A method for calculating acoustic pressure of a high-frequency transformer

Abstract

The application discloses a high-frequency transformer sound pressure calculation method, and relates to the technical field of circuits, which comprises the following steps: obtaining the microstructure parameters of a gap beryllium copper reed to generate a microstructure feature vector; based on the microstructure feature vector, calculating the magnetostrictive strain regulated by the microstructure to determine the coupling field distribution description of the magnetostrictive regulation; based on the coupling field distribution description, carrying out finite element grid division on the air domain around the high-frequency transformer to generate an air domain finite element grid; equivalent the gap beryllium copper reed to an acoustic impedance boundary condition, and defining a pseudo-density variable in the air domain finite element grid to determine the optimal sound wave propagation path tracking in the air domain; and based on the tracked sound wave propagation path, calculating the sound pressure of the high-frequency transformer. The application generates a feature vector by obtaining structure parameters, accurately calculates magnetostrictive strain, optimizes grid division, processes boundary conditions, and improves the calculation accuracy through multi-link cooperation, thereby providing a reliable basis for product design.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of circuit, in particular to a high-frequency transformer sound pressure calculation method. BACKGROUND

[0002] As a key device, the noise problem generated in the operation process of the high-frequency transformer is concerned. Accurate calculation of the sound pressure of the high-frequency transformer is of great significance for evaluating the noise level, optimizing the design and meeting the relevant noise standards.

[0003] At present, the existing high-frequency transformer sound pressure calculation scheme is mainly based on the combination of macroscopic electromagnetic parameters and acoustic empirical model. This scheme first obtains some macroscopic electromagnetic parameters of the high-frequency transformer through experimental measurement or theoretical calculation, such as the magnetic permeability of the core, the resistance and inductance of the winding, etc. These macroscopic parameters reflect the overall electromagnetic characteristics of the transformer, but ignore the complex influence of the internal electromagnetic-acoustic conversion process of the transformer. Specifically, in the actual high-frequency transformer, the structure (such as the gap beryllium copper reed) will have a significant effect on the magnetostriction effect, and then affect the distribution of electromagnetic force and the excitation and propagation of sound waves. Since the existing scheme does not consider these factors, there is a large error in the calculation of sound pressure, which cannot accurately reflect the real noise condition of the transformer, and it is difficult to meet the actual demand for increasing the control precision of the transformer noise. SUMMARY

[0004] In order to solve the above technical problems, the present application provides a high-frequency transformer sound pressure calculation method to at least alleviate the above technical problems.

[0005] The technical scheme provided by the embodiments of the present application is as follows:

[0006] A high-frequency transformer sound pressure calculation method comprises:

[0007] Step 1, obtaining the microstructure parameters of the gap beryllium copper reed to generate a microstructure feature vector;

[0008] Step 2, based on the microstructure feature vector, calculating the magnetostrictive strain regulated by the microstructure to determine the coupling field distribution description of the magnetostriction regulated by the microstructure;

[0009] Step 3, based on the coupling field distribution description, dividing the air domain around the high-frequency transformer into finite element grids to generate air domain finite element grids;

[0010] Step 4, equivalent the gap beryllium copper reed to acoustic impedance boundary condition, and define pseudo-density variable in the air domain finite element grid to determine the optimal sound wave propagation path tracking in the air domain;

[0011] Step 5, based on the tracked sound wave propagation path, calculating the sound pressure of the high-frequency transformer.

[0012] The high-frequency transformer sound pressure calculation method provided in the application has the following technical advantages:

[0013] The high-frequency transformer sound pressure calculation method provided in the application has the following technical advantages:

[0014] (1) By obtaining the microstructure parameters of the gap beryllium copper reed to generate a microstructure feature vector, the structural information of key components such as the gap beryllium copper reed is fully considered. These structural parameters can accurately describe the grain structure, defect distribution and other characteristics of the components, which are completely ignored in existing schemes. By incorporating these structural information into the calculation system, the physical state inside the transformer can be more realistically reflected, laying a foundation for subsequent accurate sound pressure calculation.

[0015] (2) Since the regulation of the magnetostrictive effect by the structure is considered, the size and distribution of the magnetostrictive strain under different structures can be accurately calculated. Compared with the existing scheme which only estimates the magnetostrictive effect based on macroscopic parameters, the accuracy of the calculation is greatly improved. The magnetostrictive effect is a key link in the electromagnetic-acoustic conversion process, and accurate calculation of the magnetostrictive strain helps to accurately determine the distribution of electromagnetic force, and then provides more reliable data support for the excitation of sound waves.

[0016] (3) Since the coupling field distribution description fully considers the magnetostrictive effect regulated by the structure, the grid can be more reasonably laid out when the air domain grid is divided. Compared with the relatively rough grid division method in the existing scheme, the finite element grid generated by the present scheme can more accurately simulate the propagation process of sound waves in the air domain, improving the calculation accuracy.

[0017] (4) The existing scheme often simplifies the treatment of transformer boundary conditions, without fully considering the complex effects of gap components and the like on sound wave propagation. The present scheme can more accurately simulate the reflection, refraction and absorption of sound waves by gap beryllium copper reed and the like components, accurately track the propagation path of sound waves in the air domain, and avoid errors in sound pressure calculation caused by inaccurate path calculation.

[0018] (5) The influence of the structure on the electromagnetic-acoustic conversion process is fully considered, the magnetostriction effect is accurately calculated, the air domain grid is reasonably divided, and the sound wave propagation path is accurately tracked, so that the finally calculated high-frequency transformer sound pressure result is closer to the actual measured value. Compared with the existing scheme, the present scheme can significantly improve the accuracy of sound pressure calculation, provide more reliable technical basis for noise evaluation and optimal design of high-frequency transformers, help to meet the increasingly stringent noise standard requirements, and improve the performance and quality of high-frequency transformers. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The flowchart of the high-frequency transformer sound pressure calculation method of the embodiments of the present application is shown. DETAILED DESCRIPTION

[0020] As shown in Figure 1 , the present application provides a high-frequency transformer sound pressure calculation method, which comprises:

[0021] Step 1, obtaining the microstructure parameters of the gap beryllium copper reed to generate a microstructure feature vector;

[0022] Step 2, based on the microstructure feature vector, calculating the magnetostriction strain regulated by the microstructure to determine the coupling field distribution description of the magnetostriction regulated by the microstructure;

[0023] Step 3, based on the coupling field distribution description, dividing the air domain around the high-frequency transformer into finite element grids to generate air domain finite element grids;

[0024] Step 4, equivalent the gap beryllium copper reed to an acoustic impedance boundary condition, and define a pseudo-density variable in the air domain finite element grid to determine the optimal sound wave propagation path tracking in the air domain;

[0025] Step 5, based on the tracked sound wave propagation path, calculating the sound pressure of the high-frequency transformer.

[0026] Optionally, step 1, obtaining the microstructure parameters of the gap beryllium copper reed to generate a microstructure feature vector, specifically includes the following steps:

[0027] Step 11, micron-level three-dimensional microstructure reconstruction is performed on the gap beryllium copper reed to generate a scanning electron microscope image and calculate the grain size distribution characteristics accordingly;

[0028] Step 12, in-situ stress testing of the gap beryllium copper reed to obtain grain boundary stress distribution and generate synchrotron X-ray diffraction data to extract grain boundary stress characteristics;

[0029] Step 13, collecting element distribution and crystal orientation data of the gap beryllium copper reed to generate electron backscatter diffraction data and generate misorientation distribution characteristics accordingly;

[0030] Step 14, establishing a probability correlation matrix between the grain size distribution characteristics, the grain boundary stress characteristics, and the misorientation distribution characteristics;

[0031] Step 15, eigenvalue decomposition of the probability correlation matrix to extract principal components with cumulative contribution rate greater than the contribution rate threshold and generate a microstructure feature vector accordingly.

