, e is the natural constant, i is the imaginary number; p(x, y, t) is the time-domain flow field data; p * (x, y, ω) is the Fourier-transformed time-domain flow field data;

[0105] 2.2) performing a second Fourier transform on the pressure fluctuations in the x and y directions in the frequency domain to obtain the pressure data P * (k x ,k y ,ω) in the wave number domain, that is:

[0106]

[0107] In the formula, k x is the wave number in the x direction, and k y is the wave number in the y direction;

[0108] 2.3) in the two-dimensional wave number domain (k x ,k y ), separating the acoustic energy and the hydrodynamic energy according to the difference in propagation speed of the acoustic energy and the hydrodynamic energy;

[0109] The distribution of the acoustic energy is as follows:

[0110]

[0111] In the formula, c is the speed of sound; f is the frequency;

[0112] The distribution of the hydrodynamic energy is as follows:

[0113]

[0114] In the formula, u is the air flow speed.

[0115] 2.4) the decomposed sound energy as the sound pressure excitation of the front side window surface of the car, and the decomposed water kinetic energy as the water pressure excitation of the front side window surface of the car; the sound pressure excitation of the front side window surface of the car and the water pressure excitation of the front side window surface of the car as the frequency domain flow field excitation data of the front side window surface of the car.

[0116] The car geometry parameter data is measured by a geometry measurement means.

[0117] The car geometry parameters include, but are not limited to, the following parameters that can describe the geometry of the A-pillar and the rearview mirror: the car width, the front spoiler angle, the A-pillar width, the A-pillar deflection, the A-pillar side step height, the A-pillar side inclination angle, the A-pillar included angle, the rearview mirror handle thickness, the rearview mirror inner side surface included angle, the rearview mirror outer side surface included angle, the rearview mirror upper deflection, the rearview mirror bottom deflection, the rearview mirror distance from the side window, the rearview mirror height, the rearview mirror width, and the like.

[0118] In step 4), the step of calculating the contribution amount of the car geometry parameters to each flow field excitation includes:

[0119] 4.1) normalizing the car geometry parameters;

[0120] 4.2) taking the normalized geometry parameters as the input variables and the car front side window flow field excitation as the output, and establishing a regression analysis expression:

[0121]

[0122] In the formula, X j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable, and β0 is a constant value. Y is the output.

[0123] 4.3) using a regression problem optimization algorithm to calculate the regression coefficients corresponding to each input variable;

[0124] 4.4) calculating the contribution amount of each input variable to the output, i.e.,

[0125]

[0126] In the formula, is the contribution amount.

[0127] In step 5), the step of taking the car geometry parameters with the contribution amount in the top n% in each sub-frequency band as the key geometry parameters is:

[0128] In each sub-frequency band, the average value of the contribution amount of each geometry parameter is calculated, and the geometry parameter with the average value of the contribution amount in the top n% is selected as the key geometry parameter of the sub-frequency band.

[0129] In step 6), the step of establishing a neural network model by using a machine learning technique comprises:

[0130] 6.1) determining a neural network infrastructure, including an input layer, a single or multiple hidden layers, and an output layer;

[0131] 6.2) determining the number of neurons in each layer of the neural network.

[0132] In step 7), the step of training the neural network model by using a sub-frequency band sample data set comprises:

[0133] 7.1) normalizing the data of the sample data set, and dividing the normalized sample data set into a test set and a training set;

[0134] 7.2) selecting a training function of the neural network model and an activation function of each hidden layer;

[0135] 7.3) dividing each sub-frequency band corresponding sample set into a training set and a test set in units of the divided multiple sub-frequency bands;

[0136] 7.4) training the neural network model by using the training set;

[0137] 7.5) testing the neural network model by using the test set, and if the error is less than a preset value, outputting a front side window flow field excitation prediction model, otherwise, re-constructing a sample data set;

[0138] 7.6) repeating steps 7.3) to 7.5) to obtain a front side window flow field excitation prediction model corresponding to each sub-frequency band.

[0139] The neural network model comprises a back propagation neural network model.

[0140] Embodiment 2:

[0141] A neural network model frequency band modeling method for predicting a front side window flow field excitation of a vehicle comprises the following steps:

[0142] 1) calculating time domain flow field data of a front side window surface of a vehicle by using a computational fluid dynamics simulation software;

[0143] 2) processing the time domain flow field data by using a wave number frequency spectrum decomposition technique to obtain frequency domain flow field excitation data of the front side window surface of the vehicle;

[0144] 3) measuring vehicle geometric parameter data, and forming a mapping with the frequency domain flow field excitation data of the front side window surface of the vehicle;

[0145] 4) using the contribution amount analysis method to calculate the contribution amount of the geometric parameters of the automobile to the excitation of each flow field, and dividing the prediction frequency band according to the contribution amount to obtain a plurality of sub-frequency bands;

