Construction method of carbon fiber composite material effective dielectric constant model

Abstract

The invention relates to the technical field of model construction, in particular to a method for constructing an effective dielectric constant model of a carbon fiber composite material. The method comprises the following steps: collecting a sample, carrying out electromagnetic orthogonal frequency sweep analysis, generating a spectrum distribution map, obtaining internal field distribution mapping data through boundary topology recombination, carrying out interval correlation analysis on the internal field data, generating a spatial distribution function, and calibrating a phase resonance point to obtain a frequency characteristic spectrum. Performing field region classification and interactive coupling evaluation according to the frequency characteristic spectrum, deducing a node link structure, performing internal and external field interactive tensor projection and constraint embedding processing based on the node link structure, constructing a composite field model, generating a dielectric frequency dispersion curve through dielectric characteristic inversion, and performing modeling based on frequency dispersion curve data to obtain an effective dielectric constant model. According to the invention, the more accurate construction method of the carbon fiber composite material effective dielectric constant model is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and particularly to a method for constructing an equivalent dielectric constant model of a carbon fiber composite material. Background Art

[0002] Due to the complex internal structure and non-uniform electromagnetic properties of carbon fiber composite materials, there are still defects in the accurate modeling of their dielectric behavior. In the existing technology for characterizing the electromagnetic properties of carbon fiber composite materials, it often relies on simplified theoretical models or extrapolation of experimental data, and it is difficult to accurately reflect the complex electromagnetic response of the material in the actual working environment. Especially under high-frequency conditions, the dielectric properties of the material will fluctuate significantly with the change of frequency, and the existing methods are difficult to capture this frequency-varying characteristic, resulting in insufficient model accuracy. In current research, there is a lack of systematic analysis of the distribution and interaction mechanism of the internal field of carbon fiber composite materials. Traditional methods usually adopt a single-scale modeling method and cannot comprehensively describe the multi-scale electromagnetic behavior of the material. In addition, when dealing with the coupling effect of the internal field of the material, the existing technology often ignores the influence of the node-link structure on the electromagnetic field distribution, resulting in a large deviation between the model prediction result and the actual measurement value. Summary of the Invention

[0003] Based on this, it is necessary to provide a method for constructing an equivalent dielectric constant model of a carbon fiber composite material to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for constructing an equivalent dielectric constant model of a carbon fiber composite material includes the following steps:

[0005] Step S1: Collect a carbon fiber composite material sample; perform electromagnetic orthogonal sweep frequency analysis on the carbon fiber composite material sample to obtain a spectrum distribution map; perform boundary topology recombination on the spectrum distribution map to generate internal field distribution mapping data;

[0006] Step S2: Perform spacing correlation analysis on the internal field distribution mapping data to generate a spatial distribution function; perform phase resonance point calibration based on the spatial distribution function to obtain a frequency characteristic spectrum;

[0007] Step S3: Perform field region classification mapping according to the frequency characteristic spectrum, and perform interaction coupling evaluation to generate field coupling strength data; determine the node-link structure based on the field coupling strength data;

[0008] Step S4: Perform internal and external field interaction tensor projection based on the node-link structure to generate multi-scale field projection data; perform constraint embedding processing on the multi-scale field projection data based on the node-link structure, and perform composite field modeling to obtain a composite field model;

[0009] Step S5: Perform dielectric property inversion on the composite field model to generate a dielectric dispersion curve; perform data modeling based on the nodal dispersion curve to obtain an equivalent dielectric constant model.

[0010] Through the electromagnetic orthogonal sweep frequency analysis and the boundary topology recombination of the spectral distribution map, the present invention accurately extracts the internal field distribution mapping data of the carbon fiber composite material, ensuring the high resolution and accuracy of the data, providing a reliable basis for subsequent analysis. Based on the spacing correlation analysis and the generation of the spatial distribution function of the internal field distribution mapping data, the refined characterization of the internal structure of the material is realized. Further, through the calibration of the phase resonance point and the generation of the frequency characteristic spectrum, the electromagnetic response characteristics of the material can be accurately captured, providing a key basis for the field region classification mapping and the interactive coupling evaluation. The generation of the field coupling strength data and the inference of the nodal link structure reveal the interaction mechanism of the internal field of the material, laying a theoretical foundation for the construction of the multi-scale field model. Through the internal and external field interaction tensor projection and the constrained embedding process of the multi-scale field model, the multi-dimensional modeling of the electromagnetic properties of the material is realized, improving the adaptability and accuracy of the model. The dielectric property inversion of the composite field model and the generation of the dielectric dispersion curve directly reflect the dielectric behavior of the material, providing data support for the construction of the equivalent dielectric constant model. Finally, the equivalent dielectric constant model generated through data modeling can accurately describe the dielectric properties of the carbon fiber composite material at different frequencies.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: Collect a carbon fiber composite material sample; perform multi-frequency electromagnetic scanning on the carbon fiber composite material sample to obtain original electromagnetic response data;

[0013] Step S12: Perform Fourier domain transformation on the original electromagnetic response data to obtain an electromagnetic response spectrum; perform spectral distribution analysis on the electromagnetic response spectrum to generate a spectral distribution map;

[0014] Step S13: Perform adaptive filtering on the spectral distribution map and detect the spectral edge to obtain a boundary feature profile;

[0015] Step S14: Perform topological segmentation on the carbon fiber composite material sample based on the boundary feature profile to generate a double-interface marker map; extract the edge microstructure description data of the double-interface marker map;

[0016] Step S15: Perform spatial reconstruction according to the edge microstructure description data to obtain internal field distribution mapping data.

[0017] Through multi - frequency electromagnetic scanning and the acquisition of original electromagnetic response data, the present invention can comprehensively cover the electromagnetic characteristics of carbon fiber composite materials at different frequencies, providing a high - precision data basis for subsequent analysis. The combination of Fourier - domain transformation and spectral distribution analysis significantly improves the resolution and analysis ability of the electromagnetic response spectrum, enabling the spectral distribution map to more accurately reflect the frequency - domain characteristics of the material. The application of adaptive filtering processing and spectral edge detection effectively removes noise interference and extracts key boundary feature contours, providing clear input data for topological segmentation. The topological segmentation based on the boundary feature contours and the generation of a double - interface marking map achieve an accurate division of the internal structure of carbon fiber composite materials. The further extracted edge microstructure description data provides key structural information for spatial reconstruction. The spatial reconstruction performed through the edge microstructure description data can faithfully restore the field distribution characteristics inside the material, generating high - resolution internal field distribution mapping data.

[0018] Preferably, step S14 includes the following steps:

[0019] Step S141: Continuously fit the curvature of the boundary feature contour to obtain a local curvature transformation matrix;

[0020] Step S142: Apply a partition - constraint mapping to the curvature mutation points according to the local curvature transformation matrix to generate an initial topological grid;

[0021] Step S143: Correct the connectivity of the initial topological grid to obtain a corrected topological grid; apply multi - scale interface cross - filtering to the corrected topological grid and construct a matrix to generate an interface coherence enhancement matrix;

[0022] Step S144: Based on the interface coherence enhancement matrix, perform topological segmentation on the carbon fiber composite material sample to generate a double - interface marking map; extract the edge microstructure description data of the double - interface marking map.

[0023] Through continuous curvature fitting of the boundary feature contour and generation of the local curvature transformation matrix, the present invention can accurately capture the geometric features of the edge of the carbon fiber composite material, provide high-precision curvature information for topological segmentation, realize effective control of the curvature mutation points based on the partition constraint mapping of the local curvature transformation matrix and generation of the initial topological grid, ensure the rationality and continuity of the topological structure, significantly improve the accuracy and stability of the topological grid by applying the connectivity correction of the modified topological grid and multi-scale interface cross-filtering, further optimize the coherence of the interface segmentation by generating the interface coherence enhancement matrix, realize the accurate division of the internal structure of the carbon fiber composite material based on the topological segmentation of the interface coherence enhancement matrix and generation of the double interface marker map, and the extracted edge microstructure description data provides key structural information for subsequent spatial reconstruction and field distribution mapping, thus laying a high-precision data foundation for the construction of the equivalent dielectric constant model.

[0024] Preferably, step S2 includes the following steps:

[0025] Step S21: Perform multi-frequency point sampling based on the internal field distribution mapping data to obtain a frequency response matrix; perform wavelength eigen-decomposition on the frequency response matrix to generate eigen-frequency data;

[0026] Step S22: Perform fiber spacing correlation analysis on the eigen-frequency data to generate a spacing-frequency relationship diagram; perform parameter fitting on the spacing-frequency relationship diagram, where the fitting parameters are set at 0.1 mm - 5 mm, to generate a spatial distribution function;

[0027] Step S23: Detect the extreme points of the spatial distribution function and perform frequency domain transformation mapping on the extreme points of the spatial distribution function to generate resonance frequency parameters;

[0028] Step S24: Apply phase reference compensation to the resonance frequency parameters to obtain calibrated frequency parameters; perform quantitative calibration of the phase resonance points based on the spatial distribution function and the calibrated frequency parameters to obtain a frequency characteristic spectrum, where the selection range of the resonance points is 1 MHz - 10 GHz, and fine scanning is performed in steps of less than 10 kHz.

