Equivalent statistical energy modeling method for accurate prediction of wind noise in automobile
The equivalent statistical energy modeling method addresses the limitations of existing methods by using CFD simulation and optimization algorithms to create a simplified SEA model for accurate wind noise prediction in automobiles, reducing time and cost while enhancing prediction accuracy.
Patent Information
- Application Number
- CN202411050362.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-12-20
AI Technical Summary
Existing methods for predicting wind noise in automobiles, such as finite element and statistical energy analysis (SEA), are limited by the need for detailed modeling of substructure components and inaccurate representation of acoustic packages, leading to poor prediction accuracy and high time and cost.
An equivalent statistical energy modeling method using CFD simulation, wind tunnel testing, and a Gray Wolf global optimization algorithm to adjust damping loss factors, allowing for a simplified SEA model that accurately predicts wind noise without detailed subsystem modeling or experimental parameter selection.
The method provides an accurate wind noise prediction model that reduces the need for detailed modeling and experimental parameter selection, saving time and cost while improving prediction accuracy, and enabling aerodynamic styling optimization.
Smart Images
Figure CN119164596A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of aeroacoustics, and in particular to an equivalent statistical energy modeling method for accurately predicting wind noise in a car. Background Art
[0002] With the advancement of automobile technology and the improvement of people's living standards, automobile users have higher and higher requirements for automobile riding comfort. The noise level inside the vehicle when the vehicle is driving has become an important indicator affecting user satisfaction. Wind noise is one of the main noise sources of automobiles, and its severity increases most significantly with the increase of vehicle speed. Usually, wind noise becomes significant at speeds above 80kph and gradually becomes the dominant part of the automobile noise perceived by people.
[0003] The simulation calculations of the sound field and its propagation in the passenger cabin mainly use finite element, boundary element, SEA (Statistical energy analysis) and other methods. The advantage of finite element is that it can calculate the noise response of a specific position in space, but due to the limitations of grid size and computing resources, this method is only suitable for solving low-frequency problems. The SEA method divides the structure into multiple vibration mode groups that store energy according to the principle of modal similarity. It does not need to calculate the response of each mode, and the calculation efficiency is much higher than FEM.
[0004] The prediction accuracy of the SEA method is affected by the refinement of the modeling. However, the car model used for noise prediction is usually composed of hundreds of substructure components and dozens of acoustic package arrangements. Among them, the acoustic performance parameters of the acoustic package need to be measured through experiments or selected based on experience, and the installation method of the actual vehicle acoustic package is often not accurately reflected in the SEA model, which makes the refined modeling of the SEA model very time-consuming. At the same time, the prediction accuracy is poor because the acoustic package cannot fully simulate the actual situation.
[0005] Therefore, it is of great significance to develop an equivalent statistical energy modeling method for accurate prediction of wind noise in automobiles. Summary of the invention
[0006] The purpose of the present invention is to provide an equivalent statistical energy modeling method for accurately predicting wind noise in a car, so as to solve the problems existing in the prior art.
[0007] The technical solution adopted to achieve the purpose of the present invention is as follows:
[0008] The equivalent statistical energy modeling method for accurate prediction of wind noise in automobiles includes the following steps:
[0009] 1) Use CFD software to simulate the flow field outside the test vehicle and extract the sound pressure and hydrodynamic pressure data on the side window surface.
[0010] 2) The sound pressure level spectrum at the external ear of the test vehicle driver is measured using a wind tunnel test.
[0011] 3) Establish a simple SEA model based on the geometric dimensions of the test vehicle and assign material properties to the subsystems.
[0012] 4) Load the side window excitation obtained in step 1) onto the simplified SEA model in step 3) and use VAone software to solve the response of the simplified SEA model.
[0013] 5) Use the Gray Wolf global optimization algorithm to adjust the damping loss factor of the simple SEA model acoustic cavity so that the response of step 4) is close to the experimental value of step 2). After a certain number of iterative optimizations, the optimal damping loss factor is obtained. Among them, each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the equivalent model acoustic cavity.
