Non-ferromagnetic metal material surface contour reconstruction method based on rotating alternating current electromagnetic field

Abstract

The invention relates to the field of metal material nondestructive testing, and provides a non-ferromagnetic metal material surface profile reconstruction method based on a rotating alternating current electromagnetic field, which comprises the following steps of: determining a detection point array of a detection area, acquiring a vertical magnetic field signal z of the detection area by adopting a TMR magnetic sensing chip, and forming an initial Bz numerical matrix Bz0; performing bilateral filtering; obtaining a Bz numerical matrix Bz2 after morphological processing; obtaining a weighted Bz numerical matrix Bz3; a positive value matrix is formed, binarization processing is conducted on the positive value matrix, and a key numerical value matrix area Tkey containing crack related feature points is formed; calculating the center-of-mass coordinates of the disconnected independent areas, which are called nodes; on the basis of an improved Boruvka algorithm, new edge weight design is carried out on connecting edges between nodes, and a generated minimum spanning tree is effective representation of the surface profile of the bifurcated crack of the metal part.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromagnetic nondestructive testing, and in particular relates to a method for reconstructing the surface contour of a bifurcated crack of a metal material at low resolution based on a rotating alternating current electromagnetic field technology. Background Art

[0002] In the field of metal part defect detection, the inventor team had previously applied for two patents, namely, a defect feature inversion and enhancement method based on an orthogonal dual-frequency magnetic yoke with application number: 202310390447.3 and a surface contour reconstruction method for bifurcated cracks in metal parts with application number: 202311648906.X.

[0003] Patent application 202310390447.3 applies low-frequency and high-frequency excitation signals to orthogonal dual-frequency yoke coils, respectively, and uses electromagnetic detection elements to scan the metal specimen to obtain coupled electromagnetic signals. The signals are then processed through frequency division, transformation, and other processing steps to invert and enhance defect features. However, this method has certain limitations. On the one hand, it relies on the electromagnetic detection element to move grid by grid to collect ordinary magnetic field signals, and its direction is fixed. The detection angle for complex-shaped metal parts or multi-directional cracks is limited, and crack features with large angles are easily missed. In addition, the differences in the propagation characteristics of the dual-frequency excitation signals make subsequent fusion difficult, affecting the accuracy of defect inversion. On the other hand, the contour reconstruction process only fuses gradient features and convolution processing to construct a two-dimensional plane map. The reconstruction method for crack branch connections and overall contours needs to be further developed. In actual applications, there may be deviations in crack branch connections or insufficient restoration of the overall contour, affecting the accuracy of defect judgment.

[0004] Patent application 202311648906.X has improved the contour reconstruction accuracy of bifurcation cracks to a certain extent through orthogonal scanning and gradient fusion algorithms, but it still has the following limitations: First, the directionality of the excitation field is limited. Specifically, the direction of the magnetic field generated by the traditional single U-shaped magnetic yoke excitation probe is fixed, and the detection sensitivity of non-vertical cracks decreases significantly with increasing angles, making it difficult to adapt to the detection needs of multi-directional cracks. Secondly, the signal processing robustness is insufficient. In low-resolution detection scenarios or strong noise environments, the single mean filtering and threshold segmentation methods are prone to blurred crack edges and missed detection of minor defects, and it is impossible to achieve a balance between noise suppression and feature retention. Finally, there are limitations to the defect reconstruction algorithm. When dealing with bifurcation cracks, traditional graph theory algorithms do not fully combine the characteristics of the magnetic field gradient distribution, and the accuracy of crack branch connections is insufficient. In addition, the time complexity is high, making it difficult to meet real-time detection requirements. Summary of the Invention

[0005] This invention improves the existing technology and proposes a method for reconstructing the surface profile of bifurcated cracks in metal materials, which has better detection effect. The specific technical solution is as follows:

[0006] A method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field is characterized in that the probe used is a double U-shaped orthogonal structure excitation probe based on AC electromagnetic field detection, including two excitation coils and a TMR magnetic sensor chip for collecting magnetic field distortion signals, and includes the following steps:

[0007] (1) A double U-shaped orthogonal structure excitation probe and a TMR magnetic sensor chip located below it are placed on the surface of the metal part to be measured, and an alternating current of a certain frequency is passed through the two excitation coils to generate a rotating AC electromagnetic field. When the rotating AC electromagnetic field acts on the metal surface, an induced current is generated inside the metal. If there is a crack on the surface, the distribution of the induced current changes, which in turn causes the surrounding magnetic field to be distorted. The detection point array of the detection area is determined. The number of detection points is set to M×N. The TMR magnetic sensor chip is used to collect the vertical magnetic field signal Bz of the detection area to form an initial M×N Bz numerical matrix Bz0. The matrix element Bz(m,n) represents the magnetic field amplitude at the detection point coordinate (m,n).

