Electromagnetic tomography method for detecting defects in carbon fiber cables

Abstract

The application relates to an electromagnetic tomography flaw detection method of a carbon fiber cable, which comprises the following steps: S1. constructing a simulation model of a carbon fiber composite cable, obtaining an initial sensitivity matrix and an initial voltage; S2. randomly setting a defect shape and position of the simulation model, applying excitation to the simulation model with the set defect, obtaining a detection voltage, and constructing a defect image; S3. repeating step S2 at least 600 times to obtain a defect image set and a detection voltage set; S4. iteratively updating the initial sensitivity matrix according to the initial voltage, the preprocessed defect image and the detection voltage; S5. constructing a defect positioning model; S6. applying a voltage to the measured carbon fiber cable to obtain a detection voltage of the measured carbon fiber cable; S7. constructing a cross-section image of the measured carbon fiber cable and inputting the trained defect positioning model, and determining a defect position according to an interface image with the defect. The application can update the sensitivity matrix, reduce the influence of micro-vibration noise on the imaging quality, and make the defect imaging clearer.

Description

Technical Field

[0001] The invention relates to an electromagnetic tomography detection method, in particular to an electromagnetic tomography flaw detection method for carbon fiber cables. Background Art

[0002] Carbon fiber composite cables overcome the heavy weight and surface corrosion challenges of traditional steel cables, meeting the load-bearing requirements of long-span cables and representing a key development direction for lightweighting ultra-large amusement rides and aerial passenger ropeways. Due to the influence of manufacturing processes and the service environment, carbon fiber composite cables inevitably experience defects such as debonding and strand breakage. The shape and location of these defects directly impact the load-bearing capacity and service life of the carbon fiber composite cables, and can even lead to cable failure and safety accidents. Therefore, conducting electromagnetic tomography of carbon fiber composite cable defects is crucial for ensuring their safe and reliable service.

[0003] Electromagnetic tomography (EMT) is a non-contact, non-invasive, nondestructive testing technology. Based on the principle of electromagnetic induction, it can visualize the conductivity distribution of the measured area and has been widely used in fields such as industrial flaw detection and multiphase flow detection. To improve the imaging quality of EM tomography systems, researchers have conducted extensive research on excitation mode selection, sensor array optimization, and image reconstruction algorithms. Image reconstruction algorithms can be divided into fast reconstruction algorithms and iterative reconstruction algorithms. Both require the prior knowledge of a sensitivity matrix, which is used in the image reconstruction calculation. Therefore, the sensitivity matrix plays a crucial role in the object field image reconstruction process. The sensitivity matrix is ​​mainly obtained through perturbation methods and field quantity extraction. The field quantity extraction method is superior in terms of acquisition speed and imaging quality. Current research shows that EM tomography has excellent imaging performance for defect detection in high-conductivity materials such as metals. However, for defect detection in low-conductivity materials such as carbon fiber cables, accurate identification of defect location and shape is difficult.

[0004] Due to the electrical anisotropy and low conductivity of carbon fiber cables, defects have little impact on the changes in the coil sensor's detection voltage, which is easily affected by micro-vibrations and mechanical noise. Furthermore, the physical field conductivity distribution obtained by traditional field quantity extraction methods, derived from the prior sensitivity matrix, is severely affected by edge artifacts, which can easily obscure the position and shape of the defect image, making defect detection and identification difficult. Summary of the Invention

[0005] The purpose of the present invention is to provide an electromagnetic tomography flaw detection method for carbon fiber cables to solve the problem of inaccurate recognition of the position and shape of defects in carbon fiber composite cables.