[0146] 5) taking the geometric parameters of the automobile with the contribution amount in the top n% in each sub-frequency band as the key geometric parameters;

[0147] According to the sub-frequency band, the frequency domain flow field excitation data is segmented and processed to obtain sub-frequency band frequency domain flow field excitation data;

[0148] A sub-frequency band sample data set is constructed with the key geometric parameter data as input and the sub-frequency band frequency domain flow field excitation data as output;

[0149] 6) using machine learning technology to establish a neural network model;

[0150] 7) using the sub-frequency band sample data set to train the neural network model to obtain the front side window flow field excitation prediction model corresponding to each sub-frequency band;

[0151] 8) combining the front side window flow field excitation prediction model corresponding to each sub-frequency band to construct a front side window flow field excitation frequency band prediction model;

[0152] 9) measuring the geometric parameter data of the vehicle to be tested and inputting it into the front side window flow field excitation frequency band prediction model to obtain the front side window flow field excitation prediction value of the vehicle.

[0153] Embodiment 3:

[0154] A neural network model frequency band modeling method for predicting the flow field excitation of the front side window of an automobile, the technical content is the same as that of embodiment 2, further, in step 1), the step of calculating the time domain flow field data of the front side window surface of the automobile comprises:

[0155] 1.1) establishing a virtual numerical wind tunnel in a fluid dynamics simulation software, the virtual numerical wind tunnel is a cuboid, wherein two wall surfaces are used as velocity inlet wall surfaces and pressure outlet wall surfaces;

[0156] 1.2) constructing a closed vehicle model to be tested;

[0157] 1.3) placing the vehicle model to be tested in the virtual numerical wind tunnel and close to the velocity inlet wall surface;

[0158] 1.4) setting a grid dense area in the front side window area of the vehicle model to be tested;

[0159] 1.5) setting the incoming flow velocity on the velocity inlet wall surface, setting the pressure on the pressure outlet wall surface to 0Pa, and setting the remaining four wall surfaces and the vehicle body wall surface as no-slip wall surfaces;

[0160] 1.6) Run the simulation software to obtain the time-domain flow field data of the front side window surface of the vehicle.

[0161] Embodiment 4:

[0162] A neural network model frequency segment modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-3, further, the virtual numerical wind tunnel has a length of 15 times the body length of the vehicle to be tested, a width of 13 times the body width, and a height of 7 times the body height;

[0163] The distance between the vehicle model to be tested and the speed inlet wall surface is 3 times the body length, and the distance between the body on both sides and the two side walls is 6 times the body width;

[0164] Embodiment 5:

[0165] A neural network model frequency segment modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-4, further, in step 2), the step of processing the time-domain flow field data of the front side window surface of the vehicle includes:

[0166] 2.1) Fourier transform the time-domain flow field data of the front side window surface of the vehicle to obtain:

[0167]

[0168] In the formula, ω is the circular frequency, x and y are spatial quantities, t is time, e is the natural constant, and i is the imaginary number;

[0169] 2.2) Perform a second Fourier transform on the pressure fluctuations in the x and y directions in the frequency domain to obtain the pressure data in the wave number domain, that is:

[0170]

[0171] In the formula, k x is the wave number in the x direction, and k y is the wave number in the y direction;

[0172] 2.3) In the two-dimensional wave number domain (k x , k y ), the acoustic energy and the hydrodynamic energy are separated according to the difference in propagation speed of the acoustic energy and the hydrodynamic energy;

[0173] The distribution of the acoustic energy is as follows:

[0174]

[0175] In the formula, c is the speed of sound; and f is the frequency;

[0176] The distribution of the hydrodynamic energy is as follows:

[0177]

[0178] where u is the air flow velocity.

[0179] 2.4) the decomposed sound energy is taken as the sound pressure excitation of the front side window surface of the car, and the decomposed water kinetic energy is taken as the water dynamic pressure excitation of the front side window surface of the car; the sound pressure excitation of the front side window surface of the car and the water dynamic pressure excitation of the front side window surface of the car are taken as the frequency domain flow field excitation data of the front side window surface of the car.

[0180] Embodiment 6:

[0181] A neural network model frequency segment modeling method for predicting the flow field excitation of the front side window of a car, the technical content of which is the same as any one of embodiments 2-5, further, the car geometry parameter data is measured by a geometry measurement means;

[0182] The car geometry parameters include but are not limited to: car width, front wind breaking angle, A-pillar width, A-pillar deflection, A-pillar side step height, A-pillar side inclination angle, A-pillar included angle, rearview mirror handle thickness, rearview mirror inner side face included angle, rearview mirror outer side face included angle, rearview mirror upper deflection, rearview mirror bottom deflection, rearview mirror distance from side window, rearview mirror height, rearview mirror width, and other parameters that can describe the geometry of the A-pillar and the rearview mirror.