[0014] 6) The optimal damping loss factor is loaded on the simple SEA model to obtain the equivalent SEA model. The equivalent SEA model is used to predict the path of air pulsation and aerodynamic noise propagation into the vehicle.
[0015] Furthermore, in step 1), the pressure pulsation data on the side window surface is subjected to wave number decomposition to obtain the acoustic pressure and the hydrodynamic pressure.
[0016] Furthermore, the incompressible fluid combined with the acoustic perturbation equation physical model of Starccm+ software is used to solve the acoustic pressure pulsation and turbulent pressure pulsation on the side window surface.
[0017] Further, in step 2), the digital artificial head is placed in the driving position and fixed on the seat with a safety belt. The doors and windows of the car are closed, and the gaps on the outer surface of the car body are sealed with tape.
[0018] Further, step 3) specifically includes the following sub-steps:
[0019] 3.1) Measure the wheelbase L, track B, overall height H and wheel radius R of the test vehicle w . Get the length a, width b and height c of the simple SEA model:
[0020] a=L#(1)
[0021] b=B#(2)
[0022] c=HR w #(3)
[0023] 3.2) By measuring the door length L D and rear overhang L b , Side window top edge height H gu and the height H of the lower edge of the side window g1 , get the side window dimensions d, e, f and h of the simple model:
[0024] d=L D #(4)
[0025] e=0.8L D #(5)
[0026] f=0.4(L b -R w )#(6)
[0027] h=c-1.5(H gu -H gl )#(7)
[0028] 3.3) Determine the top size g of the simple model:
[0029] g=0.5W R #(8)
[0030] Where W R The width of the roof of the vehicle.
[0031] Furthermore, in step 4), the hydrodynamic pressure and the acoustic pressure are loaded on the turbulent boundary layer sound source model and the diffuse sound field sound source model. The turbulent boundary layer and the diffuse sound field sound source model are only loaded on the left and right front side windows.
[0032] Furthermore, in step 5), the wind tunnel measured spectrum is used as a benchmark value, and the Grey Wolf global optimization algorithm is used to adjust the acoustic cavity damping loss factor of the simple SEA model.
[0033] Further, step 5) specifically includes the following sub-steps:
[0034] 5.1) Randomly initialize wolf individuals in Matlab. Each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the simple SEA model vocal cavity, with a total of 16 dimensions. The optimization range under each dimension is 0 to 1. Select the first round of α wolf, β wolf and δ wolf according to the fitness formula.
[0035] 5.2) In the new round of iterations, individuals in the population move according to the positions of the α wolf, β wolf, and δ wolf selected in the previous round.
[0036] 5.3) The moved individuals are assigned to the equivalent acoustic vibration model through the interface software, and the VA one solver is called to calculate the predicted value of the sound pressure level spectrum response in the car under the given configuration of the individual. Each individual represents a set of DLF spectra of the acoustic cavity.
[0037] 5.4) The root mean square error between the predicted value of the in-car sound pressure level spectrum response and the benchmark value is taken as the fitness of the individual, and the calculation formula is as follows:
[0038]
[0039] In the formula, represents the predicted sound pressure level of the ith center frequency band, It represents the measured sound pressure level of the ith center frequency band, and n is the number of frequency bands.
[0040] 5.5) According to the individual fitness, select the α wolf, β wolf and δ wolf generated in this iteration as the reference direction for the wolf pack to move in the next iteration. Stop the optimization until the maximum number of iterations is reached.