[0006] The purpose of the present application is achieved in that:

[0007] An electromagnetic tomography method for detecting defects in a carbon fiber cable, comprising the following steps:

[0008] S1. Constructing an electromagnetic simulation model of the carbon fiber cable according to the specifications of the measured carbon fiber cable, applying current excitation to the three-dimensional carbon fiber cable electromagnetic simulation model to obtain an initial voltage, and establishing an initial sensitivity matrix of the carbon fiber cable cross-section object field region by field extraction method;

[0009] S2. Randomly setting the shape and position of the defect in the simulation model, applying excitation to the simulation model with the defect set, obtaining the detection voltage; according to the defect shape and position set in the simulation model, drawing the cross-sectional image of the simulation model where the defect is located to obtain the defect image; performing edge gradual filling on the defect image and adding Gaussian noise to the detection voltage;

[0010] S3. Repeating step S2 at least 600 times to obtain a defect image set and a detection voltage set of the carbon fiber cable electromagnetic simulation model;

[0011] S4. Iteratively updating the initial sensitivity matrix according to the initial voltage, the defect image set and the detection voltage set;

[0012] S5. Constructing a defect positioning model for detecting internal defects in the carbon fiber cable;

[0013] S6. Setting a detection probe, the structure of which is that a plurality of annular coils pointing to the center of the circle are arranged on the inner ring surface of the circular carrier, which are divided into detection coils and excitation coils, the number of detection coils is the same as that of excitation coils, and they are uniformly distributed in double-layer annular around the object field, the inner layer coils are set as detection coils, and the outer layer coils are set as excitation coils; the detection probe moves from one end to the other end of the measured carbon fiber cable at a preset speed, and completes the collection of detection voltage every 300 milliseconds to obtain a set of detection voltages of the measured carbon fiber cable;

[0014] S7. According to the set of detection voltages of the measured carbon fiber cable and the iteratively updated sensitivity matrix, calculating the gray value of each pixel of the cross-section of the measured carbon fiber cable, generating the cross-sectional image of the measured carbon fiber cable according to the gray value of each pixel, inputting the cross-sectional image of the measured carbon fiber cable into the trained defect positioning model, and outputting the clear cross-sectional image of the measured carbon fiber cable, when there is a cross-sectional image with defects, determining the defect position through the collection time of the detection voltage corresponding to the cross-sectional image with defects.

[0015] Further, the specific way of iteratively updating the initial sensitivity matrix in step S4 is:

[0016] S4-1. The initial sensitivity matrix is iteratively updated, and the formula for iterative updating is:

[0017]

[0018] wherein S0 is the initial sensitivity matrix, S1 is the matrix after the first iterative update, G(S0) is the normalized pixel gray value vector data of the defect image corresponding to the sensitivity matrix S0;

[0019] The calculation formula of G(S0) is:

[0020]

[0021] wherein S0 T is the transpose matrix of S0, U * is a set of detection voltages obtained by randomly setting the defects of the simulation model, is the initial voltage;

[0022] S4-2. The normalized pixel gray value vector data of the defect image corresponding to the iteratively updated sensitivity matrix is calculated, and the calculation formula is:

[0023]

[0024] S4-3. The loss function is calculated according to the normalized pixel gray value vector data of the defect image:

[0025] Loss||G * -G(S1)||

[0026] wherein G * is a set of normalized pixel gray value vector data of the defect image obtained by randomly setting the defects of the simulation model;

[0027] S4-4. It is judged whether the loss function obtained in step S4-3 is lower than 1x10 -3 or the iteration is updated to a preset iteration number, when the judgment result is yes, the iteration update is stopped, and the updated sensitivity matrix S1 is output, when the judgment result is no, steps S4-5-S4-8 are executed;

[0028] S4-5. The sensitivity matrix is iteratively updated, and the formula for iterative updating is:

[0029]

[0030] wherein S i is the sensitivity matrix obtained after the i-th iterative update, S i-1 is the sensitivity matrix obtained after the i-1-th iterative update, a is the learning rate, indicating the step size of iteration, G(S i-1) is the sensitivity matrix S i-1 corresponding to the normalized pixel gray value vector data of the defect image;

[0031] G(S i-1 ) is calculated as follows:

[0032]

[0033] wherein S i T is the transpose matrix of S i , U * is a set of detection voltages obtained from defects of a randomly set simulation model, and is an initial voltage.