[0183] Embodiment 7:

[0184] A neural network model frequency segment modeling method for predicting the flow field excitation of the front side window of a car, the technical content of which is the same as any one of embodiments 2-6, further, in step 4), the step of calculating the contribution amount of the car geometry parameters to each flow field excitation includes:

[0185] 4.1) normalizing the car geometry parameters;

[0186] 4.2) taking the normalized geometry parameters as input variables and the flow field excitation of the front side window of the car as output, an expression in the form of regression analysis is established:

[0187]

[0188] where X j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable, and β0 is a constant value.

[0189] 4.3) the regression problem optimization algorithm is used to calculate the regression coefficients corresponding to each input variable;

[0190] 4.4) the contribution amount of each input variable to the output is calculated, that is:

[0191]

[0192] Embodiment 8:

[0193] A frequency segment modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-7, further, in step 5), the step of taking the vehicle geometric parameters with the contribution amount in the top n% in each sub-frequency segment as the key geometric parameters is:

[0194] In each sub-frequency segment, the average value of the contribution amount of each geometric parameter is calculated, and the geometric parameter with the average value of the contribution amount in the top n% is selected as the key geometric parameter of the sub-frequency segment.

[0195] Embodiment 9:

[0196] A frequency segment modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-8, further, in step 6),

[0197] The step of establishing a neural network model using machine learning technology includes:

[0198] 6.1) Determine the neural network infrastructure, including an input layer, a single or multiple hidden layers, and an output layer;

[0199] 6.2) Determine the number of neurons in each layer of the neural network.

[0200] Embodiment 10:

[0201] A frequency segment modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-9, further, in step 7), the step of training the neural network model using the sub-frequency segment sample data set includes:

[0202] 7.1) Normalize the data of the sample data set, and divide the normalized sample data set into a test set and a training set;

[0203] 7.2) Select the training function of the neural network model and the activation function of each hidden layer;

[0204] 7.3) Divide the sample set corresponding to each sub-frequency segment into a training set and a test set in units of multiple sub-frequency segments;

[0205] 7.4) Train the neural network model using the training set;

[0206] 7.5) Test the neural network model using the test set, if the error is less than a preset value, output the front side window flow field excitation prediction model, otherwise, re-construct the sample data set;

[0207] 7.6) Repeat steps 7.3)-7.5) to obtain the front side window flow field excitation prediction model corresponding to each sub-frequency band.

[0208] Embodiment 11:

[0209] A neural network model frequency band modeling method for predicting the flow field excitation of the front side window of a vehicle, the technical content of which is the same as any one of embodiments 2-10, further, the neural network model comprises a back propagation neural network model.

[0210] Embodiment 12:

[0211] Referring to Figures 1 to 9 A neural network model frequency band modeling method for predicting the flow field excitation of the front side window of a vehicle, comprising the following steps:

[0212] 1) Obtain the time-domain flow field data of the surface of the front side window of the vehicle by solving with computational fluid dynamics (CFD) simulation software;

[0213] 2) Process the time-domain flow field data using wavenumber frequency spectrum decomposition technology (WFS) to obtain the frequency-domain flow field excitation data of the surface of the front side window of the vehicle;

[0214] 3) Measure the geometric parameters of the vehicle using geometric measurement means;

[0215] 4) Obtain the contribution amount of the geometric parameters of the vehicle to each flow field excitation using a contribution amount analysis method, and divide the prediction frequency band according to the contribution amount analysis result;

[0216] 5) Select the top-ranked parameters in each frequency band as the key geometric parameters of the frequency band, and process the flow field excitation data by frequency band;

[0217] 6) Establish a neural network model using machine learning technology;

[0218] 7) Take the key geometric parameters in each frequency band as the input of the neural network model, and take the excitation data corresponding to the frequency band as the output of the neural network model. Train the neural network model by frequency band.

[0219] 8) Combine the prediction results of the neural network models of each frequency band to establish a neural network model for quickly predicting the flow field excitation of the front side window in the entire prediction frequency band;

[0220] 9) measuring the geometric parameter data of the vehicle to be tested, inputting the front side window flow field excitation frequency band prediction model established in step 8) to obtain the front side window flow field excitation prediction value of the vehicle to be tested.

