Metal wire data analysis method and system

Abstract

The present application relates to the technical field of data processing, and more particularly to a metal wire data analysis method and system. The method comprises the following steps: obtaining wire interface data of a composite metal; collecting a wire interface feature matrix based on the wire interface data; constructing a wire interface feature model according to the wire interface feature matrix; performing differential signal conversion processing on the wire interface feature model, extracting wire interface change features based on the differential signal conversion result; constructing a wire interface defect preliminary indication map according to the wire interface change features; performing multi-scale decomposition processing based on the wire interface defect preliminary indication map to obtain interface defect feature data; and identifying wire interface defect type features according to the interface defect feature data. The present application establishes a closed-loop optimization mechanism of defect features and production process parameters, significantly improving the interface quality and reliability of high-performance composite metal wires.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for analyzing data of metal wires. Background Technology

[0002] Metal wire is a slender metallic material with a relatively small cross-sectional area and a relatively long length, typically in the form of a wire. It can be made from a single metal (such as copper, aluminum, iron, etc.) or an alloy (such as stainless steel, nickel-titanium alloy, etc.). The diameter of metal wire is usually less than or equal to 100 mm. Bimetallic composite wire is a composite wire composed of two different metals or alloys, with its core and outer layer made of different metallic materials. This material combines the characteristics of both the core and outer layer materials, saving high-value metals during use while compensating for the shortcomings of single materials. In the production of bimetallic composite wire (such as copper-clad aluminum wire), complex defects exist at the interface (such as peeling, porosity, and inclusions occurring simultaneously). These defects are usually not isolated but multiple and superimposed, resulting in highly coupled signals; single defect identification algorithms fail, requiring simultaneous modeling of the complex relationships between multiple defect features, which is extremely difficult to annotate. Copper and aluminum often have a diffusion layer or intermediate layer structure at the composite interface, leading to blurred boundaries or even a gradual transition in imaging or spectral analysis. Traditional algorithms based on edge recognition and texture differences struggle to clearly extract defect contours or locate defect areas. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a method and system for analyzing metal wire data to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for analyzing metal wire data includes the following steps:

[0005] Step S1: Obtain wire interface data of composite metals; collect wire interface feature matrix based on wire interface data; construct wire interface feature model based on wire interface feature matrix;

[0006] Step S2: Perform differential signal conversion processing on the wire interface feature model, and extract the wire interface change features based on the differential signal conversion results; construct a preliminary indicator map of wire interface defects based on the wire interface change features;

[0007] Step S3: Perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data.

[0008] Step S4: Establish a defect distribution map of the wire interface based on the accurate defect characterization data; calculate the severity of the wire interface defects based on the defect distribution map; analyze the wire production process related data based on the severity of the wire interface defects.

[0009] Step S5: Generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters and verify wire interface quality to obtain wire interface performance evaluation data.

[0010] The present invention also provides a metal wire data analysis system for performing the metal wire data analysis method described above, the metal wire data analysis system comprising:

[0011] The data acquisition and modeling module is used to acquire wire interface data of composite metals; acquire wire interface feature matrix based on wire interface data; and construct wire interface feature model based on wire interface feature matrix.

[0012] The differential signal processing module is used to perform differential signal conversion processing on the wire interface feature model, extract the wire interface change features based on the differential signal conversion results, and construct a preliminary indicator map of wire interface defects based on the wire interface change features.

[0013] The multi-scale decomposition module is used to perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; and match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data.

[0014] The defect distribution module is used to establish a defect distribution map of the wire interface based on the precise defect characterization data; calculate the severity of wire interface defects based on the defect distribution map; and analyze the wire production process related data based on the severity of wire interface defects.

[0015] The quality verification module is used to generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters, and verify wire interface quality to obtain wire interface performance evaluation data.

[0016] This invention significantly improves the accuracy of the production process and the stability of wire quality by comprehensively analyzing and optimizing interface defects in the metal wire manufacturing process. First, by establishing a physical constraint model library of wire interface defects and comparing it with the parameterized characteristics of defects, it is possible to accurately identify and characterize different types of defects, helping to identify the generation mechanisms of various defects and their impact on wire quality. Each defect type can be finely classified according to its characteristics, and corresponding optimization strategies can be adopted for different defects, thereby effectively reducing product quality fluctuations caused by defects. Second, by combining spatial distribution data of wire interface defects, a defect distribution map can be accurately drawn, providing an intuitive basis for subsequent quality assessment. This distribution map not only reflects the occurrence patterns of various defects but also reveals the changing trends of defects under different production process parameters. This combination of spatial positioning and distribution analysis helps to more accurately assess the overall and local quality levels of the wire, providing early warning of potential quality problems that may occur during the production process. During the quality assessment process, this method, through correlation analysis of the relationship between defect types and production process parameters, can gain a deeper understanding of the direct impact of process parameter changes on wire interface quality. This analysis clearly identifies which process parameters are key factors leading to specific defects, providing a theoretical basis for subsequent process adjustments. More importantly, the results of this analysis enable the optimization of production process parameters, avoiding new defects caused by over- or inappropriate adjustments, thereby improving overall production efficiency and product quality consistency. Furthermore, this method demonstrates high flexibility and scientific rigor in process parameter optimization. By accurately identifying key process parameters affecting interface quality and optimizing them accordingly, it effectively reduces defect types caused by improper process parameters, especially defects such as peeling and porosity that significantly impact wire performance. This process not only improves production efficiency but also reduces resource waste and lowers production costs. After simulation verification of the actual process parameter adjustment scheme, the adjustment effect can be predicted in advance, providing more reliable guidance for the production line. This adjustment verification based on predictive data avoids the uncertainty caused by blind adjustments and further reduces costs during the experimental phase. By translating the optimized process parameters into production equipment control commands, the optimized scheme can be executed more precisely during production, and process parameters can be fine-tuned in real time to ensure that each stage operates in an optimal state. By adjusting production process parameters in real time and collecting relevant process data, this method enables precise control of various process parameters during production. The real-time data feedback mechanism not only ensures continuous optimization of parameters during production but also provides detailed process data support for quality control. In this process, precise process adjustment and data acquisition mechanisms provide a reliable basis for subsequent quality inspection and offer rapid response capabilities to anomalies in the production process.Finally, by combining the interface quality detection results, this method achieves a quantitative evaluation of the process optimization effect. By detecting the type, quantity, size, and distribution of surface and internal defects in the wire, the improvement effect of process optimization on wire quality can be comprehensively evaluated. This process ensures that various process adjustments during production meet the expected goals and provides a feedback mechanism for subsequent process adjustments. This feedback mechanism not only helps the production process team continuously optimize the process but also improves the quality standards of wire production through regular evaluation. In summary, this method, by accurately analyzing the generation mechanism of wire interface defects, combining real-time data from the production process, deeply explores the influence relationship between process parameters and defects, and through scientific process optimization and real-time adjustments, ensures that every stage of wire production operates at its optimal state. Through this series of interconnected optimization processes, the overall quality of wire products can be significantly improved, the number of defective products caused by defects can be reduced, thereby increasing production efficiency and reducing production costs, demonstrating significant practical application value. Attached Figure Description

[0017] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0018] Figure 1 This is a schematic diagram of the steps of a metal wire data analysis method according to the present invention;

[0019] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0020] Figure 3 for Figure 1 A detailed flowchart of step S2. Detailed Implementation

[0021] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0023] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a method for analyzing metal wire data, the method comprising the following steps:

[0025] Step S1: Obtain wire interface data of composite metals; collect wire interface feature matrix based on wire interface data; construct wire interface feature model based on wire interface feature matrix;

[0026] In this embodiment of the invention, when analyzing data of a composite metal wire, high-resolution image data of the wire cross-section is first acquired using scanning electron microscopy (SEM) combined with focused ion beam (FIB) technology. This allows for the collection of microscopic interface information between different metal layers within the wire, and the acquired image data constitutes the wire interface data. Next, a grayscale gradient detection method based on image processing is used to extract interface contour features. Simultaneously, texture direction matrices (such as grayscale co-occurrence matrices) and multi-channel color distribution statistics are combined to encode the interface material properties, thereby constructing a wire interface feature matrix. This feature matrix includes, but is not limited to, multi-dimensional feature parameters such as interface roughness, inclusion distribution, fusion boundary continuity, and metal particle orientation. Furthermore, principal component analysis (PCA) is used to reduce dimensionality and retain the main changing dimensions, establishing a wire interface feature model. This model expresses the spatial structural and material property variation patterns of different interface regions in a vectorized form, providing high-quality input data for subsequent defect identification.

[0027] Step S2: Perform differential signal conversion processing on the wire interface feature model, and extract the wire interface change features based on the differential signal conversion results; construct a preliminary indicator map of wire interface defects based on the wire interface change features;

[0028] In this embodiment of the invention, after obtaining the wire interface feature model, a differential signal conversion processing method combining time-domain and spatial-domain differential methods is employed to enhance the sensitivity to interface structural changes. Time-domain differential is used to analyze the microstructural changes of the same region in continuous production samples over time, while spatial-domain differential is used to characterize the non-uniform distribution of different wire cross-sections on the interface morphology at the same moment. The signal data obtained after differential processing presents a "response curve" form of interface feature changes. By extracting the envelope and statistically analyzing the peak density of these differential responses, interface abrupt change points and their degree of change are identified, thereby extracting wire interface change characteristics, such as the location of microcrack initiation points and metal layer separation trends. Subsequently, a preliminary defect indicator map is constructed based on these change characteristics. This map uses a pseudo-color heatmap to mark areas in the wire cross-section where defects may exist, with heat values ​​indicating the severity of changes in potential defect areas, forming the basis for the first step of defect manifestation.

[0029] Step S3: Perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data.

