Steel wire rope nondestructive testing method based on weak magnetic flaw detection

Abstract

The invention discloses a steel wire rope nondestructive testing method based on weak magnetic flaw detection, which comprises the following steps: acquiring a magnetic field signal in a steel wire rope nondestructive testing device, and preprocessing; performing noise reduction processing on the preprocessed magnetic field signal; converting the image into a magnetic field grayscale image through a predetermined rule, and improving the definition of the magnetic field grayscale image in combination with a bicubic interpolation algorithm; performing key feature extraction by using a convolutional neural network and a feature pyramid network; analyzing the key features, and recognizing the damage condition of the steel wire rope in combination with an artificial neural network; based on the damage condition of the steel wire rope, a damage condition evaluation report is generated and graphically presented. According to the method, through an efficient signal processing technology and an advanced image analysis and machine learning algorithm, accurate recognition and evaluation of the damage of the steel wire rope can be achieved, the detection sensitivity and reliability are improved, and the intelligent level and the working efficiency of an automatic detection system are greatly improved.

Description

Technical Field

[0001] The invention relates to the field of nondestructive testing of steel wire ropes, and more particularly to a nondestructive testing method for steel wire ropes based on weak magnetic flaw detection. Background Art

[0002] Nondestructive testing of wire ropes based on weak magnetic flaw detection identifies potential damage, such as cracks and broken wires, by detecting changes in the magnetic field inside the wire rope. Compared with traditional strong magnetic detection technology, weak magnetic flaw detection has higher sensitivity and faster detection speed, and can achieve online monitoring, which is especially suitable for real-time detection in occasions such as mine hoists. However, although this technology has made certain progress, it still faces some challenges and limitations in practical applications, especially in terms of detection accuracy, degree of automation and intelligent analysis. Specifically, the main defects of the existing technology include:

[0003] The on-site working environment is complex, with various electromagnetic interference sources (such as motors, inverters, etc.), which will be mixed into the collected magnetic field signal, resulting in low quality of the original data. In addition, the magnetic field changes caused by tiny defects inside the wire rope are very weak, which makes it very difficult to separate useful signals from background noise.

[0004] When the existing technology uses linear filters (such as low-pass filters) for noise reduction, it may cause distortion to the signal, especially when the signal contains nonlinear components. For different application scenarios, the selection of the best filter parameters has great uncertainty and requires a lot of manual adjustments to optimize.

[0005] In addition, some important information may be lost in the process of converting one-dimensional signals into two-dimensional images, especially information about the shape and location of defects. There is a lack of unified standards to guide how to effectively convert magnetic field signals into grayscale images, and different conversion methods will result in different image quality.

[0006] For weak magnetic signals, it is very difficult to manually design effective features because it is necessary to take into account multiple types of damage and complex background conditions, and it has poor adaptability to newly emerging damage types.

[0007] In the absence of intelligent analysis tools, damage assessment often relies on the experience of experts, which limits the efficiency and reliability of detection to a certain extent. The lack of a complete set of standardized damage assessment processes may lead to large differences in the assessment results between different inspectors.

[0008] Therefore, how to design a nondestructive testing method for wire ropes based on weak magnetic flaw detection to achieve accurate identification and evaluation of wire rope damage and improve the sensitivity and reliability of detection is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0009] In view of this, the present invention provides a non-destructive detection method for wire ropes based on weak magnetic flaw detection. Through efficient signal processing technology and advanced image analysis and machine learning algorithms, it can achieve accurate identification and evaluation of wire rope damage, which is of great significance for ensuring the safe operation of important facilities such as lifting equipment and elevators.

[0010] In order to achieve the above object, the present invention adopts the following technical solution:

[0011] A nondestructive testing method for steel wire rope based on weak magnetic flaw detection comprises the following steps:

[0012] S1. Obtaining a magnetic field signal in a nondestructive testing device for a steel wire rope, and performing preprocessing to obtain a preprocessed magnetic field signal;

[0013] S2, performing noise reduction processing on the preprocessed magnetic field signal to obtain a noise-reduced magnetic field signal;

[0014] S3, converting the denoised magnetic field signal into a magnetic field grayscale image according to a predetermined rule, and improving the clarity of the magnetic field grayscale image by combining a bicubic interpolation algorithm;

[0015] S4, based on the magnetic field grayscale image with improved clarity, the key features are extracted using convolutional neural network and feature pyramid network;

[0016] S5. Analyze the key features and identify the damage status of the wire rope in combination with an artificial neural network;

[0017] S6. Based on the damage condition of the wire rope, a damage condition assessment report is generated and presented graphically.

