Rotating electromagnetic field structure surface corrosion defect three-dimensional reconstruction method

Abstract

The invention provides a rotating electromagnetic field structure surface corrosion defect three-dimensional reconstruction method. The method comprises the following steps: firstly, establishing a rotating electromagnetic field structure surface corrosion defect magnetic field image sample library through simulation and experiments; then, a structure surface corrosion defect three-dimensional reconstruction network is constructed, the network structure adopts Pix2Pix, and a loss function adopts the combination of a data loss function and a physical loss function; thirdly, the established reconstruction network is trained, and an optimal solution is obtained; and finally deploying the trained network on a rotating electromagnetic field detection system. According to the method, firstly, an effective method is provided for three-dimensional reconstruction of the corrosion defects on the surface of the structure; moreover, as a large number of samples are difficult to obtain for the real structure surface corrosion defect and the manufacturing is expensive, the precise reconstruction of the three-dimensional contour of the structure surface corrosion defect under the small sample is realized by adopting a physical and data dual-drive mode, and reliable data support is provided for safety assessment and maintenance decision of the structure.

Description

Technical Field

[0001] The invention relates to the field of nondestructive testing signal processing, and in particular to a three-dimensional reconstruction method for corrosion defects on the surface of a rotating electromagnetic field structure. Background Art

[0002] 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 has the advantages of non-contact detection and high quantitative accuracy. At the same time, it makes up for the disadvantage of directional sensitivity of conventional electromagnetic nondestructive testing technology and can realize high-sensitivity detection of defects in any direction of the structure.

[0003] However, there are multiple nonlinear mapping relationships between the three-dimensional contours of corrosion defects on the surface of structures and the spatial magnetic field response signals of rotating electromagnetic fields. The equations for solving conventional inverse problems are ill-conditioned equations and do not have unique solutions. The rapid development of artificial intelligence technology has provided a new idea for the reconstruction of three-dimensional contours of corrosion defects on the surface of structures, but it relies on a large number of sample training, there are few industrial field samples of structural corrosion defects, and the cost of manual production is high. In response to the above problems, the present invention proposes a new three-dimensional reconstruction method for corrosion defects on the surface of structures with rotating electromagnetic fields. For the first time, the Pix2Pix network is applied to the three-dimensional reconstruction of corrosion defects in rotating electromagnetic fields. Then, according to the characteristics of the three-dimensional morphology of corrosion defects, physical and data loss functions are constructed, and finally the three-dimensional reconstruction and accurate evaluation of corrosion defects under small samples are achieved. Summary of the invention

[0004] The problem to be solved by the present invention is to propose a three-dimensional reconstruction method for surface corrosion defects of a rotating electromagnetic field structure, so as to realize the accurate reconstruction of the three-dimensional contour of structural corrosion defects under small samples, provide effective data support for the safety assessment of the structure, and effectively improve the intrinsic safety level of the equipment.

[0005] The present invention provides a three-dimensional reconstruction method for corrosion defects on the surface of a rotating electromagnetic field structure, the steps of which include: Step 1: Establish a sample library of magnetic field images of corrosion defects on the surface of rotating electromagnetic field structures. The sample library includes magnetic field images and real defect images. Step 2: Establish a 3D reconstruction network for corrosion defects on the surface of the structure. Since Pix2Pix can automatically capture the data distribution of the real sample set and obtain the mapping relationship between the input image and the output image, it has been widely used in image restoration, handwritten digit generation, face generation, image inversion, etc., and both the magnetic field image and the defect image can be obtained in this method, so the main network structure adopts Pix2Pix. The input of the network training is the magnetic field image, and the output is the generated defect image. The original loss function of the Pix2Pix network is not completely suitable for the reconstruction of the 3D contour of corrosion defects. Due to the limited samples of defects, the use of the original loss function will produce large errors. In order to obtain high-precision 3D contour reconstruction results of structural corrosion defects under small samples, according to the characteristics of 3D contour reconstruction of corrosion, physical and data loss functions are established to constrain the optimization direction of the network. The total loss function of the network is: in, , and is the weight coefficient, and is the data loss function, and For the physical loss function, the weight coefficient can be used to adjust the proportion of the data loss function or the physical loss function in the total loss function; The main function is to establish the adversarial relationship of the network, using the cross entropy loss function. The main function is to obtain the difference between the generated image and the real image, using the L1 loss function; The expression is:

