Alloy steel wire rope defect intelligent detection and identification method based on deep learning

Abstract

The invention discloses an alloy steel wire rope defect intelligent detection and identification method based on deep learning, and relates to the technical field of steel wire rope detection, and the method comprises the steps: obtaining standard edge parameters and standard core parameters through a standard torsion analysis method, inputting the standard torsion analysis method into a deep neural network, and obtaining a recognizable network; detecting the defects of the alloy steel wire rope by using a recognizable network; the method is used for solving the problems that in an existing alloy steel wire rope defect intelligent detection and identification method, the steel wire rope needs to be processed in the detection process, so that defect detection cannot be visually and efficiently conducted on the steel wire rope, and hidden dangers affecting the performance of the steel wire rope exist after the steel wire rope is processed; and meanwhile, for the twisted steel wire rope, the detection difficulty is increased, and the detection efficiency is reduced.

Description

Technical Field

[0001] This invention relates to the field of wire rope inspection technology, specifically to a deep learning-based intelligent detection and identification method for defects in alloy wire ropes. Background Technology

[0002] Alloy steel wire rope is a flexible component made of alloy steel wire twisted together. It has high strength, corrosion resistance and fatigue resistance. The defect detection and identification of alloy steel wire rope mainly adopts non-destructive testing technology, including eddy current testing, magnetic memory testing and ultrasonic testing, focusing on detecting safety hazards such as broken wires, corrosion and cracks.

[0003] Existing methods for intelligent detection and identification of defects in alloy steel wire ropes typically involve heating and cooling the wire rope to obtain thermal imaging images. These images are then cropped and feature extracted to obtain sample data, which is used to train a detection model. While this improved method compensates for the lack of dynamic feature information in wire rope defect detection, the need for pre-processing makes direct and efficient defect detection difficult. Furthermore, processing the wire rope can potentially affect its performance. For twisted wire ropes, this increases detection difficulty and reduces efficiency. For example, patent application CN119666929A discloses a method for intelligent detection and identification of defects in alloy steel wire ropes. The present invention relates to a flaw detection method and system for alloy steel wire rope production. This method involves heating and cooling the alloy steel wire rope, acquiring thermal imaging images of the rope, constructing an initial detection model, and training it to obtain a new alloy steel wire rope detection model. Flaw detection is then performed on the alloy steel wire rope based on this model. However, other improvements to intelligent defect detection and identification methods for alloy steel wire ropes typically focus on real-time detection during production. These improvements fail to address the need for prior treatment of the wire rope before inspection, hindering intuitive and efficient defect detection. Furthermore, the treatment of the wire rope can potentially affect its performance, and for twisted wire ropes, it increases inspection difficulty and reduces efficiency. Therefore, it is necessary to improve existing intelligent defect detection and identification methods for alloy steel wire ropes. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in the prior art by proposing a deep learning-based intelligent detection and identification method for defects in alloy steel wire ropes. This method addresses the issue that existing intelligent detection and identification methods for defects in alloy steel wire ropes require pretreatment of the wire rope before detection, making it impossible to intuitively and efficiently detect defects. Furthermore, pretreatment of the wire rope may affect its performance, and for twisted wire ropes, it increases the difficulty of detection and reduces detection efficiency.

[0005] To achieve the above objectives, this application provides a deep learning-based intelligent detection and identification method for defects in alloy steel wire ropes, comprising the following steps: The number of strands in the alloy steel wire rope during torsion is obtained based on the torsion standard of the alloy steel wire rope and denoted as L; the alloy steel wire rope is processed using the standard torsion analysis method based on L, and the standard edge parameters and standard core parameters are obtained based on the processing results. The standard torsion analysis method is input into a deep neural network; the process of obtaining standard edge parameters and standard core parameters of the standard torsion analysis method is trained multiple times based on the deep neural network, and the standard edge parameters and standard core parameters are optimized based on the training results, and the optimized deep neural network is recorded as a recognizable network. Obtain the cross-section of the alloy steel wire rope to be inspected, and record it as the inspection section; use an identifiable network to process the inspection section, and record the obtained standard edge parameters and standard core parameters as the inspection parameters of the inspection section; detect defects in the alloy steel wire rope based on the inspection parameters, standard edge parameters, and standard core parameters.