[0032] Preferably, for step 11, the three-dimensional structure information of the internal grains of the gap beryllium copper reed is obtained by imaging technology. First, the reed is cut layer by layer using a focused ion beam (FIB), and the cutting thickness can be controlled to be in the micron level, while a scanning electron microscope (SEM) is used to scan and image each layer of cross section to obtain a high-resolution two-dimensional image sequence. Then, with the help of three-dimensional reconstruction algorithm, these two-dimensional images are spliced to construct a three-dimensional model of the internal grains of the reed. In this process, in order to accurately calculate the grain size distribution characteristics, each grain in the three-dimensional model needs to be identified and segmented, and the method based on image gradient and edge detection is usually used to determine the boundary of the grain. Then, by calculating the equivalent diameter (such as area equivalent circle diameter or volume equivalent sphere diameter) of each grain, the distribution of grain size is counted to obtain the grain size distribution characteristics, such as the average grain size, the standard deviation of grain size, etc.

[0033] Preferably, for step 12, the in-situ stress testing is to obtain the stress distribution of the internal grain boundaries of the reed under actual working conditions. The gap beryllium copper reed is installed on an in-situ tensile / compressive testing machine, and the structure changes of the grain boundary region are monitored in real time by SEM during the application of a certain stress. Through digital image correlation (DIC) technology, the SEM images are analyzed to calculate the strain at the grain boundary, and then the grain boundary stress is calculated according to the elastic modulus of the material using Hooke's law (σ=E·ε). At the same time, in order to more accurately obtain the grain boundary stress characteristics, synchrotron X-ray diffraction technology is used. The high-intensity and high-collimation X-ray beam generated by the synchrotron light source penetrates the reed and irradiates the grain boundary region, and the diffraction pattern is collected. According to the Bragg equation (2d·sinθ=nλ), the interplanar spacing information is extracted from the diffraction pattern, and then the lattice strain is calculated, and finally the grain boundary stress characteristics such as the distribution of stress concentration region and stress gradient are obtained.

[0034] σ = E · ε where σ: represents Stress, the force per unit area within a body, unit is Pascal (Pa), here specifically refers to the stress at the reed grain boundary. E: represents Elastic Modulus, also known as Young's Modulus, a physical quantity that measures the ability of a material to resist elastic deformation, unit is Pascal (Pa), its value is determined by the inherent properties of the material (such as the inherent properties of gap beryllium copper). ε: represents Strain, the relative deformation of an object under external force, is a dimensionless quantity (usually expressed in percentage or decimal), here refers to the strain at the grain boundary calculated by Digital Image Correlation (DIC) technology.

[0035] 2d · sin θ = nλ where d: represents Interplanar Spacing, the perpendicular distance between two adjacent parallel crystal planes in a crystal, unit is meter (m) or nanometer (nm), is an important parameter of crystal structure. θ: represents Bragg Angle, also known as grazing angle, is the angle between the incident X-ray beam and the crystal plane, unit is degree (°). n: represents Diffraction Order, is a positive integer (n = 1, 2, 3,...), represents the number of times X-ray is reflected between crystal planes. λ: represents Wavelength of incident X-ray, unit is meter (m) or nanometer (nm), determined by the characteristics of the synchrotron radiation source, here is the wavelength of the high collimation X-ray beam.

[0036] Preferably, for step 13, the spatial correlation and interactive analysis of element distribution and crystal orientation data need to be established. Using SEM equipped with EDS and EBSD combined system, the same area of gap beryllium copper reed is scanned synchronously: through EDS, the element distribution spectrum is obtained, and the composition difference of different micro areas (such as alloy element enrichment area or impurity phase distribution) is clear; at the same time, the reed is tilted to 70°, and the Kikuchi diffraction pattern of the area is collected by EBSD, and the grain orientation is preliminarily determined. Through data superposition technology, the element distribution characteristics and diffraction pattern are spatially matched - for example, identifying the element composition mutation area (which may correspond to the grain boundary or second phase), assisting the EBSD software to more accurately divide the grain boundary; for the composition uniform grain interior, combined with the influence of element species on crystal structure (such as the fine tuning of solid solution atoms on lattice parameters), the indexing algorithm of diffraction pattern is optimized, and the accuracy of crystal orientation determination is improved. Based on this interactive corrected EBSD data, when calculating the orientation difference of adjacent grains, the orientation difference characteristics of element segregation area (such as whether the element enriched grain boundary is more prone to form large angle orientation difference) can be focused on, and finally the orientation difference distribution characteristics containing composition-structure correlation information are obtained, such as the proportion difference of small angle grain boundary and large angle grain boundary in specific element distribution area, etc.

[0037] Preferably, for step 15, the numpy.linalg.eig function is used to perform eigenvalue decomposition on the probability association matrix, the eigenvalues and eigenvectors are solved by the QR iterative algorithm, the orthogonal matrix V composed of eigenvectors and the diagonal matrix Λ containing eigenvalues are obtained, and it is ensured that the eigenvalues are arranged in descending order. An eigenvalue accumulator is constructed, the proportion of each eigenvalue in the total sum of all eigenvalues is calculated in turn, and the cumulative sum is accumulated to generate a cumulative contribution rate sequence, while the change rate of adjacent contribution rates is calculated to identify the contribution rate mutation point. A dynamic contribution rate threshold interval [85%, 95%] is set, the optimal threshold is automatically determined according to the change rate of the cumulative contribution rate sequence, the first k eigenvalues whose cumulative contribution rate exceeds the dynamic contribution rate threshold interval are selected, and the corresponding eigenvectors are extracted as principal components to form a principal component matrix. Based on the eigenvalue size, an exponential decay weight function is constructed, each vector in the principal component matrix is weighted, a weighted principal component vector is generated by vector dot product operation, and the microstructure feature vector is output after L2 norm normalization.

[0038] Optionally, step 14, establishing a probability association matrix among the grain size distribution characteristics, the grain boundary stress characteristics, and the orientation difference distribution characteristics, specifically includes the following steps:

[0039] Step 141, taking the grain size distribution characteristics, the grain boundary stress characteristics, and the orientation difference distribution characteristics as input variables, constructing a joint probability distribution function of the three characteristics and generating a joint distribution model describing the dependence relationship among the characteristics according to the joint probability distribution function;

[0040] Step 142, according to the joint distribution model, calculating the Kendall rank correlation coefficient between the characteristics to determine the conditional probability distribution between the characteristics;

[0041] Step 143, mapping the probability association index between the characteristics in the conditional probability distribution to matrix elements to generate a probability association matrix.

[0042] Preferably, for step 141, the grain size distribution characteristics, the grain boundary stress characteristics, and the orientation difference distribution characteristics are taken as input variables to construct a joint probability distribution function among them. By collecting a large amount of sample data, the parameters of the joint distribution function such as the mean vector and the covariance matrix are estimated by using methods such as maximum likelihood estimation. In this way, a joint distribution model describing the dependence relationship among the three characteristics is obtained. This model can reflect the change trend and probability distribution of the other two characteristics when one of the characteristics changes, providing a basis for subsequent analysis of the correlation between the characteristics.

[0043] Preferably, for step 142, after obtaining the joint distribution model, the Kendall rank correlation coefficient is used to calculate the correlation between the grain size distribution characteristics, the grain boundary stress characteristics and the orientation difference distribution characteristics. The Kendall rank correlation coefficient is a non-parametric statistic that measures the correlation between variables based on the rank of the data. For each pair of features, such as grain size and grain boundary stress, the sample data are sorted, their ranks are calculated, and then the number of consistent pairs and inconsistent pairs is counted according to the calculation formula of the Kendall rank correlation coefficient to obtain the correlation coefficient between them. The value range of this correlation coefficient is between -1 and 1. The larger the absolute value, the stronger the correlation between the two features. By calculating the Kendall rank correlation coefficient between each pair of features, the conditional probability distribution between them can be determined, that is, the probability distribution of another feature when one feature takes a specific value.

[0044] Preferably, for step 143, the grain size distribution characteristics, grain boundary stress characteristics, and orientation difference distribution characteristics are preprocessed, the isolation forest algorithm is used to identify and eliminate outliers, the random forest regression model is used to fill missing values, and the Z-score normalization method is used to convert each feature data into a standard normal distribution with a mean of 0 and a standard deviation of 1 to obtain a standardized data set. Based on the standardized data set, the kendalltau function in the scipy.stats library is used to build a parallel computing framework to calculate the Kendall rank correlation coefficients between the grain size distribution characteristics and the grain boundary stress characteristics, the grain size distribution characteristics and the orientation difference distribution characteristics, and the grain boundary stress characteristics and the orientation difference distribution characteristics, respectively, to generate a set of correlation coefficient triplets. According to the symmetry principle of the Kendall rank correlation coefficient, a 3×3 symmetric matrix structure is constructed, and the elements in the correlation coefficient triplets are mapped to the corresponding non-diagonal positions of the matrix. At the same time, the matrix diagonal elements are uniformly assigned a value of 1, and finally a probability association matrix is ​​generated.