[0041] The technical effect of the present invention is unquestionable: based on a small amount of aerodynamic acoustic wind tunnel test data of the prototype vehicle, a wind noise equivalent SEA model of the vehicle is constructed. It overcomes the difficulty of traditional statistical energy models in accurately capturing the impact of the optimization of the exterior styling of the car on the wind noise in the passenger compartment. In addition, there is no need to carry out detailed modeling of subsystems such as body panels, nor is there any need to select the acoustic parameters of sound-absorbing and insulating materials through experience or test measurement, thus saving a lot of time and cost. Moreover, the prediction accuracy of the model does not depend on the user's modeling level. The method provided by the present invention provides an accurate passenger compartment wind noise prediction model for the aerodynamic styling optimization of the prototype vehicle, reduces the number of wind tunnel tests, saves economic costs, and has a high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is the CFD calculation domain of the test vehicle;
[0043] Figure 2 Set up mesh encryption for the test vehicle;
[0044] Figure 3 The acoustic pressure and hydrodynamic pressure results are decomposed by wave number;
[0045] Figure 4 How to install the in-vehicle noise test equipment;
[0046] Figure 5 The spectral response of the driver's external ear sound pressure level in the car measured by wind tunnel test;
[0047] Figure 6 Mapping the geometric relationship between the simple SEA model and the test vehicle;
[0048] Figure 7 is the equivalent SEA model of the test vehicle;
[0049] Figure 8 Optimization process for acoustic cavity damping loss factor of simple SEA model;
[0050] Fig. 9 is the prediction effect of the equivalent SEA model;
[0051] Fig.10This is a flow chart of the rapid prediction method for wind noise inside the vehicle. DETAILED DESCRIPTION
[0052] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.
[0053] Embodiment 1:
[0054] See also Fig.10 This embodiment provides an equivalent statistical energy modeling method for accurately predicting wind noise in a car, comprising the following steps:
[0055] 1) Use CFD software to simulate the flow field outside the test vehicle, and extract the acoustic pressure and hydrodynamic pressure data on the side window surface. In this embodiment, the pressure pulsation data on the side window surface is decomposed by wave number to obtain the acoustic pressure and hydrodynamic pressure. The incompressible fluid combined with the acoustic perturbation equation physical model of Starccm+ software is used to solve the acoustic pressure pulsation and turbulent pressure pulsation on the side window surface.
[0056] See also Figure 1 to Figure 3 The external flow field simulation uses a rectangular calculation domain with geometric dimensions of 12 times the vehicle length, 9 times the vehicle width, and 7 times the vehicle height. The outlet, inlet, side, and top surfaces use free flow boundaries. The ground is set as a sliding wall. The grid distribution is as follows: Figure 2 As shown in the figure, the sound source areas near the A-pillar, rearview mirror and side window are locally encrypted, with a minimum grid size of 2 mm and a grid number of 32.7 million. The velocity amplitude at the wind tunnel entrance is 120 km / h, and the incompressible SST k-omega detached vortex model is used to simulate the external flow field of the car. The detached flow solver is used, the time step is 0.05 ms, and the total simulation time is 0.25 s. The momentum equation uses a bounded central difference format, and the energy equation uses a second-order convection format for discretization. Full Y+ wall treatment is performed, and the aeroacoustic model is the acoustic perturbation equation.
[0057] 2) Use wind tunnel test to measure the sound pressure level spectrum at the driver's outer ear of the test vehicle. Put the digital artificial head into the driver's seat and fasten it to the seat with a seat belt. Close the doors and windows of the car and seal the gaps on the outer surface of the car body with tape.
[0058] 3) A simple SEA model is established according to the geometric dimensions of the test vehicle, and material properties are assigned to the subsystems. The SEA board subsystem of the simple SEA model includes the front windshield, the front panel, four side windows, four door panels, the rear window, the top cover and the floor. The sound cavity subsystem includes four sound cavities, namely the front head, the rear head, the front leg and the rear leg. The subsystems are interconnected to transfer sound energy. Among them, the two front side windows use 3.96mm laminated glass, and the measured DLF spectrum is used as the damping term of the side window subsystem. The remaining SEA panels are made of 1mm thick aluminum, and the damping term is set to 1% of the DLF spectrum. The damping terms of the four sound cavity subsystems are also set to 1% of the DLF. The coupling loss factor spectrum between the subsystems is also 1%.