[0034] S4-6. The normalized pixel gray value vector data of the defect image corresponding to the updated sensitivity matrix is calculated as follows:

[0035]

[0036] S4-7. The loss function is calculated according to the normalized pixel gray value vector data of the defect image:

[0037] Loss||G * -G(S i )|

[0038] wherein G * is a set of normalized pixel gray value vector data of the defect image obtained from defects of a randomly set simulation model;

[0039] S4-8. It is determined whether the loss function output by step S4-7 is lower than 1x10 -3 or the iteration is updated to a preset iteration number, and when the determination result is no, steps S4-5-S4-7 are repeated until the determination result is yes, the iteration update is stopped, and the sensitivity matrix updated last time is output.

[0040] Further, the specific way of constructing the defect positioning model is as follows:

[0041] S5-1. The U-Net model is improved;

[0042] S5-2. The defect image set obtained in step S3 is labeled to obtain a labeled defect image set;

[0043] S5-3. According to the updated sensitivity matrix and the set of detection voltages, the pixel values of the pixel points of each image in the defect image set obtained in step S3 are recalculated, and the defect image set is reconstructed;

[0044] S5-4. The images in the image set labeled with defects and the image set after reconstruction of defects are input into the improved U-Net model for training to obtain a defect positioning model for detecting internal defects of the carbon fiber material cable.

[0045] Further, the specific way of improving the U-Net model is that the double convolution of the coding part of the U-Net deep learning network is replaced by a residual module, and the structure of each residual module is that the input of the residual module is divided into two paths, one path passes through a first convolution layer, a first batch normalization layer, an activation function, a second convolution layer and a second batch normalization layer, and the other path passes through a 1*1 convolution, and the outputs of the two paths are added and output after passing through an activation function.

[0046] Further, the iteration period of the defect positioning model training is 100 epochs, the Batch Size is set to 16, the initial learning rate is set to 0.0001, and the loss function adopts a cross-entropy loss function and a Dice coefficient.

[0047] The sensitivity matrix is first updated by using the random gradient descent method, the updated sensitivity matrix is more suitable for imaging of specific object field information, the influence of micro-vibration noise on the imaging quality is reduced, and the reconstructed defect image is clearer;Then, the U-Net network is used to construct a defect positioning model, improve the accuracy of the model for defect positioning, and position the defect image of the measured carbon fiber cable, weaken the influence of the artifact noise, and more intuitively display the position and size information of the defect, which is beneficial to the quantitative evaluation of the defect of the carbon fiber composite material cable. The two stages of sensitivity optimization and defect image optimization are combined to realize the quantitative detection and visual characterization of the defect of the low-conductivity carbon fiber material cable.

[0048] The present application is suitable for the electromagnetic tomography method for defect of low-conductivity carbon fiber composite material cable, which can weaken the micro-vibration and mechanical noise, reduce the influence of edge artifact, and realize the accurate identification of the position and shape of the defect of the carbon fiber composite material cable. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is the flow chart of the present application.

[0050] Figure 2 is the flow chart of randomly setting defects to the carbon fiber composite material cable model.

[0051] Figure 3 is the sensitivity matrix optimization flow chart.

[0052] Figure 4 is the second stage defect image reconstruction flow chart.

[0053] Figure 5 is the schematic diagram of detecting the measured carbon fiber cable.

[0054] Figure 6 It is a comparison chart of defect imaging at each stage. DETAILED DESCRIPTION

[0055] The present invention is described in further detail below.

[0056] like Figure 1 As shown, the electromagnetic tomography flaw detection method for carbon fiber cables provided by the present invention comprises the following steps:

[0057] S1. Construct an electromagnetic simulation model of the carbon fiber cable according to the specifications of the tested carbon fiber cable. Apply current excitation to the three-dimensional electromagnetic simulation model of the carbon fiber cable to obtain an initial voltage. Establish the initial sensitivity matrix of the material field region of the carbon fiber cable cross section through the field quantity extraction method.

[0058] The constructed simulation model is an ideal damage-free model, and the initial voltage obtained is the full-field voltage, and the specifications are the same as the parameters of the tested carbon fiber cable.

