A method for reconstructing the surface defect contour of magneto-optical imaging based on multi-directional detection

Abstract

The present invention discloses a method for reconstructing the surface defect contour based on multi-directional detection by magneto-optical imaging. The method uses a magneto-optical imaging detection device to detect and collect the surface defect information of ferromagnetic materials online, and establishes a magnetic field intensity distribution matrix based on the extracted defect signals. Then, by calculating the partial derivative of the signal obtained by the forward simulation of the physical model with respect to the surface magnetic resistance distribution of the measured part, a sensitivity matrix is established. Next, an iterative inversion method is introduced to invert the magnetic field distribution signal to obtain the magnetic resistance distribution data, realizing the high-precision reconstruction of the defect contour. The operation is simple, and compared with the traditional defect contour reconstruction methods, it has obvious advantages in terms of accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of defect detection, and more specifically, relates to a method for reconstructing the surface defect contour of magneto-optical imaging based on multi-directional detection. Background Art

[0002] Magnetic flux leakage detection refers to that after a ferromagnetic material is magnetized, a magnetic flux leakage field is formed on its surface due to defects on the surface or near the surface of the test piece, and the shape of the defect is reconstructed by detecting the magnetic field leakage on the defect surface. However, the current magnetic flux leakage detection only magnetizes along a single direction. Due to the magnetic flux leakage distortion of complex defects, it shows complex diffusion and distortion at the defect edge, resulting in inaccurate reconstruction of the notch contour.

[0003] The existing solutions currently include using low-frequency orthogonal excitation to generate a magnetic field whose direction changes with time. By analyzing the magneto-optical images at different moments within a period, defect information can be accurately extracted and identified. However, this type of method has a complex magnetization device and a large amount of data, and the parameters need to be re-adjusted for different scenarios; or defect images are obtained from different angles and directions, and then the enhanced ant colony algorithm and the average image method are used to reconstruct the defect. However, this type of method is easily affected by background noise and cannot effectively display the edge shape in the defect. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for reconstructing the surface defect contour of magneto-optical imaging based on multi-directional detection. By applying the iterative inversion method to the shape reconstruction of multi-directional magnetization magnetic flux leakage, the defect contour can be quickly identified, with high inversion efficiency, simple operation, and good performance in terms of speed and accuracy.

[0005] To achieve the above-mentioned invention purpose, a method for reconstructing the surface defect contour of magneto-optical imaging based on multi-directional detection according to the present invention is characterized by including the following steps:

[0006] (1). Use the MOI magneto-optical imaging technology to collect the magnetic flux leakage images of the test piece under multiple magnetization directions, denoted as M1, M2, …, M , d ,

[0008] , N ,

[0007] ,

[0006] ,

[0005] , d , d ,

[0009] , d , , , , , …, M N , where M d represents the magnetic flux leakage image collected in the d-th direction, N is the number of directions for collecting the magnetic flux leakage images, and the size of each magnetic flux leakage image is m×n;

[0007] First, convert the magnetic flux leakage image collected in each direction into a one-dimensional vector, and then construct a matrix λ:

[0008]

[0009] where v d represents the one-dimensional vector after converting the magnetic flux leakage image M d ;

[0010] (2) Construct a zero matrix \(g_0\) of size \(m\times n\), where the value of each element in \(g_0\) represents the initial magnetoresistance magnitude;

[0011] (3) Set the number of iterations \(k\), initialize \(k = 1\); set the threshold \(e\); set the forward transformation function \(F\) from the magnetoresistance matrix to the magnetic field distribution matrix in the \(d\)-th direction d (g);

[0012] (4) In the \(k\)-th iteration, change the magnetoresistance at \((i, j)\) in the \(d\)-th direction in the matrix \(g\) k-1 from \(R\) before to \(R\) after , and keep the magnetoresistance at other positions unchanged to obtain the modified magnetoresistance distribution in the \(d\)-th direction where \(R\) before represents the magnetoresistance before modification, and \(R\) after represents the magnetoresistance after modification; substitute the magnetoresistance distribution into the forward transformation function to obtain the magnetic field distribution in the \(d\)-th direction