[0045] Optionally, step 2, calculating the magnetostrictive strain regulated by the microstructure based on the microstructure characteristic vector to determine a coupled field distribution description of the magnetostriction regulated by the microstructure, specifically comprises the following steps:

[0046] Step 21: using the microstructure characteristic vector as an input parameter and inputting it into the constructed microstructure-magnetostriction nonlinear constitutive model to calculate the magnetostrictive strain of the synergistic effect of grain size, grain boundary stress and misorientation;

[0047] Step 22: Generate a multi-physics field coupling interaction matrix based on the magnetostrictive strain;

[0048] Step 23: Perform spatiotemporal fusion processing on the multi-physics field coupling interaction matrix to determine the coupling field distribution description of the microstructure-controlled magnetostriction.

[0049] Optionally, in step 21, the microstructure feature vector is input into the constructed microstructure-magnetostrictive nonlinear constitutive model as an input parameter to calculate the magnetostrictive strain synergistically affected by the grain size, grain boundary stress and misorientation, specifically including the following steps:

[0050] Step 211, map the microstructure feature vector to the crystal elasticity theory framework to generate the microstructure-modified elastic constant matrix;

[0051] Step 212, input the microstructure-modified elastic constant matrix into the microstructure-magnetostrictive nonlinear constitutive model for thermodynamic potential function expansion to obtain the nonlinear magnetostrictive strain expression;

[0052] Step 213, generate the magnetostrictive strain tensor containing microstructure parameters based on the nonlinear magnetostrictive strain expression;

[0053] Step 214, perform multi-scale consistency correction and physical constraint optimization processing on the magnetostrictive strain tensor to calculate the magnetostrictive strain synergistically affected by the grain size, grain boundary stress and misorientation.

[0054] Preferably, for step 211, first decouple the microstructure feature vector, map the grain size distribution feature to the elastic modulus correction coefficient, convert the grain boundary stress feature to the shear modulus adjustment parameter, and generate the anisotropy tensor from the misorientation distribution feature. Based on the crystal periodic structure, the interatomic force is converted into macroscopic elastic parameters and the mechanical properties determined by the structure such as anisotropy, a crystal elasticity theory framework is constructed to generate a microstructure-sensitive elastic constant matrix model, and the anisotropy tensor is processed by multi-scale weighted fusion and physical constraint correction: combine the elastic modulus correction coefficient corresponding to the grain size to scale the principal component of the anisotropy tensor to obtain the corresponding shear component; introduce the shear modulus adjustment parameter associated with the grain boundary stress to dynamically correct the shear component to obtain the corrected shear component; at the same time, embed the crystal symmetry constraint condition to perform rational symmetry projection and thermodynamic stability constraint processing on the corrected tensor, so as to integrate the multi-dimensional microstructure features into the elastic constant matrix, and generate the elastic constant matrix containing the microstructure information of grain size, grain boundary stress and misorientation.

[0055] Preferably, in step 212, a microstructure-magnetostrictive nonlinear constitutive model is constructed based on the Landau free energy theory, and the microstructure-modified elastic constant matrix is taken as an input parameter of the model. In the model, the magnetocrystalline anisotropy energy, the elastic energy, and the magnetic-elastic coupling energy are introduced to construct a complete thermodynamic potential function. In view of the influence of the microstructure on the magnetic domain wall movement and the magnetization rotation, a microstructure-sensitive energy correction term is introduced to optimize the thermodynamic potential function. The optimized thermodynamic potential function is variational expanded, and the nonlinear magnetostrictive strain expression containing the microstructure parameters is derived by combining the Maxwell relationship and the constitutive equation. The nonlinear magnetostrictive strain expression can reflect the complex nonlinear relationship among the magnetic field, the stress, and the microstructure parameters.

[0056] Preferably, in step 213, the magnetic field strength, direction, and external stress conditions in the actual working condition are substituted into the nonlinear magnetostrictive strain expression to provide input parameters for subsequent calculation. Based on the input parameters, the finite element discretization method is used to divide the calculation domain into regular grids to obtain a discrete calculation model composed of multiple grid elements. For each grid element in the discrete model, the magnetostrictive strain components of the grid element are calculated according to its position and orientation and in combination with the local microstructure parameters in the microstructure feature vector. Based on the strain components, the three-dimensional strain tensor structure is constructed by considering the anisotropy characteristics of the material, and the strain components in each direction are filled into the corresponding positions of the tensor. The strain values at the grid boundaries in the three-dimensional strain tensor are processed using an interpolation algorithm to obtain a strain tensor with continuity and physical rationality. The processed strain tensor finally generates a magnetostrictive strain tensor containing spatial distribution information, which fully reflects the regulation of the microstructure parameters on the magnetostrictive strain.

[0057] Preferably, in step 214, a cross-scale correlation model is established to couple the grain structure and the macro-scale strain response to obtain a physically consistent strain tensor at different scales. For the strain tensor, a grain boundary constraint condition is introduced to reflect the hindering effect of the grain boundary on the grain deformation, and the strain components near the grain boundary are corrected. Based on the corrected strain components, a physical constraint optimization objective function is constructed based on the energy minimization principle, so that the function can reflect the constraints of the basic laws of elastic mechanics and the physical characteristics of the material on the tensor. The constraint conditions in the objective function are processed using the Lagrange multiplier method, and the optimal strain distribution is obtained by solving through an iterative optimization algorithm. The optimal strain distribution is further processed to finally obtain the magnetostrictive strain that has been modified for multi-scale consistency and optimized for physical constraints, and the accurate calculation of the synergistic effect of the grain size, the grain boundary stress, and the orientation difference is realized.

[0058] Optionally, in step 22, a multi-physical field coupling interaction matrix is generated based on the magnetostrictive strain, specifically including the following steps:

[0059] Step 221, deduce the permeability variation tensor caused by magnetostrictive strain according to the influence of magnetostrictive strain on magnetic field distribution;

[0060] Step 222, calculate the Joule heat and mechanical energy dissipation variation caused by magnetostrictive strain according to the coupling coefficient between stress component and temperature gradient, and generate stress-temperature field variation tensor according to the variation;

[0061] Step 223, perform tensor fusion and nonlinear correction on the permeability variation tensor and the stress-temperature field variation tensor to generate the multi-physical field coupling interaction matrix.

[0062] Preferably, for step 221, analyze the coupling relationship between magnetostrictive strain and magnetic field distribution, and determine the influence mechanism of internal magnetic domain structure rearrangement on magnetization characteristics when the material generates magnetostrictive strain. Based on this influence mechanism, a nonlinear mapping model between magnetostrictive strain and permeability is established based on magnetoelastic coupling theory. According to this nonlinear mapping model, the magnetic domain distribution variation under different magnetostrictive strains is simulated by finite element method, and the spatial distribution characteristics of magnetic domain orientation and magnetization intensity are extracted. For these spatial distribution characteristics, the magnetostrictive strain is decomposed into components in different directions, and the influence law of each component on permeability anisotropy is analyzed. Based on the micromagnetic theory and the influence law of each component on permeability anisotropy, the permeability tensor containing magnetostrictive strain effect is deduced, which can reflect the anisotropy variation of permeability caused by strain, and realize the quantitative description of the influence of magnetostrictive strain on magnetic field distribution.