[0221] The step of solving the time-domain flow field data of the surface of the front side window of the vehicle by using the CFD simulation software comprises:

[0222] 1.1) establishing a virtual numerical wind tunnel in the CFD simulation software, the virtual numerical wind tunnel has a length of about 15 times the length of the vehicle body of the vehicle to be tested, a width of about 13 times the width of the vehicle body, and a height of about 7 times the height of the vehicle body. The virtual numerical wind tunnel is composed of a velocity inlet wall, a pressure outlet wall, a ground boundary, a top boundary, and the remaining two opposite wall boundaries;

[0223] 1.2) performing geometric processing on the vehicle model to be tested, only retaining the outermost geometric surface and strictly ensuring that the entire vehicle interior is a closed space;

[0224] 1.3) placing the vehicle to be tested in the virtual numerical wind tunnel, the placement position is close to the end of the velocity inlet wall, the vehicle head of the vehicle to be tested is about 3 times the length of the vehicle body away from the inlet wall, and the vehicle body is about 6 times the width of the vehicle body away from the two side walls;

[0225] 1.4) setting a grid dense area to improve the accuracy of the obtained flow field data, the grid dense area is located in the front side window area of the vehicle to be tested, and the grid dense area is set to be an octagonal prism grid with a boundary layer thickness of 2 mm;

[0226] 1.5) performing simulation setting on the numerical simulation software, setting the incoming flow velocity on the velocity inlet wall, setting the pressure outlet on the pressure outlet wall, and setting the remaining walls including the vehicle body wall as no-slip walls;

[0227] 1.6) running the simulation software to obtain the time-domain flow field data of the vehicle to be tested and save it.

[0228] The step of solving the frequency-domain flow field excitation of the surface of the front side window of the vehicle by using the WFS method comprises:

[0229] 2.1) converting the time-domain flow field data (pressure fluctuation p(x, y, t)) obtained by the CFD simulation software into the frequency domain by Fourier transform:

[0230]

[0231] In the formula, ω is the circular frequency, x and y are spatial quantities, t is time, e is a natural constant, and i is an imaginary number.

[0232] 2.2) converting the pressure fluctuation in the frequency domain into pressure data in the wave number domain by performing Fourier transform in the x direction and the y direction:

[0233]

[0234] where k x is the wave number in x direction, k y is the wave number in y direction.

[0235] 2.3) According to the difference of the propagation speed of acoustic energy and hydrodynamic energy, the two kinds of energy can be separated in two-dimensional wave number domain (k x ,k y ).

[0236] The distribution of acoustic energy is:

[0237]

[0238] The distribution of hydrodynamic energy is:

[0239]

[0240] where c is the sound speed, f is the frequency, and u is the air flow speed.

[0241] 2.4) The acoustic energy obtained by the WFS method is the sound pressure excitation on the surface of the front side window of the car, and the hydrodynamic energy obtained by the WFS method is the hydrodynamic pressure excitation on the surface of the front side window of the car.

[0242] The geometric parameters of the car measured by geometric measurement methods include but are not limited to: car width, front wind breaking angle, A-pillar width, A-pillar deflection, A-pillar side step height, A-pillar side inclination angle, A-pillar included angle, rearview mirror handle thickness, rearview mirror inner side surface included angle, rearview mirror outer side surface included angle, rearview mirror upper deflection, rearview mirror bottom deflection, rearview mirror distance from side window, rearview mirror height, rearview mirror width, and other parameters that can describe the geometric shape of the A-pillar and the rearview mirror.

[0243] The contribution amount analysis method is used to obtain the contribution amount of the geometric parameters of the car to each flow field excitation, and the prediction frequency band is divided according to the contribution amount analysis result. The steps of obtaining the contribution amount of the geometric parameters of the car by the contribution amount analysis method include:

[0244] 4.1) The maximum-minimum normalization method is used to process the geometric parameters of the car to the interval [0, 1]:

[0245]

[0246] where x is the normalized value, i represents the i-th sample, i = 1, 2, …, p, m represents the m-th feature in the same sample, m = 1, 2, …, n, and a im represents the m-th feature value in the i-th sample.

[0247] 4.2) Calculate the normalized geometric parameter α j As the input variable, the automobile front side window flow field excitation Y is taken as the output value, and the expression in the form of regression analysis is established:

[0248]

[0249] In the formula, α j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable, and β0 is a constant value.

[0250] 4.3) The regression coefficients corresponding to each input variable are solved by using the ridge regression method, and the loss function of the ridge regression is:

[0251]

[0252] In the formula, α j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable, and β0 is a constant value, Y i represents the output value of the ith sample, i = 1, 2, …, n, and λ is a regularization parameter.

[0253] The loss function is minimized to obtain the regression coefficients corresponding to each input variable.

[0254] 4.4) The contribution of each input variable to the output is calculated by the following formula:

[0255]

[0256] In the formula, α j represents the jth input variable, j = 1, 2, …, m, β j represents the regression coefficient corresponding to the jth input variable.

[0257] According to the variation characteristics of the geometric parameter contribution with frequency, the prediction frequency range is divided into several small frequency ranges.