[0006] Furthermore, the standard torsion analysis method includes: The steel wire rope used for torsion in the alloy steel wire rope is denoted as the sub-steel wire rope, and the diameter of the sub-steel wire rope is denoted as L0. The alloy steel wire rope is a steel wire rope obtained by torsion of L strands of sub-steel wire rope. A cross-sectional analysis was performed on the alloy steel wire rope obtained by twisting L strands of steel wire rope, and the standard edge parameters and standard core parameters of the alloy steel wire rope were obtained based on the results of the cross-sectional analysis.

[0007] Furthermore, the cross-sectional analysis process includes: The image corresponding to the cross-section of the alloy steel wire rope is obtained using a camera and recorded as the cross-section image; the cross-section image is grayscale processed and recorded as the gray cross-section image; the edge of the sub-steel wire rope in the gray cross-section image is identified using the contour acquisition method based on AI, and the cross-sectional contour and gap contour in the gray cross-section image are obtained based on the recognition results. The contour acquisition method includes: using AI to mark any point in the area where the sub-steel wire rope is located in the cross-sectional image and recording it as a sampling point; recording the gray value of the pixel in the gray cross-sectional image that coincides with the sampling point as the gray value of the gray rope.

[0008] Furthermore, contour acquisition methods also include: For any pixel α in the grayscale image with a grayscale value of grayline grayscale: when the grayscale values ​​of all pixels corresponding to the eight neighbors of pixel α are grayline grayscale values, pixel α is recorded as an inner pixel; when the grayscale value of any pixel among the eight neighbors of pixel α is not grayline grayscale value, pixel α is recorded as an outer pixel. Connect all adjacent outer pixels in the gray cross-section image, and record all closed contours formed by the connected pixels as edge contours; record the contour with the longest perimeter among all edge contours as the cross-sectional contour, and record all edge contours other than the cross-sectional contour as gap contours.

[0009] Furthermore, the cross-sectional analysis process also includes: For any pixel β in the cross-sectional contour: Pixels with gray values ​​of gray line gray value in the eight neighborhood of pixel β are recorded as same color pixels, and pixels with gray values ​​other than gray line gray value are recorded as different color pixels, and the area where the different color pixels are located is recorded as different color area. Let the pixels of the same color adjacent to the opposite-color region be denoted as outer corner pixel A and outer corner pixel B, respectively; let the line segment Z1 connect the center of outer corner pixel A and the center of pixel β, and let the line segment Z2 connect the center of outer corner pixel B and the center of pixel β; let the angle formed by line segment Z1 and line segment Z2 at pixel β be denoted as the pixel outer angle, where the arc corresponding to the pixel outer angle intersects the opposite-color region; Obtain the pixel exterior angles of all pixels in the cross-sectional contour, and denote the smallest and largest pixel exterior angles as the small extrusion angle and the large extrusion angle, respectively. Denote the small extrusion angle and the large extrusion angle as the standard edge parameters.

[0010] Furthermore, the cross-sectional analysis process also includes: For any gap profile: the line segment connecting the leftmost and rightmost points of the gap profile is called the horizontal gap line, and the line segment connecting the highest and lowest points of the gap profile is called the vertical gap line; the length of the shortest line segment of the horizontal and vertical gap lines divided by the length of the longest line segment is called the inclination ratio, where the inclination ratio is 1 when the lengths of the horizontal and vertical gap lines are equal. When the horizontal and vertical lines of the gap are perpendicular to each other, the inclination angle of the gap profile is recorded as 90°. When the horizontal and vertical lines of the gap are not perpendicular to each other, the inclination angle of the gap profile is recorded as the acute angle obtained by the intersection of the horizontal and vertical lines of the gap. Obtain the tilt ratio and tilt angle of all gap profiles, and denote the closed interval formed by the maximum and minimum values ​​of all tilt ratios as the gap ratio interval, and denote the closed interval formed by the maximum and minimum values ​​of all tilt angles as the gap angle interval; denote the gap ratio interval and gap angle interval as the standard core parameters.

[0011] Furthermore, the standard torsion analysis method is incorporated into a deep neural network; the process of obtaining standard edge parameters and standard core parameters using the standard torsion analysis method is trained multiple times based on the deep neural network; the standard edge parameters and standard core parameters are optimized based on the training results; and the optimized deep neural network is recorded as a recognizable network including: Obtain the deep neural network, set the input layer of the deep neural network to a cross-sectional image, and set the output layer of the deep neural network to standard edge parameters and standard core parameters respectively; input the standard torsional analysis method into the hidden layer of the deep neural network; Deep neural networks are used for torsion training until the standard edge parameters and standard core parameters of the alloy steel wire rope are completely identical in adjacent torsion training sessions. At this point, the deep neural network is recorded as a recognizable network, and the standard edge parameters and standard core parameters of all stacked rope groups obtained from the latest torsion training are recorded as optimized standard edge parameters and standard core parameters.