[0018] Wherein, in said S1, the preprocessing includes:

[0019] S11, differentially amplifying the magnetic field signal in the steel wire rope nondestructive testing device through a differential amplifier; the differential amplifier includes a plurality of amplifier circuits with adjustable gains;

[0020] S12. Perform analog-to-digital conversion on the magnetic field signal after differential amplification using an analog-to-digital converter to convert it into a digital signal; wherein the sampling rate of the analog-to-digital converter is not less than twice the frequency of the magnetic field signal.

[0021] Furthermore, the S12 further includes performing a baseline adjustment on the digital signal:

[0022] S'(t)=S(t)-B+B0

[0023] Wherein, S(t) is the original digital signal, B is the average value of the digital signal, and B0 is the desired fixed level.

[0024] Furthermore, the S2 includes:

[0025] S21, using an empirical mode decomposition algorithm to decompose the preprocessed magnetic field signal into multiple intrinsic mode functions;

[0026] S22, removing high-frequency noise from each intrinsic mode function by using a wavelet soft threshold denoising algorithm; the wavelet soft threshold denoising algorithm includes wavelet basis function and threshold selection:

[0027]

[0028] Among them, C j,k is the wavelet coefficient of the jth layer, k is the time index, and λ is the threshold;

[0029] S23, reconstructing the intrinsic mode function after high-frequency noise is removed to obtain a magnetic field signal after noise reduction processing; wherein, Hilbert transform is used in the reconstruction process to restore the instantaneous frequency characteristics of the signal.

[0030] Furthermore, in S3, the predetermined rules include: constructing a signal-to-image mapping function, constructing a grayscale allocation function, and defining an image size and ratio.

[0031] Furthermore, in S3, the clarity of the magnetic field grayscale image is improved by combining a bicubic interpolation algorithm, including:

[0032] S31, determining the magnification factor W of the image;

[0033] S32, magnifying the low-resolution magnetic field grayscale image by W times to obtain a blank image H;

[0034] S33, based on each pixel position in the blank image H, calculating the corresponding position of each pixel position in the low-resolution magnetic field grayscale image L;

[0035] S34, using an interpolation kernel function to calculate weight values ​​of four pixels near each corresponding position;

[0036] S35 , based on the weight values ​​of four pixels near each corresponding position, calculate the grayscale value of the blank image H to obtain a magnetic field grayscale image with improved clarity.

[0037] Furthermore, the interpolation kernel function is:

[0038]

[0039] Where a is a constant, e is the distance of the target pixel relative to the nearest integer pixel position; the kernel function B(t) is non-zero in the intervals 0≤e<1 and 1≤e<2, and is zero in other intervals.

[0040] Furthermore, the S4 includes:

[0041] S41, performing standardization processing on the magnetic field grayscale image;

[0042] S42, inputting the standardized magnetic field grayscale image into the optimized deep convolutional neural network for feature extraction; wherein the deep convolutional neural network integrates an attention module;

[0043] S43. Combine the feature pyramid network to fuse the extracted features at different levels to obtain key features; the feature pyramid network adopts a top-down path and lateral connections.

[0044] Furthermore, in S5, the artificial neural network uses a pre-trained recurrent neural network as a feature extractor;

[0045] The pre-trained recurrent neural network is added with several fully connected layers and fine-tuned, and trained using a cross entropy loss function and an Adam optimizer.

[0046] Furthermore, the S5 also includes calculating the approximate location of the damage according to the sampling time of the magnetic field signal and the moving speed of the wire rope.

[0047] Furthermore, in S6, the damage condition assessment report includes damage type, severity and location information; and the graphical presentation includes damage image annotation and damage location heat map.

[0048] It can be seen from the above technical solution that compared with the prior art, the technical solution of the present invention has the following advantages:

[0049] Beneficial effects:

[0050] 1. By preprocessing the collected magnetic field signal, noise and interference can be effectively eliminated to ensure the effectiveness of subsequent processing steps. At the same time, the effective components in the signal are highlighted, making the noise-reduced signal closer to the actual damage situation and improving the detection accuracy.