[0258] In each small frequency range, the average value of the contribution of each geometric parameter in the range is calculated, and the geometric parameters are arranged in descending order of average contribution. The top several geometric parameters in each frequency range are selected as the key geometric parameters of the frequency range.

[0259] In each small frequency range, the automobile front side window flow field excitation data (sound pressure excitation and hydrodynamic pressure excitation) is divided and processed according to the frequency range, as the excitation output data of the frequency range.

[0260] The steps of establishing a neural network model using machine learning technology include:

[0261] 6.1) Determine the neural network infrastructure, i.e. an input layer, single or multiple hidden layers, and an output layer;

[0262] 6.2) Determine the number of neurons in each layer of the neural network;

[0263] 6.3) Normalize the input and output sample data, using the maximum-minimum normalization method of step 4.1) to process the data to the interval [0, 1];

[0264] 6.4) Select the training function of the neural network, which is the trainlm function, i.e. the Levenberg-Marquardt (LM) algorithm. For the following nonlinear least squares problem:

[0265]

[0266] where H(x) is the sum of squared residuals, P i ′ is the network model prediction vector, P i is the expected output vector, n is the number of samples, e i (x) is the residual, and a is a vector composed of network weights and thresholds.

[0267] The iteration step of the LM algorithm is:

[0268] I LM =-(J T (a)J(a)+μI) -1 J T (a)e(a) (10)

[0269] where μ is the damping factor, I is the identity matrix, and J(a) is the Jacobian matrix.

[0270] 6.5) Select the activation function of each hidden layer of the neural network, which is the Sigmoid function. Its mathematical definition is:

[0271]

[0272] where f(b) is the output value of the Sigmoid function, e is the natural constant, and b is the input value of the previous layer of neurons.

[0273] 6.6) The performance of the neural network model is represented by the Mean Squared Error (MSE):

[0274]

[0275] where n is the number of samples, P i ′ is the network model prediction vector, P i is the expected output vector.

[0276] The key geometric parameters in each frequency band are taken as the input of the neural network model, and the excitation data corresponding to the frequency band are taken as the output of the neural network model. In frequency band units, the steps of training the neural network model include:

[0277] 7.1) Divide the sample data into a training sample set and a test sample set, and the training sample set is used for training the network model. The test sample set is used to verify the model prediction result, and cannot be used for network training;

[0278] 7.2) In units of several small frequency bands divided, the key geometric parameters of the frequency band are taken as the input sample, and the surface sound pressure excitation and hydrodynamic pressure excitation data of the automobile front side window of the frequency band are taken as the output sample, the network model is trained, and the network prediction model corresponding to the frequency band is obtained;

[0279] 7.3) The test set data is used to verify the model prediction effect;

[0280] The network prediction model of each frequency band is combined to obtain the prediction model of the automobile front side window flow field excitation in the full frequency band.

[0281] Embodiment 13:

[0282] Referring to Figures 1 to 13 A neural network model frequency band modeling method for predicting automobile front side window flow field excitation includes the following steps:

[0283] 1) Time-domain flow field data of the automobile front side window surface is obtained by solving using computational fluid dynamics (CFD) simulation software;

[0284] 2) The time-domain flow field data is processed using wavenumber frequency spectrum decomposition (WFS) to obtain frequency-domain flow field excitation data of the automobile front side window surface;

[0285] 3) Geometric parameters of the automobile are measured using a geometric measurement method;

[0286] 4) The contribution amount of the geometric parameters of the automobile to each flow field excitation is obtained using a contribution amount analysis method, and the prediction frequency band is divided according to the contribution amount analysis result;

[0287] 5) Screen out the top-ranking parameters in each frequency band as the key geometric parameters of the frequency band in terms of frequency bands, and process the flow field excitation data in segments according to the frequency bands;

[0288] 6) Establish a neural network model by using a machine learning technique;

[0289] 7) Take the key geometric parameters in each frequency band as the input of the neural network model, and take the excitation data of the corresponding frequency band as the output of the neural network model. Train the neural network model in terms of frequency bands.

[0290] 8) Combine the prediction results of the neural network models of the respective frequency bands to establish a neural network model for rapidly predicting the front side window flow field excitation in the entire prediction frequency band;

[0291] 9) Measure the geometric parameter data of the vehicle to be tested, input the front side window flow field excitation frequency-band prediction model established in step 8), and obtain the prediction value of the front side window flow field excitation of the vehicle to be tested.