[0012] Furthermore, twisting exercises include: A cross-sectional image of the alloy steel wire rope without torsion defects is obtained based on cross-sectional analysis and input into the input layer of a deep neural network. Based on the standard torsion analysis method in the hidden layer of the neural network, the standard edge parameters and standard core parameters of the alloy steel wire rope corresponding to the output layer of the deep neural network are obtained.

[0013] Furthermore, the detection cross-section is processed using a recognizable network, and the obtained standard edge parameters and standard core parameters are recorded as the detection parameters of the detection cross-section, including: The detection cross section is entered into the input layer of the recognizable network, and the standard edge parameters and standard core parameters obtained from the output layer of the recognizable network are recorded as the detection parameters of the detection cross section. The latest optimized standard edge parameters and standard core parameters obtained from the recognizable network are denoted as the alignment edge parameters and alignment core parameters, respectively.

[0014] Furthermore, the detection of defects in alloy steel wire ropes based on testing parameters, standard edge parameters, and standard core parameters includes: When the small outward extrusion angle in the detection parameters of the detection section is not equal to the small outward extrusion angle in the comparison edge parameters, or when the large outward extrusion angle in the detection parameters of the detection section is not equal to the large outward extrusion angle in the comparison edge parameters, an early warning of loose edge defects of alloy steel wire rope is sent. When the gap ratio range in the detection parameters of the detection section is not completely within the gap ratio range of the core parameters, or when the gap angle range in the detection parameters of the detection section is not completely within the gap angle range of the core parameters, an early warning of internal extrusion defects in the alloy steel wire rope is sent. When no warning of loose edge defects in alloy steel wire rope is sent, the defect detection result of alloy steel wire rope is recorded as no defect exists.

[0015] The beneficial effects of this invention are as follows: First, this application obtains the number of strands of the alloy steel wire rope during torsion based on the torsion standard of the alloy steel wire rope, and denoted as L. Based on L, the alloy steel wire rope is processed using the standard torsion analysis method. Based on the processing results, standard edge parameters and standard core parameters are obtained. The advantage of this is that by using the standard torsion analysis method to obtain the standard edge parameters and standard core parameters, the characteristics of the outer contour and the internal gap of the alloy steel wire rope after being torsion by the sub-steel wire rope can be obtained. This facilitates the detection of whether there are defects in the alloy steel wire rope and the approximate location of the defects in subsequent analysis based on the standard edge parameters and standard core parameters, thereby ensuring the accuracy of defect detection. This application also incorporates the standard torsion analysis method into a deep neural network, optimizing the standard edge parameters and standard core parameters based on the deep neural network. Finally, the cross-section of the alloy steel wire rope to be inspected is obtained; the inspection cross-section is processed using a recognizable network; and defects in the alloy steel wire rope are detected based on the inspection parameters, standard edge parameters, and standard core parameters. The advantage of this approach is that by using a recognizable network to detect defects in the alloy steel wire rope, after obtaining the alloy steel wire rope to be inspected, the defect detection result can be obtained directly by acquiring the image of the cross-section and inputting the image into the recognizable network. This achieves the goal of ensuring the accuracy of defect detection through standard edge parameters and standard core parameters, while also providing intuitive and efficient defect detection of torsioned steel wire ropes without affecting the performance of the steel wire rope itself. This avoids the problems of increased detection difficulty and reduced detection efficiency caused by steel wire rope torsion. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of the method of the present invention; Figure 2 This is a schematic diagram of the stacking of the alloy steel wire rope of the present invention; Figure 3 This is a schematic diagram of the cross-sectional profile of the present invention; Figure 4This is a schematic diagram illustrating the acquisition of the gap profile according to the present invention; Figure 5 This is a schematic diagram illustrating the acquisition of the outer corner of a pixel according to the present invention; Figure 6 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 As shown, this application provides a deep learning-based intelligent detection and identification method for defects in alloy steel wire ropes, including the following steps: Step S1: Obtain the number of strands of the alloy steel wire rope during torsion based on the torsion standard of the alloy steel wire rope, and denot it as L; process the alloy steel wire rope using the standard torsion analysis method based on L, and obtain the standard edge parameters and standard core parameters based on the processing results. In the specific implementation process, for example, if the alloy steel wire rope is obtained by twisting 7 steel wires in one analysis, the value of L can be recorded as 7, and in subsequent analyses, the 7 steel wire rope can be recorded as the sub-steel wire rope. The standard torsion analysis method includes: step S101, where the steel wire rope used for torsion in the alloy steel wire rope is denoted as the sub-steel wire rope, and the diameter of the sub-steel wire rope is denoted as L0, wherein the alloy steel wire rope is a steel wire rope obtained by torsion of L strands of sub-steel wire rope; In the specific implementation process, for example, during a data analysis, the diameter of the sub-steel wire rope is 5mm and the number of sub-steel wire ropes is 7. That is, the alloy steel wire rope is formed by twisting one sub-steel wire rope in the middle and six sub-steel wire ropes around it. Therefore, the diameter of the alloy steel wire rope should be less than or equal to 15mm. After obtaining L0, the torsion state of the alloy steel wire rope can be preliminarily estimated based on the diameter of the alloy steel wire rope. Step S102: Perform cross-sectional analysis on the alloy steel wire rope obtained by twisting the L strands of steel wire rope, and obtain the standard edge parameters and standard core parameters of the alloy steel wire rope based on the results of the cross-sectional analysis.