[0059] like Figure 2 As shown, the electromagnetic simulation model establishment and excitation detection process are implemented using COMSOL Multiphysics software. The electromagnetic simulation model of the carbon fiber cable mainly consists of a carbon fiber cable 1, an outer coil 2, and an inner coil 3. The carbon fiber cable 1 is composed of 7 carbon fiber rods, each of which is 100 mm long and 8 mm in diameter. The axial conductivity is set to 10060 (S / M) and the radial conductivity is set to 138 (S / M). The outer coil 2 is the excitation coil, and the inner coil 3 is the detection coil. The inner and outer coils have an inner diameter of 5 mm and an outer diameter of 10 mm. Both the outer and inner coils are wound with copper wire. One inner coil corresponds to one outer coil, and the central axes of the corresponding inner and outer coils are on the same straight line.

[0060] The excitation current of the excitation coil is set to 1A and the excitation frequency is set to 1MHz. When each excitation coil is excited, the seven detection coils except the one below the excitation coil (that is, except the detection coil corresponding to the excitation coil) are detected and voltages are collected. During each detection process, the eight excitation coils are excited in turn in a clockwise direction, so 56 voltage values ​​can be obtained in one detection process.

[0061] In the simulation model, a unit current is applied to each excitation coil and detection coil to extract the vector magnetic potential distribution of the measured object field area. The extracted independent vector magnetic potential distribution data is then combined with the excitation coil and detection coil in sequence to calculate the conductivity sensitivity matrix. The formula is as follows:

[0062] S=-ω 2 AA A B

[0063] Wherein, ω represents the working frequency of the electromagnetic detection system, A is the outer coil, A A is the vector magnetic potential distribution of the object field under A excitation, B is the inner coil, A B is the vector magnetic potential distribution of the object field under B excitation.

[0064] S2. Randomly set the shape and position of the defect of the simulation model, apply excitation to the simulation model with the set defect, obtain the detection voltage; draw the cross-sectional image of the simulation model where the defect is located according to the shape and position of the defect set in the simulation model, obtain the defect image; perform edge gradual filling on the defect image, and add Gaussian noise to the detection voltage.

[0065] In the simulation model, a defect is set at the cross section of the carbon fiber cable surrounded by the detection coil, a cuboid structure of a preset specification is subtracted from the geometric structure of the carbon fiber cable, the internal boundary is saved, a carbon fiber composite cable structure model containing the defect is formed, and the axial conductivity is set to 1006 S / m and the radial conductivity is set to 138 S / m. The preset specification of the cuboid structure used in the application is: 8mm in length, 4mm in width and 2mm in height.

[0066] As Figure 2 shown, the shape and position of the defect of the three-dimensional carbon fiber material cable simulation model are randomly set, the detection voltage value data of the multi-coil turn-by-turn excitation is collected, the detection voltage is obtained, the cross-sectional defect image of the simulation model is drawn according to the shape and position of the set defect, the correspondence between the defect image and the detection voltage is established, and the detection voltage and the defect image correspond to each other.

[0067] The randomly set defect is of the same type as the defect of the measured carbon fiber cable, for example, a broken strand defect.

[0068] The edge gradual filling on the defect image can effectively improve the smoothness of the defect imaging edge, and the addition of Gaussian noise to the detection voltage can increase the robustness of the defect positioning model.

[0069] S3. Repeat step S2 at least 600 times to obtain a defect image set of the carbon fiber cable electromagnetic simulation model and a detection voltage set corresponding to various defects.

[0070] Multiple defects are set to facilitate updating of the sensitivity matrix and providing samples for subsequent model training.

[0071] The spatial position of the cuboid structure of step S2 is rotated and moved, a plurality of defect type carbon fiber composite cable models are formed after difference set, the detection voltage value data of the multi-coil turn-by-turn excitation is collected, and the detection voltage set corresponding to various defects is obtained.

[0072] S4. Iteratively updating the initial sensitivity matrix according to the initial voltage, the pre-processed defect image and the detection voltage.

[0073] As Figure 3 shown, in order to solve the problem that the sensitivity matrix obtained by the traditional field extraction method is not applicable to the detection of low conductivity materials, the gradient descent idea is adopted to optimize the initial sensitivity matrix S0, and the sensitivity matrix more suitable for defect imaging of low conductivity carbon fiber material cable is established.