[0008] (2) Perform bilateral filtering on the initial Bz numerical matrix Bz0;

[0009] (3) Performing opening and closing operations on the Bz numerical matrix Bz1 after bilateral filtering to obtain the Bz numerical matrix Bz2 after morphological processing;

[0010] (4) Calculate the x and y direction gradient matrix G for the Bz numerical matrix Bz2 after morphological processing x , G y， Get the gradient magnitude matrix Construct a composite weight function W based on the gradient amplitude matrix, and multiply Bz2 and W element by element to obtain the weighted Bz numerical matrix Bz3 to enhance the crack edge and core features;

[0011] (5) Take the absolute value of the weighted Bz numerical matrix Bz3 to form a positive value matrix. Determine the threshold k based on the detection scene experience, and perform binarization on the positive value matrix: when the element value ≥ k, it is set to 1, representing the potential crack feature point; otherwise, it is set to 0, representing the background area. Extract all connected areas with values ​​1 to form the key numerical matrix area T containing the crack-related feature points. key ;

[0012] (6) The key numerical matrix area T after binarization key The coordinates of the centroid of each independent area in the unconnected area are calculated. The centroid is used as the key feature point for detecting whether there is a crack branch and is called a node.

[0013] (7) Based on the improved Boruvka algorithm, a new edge weight is designed for the connecting edges between nodes, and gradient information reflecting the clarity of the crack edge is introduced. If there is a bifurcation crack in the metal part, the generated minimum spanning tree is an effective representation of the surface contour of the bifurcation crack in the metal part.

[0014] Furthermore, in step (4), the formula of the composite weight function W based on the gradient magnitude matrix is:

[0015] W=exp(-βG)+γBz 2 2, where β and γ are weight coefficients.

[0016] Furthermore, in step (4), β=0.2 and γ=0.8.

[0017] Furthermore, let the two nodes be (x i ,y i ) and (x j ,y j ), calculate the mean gradient G of the edge midpoint according to the weighted Bz numerical matrix Bz3 matrix mean The method is as follows: construct a detection point grid of the detection area based on the detection point array, where each grid contains one detection point. Connect the detection point grids where two nodes are located with a straight line. Select corresponding detection points based on the detection point grids that this straight line passes through. Average the gradient amplitudes corresponding to the selected detection points and use them as the gradient mean of the midpoint of side AB.

[0018] Furthermore, in step (7), the two nodes (x i ,y i ) and (x j ,y j ) is used to measure the actual distance between nodes in space, and the formula is:

[0019]

[0020] Furthermore, in step (7), the weight of each edge W edge From the Euclidean distance D and the mean gradient G of the edge midpoint mean Jointly determined, the calculation formula is W edge =(1+αG mean )·D, where G mean It is the mean gradient of the edge midpoint calculated based on the weighted Bz numerical matrix Bz3 matrix. D is the Euclidean distance between two nodes, and α is the gradient weight coefficient, which ranges from 0.5 to 1.0.

[0021] Furthermore, in step (7), the implementation steps of the improved Boruvka algorithm are as follows:

[0022] 1) Initialization: All centroids are treated as independent nodes, and each node constitutes an initial connected component, that is, T = (V, {}), where V is the node set;

[0023] 2) Iterative merging, the method is as follows:

[0024] a. Select the minimum cross-component edge for each connected component: For all current connected components, traverse the edges between them and all other connected components and calculate the weight W of each edge edge , and select for each connected component an edge with the minimum weight connecting it to other connected components;

[0025] b. Filter invalid edges and merge components: After collecting the minimum weight edges selected for all connected components, only the edges connecting different connected components are retained, which are called valid edges. For each valid edge, the two connected components connected by it are merged into a new connected component.

[0026] c. Iterate until fully connected: Check whether all nodes belong to the same connected component. If not, repeat steps a and b for the remaining connected components, recalculate the edge weights and select the minimum weight edges for merging. Finally, when all nodes are connected, the minimum spanning tree generated at this time is a valid representation of the surface contour of the bifurcation crack in the metal part.

[0027] Furthermore, the two U-shaped magnetic cores of the double U-shaped orthogonal structure excitation probe are perpendicular to each other, an excitation coil is wound on each U-shaped magnetic core, and the TMR magnetic sensor chip is arranged below the excitation probe, and its sensitive axis direction is perpendicular to the surface of the metal part being measured.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows: traditional probes have certain limitations when detecting bifurcation cracks. In terms of signal processing, conventional filtering methods have limited processing capabilities for complex magnetic field signals, and traditional algorithms have high time complexity when reconstructing defect contours, which has a certain impact on detection efficiency. The dual U-shaped orthogonal structure excitation probe adopted by the present invention can generate and detect magnetic fields in different directions, reduce the influence of crack direction on detection, and effectively collect magnetic field distortion signals related to bifurcation cracks through the synergistic effect of rotating AC electromagnetic field and TMR magnetic sensor chip. In the data processing process, the numerical matrix acquisition, bilateral filtering smoothing, morphological opening and closing operations, gradient fusion algorithm and Boruvka algorithm of the present invention cooperate with each other to overcome the challenges brought by low resolution and bifurcation crack shape, and improve the accuracy of surface contour reconstruction of bifurcation cracks in metal materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The double U-shaped orthogonal structure excitation probe used in the experimental process of the present invention;