[0030] This invention performs multi-scale decomposition processing on the generated preliminary indicator map of wire interface defects. A wavelet packet decomposition algorithm is used to decompose the spatial features of each potential defect region in the image at multiple scales to separate local detail features from global trend features. At each scale, defect shape features (such as ellipticity and edge angles), spatial distribution features (such as orientation consistency and concentration), and material response features (such as absorptivity changes and density gradients) are extracted. An interface defect feature dataset is formed based on these features. Furthermore, a support vector machine (SVM) classifier combined with an expert experience sample database is used to identify defect type features, such as poor fusion, foreign matter inclusions, crack initiation, and interlayer debonding. Based on this, a pre-set metal physics constraint model is used to constrain and correct these defect features. This model is based on parameters such as the coefficient of thermal expansion, Young's modulus, and stress concentration factor of the metal, ensuring that the detected defect types have physical rationality from a thermodynamic and material mechanics perspective. After correction, accurate defect characterization data is generated to ensure the accuracy of subsequent evaluations.

[0031] Step S4: Establish a defect distribution map of the wire interface based on the accurate defect characterization data; calculate the severity of the wire interface defects based on the defect distribution map; analyze the wire production process related data based on the severity of the wire interface defects.

[0032] This invention maps precise defect characterization data onto a three-dimensional wire model, constructing a defect distribution map of the wire interface. The map uses a voxel representation structure, with each voxel recording the defect type, location coordinates, and degree of impact. A volume-weighted statistical method is used to analyze the defect map, calculating a defect severity score for each wire segment. This score is based on a comprehensive calculation of three dimensions: defect volume density, stress concentration area superposition, and defect distribution symmetry. Subsequently, the system extracts wire production process parameters corresponding to the defect areas, such as rolling temperature curves, alloy ratio records, cooling rates, and drawing tension changes. Cluster analysis is used to correlate and match defect severity with corresponding process data, identifying key process variables that lead to defect occurrence and prioritizing them to provide a basis for process optimization.

[0033] Step S5: Generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters and verify wire interface quality to obtain wire interface performance evaluation data.

[0034] This invention addresses key process variables identified, such as rolling temperature range shifts and high-stress drawing stages. Based on machine learning optimization algorithms (e.g., genetic algorithms combined with random forest regression), it generates a set of optimal wire rod manufacturing process parameters, forming a wire rod manufacturing process optimization dataset. This dataset includes process adjustment suggestions for various production scenarios, such as reducing drawing tension by 10%, increasing intermediate annealing temperature by 30°C, and controlling metal powder particle size to below 10 micrometers. The optimized data is loaded into the production line's process control system in real time, automatically adjusting the parameters of the temperature control module, tension controller, and mixing unit. Simultaneously, an online interface imaging unit is deployed during production to verify the interface quality change trend in real time. By comparing the changes in interface defect distribution and the decrease in defect severity scores before and after optimization, wire rod interface performance evaluation data is obtained. This evaluation data includes indicators such as defect reduction rate, improvement in interface uniformity, and overall material strength improvement ratio, verifying the effectiveness of this data analysis method in improving the actual production line.

[0035] Preferably, step S1 includes the following steps:

[0036] Step S11: Set up a wire detection device to fix the composite metal wire to be tested within the detection area of ​​the sensor array, thereby acquiring cross-sectional image data and surface feature data of the wire;

[0037] This invention deploys a wire inspection device on an experimental testing platform. The device includes a high-precision fixture, an industrial camera system, and a multi-channel surface sensor array. The composite metal wire to be tested (e.g., a 2.5 mm diameter, nickel-titanium alloy copper-clad structure) is fixed at the center of the sensor array, with the wire surface perpendicular to the array at a 90-degree angle. The industrial camera captures cross-sectional images of the wire at a sampling rate of 60 frames per second, using a ring-shaped LED array to reduce shadow interference. Simultaneously, a surface contour laser scanning module acquires surface feature data such as surface waviness, scratch depth, and oxide layer distribution. This step provides high-precision two-dimensional image data and one-dimensional surface attribute information, providing a multi-source input basis for subsequent morphology reconstruction and interface analysis.

[0038] Step S12: Perform morphological feature reconstruction processing on the cross-sectional image data and surface feature data of the wire to obtain the wire interface data of the composite metal;

[0039] This invention, based on the wire cross-sectional image and surface feature data obtained in step S11, utilizes a three-dimensional morphology reconstruction algorithm for data fusion and reconstruction. First, image segmentation technology is used to extract the metal material contour and the boundaries of each layer. Then, a frequency domain compensation algorithm based on Fourier transform is applied to correct deformation deviations caused by curved surface projection. Subsequently, a structured light three-dimensional modeling method combined with image overlay stitching technology is used to fuse data collected from different perspectives to generate a complete three-dimensional structural model. This model not only presents the spatial distribution boundaries of different materials in the composite metal but also reflects the regional differences in interface bonding strength based on surface roughness, thereby generating wire interface data for the composite metal, including metal interface layer thickness, morphological curvature, and interlayer boundary clarity indicators, providing a basis for subsequent detection parameter selection.

[0040] Step S13: Determine the ultrasonic transmission position and parameters based on the wire interface data, transmit multi-band ultrasonic signals to the fixed composite metal wire, receive the echo signals from the wire interface, and calculate the ultrasonic response parameters based on the echo signal intensity, phase, and delay time.

[0041] Based on the wire interface data extracted in step S12, this embodiment of the invention identifies material interface regions and potential structural discontinuities in a three-dimensional model, determines that the ultrasonic emission points are located in these critical interface regions, and sets appropriate multi-band ultrasonic parameters. Specifically, a 5MHz to 15MHz tunable ultrasonic transducer is used for non-contact excitation, with the emission angle controlled at a 30-degree angle to the wire axis to enhance the interface wave coupling effect. During the acquisition process, the intensity (reflecting interface reflectivity), phase (reflecting material impedance matching), and delay time (reflecting interface thickness changes) of the echo signal are recorded using an array-type receiving transducer. Based on the formula for calculating sound wave propagation time and velocity, a set of ultrasonic response parameters corresponding to each emission point is comprehensively extracted to quantify the bonding status and internal defect clues at each interface location.

[0042] Step S14: Determine the eddy current detection path based on the wire interface data, scan the wire using a multi-frequency eddy current sensor, measure the electromagnetic field response changes at different locations on the wire, and calculate the electromagnetic conductivity distribution data based on the electromagnetic field response changes.

[0043] This embodiment of the invention continues to utilize the wire interface data from step S12, selecting areas with complex interface structures and significant material variations as key detection paths. Eddy current detection paths are designed, and multi-frequency eddy current sensors are deployed. A sweep frequency excitation method with a frequency range between 10kHz and 1MHz is used to perform a circular scan of the wire, with a detection period of 5 milliseconds for each scan point. Since different metal layers have different electromagnetic conductivities, the changes in induced voltage collected by the sensors during the detection process are used to infer their local electromagnetic properties. The response curve of the induced electric field as a function of frequency is calculated through Fourier analysis, thereby obtaining electromagnetic conductivity distribution data. This data expresses the spatial variation of electromagnetic properties in each region in the form of a two-dimensional matrix, reflecting latent defect characteristics such as inclusions, interface debonding, and material impurity distribution.

[0044] Step S15: Set X-ray scanning parameters based on wire interface data, use low-energy X-rays to perform tomographic scanning on the wire, and generate an interface density gradient map;

[0045] Referring to the interface model obtained in step S12, the X-ray scanning parameters for the wire are set, prioritizing low-energy X-rays (energy range of 20keV to 50keV) to enhance the imaging capability of interface density details while reducing radiation damage to equipment and materials. The wire is placed in a microfocus X-ray tomography scanner, with a rotation angle step set to 1 degree and a scanning resolution set to 5 micrometers, to acquire a complete attenuation image sequence of multiple cross-sections of the wire. A volume rendering method based on a Bayesian reconstruction algorithm is used to perform tomographic reconstruction of the slice data, generating a density gradient map. This map shows the changing trend of density transition zones between different metal layers, with density abrupt change regions typically corresponding to interface voids, lack of fusion, or areas of high internal stress accumulation. This map further improves the observation perspective of the interface physical state.

[0046] Step S16: Combine the ultrasonic response parameters, electromagnetic conductivity distribution data, and interface density gradient map to obtain the wire interface feature matrix;

[0047] In this embodiment of the invention, the ultrasonic response parameters (such as echo intensity spectra and time delay differences at each interface) in step S13, the electromagnetic conductivity distribution data (reflecting the spatial distribution of material conductivity differences) in step S14, and the interface density gradient map (describing density continuity) in step S15 are uniformly fused into a feature fusion process. The fusion method employs feature vector concatenation and normalization operations to ensure that all indicators are compared on the same scale, generating a wire interface feature matrix. Each row of this matrix corresponds to a record of the composite metal's multiple physical properties at a specific cross-section or scanning path of the wire, including but not limited to local bonding strength indicators, electromagnetic shielding capability assessment values, and structural consistency scores, forming a data vector set with dozens of characterization dimensions, which serves as the foundational data for subsequent modeling.

[0048] Step S17: Construct a wire interface feature model based on the wire interface feature matrix.

[0049] Based on the wire interface feature matrix generated in step S16, this embodiment of the invention constructs a wire interface feature model. It is trained using a self-organizing map (SOM) neural network from machine learning. SOM can map a high-dimensional feature matrix to a two-dimensional topological graph in an unsupervised manner, reflecting the similarity and anomalies of the interface structure in different regions. During training, each feature vector is used as an input node to the network, and the network gradually clusters the spatial distribution feature patterns of the wire interface through weight adjustments. The final constructed wire interface feature model has visualization capabilities and can mark abnormal clustering areas as potential defect areas of interest, providing a structured and scalable data representation framework for subsequent defect identification, model correction, and performance evaluation.