[0051] 2. Convert the magnetic field signal into a grayscale image to make the detection result more intuitive. The image processing method facilitates the application of image processing technology to further improve the detection efficiency and accuracy.

[0052] 3. Bicubic interpolation can effectively increase the resolution of the image and enhance the image details. Compared with simple neighbor interpolation, bicubic interpolation can reduce the staircase effect and other artifacts that may occur during the magnification process and maintain the authenticity of the image.

[0053] 4. Convolutional neural networks can automatically learn and extract complex features from signals, and feature pyramid networks can capture feature information at different scales. Combined with graphical reports, they can intuitively display the location, type, and severity of damage, making it easier for technicians to quickly understand and make decisions. Quickly understand and make decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0055] Figure 1 A flow chart of a nondestructive testing method for steel wire ropes based on weak magnetic flaw detection provided by an embodiment of the present invention;

[0056] Figure 2 A diagram of a preprocessing process provided by an embodiment of the present invention;

[0057] Figure 3 A diagram of a process of performing noise reduction processing on the pre-processed magnetic field signal provided in an embodiment of the present invention;

[0058] Figure 4 A diagram showing a process of improving the clarity of a magnetic field grayscale image by combining a bicubic interpolation algorithm provided in an embodiment of the present invention;

[0059] Figure 5 A diagram of the key feature extraction process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] like Figure 1 As shown, this embodiment provides a nondestructive testing method for steel wire rope based on weak magnetic flaw detection, comprising the following steps:

[0062] S1. Obtaining a magnetic field signal in a nondestructive testing device for a steel wire rope, and performing preprocessing to obtain a preprocessed magnetic field signal;

[0063] S2, performing noise reduction processing on the preprocessed magnetic field signal to obtain a noise-reduced magnetic field signal;

[0064] S3, converting the denoised magnetic field signal into a magnetic field grayscale image according to a predetermined rule, and improving the clarity of the magnetic field grayscale image by combining a bicubic interpolation algorithm;

[0065] S4, based on the magnetic field grayscale image with improved clarity, the key features are extracted using convolutional neural network and feature pyramid network;

[0066] S5. Analyze the key features and identify the damage status of the wire rope in combination with an artificial neural network;

[0067] S6. Based on the damage condition of the wire rope, a damage condition assessment report is generated and presented graphically.

[0068] This method uses efficient signal processing technology and advanced image analysis and machine learning algorithms to accurately identify and evaluate wire rope damage, which not only improves the sensitivity and reliability of detection, but also greatly enhances the intelligence level and work efficiency of the automated detection system.

[0069] The above steps and related technical features are further described in detail below:

[0070] like Figure 2 As shown, in this embodiment S1, the preprocessing includes:

[0071] S11. The magnetic field signal in the steel wire rope nondestructive testing device is differentially amplified by a differential amplifier; the differential amplifier includes a plurality of amplifier circuits with adjustable gain; the common mode noise can be effectively suppressed, the signal-to-noise ratio of the signal can be improved, and the gain can be flexibly adjusted according to actual needs to optimize the signal quality.

[0072] S12. Perform analog-to-digital conversion on the magnetic field signal after differential amplification by using an analog-to-digital converter to convert it into a digital signal; wherein the sampling rate of the analog-to-digital converter is not less than twice the frequency of the magnetic field signal, and the Nyquist-Shannon sampling theorem is followed to ensure that all information in the signal can be fully captured. At the same time, a high-precision analog-to-digital converter is used to reduce quantization errors and improve the accuracy of the digital signal, thereby significantly improving the effect of subsequent processing steps and the accuracy of damage identification.

[0073] In weak magnetic flaw detection technology, DC offset caused by system noise or signal processing may occur, thus affecting subsequent data processing and analysis. Therefore, in this embodiment, the digital signal is baseline adjusted by adding a baseline raising circuit, and the average level of the signal is raised to a fixed value by adding a fixed DC bias voltage, thereby improving the signal quality.