[0063] Preferably, for step 222, consider the coupling effect of stress field and temperature field caused by magnetostrictive strain, first determine the coupling coefficient between stress component and temperature gradient, which is obtained by comprehensive experimental data and thermodynamic theory. Based on the coupling coefficient, analyze the physical phenomena when the material generates magnetostrictive strain: stress concentration is generated inside, and hysteresis loss and eddy current loss are converted into Joule heat. For these phenomena, an energy conservation equation is established to convert the mechanical work and magnetic energy dissipation generated by magnetostrictive strain into heat energy, and the temperature field distribution inside the material is calculated. Based on the temperature field distribution, the conduction process of temperature gradient in the material is simulated by finite element heat conduction analysis, and the influence of stress field on thermal conductivity coefficient is also taken into account. The variation of stress component and temperature gradient obtained by conduction analysis is taken as a tensor element, and a stress-temperature field variation tensor is constructed, which can reflect the coupling variation characteristics of stress field and temperature field caused by magnetostrictive strain.

[0064] Preferably, for step 223, a multi-physical field tensor fusion model is established to consider the nonlinear coupling relationship between the permeability change tensor and the stress-temperature field change tensor. The dimensions of the two tensors are unified and the coordinates are transformed to obtain tensors with comparability in the same physical space. The tensors with comparability are fused by tensor product operation and weighted summation method (the weight coefficients are calibrated by experimental data according to the coupling strength of each physical field), to obtain a preliminary fused tensor. For the preliminary fused tensor, a nonlinear correction term is introduced (to consider the nonlinear changes of physical fields caused by magnetostrictive effect, such as the saturation characteristics of permeability and the influence of temperature on material elastic modulus), to obtain a fused tensor with nonlinear factors. A nonlinear correction function is constructed to optimize the fused tensor with nonlinear factors, to obtain an optimized tensor that can accurately reflect the coupling interaction between multiple physical fields. The optimized tensor is generated as a multi-physical field coupling interaction matrix, which can comprehensively describe the interaction relationship between the magnetic field, stress field and temperature field caused by magnetostrictive strain.

[0065] Optionally, step 23, the multi-physical field coupling interaction matrix is subjected to space-time scale fusion processing to determine the coupling field distribution description of microstructure regulated magnetostriction, specifically including the following steps:

[0066] Step 231, the multi-physical field coupling interaction matrix is subjected to time scale decomposition, and the wavelet packet transform is adopted to decompose the matrix elements in the time dimension into dynamic components and steady components to obtain a time scale decomposition tensor;

[0067] Step 232, the time scale decomposition tensor is subjected to spatial fractal interpolation processing to implement non-uniform grid division on the interface region of the iron core and the gap beryllium copper reed to generate a spatial scale feature matrix;

[0068] Step 233, based on the spatial scale feature matrix, a three-dimensional coupling operator containing time-space-physical field is constructed to determine the coupling field distribution description of microstructure regulated magnetostriction.

[0069] Preferably, for step 231, the change characteristics of each element in the multi-physical field coupling interaction matrix over time are analyzed to determine the dynamic characteristic differences of magnetostrictive effect at different time scales. Based on these dynamic characteristic differences, a time scale decomposition model is established to provide a framework for subsequent time-frequency analysis. The multi-physical field coupling interaction matrix is incorporated into this model, and wavelet packet transform technology is used to perform time-frequency analysis on it, decomposing the time series signal of the matrix into sub-signals of different frequency bands to achieve multi-resolution analysis of the time characteristics of the matrix elements. For the time dimension of the matrix, a suitable wavelet basis function (such as Daubechies wavelet) is selected for layer-by-layer decomposition to separate high-frequency dynamic components (such as instantaneous coupling caused by rapid domain flipping) and low-frequency steady-state components (such as slow coupling caused by thermal accumulation). Based on each frequency sub-band obtained by decomposition, a time scale decomposition tensor is constructed, which retains the dynamic characteristics of physical field coupling at different time scales.

[0070] Preferably, for step 232, the physical field gradient characteristics (such as stress concentration and magnetic field distortion) of the interface region between the core and the beryllium copper spring sheet are analyzed to determine the necessity of fine spatial scale description in this region. The time scale decomposition tensor is used as the analysis object, and the fractal geometry theory is used to analyze its spatial distribution characteristics. By calculating the fractal dimension, the irregularity of the physical field in space is characterized. For the physical field parameters of the interface region, adaptive interpolation is performed based on the fractal interpolation algorithm - increasing the interpolation density in regions with large gradients (such as grain boundaries and contact surfaces) and reducing the density in non-critical regions to achieve non-uniform spatial scale description. The spatial distribution characteristics after interpolation are combined with the finite element mesh division principle, and a non-uniform mesh division strategy is used - using sub-micron fine mesh in the interface region and using regular mesh in the non-interface region to construct a spatial scale characteristic matrix. This spatial scale characteristic matrix establishes a correlation between mesh density and physical field gradient, achieving multi-resolution characterization in spatial scale, and accurately capturing the strong coupling effect in the interface region.

[0071] Preferably, for step 233, a three-dimensional coupling model of time-space-physical field is established based on the spatial scale feature matrix and the time scale decomposition tensor, the former is taken as the input parameter of spatial dimension, and the latter provides time dimension information, and a multi-scale coupling operator is constructed. For the multi-scale coupling operator, an interpolation kernel function in three-dimensional space is defined, and the physical field coupling parameters in different regions are weighted according to the grid density of the spatial scale feature matrix, so that the fine grid region obtains higher coupling weight. Taking the weighted physical field coupling parameters as the object, a time evolution operator is introduced, and the dynamic component and the steady component in the time scale decomposition tensor are associated with the spatial field distribution through convolution operation, and dynamic coupling in time dimension is realized. For the model after dynamic coupling, the interaction rules among magnetic field, stress field and temperature field (such as stress-magnetic field coupling caused by magnetostrictive strain and temperature-stress coupling caused by Joule heat) are designed, and these rules are embedded into the three-dimensional coupling operator to improve the description ability of the operator to the interaction of multi-physical field. Using the improved three-dimensional coupling operator, the multi-physical field coupling interaction matrix is transformed in time and space scales, the coupling field distribution description containing microstructure regulation effect is generated, and full-scale coupling characterization from time dynamics to spatial distribution is realized.

[0072] Optionally, step 3, based on the coupling field distribution description, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids, specifically including the following steps:

[0073] Step 31, based on the coupling field distribution description of microstructure regulation magnetostriction, the iron core vibration displacement gradient and the severe region of sound pressure fluctuation are extracted as key site features through the constructed air domain adaptive grid division criterion;

[0074] Step 32, based on the key site features, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids.

[0075] Optionally, step 31, based on the coupling field distribution description of microstructure regulation magnetostriction, the iron core vibration displacement gradient and the severe region of sound pressure fluctuation are extracted as key site features through the constructed air domain adaptive grid division criterion, specifically including the following steps:

[0076] Step 311, based on the physical field gradient energy density, a grid density distribution function is constructed as an air domain adaptive grid division criterion;

[0077] Step 312, according to the coupling field distribution description, the magnetic field gradient energy density, the stress gradient energy density and the sound pressure gradient energy density are calculated to form a comprehensive gradient energy density field;

[0078] Step 313, based on the grid density distribution function, the fractal features of the iron core surface and the air domain interface are extracted, and the fractal dimension is calculated.

[0079] Step 314, according to the fractal dimension and the comprehensive gradient energy density field, the core vibration displacement gradient and the sound pressure fluctuation region are calculated as the key position characteristics.

[0080] Preferably, for step 311, the gradient change rate of the magnetic field, stress field and sound pressure field is taken as the input parameter to establish the mapping relationship between the physical field gradient energy density and the grid density. Based on the mapping relationship, the influence weight of each physical field gradient energy density on the grid density is determined through experimental data calibration and numerical simulation verification. Using these influence weights, a nonlinear mapping function is designed to allocate higher grid density in the region where the physical field gradient changes sharply and lower grid density in the region where the physical field gradient changes gently. For this nonlinear mapping function, a regularization term is introduced to ensure its smoothness and avoid sudden changes in grid density. The processed nonlinear mapping function forms a grid density distribution function, which can automatically adjust the grid distribution according to the physical field characteristics to balance the calculation accuracy and efficiency.

[0081] Preferably, for step 312, the spatial distribution data of the magnetic field, stress field and sound pressure field are extracted based on the coupling field distribution description obtained in step 2. Spatial differentiation operation is performed on these spatial distribution data to calculate the gradient change rate of each physical field, and the square of the gradient amplitude is taken as the energy density index. According to the characteristics of different physical fields, a normalization processing method is designed to process the energy density index to eliminate the dimensional difference. The normalized gradient energy densities of the three physical fields are fused in a weighted sum manner (the weight coefficients are determined according to the contribution of each physical field to sound radiation) to form a comprehensive gradient energy density field.