[0292] 1) Obtain the time-domain flow field data of the surface of the front side window of the vehicle by solving with a computational fluid dynamics (CFD) simulation software;

[0293] The step of obtaining the time-domain flow field data of the surface of the front side window of the vehicle by solving with the CFD simulation software comprises:

[0294] 1.1) Establish a virtual numerical wind tunnel in the CFD simulation software, with the virtual numerical wind tunnel having a length of 60 m, a width of 24 m and a height of 12 m. The virtual numerical wind tunnel is composed of a velocity inlet wall, a pressure outlet wall, a ground boundary, a top boundary and the remaining two opposite wall boundaries;

[0295] 1.2) Geometrically process the vehicle model to be tested, and only retain the outermost geometric surface while strictly ensuring that the entire vehicle interior is a closed space;

[0296] 1.3) Place the vehicle to be tested in the virtual numerical wind tunnel, with the placement position being close to the end of the velocity inlet wall, the vehicle to be tested being about 3 times the vehicle body length away from the inlet wall, and the vehicle to be tested being equidistant from the two side walls;

[0297] 1.4) Set a grid-intensive region to improve the accuracy of the obtained flow field data, with the grid-intensive region being located in the front side window region of the vehicle to be tested, and the grid-intensive region being set as a four-prism grid with 8 layers of boundary layer thickness of 2 mm and a grid thickness growth rate of 1.5;

[0298] 1.5) Set up the simulation in the numerical simulation software, set the incoming flow velocity on the velocity inlet wall surface, and set the wind speed to 120 km / h. Set the pressure outlet on the pressure outlet wall surface, and the size is 0 Pa. The remaining wall surfaces including the vehicle body wall surface are set to no-slip wall surfaces;

[0299] 1.6) The turbulence model is the two-equation SST k-omega model for numerical simulation, and the semi-implicit method for pressure linked equations (SIMPLE) is used to solve the pressure-velocity coupling term. Separated vortex simulation is used for transient simulation. Some other settings of the solver are shown in the following table.

[0300]

[0301] 1.7) Run the simulation software, solve a total of 6 different vehicle models, and save a total of 46 groups of front side window flow field data of different shapes.

[0302] 2) The wavenumber frequency spectrum decomposition (WFS) method is used to process the time domain flow field data to obtain the frequency domain flow field excitation data on the surface of the front side window of the vehicle;

[0303] The steps for solving the frequency domain flow field excitation on the surface of the front side window of the vehicle using the WFS method include:

[0304] 2.1) The time domain flow field data (pressure fluctuation p(x, y, t)) solved by the CFD simulation software is converted to the frequency domain by Fourier transform:

[0305]

[0306] where ω is the circular frequency, x and y are spatial quantities, t is time, e is the natural constant, and i is the imaginary number.

[0307] 2.2) The pressure fluctuation in the frequency domain can be converted to the pressure data in the wavenumber domain by performing Fourier transform in the x and y directions:

[0308]

[0309] where k x is the wave number in the x direction, and k y is the wave number in the y direction.

[0310] 2.3) According to the difference in the propagation speed of acoustic energy and hydrodynamic energy, these two kinds of energy can be separated in the two-dimensional wavenumber domain (k x , k y ).

[0311] The distribution of acoustic energy is:

[0312]

[0313] The distribution of hydrodynamic energy is:

[0314]

[0315] where c is the sound speed and u is the air flow speed.

[0316] 2.4) The acoustic energy obtained by the WFS method is the sound pressure excitation of the front side window surface of the car, and the hydrodynamic energy obtained by the WFS method is the hydrodynamic pressure excitation of the front side window surface of the car, and both are saved as one-third octave results.

[0317] 3) Measure the geometric parameters of the car using geometric measurement means;

[0318] The geometric parameters of the car measured by geometric measurement means are shown in the following table, a total of 15 geometric parameters.

[0319]

[0320] 4) Use the contribution analysis method to obtain the contribution of the geometric parameters of the car to each flow field excitation, and divide the prediction frequency band according to the contribution analysis results;

[0321] 4.1) The contribution analysis results are shown in Figures 7-8

[0322] 4.2) According to the contribution analysis results, the prediction frequency band is divided into four small frequency bands (one-third octave). Respectively: 100Hz-250Hz, 315Hz-800Hz, 1000Hz-2500Hz, 3150Hz-8000Hz.

[0323] 5) Calculate the average value of the contribution of each geometric parameter in each small frequency band in units of frequency band, and arrange them in order of average contribution from large to small. The figure shows the average contribution of each geometric parameter to the sound pressure excitation and the hydrodynamic pressure excitation in the first frequency band (100Hz-250Hz). Select the top 5 geometric parameters in each frequency band as the key geometric parameters of the frequency band. The top 5 key geometric parameters in each frequency band are shown in the following table.

[0324]

[0325] The car front side window flow field excitation data (sound pressure excitation and hydrodynamic pressure excitation) is divided and processed by frequency band as the excitation output data of the frequency band.​

[0326] 6) Establish a Back Propagation Neural Network (BPNN) model using machine learning techniques.

