A cable surface production defect detection and repair system and method

Abstract

The application relates to the technical field of cable visual identification, and discloses a cable surface production defect detection and repair system and method, which detects defects on the surface of a cable through an industrial camera detection device, uses an injection molding device to injection mold the detected defect positions, and then uses a trimming mechanism to trim and process the generated convex and concave block defects, and finally polishes the defects. The method can effectively solve the problems of long time consumption, poor effect and low accuracy of artificial detection of scratch defects in the automatic flow operation of cable production on the market, the visual detection device in the method adopts deep learning to analyze defects, first outputs a defect target position through a detection module, combines an injection molding module, a trimming module and a polishing module to repair defects, and effectively solves the problems of low defect detection efficiency on the surface of the cable, complicated repair, high error rate of artificial detection and repair and high labor cost.

Description

Technical Field

[0001] This invention relates to the field of cable visual recognition technology, specifically to a cable surface manufacturing defect detection and repair system and method. Background Technology

[0002] Cables play a vital role in industrial applications such as power transmission, signal transmission, and electromagnetic energy conversion. The surface quality of cables is a crucial factor in determining whether their various performance indicators meet standards. During cable production, improper manufacturing processes, inadequate mold configurations, and poor extrusion process control can lead to various defects, such as bulges, spots, bumps, bubbles, pores, wrinkles, and uneven outer diameter. Most cable surface defect inspection instruments perform these tests after cable production but before the cable sheath is shaped and wound.

[0003] When inspecting cables, different types of inspection equipment have emerged on the market. For example, some equipment can detect lumps, wrinkles, and dents that occur during cable production. When a defect detector detects a relevant type of target, manual repair is then carried out at the designated defect location. Although these devices can detect defects, their detection functions are limited. When encountering some defects, such as lumps and bumps, manual repair is required, which is time-consuming and the repair quality is difficult to guarantee. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a system and method for detecting and repairing manufacturing defects on the surface of cables, which has the advantages of real-time detection of cable surfaces and automatic defect repair, thus solving the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cable surface manufacturing defect detection and repair system, comprising a cable feeding module, a defect detection module, an injection molding module, a first trimming module, a second trimming module, a polishing module, and a take-up module. The cable feeding module is used to feed the manufactured cable to the defect detection module. The defect detection module is used for real-time image acquisition and detection of concave and convex defects in the acquired images. The injection molding module is used to inject mold the cable when the defect detection module detects concave defects on the cable surface. The first trimming module is used to trim convex defects on the cylindrical surface of the cable. The second trimming module is used to trim convex defects on the cylindrical surface of the cable. The polishing module is used to polish the surface of the injected or trimmed cable. The take-up module is used to take up the detected and repaired cable.

[0006] As a preferred embodiment of the present invention, the cable feeding module includes a cable frame, a support frame, a first motor, a motor connecting wire, and a motor switch. The support frame is mounted on the cable frame, and the first motor and the motor switch are connected by the motor connecting wire and control the cable frame to feed the cable.

[0007] As a preferred embodiment of the present invention, the defect detection module includes a detection platform, a camera bracket, a guide ring, a power interface, a first support plate, a camera mounting plate, an industrial camera, a light source fixing block, a light source, and a guide ring upright plate. The camera bracket and the guide ring upright plate are mounted on the detection platform. The camera bracket is provided with a first support plate along the cable travel direction. The first support plate is hexagonal, and a camera mounting plate is installed every 120 degrees. Each camera mounting plate is provided with an industrial camera, a light source fixing block, and a light source facing the cable travel direction. The light source is located on the light source fixing block and has a power interface. The cable passes through the guide ring, passes through the first support plate, and then passes through the guide ring upright plate.

[0008] As a preferred technical solution of the present invention, the injection molding module includes an upper injection molding assembly and a lower clamping assembly. The upper injection molding assembly includes a needle, a nozzle sleeve fixing block, a nozzle sleeve, a glue tube knob, a glue tube sleeve, a glue tube, a buffer assembly knob, a glue tube fixing block, and a buffer assembly. The glue tube stores molten plastic and is injected into the corresponding position of the defect on the cable surface through the needle. The needle is fixed by installing a nozzle sleeve in the nozzle sleeve fixing block. The glue tube is fixed by installing a glue tube sleeve in the glue tube fixing block. At the same time, a glue tube knob is installed on the side to control the tightness of the glue tube. A buffer assembly knob is installed on the top of the buffer assembly. The lower clamping assembly includes a second support plate, a fixing block, a guide block, a finger cylinder, and a base. The fixing block is installed at the bottom of the second support plate. The fixing block is controlled by adjusting the finger cylinder to clamp and tighten the guide block. The finger cylinder is installed on the base.