[0059] Step 3) specifically includes the following sub-steps:
[0060] 3.1) Measure the wheelbase L, track B, overall height H and wheel radius R of the test vehicle w . Get the length a, width b and height c of the simple SEA model:
[0061] a=L#(1)
[0062] b=B#(2)
[0063] c=HR w #(3)
[0064] 3.2) By measuring the door length L D and rear overhang L b , Side window top edge height H gu and the height H of the lower edge of the side window g1 , get the side window dimensions d, e, f and h of the simple model:
[0065] d=L D #(4)
[0066] e=0.8L D #(5)
[0067] f=0.4(L b -R w )#(6)
[0068] h=c-1.5(H gu -H gl )#(7)
[0069] 3.3) Determine the top size g of the simple model:
[0070] g=0.5W R #(8)
[0071] Where W R The width of the roof of the vehicle.
[0072] 4) Load the side window excitation obtained in step 1) on the simple SEA model in step 3), and use VAone software to solve the response of the simple SEA model. Load the hydrodynamic pressure and sound pressure on the turbulent boundary layer sound source model and the diffuse sound field sound source model. Among them, the turbulent boundary layer and diffuse sound field sound source models are only loaded on the left and right front side windows.
[0073] 5) Use the Gray Wolf global optimization algorithm to adjust the damping loss factor of the simple SEA model acoustic cavity so that the response of step 4) is close to the experimental value of step 2). The optimal damping loss factor is obtained through a certain number of iterative optimizations. Among them, each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the equivalent model acoustic cavity. Take the wind tunnel measured spectrum as the benchmark value, and use the Gray Wolf global optimization algorithm to adjust the damping loss factor of the simple SEA model acoustic cavity. Step 5) specifically includes the following sub-steps:
[0074] 5.1) Randomly initialize wolf individuals in Matlab. Each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the simple SEA model vocal cavity, with a total of 16 dimensions. The optimization range under each dimension is 0 to 1. Select the first round of α wolf, β wolf and δ wolf according to the fitness formula.
[0075] 5.2) In the new round of iterations, individuals in the population move according to the positions of the α wolf, β wolf, and δ wolf selected in the previous round.
[0076] 5.3) The moved individuals are assigned to the equivalent acoustic vibration model through the interface software, and the VA one solver is called to calculate the predicted value of the sound pressure level spectrum response in the car under the given configuration of the individual. Each individual represents a set of DLF spectra of the acoustic cavity.
[0077] 5.4) The root mean square error between the predicted value of the in-car sound pressure level spectrum response and the benchmark value is taken as the fitness of the individual, and the calculation formula is as follows:
[0078]
[0079] In the formula, represents the predicted sound pressure level of the ith center frequency band, It represents the measured sound pressure level of the ith center frequency band, and n is the number of frequency bands.
[0080] 5.5) According to the individual fitness, select the α wolf, β wolf and δ wolf generated in this iteration as the reference direction for the wolf pack to move in the next iteration. Stop the optimization until the maximum number of iterations is reached.
[0081] 6) The optimal damping loss factor is loaded on the simple SEA model to obtain the equivalent SEA model. The equivalent SEA model is used to predict the path of air pulsation and aerodynamic noise propagation into the vehicle.
[0082] Embodiment 2:
[0083] The main contents of this embodiment are the same as those of embodiment 1, wherein the wind tunnel test environment conditions are as follows: the temperature of the aeroacoustic wind tunnel is strictly controlled at 25°C ± 1°C. The wind noise inside the vehicle is measured at a yaw angle of 0° and an incoming flow velocity of 120km / h. After the main fan reaches the target wind speed and stabilizes, the front end of the HEAD lab starts to collect data, with a collection time of 10s and a sampling frequency of 48000Hz. The time domain signal obtained from the experiment is Fourier transformed.
[0084] Embodiment 3:
[0085] The main contents of this embodiment are the same as those of embodiment 1, wherein see Figures 1 to 8 , this embodiment includes the following steps:
[0086] 1) CFD software was used to simulate the flow field outside the vehicle body for five configurations (Base1, Base2, Body1, Body2 and Body3) of the test vehicle, with an airflow speed of 120 km / h. The unsteady physical model used the SST k-omega detached vortex model and the acoustic disturbance equation, with a time step of 5×10 -5 s. The pressure pulsation data on the side window surface are decomposed by wave number to obtain the acoustic pressure and hydrodynamic pressure.