[0030] Figure 2 The positional relationship between the probe used in the experimental process of the present invention and the metal being measured;

[0031] Figure 3 A schematic diagram of the specific connection architecture and layout of the system used in the present invention;

[0032] Figure 4 Schematic diagram of the size of bifurcated cracks on the surface of the metal part being tested in the experiment of the present invention;

[0033] Figure 5 The numerical matrix images of the initial Bz signals of a T-type bifurcation defect and a Y-type bifurcation defect are shown (the left side is a T-type crack, and the right side is a Y-type crack);

[0034] Figure 6 The numerical matrix images of T-type and Y-type bifurcation cracks after smoothing are shown (T-type crack on the left and Y-type crack on the right).

[0035] Type crack);

[0036] Figure 7 The images of T-type and Y-type bifurcation cracks after morphological processing are shown (T-type crack on the left and Y-type crack on the right);

[0037] Figure 8 The image shows the gradient fusion of T-type and Y-type bifurcation cracks (T-type crack on the left and Y-type crack on the right);

[0038] Figure 9 The positive matrix images of the absolute values ​​of T-type and Y-type bifurcation cracks are shown (T-type crack on the left and Y-type crack on the right);

[0039] Figure 10 The images of T-shaped and Y-shaped bifurcation cracks after threshold segmentation are shown (T-shaped crack on the left and Y-shaped crack on the right).

[0040] Type crack);

[0041] Figure 11 The centroid positions of the local areas of T-shaped and Y-shaped bifurcation cracks are shown (T-shaped crack on the left and Y-shaped crack on the right); Figure 12 The surface contours of T-shaped and Y-shaped bifurcation cracks reconstructed using the Boruvka algorithm are shown (T-shaped crack on the left and Y-shaped crack on the right); In the figure: 1-1 upper excitation coil, 1-2 lower excitation coil; 2-1 upper U-shaped magnetic yoke, 2-2 lower U-shaped magnetic yoke; 303-TMR magnetic sensor chip; 4-aluminum metal material to be measured. Figure 13 Flowchart of the program of the present invention. DETAILED DESCRIPTION

[0042] The invention is further described in detail below with reference to the accompanying drawings and specific implementation methods. The examples given here are only used to illustrate and explain the invention and are not used to limit the invention.

[0043] First, the basic scheme of the present invention will be described.

[0044] The present invention's method for reconstructing the surface profile of a bifurcated crack in a metal material uses a dual-U-shaped orthogonal excitation probe based on AC electromagnetic field detection, including an excitation coil and a TMR magnetic sensor chip for acquiring magnetic field distortion signals. The method specifically includes the following steps:

[0045] (1) Determine the location of the bifurcation crack and the specific location of the initial detection. The double U-shaped orthogonal structure excitation probe generates a rotating AC electromagnetic field by passing an AC current with specific parameters into its excitation coil. When the rotating AC electromagnetic field (RACFM) acts on the metal surface, an induced current is generated inside the metal. If there is a crack on the surface, the induced current distribution will change, thereby causing the surrounding magnetic field distortion. The TMR magnetic sensor chip is used to collect the vertical magnetic field signal Bz of the detection area point by point and line by line through C scanning, forming an initial M×N Bz numerical matrix Bz0. The matrix element Bz(m,n) represents the magnetic field amplitude at the coordinate (m,n).

[0046] (2) Perform bilateral filtering on the initial Bz numerical matrix Bz0. Bilateral filtering is a filtering method that considers both the spatial domain and the value domain. It can effectively retain the edge features of the magnetic field distortion signal while removing noise interference. After bilateral filtering, a new Bz numerical matrix Bz1 is obtained. This matrix retains the edge features of the magnetic field distortion signal while removing noise interference, avoiding the problem of blurred crack edges in traditional filtering noise reduction, making the data smoother and more stable.

[0047] (3) Opening and closing operations are performed on the bilaterally filtered Bz numerical matrix Bz1. The opening operation first erodes and then dilates to remove small noise points and reduce interference with crack feature analysis; the closing operation first dilates and then erodes to fill holes in the data. After these two steps, a further optimized Bz numerical matrix Bz2 is obtained, which has higher data quality and can more accurately reflect crack characteristics.