[0029] Through multi-frequency point sampling based on internal field distribution mapping data and the generation of a frequency response matrix, the present invention can comprehensively cover the electromagnetic response characteristics of carbon fiber composite materials at different frequencies, providing high-resolution frequency-domain data for subsequent analysis. The generation of wavelength feature decomposition and characteristic frequency data significantly improves the analytical ability of the frequency response matrix, enabling the accurate extraction of key frequency characteristics. The fiber spacing correlation analysis and the generation of the spacing-frequency relationship diagram reveal the quantitative relationship between the internal fiber spacing and the electromagnetic response of the material, providing a key basis for the construction of the spatial distribution function. The application of parameter fitting technology enables the efficient conversion of the spacing-frequency relationship diagram into a spatial distribution function, further improving the accuracy and applicability of the model. The combination of the detection of the extreme points of the spatial distribution function and the frequency-domain transformation mapping can accurately capture the resonance characteristics of the material, generating high-precision resonance frequency parameters. The introduction of phase reference compensation effectively corrects the phase error of the resonance frequency parameters, ensuring the accuracy of the calibrated frequency parameters. Based on the quantitative calibration of the phase resonance points of the spatial distribution function and the calibrated frequency parameters, a fine scan can be performed within the range of 1 MHz - 10 GHz with a step size of less than 10 kHz, generating a high-resolution frequency characteristic spectrum.

[0030] Preferably, the field region classification mapping based on the frequency characteristic spectrum and the interactive coupling evaluation in step S3 include:

[0031] Performing feature dimensionality reduction on the frequency characteristic spectrum to generate field region principal component data;

[0032] Performing clustering analysis on the field region principal component data to obtain field region categories;

[0033] Constructing a field region identifier based on the field region categories;

[0034] Performing internal and external field region mapping based on the field region identifier to obtain field region classification data;

[0035] Constructing a field strength gradient field for the field region classification data; performing polarization vector analysis on the field strength gradient field to obtain polarization field description data;

[0036] Performing interactive coupling evaluation based on the polarization field description data to obtain field coupling strength data.

[0037] Through the feature dimensionality reduction of the frequency characteristic spectrum and the generation of field region principal component data, the present invention can effectively extract key electromagnetic features, significantly reduce the data complexity and improve the analysis efficiency. The clustering analysis and the generation of field region categories realize the precise division of the internal field regions of the carbon fiber composite material, providing a clear classification basis for the subsequent field region mapping. The field region identifiers constructed based on the field region categories can efficiently identify the electromagnetic characteristics of different field regions, further improving the accuracy of field region classification. The internal and external field region mapping and the generation of field region classification data comprehensively reflect the distribution characteristics of the internal field of the material, laying a data foundation for the construction of the field strength gradient field. The combination of the construction of the field strength gradient field and the polarization vector analysis can accurately describe the polarization characteristics of the internal field of the material, generating high-precision polarization field description data. The interactive coupling evaluation based on the polarization field description data can quantitatively analyze the interaction mechanism of the internal field of the material, generating highly reliable field coupling strength data.

[0038] Preferably, the determination of the node link structure based on the field coupling strength data in step S3 includes:

[0039] Reconstruct the neighborhood relationship according to the field coupling strength data to obtain the node neighborhood relationship;

[0040] Quantify the link connection strength based on the node neighborhood relationship to generate link strength quantization data;

[0041] Perform hierarchical clustering division on the link strength quantization data to obtain a hierarchical clustering network;

[0042] Based on the hierarchical clustering network, screen the association paths of the node neighborhood relationship to obtain the key node connection links;

[0043] Reshape the link structure of the key node connection links based on the link strength quantization data to generate a node link structure.

[0044] Through the reconstruction of the neighborhood relationship and the generation of the node neighborhood relationship for the field coupling strength data, the present invention can accurately describe the interaction mode of the internal field of the carbon fiber composite material, providing clear neighborhood information for the quantization of the link connection strength. The generation of the link connection strength quantization data realizes the quantitative characterization of the connection strength between nodes, significantly improving the accuracy of the node link structure analysis. The hierarchical clustering division and the construction of the hierarchical clustering network can effectively identify the hierarchical distribution characteristics of the internal field of the material, providing structured data support for the screening of the key node connection links. The association path screening based on the hierarchical clustering network can efficiently extract the key node connection links, further optimizing the analysis efficiency of the node link structure. The reshaping of the link structure of the key node connection links by the link strength quantization data realizes the high-precision reconstruction of the node link structure, generating a highly reliable node link structure.

[0045] Preferably, the tensor projection for internal and external field interaction based on the node-link structure in step S4 includes:

[0046] Performing tensor eigen-decomposition on the node-link structure to obtain node tensor features;

[0047] Performing field space weight assignment on the node tensor features according to preset spatial weight parameters to obtain field space weight data;

[0048] Performing tensor field interaction analysis based on the field space weight data to generate tensor field interaction relationships;

[0049] Performing projection space reconstruction on the tensor field interaction relationships to obtain internal and external field projection data;

[0050] Performing multi-scale feature fusion on the internal and external field projection data to generate multi-scale field features;

[0051] Performing tensor reconstruction based on the multi-scale field features to generate a multi-scale field model.

[0052] Through the tensor eigen-decomposition of the node-link structure and the generation of node tensor features in the present invention, the multi-dimensional features of the internal field of carbon fiber composites can be accurately extracted, providing high-precision input data for field space weight assignment. The field space weight assignment based on preset spatial weight parameters realizes the spatial optimization of node tensor features, significantly improving the accuracy of tensor field interaction analysis. The generation of tensor field interaction relationships comprehensively describes the interaction mode of the internal field of the material, providing key relationship data for projection space reconstruction. The generation of internal and external field projection data realizes the high-precision mapping of the internal and external field distributions of the material, further optimizing the input quality of multi-scale feature fusion. The application of multi-scale feature fusion technology can effectively integrate field features at different scales, generating high-resolution field feature data, providing reliable data support for the construction of a multi-scale field model. The tensor reconstruction based on multi-scale field features realizes the multi-dimensional modeling of the electromagnetic characteristics of the material, generating a high-precision multi-scale field model.

[0053] Preferably, the constraint embedding process for multi-scale field projection data based on the node-link structure in step S4 and the composite field modeling include:

[0054] Extracting the topological features of the node-link structure;

[0055] Performing hierarchical coupling deconstruction on the topological features to obtain multi-level connection data;

[0056] Performing boundary gradient detection on the multi-level connection data to obtain a structure constraint boundary;

[0057] Performing manifold embedding transformation on the structure constraint boundary to generate constraint manifold data;

[0058] Perform constrained embedding processing on the multi-scale field projection data based on the constrained manifold data, and perform composite field modeling to obtain a composite field model.

[0059] By extracting the topological features of the node-link structure, the present invention can accurately describe the connection mode of the internal field of the carbon fiber composite material, provide high-precision input data for hierarchical coupling deconstruction, and generate hierarchical coupling deconstruction and multi-level connection data, realizing multi-scale analysis of topological features, significantly improving the detection accuracy of the structural constraint boundary. The application of the boundary gradient detection technology can efficiently identify the structural constraint boundary of the internal field of the material, provide clear boundary data for manifold embedding transformation. The manifold embedding transformation based on the structural constraint boundary generates high-precision constrained manifold data, further optimizing the constrained embedding processing of the multi-scale field model. The constrained embedding processing of the multi-scale field model by the constrained manifold data realizes high-fidelity modeling of the electromagnetic properties of the material and generates a highly reliable composite field model.

[0060] Preferably, the dielectric property inversion of the composite field model in step S5 includes:

[0061] Perform frequency grid division on the composite field model to obtain frequency grid data;

[0062] Perform logarithmic interval optimization on the frequency grid data, set 200 sampling points in the range of 10 MHz - 100 GHz to obtain optimized frequency point data;

[0063] Initialize the anisotropic parameters of the optimized frequency point data, set the initial dielectric constant matrix to obtain the initial dielectric parameters;

[0064] Perform gradient descent inversion on the initial dielectric parameters to generate a dielectric gradient iteration tensor;

[0065] Perform principal axis transformation on the dielectric gradient iteration tensor to obtain the principal axis dielectric parameters;

[0066] Constrain the dielectric loss coefficient of the principal axis dielectric parameters and limit the dielectric loss coefficient constraint within the range of 0.001 - 0.1 to obtain the loss-constrained dielectric data;

[0067] Perform frequency domain correlation on the loss-constrained dielectric data to generate a complex dielectric parameter distribution;

[0068] Perform dispersion mapping processing on the complex dielectric parameter distribution and perform curve fitting processing to obtain the dielectric dispersion curve.