[0074] S4-1. Iteratively updating the initial sensitivity matrix, and the formula of iterative updating is:

[0075]

[0076] Wherein, S0 is the initial sensitivity matrix, S1 is the matrix after the first iterative update, G(S0) is the normalized pixel gray value vector data of the defect image corresponding to the sensitivity matrix S0;

[0077] The calculation formula of G(S0) is:

[0078]

[0079] Wherein, S0 T is the transpose matrix of S0, U * is a set of detection voltages obtained by randomly setting defects of a simulation model, is the initial voltage;

[0080] S4-2. Calculate the normalized pixel gray value vector data of the defect image corresponding to the sensitivity matrix after iterative updating, and the calculation formula is:

[0081]

[0082] S4-3. Calculate the loss function according to the normalized pixel gray value vector data of the defect image:

[0083] Loss||G * -G(S1)|| (4)

[0084] Wherein, G * is a set of normalized pixel gray value vector data of the defect image obtained by randomly setting defects of a simulation model;

[0085] S4-4. Determine whether the loss function obtained in step S4-3 is lower than 1x10 -3or iteration is updated to a preset iteration number, when the judgment result is yes, the iteration update is stopped, and the updated sensitivity matrix S1 is output, when the judgment result is no, steps S4-5-S4-8 are executed;

[0086] S4-5. The sensitivity matrix is iteratively updated, and the formula for iterative update is:

[0087]

[0088] wherein, S i is the sensitivity matrix obtained after the i-th iteration update, S i-1 is the sensitivity matrix obtained after the (i-1)-th iteration update, a is the learning rate, indicating the step size of iteration, and G(S i-1 ) is the normalized pixel gray value vector of the defect image corresponding to the sensitivity matrix S i-1 .

[0089] The calculation formula of G(S i-1 ) is:

[0090]

[0091] wherein, S i T is the transpose matrix of S i , U * is a set of detection voltages obtained by randomly setting defects of a simulation model, and the detection voltage of each column is the detection voltage of each time the defect is set, is the initial voltage, each column of data is the same, and the number of columns is the same as the number of columns of U * .

[0092] S4-6. The sensitivity matrix calculated according to formula (5) is used to calculate the corresponding defect image, and the calculation formula is:

[0093]

[0094] S4-7. The loss function is calculated according to the defect image:

[0095] Loss||G * -G(S i )|| (8)

[0096] wherein, G * is a set of normalized pixel gray value vector data of defect images obtained by randomly setting defects of a simulation model, i.e., a set of normalized pixel gray value vectors of all defect images after preprocessing in step S2.

[0097] The loss function of formula (4) can also use a smoothed L1 loss function and an L2 loss function, and the formulas of the smoothed L1 loss function and the L2 loss function are as follows:

[0098]

[0099] Wherein, β is the switching threshold of the L1 loss function and the L2 loss function.

[0100] The smoothed L1 loss function and the L2 loss function can be very robust to voltage signals containing noise interference, and can also take into account the influence of small errors.

[0101] S4-8. Determine whether the loss function output by step S4-7 is lower than 1x10 -3 or the preset number of iterations, when the determination result is no, repeat steps S4-5-S4-7 until the determination result is yes, stop the iterative update, and output the sensitivity matrix of the last iteration update.

[0102] In each iteration process, the gradient of the parameter S i is calculated, and the sensitivity matrix is updated in the opposite direction of the gradient until the loss function is lower than 1x10 -3 or the preset number of iterations stops the loop, and outputs the optimal sensitivity matrix S i . The damage function here can select other loss functions suitable for graph similarity evaluation, which can further improve the effect of sensitivity matrix optimization.

[0103] According to the set of normalized pixel gray value vectors corresponding to the sensitivity matrix obtained by finally stopping the loop, the defect image set is reconstructed.

[0104] S5. Construct a defect positioning model for detecting defects of carbon fiber material cable.

[0105] S5-1. Improve the U-Net model.