[0019] The cross-section analysis process includes: step S1021, using a camera to acquire an image corresponding to the cross-section of the alloy steel wire rope and recording it as a cross-section image; performing grayscale processing on the cross-section image and recording the grayscale processed cross-section image as a gray cross-section image; using AI to identify the edge of the sub-steel wire rope in the gray cross-section image using the contour acquisition method, and obtaining the cross-sectional contour and gap contour in the gray cross-section image based on the identification results. In the specific implementation process, if there are elements such as lighting that affect the color of objects in the cross-sectional image during the first cross-sectional analysis, the position and orientation of these elements should be kept consistent with those in the first cross-sectional analysis during subsequent cross-sectional analyses to prevent affecting the accuracy of defect detection.

[0020] The contour acquisition method includes: Step V1: Use AI to mark any point in the area where the sub-steel wire rope is located in the cross-sectional image and record it as a sampling point; record the gray value of the pixel in the gray cross-sectional image that coincides with the sampling point as the gray value of the gray rope. Step V2: For any pixel α in the grayscale image with a grayscale value of grayline grayscale: when the grayscale values ​​of all pixels corresponding to the eight neighboring pixels of pixel α are grayline grayscale values, pixel α is recorded as an inner pixel; when the grayscale value of any pixel among the eight neighboring pixels of pixel α is not grayline grayscale value, pixel α is recorded as an outer pixel. In this embodiment, it is assumed that all pixels in the area where the alloy steel wire rope is located in the gray cross image have the same gray value. Therefore, by comparing the gray values ​​of all pixels in the eight neighborhoods of pixel α, the pixels at the edge of the alloy steel wire rope in the gray cross image can be obtained for subsequent data analysis. Step V3: Connect all adjacent outer pixels in the gray cross-section image, and record all closed contours formed by the connected pixels as edge contours; record the contour with the longest perimeter among all edge contours as the cross-sectional contour, and record all edge contours other than the cross-sectional contour as gap contours. In specific implementation processes, such as during a single data processing step, the stacking status of the alloy steel wire ropes obtained is as follows: Figure 2 The image shown is composed of all solid-line circles, and Figure 2 Each solid circle in the image corresponds to the cross-section of a sub-wire rope; analysis of the gray cross-section image reveals that... Figure 2 The corresponding cross-sectional profile is as follows Figure 3 As shown in the outline, Figure 2 A gap profile within the dashed circle, as shown Figure 4 The outline JL in the figure is shown; To account for most cases, this embodiment assumes that there are gaps between the stacked alloy steel wire ropes. However, the actual twisted alloy steel wire rope may be too compact or too loose due to the twisting requirements, which may cause the gaps between the sub-wire ropes to become larger, smaller or even disappear during twisting. If there are no gaps between the sub-wire ropes during the actual analysis, then only the cross-sectional profile can be analyzed in the subsequent analysis, so that defect detection can be performed only by standard edge parameters.