[0069] Through frequency grid division and logarithmic interval optimization of the composite field model, the present invention can efficiently set 200 sampling points in the range of 10 MHz - 100 GHz, significantly improving the coverage and accuracy of frequency point data. The initialization of anisotropic parameters and the setting of the initial dielectric constant matrix provide high-precision initial conditions for dielectric property inversion. The gradient descent inversion and the generation of dielectric gradient iterative tensors achieve efficient optimization of dielectric parameters, further improving the accuracy of the inversion results. The principal axis transformation processing and the generation of principal axis dielectric parameters can accurately capture the principal axis direction characteristics of the material's dielectric properties, providing key data support for loss constraints. The constraint of the dielectric loss coefficient tanδ and the generation of loss constraint dielectric data ensure that the dielectric parameters fluctuate within a reasonable range, significantly improving the reliability of the model. The frequency domain correlation and the generation of complex dielectric parameter distributions comprehensively reflect the dielectric behavior of the material at different frequencies, providing high-precision input data for dispersion mapping processing. The dispersion mapping processing and the generation of dielectric dispersion curves can accurately describe the frequency-varying dielectric properties of carbon fiber composites.

[0070] Preferably, the data modeling based on the nodal dispersion curve in step S5 includes:

[0071] Extracting the key nodes of the dielectric dispersion curve;

[0072] Performing linear interpolation on the key nodes, setting the dielectric constant range to be constrained within 2 - 12, and plotting the dispersion curve to obtain a normalized nodal dispersion curve;

[0073] Performing polynomial expansion projection based on the normalized nodal dispersion curve to generate polynomial coefficients;

[0074] Implementing regularization constraints on the polynomial coefficients to obtain a regularized equivalent model;

[0075] Reconstructing the frequency response function based on the regularized equivalent model to generate equivalent dielectric response data;

[0076] Performing model compression on the equivalent dielectric response data to obtain an equivalent dielectric constant model.

[0077] By extracting the key nodes of the dielectric dispersion curve and removing outliers, the present invention can effectively remove noise interference, significantly improve the data quality of the node dispersion curve. The imposition of the dielectric constant range constraint from 2 to 12 and the generation of the normalized node dispersion curve ensure that the data fluctuates within a reasonable range, further improving the reliability of the model. The polynomial expansion projection based on the normalized node dispersion curve and the generation of polynomial coefficients achieve a high-precision fitting of the dispersion curve, providing key data support for subsequent regularization constraints. The generation of regularization constraints and the regularization equivalent model can effectively prevent overfitting and significantly improve the generalization ability of the model. The reconstruction of the frequency response function based on the regularization equivalent model and the generation of equivalent dielectric response data comprehensively reflect the dielectric behavior of the material at different frequencies, providing high-precision input data for model compression. The application of model compression technology can efficiently simplify the complexity of the equivalent dielectric response data and generate a high-precision equivalent dielectric constant model. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a schematic flow chart of the steps of a method for constructing an equivalent dielectric constant model of a carbon fiber composite material;

[0079] Figure 2 It is a schematic detailed implementation step flow chart of step S2;

[0080] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art within the scope of the present invention without creative work based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0082] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0083] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0084] To achieve the above object, please refer to Figures 1 to 2 , a method for constructing an equivalent dielectric constant model of a carbon fiber composite material, comprising the following steps:

[0085] Step S1: Collect carbon fiber composite material samples; perform electromagnetic orthogonal sweep frequency analysis on the carbon fiber composite material samples to obtain a spectrum distribution map; perform boundary topology recombination on the spectrum distribution map to generate internal field distribution mapping data;

[0086] Step S2: Perform spacing correlation analysis on the internal field distribution mapping data to generate a spatial distribution function; perform phase resonance point calibration based on the spatial distribution function to obtain a frequency characteristic spectrum;

[0087] Step S3: Perform field region classification mapping according to the frequency characteristic spectrum, and perform interactive coupling evaluation to generate field coupling strength data; determine the node link structure based on the field coupling strength data;

[0088] Step S4: Perform internal and external field interaction tensor projection based on the node link structure to generate multi-scale field projection data; perform constrained embedding processing on the multi-scale field projection data based on the node link structure, and perform composite field modeling to obtain a composite field model;

[0089] Step S5: Perform dielectric property inversion on the composite field model to generate a dielectric dispersion curve; perform data modeling based on the node dispersion curve to obtain an equivalent dielectric constant model.

[0090] Through electromagnetic orthogonal swept-frequency analysis and boundary topological recombination of the spectral distribution map, the present invention accurately extracts the internal field distribution mapping data of carbon fiber composite materials, ensuring the high resolution and accuracy of the data, providing a reliable basis for subsequent analysis. Based on the spacing correlation analysis and spatial distribution function generation of the internal field distribution mapping data, the refined characterization of the internal structure of the material is realized. Further, through the calibration of phase resonance points and the generation of frequency characteristic spectra, the electromagnetic response characteristics of the material can be accurately captured, providing a key basis for field region classification mapping and interaction coupling evaluation. The generation of field coupling strength data and the inference of node link structures reveal the interaction mechanism of the internal fields of the material, laying a theoretical foundation for the construction of multi-scale field models. Through the projection of the internal and external field interaction tensors and the constrained embedding process of the multi-scale field model, the multi-dimensional modeling of the electromagnetic characteristics of the material is realized, improving the adaptability and accuracy of the model. The inversion of the dielectric properties of the composite field model and the generation of the dielectric dispersion curve directly reflect the dielectric behavior of the material, providing data support for the construction of the equivalent dielectric constant model. Finally, the equivalent dielectric constant model generated through data modeling can accurately describe the dielectric properties of carbon fiber composite materials at different frequencies.

[0091] In an embodiment of the present invention, the method for constructing the equivalent dielectric constant model of the carbon fiber composite material includes the following steps:

[0092] Step S1: Collect a carbon fiber composite material sample; perform electromagnetic orthogonal swept-frequency analysis on the carbon fiber composite material sample to obtain a spectral distribution map; perform boundary topological recombination on the spectral distribution map to generate internal field distribution mapping data;

[0093] In this embodiment, a representative composite material is selected for cutting to ensure that the size of the sample is 10 cm × 10 cm × 1 cm. The surface of the sample is cleaned and ensured to have no significant surface contaminants. A high-precision electromagnetic wave detector (such as a microwave vector network analyzer) is used to perform electromagnetic orthogonal swept-frequency analysis on the sample. The detection frequency range is from 1 GHz to 20 GHz, and the data acquisition accuracy is 0.01 dB. A spectral distribution map is generated in real time through computer software. Analyze the signal intensity distribution at different frequency points in the map. The range of the map covers various electromagnetic response signals from low frequency to high frequency. Then, perform boundary topological recombination on the spectral distribution map, identify the regions with large signal changes, and calculate the corresponding internal field distribution mapping data based on the obtained map data. Use an algorithm to convert the mapping data into a spatial field distribution pattern, and determine the electromagnetic wave propagation path and signal intensity between each characteristic point.

[0094] Step S2: Perform spacing correlation analysis on the internal field distribution mapping data to generate a spatial distribution function; perform phase resonance point calibration based on the spatial distribution function to obtain a frequency characteristic spectrum;

[0095] In this embodiment, the generated internal field distribution mapping data is zonally divided in terms of spatial distribution. The K-means clustering algorithm is used to group the data, with the number of clusters set to 4. The boundaries of the field distribution regions are determined by calculating the shortest distances between data points. Then, correlation analysis of the distance intervals is performed on each group of data, the distances between each data point are calculated, and a spatial distribution function is generated based on these data. The distribution trend and variation law of the field are analyzed using this function. On this basis, the phase resonance points are calibrated. During calibration, precise calculations are carried out according to the electromagnetic characteristics of known materials to determine the optimal matching points of the phase, and a frequency characteristic spectrum is obtained. The frequency range is set from 1 GHz to 15 GHz, and the data accuracy is 0.001 GHz, obtaining a series of frequency characteristic data.

[0096] Step S3: Perform field region classification mapping based on the frequency characteristic spectrum, and conduct cross-coupling evaluation to generate field coupling strength data; determine the node link structure based on the field coupling strength data;

[0097] In this embodiment, based on the frequency characteristic spectrum, classification mapping of the field regions is performed. The support vector machine (SVM) method is used to classify the frequency characteristics. Multiple field regions are divided according to the calibrated phase resonance points, and cross-coupling evaluation is conducted according to the different characteristics of the field regions. During evaluation, the transmission matrix method is used to calculate the coupling strength between the field regions, obtaining the field coupling strength data, which indicate the electromagnetic field coupling relationship between different regions. Then, the node link structure is inferred based on the field coupling strength data, and the connection relationship between different nodes is determined using the shortest path algorithm in graph theory, forming a preliminary node connection network.

[0098] Step S4: Perform internal and external field interaction tensor projection based on the node link structure to generate multi-scale field projection data; perform constrained embedding processing on the multi-scale field projection data based on the node link structure, and conduct composite field modeling to obtain a composite field model;

[0099] In this embodiment, when performing internal and external field interaction tensor projection, first, based on the previously obtained node link structure, the electromagnetic field interaction between nodes is modeled. The finite element method (FEM) is used to numerically solve the model. The node link relationship is transformed into a high-dimensional interaction tensor and projected to obtain the projection result of the internal and external field interaction tensor, and a multi-scale field model is generated. Then, constrained embedding processing is performed on this multi-scale field model based on the node link structure. During the constraint process, additional electromagnetic boundary conditions are added, and the layout of the node link structure is optimized using an iterative algorithm to ensure the adaptability of the multi-scale field model at different scales, finally obtaining a composite field model.