[0327] The steps for establishing a BPNN model using machine learning techniques include:

[0328] 6.1) Determine the neural network infrastructure, i.e., one input layer, three hidden layers, and one output layer;

[0329] 6.2) Determine the number of neurons in each layer of the neural network, with 5 neurons in the input layer and 20 neurons in each hidden layer;

[0330] 6.3) Normalize the input and output sample data using the maximum-minimum normalization method to process the data to the interval [0, 1];

[0331] 6.4) Select the training function for the neural network, which is the trainlm function. Select the activation function for each hidden layer of the neural network, which is the Sigmoid function. The performance of the neural network model is represented by the Mean Squared Error (MSE).

[0332] 7) Use the key geometric parameters in each frequency band as the input to the neural network model and the excitation data corresponding to the frequency band as the output of the neural network model. Train the neural network model in frequency bands. The steps include:

[0333] 7.1) Divide the 46 sample data sets into 44 training sample sets and 2 test sample sets, with the training sample sets used for training the network model. The test sample sets are used to verify the prediction results of the model and cannot be used for network training;

[0334] 7.2) Divide the key geometric parameters in each frequency band into several small frequency bands as input samples, and the surface sound pressure excitation and hydrodynamic pressure excitation data in the frequency band as output samples to train the network model and obtain the neural network prediction model in each frequency band;

[0335] 7.3) Input the geometric parameters of the 2 test sample sets into the prediction model for each frequency band according to the corresponding key geometric parameters to obtain the excitation prediction results for each frequency band;

[0336] 8) Combine the excitation prediction results for each frequency band to obtain the excitation prediction results for the entire prediction frequency range. Compare the predicted data with the expected data of the test sample set to calculate the average sound pressure level error and the relative error, which is defined as follows:

[0337]

[0338] 9) The error of the predicted data of the front side window excitation compared with the expected data is shown in the following table.

[0339]

[0340] It can be seen that the average sound pressure level error is within 1 dB, and the relative error is controlled within 1.5%, whether it is the predicted sound pressure excitation or the hydrodynamic pressure excitation. Further verification of the accuracy of the proposed neural network model for predicting the flow field excitation of the front side window of the automobile.

Claims

1. A frequency-segment modeling method for predicting flow field excitation in a vehicle's front side window, characterized in that, Includes the following steps: 1) Calculate the time-domain flow field data on the surface of the front side window of a car using computational fluid dynamics simulation software; 2) The time-domain flow field data is processed using wavenumber frequency spectrum decomposition technology to obtain the frequency-domain flow field excitation data of the front side window surface of the car. 3) Measure the vehicle's geometric parameters and map them to the frequency domain flow field excitation data on the surface of the vehicle's front side window; 4) Calculate the contribution of vehicle geometric parameters to each flow field excitation using the contribution analysis method, and divide the predicted frequency band according to the contribution magnitude to obtain multiple sub-frequency bands. 5) The vehicle geometric parameters whose contribution is in the top n% within each sub-frequency band are taken as key geometric parameters; The frequency domain flow field excitation data is segmented according to sub-frequency bands to obtain sub-frequency band frequency domain flow field excitation data. Construct a sub-frequency band sample dataset with key geometric parameter data as input and sub-frequency band frequency domain flow field excitation data as output; 6) Utilize machine learning techniques to build neural network models; 7) Train the neural network model using the sub-frequency band sample dataset to obtain the front window flow field excitation prediction model for each sub-frequency band. 8) Combine the front window flow field excitation prediction model corresponding to each sub-frequency band to construct a front window flow field excitation frequency band prediction model; 9) Measure the geometric parameters of the vehicle under test and input them into the frequency segment prediction model of the front side window flow field excitation to obtain the predicted value of the front side window flow field excitation.

2. The method for frequency segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, Step 1), the steps for calculating the time-domain flow field data on the surface of the car's front side window include: 1.1) Establish a virtual numerical wind tunnel in the fluid dynamics simulation software. The virtual numerical wind tunnel is a cuboid, with two walls serving as the velocity inlet wall and the pressure outlet wall. 1.2) Construct a closed model of the vehicle to be tested; 1.3) Place the vehicle model under test inside the virtual numerical wind tunnel and close to the velocity inlet wall; 1.4) Set a densely meshed area in the front window area of ​​the vehicle model under test; 1.5) Set the inflow velocity on the inlet wall, set the pressure on the outlet wall to 0 Pa, and set the remaining four walls and the vehicle body wall to non-slip walls. 1.6) Run the simulation software to obtain the time-domain flow field data of the front side window surface of the car.

