Material sound absorption performance measurement method based on space Fourier transform

Abstract

The invention discloses a material sound absorption performance measurement method based on spatial Fourier transform, which comprises the following steps: acquiring acoustic data of limited measurement points of a target material, and outputting spatial sound field data of the target material by utilizing a fitting model based on the acoustic data of the limited measurement points, calculating a sound absorption coefficient and a reflection coefficient of a target material according to the output space sound field data; therefore, the mapping relation between the limited measuring point data and the global complete measuring point data is learned through the neural network model, accurate space Fourier transform can be achieved under the condition that only a small number of limited measuring points are used, acoustic quantities such as the reflection coefficient and the sound absorption coefficient of the material can be calculated through the value output by the model, and the accuracy of acoustic measurement is improved. And the measurement efficiency and the economical efficiency are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of acoustic material performance measurement, and in particular to a material sound absorption performance measurement method based on spatial Fourier transform. BACKGROUND

[0002] With the deepening of industrialization and the improvement of people's living standards, noise pollution has become increasingly prominent, and the resulting noise control problem has gradually attracted people's attention. Transmission path control is one of the most commonly used and effective means of noise prevention, and its core principle is: in the process of sound propagation, sound-absorbing or damping materials are set to absorb or dissipate sound energy when it contacts the material, thereby reducing the sound pressure level transmitted to the receiving end. In practical applications, reasonable selection and layout of the type, thickness and installation position of sound-absorbing materials can significantly improve the noise control effect.

[0003] In order to scientifically design and optimize the application scheme of these materials, it is essential to have a deep understanding of the acoustic characteristics of the materials, including the sound absorption coefficient, impedance matching characteristics, and frequency response characteristics. There are various methods for measuring material characteristics, including impedance tube method, reverberation chamber method, and spatial Fourier transform method. Among them, the spatial Fourier transform method can simultaneously obtain the sound absorption performance of the material at different angles, achieving high-resolution characterization of the sound field, so it has been widely used in scientific research and engineering practice.

[0004] However, the traditional spatial Fourier transform method usually needs to be operated on a large measurement surface. On the one hand, in order to obtain a higher angle resolution, the length of the measurement surface needs to be extended; on the other hand, in order to expand the angle range of the measurement, the spacing between the measurement points needs to be reduced. This results in the need to use a large number of microphones in the conventional method, making the measurement system costly and complex in structure, increasing the difficulty of the experiment and the complexity of the operation, and also limiting the efficiency and practicality of the measurement. SUMMARY

[0005] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application proposes a material sound absorption performance measurement method based on spatial Fourier transform, which uses a small amount of front-end measurement point data to predict the sound absorption performance of the full domain complete measurement point by constructing a data-driven model, thereby breaking through the dependence of the traditional spatial Fourier transform method on a large number of measurement point arrangements. And in the model design, dynamic over-parameterization weights are introduced to enhance the representation ability, and a loss function combined with physical information constraints is used to guide the training process to fully utilize the acoustic characteristics of the material. This method not only significantly improves the training effect of the model, but also effectively improves the accuracy and measurement efficiency of the sound absorption material test, achieving rapid and reliable evaluation of the sound absorption performance of the material.

[0006] A method for measuring the sound absorption performance of materials based on spatial Fourier transform according to an embodiment of the present invention includes: Acquire acoustic data from a limited number of measurement points on the target material; Based on the acoustic data from the limited measurement points, the spatial acoustic field data of the target material is output using a fitting model. Calculate the sound absorption coefficient and reflection coefficient of the target material based on the output spatial sound field data; The construction of the fitting model includes obtaining spatial acoustic datasets collected from a limited number of local measurement points under different material properties. and spatial acoustic datasets collected from complete measurement points across the entire domain ; with the aforementioned spatial acoustic dataset As input to the model, a spatial acoustic dataset is used. The labels are used as model labels for training. The optimal weights of the fitted model are obtained by minimizing the error between the fitting results and the corresponding labels. Based on the model weights, the sound pressure mapping relationship F between local finite measurement points and global complete measurement points is constructed, and the fitted model is obtained.