[0082] Preferably, for step 313, based on the grid density distribution function, the interface geometry details of the core surface and the air domain interface are extracted by spatial scanning. Fractal analysis is performed on these details using the box counting method to count the complexity of the interface at different scales. The relationship between the box count at different scales and the scale is calculated, and the fractal dimension is obtained by linear regression fitting.

[0083] Optionally, step 314, according to the fractal dimension and the comprehensive gradient energy density field, the core vibration displacement gradient and the sound pressure fluctuation region are calculated as the key position characteristics, specifically including the following steps:

[0084] Step 3141, the fractal dimension is substituted into the grid density distribution function to obtain the grid density distribution based on the geometric complexity;

[0085] Step 3142, set the physical field change threshold in combination with the comprehensive gradient energy density field, establish a multi-dimensional correlation model, to carry out tensor fusion on the grid density distribution and the physical field change threshold to obtain a physical-geometric coupling tensor field;

[0086] Step 3143, carry out adaptive threshold segmentation on the physical-geometric coupling tensor field to extract the region where the iron core vibration displacement gradient exceeds the set gradient threshold and the intense region where the sound pressure fluctuation energy density is higher than the fluctuation energy threshold, so as to determine the key part features.

[0087] Preferably, for step 3141, the fractal dimension of the iron core surface and the air domain interface calculated in step 313 is taken as the geometric complexity index and input into the grid density distribution function constructed in step 311. The function outputs the corresponding grid density value according to the numerical size of the fractal dimension (the higher the fractal dimension, the more complex the geometric form of the region, and the more intense the physical field change), to generate the grid density distribution. Higher grid density is allocated to geometrically complex regions to capture physical field changes in detail, and relatively sparse grids are allocated to regions with lower fractal dimension and simple geometric form.

[0088] Preferably, for step 3142, the comprehensive gradient energy density field obtained in step 312 is statistically analyzed, and the physical field change threshold is set according to its data distribution characteristics (such as mean, standard deviation), to distinguish between regions with intense physical field changes and regions with gentle changes. With the physical field change threshold as a reference, a multi-dimensional correlation model is established to correlate the grid density distribution based on geometric complexity with the threshold. Through tensor operation, the grid density distribution and the physical field change threshold tensor are fused, so that the grid density of each spatial position is combined with the physical field change characteristics of the position. In the fusion process, higher weight is given to the region with intense physical field change to highlight its importance in grid division, and a physical-geometric coupling tensor field is formed. The physical-geometric coupling tensor field integrates the dual information of geometric form and physical field change, and can more comprehensively reflect the key regions in the air domain that need fine grid division.

[0089] Preferably, for step 3143, an adaptive threshold segmentation algorithm (such as Otsu algorithm) is used to process the physical-geometric coupling tensor field, which automatically calculates the optimal segmentation threshold according to the data distribution of the physical-geometric coupling tensor field, and divides the physical-geometric coupling tensor field into different regions. For the core vibration displacement gradient in these regions, compare the gradient value of each spatial position with the pre-set gradient threshold, and select the regions whose vibration displacement gradient exceeds the threshold (such regions have larger vibration due to magnetostrictive effect, and have significant influence on acoustic radiation). For the sound pressure fluctuation energy density in the region, compare the energy density value of each position with the fluctuation energy threshold, and extract the intense fluctuation region whose energy density is higher than the threshold. Through the above processing, the region with large core vibration displacement gradient and intense sound pressure fluctuation is accurately identified, which is determined as the key feature in the air domain grid division.

[0090] Optionally, step 32, based on the key feature, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids, specifically including the following steps:

[0091] Step 321, a grid density mapping model based on key feature is constructed to establish a nonlinear mapping relationship between physical field gradient energy density and grid size;

[0092] Step 322, based on the nonlinear mapping relationship, a boundary layer grid is generated at the interface between the core and the air domain to capture the energy transfer effect feature of structure vibration to sound field;

[0093] Step 323, based on the energy transfer effect feature, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids.

[0094] Preferably, for step 321, the internal relationship between the key feature and the physical field change is analyzed to determine the demand of the region with large core vibration displacement gradient and intense sound pressure fluctuation on the grid division accuracy. Simulation and experimental data under different physical field gradient energy densities are collected, and a nonlinear mapping model is constructed by using support vector regression (SVR) algorithm or artificial neural network (ANN) in machine learning. The physical field gradient energy density is used as the input of the model, and the target grid size is used as the output. The model is trained by a large number of sample data, so that it learns the complex mapping relationship between the degree of physical field change and the reasonable grid size. A regularization term is introduced into the model to avoid overfitting, so that it can accurately output the optimal grid density corresponding to the physical field gradient under different working conditions, providing a quantitative basis for subsequent grid division.

[0095] Preferably, for step 322, according to the nonlinear mapping relationship established in step 321, a progressive mesh refinement strategy is adopted for the key energy transfer region of the interface between the core and the air domain. Based on the energy density of the physical field gradient near the interface, the initial mesh size of the boundary layer is determined through the mapping relationship. Based on the initial mesh size, the boundary layer mesh is generated by expanding layer by layer into the air domain according to a preset growth rate (such as geometric progression), so that the mesh near the core surface is finer, so as to accurately capture the high-frequency response generated by structural vibration and the transient change when energy is transmitted to the sound field. The generated boundary layer mesh is quality evaluated and optimized by using a mesh quality checking algorithm (such as Jacobian determinant checking) to ensure the regularity of the shape of the mesh element and avoid calculation errors caused by mesh distortion, effectively improving the simulation accuracy of energy transfer effect.

[0096] Preferably, for step 323, starting from the boundary layer mesh, the front propagation method or the Delaunay triangulation algorithm is used for overall mesh division in combination with the physical field gradient energy density distribution in different regions of the air domain. For regions far from the core and with gentle physical field changes, according to the mapping relationship, sparse meshes are assigned to reduce the calculation amount on the premise of ensuring the calculation accuracy; for regions with physical field changes identified by key part features, fine meshes are generated according to the mapping relationship. During the division process, the matching degree of mesh density and physical field gradient is monitored in real time, and local mesh redivision technology is used to correct regions that do not meet the accuracy requirements. The finally generated air domain finite element mesh can accurately capture the sound propagation characteristics and energy transfer process caused by core vibration, and can reasonably control the number of meshes in non-key regions, achieving a balance between calculation efficiency and simulation accuracy.

[0097] Optionally, step 4, the gap beryllium copper reed is equivalent to an acoustic impedance boundary condition, and a pseudo-density variable is defined in the air domain finite element mesh to determine the optimal sound wave propagation path tracking in the air domain, which comprises the following steps:

[0098] Step 41, based on the influence of grain size distribution characteristics and grain boundary stress characteristics on the acoustic impedance of the material, the dynamic acoustic impedance of the gap beryllium copper reed is derived to equivalent the gap beryllium copper reed to an acoustic impedance boundary condition;

[0099] Step 42, according to the distance relationship between the air domain finite element mesh and the boundary of the gap beryllium copper reed, a pseudo-density variable is defined in the air domain finite element mesh based on the dynamic acoustic impedance to obtain the pseudo-density distribution characteristics;

[0100] Step 43, regarding each element of the air domain finite element mesh as a state space and the pseudo-density distribution characteristics as an environmental parameter, a reward function is defined with the minimum sound wave propagation loss as the target;

[0101] Step 44, according to the reward function, determine the optimal sound wave propagation path tracking in the air domain.

[0102] Preferably, for step 41, beryllium copper reed is taken as the research object, and the sound speed and density data under different grain sizes and grain boundary stress states are measured by experiment to establish a microstructure parameter-acoustic impedance database. Based on the database, combined with the theory of mechanics, the influence of grain size on the scattering effect of sound waves and the change of grain boundary stress on the elastic modulus and damping characteristics of the material are analyzed to clarify the mechanism of these factors on the acoustic impedance. Using the random forest regression algorithm in machine learning, the grain size distribution characteristics and grain boundary stress characteristics are taken as input variables to fit the nonlinear relationship model between acoustic impedance and microstructure parameters. For this model, a frequency domain correction factor is introduced to consider the dynamic effect of sound wave frequency on acoustic impedance, and a dynamic acoustic impedance expression is derived that can reflect the characteristics under different working conditions. The dynamic acoustic impedance expression equivalent to the gap beryllium copper reed has frequency and microstructure sensitivity.