[0087] 2) Use an artificial head in a wind tunnel test to measure the sound pressure level spectrum at the driver's outer ear of the test vehicle. Place the digital artificial head HMS IV.0 in the driver's seat and secure it to the seat with a seat belt. Connect the digital artificial head to the HEAD lab module. When all the equipment is connected, close the doors and windows and seal the gaps on the outer surface of the car body with tape. During the test, the temperature of the aeroacoustic wind tunnel is strictly controlled at 25℃±1℃. Measure the wind noise inside the car under the conditions of 0° yaw angle and 120km / h incoming flow velocity. After the main fan reaches the target wind speed and stabilizes, the front end of the HEAD lab starts to collect data with a collection time of 10s and a sampling frequency of 48000Hz. Perform Fourier transform on the time domain signal collected in the experiment.
[0088] 3) Establish a simple SEA model based on the geometric dimensions of the test vehicle and assign material properties to the subsystem. The steps to establish a simple SEA model include:
[0089] 3.1) Measure the wheelbase L, track B, overall height H and wheel radius R of the test vehicle w The length (a), width (b) and height (c) of the simple SEA model can be obtained, namely:
[0090] a=L#(1)
[0091] b=B#(2)
[0092] c=HR w #(3)
[0093] 3.2) By measuring the door length L D and rear overhang L b , Side window top edge height H gu and the height H of the lower edge of the side window g1 , we can get the side window dimensions d, e, f and h of the simple model:
[0094] d=L D #(4)
[0095] e=0.8L D #(5)
[0096] f=0.4(L b -R w )#(6)
[0097] h=c-1.5(H gu -H gl )#(7)
[0098] 3.3) By measuring the width W of the roof R Determine the top dimensions of the simple model g
[0099] g=0.5W R #(8)
[0100] Through calculation, the key geometric dimensions of the simple SEA model can be obtained, as shown in the following table.
[0101] Table 1
[0102]
[0103] Furthermore, when assigning material properties to the acoustic subsystem of the simple SEA model, the two front side windows use 3.96 mm laminated glass, and the measured DLF spectrum is used as the damping term of the side window subsystem. The remaining SEA panels are all made of 1 mm thick aluminum, and the damping term is set to 1% of the DLF spectrum. The damping terms of the four acoustic cavity subsystems are also set to 1% of DLF.
[0104] 4) Load the side window excitation obtained by the simulation of the Base1 configuration onto the simplified SEA model in step 3). Initialize the wolf pack individuals in Matlab, and each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the equivalent model acoustic cavity. The RMSE of the predicted spectrum of the simplified model and the experimental spectrum of the Base1 configuration is used as the fitness of the individual.
[0105] 5) In the new round of iterations, the individuals in the population move according to the positions of the α wolf, β wolf and δ wolf selected in the previous round. The moved individuals are then assigned to the equivalent acoustic vibration model through the interface, and the VA one solver is called to calculate the fitness of the individual. Finally, the α wolf, β wolf and δ wolf generated in this iteration are selected according to the individual fitness as the reference direction for the movement of the wolf pack in the next iteration. After a certain number of iterative optimizations, the optimal damping loss factor is obtained, as shown in the following table. Load it on the simple SEA model to obtain the equivalent SEA model.
[0106] Table 2
[0107]
[0108] 6) See Fig. 9 , using the Base1 configuration to predict the wind noise response in the car under the other body configurations (Body1, Body2, and Body3). The relative error of the total sound pressure level and the RMSE of the 1 / 3 octave spectrum are shown in the following table.
[0109] Table 3
[0110]
[0111]
[0112] It can be seen that in the mid-to-high frequency band (250Hz-8000Hz), the total sound pressure level prediction error finally obtained by this prediction method is controlled within 3%, and the RMSE is controlled at around 2dB, which further verifies the practicality and accuracy of the proposed equivalent statistical energy modeling method for accurate prediction of wind noise in automobiles.