[0007] According to some embodiments of the present invention, the step of acquiring spatial acoustic data of a limited number of measurement points of the target material includes: firstly constructing a sound field model of the target material, setting a first measurement surface and a second measurement surface above the surface of the target material, the second measurement surface being set above the first measurement surface, setting multiple microphones at intervals along the same radial direction of the first measurement surface and the second measurement surface, and acquiring acoustic data at corresponding positions through the microphones.

[0008] According to some embodiments of the present invention, in the step of constructing the fitting model, the properties of different materials are obtained through a porous medium model, wherein the expression of the porous medium model is: ; in, and This represents the characteristic impedance and complex wavenumber of a material. and These represent fluid density and the speed of sound in air, respectively. Represents frequency, The value represents the flow resistance of the material, and j is the imaginary unit.

[0009] According to some embodiments of the present invention, the acquisition of spatial acoustic datasets collected from a limited number of local measurement points under different material properties is described. and spatial acoustic datasets collected from complete measurement points across the entire domain The steps include the following: the spatial acoustic data expression is: ; in, Indicates the sound pressure at the measuring point. This indicates the distance between the sound source and the measuring point. Indicates at the angle of incidence The reflectance coefficient below, Indicates the wave number in the air. Indicates radial wavenumber, and These represent the heights of the sound source and the measuring surface, respectively. Represents the Bessel function. Indicates the radial length of the measured surface. and The characteristic impedance and complex wave number of the material are represented by j, which is the imaginary unit. Spatial acoustic datasets acquired from a limited number of local measurement points and a spatial acoustic dataset acquired from all measurement points across the entire domain The expressions are as follows: ; ; in, and These represent the sets of sound pressure levels at different measuring points on measuring surfaces 1 and 2, respectively, under a finite arrangement of measuring points. and These represent the sound pressure at the nth measuring point on measuring surface 1 and measuring surface 2, respectively, under a finite arrangement of measuring points; and These represent the sound pressure sets at different measuring points on measuring surface 1 and measuring surface 2, respectively, under a complete set of measuring points across the entire domain. and These represent the sound pressure at the m-th measuring point on measuring surface 1 and measuring surface 2, respectively, under the complete measurement point layout across the entire region.

[0010] According to some embodiments of the present invention, the step of training the fitting model includes constructing a deep learning model and associating the input layer of the deep learning model with the output of the next layer through a weight matrix and a bias matrix. The deep learning model consists of several layers, each layer performing a linear transformation on the input features through a weight matrix and a bias vector, and then mapping them through an activation function. The recursive relationship between layers can be expressed as follows: ; in, Indicates input variables, Indicates the output variable of the next level. Indicates the activation function; The weight matrix includes a low-rank matrix, a diagonal matrix, a learnable perturbation matrix, and a basic weight matrix. The expression for the weight matrix is ​​as follows: ; in, Representing a low-rank matrix using the fundamental matrix and Reconstructed to obtain; Represents a fixed diagonal matrix; It is a learnable perturbation matrix; Represents the basic weight matrix; The kernel size is... and These represent the number of input and output channels, respectively. Einstein was referring to the summation. This represents a reconstruction of the tensor dimension.

[0011] According to some embodiments of the present invention, in the step of obtaining the optimal weights of the fitted model by minimizing the error between the fitting result and the corresponding label, a loss function is constructed. and in the loss function The physical loss function is incorporated into it, based on the loss function The difference between the model output and the corresponding label is calculated, and the parameters are updated using a chain rule, which is then incorporated into the loss function; the expression for the weight matrix is: ; in, ; ; In the formula, the parameter Lphysics is the physical information loss function. The importance coefficient of the physical loss term, where n and N represent the measurement point index and the number of measurement points, respectively; and These represent the sets of spatial sound field data for the model output and the corresponding labels, respectively. and These represent the reflection coefficient values ​​calculated from the model output and the corresponding labels, respectively. and These represent the spatial spectral information of measurement surface 1 calculated through model output and corresponding labels, respectively. and These represent the spatial spectral information of measurement surface 1 obtained through model output and corresponding labels, respectively. Indicates the transverse wave number. and These represent the heights of measuring surface 1 and measuring surface 2, respectively.