[0103] Preferably, for step 42, a distance field is established to calculate the shortest distance from the center position of each element in the air domain finite element mesh to the gap beryllium copper reed boundary. Based on the distance field, combined with the dynamic acoustic impedance obtained in step 41, an attenuation function is designed so that the closer the grid element is to the boundary, the closer the pseudo-density variable is to the dynamic acoustic impedance of the reed; as the distance increases, the value of the pseudo-density variable gradually decreases according to the attenuation function until it approaches the acoustic impedance of air. Through this attenuation function, each element in the entire air domain finite element mesh is assigned a corresponding pseudo-density value, forming a pseudo-density distribution characteristic that reflects the boundary acoustic influence, providing environmental parameters for subsequent sound wave propagation path tracking.

[0104] Preferably, for step 44, a path search algorithm in reinforcement learning (such as Q-learning or deep Q network DQN) is used, taking the elements of the air domain finite element mesh as state nodes, and considering the propagation of sound waves between grid elements as a state transition process. In the algorithm initialization stage, a starting element is randomly selected as the starting point of the sound wave. For each state transition process, according to the pseudo-density distribution characteristics of the current element, combined with the reward function, the reward values of all adjacent elements are calculated, which reflect the size of the sound wave loss when propagating in that direction. The algorithm preferentially selects the adjacent element with the highest reward value (i.e., the smallest sound wave propagation loss) as the next state, realizing the gradual extension of the path. Through multiple iterations of training, the algorithm continuously optimizes the path selection strategy and gradually converges to the optimal propagation path from the starting point to the target point. In the training process, the experience replay mechanism and target network update strategy are introduced to avoid the algorithm falling into a local optimal solution, and finally the optimal path of the sound wave propagating in the air domain considering the influence of the gap beryllium copper reed is obtained, which is the optimal sound wave propagation path.

[0105] Optionally, step 43, taking each element of the air domain finite element mesh as a state space, and taking the pseudo-density distribution characteristics as environmental parameters, to define a reward function aiming at minimizing the sound wave propagation loss, specifically including the following steps:

[0106] Step 431, taking each element of the air domain finite element mesh as a state space, and taking the pseudo-density distribution characteristics as environmental parameters, to calculate the sound wave energy attenuation per unit distance in each element of the air domain finite element mesh;

[0107] Step 432, according to the sound wave energy attenuation per unit distance in each element of the air domain finite element mesh, determining the sound wave propagation direction, the acoustic impedance difference between elements, and the pseudo-density gradient;

[0108] Step 433, according to the sound wave propagation direction, the acoustic impedance difference between elements, and the pseudo-density gradient, defining a reward function aiming at minimizing the sound wave propagation loss.

[0109] Preferably, for step 431, taking each element of the air domain finite element mesh as an independent analysis element, and extracting its pseudo-density value as an environmental parameter, which can reflect the acoustic influence intensity of the gap beryllium copper reed boundary on the air domain (the closer to the boundary, the higher the pseudo-density value of the element, and the greater the sound wave propagation loss). Based on the acoustic energy attenuation theory, the sound wave attenuation rate under different pseudo-density values is calibrated through experiments to form a lookup table or an interpolation model, establishing a mapping relationship between the pseudo-density and the sound wave attenuation coefficient. The pseudo-density value of each element is substituted into the mapping model to calculate the sound wave energy attenuation coefficient per unit distance in the element, which represents the energy loss proportion per 1 meter of propagation when the sound wave passes through the element.

[0110] Preferably, for step 432, taking the current element as the center, analyzing its spatial positional relationship with adjacent elements to determine the possible sound wave propagation directions (such as 6 adjacent directions in three-dimensional space). For each propagation direction, calculate the pseudo-density difference value between the current element and the adjacent element, which reflects the "acoustic impedance difference" between the elements (the more drastic the pseudo-density change, the greater the sound wave reflection and scattering loss). Through finite difference method, calculate the pseudo-density gradient of the current element in each propagation direction, and the greater the absolute value of the gradient, the more significant the acoustic environment change in that direction. Normalize the three types of parameters: propagation direction, pseudo-density difference value (i.e. equivalent acoustic impedance difference), and pseudo-density gradient, to obtain parameters with uniform dimensions. Based on these parameters with uniform dimensions, form a multi-dimensional feature vector to quantitatively describe the acoustic loss characteristics of each propagation direction.

[0111] Preferably, for step 433, the propagation direction feature vector obtained in step 432 is used as the basis to design a nonlinear reward function. The acoustic impedance difference between units and the pseudo-density gradient are taken as negative reward factors (the greater the difference, the lower the reward value), and the energy attenuation coefficient in the propagation direction is taken as the core penalty term. The function form is constructed in the form of weighted summation: reward value = -a x (acoustic impedance difference) -b x (pseudo-density gradient) -g x (energy attenuation per unit distance), where a, b, g are weight coefficients calibrated through experimental data, reflecting the priority of the influence of each factor on sound wave loss. The function value is standardized to make the reward value distributed in a reasonable interval (such as [-1, 0]), and the smaller the negative value, the smaller the sound wave loss in the propagation direction, and the higher the reward value. The reward function formed finally can guide the algorithm to preferentially select the propagation path with gentle acoustic impedance change, small pseudo-density gradient and low energy attenuation.

[0112] Optionally, step 5, based on the tracked sound wave propagation path, calculates the sound pressure of the high-frequency transformer, specifically including the following steps:

[0113] Step 51, discretize the sound wave propagation path into a series of acoustic propagation units to calculate the sound energy transmission coefficient between adjacent acoustic propagation units and generate a path energy attenuation coefficient matrix accordingly;

[0114] Step 52, based on the path energy attenuation coefficient matrix, determine the sound pressure distribution considering the multi-path interference effect;

[0115] Step 53, based on the coupling field distribution description of the microstructure regulating magnetostriction, extract the vibration velocity distribution of the surface of the gap beryllium copper reed;

[0116] Step 54, perform acoustic-structure coupling boundary integral processing on the sound pressure distribution of the multi-path interference effect and the vibration velocity distribution of the surface of the gap beryllium copper reed to calculate the sound pressure of the high-frequency transformer.

[0117] Preferably, for step 51, the tracked continuous sound wave propagation path is discretized according to spatial position, and the path is divided into a series of connected acoustic propagation units according to the size and shape of the air domain finite element grid, each unit being regarded as an independent acoustic wave propagation medium, the characteristics of which are determined by the physical parameters such as air density and sound speed at the location. For adjacent acoustic propagation units, the energy transmission characteristics of sound waves at the unit interface are analyzed based on acoustic interface transmission theory. Due to the difference in acoustic impedance between different units, reflection and transmission phenomena occur when sound waves propagate to the interface. By calculating the acoustic impedance ratio of adjacent units, combining Snell's law and acoustic boundary conditions, the sound energy transmission coefficient of sound waves at the interface is determined, which represents the proportion of energy that can continue to propagate forward when the sound wave crosses the interface. The higher the transmission coefficient, the smaller the energy loss. Fill all the sound energy transmission coefficients between adjacent units in the matrix according to the order of the propagation path to form a path energy attenuation coefficient matrix. The rows and columns of the matrix correspond to different acoustic propagation units, respectively. The elements on the diagonal line are 1 by default (indicating no energy attenuation inside the unit), and the non-diagonal elements record the sound energy transmission coefficients between adjacent units.

[0118] Optionally, step 52, based on the path energy attenuation coefficient matrix, determine the sound pressure distribution considering the multi-path interference effect, specifically including the following steps:

[0119] Step 521, based on the physical parameters of each acoustic wave propagation unit in the path energy attenuation coefficient matrix, calculate the propagation time of the sound wave on the sound wave propagation path to determine the phase delay and generate a phase delay matrix accordingly;

[0120] Step 522, according to the path energy attenuation coefficient matrix, calculate the sound pressure attenuation factor and perform tensor product operation with the phase delay matrix to obtain the sound pressure description data;

[0121] Step 523, take each grid node in the air domain finite element grid as a sound wave receiving point to generate a sound pressure distribution matrix considering the multi-path interference effect based on the sound pressure description data.