[0050] Of particular importance, step S15 includes the following steps:

[0051] Step S151: Analyze the material composition information in the wire interface data, determine the optimal X-ray energy range and dose parameters, and generate X-ray scanning configuration data;

[0052] This invention analyzes previously collected wire interface data, focusing on extracting information about different constituent materials in metal composite wires, such as the types and relative thicknesses of common composite materials like aluminum, copper, and nickel. Based on the differences in X-ray penetration capabilities in different metals, an energy spectrum calculation model is used to select the optimal X-ray energy range. For example, for a composite wire with an aluminum outer layer and a copper core, to ensure X-rays can penetrate the copper core without overexposing the aluminum, an energy range of 80 to 120 kiloelectron volts is selected, with the dose controlled at 3 mA per second to prevent material heating and image overexposure. Simultaneously, considering the wire's maximum outer diameter (e.g., 2 mm) and scanning time requirements (not exceeding 20 seconds), the scanning speed and exposure time are set, ultimately generating "X-ray scan configuration data" containing radiation energy, dose, voltage, current, and time window, which serves as the basis for setting subsequent detection system parameters.

[0053] Step S152: Adjust the emission angle and spot size of the X-ray source according to the X-ray scanning configuration data to adapt to the geometric characteristics of the wire, thereby forming X-ray detection parameters;

[0054] Based on the X-ray scanning configuration data obtained in step S151, this embodiment of the invention adjusts the geometric emission parameters of the X-ray source in the detection system. Considering the slender cylindrical structure of the wire, to obtain high-resolution interface images, the emission angle of the X-ray source is set to a 360-degree circular scanning mode perpendicular to the wire axis. To accommodate the small diameter structure of the wire, the X-ray focal spot size (i.e., "spot size") is set between 0.1 mm and 0.2 mm to improve imaging clarity and interface recognition capability. In specific operation, an X-ray source with micro-focal point control capability is used, and the scanning path and focal point parameters are set in the control system to form X-ray detection parameters including the emission angle range, step size per angle, spot size, etc., to guide the subsequent scanning and acquisition process.

[0055] Step S153: Perform a circular rotational scan on the composite metal wire based on the X-ray detection parameters to collect X-ray transmission intensity data at different angles of the wire;

[0056] In this embodiment of the invention, based on pre-set X-ray detection parameters, a metal wire is fixed in a rotating platform and rotated 360 degrees around its own axis. An X-ray source and detector perform a circular scan. One frame of image data is acquired for every 1-degree rotation, for a total of 360 frames of X-ray transmission images. Each frame records the transmission intensity distribution along the X-ray direction at that angle. A high-sensitivity flat panel detector with a pixel resolution of 100 pixels per millimeter is used to ensure detailed capture of the interface structure. Dynamic exposure control is required during this process to ensure uniform exposure images at different thicknesses. The final X-ray transmission intensity data is represented as a two-dimensional grayscale image. The grayscale value in each frame corresponds to the attenuation degree of the ray at that location, thus indirectly reflecting the material density and interface conditions at that point.

[0057] Step S154: Denoise the X-ray transmission intensity data and enhance the contrast to obtain the original density image data of the internal structure of the wire.

[0058] This embodiment of the invention performs image denoising and enhancement processing on the X-ray transmission intensity image data acquired in step S153. First, a median filtering algorithm is used to remove high-frequency noise, especially filtering the image edge regions to preserve edge characteristics. Then, a contrast-limited adaptive histogram equalization (CLAHE) method is applied to enhance the grayscale contrast of the interface region, making the interface density abrupt change regions more clearly distinguishable in the image. During this process, image artifact suppression processing is also performed to eliminate stripe interference caused by mechanical offset during rotational scanning. The processed image set is the "original density image data," which can be used as input for subsequent tomographic reconstruction, preserving information on minute structural changes within the wire.

[0059] Step S155: Perform tomographic reconstruction calculations based on the original density image data to generate a density distribution tomographic map of the wire cross-section;

[0060] This invention utilizes the original density image data obtained in step S154 for tomographic reconstruction calculations. Specifically, it reconstructs the two-dimensional density distribution image of each cross-section of the wire from multi-angle projected images using a back-projection algorithm or filtered back-projection (FBP) method. Specifically, pixels with the same scanning angle in each frame are integrally back-projected along a set axis, while bandpass filtering is performed to suppress low-frequency drift, reconstructing several cross-sectional tomographic images. The image size is set to 512×512 pixels, achieving a spatial resolution of 20 micrometers per pixel. In a test of a composite aluminum-copper wire, a total of 200 cross-sectional density distribution tomographic images were generated, effectively reflecting the transition zone structure between the central copper core and the outer aluminum cladding of the wire.

[0061] Step S156: Extract the metal interface region from the density distribution tomographic map, analyze the density change characteristics of the materials on both sides of the interface, and construct the characteristic data of the interface density transition zone.

[0062] This invention extracts interface region information from tomographic reconstruction images, primarily identifying the geometric location and transition characteristics of the material interfaces. Using image gradient analysis, regions of abrupt grayscale changes are identified in the tomographic image, representing the boundaries between two metallic materials. Subsequently, grayscale variation curves within a certain width (e.g., ±50 micrometers) are extracted from the left and right regions adjacent to the interface, and their density variation trends are fitted to construct "interface density transition zone feature data." This data reflects the smoothness of density transition when materials transition from one metal to another, serving as a crucial basis for judging interface fusion quality and weld integrity. In specific applications, when the width of the interface density transition zone exceeds 100 micrometers, insufficient welding or interface embrittlement may occur.

[0063] Step S157: Calculate the density change rate of the interface region based on the interface density transition zone feature data, and generate an interface density gradient map.

[0064] This invention, based on interfacial density transition zone characteristic data, calculates the rate of density change with location, i.e., the density gradient, using numerical differentiation methods. The density change rate at each point is calculated using central difference or third-order finite difference methods, thereby constructing an "interfacial density gradient map" for the entire interfacial region. This map is presented as a two-dimensional thermogram, with high gradient regions typically representing locations of abrupt material interface changes or defects. In specific examples, if the density gradient of a certain region exceeds a set threshold (e.g., a change rate greater than 500 kg / m³ per millimeter), it is marked as a potential anomaly point, which can be subsequently located and confirmed using other physical detection methods. This interfacial density gradient map not only provides quantitative basis for subsequent interfacial feature analysis but is also used in process control to evaluate welding process stability and material compatibility.

[0065] Preferably, step S17 includes the following steps:

[0066] Step S171: Perform data standardization processing on the wire interface feature matrix, incorporate different physical quantities into a unified measurement system, and generate standardized interface feature data.

[0067] In this embodiment of the invention, after obtaining the "interface density gradient map" generated in step S157, key feature data for subsequent analysis and modeling are first extracted to form a "wire interface feature matrix". This matrix typically includes multiple physical quantity dimensions, such as: interface density transition zone width, maximum density gradient value, mean density gradient, transition region symmetry index, interface position offset in the tomographic map, interface roughness factor, etc. Each row represents a wire sample, and each column represents a feature dimension. Because the physical units of these feature dimensions differ significantly—for example, density is measured in kilograms per cubic meter, position in micrometers, and gradient is the ratio of density to length—direct use in modeling and analysis would result in inconsistent dimensions and dimensional imbalance. To address this issue, the feature matrix needs to be standardized. The Z-score standardization method is used, which involves subtracting the sample mean from each column of feature data and dividing by the standard deviation, so that each feature dimension has a mean of 0 and a unit standard deviation after standardization. This process is implemented using the Pandas library in Python combined with the StandardScaler module in Scikit-learn. Taking a batch of composite aluminum-copper wire as an example, the original range of the interface roughness factor is between 0.2 and 3.8, while the density gradient value is between 500 and 3000. Without standardization, the model will be more biased in explaining the features with a large range of variation, leading to a weight bias. After standardization, the physical characteristics of the above different units are unified into dimensionless standard values, thereby generating "standardized interface feature data," which provides a reliable and balanced data input basis for subsequent interface quality assessment models or cluster analysis.

[0068] Step S172: Establish a three-dimensional coordinate system based on standardized interface feature data, and map the wire geometry information to the spatial coordinate system to form wire interface spatial mapping data;

[0069] In this embodiment of the invention, the wire interface feature matrix obtained in step S16 is standardized. Specifically, the numerical ranges of various data types, including ultrasonic response parameters, electromagnetic conductivity distribution data, and interface density gradient maps, are unified, allowing data from different physical quantity dimensions to be spatially mapped on the same analytical scale. Subsequently, a three-dimensional coordinate system is established with the wire's central axis as the Z-axis and the cross-sectional direction as the XY-axis. Each type of standardized interface feature data is mapped to the three-dimensional coordinate system according to its spatial location, forming "wire interface spatial mapping data" with spatial positioning attributes. This data can be represented in sparse tensor form, with each tensor unit recording multiple interface feature values ​​at that spatial point. In a specific application scenario, for a composite metal wire with a diameter of 1.2 mm and a length of 200 mm, the spatial coordinate system is divided into 200 equally spaced sections. Each section is further divided into 36 angular directions and 10 radial layers, resulting in a total of 72,000 spatial mapping points.

[0070] Step S173: Construct the basic topology of the interface using the wire interface space mapping data, determine the interface continuity and morphological characteristics, and generate interface topology data.

[0071] This invention utilizes the constructed spatial mapping data and employs regional connectivity analysis and isosurface reconstruction methods to extract continuous interface feature regions. A three-dimensional boundary tracking algorithm identifies the boundary surfaces of different interface morphologies, establishing a basic interface topology. The "basic interface topology" refers to a graph structure model reflecting the overall morphology of the interface, constructed based on the connectivity, adjacency, and curvature information of the interface features of different materials in the wire. During this process, the characteristic gradient changes between each spatial point and its neighboring points are calculated to determine whether they belong to the same topologically connected region, thereby dividing different interface topological blocks. In practical applications, if the electromagnetic conductivity gradient exceeds a set threshold of 0.2 Siemens / mm, it is considered a boundary point between different interfaces. The interface topology data constructed in this way can effectively reflect the continuity, nesting, and local abnormal deformation of the material interfaces.

[0072] Step S174: Analyze the spatial distribution pattern of interface features based on the interface topology data, identify the feature change patterns of interface regions, and form interface region feature clustering data.