[0048] (4) Calculate the x and y direction gradient matrix G for the Bz numerical matrix Bz2 after morphological processing x , G y， Get the gradient magnitude matrix Construct a composite weight function W = exp(-βG) + γBz 2 2, where β = 0.2 and γ = 0.8. Multiply Bz2 and W element-wise to get Bz3, which enhances the crack edge and core features.

[0049] (5) Take the absolute value of the weighted Bz numerical matrix Bz3 to form a positive value matrix. Determine the threshold k based on the detection scene experience, and perform binarization on the positive value matrix: when the element value ≥ k, it is set to 1 (representing the potential crack feature point), otherwise it is set to 0 (background area). Extract all connected areas with a value of 1 to form the key numerical matrix area T containing the crack-related feature points. key The matrix can filter out background noise, retain the endpoints, branch intersections and edge contour points of bifurcated cracks, and reduce interference from irrelevant data.

[0050] (6) The key matrix T after binarization key The centroid coordinates of the independent regions in the crack are calculated. The centroid can be used as the characteristic key point of each crack branch to provide the coordinate basis for subsequent contour reconstruction.

[0051] (7) The traditional Boruvka algorithm aims to construct a minimum spanning tree (MST). Its core idea is to find the shortest edge connecting to other components for each connected component (initially each point is an independent connected component) in an undirected graph through iteration, and gradually connect all points into a loop-free network so that the total edge length of this network is minimized.

[0052] The improved Boruvka algorithm in this patent uses a new edge weight design when determining the connecting edges between nodes (i.e., the centroid points obtained in step (6) above). Instead of relying solely on Euclidean distance to measure edge weights, it introduces gradient information that reflects the clarity of the crack edge.

[0053] Specifically, the weight W of each edge edge From the Euclidean distance D and the mean gradient G of the edge midpoint mean Jointly determined, the calculation formula is W edge =(1+αG mean )·D. Among them, G mean is the mean gradient of the edge midpoint in the Bz3 matrix after a series of processing. The higher the gradient value, the more obvious the magnetic field distortion at the crack edge at that location, that is, the clearer the crack edge; D is the Euclidean distance between two nodes, and the calculation formula is

[0054]

[0055] Used to measure the actual distance between nodes in space; α is the gradient weight coefficient, and its value range is generally between 0.5 and 1.0.

[0056] Two nodes (x i ,y i ) and (x j ,y j ) of the edge midpoint gradient mean Gmean The acquisition method is as follows: construct a grid based on the detection points of the C scan, where each grid contains one detection point. Connect the grids where the two centroids are located with a straight line. Select the corresponding detection points based on the grids that the straight line passes through. Average the gradient amplitudes corresponding to the corresponding detection points and use them as the gradient mean of the midpoint of side AB.

[0057] In this embodiment, α is set to 0.7, which is used to adjust the influence of gradient information on edge weights. Taking two centroid points A(x1, y1) and B(x2, y2) as an example, first calculate the Euclidean distance between them Assume that the coordinates of the midpoint of side AB are (x m ,y m ), by obtaining the gradient value G of the midpoint in the Bz3 matrix m .

[0058] The algorithm implementation steps are as follows:

[0059] Initialization: Treat all centroids as independent nodes, and each node constitutes an initial connected component, that is, T = (V, {}), where V is a node set.

[0060] Iterative merging:

[0061] 1. Select the minimum cross-component edge for each connected component: For all current connected components (each node is an independent component initially), traverse the edges between it and all other connected components and calculate the weight W of each edge edge , and for each component, select an edge with minimum weight connecting to other components.

[0062] Example: If there are currently connected components {A}, {B}, and {C}, {A} will calculate the edge weights with {B} and {C} and select the smallest edge (such as AB edge); {B} will similarly select the smallest edge (possibly AB edge); {C} will select the smallest edge (such as BC edge).

[0063] 2. Filter invalid edges and merge components: After collecting all the edges selected by the components, only keep the edges that connect different connected components (if both ends of an edge already belong to the same component, skip it). For each valid edge, merge the two connected components it connects into a new component.

[0064] Example: If edge AB connects {A} and {B}, they are merged into {A,B}; if edge BC connects {C} and {A,B}, they are merged into {A,B,C}.

[0065] 3. Loop until fully connected: Check whether all nodes belong to the same connected component. If not, repeat steps 1-2 for the remaining connected components (including the newly merged component), recalculate the edge weights, and select the minimum edge merge. Finally, when all nodes are connected, the minimum spanning tree generated is a valid representation of the surface profile of the bifurcation crack in the metal part.

[0066] Description of the changes in the image before and after processing:

[0067] Before processing: Binarized key matrix T key It appears as a series of scattered white pixel blocks (representing crack-related areas) distributed on a black background (representing the background area). There is no obvious connection between the pixel blocks, making it difficult to intuitively judge the overall outline and branching direction of the crack.