[0106] As shown in Figure 4 , replace the double convolution of the U-Net deep learning network with a residual module. The input of each residual module flows to two branches, one of which sequentially passes through a first convolution layer, a first batch normalization layer, an activation function, a second convolution layer and a second batch normalization layer, and the other branch passes through a first convolution layer. The outputs of the two branches are added and passed through an activation function to obtain the output of the residual module.

[0107] The residual module can accelerate the training of the defect positioning model and improve the stability of the defect positioning model to reduce the risk of overfitting.

[0108] S5-2. Label the defect image set obtained in step S3 to obtain a labeled defect image set.

[0109] The pixel value of the defect position is set to 255, and the pixel value of the non-defect position is set to 1, to obtain a set of defect-labeled images.

[0110] The defect-labeled image, the detection voltage, and the defect image correspond to each other in pairs.

[0111] S5-3. Reconstruct the defect image of step S3 using the initial voltage, the preprocessed detection voltage, and the iteratively updated sensitivity matrix.

[0112] The defect image corresponding to each column of data in the set G(S i ) of normalized pixel gray value vectors of the defect image reconstructed by the LBP algorithm in the last iteration of the updated sensitivity matrix is reconstructed according to the pixel value.

[0113] S5-4. The images in the set of defect-labeled images and the set of reconstructed defect images are input into the improved U-Net model for training to obtain a defect positioning model for detecting internal defects of carbon fiber material cables.

[0114] First, the defect positioning model is pre-trained and iteratively trained on the ImageNet large data set using transfer learning. Then, the defect-labeled images and the reconstructed defect images are input into the defect positioning model after the first training for training. The hyperparameters of the model training are set as follows: the iteration period is 100 epochs, the Batch Size is set to 16, the initial learning rate is set to 0.0001, the learning rate is adaptively adjusted according to the Batch Size size and the optimizer type, the loss function uses two loss functions, Cross Entropy Loss and Dice coefficient, and the learning parameters are optimized using the Adam optimizer.

[0115] S6. Set the detection probe, which is structured to have a number of annular coils pointing to the center of the circle on the inner ring surface of the circular ring carrier, which are divided into detection coils and excitation coils. The number of detection coils is the same as the number of excitation coils, and they are evenly distributed in double layers around the object field. The inner layer coils are set as detection coils, and the outer layer coils are set as excitation coils. The detection probe moves at a preset speed from one end of the measured carbon fiber cable to the other end, and completes the collection of detection voltage every 300 milliseconds to obtain a set of detection voltages of the measured carbon fiber cable.

[0116] As Figure 5As shown, the circular carrier in the application is a circuit board, and the front-end circuit such as filtering and amplification is designed on the circuit board to reduce the influence of noise and weak signals and control the excitation and collection of the coil. The excitation coil and the detection coil are both annular coils wound by copper wires, and the distribution positions of the excitation coil and the detection coil are opposite. The inner diameters of the detection coil and the excitation coil are both 5 mm, and the outer diameters are both 10 mm. The outer excitation coil is fixed on the circuit board, and the insulator material passes through the annular detection coil and the excitation coil to fix the center shaftes on the same straight line. The middle object field area surrounded by the coil arrangement can place the measured carbon fiber cable. The number of the detection coil and the excitation coil can be defined by oneself, and the number of the detection coil and the excitation coil depends on the imaging accuracy of the object field, the size of the object field and the coil arrangement. The number of the detection coil and the excitation coil in the application is 8.

[0117] The detection probe is sleeved on the measured carbon fiber cable, the detection probe is perpendicular to the axial direction of the measured carbon fiber cable, the excitation current of the excitation coil is set to 1 A, the excitation frequency is 1 MHz, the voltage of the detection coil is collected and recorded as the detection voltage. The detection probe moves at a preset speed from one end of the measured carbon fiber cable to the other end, and the collection of the detection voltage is completed every 300 milliseconds.

[0118] The preset speed in the application can be 10 mm / s.