[0100] Step S5: Perform dielectric property inversion on the composite field model to generate a dielectric dispersion curve; perform data modeling based on the nodal dispersion curve to obtain an equivalent dielectric constant model.

[0101] In this embodiment, when performing dielectric property inversion on the composite field model, first, by measuring the reflectivity and transmittance data of the composite material, the electromagnetic wave response of the composite material at different frequencies is obtained. Then, the dielectric property inversion is performed using the reflectivity and transmittance data, and an inversion algorithm such as the Genetic Algorithm (GA for short) is used to optimize the inversion process to calculate the dielectric dispersion curve. The dispersion curve reflects the variation law of the dielectric properties of the material with frequency. Based on the nodal dispersion curve, data modeling is performed, and the polynomial fitting method is used to smooth the curve, and finally, an equivalent dielectric constant model is obtained.

[0102] Preferably, step S1 includes the following steps:

[0103] Step S11: Collect carbon fiber composite material samples; perform multi-frequency electromagnetic scanning on the carbon fiber composite material samples to obtain original electromagnetic response data;

[0104] Step S12: Perform Fourier domain transformation on the original electromagnetic response data to obtain an electromagnetic response spectrum; perform spectral distribution analysis on the electromagnetic response spectrum to generate a spectral distribution map;

[0105] Step S13: Perform adaptive filtering on the spectral distribution map and detect the spectral edge to obtain a boundary feature profile;

[0106] Step S14: Perform topological segmentation on the carbon fiber composite material samples based on the boundary feature profile to generate a double-interface labeled map; extract the edge microstructure description data of the double-interface labeled map;

[0107] Step S15: Perform spatial reconstruction according to the edge microstructure description data to obtain internal field distribution mapping data.

[0108] In this embodiment, a carbon fiber composite material that is uniform and has no significant defects is selected from a representative material batch. The sample size is 10 cm × 10 cm × 1 cm. Multi-frequency electromagnetic scanning is performed using a high-frequency electromagnetic wave detection device. The frequency range of the device is set from 1 GHz to 20 GHz. During the scanning process, the scanning step size is set to 0.1 GHz. Each time a scan is performed, a vector network analyzer (VNA) is used to measure the sample to obtain the original electromagnetic response data, which includes the reflection coefficient and transmission coefficient data of the sample at different frequencies. The original electromagnetic response data is converted from time-domain data to frequency-domain data using the fast Fourier transform (FFT) algorithm for Fourier transformation, converting the time signal into a frequency signal to obtain the electromagnetic response spectrum. This spectrum covers the signal intensities at each frequency point in the range from 1 GHz to 20 GHz. Each frequency point in the spectrum data reflects the electromagnetic response characteristics of the sample at that frequency. The conversion result is a set of spectrum data, which contains the electromagnetic characteristics of the composite material sample at different frequencies. When performing spectrum distribution analysis on the electromagnetic response spectrum, spectral analysis methods are used to process the spectrum data. First, the spectrum data is smoothed to reduce the influence of noise. A low-pass filter is used to filter out high-frequency noise, and the cut-off frequency is set to 18 GHz. Then, the smoothed spectrum data is subjected to distribution analysis to identify the main frequency and secondary frequencies in the spectrum. By analyzing the intensity peaks in the spectrum diagram, the frequency response characteristics of the sample in different frequency bands are identified, and a spectrum distribution map is generated. The map indicates the response intensity at each frequency point and the characteristic changes of the sample in each frequency band. When performing adaptive filtering on the spectrum distribution map, an adaptive filtering algorithm is used to process the low-frequency and high-frequency noise in the spectrum map. The weighted average method is used to smooth the spectrum map, and the weight of the adaptive filter is set to the inverse of the frequency. After filtering, smoothed spectrum data is obtained. Then, spectral edge detection is performed on the filtered spectrum. The Canny edge detection algorithm is used to identify the edge features in the spectrum diagram, generating a spectral edge feature map. The edge features in the map are clearly marked, and these edge features correspond to significant electromagnetic wave response changes of the carbon fiber composite material at different frequencies. By analyzing the edge feature contours, two main interfaces of the sample are identified. The sample is divided into regions using an image segmentation algorithm, and the level set method is used to track and segment the boundaries. The interface of the material is divided into an inner part and an outer part, generating a double-interface marking map. The map indicates the positions of the inner and outer interfaces of the composite material, and each interface in the double-interface marking map is described in detail. The microstructural features at the double interface are extracted. When extracting, an edge detection algorithm is used to identify the microscopic morphology on the boundary, and microscopic structure description parameters are extracted through image processing techniques. The extracted microstructural data is used as input,The edge microstructure is spatially reconstructed using a three-dimensional reconstruction algorithm (such as Delaunay triangulation). During the reconstruction process, the internal and external interface characteristics of the sample are considered, and the data points are distributed in three-dimensional space to construct an internal field distribution model of the sample. The spatial distribution map is generated using the reconstructed data, and through this map, the electromagnetic field distribution of the composite material in different regions can be intuitively represented, and finally, the internal field distribution mapping data of the carbon fiber composite material is obtained.

[0109] Preferably, step S14 includes the following steps:

[0110] Step S141: Continuously fit the boundary feature contour to obtain a local curvature transformation matrix;

[0111] Step S142: Apply a partition constraint mapping to the curvature mutation points according to the local curvature transformation matrix to generate an initial topological grid;

[0112] Step S143: Correct the connectivity of the initial topological grid to obtain a corrected topological grid; Apply multi-scale interface cross-filtering to the corrected topological grid and perform matrix construction to generate an interface coherence enhancement matrix;

[0113] Step S144: Based on the interface coherence enhancement matrix, perform topological segmentation on the carbon fiber composite material sample to generate a double-interface labeled map; Extract the edge microstructure description data of the double-interface labeled map.

[0114] In this embodiment, the region with curvature change is extracted from the spectral edge feature map. The polynomial fitting method is adopted to finely fit the boundary region with abrupt curvature change. The quadratic curve fitting algorithm is used to continuously fit the edge feature points, and the curvature value of each fitting point is calculated. The local curvature transformation matrix is generated using the curvature change information. Each item in the matrix represents the curvature change of the boundary point. The least squares method is used to optimize the fitting parameters during the fitting process to ensure the minimum error between the fitting curve and the actual boundary, so as to obtain accurate curvature data, identify the positions of the curvature mutation points, determine the distribution positions of the mutation points by analyzing the extreme points in the local curvature transformation matrix, and then divide the boundary feature contour into multiple regions according to these mutation points. The curvature change in each region is relatively stable. Each partition is constrained by the partition constraint mapping algorithm to ensure that the boundary features of each region satisfy certain curvature change rules, thereby generating an initial topological grid. Each grid cell in the topological grid corresponds to a local region, and the grid is divided according to the continuity of the curvature. The initial topological grid is checked to check whether the nodes in the grid are connected normally. The connectivity analysis algorithm in graph theory is used to identify the isolated nodes and discontinuous connection parts in the grid, and the connectivity of the grid is corrected by adding connection lines or readjusting the node positions to ensure that the grid is connected in the topological structure. After the correction is completed, multi-scale interface cross filtering is applied to filter the nodes and boundaries in the grid. The Gaussian filter is used to smooth the grid, and the scale parameter of the filter is set to 2 mm to ensure the enhancement of the interface details at multiple scales. Finally, an interface coherence enhancement matrix is generated. Each element in the matrix represents the connectivity and interface smoothness of adjacent grids. Using the coherence information in the enhancement matrix, the topological grid is finely segmented. The gradient-based segmentation algorithm is adopted to segment the grid according to the interface features, and the interface features of each grid are ensured to be consistent during the segmentation process to generate a double interface marking map. The inner and outer interfaces and different regions of the sample are clearly marked in the marking map. By further analyzing each region in the marking map, the edge microstructure description is extracted. The microstructure description data includes tiny cracks and irregular surface morphologies on the boundary. The geometric features of the microstructure are extracted by image processing methods, and finally the edge microstructure description data is formed.

[0115] Preferably, step S2 includes the following steps:

[0116] Step S21: Perform multi-frequency point sampling based on the internal field distribution mapping data to obtain a frequency response matrix; perform wavelength eigen-decomposition on the frequency response matrix to generate eigen-frequency data;

[0117] Step S22: Conduct fiber spacing correlation analysis on the characteristic frequency data to generate a spacing-frequency relationship graph; perform parameter fitting on the spacing-frequency relationship graph, where the fitting parameters are set between 0.1 mm and 5 mm, to generate a spatial distribution function;

[0118] Step S23: Detect the extreme points of the spatial distribution function and perform frequency domain transformation mapping on the extreme points of the spatial distribution function to generate resonance frequency parameters;

[0119] Step S24: Apply phase reference compensation to the resonance frequency parameters to obtain corrected frequency parameters; perform quantitative calibration of the phase resonance points based on the spatial distribution function and the corrected frequency parameters to obtain a frequency characteristic spectrum, where the selection range of the resonance points is between 1 MHz and 10 GHz, and fine scanning is performed in steps of less than 10 kHz.