[0013] (5) Calculate the change rate of the magnetic field distribution generated by the test piece at \((i, j)\) in the \(d\)-th direction

[0014]

[0015] where represents the magnetic field distribution when the magnetoresistance of all pixel points in the \(d\)-th direction is \(R\) before ;

[0016] (6) According to steps (4) and (5), calculate the change rate of the magnetic field distribution generated by the test piece at each pixel position in different magnetization directions;

[0017] (7) Construct the sensitivity matrix \(S\) after the \(k\)-th iteration k-1 :

[0018]

[0019] (8) Construct the magnetic field distribution matrix \(\lambda\) after the \(k\)-th iteration k-1 :

[0020]

[0021] (7) Substitute the sensitivity matrix \(S\) k-1 and the magnetic field distribution matrix \(\lambda\) k-1 into the iteration equation of magnetoresistance:

[0022]

[0023] Wherein, the superscript T represents transpose, β is a constant, and E represents an identity matrix of size m×n rows and m×n columns;

[0024] (8) Convert the iteration result from a one-dimensional vector to a matrix Δg1 of m rows and n columns, and then partition Δg1 to obtain the updated inversion result Δg;

[0025] (9) Calculate the k-th iteration result g k = g k-1 +Δg, and then change the values greater than 1 in g k to 1 and the values less than 0 to 0, so as to obtain the final result g k after the k-th iteration;

[0026] (10) Judge whether holds. If it holds, go to step (11); otherwise, let k = k + 1, and then return to step (4) for the next round of iteration;

[0027] (11) Output the magnetoresistance distribution matrix g = g k , and then convert g into an image, thus completing the reconstruction of the defect contour.

[0028] The invention purpose of the present invention is realized as follows:

[0029] The magneto-optical imaging surface defect contour reconstruction method based on multi-directional detection of the present invention detects and collects the surface defect information of ferromagnetic materials online through a magneto-optical imaging detection device, and establishes a magnetic field intensity distribution matrix according to the extracted defect signals; then, by calculating the partial derivative of the signal of the physical model forward simulation with respect to the magnetoresistance distribution on the surface of the measured part, a sensitivity matrix is established; then, an iterative inversion method is introduced, and by inverting the magnetic field distribution signal, magnetoresistance distribution data is obtained, realizing high-precision reconstruction of the defect contour, with simple operation and obvious advantages in accuracy compared with traditional defect contour reconstruction methods.

[0030] Meanwhile, the magneto-optical imaging surface defect contour reconstruction method based on multi-directional detection of the present invention also has the following

[0031] beneficial effects:

[0032] (1) The present invention introduces an iterative inversion method applied to the shape reconstruction of multi-directional magnetizing magnetic flux leakage, which can efficiently reconstruct complex defect contours. For the scenario where multiple crack defects are coupled with each other, the present invention has obvious advantages in accuracy;

[0033] (2) In the actual application process of this method, due to the high-resolution magnetic field imaging effect of magneto-optical imaging, the sensitivity matrix can be reused among the same type of data, so the recognition speed is relatively fast;

[0034] (3) In the iterative process of this method, the inversion results updated each time are segmented by a threshold method based on data statistics, which greatly improves the iteration speed and the approximation speed to the true contour.

[0035] (4) This method uses a physical model as the forward model, which has certain stability and accuracy, can better meet the requirements of reconstructing the surface defect contours of ferromagnetic materials based on magneto-optical imaging magnetic flux leakage detection, and has broad application scenarios. Description of the Drawings

[0036] Figure 1 is the flowchart of the magneto-optical imaging surface defect contour reconstruction method based on multi-directional detection;

[0037] Figure 2 is the schematic diagram of the test piece;

[0038] Figure 3 is the schematic diagram of the magneto-optical imaging device;

[0039] Figure 4 is the schematic diagram of the magnetic flux leakage image, where (a) is the magneto-optical image in the X direction and (b) is the magneto-optical image in the Y direction;

[0040] Figure 5 is the result diagram of defect reconstruction. Detailed Embodiments

[0041] The following describes the detailed embodiments of the present invention with reference to the drawings, so that those skilled in the art can better understand the present invention. It should be particularly noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0042] Embodiment

[0043] Figure 1 is the flowchart of the magneto-optical imaging surface defect contour reconstruction method based on multi-directional detection of the present invention.