[0119] S7. According to the set of detection voltages of the measured carbon fiber cable and the iteratively updated sensitivity matrix, the gray value of each pixel of the cross section of the measured carbon fiber cable is calculated, the cross section image of the measured carbon fiber cable is generated according to the gray value of each pixel, the cross section image of the measured carbon fiber cable is input into the trained defect positioning model, and the clear cross section image of the measured carbon fiber cable is output. When there is a defective cross section image, the defect position is determined through the collection time of the detection voltage corresponding to the defective cross section image.

[0120] The detection voltage of the carbon fiber cable without defects required for defect imaging can be obtained by applying excitation to the model, or can be obtained by applying electric excitation to the real object (i.e. the carbon fiber cable without defects). It can be the initial voltage of step S1, or the detection voltage of the carbon fiber cable without defects with the same specification as the measured carbon fiber cable.

[0121] According to the detection voltage of the measured carbon fiber cable, the detection voltage of the carbon fiber cable without defects and the iteratively updated sensitivity matrix, the gray value of each pixel of the cross section of the measured carbon fiber cable is calculated by using formula (7), and the detection voltage of each acquisition in the set of detection voltages of the measured carbon fiber cable is calculated once. According to the gray value of each pixel of the cross section of the measured carbon fiber cable, the cross section image of the measured carbon fiber cable is generated, and the cross section image of the measured carbon fiber cable is input into the trained defect positioning model, so that the clear cross section image can be obtained, and when there is a defect, the clear internal defect imaging can be obtained.

[0122] When the cross section image with defects appears, the position of the defect of the measured carbon fiber cable can be calculated according to the acquisition time of the detection voltage corresponding to the cross section image.

[0123] As shown in Figure 6 , the position and shape of the defect are clearer and the contour is clearer after the first imaging and reconstruction imaging. As can be seen from the comparison results of Figure 6 , the defects imaged by the traditional LBP algorithm are submerged in the artifacts, and only the approximate defect position can be represented, and it is difficult to reflect the accurate size and position information of the defect. However, the present application can realize the accurate visualization representation of the defects of the carbon fiber material cable.

Claims

1. A method of electromagnetic tomography for detecting defects in a carbon fiber cable, characterized by, Comprise the following steps: S1. Constructing an electromagnetic simulation model of the carbon fiber material cable according to the specifications of the measured carbon fiber cable, applying current excitation to the three-dimensional carbon fiber material cable electromagnetic simulation model to obtain an initial voltage, and establishing an initial sensitivity matrix of the carbon fiber material cable cross-section field region by field extraction method; S2. Randomly setting the shape and position of the defect on the simulation model, applying excitation to the simulation model with the defect set, and obtaining the detection voltage; according to the defect shape and position set on the simulation model, draw the cross-sectional image of the simulation model where the defect is located to obtain the defect image; Edge gradual filling is performed on the defect image, and Gaussian noise is added to the detection voltage; S3. Repeat step S2 at least 600 times to obtain a defect image set and a detection voltage set of the carbon fiber cable electromagnetic simulation model; S4. According to the initial voltage, the defect image set and the detection voltage set, the initial sensitivity matrix is iteratively updated; S5. Constructing a defect positioning model for detecting the internal defects of the carbon fiber material cable; S6. Setting a detection probe, the structure of which is that a plurality of annular coils pointing to the center are arranged on the inner ring surface of the circular ring carrier, which are divided into detection coils and excitation coils, the number of detection coils is the same as that of excitation coils, and they are uniformly distributed in double-layer annular around the object field, the inner layer coils are set as detection coils, and the outer layer coils are set as excitation coils; the detection probe moves from one end to the other end of the measured carbon fiber cable at a preset speed, and the collection of detection voltage is completed every 300 milliseconds to obtain a set of detection voltages of the measured carbon fiber cable; S7. According to the set of detection voltages of the measured carbon fiber cable and the iteratively updated sensitivity matrix, the gray value of each pixel of the cross section of the measured carbon fiber cable is calculated, and the cross-sectional image of the measured carbon fiber cable is generated according to the gray value of each pixel, the cross-sectional image of the measured carbon fiber cable is input into the trained defect positioning model, and the clear cross-sectional image of the measured carbon fiber cable is output, when there is a cross-sectional image with defects, the defect position is determined through the collection time of the detection voltage corresponding to the cross-sectional image with defects.