[0120] In this embodiment, the internal field distribution mapping data of the carbon fiber composite material is obtained. A vector network analyzer (VNA) is used to irradiate the sample with electromagnetic waves in the frequency range of 1 MHz to 10 GHz to obtain the electric field distribution data. The measurement point spacing is set at 0.1 mm to ensure the spatial resolution. Based on the complex permittivity data at the measurement points, the Fourier Transform is used to extract the response signals at different frequency points, and the corresponding amplitude and phase information are recorded to form a complete frequency response matrix. In the obtained frequency response matrix, the spatial electric field distribution at each frequency point is normalized, and the Singular Value Decomposition (SVD) is used to decompose the matrix to extract the dominant characteristic frequencies. The eigenvector truncation value is set at 0.01 to remove the noise influence and generate the characteristic frequency data. The fiber spacing correlation analysis is performed on the extracted characteristic frequency data. A high-precision optical microscopy system is used to collect the microscopic distribution images of the carbon fibers, and an image processing algorithm is used to measure the fiber spacing. The measured spacing data is correlated and matched with the characteristic frequency data to construct the spacing-frequency data pairs. In the correlation analysis process, the Least Squares Method is used for data fitting, and the fitting range is set at 0.1 mm to 5 mm to ensure that the data covers all possible fiber spacing ranges. During the fitting process, a high-order polynomial function is selected for curve fitting to improve the fitting accuracy. Finally, a spatial distribution function is obtained. The spatial distribution function is used to describe the mathematical relationship between the characteristic frequency and the fiber spacing. On the generated spatial distribution function, the first derivative is used to solve for the local extreme points, all the spacing values that meet the extreme value conditions are extracted, and the corresponding characteristic frequencies are obtained. During the process of solving for the extreme points, the finite difference method is used to calculate the derivative to ensure numerical stability. Based on the extracted extreme point data, wavelet transform is used to perform a frequency domain transformation on the signal to improve the frequency resolution and eliminate the interference of non-physical resonance points. During the frequency domain mapping process, the Brillouin scattering mode is used for matching to ensure that the mapping process conforms to the electromagnetic theory. Finally, the resonant frequency parameters are generated. The resonant frequency parameters are used to describe the specific resonance characteristics of the carbon fiber composite material under different spacing conditions. Phase reference compensation is performed on the obtained resonant frequency parameters. First, the reference phase is obtained through time domain reflectometry (TDR), and the time domain response of the signal is restored using the inverse Fourier transform. During the compensation process, the relative phase shift of the resonant frequency parameters is calculated, and the linear phase correction method is used for phase adjustment to ensure the phase consistency of each resonant point. After the correction is completed, based on the spatial distribution function and the corrected resonant frequency parameters, a high-precision radio frequency measurement system is used to perform a fine scan of the resonant points in the frequency range of 1 MHz to 10 GHz with a step size of less than 10 kHz. During the scan process, fast Fourier transform (FFT) is used for real-time spectrum analysis, and the peak detection algorithm is used to extract the final frequency characteristic spectrum. The frequency characteristic spectrum is used to characterize the dielectric response characteristics of the carbon fiber composite material.

[0121] Preferably, the field area classification mapping and interactive coupling evaluation according to the frequency characteristic spectrum in step S3 include:

[0122] Perform feature dimensionality reduction on the frequency characteristic spectrum to generate field area principal component data;

[0123] Perform clustering analysis on the field area principal component data to obtain the field area categories;

[0124] Construct a field area identifier based on the field area categories;

[0125] Perform internal and external field area mapping based on the field area identifier to obtain field area classification data;

[0126] Construct a field strength gradient field for the field area classification data; perform polarization vector analysis on the field strength gradient field to obtain polarization field description data;

[0127] Perform interactive coupling evaluation based on the polarization field description data to obtain field coupling strength data.

[0128] In this embodiment, the frequency characteristic spectrum data is arranged according to frequency points, and the response values of each frequency point are normalized. The principal component analysis (PCA) algorithm is used to reduce the dimension of the normalized frequency characteristic data, and the first 5 principal components are selected. These principal components can retain the information in the frequency characteristic spectrum to the greatest extent. By calculating the contribution rate of the principal components, the effectiveness of the data after dimension reduction is ensured. The finally generated main component data of the field region contains the main information in the frequency characteristic spectrum. The K-Means Clustering algorithm is used to analyze the data after dimension reduction, and the number of clusters is set to 3, that is, the field region is classified into three categories, and each category corresponds to different frequency response characteristics. During the clustering process, first, the main component data is initialized, the initial centroids are randomly selected, and the distance from each data point to the centroid is calculated. Iterative calculation is performed until the data points are assigned to the most suitable clusters. The finally obtained field region categories reflect the aggregation of different frequency response characteristic regions, and each category represents a typical response pattern. A unique identifier is assigned to each category, and the category information obtained by clustering analysis is mapped to the spatial position of the field region to form a set of identifiers. Each identifier corresponds to a specific field region category, and these identifiers are the key characteristic information describing different field regions. According to the field region identifiers, the internal and external regions of the material are divided, and the identifiers of the internal field region and the external field region are respectively mapped to the inside and outside of the material. The internal field region refers to a specific region inside the material, and the external field region refers to a specific region outside the material. The electric field strength of each field region is calculated, and according to the change of the electric field strength, a field strength gradient field is constructed. By using the finite element analysis (FEA) method to perform spatial interpolation on the field strength, a three-dimensional field strength distribution map is obtained. The field strength gradient field shows the change law of the electric field strength in different regions inside and outside the material. The direction of the electric field strength at each point is calculated, and through the analysis of the field strength gradient field, the polarization vector of each region is obtained. These polarization vectors reflect the electric field response state of each point in the field region. The calculation of the polarization vector uses the eigen-equation of the electromagnetic field, and by solving these equations, the polarization direction and intensity of each point are obtained. The obtained polarization field description data is combined with the dielectric constant characteristics of the material, and the numerical calculation method (such as Monte Carlo simulation) is used to perform interactive coupling analysis on the electric field responses in different regions. The calculated field coupling strength data reflects the electromagnetic coupling strength between different regions inside the material.

[0129] Preferably, the determining the node link structure based on the field coupling strength data in step S3 includes:

[0130] Reconstruct the neighborhood relationship according to the field coupling strength data to obtain the node neighborhood relationship;

[0131] Quantify the link connection strength based on the node neighborhood relationship to generate link strength quantization data;

[0132] Perform hierarchical clustering on the link strength quantization data to obtain a hierarchical clustering network;

[0133] Based on the hierarchical clustering network, screen the association paths of the node neighborhood relationship to obtain the key node connection links;

[0134] Reshape the link structure of the key node connection links based on the link strength quantization data to generate a node link structure.

[0135] In this embodiment, the field coupling strength data is standardized to map the coupling strength values of each node to a unified numerical range. Subsequently, the coupling strength between each pair of nodes is calculated. By setting a threshold, it is determined whether two nodes are neighborhood nodes. If the coupling strength between two nodes exceeds the threshold, it is considered that they have a neighborhood relationship. The relationship between nodes is represented by an adjacency matrix in graph theory. The elements in this matrix represent the coupling strength between node pairs. In this way, the neighborhood relationship of nodes is constructed. Calculate the connection strength between each pair of neighborhood nodes, convert the non-zero elements in the neighborhood relationship matrix into connection strength values, and quantify the connection strength between each pair of nodes by applying a weighted network model. The specific operation of connection strength quantization is to weight each pair of nodes according to the coupling strength, and the setting of the weight is adjusted according to the distance between nodes or the magnitude of the coupling strength. Use the hierarchical clustering method (Hierarchical Clustering) to analyze the link strength quantization data, use the Euclidean distance to measure the similarity between each pair of nodes, initialize all nodes as independent clusters, aggregate them in descending order of connection strength, and gradually merge pairs of nodes with higher similarity until the set number of clusters or similarity threshold is reached. Finally, a hierarchical clustering network is obtained, where each cluster represents a group of nodes with similar coupling strengths. According to the hierarchical clustering results, screen out the node pairs in different clusters, determine the roles and status of the key nodes in each cluster, evaluate the relationship between the nodes within each cluster, and screen out the links that play an important role in electromagnetic coupling. These links represent the most critical interaction paths between nodes. Use graph analysis techniques and the Max Flow Min Cut algorithm to evaluate the importance of the paths between nodes, screen out the key node connection links, and adjust the connection method between each node according to the link strength quantization data of the key node connection links, and perform weighted reconstruction on the links. Use graph optimization algorithms (such as the shortest path algorithm) to adjust the connection relationship between nodes. Through optimization and adjustment, the connection structure is made more reasonable, ensuring that while maintaining the key node connection links, unnecessary redundant connections are reduced, and finally a new node link structure is generated.