[0012] According to some embodiments of the present invention, the mapping relationship between the local finite measurement points and the global complete measurement points is as follows: ; in, and These represent the sound pressure values ​​at the measuring points along the radial measuring lines of measuring surface 1 and measuring surface 2, respectively. and These represent spatial sound field datasets captured by finite measurement points and global complete measurement points under different material properties, respectively.

[0013] According to some embodiments of the present invention, the step of outputting the spatial sound field data of the target material using a fitting model includes: performing a two-dimensional Fourier transform on the x and y planes using a Fourier-Bessel transform to obtain the spatial sound field information of the target material, specifically as follows: ; in, and These represent the global measurement point data on different measurement lines output by the model. and This represents the corresponding spatial spectrum information. Represents the Bessel function. Indicates the radial length of the measured surface. Indicates the radial wave number.

[0014] According to some embodiments of the present invention, the step of calculating the sound absorption coefficient and reflection coefficient of the target material based on the output spatial sound field data includes: reflection coefficient and sound absorption coefficient The expression is: ; Among them, spatial sound field incident wave and reflected waves The expression is as follows: ; In the formula, Indicates the transverse wave number. and These represent the heights of measuring surface 1 and measuring surface 2, respectively.

[0015] Beneficial effects: This invention learns the mapping relationship between finite measurement point data and complete measurement point data across the entire domain through a neural network model. It can achieve accurate spatial Fourier transform using only a small number of finite measurement points, reconstruct the complete spatial sound field, and directly calculate the sound absorption characteristics of the material as a function of the incident angle from the spatial sound field data output by the model, including acoustic quantities such as the material's reflection coefficient and sound absorption coefficient, thereby effectively improving measurement efficiency and economy.

[0016] In this invention, the traditional method of constructing the weight matrix in the network model is changed. The weight matrix is ​​reconstructed into a multi-degree-of-freedom learnable matrix composed of a low-rank matrix, a diagonal matrix, a learnable perturbation matrix, and a basic weight matrix, thereby enhancing the feature learning ability and realizing a more refined representation of input features and deep feature extraction.

[0017] This invention introduces dynamic overparameterized weights to enhance representation capabilities during model training and combines them with a loss function constrained by physical information to guide the training process to fully utilize the acoustic properties of materials. This allows the model prediction results to follow physical laws, thereby improving the accuracy and reliability of predictions. This not only significantly improves the training effect of the model but also effectively enhances the accuracy and measurement efficiency of sound-absorbing material testing, enabling rapid and reliable evaluation of the sound absorption performance of materials. Attached Figure Description

[0018] Figure 1 This is a flowchart of a method for measuring the sound absorption performance of materials based on spatial Fourier transform; Figure 2 This is a schematic diagram of a material sound absorption performance measuring device; Figure 3 This is a comparison diagram of the spatial spectrum of the sound pressure field obtained by the method of this invention and the existing technology. Figure 4 This is a comparison chart showing the results of extracting the material reflection coefficient using the method of this invention and existing techniques. Figure 5 This is a comparison chart showing the results of extracting the sound absorption coefficient of materials using the method of this invention and existing methods. Detailed Implementation

[0019] The technical solutions of the embodiments disclosed in this application will be clearly and completely described below with reference to the accompanying drawings. The descriptions of the embodiments are merely illustrative and exemplary, and are not intended to limit the scope of this disclosure or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort should fall within the scope of protection of this disclosure. Furthermore, techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification.

[0020] According to the reference below Figure 1 A method for measuring the sound absorption performance of materials based on spatial Fourier transform according to an embodiment of the present invention includes the following steps: Acquire spatial acoustic data of the target material, use a model to fit and output spatial sound field data based on the spatial acoustic data, and calculate the sound absorption coefficient and reflection coefficient of the material based on the output spatial sound field data.

[0021] Specifically, such as Figure 2 As shown, to facilitate understanding of the solution in this application, the solution is further elaborated through modeling. Figure 2 The sound field model shown includes a sound source, measurement surface 1, measurement surface 2, and a material surface. The sound source is located 0.1m above the material, and the heights of measurement surface 1 and measurement surface 2 are 0.01m and 0.03m, respectively.

