A method for intelligent inversion and size measurement of surface cracks in rotating electromagnetic field structures

Abstract

The present invention provides a method for intelligent inversion and size measurement of cracks on the surface of a rotating electromagnetic field structure. First, the Bx, By, and Bz matrices of the inspected structure are obtained by scanning, and then the Bz magnetic field image is identified by using a crack identification artificial neural network to obtain the presence or absence of cracks; further, Bx, By, and Bz are used as three channels to form a new B color image, and the mapping relationship between the crack image and the B color image is obtained by using a crack inversion artificial neural network, thereby obtaining information on the length, depth, and direction of cracks on the surface of the structure. The present invention can realize intelligent identification of cracks on the surface of the structure, reduce the misjudgment rate of manual judgment, use magnetic field signals in three directions to invert crack morphology, and improve accuracy; the length, depth, and direction of cracks in any direction on the surface of the structure can be accurately obtained, providing data support for safety assessment and maintenance decisions of the structure.

Description

Technical Field

[0001] The invention relates to the field of nondestructive testing signal processing, and in particular to a method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure. Background Art

[0002] The rotating electromagnetic field is a new type of electromagnetic nondestructive testing technology developed on the basis of AC electromagnetic field, eddy current and other detection technologies. It adopts orthogonal excitation to achieve high-sensitivity detection of cracks in any direction and has broad application prospects in the field of electromagnetic nondestructive testing.

[0003] Cracks are the most common manifestation of structural damage. They are very easy to expand under conditions such as alternating loads and vibrations, which in turn cause structural failure, seriously threatening the safe operation of equipment and human life. Therefore, it is necessary to perform crack detection and size measurement on the structure to obtain information on cracks in the structure and provide data support for equipment safety assessment and maintenance decisions. However, currently rotating electromagnetic fields can only realize manual identification of cracks, with a high misjudgment rate, and are unable to obtain crack size information, which limits the promotion and application of this technology. Summary of the invention

[0004] The problem to be solved by the present invention is to propose a method for intelligent inversion and size measurement of surface cracks in a rotating electromagnetic field structure, to realize intelligent identification and inversion of surface cracks, to obtain information on the length, depth and direction of the cracks, and to provide data support for safety assessment and maintenance decisions of the structure.

[0005] The present invention provides a method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure, the steps of which include:

[0006] S1: The manipulator clamps the rotating electromagnetic field probe to complete the scanning of the measured structure, and collects the Bx, By, and Bz matrices, where the sizes of the three matrices are unified to M*M; Bx is the magnetic flux density in the X direction; By is the magnetic flux density in the Y direction, and Bz is the magnetic flux density in the Z direction, where the direction parallel to the scanning direction is the X direction, the direction perpendicular to the scanning direction and parallel to the surface of the measured structure is the Y direction, and the direction perpendicular to the scanning direction and perpendicular to the surface of the measured structure is the Z direction;

[0007] S2: Since the Bx and By of the rotating electromagnetic field are more sensitive to interference factors such as lift-off jitter, while Bz is only sensitive to cracks, Bz is selected for crack identification. First, divide the Bz matrix by Bz max , to achieve normalization, the Bz max The maximum value of the magnetic flux density in the Z direction that can be collected by the rotating electromagnetic field probe is then multiplied by 255, the normalized matrix is ​​mapped to the pixel interval, and finally converted into a Bz grayscale image;

[0008] S3: The Bz grayscale image is identified by a pre-trained crack identification artificial neural network. Compared with manual defect determination, this method can eliminate subjective factors and significantly reduce the misjudgment rate. If it is identified as a crack, the next step is performed; if it is not identified as a crack, the process ends.

[0009] S4: Divide the Bx matrix by Bx max , the Bx max is the maximum value of the magnetic flux density in the X direction that can be collected by the rotating electromagnetic field probe, then multiplied by 255 and finally converted into a Bx grayscale image; divide the By matrix by By max , said By max is the maximum value of the magnetic flux density in the Y direction that can be collected by the rotating electromagnetic field probe, then multiplied by 255, and finally converted into a By grayscale image; since the length, depth, and direction of the crack have an impact on Bx, By, and Bz of the rotating electromagnetic field, and the impacts are mutually crossed, it is difficult for a single direction magnetic field to accurately obtain comprehensive crack information, so the Bx, By, and Bz grayscale images are divided into one, two, and three channels to form a B color image;

[0010] S5: input the B color image into a pre-trained crack inversion artificial neural network, output a surface crack grayscale image, then divide the pixel value of the surface crack grayscale image by 255, and then multiply it by P to convert it into a crack depth matrix, the size of the matrix is ​​M*M;