Claims
1. An equivalent statistical energy modeling method for accurate prediction of wind noise in automobiles, characterized in that: The following steps are involved: 1) Use CFD software to simulate the flow field outside the test vehicle and extract the sound pressure and hydrodynamic pressure data on the side window surface; 2) The sound pressure level spectrum at the external ear of the test vehicle driver is measured by wind tunnel test; 3) Establish a simple SEA model based on the geometric dimensions of the test vehicle and assign material properties to the subsystems; 4) Loading the side window excitation obtained in step 1) onto the simplified SEA model in step 3), and using VAone software to solve the response of the simplified SEA model; 5) Use the Gray Wolf global optimization algorithm to adjust the damping loss factor of the simple SEA model acoustic cavity so that the response of step 4) is close to the experimental value of step 2); obtain the optimal damping loss factor through a certain number of iterative optimizations; wherein each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the equivalent model acoustic cavity; 6) The optimal damping loss factor is loaded on the simple SEA model to obtain an equivalent SEA model; the equivalent SEA model is used to predict the path of air pulsation and aerodynamic noise propagation into the vehicle.
2. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 1 is characterized in that: In step 1), the pressure pulsation data on the side window surface is subjected to wave number decomposition to obtain the acoustic pressure and the hydrodynamic pressure.
3. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 2 is characterized in that: The incompressible fluid combined with the acoustic perturbation equation physical model of Starccm+ software is used to solve the acoustic pressure pulsation and turbulent pressure pulsation on the side window surface.
4. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 1 is characterized in that: In step 2), the digital artificial head is placed in the driving seat and secured to the seat with a safety belt; the doors and windows are closed, and the gaps on the outer surface of the vehicle body are sealed with tape.
5. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 1 is characterized in that: Step 3) specifically includes the following sub-steps: 3.1) Measure the wheelbase L, track B, overall height H and wheel radius R of the test vehicle w ; Get the length a, width b and height c of the simple SEA model: a＝l#(1) b=B#(2) c=H-R w #(3) 3.2) By measuring the door length L D and rear overhang L b , Side window top edge height H gu and the height H of the lower edge of the side window gl , get the side window dimensions d, e, f and h of the simple model: d＝l D #(4) e=0.8L D #(5) f=0.4(l b -R w )#(6) h=c-1.5(H gu -H gl )#(7) 3.3) Determine the top size g of the simple model: g=0.5W R #(8) Where W R The width of the roof of the vehicle.
6. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 1 is characterized in that: In step 4), the hydrodynamic pressure and the acoustic pressure are loaded on the turbulent boundary layer sound source model and the diffuse sound field sound source model; wherein the turbulent boundary layer and the diffuse sound field sound source model are only loaded on the left and right front side windows.
7. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 1 is characterized in that: In step 5), the wind tunnel measured spectrum is used as the benchmark value, and the Grey Wolf global optimization algorithm is used to adjust the acoustic cavity damping loss factor of the simple SEA model.
8. The equivalent statistical energy modeling method for accurate prediction of wind noise in a vehicle according to claim 7 is characterized in that: Step 5) specifically includes the following sub-steps: 5.1) Randomly initialize wolf individuals in Matlab. Each dimension of the individual represents a central frequency of the 1 / 3 octave DLF spectrum of the simple SEA model vocal cavity, with a total of 16 dimensions. The optimization range under each dimension is 0 to 1. Select the first round of α wolf, β wolf and δ wolf according to the fitness formula; 5.2) In the new round of iterations, the individuals in the population move according to the positions of the α wolf, β wolf, and δ wolf selected in the previous round; 5.3) Assign the moved individuals to the equivalent acoustic vibration model through the interface software, call the VAone solver, and calculate the predicted value of the sound pressure level spectrum response in the car under the given configuration of the individual; a set of DLF spectra of the acoustic cavity represented by each individual; 5.4) The root mean square error between the predicted value of the in-car sound pressure level spectrum response and the benchmark value is taken as the fitness of the individual, and the calculation formula is as follows: In the formula, represents the predicted sound pressure level of the ith center frequency band, represents the measured sound pressure level of the i-th central frequency band, and n is the number of frequency bands; 5.5) According to the individual fitness, select the α wolf, β wolf and δ wolf generated in this iteration as the reference direction for the wolf pack to move in the next iteration. Stop the optimization until the maximum number of iterations is reached.