[0074] Specifically:

[0075] S'(t)=S(t)-B+B0

[0076] Wherein, S(t) is the original digital signal, B is the average value of the digital signal, and B0 is the desired fixed level.

[0077] like Figure 3 As shown, in this embodiment S2, it includes:

[0078] S21, using an empirical mode decomposition algorithm to decompose the preprocessed magnetic field signal into multiple intrinsic mode functions;

[0079] S22, high-frequency noise is removed from each intrinsic mode function by a wavelet soft threshold denoising algorithm; the high-frequency noise can be accurately removed while retaining useful information in the signal, and the wavelet soft threshold denoising algorithm includes wavelet basis function and threshold selection:

[0080]

[0081] Among them, C j,k is the wavelet coefficient of the jth layer, k is the time index, and λ is the threshold;

[0082] S23, reconstruct the intrinsic mode function after high-frequency noise removal to obtain the magnetic field signal after noise reduction; in the reconstruction process, Hilbert transform is used to restore the instantaneous frequency characteristics of the signal. This significantly improves the purity of the signal and the effect of subsequent processing steps, ensuring the accuracy of damage identification.

[0083] In this embodiment S3, the predetermined rules include: constructing a signal-to-image mapping function, constructing a grayscale allocation function, and defining image size and ratio. Specifically, they include:

[0084] Construct a signal-to-image mapping function: define a mapping function f that accepts the denoised magnetic field signal

[0085] s(t) is taken as input and converted into a two-dimensional image I(x,y). The mapping can be based on the time series characteristics of the signal, such as mapping different time points of the signal to different rows of the image, and magnetic field signals of different intensities are mapped to different gray levels.

[0086] I(x,y)=f(s(t))

[0087] Here, x and y are image coordinates, and t is the position on the time axis.

[0088] Constructing a grayscale allocation function: Determine a threshold function g that maps the magnetic field signal strength S(t) to the grayscale G. Use linear mapping or nonlinear mapping to convert the magnetic field signal strength to grayscale.

[0089] G = g(s(t))

[0090] Grayscale is a value between 0 and 255, where 0 represents black and 255 represents white.

[0091] Define image size and ratio: Define the image size W×H and the mapping ratio factor r between the time series and image coordinates x and r y , ensuring that the image can accurately represent the signal information.

[0092] like Figure 4 As shown, the clarity of the magnetic field grayscale image is improved by combining the bicubic interpolation algorithm. When calculating the value of a pixel in the target image, the corresponding position of the pixel in the original image is first determined, and then the closest integer pixel position needs to be found; specifically, the following steps are included:

[0093] S31, determining the magnification factor W of the image;

[0094] S32, magnifying the low-resolution magnetic field grayscale image by W times to obtain a blank image H;

[0095] S33, based on each pixel position in the blank image H, calculating the corresponding position of each pixel position in the low-resolution magnetic field grayscale image L;

[0096] S34, using an interpolation kernel function to calculate weight values ​​of four pixels near each corresponding position;

[0097] S35 , based on the weight values ​​of four pixels near each corresponding position, calculate the grayscale value of the blank image H to obtain a magnetic field grayscale image with improved clarity.

[0098] The interpolation kernel function is:

[0099]

[0100] Where a is a constant, e is the distance of the target pixel relative to the nearest integer pixel position; the kernel function B(t) is non-zero in the intervals 0≤e<1 and 1≤e<2, and is zero in other intervals.

[0101] In this step, the interpolation kernel function can effectively calculate the weight value between the target pixel and its surrounding pixels, achieving a high-quality image magnification effect, thereby significantly improving the clarity of the magnetic field grayscale image and the accuracy of subsequent feature extraction.

[0102] like Figure 5 As shown, this embodiment S4 includes:

[0103] S41. Standardize the magnetic field grayscale image; the standardization is to reduce the differences between images and make the model converge more easily. The training efficiency can be improved by scaling the pixel values ​​to the same range (for example, between 0 and 1). Specifically including: Zero mean normalization: This can be achieved by subtracting the mean value of the image. Unit variance normalization: This can be achieved by dividing by the standard deviation. Minimum maximum normalization: Each pixel value can be mapped to the interval [0, 1].