[0011] S6: Since the crack is a narrow and long defect, the depth value of the crack area output by the crack inversion artificial neural network is discrete, and there is a large deviation between the maximum value and the actual maximum depth of the crack. Therefore, it is necessary to take the average value of the largest N numbers in the crack depth matrix as the maximum depth of the surface crack, take the distance between the two ends of the crack as the length of the crack, and obtain the direction of the crack through the two end points of the crack;

[0012] The M is the number of rows or columns of the matrix;

[0013] N=M / 10, if it is not an integer, it is rounded up;

[0014] The P is the maximum value of the detected crack depth.

[0015] Furthermore, the crack identification artificial neural network includes Yolo, SSD and Faster-RCNN.

[0016] Furthermore, the crack inversion artificial neural network includes Pix2Pix and Unet; the network input is the B color image, and the output is the surface crack grayscale image, wherein the construction process of the surface crack grayscale image in the database during training is as follows:

[0017] First, a matrix of size M*M with all values ​​being 0 is constructed, then the values ​​of the crack areas are changed to the crack depth values, the matrix is ​​divided by P, multiplied by 255, and finally converted into the surface crack grayscale image.

[0018] Furthermore, the rotating electromagnetic field probe adopts an orthogonal excitation mode.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] (1) This invention applies artificial intelligence methods to the inversion and size measurement of surface cracks in rotating electromagnetic field structures for the first time, which significantly improves the efficiency of defect identification, reduces the error rate of manual judgment, and can accurately obtain information on the length, depth and direction of cracks.

[0021] (2) The present invention selects Bz for crack identification, which can greatly reduce the impact of factors such as lift-off jitter and greatly improve the accuracy of identification.

[0022] (3) Since the length, depth and direction of the crack have an impact on the Bx, By and Bz of the rotating electromagnetic field, and the influences are cross-linked, it is difficult to accurately obtain comprehensive crack information with a single-direction magnetic field. The present invention divides the Bx, By and Bz grayscale images into one, two and three channels to form a B color image, and then obtains the crack image through a crack inversion artificial neural network, and then obtains the crack size information. This method improves the accuracy of crack inversion and size measurement through three-dimensional magnetic field signals in space.

[0023] (4) Since cracks are narrow and long defects, the depth value of the crack area output by the crack inversion artificial neural network is discrete, and there is a large deviation between the maximum value and the actual maximum crack depth. The present invention uses the average value of the largest N numbers in the crack depth matrix as the maximum depth of the surface crack, which can greatly improve the accuracy of crack depth measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of intelligent inversion and size measurement of surface cracks in a rotating electromagnetic field structure in an embodiment of the present application.

[0025] Figure 2 It is the Bz grayscale image in the embodiment of the present application.

[0026] Figure 3 It is the Bx grayscale image in the embodiment of the present application.

[0027] Figure 4 It is a By grayscale image in the embodiment of the present application.

[0028] Figure 5 This is a grayscale image of a surface crack in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.

[0030] The present invention provides a method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure, the steps of which include:

[0031] S1: The manipulator clamps the rotating electromagnetic field probe to complete the scanning of the structure under test, and collects the Bx, By, and Bz matrices, where the sizes of the three matrices are unified to 100*100; Bx is the magnetic flux density in the X direction; By is the magnetic flux density in the Y direction, and Bz is the magnetic flux density in the Z direction, where the direction parallel to the scanning direction is the X direction, the direction perpendicular to the scanning direction and parallel to the surface of the structure under test is the Y direction, and the direction perpendicular to the scanning direction and perpendicular to the surface of the structure under test is the Z direction;

[0032] S2: Since the Bx and By of the rotating electromagnetic field are more sensitive to interference factors such as lift-off jitter, while Bz is only sensitive to cracks, Bz is selected for crack identification; first divide the Bz matrix by 5.2 to achieve normalization, 5.2 is the maximum value of the magnetic flux density in the Z direction that can be collected by the rotating electromagnetic field probe, then multiply by 255, map the normalized matrix to the pixel interval, and finally convert it into a Bz grayscale image, such as Figure 2 As shown;

[0033] S3: The Bz grayscale image is identified through the pre-trained crack identification Yolo network. Compared with manual defect determination, this method can eliminate subjective factors and significantly reduce the misjudgment rate; cracks are identified in the Bz grayscale image;