[0073] This invention employs spatial statistical modeling based on interface topology data. It utilizes density clustering algorithms (such as density-based spatial clustering DBSCAN) to identify groups of regions in three-dimensional space exhibiting similar trends in interface feature changes; these are termed "interface region feature clustering data." During clustering, the interface feature vectors of each point are used as input, and the Euclidean distance between points and a feature gradient change threshold are combined to determine whether they belong to the same class. In specific implementations, regions with continuous interface density changes of less than 5% in the thickness direction but significant density abrupt changes (exceeding 15%) in the lateral direction are clustered into one class to identify potential delamination, cracks, and dissimilar metal contact defects. The final output is a structured dataset containing cluster center locations, region shape descriptions, and corresponding interface type labels.

[0074] Step S175: Based on the interface region feature clustering data, filter the key feature parameters that are sensitive to interface quality to obtain the set of key interface feature parameters;

[0075] This invention further performs sensitivity analysis on the clustering results of the identified interface regions, selecting feature dimensions that significantly affect changes in quality indicators such as interface strength, interface integrity, or interface corrosion resistance as key feature parameters. For example, principal component regression (PCR) is used to analyze the correlation between each interface feature and the experimental results of interface strength of known samples, identifying ultrasonic phase delay, electromagnetic conductivity gradient, and density abrupt change rate as the main influencing factors. The resulting "set of key interface feature parameters" is the core variable set used for interface quality modeling, and in practice includes parameters such as "location of maximum interface conductivity abrupt change," "segment with maximum interface thickness change rate," and "length of minimum interface density gradient," totaling approximately 15 dimensions.

[0076] Step S176: Construct a feature vector space using the set of key interface feature parameters, and perform dimensionality reduction processing to generate dimensionality-reduced interface feature data.

[0077] This invention utilizes the aforementioned set of key interface feature parameters to construct a feature vector space, where each sample corresponds to a vector composed of 15 feature parameters. To improve model efficiency and reduce dimensionality redundancy, Linear Discriminant Analysis (LDA) is employed for dimensionality reduction, balancing discriminative power and information integrity to generate 3-5 dimensionality-reduced interface feature data. In practice, covariance matrix analysis is performed on the feature parameters to extract principal axis directions, and the top few principal axes are selected based on an information retention rate of no less than 95%. The dimensionality-reduced feature data is readily available for subsequent mathematical modeling and physical simulation, improving analytical efficiency and accuracy.

[0078] Step S177: Establish a mathematical model of the wire interface based on the dimensionality reduction data of the interface features, and impose physical property constraints on the mathematical model of the wire interface to obtain the feature model of the wire interface.

[0079] This invention constructs a mathematical model, or "wire interface mathematical model," based on dimensionality reduction data of interface features to describe the spatial characteristics and physical properties of wire interfaces. This model can be a multivariable function combination, describing the functional relationship between key parameters of the interface in its spatial distribution and the interface stress field and bonding strength. Next, physical constraints are applied to the model based on the physical properties of the metal, such as interface continuity constraints (i.e., the interface strength function is continuous within the topological structure) and material mechanics constraints (combining Hooke's law and the interface bonding model), ensuring that the constructed model has practical physical meaning. The resulting "wire interface feature model" can be used for applications such as interface performance prediction, production process control, and defect location. In the quality assessment of a certain aluminum-copper composite wire, the model's prediction of the interface debonding zone showed a consistency rate of over 93% with the actual ultrasonic detection results, demonstrating strong practicality and accuracy.

[0080] Preferably, step S2 includes the following steps:

[0081] Step S21: Establish the benchmark reference area for the wire interface feature model, select an interface segment without obvious defects as the reference standard, and generate wire interface benchmark feature data.

[0082] After constructing the interface feature model (as shown in step S177), this embodiment of the invention first needs to define a representative benchmark reference area. This area is typically selected from sample segments of a large batch of wires that have been confirmed by manual or automatic inspection to have no obvious defects, uniform interface structure, stable density gradient, and good surface continuity. In specific implementation, a sample screening method can be used, combining the previously standardized feature data and the original tomographic image, using the stable value range of features as screening conditions (e.g., density gradient between 800 and 1200, interface roughness factor less than 1.0), and combining the three-dimensional image to judge the interface continuity. Finally, 5 to 10 length segments are selected from each wire for manual annotation and confirmation, and their interface feature model parameters are extracted after merging. The average of these parameters forms the wire interface benchmark feature data, which will serve as the reference template for all subsequent inspection samples.

[0083] Step S22: Compare each detection area in the wire interface feature model with the wire interface baseline feature data point by point, calculate the feature difference value, and form the original differential data of the wire interface.

[0084] In step S22 of this embodiment of the invention, the reference data generated in S21 is used as a guide to sequentially traverse all detection areas in the wire interface feature model, and a point-to-point feature comparison mechanism is constructed for each area. This mechanism refers to comparing each dimension of the standard feature vector with the feature vector of the area to be tested at the same position or at equal intervals, and calculating the feature difference value. The difference value can be calculated using either absolute difference or relative difference, depending on the detection purpose. Taking aluminum-copper composite wire as an example, if the detection target is to identify slight delamination or interface discontinuity, relative difference is more sensitive. Finally, all detection segments generate a set of original differential data of the wire interface, with the data structure being a one-dimensional or two-dimensional matrix, where each item represents the degree of difference between a certain detection position and the reference standard.

[0085] Step S23: Perform statistical normalization on the original differential data of the wire interface and perform neighborhood enhancement processing to obtain enhanced differential data of the wire interface;

[0086] To eliminate misjudgments caused by deviations in individual regions, this invention requires statistical normalization of the original differential data of the wire interface. This process standardizes the original difference values ​​according to the overall distribution of the regions, unifying them to a range of 0 to 1 or -1 to 1. A common approach is to subtract the mean from the difference data and then divide by the standard deviation or maximum value, ensuring the comparability of difference values ​​between different batches. After normalization, to highlight the spatial continuity and regional clustering effect of the difference signal, neighborhood enhancement processing is also implemented. This involves using a sliding window to calculate the mean or median of the difference in local regions on the difference map, thereby enhancing regional signals with spatial consistency characteristics. The final output of enhanced differential data for the wire interface is a spatially enhanced representation of the original differences, which can more accurately reflect the significance of potential abnormal regions at the interface.

[0087] Step S24: Set multi-level threshold conditions based on the wire interface enhancement differential data, and perform hierarchical filtering of differential signals to form wire interface differential signal conversion data;

[0088] This invention employs a tiered processing method to enhance the strength of differential data. Specifically, it sets multi-level threshold conditions, for example, three levels: low difference (less than 0.3), medium difference (0.3–0.6), and high difference (greater than 0.6). The spatial distribution of each level of difference is then labeled, forming a tiered signal map. In practical applications, the setting of multi-level thresholds is based on extensive prior statistical analysis and experimental verification of numerous samples to ensure the ability to detect both minor anomalies and severe defects. Subsequently, differential signal conversion is performed on the difference regions at each level, converting the difference signals into labeled results that indicate the anomaly level at that location. This yields differential signal conversion data for the wire interface, providing a discretized basis for subsequent pattern recognition and defect determination.

[0089] Step S25: Identify regions with similar change patterns based on differential signal conversion data of the wire interface, thereby extracting the change features of the wire interface;

[0090] This invention utilizes differential signal conversion data and applies clustering analysis methods such as K-means or DBSCAN to extract groups of regions with similar variation patterns. The core of this step is to find spatially continuous or similarly trending regions to determine the presence of systematic interface anomalies. The specific process includes: extracting the spatial location, intensity of difference, and morphological characteristics (such as length, area, and orientation) of each differing region, and performing multi-dimensional clustering to ultimately form several interface variation pattern groups. These groups are considered candidate regions for potential defect evolution trends at the wire interface, and their statistical characteristics will be used to further assess the reliability and risk level of the defects.

[0091] Step S26: Perform spatial mapping based on the characteristics of wire interface changes, determine the boundaries and ranges of high-confidence defect areas, and construct a preliminary indicator map of wire interface defects.

[0092] In this embodiment of the invention, the interface change groups identified in step S25 are mapped in three-dimensional space. This involves restoring the two-dimensional differential analysis results to the initial interface feature model or three-dimensional tomographic map. By calculating the spatial boundary, volume, and relative position to the baseline region for each change group, it is comprehensively determined whether it belongs to a high-confidence defect region. If a group exhibits high degree of difference, strong spatial extensibility, and a structural morphology consistent with typical defects (e.g., extending in a strip-like pattern along the interface direction), its boundary is extracted to form a preliminary indicator map of wire interface defects. This map is an important intermediate result for wire interface quality assessment and can be used in various application scenarios such as subsequent manual review, automatic defect annotation, or machine learning model training.

[0093] Preferably, the multi-scale decomposition processing based on the preliminary indicator map of wire interface defects in step S3 includes:

[0094] Multi-resolution segmentation is performed on the preliminary indicator map of defects at the wire interface, dividing the image into detection areas at macro, meso, and micro scales to generate multi-level defect decomposition data.

[0095] Morphological dilation processing is performed on the macroscopic detection area to enhance the continuity characteristics of interface cracks and generate macroscopic defect connected domain data.

[0096] Adaptive threshold segmentation is performed on the mesoscale detection region to separate grain boundary slip bands from second-phase precipitation features, generating mesoscale defect separation data.

[0097] Laplacian sharpening filtering is applied to the microscale detection area to generate enhanced data of micro-defects.

[0098] Multi-level defect decomposition data, macroscopic defect connected domain data, mesoscopic defect separation data, and microscopic defect enhancement data are input into a multi-channel data fusion network to generate cross-scale defect correlation data.

[0099] A three-dimensional defect space mapping is constructed based on cross-scale defect association data, the distribution density of defects along the X / Y / Z axes is labeled, and a defect space distribution vector is generated.

[0100] Principal component analysis is performed on the spatial distribution vector of defects to generate a dataset of core defect features.

[0101] The core feature dataset of defects is matched with a pre-defined defect type database using cosine similarity. Feature combinations with a similarity greater than 85% are selected to generate interface defect feature data.