[0122] Preferably, for step 521, based on the path energy attenuation coefficient matrix, the physical parameters corresponding to each acoustic propagation unit in the matrix (including air density, sound speed and other information within the unit) are extracted, which directly affect the propagation speed of sound waves in the unit (the faster the sound speed, the shorter the time required for the sound wave to cross the unit). The length of each unit is divided by the corresponding sound speed to calculate the propagation time of the sound wave in each unit. The propagation times of each unit along the path are sequentially added to obtain the total propagation time of the sound wave from the starting point to the end point. According to the relationship between time and phase in the wave equation (the phase change of the sound wave is directly related to its propagation time), the total propagation time is converted into a phase delay, which reflects the phase change of the sound wave due to time accumulation during propagation, and is a key parameter for analyzing the multi-path interference effect. The phase delay corresponding to each propagation unit is filled into the corresponding position of the matrix to generate a phase delay matrix.

[0123] Preferably, for step 522, the path energy attenuation coefficient matrix is taken as the analysis object, and the physical relationship between energy and sound pressure (the more the energy attenuation, the more obvious the decrease in sound pressure amplitude) is used to convert the sound energy transmission coefficient between each unit in the matrix into a corresponding sound pressure attenuation factor, which represents the degree of attenuation of the sound pressure amplitude after the sound wave passes through a specific unit or interface relative to the initial value. For these sound pressure attenuation factors, perform a tensor product operation with the phase delay matrix generated in step 521 to fuse the sound pressure attenuation information and the phase delay information through the tensor product, and simultaneously consider the influence of amplitude attenuation and phase change on the sound pressure. Through this operation, the sound pressure attenuation factor on each propagation path is combined with the corresponding phase delay to generate sound pressure description data containing amplitude and phase information (the data is in the form of a complex number, the real part corresponds to the sound pressure amplitude, and the imaginary part corresponds to the phase information).

[0124] Optionally, step 523, taking each grid node in the air domain finite element grid as a sound wave receiving point, generates a sound pressure distribution matrix considering the multi-path interference effect based on the sound pressure description data, specifically including the following steps:

[0125] Step 5231, traverse all grid nodes in the air domain finite element grid, and take each grid node as a sound wave receiving point;

[0126] Step 5232, calculate the path length difference according to the spatial position relationship between the sound wave receiving point and the sound wave propagation path, and then obtain the additional phase difference between the paths;

[0127] Step 5233, coherently superimpose the additional phase difference and the sound pressure description data to generate a sound pressure distribution matrix considering the multi-path interference effect.

[0128] Optionally, step 53, based on the coupling field distribution description of the microstructure regulating magnetostriction, extract the vibration velocity distribution of the surface of the gap beryllium copper reed.

[0129] Step 531, based on the microstructure-regulated magnetostriction coupling field distribution description, the grain orientation difference distribution and the grain boundary stress concentration coefficient of the gap beryllium copper reed are determined;

[0130] Step 532, based on the grain orientation difference distribution and the grain boundary stress concentration coefficient, the representative volume element model is used to determine the vibration response of the gap beryllium copper reed;

[0131] Step 533, based on the vibration response, the dynamic strain energy density of each point on the surface of the reed is calculated to extract the vibration velocity distribution on the surface of the gap beryllium copper reed.

[0132] Preferably, for step 531, based on the microstructure-regulated magnetostriction coupling field distribution description, the crystallographic orientation data of each grain is obtained by scanning with material characterization techniques such as electron backscatter diffraction (EBSD) for the gap beryllium copper reed. By comparing the crystallographic orientation data of adjacent grains, the grain orientation difference is calculated, and then the grain orientation difference distribution map of the entire reed is formed, which can intuitively reflect the irregularity of the grain arrangement. For the calculation of the grain boundary stress concentration coefficient, a model is constructed using finite element analysis method, and the microstructure characteristics (such as grain size, grain boundary shape) and magnetostriction effect are introduced into the model. Based on the model, the stress distribution inside the reed under the action of magnetic field is simulated, and the grain boundary area is focused on (because the atomic arrangement is irregular at the grain boundary, the stress generated by magnetostriction is easy to concentrate here). Through stress analysis of the grain boundary area, the stress concentration degree is calculated, and the grain boundary stress concentration coefficient is obtained to quantify the amplification effect of the grain boundary on the stress.

[0133] Preferably, for step 532, the representative volume element (RVE) model is selected, which contains enough grains and grain boundaries to accurately reflect the structural characteristics of the reed. The grain orientation difference distribution and the grain boundary stress concentration coefficient obtained in step 531 are assigned to the RVE model as input parameters, and a multi-physical field coupling equation is established by combining the magnetostriction effect and the electromagnetic force action. The multi-physical field coupling equation is solved by numerical calculation method to simulate the deformation and vibration caused by magnetostriction inside the RVE model under the excitation of high-frequency magnetic field. The dynamics response of the model is analyzed to obtain the displacement, velocity and acceleration of the RVE at different times. The vibration parameters of the RVE model are statistically averaged and extrapolated, and the results are mapped to the entire gap beryllium copper reed to determine the overall vibration response of the reed.

[0134] Preferably, for step 533, the reed surface is divided into multiple tiny regions according to the reed vibration response obtained in step 532. For each tiny region, the strain of the region is calculated according to the displacement change in the vibration process through geometric relations and elastic mechanics theory. Based on the strain result, the dynamic strain energy density of each region (which reflects the energy stored in unit volume in the vibration process) is calculated in combination with parameters such as the elastic modulus of the material. Based on the energy conservation principle and the dynamics equation, the conversion relationship between the dynamic strain energy density and the vibration velocity is established by using the inherent relationship between the dynamic strain energy density and the vibration velocity. Through the conversion relationship, the dynamic strain energy density of each region is converted into the corresponding vibration velocity. The vibration velocity of the discrete region is extended to the entire reed surface by using the spatial interpolation method, and a continuous and accurate vibration velocity distribution is generated.

[0135] Optionally, step 54, the sound pressure distribution of the multi-path interference effect and the vibration velocity distribution of the gap beryllium copper reed surface are subjected to sound-solid coupling boundary integral processing to calculate the sound pressure of the high-frequency transformer, specifically including the following steps:

[0136] Step 541, the vibration velocity distribution is introduced as an acoustic boundary condition into a three-dimensional free space Green function to establish a frequency domain sound pressure integral expression;

[0137] Step 542, based on the frequency domain sound pressure integral expression, the sound pressure distribution of the multi-path interference effect is subjected to sound-solid coupling boundary integral processing to generate a sound-solid coupling correction term;

[0138] Step 543, the sound-solid coupling correction term and the sound pressure distribution are subjected to complex superposition processing to obtain the sound pressure field of the high-frequency transformer.

[0139] Preferably, for step 541, the vibration velocity distribution is defined as an acoustic boundary condition to clearly define the continuity relationship between the reed surface vibration velocity and the normal velocity of the air domain particle. Based on this continuity relationship, a three-dimensional free space Green function is introduced for the air domain sound field to represent the sound pressure response characteristics of a unit point sound source at any position in space. By combining the Green function with the acoustic boundary condition, the sound pressure contributions of each vibration point on the reed surface are superimposed through integral operation to obtain the quantitative relationship between the sound pressure of any field point in the air domain and the vibration velocity distribution of the reed. The quantitative relationship is presented in integral form, forming a frequency domain sound pressure integral expression describing the sound field radiated by the entire vibrating surface.

[0140] Preferably, for step 542, after the frequency domain sound pressure integral expression is established, the sound pressure distribution considering the multipath interference effect obtained in step 52 is combined to perform the sound-structure coupling boundary integral processing. The multipath interference effect reflects the phenomenon of sound waves propagating along different paths and superimposing on each other in the air domain, and the sound-structure coupling considers the influence of the reed vibration on the sound field. In specific operation, the sound pressure distribution of the multipath interference effect is taken as the initial sound field, and on this basis, the additional sound pressure contribution caused by the reed vibration is calculated using the frequency domain sound pressure integral expression in step 541. Through the boundary integral method, the vibration velocity information on the surface of the reed is coupled with the initial sound field to quantify the correction effect of vibration on the sound field. This process considers the interaction between vibration and sound wave propagation, and finally generates a sound-structure coupling correction term, which reflects the adjustment of the sound-structure coupling effect on the original sound pressure distribution.