[0034] S4: Divide the Bx matrix by 4.5, where 4.5 is the maximum value of the magnetic flux density in the X direction that can be collected by the rotating electromagnetic field probe, then multiply it by 255 and finally convert it into a Bx grayscale image. Figure 3 As shown; divide the By matrix by 4.5, 4.5 is the maximum value of the Y-direction magnetic flux density that the rotating electromagnetic field probe can collect, then multiply it by 255, and finally convert it into a By grayscale image as shown Figure 4 As shown in the figure; since the length, depth and direction of the crack have an impact on the Bx, By and Bz of the rotating electromagnetic field, and the impacts are cross-linked, it is difficult for a single direction magnetic field to accurately obtain comprehensive crack information; the Bx, By and Bz grayscale images are divided into one, two and three channels to form the B color image;

[0035] S5: Input the color image B into the pre-trained crack inversion Unet network, and output the surface crack grayscale image as Figure 5 As shown, the pixel value of the surface crack grayscale image is then divided by 255 and then multiplied by 6 to convert it into a crack depth matrix with a matrix size of 100*100;

[0036] S6: Since the crack is a narrow and long defect, the depth value of the crack area output by the crack inversion artificial neural network is discrete, and there is a large deviation between the maximum value and the actual maximum crack depth. Therefore, it is necessary to take the average value of the 10 largest numbers in the crack depth matrix as the maximum surface crack depth, and the depth value is 5.5 mm. The distance between the two ends of the crack is taken as the length of the crack, and the crack length is 10.5 mm. The direction of the crack is obtained through the two end points of the crack, and the direction is 0°.

[0037] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent inversion and size measurement of surface cracks in a rotating electromagnetic field structure, characterized in that: S1: The manipulator holds the rotating electromagnetic field probe to scan the structure under test and collect Bx, By, and Bz matrices, where the sizes of the three matrices are all unified to M*M; S2: First divide the Bz matrix by Bz max , then multiply by 255 and finally convert to Bz grayscale image; S3: Identify the Bz grayscale image through a pre-trained crack recognition artificial neural network. If it is identified as a crack, proceed to S4; if it is not identified as a crack, end; S4: Divide the Bx matrix by Bx max , then multiply by 255, and finally convert to Bx grayscale image; divide the By matrix by By max , then multiply by 255 and finally convert to By grayscale image; use Bx, By, Bz grayscale images as the first, second, and third channels respectively to form B color image; S5: input the B color image into a pre-trained crack inversion artificial neural network, output a surface crack grayscale image, then divide the pixel value of the surface crack grayscale image by 255, and then multiply it by P to convert it into a crack depth matrix, the size of the matrix is ​​M*M; S6: taking the average value of the largest N numbers in the crack depth matrix as the maximum depth of the surface crack, taking the distance between the two ends of the crack as the length of the crack, and obtaining the direction of the crack through the two end points of the crack; Bx is the magnetic flux density in the X direction; By is the magnetic flux density in the Y direction; Bz is the magnetic flux density in the Z direction, wherein the direction parallel to the scanning direction is the X direction, the direction perpendicular to the scanning direction and parallel to the surface of the structure being measured is the Y direction, and the direction perpendicular to the scanning direction and perpendicular to the surface of the structure being measured is the Z direction; The Bz max is the maximum value of the magnetic flux density in the Z direction that can be collected by the rotating electromagnetic field probe, and Bx max is the maximum value of the magnetic flux density in the X direction that can be collected by the rotating electromagnetic field probe, and By max is the maximum value of the magnetic flux density in the Y direction that can be collected by the rotating electromagnetic field probe; The M is the number of rows or columns of the matrix; the P is the maximum value of the detected crack depth.

2. The method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure according to claim 1 is characterized in that: The N=M / 10 is rounded up if it is not an integer.

3. The method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure according to claim 1 is characterized in that: The crack recognition artificial neural networks include Yolo, SSD and Faster-RCNN.

4. The method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure according to claim 1 is characterized in that: The crack inversion artificial neural network includes Pix2Pix and Unet; the network input is the B color image, and the output is the surface crack grayscale image, wherein the construction process of the surface crack grayscale image in the database during training is as follows: First, a matrix of size M*M with all values ​​being 0 is constructed, then the values ​​of the crack areas are changed to crack depth values, the matrix is ​​divided by P, multiplied by 255, and finally converted into the surface crack grayscale image.

5. The method for intelligent inversion and size measurement of surface cracks of a rotating electromagnetic field structure according to claim 1 is characterized in that: The rotating electromagnetic field probe adopts an orthogonal excitation mode.

Images