[0104] S42. Input the standardized magnetic field grayscale image into the optimized deep convolutional neural network for feature extraction; wherein the deep convolutional neural network integrates an attention module; the deep convolutional neural network adopts the ResNet-50 architecture, including at least 50 convolutional layers, and uses residual connections during training to alleviate the gradient vanishing problem; the attention module is the Squeeze-and-Excitation module in the squeeze-excitation network, which can automatically learn and adjust the weights between channels to enhance the network's attention to key features.

[0105] S43, combining the feature pyramid network to fuse the extracted features at different levels to obtain key features; the feature pyramid network adopts a top-down path, which starts from a high-resolution feature map and gradually reduces the resolution while maintaining or increasing the number of channels. It also includes a lateral connection, where the low-resolution feature map is connected to the high-resolution feature map through an upsampling operation to fuse information at different levels.

[0106] In this embodiment S5, the artificial neural network uses a pre-trained recurrent neural network as a feature extractor;

[0107] The pre-trained recurrent neural network is added with several fully connected layers and fine-tuned. The fully connected layers are used to map the features extracted from the recurrent neural network into a low-dimensional space, which can be used to distinguish different damage categories.

[0108] Fine-tuning means further training the model based on pre-training using a task-specific dataset to adapt to specific task requirements. During fine-tuning, some of the pre-trained model layers are frozen and only the newly added layers are trained, or the entire model is completely unfrozen for end-to-end training.

[0109] The cross entropy loss function is used to measure the distance between the predicted probability distribution and the true label. The Adam optimizer combines the advantages of momentum and adaptive learning rate (Adagrad / RMSprop), and can automatically adjust the learning rate of each parameter during training.

[0110] The step S5 also includes calculating the approximate location of the damage according to the sampling time of the magnetic field signal and the moving speed of the wire rope.

[0111] When the artificial neural network identifies the damage feature, it records the timestamp of the corresponding magnetic field signal. The timestamp is associated with the moving speed of the wire rope to calculate the length position of the wire rope when the damage occurs. If more precise position information is required, for example: the speed of the hoist is 2m / s, then the wire rope moves 5.2×2=10.4 meters in 5.2 seconds. Assuming that the starting position of the wire rope is 0 meters when the hoist starts, the damage is located at the position of 10.4 meters.

[0112] For more precise positioning, one can consider using an encoder to record the absolute position of the wire rope, or adding a time synchronization mechanism to the detection system to ensure that the timestamp of the magnetic field signal matches the actual position of the wire rope.

[0113] In this embodiment S6, the damage status assessment report includes damage type, severity and location information; the graphical presentation includes damage image annotation and damage location heat map. Specifically:

[0114] Damage type: Based on the identified damage characteristics, the report should clearly indicate the specific type of damage, such as broken wire, wear, rust, etc. Severity: The severity of the damage is assessed based on parameters such as the area and depth of the damage, which can be divided into several levels such as slight, moderate, and severe. Location information: Record the specific location of the damage, including the location along the length of the wire rope and the circumferential location, so that maintenance personnel can accurately locate it.

[0115] Damage image annotation: Use different colors or marks to distinguish different types of damage. For each damage, provide a brief description, such as damage type, location, severity, etc. The annotation should be clearly visible and not affect the information display of the original image.

[0116] Damage Location Heat Map: Uses a color gradient to indicate the severity of damage, with darker colors indicating more severe damage. Provides a legend to explain the relationship between color and severity. Heat maps can cover the entire length of the rope or zoom in on certain key areas.

[0117] The damage assessment report and graphical presentation in this step can provide comprehensive and detailed damage information, helping relevant personnel to quickly and accurately understand the damage status and take appropriate measures.

[0118] In this implementation, the nondestructive testing method for wire ropes based on weak magnetic flaw detection combines multi-level signal processing with advanced image and deep learning technologies to achieve efficient and accurate detection and evaluation of wire rope damage. Differential amplification and analog-to-digital conversion are used to improve signal quality, and empirical mode decomposition and wavelet soft threshold denoising are combined to effectively remove noise interference; at the same time, bicubic interpolation is used to enhance image clarity, and optimized deep convolutional neural networks and feature pyramid networks are used to accurately extract damage features; finally, the damage status is identified and graphically presented through artificial neural networks, which significantly improves detection accuracy and efficiency.