[0141] Preferably, for step 543, since the sound pressure distribution and the sound-structure coupling correction term are both represented in complex form in the frequency domain (the real part represents the sound pressure amplitude, and the imaginary part represents the phase information), in order to obtain the final sound pressure field, complex superposition processing is required. Complex superposition can consider the changes in amplitude and phase at the same time, and accurately reflect the sound pressure distribution under the joint action of the sound-structure coupling effect and the multipath interference effect. In specific implementation, the sound-structure coupling correction term generated in step 542 is added point by point to the sound pressure distribution considering the multipath interference effect obtained in step 52. For each node in the air domain finite element mesh, the complex number of the correction term at the corresponding position is added to the complex number of the original sound pressure distribution to obtain a new complex sound pressure value. Through this operation, the dual influence of vibration radiation and sound wave interference is integrated, and finally the high-frequency transformer sound pressure field containing the sound-structure coupling effect is obtained, realizing more accurate simulation of the actual sound field.

[0142] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various changes and modifications to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of calculating acoustic pressure of a high-frequency transformer, characterized by, The method comprises the following steps: Step 1, obtaining the microstructure parameters of the gap beryllium copper reed to generate a microstructure feature vector; Step 2, based on the microstructure feature vector, calculating the magnetostriction strain regulated by the microstructure to determine the coupling field distribution description of the magnetostriction regulated by the microstructure; Step 3, based on the coupling field distribution description, carrying out finite element grid division on the air domain around the high-frequency transformer to generate an air domain finite element grid; Step 4, equivalent the gap beryllium copper reed to an acoustic impedance boundary condition, and defining a pseudo-density variable in the air domain finite element grid to determine the optimal sound wave propagation path tracking in the air domain; Step 5, based on the tracked sound wave propagation path, calculating the sound pressure of the high-frequency transformer.

2. The method of claim 1, wherein, Step 1 specifically comprises the following steps: Step 11, micron-level three-dimensional microstructure reconstruction is performed on the gap beryllium copper reed to generate a scanning electron microscope image and calculate grain size distribution characteristics based on the scanning electron microscope image; Step 12, in-situ stress testing is performed on the gap beryllium copper reed to obtain grain boundary stress distribution and generate synchrotron X-ray diffraction data based on the grain boundary stress distribution to extract grain boundary stress characteristics; Step 13, collecting element distribution and crystal orientation data of the gap beryllium copper reed to generate electron backscatter diffraction data and generate orientation difference distribution characteristics based on the electron backscatter diffraction data; Step 14, establishing a probability correlation matrix among the grain size distribution characteristics, the grain boundary stress characteristics, and the orientation difference distribution characteristics; Step 15, performing eigenvalue decomposition on the probability correlation matrix to extract principal components with cumulative contribution rate greater than a contribution rate threshold and generate a microstructure feature vector based on the principal components.

3. The method of claim 1, wherein, Step 2 specifically comprises the following steps: Step 21, inputting the microstructure feature vector as an input parameter into a constructed microstructure-magnetostriction nonlinear constitutive model to calculate the magnetostriction strain of the grain size, the grain boundary stress, and the orientation difference synergistic effect; Step 22, generating a multi-physical field coupling interaction matrix based on the magnetostriction strain; Step 23, performing time-space scale fusion processing on the multi-physical field coupling interaction matrix to determine the coupling field distribution description of the magnetostriction regulated by the microstructure.

4. The method of claim 3, wherein, Step 21 specifically comprises the following steps: Step 211, mapping the microstructure feature vector to a crystal elasticity theory framework to generate a microstructure-modified elastic constant matrix; Step 212, inputting the microstructure-modified elastic constant matrix into the microstructure-magnetostriction nonlinear constitutive model to perform thermodynamic potential function expansion to obtain a nonlinear magnetostriction strain expression; Step 213, generating a magnetostriction strain tensor containing microstructure parameters based on the nonlinear magnetostriction strain expression; Step 214, performing multi-scale consistency correction and physical constraint optimization processing on the magnetostriction strain tensor to calculate the magnetostriction strain of the grain size, the grain boundary stress, and the orientation difference synergistic effect.

5. The method of claim 3, wherein, Step 22 specifically comprises the following steps: Step 221, deducing a permeability change tensor caused by the magnetostriction strain according to the influence of the magnetostriction strain on the magnetic field distribution; Step 222, calculating the change of the mechanical energy dissipation and the Joule heat generated by the magnetostriction strain according to the coupling coefficient of the stress component and the temperature gradient and generating a stress-temperature field change tensor based on the change; Step 223, tensor fusion and nonlinear correction are performed on the permeability change tensor and the stress-temperature field change tensor to generate a multi-physical field coupling interaction matrix.

6. The method of claim 1, wherein, Step 3 specifically includes the following steps: Step 31, based on the constructed air domain adaptive grid division criterion, the key site features of the iron core vibration displacement gradient and the sound pressure fluctuation region are extracted based on the coupling field distribution description of microstructure regulated magnetostriction; Step 32, based on the key site features, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids.

7. The method of claim 6, wherein, Step 31 specifically includes the following steps: Step 311, based on the physical field gradient energy density, a grid density distribution function is constructed as an air domain adaptive grid division criterion; Step 312, according to the coupling field distribution description, the magnetic field gradient energy density, the stress gradient energy density, and the sound pressure gradient energy density are calculated to form a comprehensive gradient energy density field; Step 313, based on the grid density distribution function, the fractal features of the iron core surface and the air domain interface are extracted, and the fractal dimension is calculated; Step 314, according to the fractal dimension and the comprehensive gradient energy density field, the key site features of the iron core vibration displacement gradient and the sound pressure fluctuation region are calculated.

8. The method of claim 6, wherein, Step 32 specifically includes the following steps: Step 321, a grid density mapping model based on key site features is constructed to establish a nonlinear mapping relationship between physical field gradient energy density and grid size; Step 322, based on the nonlinear mapping relationship, a boundary layer grid is generated at the interface between the iron core and the air domain to capture the energy transfer effect feature of structure vibration to sound field; Step 323, based on the energy transfer effect feature, the air domain around the high-frequency transformer is divided into finite element grids to generate air domain finite element grids.

9. The method of claim 2, wherein, Step 4 specifically includes the following steps: Step 41, based on the influence of grain size distribution characteristics and grain boundary stress characteristics on material acoustic impedance, the dynamic acoustic impedance of the gap beryllium copper reed is derived to equivalent the gap beryllium copper reed as an acoustic impedance boundary condition; Step 42, according to the distance relationship between the air domain finite element grid and the boundary of the gap beryllium copper reed, a pseudo-density variable is defined in the air domain finite element grid based on the dynamic acoustic impedance to obtain the pseudo-density distribution feature; Step 43, each element of the air domain finite element grid is regarded as a state space, and the pseudo-density distribution feature is regarded as an environmental parameter to define a reward function with the minimum sound wave propagation loss as the target; Step 44, according to the reward function, the optimal sound wave propagation path tracking in the air domain is determined.

10. The method of claim 1, wherein, Step 5 specifically includes the following steps: Step 51, the sound wave propagation path is discretized into a series of acoustic propagation units to calculate the sound energy transmission coefficient between adjacent acoustic propagation units and generate a path energy attenuation coefficient matrix accordingly; Step 52, based on the path energy attenuation coefficient matrix, the sound pressure distribution considering the multi-path interference effect is determined; Step 53, based on the coupling field distribution description of microstructure regulated magnetostriction, the vibration velocity distribution on the surface of the gap beryllium copper reed is extracted; Step 54, the sound pressure distribution of the multi-path interference effect and the vibration velocity distribution of the gap beryllium copper reed surface are subjected to sound-solid coupling boundary integral processing to calculate the sound pressure of the high-frequency transformer.

Images