[0006] The expression is:

[0007] The physical loss function is constructed based on the evaluation index of the defect. For corrosion defects, the main evaluation indexes are the volume and maximum depth of corrosion. Since the collected magnetic field image contains noise, it is easy for pixels with large deviations to appear in the generated defect image. In order to reduce the maximum depth error, the average value of the maximum 5 pixels in the image is taken as the maximum depth of the defect. The deviation between the generated defect depth and the actual defect depth is calculated, and the weight coefficient is added for correction to construct the Loss function; For the defect volume, the basic loss function only calculates the deviation between the real corrosion and the generated corrosion. Since the values ​​of non-defective areas are all 0, the pixel values ​​with large deviations are generally located in the corroded area. In order to obtain a more accurate defect volume, the weights of the 15 pixels with the largest deviations are increased to form Loss function. The corrosion defect loss function finally constructed is: The construction process is as follows: Convert the real defect image (M*M) and the generated defect image (M*M) into an array of 1 column and M*M rows, and arrange the two arrays in descending order; Take the first 5 rows of the two arrays after rearrangement to form a new array A (1*5) and B (1*5), The calculation process is:

[0008] In the formula, Representing an array A The i-th element of Representing an array B The i-th element of Indicates taking the absolute value; The construction process is as follows: The deviation matrix between the real defect image and the generated defect image V It can be expressed as:

[0009] In the formula, is the real defect image, is the defect image generated, x is the magnetic field image, z is a noise signal; due to the deviation matrix V Not all are positive numbers, so we need to find their absolute values ​​to get the absolute deviation matrix :

[0010] Convert the absolute deviation matrix into an array of 1 column and M rows, and arrange them in descending order, and select the first 15 rows to form a new array , The calculation process is:

[0011] In the formula, Represents the i-th row of array C; Step 3: Use the established rotating electromagnetic field structure surface corrosion defect magnetic field image sample library to train the established structure surface corrosion defect three-dimensional reconstruction network to obtain the optimal solution; Step 4: Deploy the trained network to the rotating AC electromagnetic field detection system.

[0012] Furthermore, the magnetic field image is constructed by the magnetic flux density perpendicular to the surface of the structure under test. Since there are few non-ferromagnetic corrosion samples in industrial sites and artificial production is expensive, the magnetic flux density is measured by a rotating electromagnetic field simulation model or experimental system.

[0013] Furthermore, the real defect image is constructed by using real three-dimensional features of the defect, and the length and width of the real defect image are consistent.

[0014] Furthermore, M is the number of pixels in length or width of the image.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) This paper applies the Pix2Pix network to the three-dimensional reconstruction of structural corrosion defects for the first time, providing an effective intelligent solution for defect inversion and making up for the deficiency that the traditional physical equation solving method does not have a unique solution.

[0016] (2) In order to address the problem that the accuracy of the three-dimensional corrosion contour reconstruction by neural network is low due to the small number of surface corrosion defect samples of industrial structures and the high cost of production, a physical loss function is added on the basis of the conventional data loss function to improve the accuracy of the three-dimensional corrosion contour reconstruction.

[0017] (3) The physical loss function is constructed based on the evaluation index of the defect. For corrosion defects, the main evaluation indexes are the corrosion volume and maximum depth. Therefore, the average value of the largest 5 pixels in the image is taken as the maximum depth of the defect. The deviation between the generated defect depth and the actual defect depth is calculated. Loss function; For the defect volume, the basic loss function only calculates the deviation between the real corrosion and the generated corrosion. Since the values ​​of non-defective areas are all 0, the pixel values ​​with large deviations are generally located in the corroded area. In order to obtain a more accurate defect volume, the weights of the 15 pixels with the largest deviations are increased to form Loss function; through The construction of the loss function improves the inversion of the maximum depth of corrosion defects by The construction of the loss function improves the inversion of the corrosion volume, and both further improve the accuracy of the three-dimensional reconstruction of corrosion defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Flow chart of three-dimensional reconstruction of structural corrosion defects in the embodiment of this application Figure 2 This is the magnetic field image in the embodiment of this application Figure 3 The defect image in the embodiment of the present application Figure 4 The three-dimensional reconstruction network of the corrosion defects on the surface of the structure in the embodiment of the present application DETAILED DESCRIPTION

[0019] 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.