[0119] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0120] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A nondestructive testing method for steel wire rope based on weak magnetic flaw detection, characterized in that: The following steps are involved: S1. Obtaining a magnetic field signal in a nondestructive testing device for a steel wire rope, and performing preprocessing to obtain a preprocessed magnetic field signal; S2, performing noise reduction processing on the preprocessed magnetic field signal to obtain a noise-reduced magnetic field signal; S3, converting the denoised magnetic field signal into a magnetic field grayscale image according to a predetermined rule, and improving the clarity of the magnetic field grayscale image by combining a bicubic interpolation algorithm; S4, based on the magnetic field grayscale image with improved clarity, the key features are extracted using convolutional neural network and feature pyramid network; S5. Analyze the key features and identify the damage status of the wire rope in combination with an artificial neural network; S6. Based on the damage condition of the wire rope, a damage condition assessment report is generated and presented graphically.

2. A nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1, characterized in that: In S1, the preprocessing includes: S11, differentially amplifying the magnetic field signal in the steel wire rope nondestructive testing device by a differential amplifier; the differential amplifier includes a plurality of amplifier circuits with adjustable gain; S12. Perform analog-to-digital conversion on the magnetic field signal after differential amplification using an analog-to-digital converter to convert it into a digital signal; wherein the sampling rate of the analog-to-digital converter is not less than twice the frequency of the magnetic field signal.

3. A nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 2, characterized in that: The S12 further includes performing a baseline adjustment on the digital signal: S'(t)=S(t)-B+B0 Wherein, S(t) is the original digital signal, B is the average value of the digital signal, and B0 is the desired fixed level.

4. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1 is characterized in that: The S2 comprises: S21, using an empirical mode decomposition algorithm to decompose the preprocessed magnetic field signal into multiple intrinsic mode functions; S22, removing high-frequency noise from each intrinsic mode function by using a wavelet soft threshold denoising algorithm; the wavelet soft threshold denoising algorithm includes wavelet basis function and threshold selection: Among them, C j,k is the wavelet coefficient of the jth layer, k is the time index, and λ is the threshold; S23, reconstructing the intrinsic mode function after high-frequency noise is removed to obtain a magnetic field signal after noise reduction processing; wherein, Hilbert transform is used in the reconstruction process to restore the instantaneous frequency characteristics of the signal.

5. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1 is characterized in that: In S3, the predetermined rules include: constructing a signal-to-image mapping function, constructing a grayscale allocation function, and defining an image size and ratio.

6. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1 is characterized in that: In S3, the clarity of the magnetic field grayscale image is improved by combining a bicubic interpolation algorithm, including: S31, determining the magnification factor W of the image; S32, magnifying the low-resolution magnetic field grayscale image by W times to obtain a blank image H; S33, based on each pixel position in the blank image H, calculating the corresponding position of each pixel position in the low-resolution magnetic field grayscale image L; S34, using an interpolation kernel function to calculate weight values ​​of four pixels near each corresponding position; S35 , based on the weight values ​​of four pixels near each corresponding position, calculate the grayscale value of the blank image H to obtain a magnetic field grayscale image with improved clarity.

7. A nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 6, characterized in that: The interpolation kernel function is: Where a is a constant, e is the distance of the target pixel relative to the nearest integer pixel position; the kernel function B(t) is non-zero in the intervals 0≤e<1 and 1≤e<2, and is zero in other intervals.

8. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1, characterized in that: The S4 comprises: S41, performing standardization processing on the magnetic field grayscale image; S42, inputting the standardized magnetic field grayscale image into the optimized deep convolutional neural network for feature extraction; wherein the deep convolutional neural network integrates an attention module; S43. Combine the feature pyramid network to fuse the extracted features at different levels to obtain key features; the feature pyramid network adopts a top-down path and lateral connections.

9. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1, characterized in that: In S5, the artificial neural network uses a pre-trained recurrent neural network as a feature extractor; The pre-trained recurrent neural network is added with several fully connected layers and fine-tuned, and trained using a cross entropy loss function and an Adam optimizer.

10. The nondestructive testing method for steel wire rope based on weak magnetic flaw detection according to claim 1, characterized in that: In S6, the damage condition assessment report includes damage type, severity and location information; the graphical presentation includes damage image annotation and damage location heat map.

Images