[0021] The cross-sectional analysis process also includes: step S1022, for any pixel β in the cross-sectional contour: the pixels with gray values ​​of gray line gray values ​​in the eight neighborhoods of pixel β are recorded as same color pixels, and the pixels with gray values ​​other than gray line gray values ​​are recorded as different color pixels, and the area where the different color pixels are located is recorded as different color area. Step S1023: The same-color pixels adjacent to the dissimilar color region are respectively denoted as outer corner pixel A and outer corner pixel B; the center of outer corner pixel A is connected to the center of pixel β and denoted as line segment Z1, and the center of outer corner pixel B is connected to the center of pixel β and denoted as line segment Z2; the angle formed by line segment Z1 and line segment Z2 at pixel β is denoted as the pixel outer angle, wherein the arc corresponding to the pixel outer angle intersects with the dissimilar color region; In specific implementation processes, such as during a data analysis, the obtained pixel β is as follows: Figure 5 As shown in the rectangle where β is located, the eight rectangles around β are the pixels in the eight-neighborhood of pixel β; in the eight-neighborhood of pixel β, TS1 to TS3 are pixels of the same color, and all pixels except TS1 to TS3 are pixels of different colors. Through analysis, we can find that TS1 and TS3 are the outer corner pixels A and B, respectively. Then the angle γ is the outer corner of the pixel where the arc intersects with the different color region. Step S1024: Obtain the pixel exterior angles of all pixels in the cross-sectional contour, and record the smallest and largest pixel exterior angles as the small extrusion angle and the large extrusion angle, respectively, and record the small extrusion angle and the large extrusion angle as the standard edge parameters.

[0022] The cross-sectional analysis process also includes: step S1025, for any gap profile: the line segment connecting the leftmost point and the rightmost point of the gap profile is recorded as the gap horizontal line, and the line segment connecting the highest point and the lowest point of the gap profile is recorded as the gap vertical line; the length of the shortest line segment of the gap horizontal line and the gap vertical line divided by the length of the longest line segment is recorded as the inclination ratio, wherein when the lengths of the gap horizontal line and the gap vertical line are equal, the inclination ratio is 1; In the specific implementation process, such as during a data analysis, the results obtained are... Figure 4The lengths of the horizontal and vertical gap lines of the middle contour JL are 5mm and 5mm respectively. Therefore, the inclination ratio is calculated to be 1. By obtaining the inclination ratio and the subsequent inclination angle, the internal gap features of different layers of alloy steel wire rope can be extracted to facilitate subsequent defect detection. Step S1026: When the horizontal gap line and the vertical gap line are perpendicular to each other, the inclination angle of the gap profile is recorded as 90°. When the horizontal gap line and the vertical gap line are not perpendicular to each other, the inclination angle of the gap profile is recorded as the acute angle obtained by the intersection of the horizontal gap line and the vertical gap line. Step S1027: Obtain the tilt ratio and tilt angle of all gap profiles, and record the closed interval formed by the maximum and minimum values ​​of all tilt ratios as the gap ratio interval, and the closed interval formed by the maximum and minimum values ​​of all tilt angles as the gap angle interval; record the gap ratio interval and gap angle interval as standard core parameters.

[0023] Step S2: Input the standard torsion analysis method into the deep neural network; train the process of obtaining standard edge parameters and standard core parameters of the standard torsion analysis method multiple times based on the deep neural network, optimize the standard edge parameters and standard core parameters based on the training results, and record the optimized deep neural network as a recognizable network. Step S2 includes: Step S201, obtaining a deep neural network, setting the input layer of the deep neural network to a cross-sectional image, and setting the output layer of the deep neural network to standard edge parameters and standard core parameters respectively; and inputting standard torsion analysis into the hidden layer of the deep neural network. Step S202: Use a deep neural network for torsion training until the standard edge parameters and standard core parameters of the alloy steel wire rope are completely identical in adjacent torsion training. At this point, the deep neural network is recorded as a recognizable network, and the standard edge parameters and standard core parameters of all stacked rope groups obtained from the latest torsion training are recorded as optimized standard edge parameters and standard core parameters.

[0024] Torsion training includes: acquiring alloy steel wire ropes without torsion defects, obtaining cross-sectional images of the alloy steel wire ropes based on cross-sectional analysis, and inputting them into the input layer of a deep neural network; and obtaining the standard edge parameters and standard core parameters of the alloy steel wire ropes corresponding to the output layer of the deep neural network based on the standard torsion analysis method in the hidden layer of the neural network. In the specific implementation process, the purpose of torsion training is to allow the deep neural network to fully learn the standard torsion analysis method, and at the same time optimize the existing standard edge parameters and standard core parameters so that the standard edge parameters and standard core parameters are more in line with the torsion condition of the alloy steel wire rope, thereby improving the accuracy of subsequent defect detection.