[0102] In this embodiment, the preliminary indicator map of wire interface defects obtained in step S26 is used as input. This map includes the three-dimensional coordinates and image intensity values ​​of the interface defect region. To better identify defect features at different scales, a strategy based on image pyramids and multi-scale structural analysis is adopted to decompose the entire image into three levels in terms of spatial resolution: the macro-scale resolution is 1 mm, used to observe the overall structural discontinuity and interface deformation trend; the meso-scale is 200 micrometers, used to identify grain boundary distortion or small-sized precipitates in the microstructure; and the micro-scale is 50 micrometers or less, used to capture fine cracks, microvoids, or atomic-level dislocation concentration areas. The segmentation method is based on sliding window block division with spatial scale constraints, and generates a corresponding set of sub-images by combining the spatial resolution of each level. Each sub-image has a label of its scale. The final output multi-level defect decomposition data includes the image block index, spatial coordinate range, scale label, and preliminary defect intensity distribution information at each scale, which are used in the subsequent multi-scale feature extraction process. The set of macro-scale detection region image blocks generated in the previous step is selected, and morphological dilation processing is performed on the initial contour of the crack in these regions to improve their structural coherence. Morphological dilation is a structural enhancement technique that improves the continuity and connectivity of the fracture path by expanding pixels around the crack edge. In this example, a 5×5 structuring element is selected as the dilation kernel. This elliptical structuring element adapts to the changing propagation direction of the interface crack. During the operation, each macroscopic image patch is binarized to extract the defect boundary, and then a dilation operation is performed to determine whether the connected domain formed after dilation crosses the main axis of the interface. If it does, its start and end coordinates, connectivity length, and connectivity strength value are recorded to generate a macroscopic defect connected domain dataset. This dataset can be used to subsequently determine whether the crack has formed a macroscopic visible defect channel, which has important reference value in the fatigue life prediction of metal wires. The mesoscopic-scale image patches divided in the previous step are processed. Each image patch covers an area of ​​approximately 200 micrometers and is mainly used to detect abnormal phenomena in the microstructure of materials. For grain boundary slip zones (often manifested as regions with directional variations in grayscale intensity) and second-phase precipitates (often manifested as local bright or dark spots) at this scale, this example employs an adaptive threshold segmentation method based on local grayscale statistics. Specifically, a sliding window process (window size 20×20 pixels) is applied to each image block, calculating the local mean and local variance. The adaptive threshold is then dynamically determined using the local mean plus or minus twice the standard deviation. Subsequently, the image is divided into background, slip zone, and precipitate regions based on the threshold. The output mesoscopic defect separation data is saved as a label map, indicating the location, size, and orientation of the three types of regions. This data is particularly suitable for evaluating the microstructural stability and interfacial composition compatibility of the heat-affected zone during interface processing. For microscale detection image blocks, the resolution of each image is 50 micrometers or less.Since defects at the microscale often manifest as low-contrast microcracks or lattice perturbations, traditional enhancement methods struggle to clearly extract edge features. Therefore, a Laplacian sharpening filter algorithm is employed. This algorithm, based on the second derivative edge enhancement mechanism, strengthens edge responses by detecting locations with high rates of change in the grayscale image. A 3×3 Laplacian operator convolution kernel is used, combined with the original grayscale image for filtering. The output is then weighted and fused with the original image (weights set to 0.7 for the original image and 0.3 for the sharpened image) to enhance defect edges while preserving the overall image structure. The microscale defect enhancement data records the spatial location index, enhancement weight parameters, and significant edge region masks of the enhanced image, used for subsequent identification of the boundary morphology features of submicron-level defects, particularly applicable in electron beam detection and X-ray computed tomography. A multi-channel convolutional fusion network (MCFN) is used to uniformly model the defect feature information at each scale. The network structure comprises four input channels, corresponding to macroscopic image patches, mesoscopic image labels, microscopic enhanced images, and scale decomposition metadata, respectively. Each channel extracts features through independent convolutional layers, and these features are then concatenated in an intermediate fusion layer. Two attention mechanisms are then used to identify cross-scale relationships, ultimately outputting a fused cross-scale feature map. During network training, manually annotated defect maps are used as supervision signals. The training loss function incorporates cross-entropy and structural similarity error terms to optimize edge fidelity and structural consistency across different scales. The cross-scale defect association data records the spatial distribution map, eigenvalue matrix, and saliency scores of each defect region across multiple scales, serving as the core input for subsequent 3D mapping and type recognition. The defect association data output by the fused network is matched with the initial 3D coordinate system, remapping each defect region onto the original wire voxel mesh to construct a defect point cloud map with spatial location attributes. By statistically analyzing the spatial coverage and density variation of each defect in the X, Y, and Z directions (unit: ratio of defect points to spatial volume), density vectors are generated along the three axes. The length of each vector equals the number of spatial units in the corresponding direction, and the vector value represents the defect point density within that unit. These vectors are then integrated to form a defect spatial distribution vector, formatted as a triplet array, recording the start and end positions, peak density, and standard deviation of high-density regions in each direction. In practical applications, this vector can be used to determine spatial evolution information such as defect aggregation direction and crack propagation tendency. The three-dimensional defect distribution vector obtained in the previous step is then dimensionality-reduced using Principal Component Analysis (PCA).The process begins by concatenating the density vectors in the X, Y, and Z directions into a long vector set, which is then centered and standardized. Next, the covariance matrix is ​​calculated to extract the first three principal component directions, retaining information that explains over 90% of the variance in the core feature data. The generated defect core feature dataset includes principal component vector coefficients, the projection values ​​of the original vectors into the principal component space, and the explanatory power of each principal component. In practical analysis, this dataset can be used to extract the most representative defect structural features, serving as crucial data input for constructing defect fingerprints or training defect recognition models. A pre-constructed defect type database is introduced, encompassing over ten defect types, including common interface delamination, metallurgical inclusions, thermal stress cracks, and grain boundary mismatches. Each defect type is represented by its typical principal component features. During the matching process, cosine similarity is used as a metric. The cosine of the angle between the current defect core feature vector and each defect vector in the database is calculated. Types with a similarity greater than 85% are selected as preliminary matching results, and their corresponding labels, similarity scores, and principal component contributions are recorded. The final generated interface defect feature data includes not only category labels, but also defect strength level, feature stability evaluation and matching confidence, which has important application value in scenarios such as quality traceability, failure analysis and process optimization.

[0103] Preferably, the step S3 of identifying the wire interface defect type characteristics based on interface defect feature data includes:

[0104] Extract the feature values ​​of defect length, width, depth and area ratio from the interface defect feature data to generate a basic geometric feature set of defects;

[0105] The geometric feature set of defects is compared one by one with the standard defect geometric templates in the defect type database to calculate the geometric similarity.

[0106] Based on geometric shape similarity, a candidate set of defect types with a matching degree >75% was selected;

[0107] Based on the candidate set of defect types, the density distribution of defects, the spacing between adjacent defects and the arrangement direction of defects in the material interface region are calculated to generate a spatial distribution feature vector of defects.

[0108] Based on the spatial distribution feature vector of defects, defect groups are divided according to density peak detection to generate defect spatial clustering data;

[0109] Based on defect spatial clustering data, defect edge morphological features are extracted, and edge roughness, fractal dimension and curvature change rate are quantified to generate a defect edge feature parameter set.

[0110] A secondary matching is performed between the defect edge feature parameter set and the defect type database to correct misclassification items in the preliminary defect classification data and generate optimized defect classification results.

[0111] Based on the preset defect type determination rules, the optimized defect classification results are determined by rules, and wire interface defect type features containing defect type codes, confidence levels and location coordinates are generated.

[0112] This invention employs an image measurement algorithm to quantify the geometric projection contour of defects in an image. Specifically, a contour analysis method based on boundary extraction is used to extract the maximum extension direction length of the defect as the defect length and the maximum width perpendicular to it as the defect width. Further, the defect depth is extracted by combining the defect extension information along the Z-axis in the three-dimensional defect spatial mapping data. Simultaneously, the ratio of the projected area of ​​the defect in the entire wire interface image to the total area of ​​the interface region is calculated to obtain the defect area percentage. Based on this, a basic geometric feature set of defects is constructed, covering length (unit: micrometer), width (unit: micrometer), depth (unit: micrometer), and area percentage (percentage), for subsequent geometric template matching analysis. Based on the generated basic geometric feature set of defects, this embodiment introduces preset defect geometric templates from a standard defect type database, such as three-dimensional geometric structure templates for typical grain boundary cracks, second-phase cluster precipitation, and slip zone cracking. A matching algorithm combining Hausdorff distance and shape context descriptors is used to calculate the geometric similarity of each defect sample. The specific operation involves first standardizing the dimensions of the defect samples and templates, then extracting their boundary shape distribution and angle information, and fitting them using a shape distribution function to obtain a similarity score between each pair of defect samples and templates. Templates with a similarity exceeding 75% are selected as a preliminary candidate set for defect types. After selecting a candidate set with a geometric shape matching degree greater than 75%, the embodiment calculates the distribution density of each type of defect per unit area of ​​the image based on the spatial coordinate information of the defects in the material interface region. Simultaneously, Euclidean distance is used to measure the center distance between defects to obtain the spacing between adjacent defects, and then the overall arrangement direction angle of the defects is calculated using a principal direction vector fitting algorithm. This method uses Local Principal Component Analysis (Local PCA) to solve for the direction vectors of multiple defect point clouds, forming a set of feature vectors reflecting the spatial structure distribution of defects, including density values ​​(number per square millimeter), average spacing (micrometers), and principal direction (degrees), serving as a structural representation of the spatial distribution of defects. Subsequently, in this embodiment, cluster analysis is performed on the aforementioned defect spatial distribution feature vectors. A density peaks clustering (DPC) method is used to identify density centers and outliers, resulting in defect grouping. This method calculates the minimum distance between each defect point and samples with higher density, thereby identifying cluster centers and achieving natural clustering of defects within the interface region. For example, in a 200×200 micrometer wire interface image, there are approximately 30 spatially continuous defect points. This algorithm can identify this as a high-density clustering region, generating defect spatial clustering data containing multiple cluster category labels.Based on the generated defect spatial clustering data, the embodiment further extracts the edge morphology features of each defect group, calculating edge roughness (assessing contour smoothness through boundary fluctuation coefficient), fractal dimension (measuring boundary complexity using box dimension method), and rate of curvature change (reflecting the degree of drastic shape change based on the standard deviation of continuous curvature change), quantifying the edge structural complexity of each defect region. These features help distinguish different defect types such as grain boundary cracks (regular edges) and micro-corrosion points (rough edges, abrupt curvature changes), forming a defect edge feature parameter set for further classification optimization. After obtaining the defect edge feature parameter set, the embodiment performs a secondary comparison with the edge morphology templates in the aforementioned standard defect type database. A support vector machine (SVM) classifier combined with the K-nearest neighbor algorithm is used to correct misclassifications of low-confidence samples in the initial classification results, especially for defects that are geometrically similar but have significantly different edge features, such as slip band cracks and grain corrosion, which have similar lengths but significantly different fractal dimensions. After comparison, the defect classification results are recalibrated to generate an optimized defect classification result dataset. In this embodiment, based on the preset defect type determination rules in the metal wire defect identification system (e.g., crack defects require a length greater than 50 micrometers and a consistent distribution direction, while corrosion defects have an area ratio of less than 3% but large edge roughness), the optimized defect classification results are determined by rules, and samples that meet the rule matching are labeled as specific defect types. The output is structured result data containing defect type codes (e.g., CRK represents crack), identification confidence (e.g., 0.93), and location coordinates (e.g., x = 112.4, y = 85.2), forming wire interface defect type features that can be used for subsequent quality control and defect prediction modeling.