[0136] Preferably, the internal and external field interaction tensor projection based on the node link structure in step S4 includes:

[0137] Performing tensor eigen-decomposition on the node link structure to obtain node tensor features;

[0138] Performing field space weight assignment on the node tensor features according to preset spatial weight parameters to obtain field space weight data;

[0139] Performing tensor field interaction analysis based on the field space weight data to generate tensor field interaction relationships;

[0140] Performing projection space reconstruction on the tensor field interaction relationships to obtain internal and external field projection data;

[0141] Performing multi-scale feature fusion on the internal and external field projection data to generate multi-scale field features;

[0142] Performing tensor reconstruction according to the multi-scale field features to generate a multi-scale field model.

[0143] In this embodiment, the node-link structure is represented as a high-order tensor. The dimensions of this tensor correspond to the electromagnetic coupling relationship between nodes and their interaction strength. Taking this tensor as the input, the singular value decomposition (SVD) method is applied to perform eigen-decomposition on the tensor. Through the decomposition, the eigenvalues and eigenvectors of the tensor are obtained. The eigenvalues represent the contributions of the coupling strength in different dimensions, and the eigenvectors represent the main feature directions in each dimension. A spatial weight is assigned to each node tensor feature. This weight is set based on factors such as the position of the node in space, the electromagnetic coupling relationship between nodes, and spatial characteristics. The spatial weight parameter can be quantified through a preset spatial coordinate system. When assigning weights, the weighted average method is used. Considering the relationship between the electromagnetic influence of each node and its position in space, the tensor features of the nodes are weighted to obtain field space weight data. Combining the field space weight data with the electromagnetic coupling strength between nodes forms a composite tensor containing spatial weights and coupling strength. On this basis, tensor decomposition technology is applied to conduct interaction analysis of the tensor field. During the analysis process, the multiplication operation of the tensor is used to calculate the electromagnetic interaction effects between nodes. By modeling the interaction of multiple tensors, a tensor field interaction relationship is generated. This relationship reveals how the electromagnetic interactions between different nodes affect the overall behavior of the composite material. According to the tensor field interaction relationship, the principal components to be retained are determined. Through the principal component analysis (PCA) method, the most representative projection space is extracted from the original high-dimensional tensor space. The reconstructed projection space contains the key information of the interaction relationship between nodes and removes redundant parts to obtain the internal and external field projection data. The internal and external field projection data are decomposed at different scales, and the wavelet transform technology is used to perform multi-scale processing on the data. Each scale corresponds to different spatial resolutions and frequency components. During the fusion process, the features of each scale are combined according to a certain weighting rule. The combined data can reflect the field features at different scales and generate multi-scale field feature data. This data contains comprehensive information on the influence of the dielectric constant of the composite material at different scales. According to the multi-scale field feature data, it is transformed into a higher-dimensional tensor structure through the tensor reconstruction method. Using the known composite material model parameters, a composite material field model containing multi-scale information is generated through the reconstruction of the multi-scale feature data.

[0144] Preferably, the constraint embedding process of the multi-scale field projection data based on the node-link structure and the composite field modeling in step S4 include:

[0145] Extract the topological features of the node-link structure;

[0146] Perform hierarchical coupling deconstruction on topological features to obtain multi-level connection data;

[0147] Perform boundary gradient detection on the multi-level connection data to obtain a structure constraint boundary;

[0148] Perform manifold embedding transformation on the structure constraint boundary to generate constrained manifold data;

[0149] Perform constrained embedding processing on the multi-scale field projection data based on the constrained manifold data and perform composite field modeling to obtain a composite field model.

[0150] In this embodiment, when extracting the topological features of the node-link structure, a three-dimensional CT scanning device is used to perform high-resolution scanning on the carbon fiber composite material to obtain complete internal microstructure data. The scanned grayscale image is converted into a binary image, and noise is removed and the boundary features between the fibers and the matrix are enhanced through morphological operations. The Watershed Algorithm is used for image segmentation to identify the distribution of carbon fiber bundles. An undirected graph is constructed based on the image processing results, where the center position of the carbon fiber bundle is used as a node, and the physical contact relationship between different fiber bundles is defined as an edge. Calculate the degree, clustering coefficient, and adjacency matrix of each node to fully characterize the topological features of the carbon fiber composite material. The PageRank algorithm is used to rank the importance of different nodes to obtain hierarchical topological structure information.

[0151] When performing hierarchical coupling deconstruction on topological features, based on the topological features constructed in the previous step, the Community Detection Algorithm is used to classify the nodes. The LabelPropagation Algorithm is selected to identify different carbon fiber bundle network regions. According to the connection strength between the nodes, the entire topological structure is decomposed into multiple subgraphs, and the hierarchical relationship of different subgraphs is defined. Each subgraph represents different scale coupling features inside the carbon fiber composite material. To further analyze the multi-scale coupling features, the Non-negative Matrix Factorization (NMF) method is used to decompose the adjacency matrix of each hierarchical subgraph to obtain multi-level connection data. Calculate the shortest path distribution, betweenness centrality, and motif structure between nodes at each level, and analyze the topological evolution trend of different hierarchical networks under the action of the electromagnetic field. Finally, a multi-scale coupling feature data set is formed.

[0152] When performing boundary gradient detection on multi-level connection data, perform boundary analysis on the stored multi-level topological data, calculate the connection density within each level and the edge weight distribution across levels, select a Gradient Boosted Filter to extract the region with the most drastic changes in the network structure, establish a Laplacian Matrix in the topological space, calculate the eigenvectors of each node based on Laplacian Eigenmaps, compare the change rates of the node features in each level, screen the region with the largest change rate by setting a threshold to obtain the structure constraint boundary, map this boundary data back to the three-dimensional microstructure, and compare the changes in electromagnetic parameters at the junctions of different carbon fiber bundles.

[0153] When performing manifold embedding transformation on the structure constraint boundary, construct boundary point cloud data based on the boundary detection results, use the Isomap algorithm to reduce the dimensionality of the boundary point cloud data to the manifold space, calculate the local geometric relationships of the boundary point cloud, use the k-Nearest Neighbors (k-NN) method to establish a local adjacency graph, calculate the similarity between different boundary regions based on the Earth Mover’s Distance (EMD), use Principal Component Analysis (PCA) to further optimize the stability of the manifold embedding, keep the topological relationships of the boundary points unchanged during the embedding process, and at the same time reduce the redundant information in the high-dimensional data to finally generate the constrained manifold data.

[0154] Perform multi-scale field projection transformation on the measured data of the equivalent dielectric constant of carbon fiber composites, project the measured electromagnetic field data into the spatial field domain at different scales, decompose the original electromagnetic field data using wavelet transform, convert it into field data at multiple scales, select the Daubechies wavelet basis for four-layer wavelet decomposition, convert the original data into approximation components and detail components, representing low-frequency and high-frequency information respectively, store the projection data at each scale, construct a constrained embedding model based on constrained manifold data, map the field projection data at different scales to the manifold space, use the Laplacian Eigenmaps method to perform dimensionality reduction on the multi-scale field projection data, construct an adjacency matrix to represent the relationship between data points, set the neighborhood parameter to 5, calculate the similarity of adjacent points using a weight function, and construct a Laplacian matrix. Use the eigenvalue decomposition method to extract the main manifold structure of the data, obtain the low-dimensional manifold embedding representation, remap the data to a unified manifold coordinate system, and use the Locally Linear Embedding (LLE) method to optimize the constraint conditions, set the embedding dimension to 3, use the composite field modeling method, construct a composite field model based on the embedded low-dimensional manifold data, set the basic unit of the field model as a three-dimensional tensor (3D Tensor), perform tensor decomposition on the embedded data, use the Higher-Order Singular Value Decomposition (HOSVD) method to represent the data as the product of a core tensor and mode matrices, perform rank constraint on the core tensor, set the rank parameter to 10 to ensure the computational efficiency of the model, and at the same time orthogonalize the mode matrices. Perform numerical calculation on the composite field model, use the Finite Difference Method (FDM) to solve the composite field model, discretize the field data into a grid form, set the grid division step size to 0.1 mm, and construct a system of difference equations. Use the Jacobi iterative method to solve the steady-state solution of the composite field, set the iteration error threshold to 1e-6 to ensure the accuracy of the calculation results, and finally obtain the composite field model data of carbon fiber composites under multi-scale field projection.

[0155] Preferably, the dielectric property inversion of the composite field model described in step S5 includes:

[0156] Perform frequency grid division on the composite field model to obtain frequency grid data;

[0157] Perform logarithmic interval optimization on the frequency grid data, set 200 sampling points in the range of 10 MHz - 100 GHz to obtain optimized frequency point data;

[0158] Initialize the anisotropic parameters for the optimized frequency point data, set the initial dielectric constant matrix, and obtain the initial dielectric parameters;

[0159] Perform gradient descent inversion on the initial dielectric parameters to generate a dielectric gradient iteration tensor;

[0160] Perform a principal axis transformation on the dielectric gradient iteration tensor to obtain the principal axis dielectric parameters;

[0161] Apply a dielectric loss coefficient constraint to the principal axis dielectric parameters and limit the dielectric loss coefficient constraint within the range of 0.001 - 0.1, thereby obtaining the loss-constrained dielectric data;

[0162] Perform frequency domain correlation on the loss-constrained dielectric data to generate a complex dielectric parameter distribution;

[0163] Perform a dispersion mapping process on the complex dielectric parameter distribution and perform curve fitting to obtain the dielectric dispersion curve.