2. The electromagnetic tomography method for inspecting the carbon fiber cable according to Claim 1, wherein The specific way of iteratively updating the initial sensitivity matrix in step S4 is: S4-1. Iteratively updating the initial sensitivity matrix, and the formula of iterative updating is: Wherein, S0 is the initial sensitivity matrix, S1 is the matrix after the first iteration update, G(S0) is the normalized pixel gray value vector data of the defect image corresponding to the sensitivity matrix S0; The calculation formula of G(S0) is: wherein S0 T is the transpose matrix of S0, U * is a set of detection voltages obtained from defects of a randomly set simulation model, is an initial voltage; S4-2. Calculate the normalized pixel gray value vector data of the defect image corresponding to the iteratively updated sensitivity matrix, and the calculation formula is: S4-3. According to the normalized pixel gray value vector data of the defect image, the loss function is calculated: Loss||G * -G(S1)|| wherein G * is a set of normalized pixel gray value vector data of defect images obtained from defects of a randomly set simulation model; S4-4. Determine whether the loss function obtained in step S4-3 is less than 1×10 -3 Or iteratively update to a preset number of iterations. When the judgment result is yes, stop iterative update and output the updated sensitivity matrix S1. When the judgment result is no, execute steps S4-5-S4-8; S4-5. Iteratively updating the sensitivity matrix, and the formula of iterative updating is: wherein S i is the sensitivity matrix obtained after the i-th iteration update, S i-1 is the sensitivity matrix obtained after the i-1-th iteration update, a is the learning rate, and indicates the step size of iteration, G(S i-1 ) is the gradient of the sensitivity matrix S i-1 corresponding to the normalized pixel gray value vector data of the defect image; G(S i-1 ) = 1 - (S / Smax) where S i T is the transpose matrix of S i , U * is a set of detection voltages obtained from defects of a randomly set simulation model, is an initial voltage; S4-6. Calculate the normalized pixel gray value vector data of the defect image corresponding to the iteratively updated sensitivity matrix, and the calculation formula is: S4-7. According to the normalized pixel gray value vector data of the defect image, the loss function is calculated: Loss||G * -G(S i )|| wherein G * is a set of normalized pixel gray value vector data of defect images obtained from defects of a randomly set simulation model; S4-8. Determine whether the loss function outputted in step S4-7 is lower than 1 x 10 -3 or the iteration is updated to a preset iteration number, when the determination result is no, repeat steps S4-5-S4-7 until the determination result is yes, stop the iteration update, and output the sensitivity matrix of the last iteration update.

3. The electromagnetic tomography method for inspecting the carbon fiber cable according to Claim 1, wherein The specific way of constructing the defect positioning model is: S5-1. Improving the U-Net model; S5-2. Labeling the defect image set obtained in step S3 to obtain a labeled defect image set; S5-3. Recalculating the pixel values of the pixel points of each image in the defect image set obtained in step S3 according to the updated sensitivity matrix and the set of detection voltages, and reconstructing the defect image set; S5-4. Inputting the images in the labeled defect image set and the reconstructed defect image set as samples into the improved U-Net model for training to obtain a defect positioning model for detecting the internal defects of the carbon fiber material cable.

4. The electromagnetic tomography method for inspecting the carbon fiber cable according to Claim 3, wherein The specific way of improving the U-Net model is: replacing the double convolution of the U-Net deep learning network coding part with a residual module, and the structure of each residual module is: the input of the residual module is divided into two paths, one path passes through a first convolution layer, a first batch normalization layer, an activation function, a second convolution layer and a second batch normalization layer, and the other path passes through a 1*1 convolution, the outputs of the two paths are added and output after passing through an activation function.

5. The electromagnetic tomography method for inspecting the carbon fiber cable according to Claim 3, wherein The iteration period of the defect positioning model training is 100 epochs, the Batch Size is set to 16, the initial learning rate is set to 0.0001, and the loss function adopts the cross-entropy loss function and the Dice coefficient.

Images