[0068] After processing: By constructing a minimum spanning tree using an improved Boruvka algorithm, the crack's bifurcation structure and overall outline are revealed by connecting the centroids of each branch (i.e., key feature points) with lines. For example, for a Y-shaped crack, the centroids of the three branches can be effectively connected to form a Y-shaped connection structure. For a T-shaped crack, the connection between the trunk and branches can also be clearly displayed, which is consistent with the actual crack morphology and effectively reconstructs the surface contour of the bifurcated crack.

[0069] The two U-shaped magnetic cores are perpendicular to each other, and the excitation coils, typically made of enameled wire, are evenly wound around the crossbars of each U-shaped core. This winding method optimizes the distribution and intensity of the excitation magnetic field, making the probe more sensitive to magnetic field distortion caused by bifurcated cracks. The TMR magnetic sensor chip is positioned below the excitation probe, with its sensitive axis perpendicular to the aluminum plate surface. The TMR magnetic sensor chip's high sensitivity and low noise enable it to accurately capture magnetic field distortion signals around bifurcated cracks, improving the resolution of weak magnetic field changes compared to traditional detection coils.

[0070] The AC electromagnetic field metal material magnetic field detection system of the present invention comprises the aforementioned dual U-shaped orthogonal excitation probe, a signal generator, a power amplifier, a data acquisition card, and a computer. The output of the signal generator is connected to the input of the power amplifier, which in turn is connected to an excitation coil, providing the excitation coil with a suitable signal to generate an AC electromagnetic field. Furthermore, a TMR magnetic sensor chip, used to sense magnetic field changes, is connected to the input of the data acquisition card to transmit the detected signal to the card. The output of the data acquisition card is connected to the computer, allowing the collected data to be further transmitted to the computer for processing and analysis.

[0071] This patent proposes an improved hardware solution based on a rotating AC electromagnetic field. Addressing the issues with patent application 202310390447.3, which relies on the grid-by-grid movement of electromagnetic detection elements to collect fixed-direction normal magnetic field signals, and patent application 202311648906.X, which uses a traditional single U-shaped magnetic yoke excitation probe, resulting in a fixed magnetic field direction, this patent designs a dual-U-shaped orthogonal excitation probe. The U-shaped yoke is made of manganese-zinc ferrite with a magnetic permeability of 7000, and 100 turns of 0.2mm diameter enameled wire are wound around the U-shaped yoke. AC current with a frequency of 10kHz and a phase difference of 90° is applied to the two coils, generating a rotating AC electromagnetic field to enhance crack detection sensitivity. Compared to traditional unidirectional excitation methods, this rotating AC electromagnetic field can induce multidirectional induced currents on the specimen surface, significantly perturbing cracks of varying orientations and reducing detection angle sensitivity. In the signal processing and defect reconstruction links, in response to the shortcomings of patent application 202310390447.3 lacking a signal processing mechanism and patent application 202311648906.X only using a single signal processing method, this patent first uses a method combining bilateral filtering and morphological processing to simultaneously consider the pixel correlation in the spatial domain and the value domain, effectively remove small noise points and fill holes, and to a certain extent achieve a good balance between noise reduction and edge retention; in terms of feature enhancement, a composite weight function is constructed through the gradient fusion algorithm, and the gradient amplitude is combined with the original magnetic field data to highlight the crack contour characteristics. In terms of defect reconstruction technology, in response to the shortcomings of patent application 202310390447.3 in constructing a two-dimensional plane graph by fusing gradient features and convolution processing, and patent application 202311648906.X in using the Kruskal algorithm and only using Euclidean distance as the minimum spanning tree condition, this patent introduces an improved Boruvka algorithm. In the process of minimum spanning tree construction, the mean gradient of the edge midpoint is introduced as a weight parameter. The optimized edge weight formula can effectively improve the accuracy of branch connections when processing bifurcated crack structures. At the same time, algorithm optimization makes the calculation process more efficient, thereby reducing time complexity. The technical solution of the present invention helps to overcome some problems of traditional methods in low-resolution detection and improve the reconstruction effect and anti-interference performance of the surface contour of bifurcated cracks.

[0072] The present invention will be described below with reference to the embodiments.