[0164] In this embodiment, when performing frequency grid division on the composite field model, first set the frequency range from 10 MHz to 100 GHz, and use the logarithmic interval method for division. Convert the entire frequency range into a logarithmic scale, and equally spaced select 200 sampling points on the logarithmic scale to form frequency grid data. Use the logspace function (logarithmic interval sampling function) of MATLAB to generate a sequence of frequency points with logarithmic intervals, and import this sequence into the finite element simulation software ANSYS HFSS (High Frequency Structure Simulator) or CST (Computer Simulation Technology). Based on this sequence of frequency points, perform frequency discretization processing on the composite field model, so that each frequency point corresponds to an independent electromagnetic simulation calculation unit. Use FDTD (Finite-Difference Time-Domain) or FEM (Finite Element Method) for grid discretization to ensure that the data of each frequency point can be used as the input data for subsequent dielectric property inversion, and obtain frequency grid data. Read the frequency grid data, calculate the logarithmic interval ratio of adjacent sampling points, and ensure that the frequency points are distributed in accordance with the geometric progression law on the logarithmic scale. Use the gradient optimization method to adjust the interval parameters. Set the initial sequence of frequency points as equally spaced sampling values between log(f0)=log(10 MHz) and log(100 GHz). Use the Broyden–Fletcher–Goldfarb–Shanno (BFGS) optimization algorithm to adjust the sampling point spacing, making the sampling points denser in the low-frequency band and sparser in the high-frequency band, ensuring more accurate capture of the dielectric property changes in the low-frequency band. Perform error analysis on the optimized frequency point data to ensure that the optimized frequency points can still cover the entire frequency range and ensure that each frequency point can converge stably in subsequent simulation calculations. Finally, obtain the optimized frequency point data. When initializing the anisotropic parameters for the optimized frequency point data, first read the optimized frequency point data. According to the structural characteristics of the carbon fiber composite material, construct an anisotropic dielectric constant matrix in the simulation software. Use the triaxial anisotropic modeling method to set the initial values of the dielectric constants in the x-axis, y-axis, and z-axis directions respectively. The initial dielectric constant in the x-axis direction is set to ε_x = 3.5, the initial dielectric constant in the y-axis direction is set to ε_y = 3.0, and the initial dielectric constant in the z-axis direction is set to ε_z = 2.8. Use Python scripts in HFSS or CST to batch import this initial matrix data to make it the initial dielectric parameter input value of the simulation model. Use FEM to solve the dielectric response at different frequencies to verify the feasibility of this initial dielectric parameter. Set a parameter optimization module in HFSS or CST, use the inversion algorithm based on gradient descent, and set the learning rate to 0.01, set the maximum number of iterations to 500, use the minimum mean square error (MSE) as the loss function, calculate the deviation between the electric field distribution at each frequency point and the target dielectric constant, calculate the gradient of the loss function to the dielectric parameters based on the back propagation algorithm, and use the Adam optimization algorithm to dynamically adjust the learning rate to ensure that the inversion process can converge stably. During the inversion process, the dielectric parameters are updated after each iteration, and the electromagnetic response of the simulation model is recalculated. The calculated dielectric parameters are stored in a three-dimensional tensor format to form a dielectric gradient iteration tensor. The gradient iteration tensor is reduced by PCA (principal component analysis), the eigenvectors in the main axis direction are calculated, the projection of the dielectric parameters in the main axis direction at each frequency point is extracted, and the main axis dielectric parameter matrix is ​​constructed. The singular value decomposition (SVD) method is used to calculate the eigendecomposition value in the main axis direction, and the eigenvalues ​​less than 0.01 are removed to ensure the main axis. The data after the axis transformation retains the main dielectric characteristics, and the dielectric parameters are transformed to the main axis direction using the rotation matrix. When the dielectric loss coefficient is constrained for the main axis dielectric parameters, the loss coefficient range is first set to 0.001 to 0.1, and the dielectric parameter data after the main axis transformation is read. The dielectric loss value at each frequency point is calculated, and the nonlinear constraint optimization method is used to adjust tanδ to ensure that the tanδ values ​​of all frequency points are within the range of 0.001 to 0.1. The Lagrange multiplier method is used to optimize the constraints of tanδ to ensure that the overall distribution of the main axis dielectric parameters is not affected while satisfying the loss constraints. Based on the optimized loss-constrained dielectric data, the dielectric constant correlation matrix at different frequency points is calculated, and the Pearson correlation coefficient is used to calculate the correlation of dielectric properties between different frequency points. The frequency domain correlation diagram is constructed, and the graph clustering algorithm (Spectral Clustering) is used to cluster the frequency points and divide the frequency points into different frequency bands. The dielectric parameters within each frequency band have a high correlation. In HFSS or CST, Python scripts are used to read the frequency domain correlation results, and the data of different frequency bands are mapped back to the simulation model to finally generate the complex dielectric parameter distribution. The complex dielectric parameter data after frequency domain correlation is read, and a dispersion model is constructed in HFSS or CST. The Drude model or Debye model is used to perform dispersion fitting on the complex dielectric parameters. During the fitting process, the plasma frequency ω_p of the Drude model is set to 2π×10^12rad / s, and the collision frequency γ is set to 2π×10^10rad / s to ensure that the dispersion characteristics of the complex dielectric parameters can be accurately described in the high frequency band. The fitted complex dielectric parameters are interpolated to generate continuous dielectric dispersion curves in the entire range of 10MHz to 100GHz. .

[0165] Preferably, the data modeling based on the node dispersion curve in step S5 includes:

[0166] Extract key nodes of dielectric dispersion curve;

[0167] Perform linear interpolation on the key nodes, set the dielectric constant range to 2 - 12 to apply constraints, and plot the dispersion curve to obtain the normalized node dispersion curve;

[0168] Based on the normalized node dispersion curve, perform polynomial expansion projection to generate polynomial coefficients;

[0169] Apply regularization constraints to the polynomial coefficients to obtain a regularized equivalent model;

[0170] Based on the regularized equivalent model, reconstruct the frequency response function to generate equivalent dielectric response data;

[0171] Perform model compression on the equivalent dielectric response data to obtain an equivalent dielectric constant model.

[0172] In this embodiment, the dielectric dispersion curve data of the composite material in the range of 10 MHz to 100 GHz is obtained, the frequency range is divided into 200 equally spaced sampling points, and the real dielectric constant (ε′) and imaginary dielectric loss (ε″) data corresponding to each sampling point are extracted. In the extraction process, an algorithm based on the curvature change rate is used to calculate the second-order derivative of the curve, and the curvature mutation points are screened as key nodes. In order to improve the extraction accuracy, the curvature change rate of each mutation point is calculated, and a threshold of 0.02 is set as the mutation judgment standard. Points with a curvature change rate greater than the threshold are regarded as key nodes. At the same time, the Euclidean distance between adjacent points is considered, and points with a distance less than 0.05 are merged to reduce redundant data. The distribution of key nodes is analyzed, the dielectric constant range of each key node is calculated, and a method based on a box plot is used to identify outliers. The interquartile range (IQR) is calculated, the upper and lower thresholds are determined, and the points exceeding the threshold range are eliminated. In order to ensure the stability of the equivalent model, the dielectric constant range constraint is imposed. After the outliers are removed, all node data are stored in the data matrix and sorted in ascending order according to the frequency to ensure that the input data of the interpolation calculation has strict monotonicity. Linear interpolation is selected as the interpolation method to construct the interpolation function. The frequency value of the known key node is used as the independent variable, and the corresponding dielectric constant is used as the dependent variable. The dielectric constant value of the interpolation point is calculated by piecewise linear interpolation. In order to prevent the interpolation process from generating abnormal data, during the interpolation calculation process, its dielectric constant (ε′) value is ensured to be within the range of 2 to 12. The data points outside the range are adjusted to the adjacent boundary values. For example, for the point with ε′<2, its value is set to 2, and for the point with ε′>12, its value is set to 12, so as to obtain the normalized node dispersion curve. The orthogonal polynomial basis function is selected, and the Laguerre polynomial is used as the basis function. Polynomial) is used as the expansion basis, and the least square fitting calculation is performed on the normalized node dispersion curve data. In order to improve the calculation accuracy, the expansion order is set to 6. The basis function weighted summation is performed on the dielectric constant of each frequency point, and the corresponding polynomial coefficients are calculated. The QR decomposition method is used to optimize the numerical stability of the solution process. In the calculation process, the coefficients of high-order terms are gradient clipped to avoid numerical overflow, thereby generating polynomial coefficients. When regularization constraints are implemented on polynomial coefficients, the L2 norm regularization (Ridge Regression) method is used to constrain all polynomial coefficients, and the regularization strength parameter λ is set to 0.01. To prevent overfitting, during the calculation process, the Laplace Smoothing method is introduced to impose an exponential decay constraint on the high-order coefficients to reduce their impact on the model. At the same time, it ensures that the low-order coefficients maintain a higher weight, thereby obtaining a regularized equivalent model. When reconstructing the frequency response function based on the regularized equivalent model, the Fourier Series Expansion method is used to convert the regularized polynomial coefficients into a frequency-domain expression form to reconstruct the distribution characteristics of the dielectric constant at different frequencies. To enhance the reconstruction accuracy, interpolation calculation is performed within the frequency range, and the Cubic Spline Interpolation method is used to fill the discrete data points to ensure the continuity of the frequency response function. During the interpolation process, boundary conditions are set so that the first derivative of the interpolation function remains continuous at the boundaries to eliminate the boundary effect, thereby generating equivalent dielectric response data. When performing model compression on the equivalent dielectric response data, the Principal Component Analysis (PCA) method is used to reduce the dimensionality of the high-dimensional equivalent dielectric response data, calculate the covariance matrix, solve for the eigenvalues and eigenvectors, and select the principal components with a cumulative contribution rate reaching 99% for retention. During the dimensionality reduction process, to reduce the computational complexity, the Singular Value Decomposition (SVD) method is introduced to decompose the eigenmatrix and truncate the low-contribution singular value terms to reduce data redundancy. Finally, numerical fitting is performed on the compressed data to generate an equivalent dielectric constant model.