[0022] During setup, a single-pole point source is used as the sound source. Microphones are arranged at equal intervals along the radial direction on measurement surface 1 and measurement surface 2, with a spacing of 0.01m. For a complete measurement point across the entire area, 101 microphones are arranged. A certain number of these measurement points are taken as finite measurement points, that is, for finite measurement points, 20 microphones are arranged.

[0023] It is important to note that a complete measurement point across the entire domain is equivalent to the theoretical measurement point of the sound field model. The sound absorption coefficient and reflection coefficient of the material can be directly calculated using the data from the complete measurement point across the entire domain. In contrast, a limited measurement point in a local area is equivalent to the actual measurement point of the sound field model during measurement.

[0024] Traditional methods often require expanding the measurement area and reducing the microphone spacing to obtain accurate measurement results. Expanding the measurement area reduces angular resolution and improves the accuracy of acoustic parameter measurements at small angles, while reducing the microphone spacing increases the angular range to obtain acoustic parameters across a wider range of angles. This undoubtedly requires a large number of loudspeakers, significantly increasing testing costs and experimental complexity. Therefore, the method proposed in this invention requires only a shorter sampling length and a smaller number of microphones to complete the measurement, thereby significantly reducing testing costs and simplifying the experimental procedure while maintaining measurement accuracy.

[0025] In practice, by changing the parameters of the porous media model on the material surface, datasets obtained from a finite number of measuring points under different measuring point arrangements are acquired. Dataset obtained from complete measurement points across the entire region Subsequently, the dataset As input to the model, the dataset As the reference label corresponding to the model output, the model is trained iteratively. During the model training process, the learning rate is 0.0007, the number of training rounds is 50, and ReLU is selected as the activation function.

[0026] The Adam optimizer is used during model training to iteratively update the network parameters. Specifically, in each training epoch, the model generates a predicted output based on the input finite measurement point data and compares it with the corresponding global complete measurement point data to calculate the loss function. Subsequently, gradient information is calculated through the backpropagation algorithm, and the Adam optimizer is used to update the network parameters, thereby progressively optimizing the network weights to best represent the mapping relationship between the finite measurement points and the global complete measurement points.

[0027] During training, the network weights that minimize the loss function are always retained as the current optimal weights, and this process is continuously iterated until the training loss function changes only slightly in several consecutive iterations. This indicates that the model performance has stabilized, and the optimal network weights that can achieve accurate mapping from finite measurement points to complete measurement points across the entire domain are finally obtained.

[0028] Of course, it is important to note the importance coefficient of the physical loss term. This will have a significant impact on the results. In order to improve the accuracy of model training, this invention sets... .

[0029] More specifically, step 1: Obtain training data. 1.1 By changing the internal parameters of the porous medium acoustic model in formula (1), material samples with different properties can be obtained.

[0030] Specifically, the porous medium model expression is as follows: (1) in, and This represents the characteristic impedance and complex wavenumber of a material. and These represent fluid density and the speed of sound in air, respectively. Represents frequency, The value represents the flow resistance of the material, and j is the imaginary unit.

[0031] 1.2 Using the combined formula (2-4) to obtain data on different material sample properties, where the dataset is obtained from a finite number of measurement points. Dataset obtained from complete measurement points across the entire region Spatial acoustic data can be obtained through the following expression (2): (2) in, Indicates the sound pressure at the measuring point. This indicates the distance between the sound source and the measuring point. Indicates at the angle of incidence The reflectance coefficient below, Indicates the wave number in the air. Indicates radial wavenumber, and These represent the heights of the sound source and the measuring surface, respectively. Represents the Bessel function. Indicates the radial length of the measured surface. and The characteristic impedance and complex wave number of the material are represented by j, which is the imaginary unit.

[0032] (3) (4) in, and These represent the sets of sound pressure levels at different measuring points on measuring surfaces 1 and 2, respectively, under a finite arrangement of measuring points. and These represent the sound pressure at the nth measuring point on measuring surface 1 and measuring surface 2, respectively, under a finite arrangement of measuring points; and These represent the sound pressure sets at different measuring points on measuring surface 1 and measuring surface 2, respectively, under a complete set of measuring points across the entire domain. and These represent the sound pressure at the m-th measuring point on measuring surface 1 and measuring surface 2, respectively, under the complete measurement point layout across the entire region.