3. The method for frequency-segment modeling of a neural network model for predicting flow field excitation in a vehicle's front side window according to claim 2, characterized in that, The virtual numerical wind tunnel is 15 times the length of the vehicle body under test, 13 times the width of the vehicle body, and 7 times the height of the vehicle body. The distance between the vehicle model under test and the velocity inlet wall is 3 times the vehicle length, and the distance between the two sides of the vehicle and the two side walls is 6 times the vehicle width.

4. The method for frequency-segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, Step 2) involves processing the time-domain flow field data on the surface of the car's front side window, including: 2.1) Perform a Fourier transform on the time-domain flow field data of the front side window surface of the car to obtain: p * (x,y,ω)=∫p(x,y,t)e (-iωt) dt (1) In the formula, ω is the angular frequency, x and y are spatial quantities, t is time, e is the natural constant, and i is the imaginary number; p(x,y,t) represents the time-domain flow field data; p * (x,y,ω) represents the time-domain flow field data after Fourier transform; 2.2) Perform a second Fourier transform on the pressure fluctuations in the frequency domain in the x and y directions to obtain the pressure data P in the wavenumber domain. * (k x ,k y ,ω), that is: In the formula, k x Let k be the wave number in the x-direction. y The wave number in the y-direction; 2.3) In the two-dimensional wavenumber domain (k x ,k y Within the range, sound energy and water kinetic energy are separated based on the difference in their propagation speeds. The distribution of sound energy is shown below: In the formula, c is the speed of sound; f is the frequency; The distribution of water kinetic energy is shown below: In the formula, u is the airflow velocity; 2.4) The decomposed acoustic energy is used as the acoustic pressure excitation of the front side window surface of the car, and the decomposed hydrodynamic energy is used as the hydrodynamic pressure excitation of the front side window surface of the car; the acoustic pressure excitation and the hydrodynamic pressure excitation of the front side window surface of the car are used as the frequency domain flow field excitation data of the front side window surface of the car.

5. The method for frequency segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, The vehicle's geometric parameters were obtained through geometric measurement methods; The vehicle geometry parameters include, but are not limited to: vehicle width, front wind angle, A-pillar width, A-pillar deflection, A-pillar side step height, A-pillar lateral tilt angle, A-pillar angle, rearview mirror handle thickness, rearview mirror inner side angle, rearview mirror outer side angle, rearview mirror upper deflection, rearview mirror bottom deflection, rearview mirror distance from side window, rearview mirror height, and rearview mirror width, etc., which are parameters that describe the geometry of the A-pillar and rearview mirror.

6. The method for frequency-segment modeling of a neural network model for predicting flow field excitation in a vehicle's front side window according to claim 1, characterized in that, Step 4), which involves calculating the contribution of vehicle geometry parameters to each flow field excitation, includes: 4.1) Normalize the vehicle's geometric parameters; 4.2) Using the normalized geometric parameters as input variables and the flow field excitation of the front side window of the car as output, establish an expression in the form of regression analysis, namely: In the formula, X j Let β represent the j-th input variable, j = 1, 2, ..., m. j Y represents the regression coefficient corresponding to the j-th input variable, where β0 is a constant; Y is the output. 4.3) Calculate the regression coefficients for each input variable using a regression problem optimization algorithm; 4.4) Calculate the contribution of each input variable to the output, i.e.: In the formula, For contribution amount.

7. The method for frequency segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, In step 5), the process of selecting the vehicle geometric parameters whose contribution ranks in the top n% within each sub-frequency band as key geometric parameters is as follows: Within each sub-frequency band, the average contribution of each geometric parameter is calculated, and the geometric parameters whose average contribution is in the top n% are selected as the key geometric parameters of that sub-frequency band.

8. The method for frequency segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, Step 6) involves the following steps in building a neural network model using machine learning techniques: 6.1) Determine the basic structure of the neural network, including one input layer, one or more hidden layers, and one output layer; 6.2) Determine the number of neurons in each layer of the neural network.

9. The method for frequency-segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that, Step 7), which involves training the neural network model using the sub-frequency band sample dataset, includes: 7.1) Normalize the data in the sample dataset and divide the normalized sample dataset into a test set and a training set; 7.2) Select the training function and activation function for each hidden layer of the neural network model; 7.3) Divide the sample set corresponding to each sub-frequency band into a training set and a test set, using the sub-frequency bands as units; 7.4) Train the neural network model using the training set; 7.5) Test the neural network model using the test set. If the error is less than the preset value, output the front window flow field excitation prediction model; otherwise, reconstruct the sample dataset. 7.6) Repeat steps 7.3)-7.5) to obtain the front window flow field excitation prediction model corresponding to each sub-frequency band.

10. The method for frequency segment modeling of a neural network model for predicting flow field excitation of a car's front side window according to claim 1, characterized in that: The neural network model includes a backpropagation neural network model.

Images