[0113] Most importantly, the specific rules for determining defect types are as follows:

[0114] If the aspect ratio of the defect is >5, the orientation angle is ∈[60°,120°], and the cluster label is a continuous group, then it is marked as an interface through crack;

[0115] If the defect area ratio is >3%, the edge roughness is >2.5μm, and the cluster density is <3 clusters / mm 2 If so, it is marked as localized material peeling;

[0116] If the defect spacing is <0.1mm, the fractal dimension is >1.6, and the cluster density is >5 clusters / mm 2 If it is, it is marked as diffuse porosity aggregation.

[0117] In this embodiment of the invention, the geometric shape characteristics of the defect are calculated based on its aspect ratio, which is obtained by measuring the ratio of the longest and shortest axes of the defect. In this example, if this value is greater than 5, and the angle of the defect's orientation is between 60° and 120°, and the cluster label of the defect is a continuous group, then the morphology of the defect corresponds to an interface penetration crack. Interface penetration cracks typically exhibit long lengths and strong directional consistency. To further determine the orientation of the defects, this embodiment uses rotation-invariant feature descriptors (such as Hough transform) to calculate the main orientation of the defects and determine whether it falls within a specified directional range. Based on this information, combined with the group connectivity identified in cluster analysis, the defect can be accurately labeled as an interface penetration crack, which is a typical defect in the metal wire processing process. For the identification of localized material delamination, this embodiment first calculates the area ratio of the defect, i.e., the proportion of the defect area in the entire image. If this area ratio is greater than 3%, and the edge roughness of the defect exceeds 2.5 μm, while the cluster density is less than 3 clusters / mm, then the defect is considered a valid defect. 2 If the defect is identified as localized material peeling, then it is determined to be a defect. Edge roughness is obtained by contour analysis of the defect edge. The specific calculation method adopts an algorithm based on edge fluctuation analysis, which uses wavelet transform to decompose the defect contour and obtain roughness features. Cluster density is obtained by calculating the number of defects per unit area within the defect group. In this scenario, localized material peeling often manifests as material surface peeling or corrosion, so the edge complexity is high, but the density is low, and the peeling area is usually large, which meets the above criteria. Through the above steps, the embodiment can accurately identify these defect types and label them. For the identification of diffuse porosity clusters, the embodiment first calculates the minimum distance between defects, which can be obtained by calculating the Euclidean distance of the center points of all defects. If the defect spacing is less than 0.1 mm, the fractal dimension is greater than 1.6, and the cluster density is greater than 5 clusters / mm, then the defect is considered a localized defect. 2 If the fractal dimension is calculated, the defect can be identified as a cluster of diffuse pores. The fractal dimension is obtained by calculating the complexity of the defect profile, using the box-count method, where the details and distribution complexity of the defect boundary are quantified into a numerical value. Clusters of diffuse pores typically exhibit a large number of small pores tightly packed together in a localized area. These pores usually have a high aggregation density; therefore, density and spacing analysis are key indicators during detection. In this example, combining the characteristics of spacing and fractal dimension effectively distinguishes these clusters of small pores and allows for accurate classification.

[0118] Preferably, step S3, which involves matching and correcting the characteristics of the wire interface defect type with a preset physical constraint model, includes:

[0119] Establish a physical constraint model library for wire interface defects, including physical property parameters and constraint conditions for peeling, porosity and inclusions;

[0120] The defect type characteristics of wire interface are compared with the physical constraint model library of wire interface defects by parameterization to generate preliminary defect type matching data.

[0121] Based on the preliminary matching data of defect types, the physical rationality of defect characteristics is assessed, and the constraints are verified to form a physical rationality assessment of the defects.

[0122] The physical rationality assessment of defects is modified based on physical constraint rules to obtain defect feature correction data.

[0123] Based on the defect feature correction data, quantitative defect characterization parameters including defect size, depth, shape and distribution characteristics are calculated to generate a set of quantitative defect characterization parameters;

[0124] The defect evolution law is matched with the defect evolution law in the physical constraint model library of wire interface defects based on the quantitative characterization parameter set of defects, and the defect formation mechanism and development trend are evaluated to generate accurate defect characterization data.

[0125] This invention establishes a physical constraint model library for wire interface defects. This library contains physical characteristic parameters and related constraint conditions for various types of metal wire interface defects (such as delamination, porosity, and inclusions). Specifically, the physical characteristics of delamination defects may involve their adhesion strength, area, depth, and mechanical behavior to the matrix material; porosity defects involve porosity, pore size distribution, and their impact on the material's mechanical properties; inclusion defects include the morphology, size, distribution, and impact on material strength and corrosion resistance. These defect characteristics are concretized into physical constraint models to aid in the evaluation and correction of defect features in subsequent steps. This model library not only includes conventional defect morphological characteristics but also covers relevant physical constraint conditions, such as the material's elastic modulus, plastic deformation behavior, and shear strength of the material interface. Next, this embodiment generates preliminary defect type matching data by parametrically comparing the wire interface defect type characteristics with the physical constraint model library. In this process, the physical characteristics of the defects are first extracted, including the defect's size, morphology, distribution density, and edge roughness. Then, these characteristics are matched with the parameters defined in the defect physical constraint model library. For example, for peeling defects, the matching conditions might include the ratio of the defect area and depth to the matrix interface strength; for porosity defects, the relationship between porosity and the material's compressive strength is considered. This comparison generates preliminary defect type matching data, identifying the defects that best match known defect types. After obtaining the preliminary matching data, the embodiment further performs a physical rationality assessment of the defect features. At this point, based on the physical rationality assessment, it checks whether the physical characteristics of each defect meet the corresponding physical constraints. For example, if the depth and width of a defect exceed the limits that the material's mechanical properties can withstand, or if the shape of some defects does not conform to the assumptions in the known physical constraint model, these defects will be marked as unreasonable. This step is mainly used to screen defect features that do not conform to physical laws, ensuring the accuracy of subsequent data analysis. Next, based on the results of the physical rationality assessment, the embodiment corrects the defect features under physical constraint rules. This step corrects the initially matched defect features by applying physical constraints. For example, if the shape of a defect is shown as unreasonable in the physical constraint model, the data can be corrected by adjusting the edge shape or size of the defect to conform to the physical constraint rules. The corrected defect data better reflects actual material behavior, enhancing the accuracy and reliability of subsequent defect analysis. After obtaining the corrected data for defect characteristics, the embodiment further calculates quantitative characterization parameters of the defects. These parameters include defect size (e.g., length, width, depth), morphology (e.g., aspect ratio, edge smoothness), and distribution characteristics (e.g., uniformity or aggregation of defects on the material surface). These quantitative parameters can be extracted using morphological analysis, geometric parameter calculation, and statistical analysis methods.In practical applications, different image analysis techniques can be combined. For example, edge detection algorithms can be used to extract the geometric shape of defects, and cluster analysis can be used to evaluate the distribution characteristics of defects, ultimately resulting in a quantitative characterization parameter set containing the geometric shape and distribution of defects. Finally, the example matches the quantitative characterization parameter set with the defect evolution laws in the physical constraint model library of wire interface defects. This step mainly involves modeling and simulating the evolution process of defects to analyze the formation mechanism and possible development trend of defects in metal wires. For example, peeling defects may gradually expand over time, while porosity defects may merge under certain conditions to form larger pores. By matching the evolution law of defects with the physical constraint model, the formation mechanism of defects can be evaluated, and their possible expansion or development trend can be predicted. Ultimately, the generated accurate defect characterization data can provide strong support for wire quality control, defect repair, and optimization.

[0126] Preferably, step S4 includes the following steps:

[0127] Step S41: Spatial mapping of the defect characterization data based on the axial and radial coordinates of the material to generate wire defect spatial location data containing three-dimensional defect distribution information;

[0128] This invention requires converting defect feature data on the interface of a metal wire into a coordinate system suitable for three-dimensional spatial positioning. Specifically, this involves mapping the lateral position, longitudinal position, and depth values ​​(such as depth, width, and length) of each defect into the three-dimensional coordinate system of the metal wire. By using three-dimensional spatial mapping technology, the spatial position of each defect on and within the material surface can be obtained. To achieve this process, precise three-dimensional geometric data of the metal wire may be acquired using equipment such as laser scanners, CT imaging, or X-ray imaging. The defect data is then precisely located based on its position and size, generating defect spatial positioning data containing three-dimensional coordinate information. This data will provide a foundation for subsequent defect distribution analysis, quality assessment, and process correlation.