[0044] In this embodiment, as Figure 1 shown, a magneto-optical imaging surface defect contour reconstruction method based on multi-directional detection of the present invention includes the following steps:

[0045] S1. In this embodiment, the test piece is an iron plate with a thickness of 5 mm, and the defect of the test piece is in a T shape, as Figure 2 shown. Based on the principle of the MOI magneto-optical imaging technology, the magnetic flux leakage images of the test piece in two magnetization directions with a 90° difference are collected according to the Figure 3 shown magneto-optical imaging device, denoted as M1 and M2. The size of each magnetic flux leakage image is m×n = 100*100, asFigure 4 As shown;

[0046] The magnetic leakage images collected in each direction are first converted into one-dimensional vectors, and then the matrix λ is constructed:

[0047]

[0048] Among them, v1 and v2 represent the one-dimensional vectors after the transformation of the magnetic leakage images M1 and M2 respectively;

[0049] S2. Construct an m×n zero matrix g0, where each element in g0 represents the initial magnetic resistance. In this embodiment, only the magnetic resistance of the test device and air is considered. Therefore, at the initial moment, the element value in g0 is 0, representing the magnetic resistance of air.

[0050] S3, set the number of iterations k, initialize k = 1; set the threshold e = 0.1; set the forward transformation function F from the magnetoresistance matrix to the magnetic field distribution matrix in the dth direction d (g), d∈[1,2];

[0051] S4. In the kth iteration, the matrix g k-1 The magnetic resistance of (i, j) in the dth direction is R before Change to R after , the magnetic resistance at other positions remains unchanged, and the modified magnetic resistance distribution in the dth direction is obtained Among them, R before Represents the magnetic resistance before modification, R after represents the modified reluctance;

[0052] In this embodiment, when k=1, we transform the matrix g k-1 The magnetic resistance of (i, j) in the dth direction is increased from 0 to 1, and the other magnetic resistances remain unchanged at 0, and then the same is applied in subsequent iterative calculations;

[0053] The magnetic resistance distribution Substituting the forward transformation function, we can get the magnetic field distribution in the dth direction:

[0054] S5. Calculate the rate of change of the magnetic field distribution generated by the test piece at (i, j) in the dth direction

[0055]

[0056] in, The magnetic resistance of all pixels in the dth direction is R before The magnetic field distribution at ;

[0057] S6. Calculate the change rate of the magnetic field distribution generated by the test piece at each pixel position in different magnetization directions according to steps S4 and S5;

[0058] S7. Construct the sensitivity matrix S after the k-th iteration k-1 :

[0059]

[0060] (8). Construct the magnetic field distribution matrix λ after the k-th iteration k-1 :

[0061]

[0062] S7. Substitute the sensitivity matrix S k-1 and the magnetic field distribution matrix λ k-1 into the iterative equation of magnetoresistance:

[0063]

[0064] where the superscript T represents transpose, β is a constant with a value of 1; E represents the identity matrix of size m×n rows and m×n columns;

[0065] S8. Convert the iterative result from a one-dimensional vector to a matrix Δg1 with m rows and n columns, and then divide Δg1 to obtain the updated inversion result Δg;

[0066] The following describes the specific process of dividing Δg1, which is as follows:

[0067] Calculate the mean μ and standard deviation σ of all elements in the matrix Δg1; construct a zero matrix Δg of the same dimension as Δg1; define the positive value L = 3;

[0068] Traverse each element Δg1(i,j) in the matrix Δg1. When μ - L×σ < Δg1(i,j) < μ + L×σ, keep the value of Δg1(i,j) unchanged; when Δg1(i,j) < μ - L×σ, modify the value of Δg1(i,j) to -1; when Δg1(i,j) > μ + L×σ, modify the value of Δg1(i,j) to 1, so as to obtain the updated inversion result Δg.