[0025] Step S3: Obtain the cross-section of the alloy steel wire rope to be inspected, and record it as the inspection section; use a recognizable network to process the inspection section, and record the obtained standard edge parameters and standard core parameters as the inspection parameters of the inspection section; detect defects in the alloy steel wire rope based on the inspection parameters, standard edge parameters, and standard core parameters. Step S3 includes: Step S301, recording the detection cross section into the input layer of the recognizable network, and recording the standard edge parameters and standard core parameters obtained from the output layer of the recognizable network as the detection parameters of the detection cross section; Step S302: The latest optimized standard edge parameters and standard core parameters obtained from the recognizable network are recorded as the alignment edge parameters and alignment core parameters, respectively.

[0026] Step S3 also includes: Step S303, when the small outward extrusion angle in the detection parameters of the detection section is not equal to the small outward extrusion angle in the comparison edge parameters, or when the large outward extrusion angle in the detection parameters of the detection section is not equal to the large outward extrusion angle in the comparison edge parameters, an early warning of loose edge defects of alloy steel wire rope is sent. In the actual implementation process, when the alloy steel wire rope is twisted, defects in the sub-wire ropes at the edge may cause the outward extrusion small angle or outward extrusion large angle in the detection parameters to be unequal to the corresponding parameters in the comparison edge parameters. Therefore, the above comparison can be used to detect defects in the sub-wire ropes at the edge after the alloy steel wire rope is twisted. Similarly, the following comparison can be used to detect internal defects of the alloy steel wire rope by comparing whether the characteristics corresponding to the gaps inside the sub-wire ropes are the same. Step S304: When the gap ratio range in the detection parameters of the detection section is not completely within the gap ratio range of the core parameters, or when the gap angle range in the detection parameters of the detection section is not completely within the gap angle range of the core parameters, an early warning of internal extrusion defects in the alloy steel wire rope is sent. Step S305: When no warning of loose edge defect of alloy steel wire rope is sent, the defect detection result of alloy steel wire rope is recorded as no defect exists.

[0027] Example 2, please refer to Figure 6 As shown, Figure 6The example illustrates the structure of an electronic device, which may include a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes these computer-readable instructions, it runs steps such as those in the deep learning-based intelligent detection and identification method for alloy steel wire rope defects to achieve the following functions: First, based on the torsion standard of the alloy steel wire rope, the number of strands in the alloy steel wire rope during torsion is obtained and denoted as L; based on L, the alloy steel wire rope is processed using the standard torsion analysis method, and standard edge parameters and standard core parameters are obtained based on the processing results; then, the standard torsion analysis method is input into a deep neural network; the process of obtaining standard edge parameters and standard core parameters using the standard torsion analysis method is trained multiple times based on the deep neural network, and the standard edge parameters and standard core parameters are optimized based on the training results, and the optimized deep neural network is denoted as the recognizable network; finally, the cross-section of the alloy steel wire rope to be detected is obtained and denoted as the detection section; the recognizable network is used to process the detection section, and the obtained standard edge parameters and standard core parameters are denoted as the detection parameters of the detection section; based on the detection parameters, standard edge parameters, and standard core parameters, defects in the alloy steel wire rope are detected.

[0028] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0029] Example 3: This application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by the computer, the computer can execute the intelligent detection and identification method for alloy steel wire rope defects based on deep learning provided by the above methods. The method includes: first, obtaining the number of strands of the alloy steel wire rope during torsion based on the torsion standard of the alloy steel wire rope, and denoting it as L; processing the alloy steel wire rope using a standard torsion analysis method based on L, and obtaining standard edge parameters and standard core parameters based on the processing results; then, inputting the standard torsion analysis method into a deep neural network; training the process of obtaining standard edge parameters and standard core parameters using the standard torsion analysis method multiple times based on the deep neural network, optimizing the standard edge parameters and standard core parameters based on the training results, and denoting the optimized deep neural network as a recognizable network; finally, obtaining the cross-section of the alloy steel wire rope to be detected, denoted as the detection section; processing the detection section using the recognizable network, and denoting the obtained standard edge parameters and standard core parameters as the detection parameters of the detection section; and detecting defects in the alloy steel wire rope based on the detection parameters, standard edge parameters, and standard core parameters.