[0073] In the double U-shaped orthogonal structure excitation probe of the present invention, the excitation coil 1-1 in the upper U-shaped magnetic yoke 2-1 and the excitation coil 1-2 in the lower U-shaped structure 2-2 are arranged orthogonally. By passing an alternating current with a frequency of 10kHz, an amplitude of 5V and a phase difference of 90° through the excitation coils 1-1 and 1-2, a rotating AC electromagnetic field is generated by utilizing the synergistic effect of the orthogonal layout and the specific phase difference. The current frequency of the two coils is set to 10kHz to ensure that the magnetic fields generated by each coil change synchronously in time, providing a prerequisite for forming a regular rotating AC electromagnetic field. The fixed phase difference of 90° is used as an excitation condition to make the two currents satisfy I1=I m sin(ωt), I2=I m cos(ωt)), the corresponding mutually perpendicular magnetic field strength components are B1=B m sin(ωt), B2=B m cos(ωt), the two are perpendicular to each other. The synthetic magnetic field strength can be obtained by the principle of vector superposition: Constant, its amplitude is constant at B m , the angle θ=90-ωt between the direction and B1 rotates with time. When there is a crack in the metal plate, the induced current distribution is disturbed. By analyzing the current changes at different times (such as 1 / 8T, 2 / 8T, etc.), the crack can be detected. Figure 1 shown.

[0074] Each double U-shaped orthogonal structure excitation probe contains specific components: the U-shaped yoke is made of manganese-zinc ferrite, which is conducive to the effective conduction and concentration of the magnetic field; the excitation coil 1 uses 0.2mm diameter enameled wire tightly wound 100 turns on the U-shaped yoke, and generates a stable and controllable magnetic field when energized; the TMR magnetic sensor chip is vertically installed directly below the center of the U-shaped yoke 2-2, with its sensitive axis direction perpendicular to the surface of the aluminum plate, and is used to sense the leakage magnetic field distortion signal caused by internal defects (such as bifurcated cracks) of the metal material being measured. The positional relationship between the probe used and the metal being measured is shown in the system structure. Figure 2 shown.

[0075] In terms of the construction of the detection system, the signal generator is connected to the input end of the power amplifier as a signal source, which can generate a sinusoidal signal with a specific waveform and parameters. Its frequency, amplitude and other parameters can be adjusted according to the actual detection needs, and serve as the system excitation source signal. The power amplifier undertakes the task of power amplification, and the output end is connected to the excitation coil 1 to provide it with energy to generate a rotating AC electromagnetic field. The TMR magnetic sensor chip senses the magnetic field distortion signal and transmits it to the input end of the data acquisition card. The data acquisition card has data acquisition and conversion functions, converting the analog signal into a digital signal and transmitting it to the computer. As the core of data analysis and processing, the computer runs the algorithm program to analyze and process the collected data, realize the detection of bifurcation cracks in metal materials and reconstruction of the surface contour. Please refer to the schematic diagram of the specific connection architecture and layout configuration of the system. Figure 3 shown.

[0076] In this embodiment, the material of the metal part 4 under test is aluminum, and the surface contains Y-shaped and T-shaped bifurcated cracks. The length of each branch crack defect is 30 mm, the width of the defect is 0.5 mm, and the depth of the defect is 4 mm. The specific dimensions are as follows: Figure 4 As shown in the figure, a three-axis scanning platform was built to record the position of the probe during scanning. The system was programmed using Labview 2018, which integrated the probe excitation acquisition and three-axis platform motion functions. The computer displayed and saved the scanning data in real time.

[0077] Compared to traditional ACFM single-excitation probes, the dual U-shaped orthogonal excitation probe overcomes the single-directional limitations of ACFM technology. This means that ACFM, based on a rotating AC electromagnetic field, can detect cracks regardless of the probe's path across the defect. Combined with the TMR magnetic sensor chip, a single scan can obtain dimensional information about cracks in any orientation, helping to reduce the likelihood of missed detections and improve crack detection rates. The probe is precisely positioned to the starting position along the preset detection path. A C-scan method is used to scan the detection range point by point and line by line, collecting magnetic field distortion signals. In this embodiment, the probe's scanning speed is 15 mm / s, and the excitation frequency is 10 kHz.

[0078] As the probe moves point by point along the detection path, the excitation magnetic field interacts with the metal material at each detection point. The dual-U-shaped orthogonal excitation probe generates a rotating AC electromagnetic field by feeding alternating current with a specific phase difference and frequency range into the excitation coil. When this rotating AC electromagnetic field acts on the metal material, it induces an induced current within the metal. If bifurcated cracks exist in the metal material, the cracks alter the distribution and path of the induced current, causing magnetic field distortion. The TMR magnetic sensor chip senses this magnetic field distortion and converts it into an electrical signal for output.

[0079] The data acquisition card collects the electrical signals output by the TMR magnetic sensor chip in real time at a pre-set sampling frequency. The collected data includes information such as magnetic field strength, magnetic field direction, and signal phase, and transmits this data to a computer for storage and subsequent processing.