[0129] Step S42: Based on the spatial location data of wire defects, perform layered display and density statistics on different types of defects to construct a wire interface defect distribution map that reflects the distribution pattern of various types of defects;

[0130] This invention, based on the spatial location data of wire defects obtained in step S41, performs layered display and density statistics on different types of defects at this stage, thereby constructing a wire interface defect distribution map reflecting the distribution patterns of various defects. Specifically, spatial clustering algorithms (such as K-means clustering or DBSCAN) are used to classify different types of defects. Based on the spatial location of each defect, the distribution density of different defects at the metal wire interface and in its depth direction is statistically analyzed. By constructing the defect distribution map, the concentration and distribution trends of different types of defects in space can be visually observed. This process typically requires analyzing the defect category, location, and defect density in the surrounding area to generate a map reflecting the defect distribution, such as a color-coded heatmap, identifying areas with dense and sparse defects. Based on the spatial distribution patterns of different defects, engineers can assess which areas require special attention, which defect types are more common, and their potential impact on overall wire quality.

[0131] Step S43: Based on the defect distribution map of the wire interface, perform weighted calculations on the impact of various defects on wire performance, assess the overall quality level of the wire and local areas, and form the severity of the wire interface defects;

[0132] Based on the wire interface defect distribution map constructed in step S42, the next step in this invention is to perform a weighted calculation on the impact of various defects on wire performance to assess the overall and local quality level of the wire and determine the severity of the interface defects. In this process, it is first necessary to define the weight of each type of defect on the performance of the metal wire. For example, some defects (such as through cracks and localized peeling) may have a significant impact on the mechanical properties of the wire (such as tensile strength and impact resistance), while other types of defects (such as micropores) may have a smaller impact on performance. By combining materials science knowledge, a corresponding weight coefficient is assigned to each defect, and then a weighted calculation is performed based on the distribution of defects on the surface and inside the metal wire to obtain the defect severity of each region. This method can assess the overall and local quality level of the metal wire, providing a quantitative basis for subsequent quality control and optimization.

[0133] Step S44: Determine the characteristic pattern and severity of the current wire defect based on the severity of the wire interface defect, and generate the wire defect rating result;

[0134] This invention determines the characteristic patterns and severity of current wire defects based on the severity of interface defects, and generates a wire defect rating result. Specifically, based on the aforementioned calculated defect severity data, a classification algorithm (such as decision tree, support vector machine, etc.) is used to rate the wire defects. For example, if the defect severity of a certain section of wire exceeds a preset threshold, that section of wire may be marked as "severe defect," while another section with fewer defects can be rated as "minor defect." This rating method allows for a comprehensive understanding of the quality status of the wire under different production batches or process conditions, helping quality control personnel in the production process to promptly identify and resolve problems. The generated defect rating result will serve as standardized evaluation data for wire quality, providing support for subsequent product inspection, customer acceptance, and quality traceability.

[0135] Step S45: Perform correlation analysis between the wire defect rating results and the time series data of process parameters recorded during the wire production process to identify the correspondence between defect formation and changes in specific process parameters, and form wire production process correlation data.

[0136] This invention employs time-series matching between process parameters (such as temperature, stretching rate, and cooling rate) recorded during production and defect data. Statistical analysis methods (such as correlation analysis or regression analysis) are used to explore the potential relationships between process parameters and defect types and severity. For example, processing a batch of wire at higher temperatures may result in more porosity defects, while excessively rapid cooling may lead to an increase in peeling defects. This correlation analysis identifies key control parameters in the process, providing data support for optimizing production processes. Ultimately, this process correlation data helps the production team adjust process flows, reduce defect rates, and thereby improve the overall quality and production efficiency of the wire.

[0137] Preferably, step S5 includes the following steps:

[0138] Step S51: Construct a process parameter optimization model based on the wire production process correlation data, identify key process parameters that affect interface quality, and generate a process parameter priority list;

[0139] This invention uses the wire manufacturing process data obtained in step S45 as input, and employs statistical analysis and machine learning algorithms (such as random forest and support vector machine) to identify key process parameters affecting interface quality. These process parameters may include temperature, stretching rate, cooling rate, and pressure. Through model training, it is possible to identify which process parameters have a significant impact on wire defect formation and interface quality, and to rank the process parameters according to their degree of influence. By calculating the impact of each parameter on quality indicators such as defect type and defect severity, a priority list of process parameters can be generated, helping engineers focus on optimizing the parameters that have the greatest impact on quality, thereby improving the consistency and performance of wire products.

[0140] Step S52: Design parameter adjustment schemes based on the process parameter priority list, formulate process parameter optimization strategies based on different types of interface defects, and generate process parameter adjustment data;

[0141] This invention designs targeted process parameter optimization strategies based on different types of interface defects (such as peeling, porosity, inclusions, etc.) and their correlation with process parameters. For each defect type, the process parameter with the greatest impact is selected and adjusted. For example, if the cooling rate is found to have a significant impact on the formation of porosity defects, the cooling rate is adjusted; if temperature has a significant impact on peeling defects, the temperature profile is optimized to reduce peeling defects. The design of the optimization strategy must consider the interaction of different process parameters during production; therefore, the synergistic effect between various parameters must be comprehensively considered to ensure that the adjustment scheme can effectively control multiple defect types. Finally, the generated process parameter adjustment data includes the specific adjustment value and adjustment strategy for each parameter, providing a basis for subsequent process verification and implementation.

[0142] Step S53: Use process parameter adjustment data to simulate and verify the production process, predict the impact of parameter adjustment on interface quality, and generate process optimization prediction data;

[0143] This invention utilizes the process parameter adjustment data generated in step S52 for production process simulation verification. First, the impact of adjusted process parameters on interface quality is simulated using a computer simulation platform or a model based on historical production data. During the simulation, based on different process parameter adjustment schemes, it predicts whether the adjusted production process can effectively reduce specific types of interface defects, especially those that significantly affect wire performance. The effectiveness of the parameter adjustment is verified by comparing the defect distribution and severity before and after the simulation. The generated process optimization prediction data includes predictions of defect quantity, severity, and distribution. This data can provide a scientific basis for process adjustments in actual production, helping the process optimization team better understand the potential impact of parameter adjustments on wire quality.

[0144] Step S54: Transform the process optimization prediction data into specific production equipment control parameters to generate wire manufacturing process optimization data;

[0145] In this embodiment of the invention, the process optimization prediction data in step S53 is transformed into specific production equipment control parameters. Based on the process optimization prediction data, the optimal process parameters are selected and transformed into specific production equipment control commands. For example, optimized parameters such as temperature, stretching rate, and cooling rate are input into the production equipment's control system. This process includes parameter quantification, standardization, and transformation into specific control values ​​executable by the production line. This optimized data will be directly applied to the automatic control system of the production equipment to ensure that the production process meets optimal process requirements, thereby reducing defects and guaranteeing wire quality. The generated wire manufacturing process optimization data includes the control parameters and their variation ranges for each production piece of equipment, used to achieve precise control of wire quality during the production process.

[0146] Step S55: Adjust the wire production process parameters in real time based on the wire manufacturing process optimization data, and collect process parameters during the production process to generate process adjustment execution data;

[0147] This embodiment of the invention utilizes the wire rod manufacturing process optimization data generated in step S54 to adjust the wire rod production process parameters in real time. This is achieved by monitoring various process parameters on the production line in real time and adjusting them based on the optimization data. For example, parameters such as the heating furnace temperature, stretching machine speed, and cooling spray system are adjusted. During the adjustment process, the production system automatically corrects the parameters based on the real-time collected process data to ensure precise control of the parameters during production. Simultaneously, various process parameters, such as temperature, pressure, speed, and tension, need to be collected in real time during production. This data will be used for subsequent analysis to determine whether the adjusted process achieves the expected results. The generated process adjustment execution data includes the process parameter values ​​and their execution status at each adjustment moment, providing fundamental data support for further analysis and optimization.

[0148] Step S56: Perform interface quality testing based on the process adjustment execution data to obtain wire interface performance evaluation data.

[0149] This invention, based on the process adjustment execution data generated in step S55, performs wire interface quality testing. Advanced non-destructive testing techniques (such as X-ray imaging, ultrasonic testing, and laser scanning) are used to inspect the interface quality of the wire after adjustments during production, evaluating the type, quantity, size, and distribution of surface and internal defects. By comparing changes in defect types and quantities, the improvement effect of process parameter adjustments on interface quality can be verified. The generated wire interface performance evaluation data includes the distribution and severity of various defects, as well as the mechanical properties of the wire (such as tensile strength and bending strength). This data provides a quantitative assessment of the process optimization effect and provides feedback for subsequent quality improvements, enabling further adjustments to the production process.

[0150] The present invention also provides a metal wire data analysis system for performing the metal wire data analysis method described above, the metal wire data analysis system comprising:

[0151] The data acquisition and modeling module is used to acquire wire interface data of composite metals; acquire wire interface feature matrix based on wire interface data; and construct wire interface feature model based on wire interface feature matrix.

[0152] The differential signal processing module is used to perform differential signal conversion processing on the wire interface feature model, extract the wire interface change features based on the differential signal conversion results, and construct a preliminary indicator map of wire interface defects based on the wire interface change features.

[0153] The multi-scale decomposition module is used to perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; and match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data.

[0154] The defect distribution module is used to establish a defect distribution map of the wire interface based on the precise defect characterization data; calculate the severity of wire interface defects based on the defect distribution map; and analyze the wire production process related data based on the severity of wire interface defects.

[0155] The quality verification module is used to generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters, and verify wire interface quality to obtain wire interface performance evaluation data.