[0020] The present invention provides a three-dimensional reconstruction method for corrosion defects on the surface of a rotating electromagnetic field structure, the steps of which include: The first step is to establish a sample library of magnetic field images of corrosion defects on the surface of rotating electromagnetic field structures. The sample library includes magnetic field images and real defect images, such as Figure 2 and 3 As shown; The second step is to establish a three-dimensional reconstruction network for corrosion defects on the surface of the structure. The main structure of the network adopts Pix2Pix. The input of the network training is the magnetic field image, and the output is the generated defect image, such as Figure 4 As shown, the length and width of the pixels are 100 respectively; The loss function is:

[0021] in, , and is the weight coefficient, which can be set to 1, 0.5, and 1.5; The expression is:

[0022] The expression is:

[0023] The construction process is as follows: Convert the real defect image (100*100) and the generated defect image (100*100) into an array of 1 column and 10,000 rows, and arrange the two arrays in descending order; Take the first 5 rows of the two arrays after rearrangement to form a new array A (1*5) and B (1*5), The calculation process is:

[0024] In the formula, Representing an array A The i-th element of Array B The i-th element of Indicates taking the absolute value; The construction process is as follows: The deviation matrix between the real defect image and the generated defect image V It can be expressed as:

[0025] In the formula, is the real defect image, is the defect image generated, x is the magnetic field image, z is a noise signal; due to the deviation matrixV Not all are positive numbers, so we need to find their absolute values ​​to get the absolute deviation matrix :

[0026] Convert the absolute deviation matrix into an array of 1 column and 100*100 rows, and arrange them in descending order, and select the first 15 rows to form a new array , The calculation process is:

[0027] In the formula, Represents the i-th row of array C; Step 3: Use the established rotating electromagnetic field structure surface corrosion defect magnetic field image sample library to build a three-dimensional reconstruction network for structural surface corrosion defects to train and obtain the optimal solution; Step 4: Deploy the trained network to the rotating AC electromagnetic field detection system.

[0028] 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. The present invention provides a three-dimensional reconstruction method for corrosion defects on the surface of a rotating electromagnetic field structure, characterized in that: S1: Establish a sample library of magnetic field images of corrosion defects on the surface of a rotating electromagnetic field structure, which includes magnetic field images and real defect images; S2: Establishing a three-dimensional reconstruction network for corrosion defects on the surface of the structure, wherein the main structure of the network adopts Pix2Pix, the input of the network training is the magnetic field image in S1, and the output is the generated defect image; The loss function is: ; in, , and is the weight coefficient; The expression is: ; The expression is: ; The construction process is as follows: Convert the real defect image (M*M) and the generated defect image (M*M) into an array of 1 column and M*M rows, and arrange the two arrays in descending order; Take the first 5 rows of the two arrays after rearrangement to form a new array A (1*5) and B (1*5), The calculation process is: ; In the formula, Representing an array A The i-th element of Representing an array B The i-th element of Indicates taking the absolute value; The construction process is as follows: The deviation matrix between the real defect image and the generated defect image V It can be expressed as: ; In the formula, is the real defect image, is the defect image generated, x is the magnetic field image, z is a noise signal; due to the deviation matrix V Not all are positive numbers, so we need to find their absolute values ​​to get the absolute deviation matrix : ; Convert the absolute deviation matrix into an array of 1 column and M*M rows, and arrange them in descending order, and select the first 15 rows to form a new array , The calculation process is: ; In the formula, Represents the i-th row of array C; S3: Use the rotating electromagnetic field structure surface corrosion defect magnetic field image sample library established in S1 to train the structure surface corrosion defect three-dimensional reconstruction network established in S2 to obtain the optimal solution; S4: Deploy the trained network to the rotating AC electromagnetic field detection system.

2. The three-dimensional reconstruction method of corrosion defects on the surface of a rotating electromagnetic field structure according to claim 1 is characterized in that: The magnetic field image is constructed by the magnetic flux density in a direction perpendicular to the surface of the structure under test, and the magnetic flux density is measured by a rotating electromagnetic field simulation model or an experimental system.

3. The three-dimensional reconstruction method of corrosion defects on the surface of a rotating electromagnetic field structure according to claim 1 is characterized in that: The real defect image is constructed by using real three-dimensional features of the defect, and the length and width of the real defect image are consistent.

4. The three-dimensional reconstruction method of corrosion defects on the surface of a rotating electromagnetic field structure according to claim 1 is characterized in that: The M is the number of pixels of the image length or width.

Images