[0080] In the electromagnetic detection technology system of the present invention, the vertical magnetic field component Bz is the core characteristic signal for inverting the crack profile. Its physical mechanism is as follows: when there is a bifurcated crack in the metal material, the electric field concentration effect at the defect endpoint will cause magnetic field distortion, causing the Bz signal amplitude to deviate from the ideal value (Bz = 0 when there is no defect), forming a characteristic extreme value. This extreme value is related to geometric parameters such as the bifurcation shape, depth and width of the crack, and can be extracted through signal analysis. The rotating AC electromagnetic field generated by the dual U-shaped orthogonal excitation probe causes the direction of the induced current to change dynamically. The interference of the crack on the induced current is related to the angle of rotation of the magnetic field. By analyzing the changes in the Bz signal under different phases, the inversion of the three-dimensional morphology of the crack can be achieved. The Bz signal is less sensitive to changes in the vertical distance (lift-off) between the probe and the surface of the specimen.

[0081] The initial Bz numerical matrix (100 × 100 pixels) obtained after scanning was processed by MATLAB 2021. The original Bz matrix images of Y-type and T-type bifurcation cracks are shown in Figure 5 As shown, its extreme value distribution reflects the location of the crack endpoint.

[0082] Bilateral filtering with spatial standard deviation σ_s=4 and grayscale standard deviation σ_r=5 is used to reduce noise while retaining edge features. The effect after filtering is as follows Figure 6 shown.

[0083] Using 3×3 structural elements, we perform 2 corrosion + 2 expansion operations to remove noise points with an area of ​​less than 2 pixels and fill holes with an area of ​​less than 5 pixels to improve data quality. The Bz data after morphological processing is as follows: Figure 7 shown.

[0084] Construct a composite weight function W = exp(-βG) + γB z 2 , where β = 0.2, γ = 0.8, the crack contour is highlighted by fusing the gradient information with the original data. The image after Bz gradient fusion is as follows Figure 8 shown.

[0085] Take the absolute value of the enhanced matrix to form a positive matrix. The positive matrix image is as follows Figure 9 As shown in the figure, a threshold is set to perform binarization on the positive value matrix. If the numerical matrix element is greater than the threshold, it is set to 1, and if it is less than the threshold, it is set to 0. The key numerical matrix area for reconstructing the surface contour of the metal bifurcation crack is extracted. The key numerical matrix area of ​​the Y-type and T-type bifurcation crack is shown in Figure 10 After obtaining the positive value matrix, the surface contour of the bifurcation crack of the metal part is reconstructed by threshold segmentation. The centroid position of each independent area that is not connected in the key value matrix area is obtained. The centroid position of the local area of ​​the Y-type and T-type bifurcation crack is shown as follows: Figure 11 As shown, get the coordinates of the center of mass in the entire matrix.

[0086] The present invention adopts the gradient-guided Boruvka algorithm to reconstruct the bifurcation crack contour. The algorithm optimizes the minimum spanning tree construction process by introducing gradient information, which significantly improves the connection accuracy of complex crack contours. The following are the specific implementation steps: First, the extracted centroid point is used as an independent node to initialize the minimum spanning tree T = (V, {}). Each node represents a key feature point of the crack contour. In the initial state, each node is independent and no connectivity is formed. After the initial spanning tree is constructed, it is necessary to calculate the weight of the edge between any two nodes. The edge weight is an important indicator to measure the degree of connection between two nodes. The present invention adopts an edge weight calculation formula that combines gradient information and Euclidean distance: W edge =(1+αG mean )·D)

[0087] α is the gradient weight coefficient, which is 0.7, where G mean is the mean gradient of the edge midpoint. D is the Euclidean distance between the two centroids, calculated as Euclidean distance measures the actual distance between two nodes in space.

[0088] Entering the iterative optimization stage, each centroid point is initially an independent connected component. The algorithm takes the connected component as the unit. In each round of iteration, for each component, the cross-component edges between it and all other components are traversed, the edge weight of the fused gradient mean and the Euclidean distance is calculated, the minimum weight edge is selected and the invalid edges connecting the same component are filtered out, the components connected by valid edges are merged, and this process is repeated until all nodes belong to the same connected component. At this time, the minimum spanning tree generated is designed through gradient guidance, combined with geometric position and feature significance, which effectively presents the surface contour structure of the bifurcation crack of the metal part. Its edge set corresponds to the reasonable connection path of the crack feature points, which can reflect the bifurcation shape and direction.