[0033] Step 2: Model training; Specifically, the limited measurement point data obtained above As input to the model, complete measurement point data across the entire domain is used as the corresponding labels for the model output. Then, the model is trained and a deep learning model is constructed. The optimal model weights are obtained iteratively through forward and backward propagation processes.

[0034] In the forward propagation, the deep learning model consists of several layers. Each layer performs a linear transformation on the input features through a weight matrix and a bias vector, and then maps them through an activation function. The recursive relationship between layers can be expressed as: (5) in, Indicates input variables, Indicates the output variable of the next level. This represents the activation function.

[0035] In this way, the input layer is associated with the output of the next layer through the weight matrix and the bias matrix.

[0036] It should be emphasized that the weight matrix Unlike existing technologies, it no longer uses a single fixed weight, but is generated by a combination of a low-rank matrix, a diagonal matrix, a learnable perturbation matrix, and a basic weight matrix, as shown below: (6) in, Representing a low-rank matrix using the fundamental matrix and Reconstructed to obtain; Represents a fixed diagonal matrix; It is a learnable perturbation matrix; Represents the basic weight matrix; The kernel size is... and These represent the number of input and output channels, respectively. Einstein was referring to the summation. This represents a reconstruction of the tensor dimension.

[0037] During backpropagation, based on the loss function The difference between the model output and the corresponding label is calculated, and the parameters are updated using a chain rule. Simultaneously, in the loss function... Physical information is incorporated into the model, enabling the prediction results to follow physical laws, thereby improving the accuracy and reliability of the prediction. Its expression is shown in formula (7-9). In addition, the importance coefficient of physical information is also introduced into the loss function. By adjusting The size of the model will improve the prediction results.

[0038] (7) in: (8) (9) In the formula, The importance coefficient of the physical loss term, where n and N represent the measurement point index and the number of measurement points, respectively; and These represent the sets of spatial sound field data for the model output and the corresponding labels, respectively. and These represent the reflection coefficient values ​​calculated from the model output and the corresponding labels, respectively. and These represent the spatial spectral information of measurement surface 1 calculated through model output and corresponding labels, respectively. and These represent the spatial spectral information of measurement surface 1 obtained through model output and corresponding labels, respectively. Indicates the transverse wave number. and These represent the heights of measuring surface 1 and measuring surface 2, respectively.

[0039] The parameter Lphysics is the physical information loss function. This loss function is equal to the difference in reflection coefficients calculated from the model's output. Without this term, the loss function only shows the difference in model output. Adding this term allows the model to further measure the difference between the output and the target using physical information indicators, namely reflection coefficient indicators, thus improving the model's fit.

[0040] Finally, through the above steps, the model completes training and obtains optimal weights while minimizing the loss function, thereby establishing a mapping relationship between finite measurement points and the complete set of measurement points across the entire domain, as shown below: (11) in, and These represent the sound pressure values ​​at the measuring points along the radial measuring lines of measuring surface 1 and measuring surface 2, respectively. and These represent spatial sound field datasets captured by finite measurement points and global complete measurement points under different material properties, respectively.

[0041] Because the sound field is axisymmetric along the z-axis, a two-dimensional Fourier transform can be performed on the x and y planes using the Fourier-Bessel transform to obtain the spatial spectrum information of the material, as shown below: (12) in, and These represent the global measurement point data on different measurement lines output by the model. and This represents the corresponding spatial spectrum information. Represents the Bessel function. Indicates the radial length of the measured surface. Indicates the radial wave number.

[0042] Based on the obtained spatial spectrum information, the incident wave of the spatial sound field can be obtained. and reflected waves The expression is as follows: (13) in, Indicates the transverse wave number. and These represent the heights of measuring surface 1 and measuring surface 2, respectively.