[0030] Example 4: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the above-mentioned intelligent detection and identification method for alloy steel wire rope defects based on deep learning, to achieve the following functions: First, based on the torsion standard of the alloy steel wire rope, the number of strands in the alloy steel wire rope during torsion is obtained and denoted as L; based on L, the alloy steel wire rope is processed using a standard torsion analysis method, and standard edge parameters and standard core parameters are obtained based on the processing results; then, the standard torsion analysis method is input into a deep neural network; the process of obtaining standard edge parameters and standard core parameters using the standard torsion analysis method is trained multiple times based on the deep neural network, and the standard edge parameters and standard core parameters are optimized based on the training results, and the optimized deep neural network is denoted as a recognizable network; finally, the cross-section of the alloy steel wire rope to be detected is obtained and denoted as the detection section; the detection section is processed using the recognizable network, and the obtained standard edge parameters and standard core parameters are denoted as the detection parameters of the detection section; the defects of the alloy steel wire rope are detected based on the detection parameters, standard edge parameters, and standard core parameters.

[0031] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0032] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An alloy steel wire rope defect intelligent detection and identification method based on deep learning, characterized in that, The method comprises the following steps: Based on the torsion standard of the alloy steel wire rope, the number of strands of the alloy steel wire rope when twisted is obtained, and is recorded as L; the alloy steel wire rope is processed based on L using the standard torsion analysis method, and the standard edge parameter and the standard core parameter are obtained based on the processing result; The standard torsion analysis method is input into a deep neural network; the process of the standard torsion analysis method for obtaining the standard edge parameter and the standard core parameter is trained multiple times based on the deep neural network, the standard edge parameter and the standard core parameter are optimized based on the training result, and the optimized deep neural network is recorded as a recognizable network; The cross section of the alloy steel wire rope to be detected is obtained, which is recorded as a detection cross section; the detection cross section is processed using the recognizable network, and the obtained standard edge parameter and standard core parameter are recorded as the detection parameter of the detection cross section; Based on the detection parameter, the standard edge parameter and the standard core parameter, the defects of the alloy steel wire rope are detected. 2.The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 1, characterized in that, The standard torsion analysis method comprises: The steel wire rope used for twisting in the alloy steel wire rope is recorded as a sub-steel wire rope, and the diameter of the sub-steel wire rope is recorded as L0, wherein the alloy steel wire rope is a steel wire rope twisted by L strands of sub-steel wire ropes; The alloy steel wire rope twisted by L strands of sub-steel wire ropes is processed by cross section analysis, and the standard edge parameter and the standard core parameter of the alloy steel wire rope are obtained based on the result of the cross section analysis. 3.The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 2, characterized in that, The cross section analysis processing comprises: An image corresponding to the cross section of the alloy steel wire rope is obtained using a camera, and is recorded as a cross section image; the cross section image is processed by grayscale, and the cross section image processed by grayscale is recorded as a gray cross section image; the edges of the sub-steel wire rope in the gray cross section image are recognized based on AI using a contour acquisition method, and the cross section contour and the gap contour in the gray cross section image are obtained based on the recognition result; The contour acquisition method comprises: using AI to mark any point in the region of the sub-steel wire rope in the cross section image, and recording it as a sampling point; the gray value of the pixel point in the gray cross section image coinciding with the sampling point is recorded as a gray rope gray value.

4. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 3, characterized in that, The contour acquisition method further comprises: For any pixel point α in the gray cross section image with a gray value of the gray rope gray value: when the gray values of all pixel points corresponding to the eight neighborhoods of the pixel point α are all the gray rope gray value, the pixel point α is recorded as an internal pixel point; when there is any pixel point in the eight neighborhoods of the pixel point α whose gray value is not the gray rope gray value, the pixel point α is recorded as an external pixel point; All adjacent external pixel points in the gray cross section image are connected, and all closed contours formed by the connected pixel points are recorded as edge contours; the contour with the longest length in all edge contours is recorded as the cross section contour, and the edge contours other than the cross section contour are recorded as the gap contour.

5. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 4, characterized in that, The cross section analysis processing further comprises: For any pixel point β in the cross section contour: the pixel points with the gray value of the gray rope gray value in the eight neighborhoods of the pixel point β are recorded as same-color pixel points, the pixel points with the gray value not being the gray rope gray value are recorded as different-color pixel points, and the region where the different-color pixel points are located is recorded as a different-color region; The same-color pixel points adjacent to the different-color area are respectively denoted as an outer corner pixel point A and an outer corner pixel point B; a center of the outer corner pixel point A and a center of the pixel point β are connected, denoted as a line segment Z1, a center of the outer corner pixel point B and the center of the pixel point β are connected, denoted as a line segment Z2; an angle formed by the line segment Z1 and the line segment Z2 at the pixel point β is denoted as a pixel outer corner, wherein an arc corresponding to the pixel outer corner intersects with the different-color area; The pixel outer corners of all the pixel points in the cross-section profile are obtained, and the minimum and maximum pixel outer corners are respectively denoted as an outer extrusion small angle and an outer extrusion large angle, and the outer extrusion small angle and the outer extrusion large angle are denoted as standard edge parameters.

6. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 5, characterized in that, The cross-section analysis processing further includes: For any one gap profile: a line segment obtained by connecting the leftmost point and the rightmost point of the gap profile is denoted as a gap horizontal line, a line segment obtained by connecting the highest point and the lowest point of the gap profile is denoted as a gap vertical line; a value obtained by dividing the length of the shortest line segment among the gap horizontal line and the gap vertical line by the length of the longest line segment is denoted as a tilt ratio, wherein when the lengths of the gap horizontal line and the gap vertical line are equal, the tilt ratio is 1; When the gap horizontal line and the gap vertical line are perpendicular to each other, the tilt angle of the gap profile is denoted as 90°, and when the gap horizontal line and the gap vertical line are not perpendicular to each other, the tilt angle of the gap profile is denoted as an acute angle obtained by intersecting the gap horizontal line and the gap vertical line; The tilt ratios and the tilt angles of all the gap profiles are obtained, and a closed interval formed by the maximum value and the minimum value among all the tilt ratios is denoted as a gap ratio interval, and a closed interval formed by the maximum value and the minimum value among all the tilt angles is denoted as a gap angle interval; the gap ratio interval and the gap angle interval are denoted as standard core parameters.

7. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 6, characterized in that, The standard torsion analysis method is input into the deep neural network; the process of obtaining the standard edge parameters and the standard core parameters by the standard torsion analysis method based on the deep neural network is trained multiple times, the standard edge parameters and the standard core parameters are optimized based on the training results, and the optimized deep neural network is denoted as a recognizable network including: The deep neural network is obtained, the input layer of the deep neural network is set as a cross-section image, and the output layer of the deep neural network is set as the standard edge parameters and the standard core parameters respectively; the standard torsion analysis method is input into the hidden layer of the deep neural network; The deep neural network is used for torsion training until the standard edge parameters of the alloy steel wire rope obtained in adjacent torsion training are completely the same and the standard core parameters are completely the same, and the deep neural network at this time is denoted as a recognizable network, and the standard edge parameters and the standard core parameters of all the stacked rope groups obtained by the latest torsion training are denoted as optimized standard edge parameters and standard core parameters. 8.The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 7, characterized in that, The torsion training includes: An alloy steel wire rope without a torsion defect is obtained, a cross-section image of the alloy steel wire rope is obtained based on the cross-section analysis processing, and is input into the input layer of the deep neural network; the standard edge parameters and the standard core parameters corresponding to the alloy steel wire rope obtained from the output layer of the deep neural network are obtained based on the standard torsion analysis method in the hidden layer of the neural network.

9. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 8, characterized in that, The standard edge parameter and the standard core parameter obtained by processing the detection section using the identifiable network are recorded as detection parameters of the detection section, including: The detection section is input into the input layer of the identifiable network, and the standard edge parameter and the standard core parameter obtained by the output layer of the identifiable network are recorded as the detection parameters of the detection section; The latest optimized standard edge parameter and the standard core parameter obtained by the identifiable network are recorded as the comparison edge parameter and the comparison core parameter respectively.

10. The deep learning-based alloy steel wire rope defect intelligent detection and identification method according to claim 9, characterized in that, Based on the detection parameters, the standard edge parameter and the standard core parameter, the defects of the alloy steel wire rope are detected, including: When the outer extrusion small angle in the detection parameters of the detection section is not equal to the outer extrusion small angle of the comparison edge parameter, or the outer extrusion large angle in the detection parameters of the detection section is not equal to the outer extrusion large angle of the comparison edge parameter, an edge looseness defect warning of the alloy steel wire rope is sent; When the gap proportion interval in the detection parameters of the detection section is not completely in the gap proportion interval of the comparison core parameter, or the gap angle interval in the detection parameters of the detection section is not completely in the gap angle interval of the comparison core parameter, an internal extrusion defect warning of the alloy steel wire rope is sent; When the edge looseness defect warning of the alloy steel wire rope is not sent or the edge looseness defect warning of the alloy steel wire rope is sent, the defect detection result of the alloy steel wire rope is recorded as no defect.

Images