[0156] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0157] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for analyzing metal wire data, characterized in that, Includes the following steps: Step S1: Obtain the wire interface data of the composite metal; The wire interface feature matrix is ​​acquired based on the wire interface data; a wire interface feature model is constructed based on the wire interface feature matrix. Step S2: Perform differential signal conversion processing on the wire interface feature model, and extract the wire interface change features based on the differential signal conversion results; construct a preliminary indicator map of wire interface defects based on the wire interface change features. Step S2 includes: Step S21: Establish the benchmark reference area for the wire interface feature model, select an interface segment without obvious defects as the reference standard, and generate wire interface benchmark feature data. Step S22: Compare each detection area in the wire interface feature model with the wire interface baseline feature data point by point, calculate the feature difference value, and form the original differential data of the wire interface. Step S23: Perform statistical normalization on the original differential data of the wire interface and perform neighborhood enhancement processing to obtain enhanced differential data of the wire interface; Step S24: Set multi-level threshold conditions based on the wire interface enhancement differential data, and perform hierarchical filtering of differential signals to form wire interface differential signal conversion data; Step S25: Identify regions with similar change patterns based on differential signal conversion data of the wire interface, thereby extracting the change features of the wire interface; Step S26: Perform spatial mapping based on the characteristics of wire interface changes, determine the boundary and range of high-confidence defect areas, and construct a preliminary indicator map of wire interface defects; Step S3: Perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data. The multi-scale decomposition process based on the preliminary indicator map of wire interface defects includes: Multi-resolution segmentation is performed on the preliminary indicator map of defects at the wire interface, dividing the image into detection areas at macro, meso, and micro scales to generate multi-level defect decomposition data. Morphological dilation processing is performed on the macroscopic detection area to enhance the continuity characteristics of interface cracks and generate macroscopic defect connected domain data. Adaptive threshold segmentation is performed on the mesoscale detection region to separate grain boundary slip bands from second-phase precipitation features, generating mesoscale defect separation data. Laplacian sharpening filtering is applied to the microscale detection area to generate enhanced data of micro-defects. Multi-level defect decomposition data, macroscopic defect connected domain data, mesoscopic defect separation data, and microscopic defect enhancement data are input into a multi-channel data fusion network to generate cross-scale defect correlation data. A three-dimensional defect space mapping is constructed based on cross-scale defect association data, the distribution density of defects along the X / Y / Z axes is labeled, and a defect space distribution vector is generated. Principal component analysis is performed on the spatial distribution vector of defects to generate a dataset of core defect features. The core feature dataset of defects is matched with a pre-set defect type database using cosine similarity. Feature combinations with a similarity greater than 85% are selected to generate interface defect feature data. Step S4: Establish a defect distribution map of the wire interface based on the accurate defect characterization data; calculate the severity of the wire interface defects based on the defect distribution map; analyze the wire production process related data based on the severity of the wire interface defects. Step S5: Generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters and verify wire interface quality to obtain wire interface performance evaluation data.

2. The metal wire data analysis method according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Set up a wire detection device to fix the composite metal wire to be tested within the detection area of ​​the sensor array, thereby acquiring cross-sectional image data and surface feature data of the wire; Step S12: Perform morphological feature reconstruction processing on the cross-sectional image data and surface feature data of the wire to obtain the wire interface data of the composite metal; Step S13: Determine the ultrasonic transmission position and parameters based on the wire interface data, transmit multi-band ultrasonic signals to the fixed composite metal wire, receive the echo signals from the wire interface, and calculate the ultrasonic response parameters based on the echo signal intensity, phase, and delay time. Step S14: Determine the eddy current detection path based on the wire interface data, scan the wire using a multi-frequency eddy current sensor, measure the electromagnetic field response changes at different locations on the wire, and calculate the electromagnetic conductivity distribution data based on the electromagnetic field response changes. Step S15: Set X-ray scanning parameters based on wire interface data, use low-energy X-rays to perform tomographic scanning on the wire, and generate an interface density gradient map; Step S16: Combine the ultrasonic response parameters, electromagnetic conductivity distribution data, and interface density gradient map to obtain the wire interface feature matrix; Step S17: Construct a wire interface feature model based on the wire interface feature matrix.

3. The metal wire data analysis method according to claim 2, characterized in that, Step S17 includes the following steps: Step S171: Perform data standardization processing on the wire interface feature matrix, incorporate different physical quantities into a unified measurement system, and generate standardized interface feature data. Step S172: Establish a three-dimensional coordinate system based on standardized interface feature data, and map the wire geometry information to the spatial coordinate system to form wire interface spatial mapping data; Step S173: Construct the basic topology of the interface using the wire interface space mapping data, determine the interface continuity and morphological characteristics, and generate interface topology data. Step S174: Analyze the spatial distribution pattern of interface features based on the interface topology data, identify the feature change patterns of interface regions, and form interface region feature clustering data. Step S175: Based on the interface region feature clustering data, filter the key feature parameters that are sensitive to interface quality to obtain the set of key interface feature parameters; Step S176: Construct a feature vector space using the set of key interface feature parameters, and perform dimensionality reduction processing to generate dimensionality-reduced interface feature data. Step S177: Establish a mathematical model of the wire interface based on the dimensionality reduction data of the interface features, and impose physical property constraints on the mathematical model of the wire interface to obtain the feature model of the wire interface.

4. The metal wire data analysis method according to claim 3, characterized in that, Step S3, which involves identifying the type of wire interface defect based on the interface defect feature data, includes: Extract the feature values ​​of defect length, width, depth and area ratio from the interface defect feature data to generate a basic geometric feature set of defects; The geometric feature set of defects is compared one by one with the standard defect geometric templates in the defect type database to calculate the geometric similarity. Based on geometric shape similarity, a candidate set of defect types with a matching degree >75% was selected; Based on the candidate set of defect types, the density distribution of defects, the spacing between adjacent defects and the arrangement direction of defects in the material interface region are calculated to generate a spatial distribution feature vector of defects. Based on the spatial distribution feature vector of defects, defect groups are divided according to density peak detection to generate defect spatial clustering data; Based on defect spatial clustering data, defect edge morphological features are extracted, and edge roughness, fractal dimension and curvature change rate are quantified to generate a defect edge feature parameter set. A secondary matching is performed between the defect edge feature parameter set and the defect type database to correct misclassification items in the preliminary defect classification data and generate optimized defect classification results. Based on the preset defect type determination rules, the optimized defect classification results are determined by rules, and wire interface defect type features containing defect type codes, confidence levels and location coordinates are generated.

5. The metal wire data analysis method according to claim 4, characterized in that, Step S3, which involves matching and correcting the characteristics of the wire interface defect type with the preset physical constraint model, includes: Establish a physical constraint model library for wire interface defects, including physical property parameters and constraint conditions for peeling, porosity and inclusions; The defect type characteristics of wire interface are compared with the physical constraint model library of wire interface defects by parameterization to generate preliminary defect type matching data. Based on the preliminary matching data of defect types, the physical rationality of defect characteristics is assessed, and the constraints are verified to form a physical rationality assessment of the defects. The physical rationality assessment of defects is modified based on physical constraint rules to obtain defect feature correction data. Based on the defect feature correction data, quantitative defect characterization parameters including defect size, depth, shape and distribution characteristics are calculated to generate a set of quantitative defect characterization parameters; The defect evolution law is matched with the defect evolution law in the physical constraint model library of wire interface defects based on the quantitative characterization parameter set of defects, and the defect formation mechanism and development trend are evaluated to generate accurate defect characterization data.

6. The metal wire data analysis method according to claim 5, characterized in that, Step S4 includes the following steps: Step S41: Spatial mapping of the defect characterization data based on the axial and radial coordinates of the material to generate wire defect spatial location data containing three-dimensional defect distribution information; Step S42: Based on the spatial location data of wire defects, perform layered display and density statistics on different types of defects to construct a wire interface defect distribution map that reflects the distribution pattern of various types of defects; Step S43: Based on the defect distribution map of the wire interface, perform weighted calculations on the impact of various defects on wire performance, assess the overall quality level of the wire and local areas, and form the severity of the wire interface defects; Step S44: Determine the characteristic pattern and severity of the current wire defect based on the severity of the wire interface defect, and generate the wire defect rating result; Step S45: Perform correlation analysis between the wire defect rating results and the time series data of process parameters recorded during the wire production process to identify the correspondence between defect formation and changes in specific process parameters, and form wire production process correlation data.

7. The method for analyzing metal wire data according to claim 6, characterized in that, Step S5 includes the following steps: Step S51: Construct a process parameter optimization model based on the wire production process correlation data, identify key process parameters that affect interface quality, and generate a process parameter priority list; Step S52: Design parameter adjustment schemes based on the process parameter priority list, formulate process parameter optimization strategies based on different types of interface defects, and generate process parameter adjustment data; Step S53: Use process parameter adjustment data to simulate and verify the production process, predict the impact of parameter adjustment on interface quality, and generate process optimization prediction data; Step S54: Convert the process optimization prediction data into specific production equipment control parameters to generate wire manufacturing process optimization data; Step S55: Adjust the wire production process parameters in real time based on the wire manufacturing process optimization data, and collect process parameters during the production process to generate process adjustment execution data; Step S56: Perform interface quality testing based on the process adjustment execution data to obtain wire interface performance evaluation data.

8. A data analysis system for metal wires, characterized in that, For performing the metal wire data analysis method as described in claim 1, the metal wire data analysis system comprises: The data acquisition and modeling module is used to acquire wire interface data of composite metals; acquire wire interface feature matrix based on wire interface data; and construct wire interface feature model based on wire interface feature matrix. The differential signal processing module is used to perform differential signal conversion processing on the wire interface feature model, extract the wire interface change features based on the differential signal conversion results, and construct a preliminary indicator map of wire interface defects based on the wire interface change features. The multi-scale decomposition module is used to perform multi-scale decomposition processing based on the preliminary indicator map of wire interface defects to obtain interface defect feature data; identify the type features of wire interface defects based on the interface defect feature data; and match and correct the type features of wire interface defects with the preset physical constraint model to generate accurate defect characterization data. The defect distribution module is used to establish a defect distribution map of the wire interface based on the precise defect characterization data; calculate the severity of wire interface defects based on the defect distribution map; and analyze the wire production process related data based on the severity of wire interface defects. The quality verification module is used to generate wire manufacturing process optimization data based on wire production process correlation data; apply the wire manufacturing process optimization data to adjust wire production process parameters, and verify wire interface quality to obtain wire interface performance evaluation data.

Images