[0173] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to encompass all changes that fall within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0174] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing an equivalent dielectric constant model of a carbon fiber composite material, characterized in that: The following steps are involved: Step S1: collecting carbon fiber composite material samples; performing electromagnetic orthogonal frequency sweep analysis on the carbon fiber composite material samples to obtain a frequency spectrum distribution map; performing boundary topological reorganization on the frequency spectrum distribution map to generate internal field distribution mapping data; Step S2: performing spacing correlation analysis on the internal field distribution mapping data to generate a spatial distribution function; performing phase resonance point calibration based on the spatial distribution function to obtain a frequency characteristic spectrum; Step S3: classify and map the field regions according to the frequency characteristic spectrum, and perform interactive coupling evaluation to generate field coupling intensity data; Determining the node link structure based on the field coupling strength data; Step S4: performing internal and external field interaction tensor projection based on the node link structure to generate multi-scale field projection data; performing constraint embedding processing on the multi-scale field projection data based on the node link structure, and performing composite field modeling to obtain a composite field model; Step S5: Perform dielectric property inversion on the composite field model to generate a dielectric dispersion curve; perform data modeling based on the node dispersion curve to obtain an equivalent dielectric constant model.

2. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting carbon fiber composite material samples; performing multi-frequency electromagnetic scanning on the carbon fiber composite material samples to obtain original electromagnetic response data; Step S12: performing Fourier domain conversion on the original electromagnetic response data to obtain an electromagnetic response spectrum; performing spectrum distribution analysis on the electromagnetic response spectrum to generate a spectrum distribution map; Step S13: performing adaptive filtering on the spectrum distribution map and detecting the spectrum edge to obtain a boundary feature contour; Step S14: topologically segmenting the carbon fiber composite material sample based on the boundary feature contour to generate a dual-interface marker map; extracting edge microstructure description data of the dual-interface marker map; Step S15: spatially reconstruct the edge microstructure description data to obtain internal field distribution mapping data.

3. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: performing continuous curvature fitting on the boundary feature contour to obtain a local curvature transformation matrix; Step S142: applying partition constraint mapping to the curvature mutation points according to the local curvature transformation matrix to generate an initial topological mesh; Step S143: correcting the connectivity of the initial topological grid to obtain a corrected topological grid; applying multi-scale interface cross filtering to the corrected topological grid, and performing matrix construction to generate an interface coherence enhancement matrix; Step S144: topologically segmenting the carbon fiber composite material sample based on the interface coherence enhancement matrix to generate a dual-interface marker map; and extracting edge microstructure description data of the dual-interface marker map.

4. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing multi-frequency sampling based on the internal field distribution mapping data to obtain a frequency response matrix; performing wavelength characteristic decomposition on the frequency response matrix to generate characteristic frequency data; Step S22: performing fiber spacing correlation analysis on the characteristic frequency data to generate a spacing-frequency relationship diagram; performing parameter fitting on the spacing-frequency relationship diagram, wherein the fitting parameters are set in the range of 0.1 mm to 5 mm, to generate a spatial distribution function; Step S23: detecting the extreme points of the spatial distribution function, and performing frequency domain transformation mapping on the extreme points of the spatial distribution function to generate resonance frequency parameters; Step S24: applying phase reference compensation to the resonant frequency parameters to obtain correction frequency parameters; performing quantitative calibration of the phase resonance points based on the spatial distribution function and the correction frequency parameters to obtain a frequency characteristic spectrum, wherein the resonance points are selected in the range of 1 MHz to 10 GHz and are finely scanned in steps of less than 10 kHz.

5. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: The step S3 of classifying and mapping the field regions according to the frequency characteristic spectrum and evaluating the interaction coupling includes: Performing feature dimension reduction on the frequency characteristic spectrum to generate the principal component data of the field area; Cluster analysis was performed on the principal component data of the field area to obtain the field area categories; constructing a field area identifier based on the field area category; Mapping the inner and outer field areas based on the field area identifiers to obtain field area classification data; Constructing a field intensity gradient field from the field area classification data; performing polarization vector analysis on the field intensity gradient field to obtain polarization field description data; Based on the polarization field description data, the interaction coupling evaluation is carried out to obtain the field coupling strength data.

6. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: Determining the node link structure based on the field coupling strength data in step S3 includes: Reconstruct the neighborhood relationship based on the field coupling intensity data to obtain the node neighborhood relationship; Quantify the link connection strength based on the node neighborhood relationship and generate link strength quantification data; Perform hierarchical clustering on the link strength quantification data to obtain a hierarchical clustering network; Based on the hierarchical clustering network, the associated paths of the node neighborhood relationships are screened to obtain the key links connecting the nodes; The link structure of the key links connecting nodes is reshaped based on the link strength quantification data to generate a node link structure.

7. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: The projection of the internal and external field interaction tensor based on the node link structure described in step S4 includes: Perform tensor feature decomposition on the node link structure to obtain node tensor features; Performing field space weight allocation on node tensor features according to preset space weight parameters to obtain field space weight data; Perform tensor field interaction analysis based on field space weight data to generate tensor field interaction relationships; Reconstruct the projection space of the tensor field interaction relationship to obtain the internal and external field projection data; Perform multi-scale feature fusion on the internal and external field projection data to generate multi-scale field features; Tensor reconstruction is performed according to the multi-scale field features to generate a multi-scale field model.

8. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: The step S4 of performing constraint embedding processing on the multi-scale field projection data based on the node link structure and performing composite field modeling includes: Extract the topological features of the node link structure; The topological features are deconstructed by hierarchical coupling to obtain multi-level connection data; Perform boundary gradient detection on multi-level connection data to obtain the structural constraint boundary; Perform manifold embedding transformation on the structural constraint boundary to generate constraint manifold data; Based on the constraint manifold data, the multi-scale field projection data is subjected to constraint embedding processing, and composite field modeling is performed to obtain a composite field model.

9. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: The dielectric property inversion of the composite field model in step S5 includes: Perform frequency grid division on the composite field model to obtain frequency grid data; The frequency grid data is logarithmically optimized, 200 sampling points are set in the range of 10MHz-100GHz, and the optimized frequency point data is obtained; Initialize the anisotropic parameters of the optimized frequency point data, set the initial dielectric constant matrix, and obtain the initial dielectric parameters; Perform gradient descent inversion on the initial dielectric parameters to generate a dielectric gradient iteration tensor; Performing principal axis transformation on the dielectric gradient iteration tensor to obtain principal axis dielectric parameters; The dielectric loss coefficient constraint is imposed on the dielectric parameters of the spindle, and the dielectric loss coefficient constraint is limited to the range of 0.001-0.1, thereby obtaining loss-constrained dielectric data; Perform frequency domain correlation on loss-constrained dielectric data to generate complex dielectric parameter distributions; The complex dielectric parameter distribution is subjected to dispersion mapping and curve fitting to obtain a dielectric dispersion curve.

10. The method for constructing an equivalent dielectric constant model of carbon fiber composite materials according to claim 1, characterized in that: The data modeling based on the node dispersion curve described in step S5 includes: Extract key nodes of dielectric dispersion curve; Linear interpolation is performed on key nodes, the dielectric constant range is set to 2-12 to impose constraints, and dispersion curves are plotted to obtain normalized node dispersion curves; Perform polynomial expansion projection based on the normalized node dispersion curve to generate polynomial coefficients; Regularization constraints are imposed on the polynomial coefficients to obtain a regularized equivalent model; Reconstructing the frequency response function based on the regularized equivalent model to generate equivalent dielectric response data; Model compression is performed on the equivalent dielectric response data to obtain an equivalent dielectric constant model.

Images