[0089] The above embodiments are merely detailed descriptions of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, replacements, improvements, etc. made in accordance with the spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field, characterized in that: The probe used is a double U-shaped orthogonal structure excitation probe based on AC electromagnetic field detection. It includes two excitation coils and a TMR magnetic sensor chip for collecting magnetic field distortion signals. The following steps are included: (1) A double U-shaped orthogonal structure excitation probe and a TMR magnetic sensor chip located below it are placed on the surface of the metal part to be measured. A certain frequency of alternating current is passed through the two excitation coils to generate a rotating AC electromagnetic field. When the rotating AC electromagnetic field acts on the metal surface, an induced current is generated inside the metal. If there is a crack on the surface, the distribution of the induced current changes, which in turn causes the surrounding magnetic field to be distorted. The detection point array of the detection area is determined. The number of detection points is set to M×N. The TMR magnetic sensor chip is used to collect the vertical magnetic field signal Bz of the detection area to form an initial M×N Bz numerical matrix Bz0. The matrix element Bz(m,n) represents the magnetic field amplitude at the detection point coordinate (m,n). (2) Perform bilateral filtering on the initial Bz numerical matrix Bz0; (3) Performing opening and closing operations on the Bz numerical matrix Bz1 after bilateral filtering to obtain the Bz numerical matrix Bz2 after morphological processing; (4) Calculate the x and y direction gradient matrix G for the Bz numerical matrix Bz2 after morphological processing x , G y， Get the gradient magnitude matrix Construct a composite weight function W based on the gradient amplitude matrix, and multiply Bz2 and W element by element to obtain the weighted Bz numerical matrix Bz3 to enhance the crack edge and core features; (5) Take the absolute value of the weighted Bz numerical matrix Bz3 to form a positive value matrix. Determine the threshold k based on the detection scene experience, and perform binarization on the positive value matrix: when the element value ≥ k, it is set to 1, representing the potential crack feature point; otherwise, it is set to 0, representing the background area. Extract all connected areas with values ​​1 to form the key numerical matrix area T containing the crack-related feature points. key ; (6) The key numerical matrix area T after binarization key The coordinates of the centroid of each independent area in the unconnected area are calculated. The centroid is used as the key feature point for detecting whether there is a crack branch and is called a node. (7) Based on the improved Boruvka algorithm, a new edge weight is designed for the connecting edges between nodes, and gradient information reflecting the clarity of the crack edge is introduced. If there is a bifurcation crack in the metal part, the generated minimum spanning tree is an effective representation of the surface contour of the bifurcation crack in the metal part.

2. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 1, characterized in that: In step (4), the formula of the composite weight function W based on the gradient magnitude matrix is: W=exp(-βG)+γBz 2 2, where β and γ are weight coefficients.

3. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 3, characterized in that: In step (4), β = 0.2, γ = 0.

8.

4. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 1, wherein The two nodes are (x i ,y i ) and (x j ,y j ), calculate the mean gradient G of the edge midpoint according to the weighted Bz numerical matrix Bz3 matrix mean The method is as follows: construct a detection point grid of the detection area based on the detection point array, where each grid contains one detection point. Connect the detection point grids where two nodes are located with a straight line. Select corresponding detection points based on the detection point grids that this straight line passes through. Average the gradient amplitudes corresponding to the selected detection points and use them as the gradient mean of the midpoint of side AB.

5. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 1, characterized in that: In step (7), the two nodes (x i ,y i ) and (x j ,y j ) is used to measure the actual distance between nodes in space, and the formula is:

6. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 1, characterized in that: In step (7), the weight of each edge W edge From the Euclidean distance D and the mean gradient G of the edge midpoint mean Jointly determined, the calculation formula is W edge =(1+αG mean )·D, where G mean It is the mean gradient of the edge midpoint calculated based on the weighted Bz numerical matrix Bz3 matrix. D is the Euclidean distance between two nodes, and α is the gradient weight coefficient, which ranges from 0.5 to 1.

0.

7. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 6, characterized in that: In step (7), the implementation steps of the improved Boruvka algorithm are as follows: 1) Initialization: All centroids are treated as independent nodes, and each node constitutes an initial connected component, that is, T = (V, {}), where V is the node set; 2) Iterative merging, the method is as follows: a. Select the minimum cross-component edge for each connected component: For all current connected components, traverse the edges between them and all other connected components and calculate the weight W of each edge edge , and select for each connected component an edge with the minimum weight connecting it to other connected components; b. Filter invalid edges and merge components: After collecting the minimum weight edges selected for all connected components, only the edges connecting different connected components are retained, which are called valid edges. For each valid edge, the two connected components connected by it are merged into a new connected component. c. Iterate until fully connected: Check whether all nodes belong to the same connected component. If not, repeat steps a and b for the remaining connected components, recalculate the edge weights and select the minimum weight edges for merging. Finally, when all nodes are connected, the minimum spanning tree generated at this time is a valid representation of the surface contour of the bifurcation crack in the metal part.

8. The method for reconstructing the surface profile of a non-ferromagnetic metal material based on a rotating AC electromagnetic field according to claim 1, characterized in that: The two U-shaped magnetic cores of the double U-shaped orthogonal structure excitation probe are perpendicular to each other, and an excitation coil is wound on each U-shaped magnetic core. The TMR magnetic sensor chip is arranged below the excitation probe, and its sensitive axis direction is perpendicular to the surface of the metal part being measured.

Images