[0043] Furthermore, the reflectance coefficient and sound absorption coefficient It can be represented as: (14) Once the model is trained, spatial sound pressure data of the target material is obtained. Using the trained model, the spatial sound pressure data of the target material can be output based on a limited number of measurement points. The complete sound field data corresponds to the distribution of each measurement point on the material surface. Therefore, the sound absorption coefficient and reflection coefficient of the material can be calculated based on the sound field data.

[0044] To verify the reliability of the method proposed in this invention, two methods are compared, as follows: Comparison Method 1: Use theoretical values ​​as a comparison. The theoretical values ​​are calculated based on the empirical model of porous media.

[0045] Comparison Method 2: The finite measurement point data is not mapped through a network, that is, the actual measured front-end finite measurement point data is used as the direct data for subsequent calculation of acoustic parameters.

[0046] During the verification process, a set of material samples that were not used in model training were selected for testing. The flow resistance of these samples was set to 106,000 Pa·s / m. 2 After inputting the sound pressure data captured from a limited number of measurement points into the trained network model, the model automatically outputs the corresponding sound pressure data captured from all measurement points across the entire domain.

[0047] Based on the output sound pressure data, the corresponding spatial spectrum information, reflection coefficient, and sound absorption coefficient information are calculated according to formulas (12-14), and the results are as follows. Figure 3 , Figure 4 and Figure 5 As shown.

[0048] Therefore, the method of this invention can map the limited measurement point data at the front end into a complete sound pressure distribution across the entire measurement area, accurately reflecting the spatial sound field of the material and calculating the precise sound absorption and reflection coefficients. Furthermore, this method eliminates the need for a large number of microphones, significantly reducing the complexity and cost of the experimental system while ensuring the accuracy and reliability of the acoustic parameter calculations.

[0049] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0050] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0051] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for measuring the sound absorption performance of a material based on spatial Fourier transform, characterized in that, The method comprises the following steps: acquiring acoustic data of limited measuring points of a target material; outputting spatial sound field data of the target material by using a fitting model based on the acoustic data of the limited measuring points; calculating the sound absorption coefficient and the reflection coefficient of the target material according to the output spatial sound field data; The fitting model is constructed by acquiring spatial acoustic data sets collected by local limited measuring points under different material properties and spatial acoustic data sets collected by global complete measuring points ; with the spatial acoustic data set as input to the model, with the spatial acoustic data set as labels to the model, the model is trained, and the optimal weights of the fitted model are obtained by minimizing the error between the fitting results and the corresponding labels, and based on the model weights, the sound pressure mapping relationship F between the local limited measurement points and the global complete measurement points is constructed, and the fitted model is obtained.

2. The method for measuring material sound absorption performance based on spatial Fourier transform according to claim 1, characterized in that, the step of acquiring the spatial acoustic data of the limited measuring points of the target material comprises the following steps: first, constructing a sound field model of the target material; then, setting a first measuring surface and a second measuring surface above the surface of the target material, the second measuring surface being set above the first measuring surface; then, setting a plurality of microphones along the same radial direction interval of the first measuring surface and the second measuring surface respectively; and finally, collecting acoustic data of corresponding positions by using the microphones.

3. The method for measuring material sound absorption performance based on spatial Fourier transform according to claim 2, characterized in that, In the step of constructing the fitting model, the properties of different materials are obtained by using a porous medium model, and the expression of the porous medium model is as follows: ； where and represent the characteristic impedance and complex wave number of the material, and represent the fluid density and sound speed in air, respectively, denotes the frequency, represents the flow resistance of the material, j is the imaginary unit.

4. The method for measuring material sound absorption performance based on spatial Fourier transform according to claim 3, characterized in that, The acquisition of spatial acoustic data sets under different material properties by local finite measurement points and by global complete measurement points comprises that the spatial acoustic data expression is ； in, Indicates the sound pressure at the measuring point. This indicates the distance between the sound source and the measuring point. Indicates at the angle of incidence The reflectance coefficient below, Indicates the wave number in the air. Indicates radial wavenumber, and These represent the heights of the sound source and the measuring surface, respectively. Represents the Bessel function. Indicates the radial length of the measured surface. and The characteristic impedance and complex wave number of the material are represented by j, which is the imaginary unit. Spatial acoustic dataset acquired by local finite measurement points and global complete measurement points are respectively: ； ； wherein, and P1and P2denote the set of sound pressures at different measurement points on the measurement surface 1 and 2, respectively, under a finite measurement point arrangement, and P1and P2denote the sound pressure at the nth measurement point on the measurement surface 1 and 2, respectively, under a finite measurement point arrangement; and P1and P2denote the set of sound pressures at different measurement points on the measurement surface 1 and 2, respectively, under a global complete measurement point arrangement, and P1and P2denote the sound pressure at the mth measurement point on the measurement surface 1 and 2, respectively, under a global complete measurement point arrangement.