[0069] S9. Calculate the k-th iteration result g k = g k-1 + Δg, and then change the values in g k that are greater than 1 to 1 and the values less than 0 to 0, so as to obtain the final result g k ;

[0070] S10. Judge Whether it holds. If it holds, proceed to step (11); otherwise, let k = k + 1, and then return to step (4) for the next round of iteration;

[0071] S11. Output the magnetoresistance distribution matrix g = g k , and then convert g into an image, as Figure 5 shown, thus completing the defect contour reconstruction.

[0072] Although the above-described illustrative specific embodiments of the present invention have been described to facilitate understanding of the present invention by those skilled in the art, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions made using the concept of the present invention are within the scope of protection.

Claims

1. A method for reconstructing the surface defect contour of magneto-optical imaging based on multi-directional detection, characterized in that, Including the following steps: (1) Use the MOI magneto-optical imaging technology to collect the magnetic flux leakage images of the test piece in multiple magnetization directions, denoted as M1, M2, …, M d , …, M N , where M d represents the magnetic flux leakage image collected in the d-th direction, N is the number of directions for collecting the magnetic flux leakage images, and the size of each magnetic flux leakage image is m×n; First, convert the magnetic flux leakage images collected in each direction into one-dimensional vectors, and then construct a matrix λ: Among them, v d represents the magnetic flux leakage image M d transformed one-dimensional vector; (2) Construct a zero matrix g0 of size m×n, where each element value in g0 represents the initial magnetoresistance magnitude; (3) Set the number of iterations k, and initialize k = 1; set the threshold e; set the forward transformation function F d (g) from the magnetoresistance matrix to the magnetic field distribution matrix in the d-th direction; (4) In the k-th iteration, matrix g k-1 the magnetoresistance in the (i, j) position in the d-th direction is changed from R before to R after , and the magnetoresistance at other positions remains unchanged, obtaining the modified magnetoresistance distribution in the d-th direction where R before represents the magnetoresistance before modification, and R after represents the magnetoresistance after modification; substituting the magnetoresistance distribution into the forward transformation function to obtain the magnetic field distribution in the d-th direction (5) Calculate the rate of change of the magnetic field distribution generated by the test piece at (i, j) in the d-th direction Among them, represents the magnetic field distribution when the magnetoresistance of all pixel points in the d-th direction is R; before time (6) According to steps (4) and (5), calculate the magnetic field distribution change rate generated by the test piece at each pixel position in different magnetization directions; (7) Construct the sensitivity matrix S after the k-th iteration k-1 : (8) Construct the magnetic field distribution matrix λ after the k-th iteration k-1 : (9), Substitute the sensitivity matrix S k-1 and the magnetic field distribution matrix λ k-1 into the iterative equation of magnetoresistance: Among them, the superscript T represents transpose, β is a constant, and E represents an identity matrix of size m×n rows and m×n columns; (10) Convert the iteration result from a one-dimensional vector into an m-row and n-column matrix Δg1, and then divide Δg1 to obtain the updated inversion result Δg; (11) Calculate the result \(g\) of the \(k\)-th iteration k = \(g\) k-1 + \(\Delta g\), and then change the values in \(g\) k that are greater than 1 to 1 and the values less than 0 to 0, so as to obtain the final result \(g\) k ; (12), Determine If it holds, go to step (11); otherwise, let k = k + 1, and then return to step (4) for the next iteration; (13), Output magnetoresistance distribution matrix g = g k , and then convert g into an image to complete the reconstruction of the defect contour.

2. The method for reconstructing the surface defect profile of magneto-optical imaging based on multi-directional detection according to claim 1, wherein The specific process of splitting Δg1 in step (10) is as follows: Calculate the mean μ and standard deviation σ of all elements in the matrix Δg1; construct a zero matrix Δg of the same dimension as Δg1; define a positive value L; Traverse each element Δg1(i,j) in the matrix Δg1. When μ - L×σ < Δg1(i,j) < μ + L×σ, keep the value of Δg1(i,j) unchanged; when Δg1(i,j) < μ - L×σ, modify the value of Δg1(i,j) to -1; when Δg1(i,j) > μ + L×σ, modify the value of Δg1(i,j) to 1, so as to obtain the updated inversion result Δg.

Images