5. A method of measuring the sound absorption properties of a material based on spatial Fourier transform according to claim 3 or 4, characterized in that, The step of training the fitting model comprises the following steps: constructing a deep learning model, and associating the input layer and the output of the next layer of the deep learning model by using a weight matrix and a bias matrix, wherein the deep learning model is composed of a plurality of layers, each layer performs linear transformation on input features by using a weight matrix and a bias vector and maps the features by using an activation function, and the recursive relationship between layers can be expressed as follows: ； wherein, represents an input variable, represents a next layer output variable, represents an activation function; The weight matrix comprises a low-rank matrix, a diagonal matrix, a learnable perturbation matrix and a basic weight matrix, and the expression of the weight matrix is as follows: ； where, denotes a low-rank matrix, which is reconstructed by a basis matrix and ; denotes a fixed diagonal matrix; is a learnable perturbation matrix; denotes a basis weight matrix; is the size of the convolution kernel, and denote the number of input and output channels, respectively; denotes Einstein summation, denotes the reconstruction of tensor dimensions.

6. The method for measuring material sound absorption performance based on spatial Fourier transform according to claim 5, characterized in that, In the step of obtaining the optimal weight of the fitting model by minimizing the error between the fitting result and the corresponding label, a loss function is constructed , and a physical loss function is incorporated into the loss function ; the difference between the model output and the corresponding label is calculated according to the loss function , and parameter updating is realized by the chain rule, and the loss function is incorporated; and the expression of the weight matrix is: ； wherein ; ； where the parameter Lphysics is a physical information loss function, the importance coefficient of the physical loss term, n and N represent the index of the measurement point and the number of measurement points, respectively; and represent the set of spatial sound field data of the model output and the corresponding label, respectively; and represent the reflection coefficient values calculated by the model output and the corresponding label, respectively; and represent the spatial spectral information of measurement surface 1 calculated by the model output and the corresponding label, respectively, and represent the spatial spectral information of measurement surface 1 calculated by the model output and the corresponding label, respectively; represents the transverse wave number, and represent the height of measurement surface 1 and measurement surface 2, respectively.

7. The method of measuring the sound absorption performance of a material based on spatial Fourier transform according to claim 6, characterized in that, The mapping relationship between the local limited measuring points and the global complete measuring points is as follows: ； wherein, and P1and P2denote the sound pressure values of the measurement points on the radial measurement lines of the measurement surface 1 and the measurement surface 2, respectively, and P1and P2denote the sound pressure values of the measurement points on the radial measurement lines of the measurement surface 1 and the measurement surface 2, respectively, 8. The method for measuring material sound absorption performance based on spatial Fourier transform according to claim 7, characterized in that, The step of outputting the spatial sound field data of the target material by using the fitting model comprises the following steps: performing two-dimensional Fourier transform on the x and y planes by using Fourier-Bessel transform, so as to obtain the spatial sound field information of the target material, and the specific expression is as follows: ； wherein, and represent global data of measurement points on different measurement lines output by the model, and represent corresponding spatial spectral information, represents a Bessel function, represents a radial length of the measurement surface, represents a radial wave number.

9. The method of measuring the sound absorption performance of a material based on spatial Fourier transform according to claim 8, characterized in that, The step of calculating the absorption coefficient and the reflection coefficient of the target material according to the output spatial sound field data comprises: calculating the reflection coefficient and the absorption coefficient The expression is: ； wherein the spatial sound field incident wave and the reflected wave are expressed as follows: ； wherein denotes the lateral wave number, and denotes the height of the measurement surface 1 and 2, respectively.

Images