[0009] As a preferred technical solution of the present invention, the first trimming module includes a first bracket, a first cylinder, a first tool fixing block, a first arc tool, a first lead wire fixing block, and a first support block. The first support block is installed at the end of the first bracket. The first support block is used to support the first cylinder. The output end of the first cylinder is connected to the first tool fixing block. One end of the first tool fixing block is connected to the first arc tool, which trims the cylindrical surface of the cable supported in the first lead wire fixing block.

[0010] As a preferred embodiment of the present invention, the second trimming module includes a second arc-shaped cutter, a second lead wire fixing block, a second support block, a second bracket, a second cylinder, and a second cutter fixing block. The second support block is installed at the end of the second bracket. The second arc-shaped cutter is connected to the second cylinder through the second cutter fixing block to trim the cylindrical surface of the cable supported in the second lead wire fixing block.

[0011] As a preferred technical solution of the present invention, the grinding and polishing module includes an encoding and measuring wheel assembly and a grinding and polishing assembly. The encoding and measuring wheel assembly includes an encoder, a third cylinder, a truss, a guide wheel, a tension spring, and a measuring wheel. The third cylinder is installed on one side of the bottom of the truss, and the output end of the third cylinder is connected to the guide wheel. The measuring wheel is installed on the side of the truss corresponding to the guide wheel. The cable passes between the measuring wheel and the guide wheel and drives the measuring wheel to rotate, and transmits the rotation data to the encoder. The tension spring connects the guide wheel and the measuring wheel. The grinding and polishing assembly includes a first ceramic eye, a support block, a first synchronous wheel, a first baffle, a second synchronous wheel, a third synchronous wheel, a polishing wheel, a grinding wheel, a cylinder liner, and a fourth cylinder. The ceramic eye is used to guide the cable and is located on one side of the support block. The ceramic eye is positioned in front of the grinding wheel and the polishing wheel. The support block has two grinding wheels and two polishing wheels installed on the same side, arranged longitudinally and correspondingly as upper and lower wheels. The fourth cylinder is assembled with the cylinder liner and connected to the support block. The first synchronous wheel is connected to the second synchronous wheel via a belt for coordinated operation. The second synchronous wheel rotates via an externally added motor, driving the lower wheel of the grinding wheel and polishing wheel to rotate. The third synchronous wheel is installed on the back of the support block and rotates via the movement of the fourth cylinder, causing the upper wheel of the grinding wheel and polishing wheel connected to the third synchronous wheel to rotate. The first baffle is connected to the back of the support block.

[0012] As a preferred embodiment of the present invention, the take-up module includes an aluminum bracket, a support, a second motor, a motor mounting plate, a reel, and a second baffle. The aluminum bracket has four supports installed at its bottom. The second motor is mounted on the aluminum bracket via the motor mounting plate. The reel is mounted on the aluminum bracket. The second baffle is mounted on the center edge of the cable.

[0013] As a preferred embodiment of the present invention, the defect detection module performs defect detection after acquiring an image of the cable surface, and the specific steps are as follows: S1: Accepts RGB images from industrial cameras; S2: Defects are identified using the onboard CableDefectNet, and pixel-level anomaly heatmaps and image-level anomaly assessments are generated, with the defect type output.

[0014] This invention also provides a method for detecting and repairing manufacturing defects on the surface of cables, comprising the following steps: Step 1: Turn on the wire feeding module, and use the defect detection module to acquire images in real time and detect concave and convex defects in the acquired images; Step 2: When the defect detection module detects a dent defect on the cable surface, the cable is injection molded using the injection molding module; Step 3: Repair the protrusion defects on the upper and lower cylindrical surfaces of the cable using the first and second trimming modules; Step 4: Start the grinding and polishing module to polish the surface of the injection-molded or repaired cable, and use the cable take-up module to take up the inspected and repaired cable.

[0015] Compared with the prior art, the present invention provides a system and method for detecting and repairing manufacturing defects on the surface of cables, which has the following beneficial effects: 1. This invention enables the testing of a large number of cables through a cable feeding module. The cable rack is simple to set up and easy to disassemble, ensuring efficient cable testing. A defect detection module detects surface bumps and depressions on the cable. This module incorporates three industrial cameras and a light source to achieve global coverage testing of the cylindrical surface of the cable. This avoids the shortcomings of using a single camera for incomplete surface testing and low accuracy in rotational testing. Furthermore, the use of automatic camera testing reduces the error rate and labor intensity of manual cable inspection by workers.

[0016] 2. This invention utilizes a trimming module to trim detected protrusion defects on the cable surface and remove excess plastic after the injection molding module has completed repairs. By setting up upper and lower trimming devices, it effectively avoids damage to the lower surface of the cable when trimming upper protrusion defects with a full-circular-arc cutter. The injection molding module automatically injects molten plastic to repair concave defects on the cable surface, reducing the difficulties, time-consuming nature, and poor results of manual repairs. The grinding and polishing section uses a motor and cylinder to drive the grinding and polishing wheels to perform post-processing of the cable surface defect repair locations. The lower wheel of the grinding and polishing wheel is also driven by a cylinder, enabling the grinding and polishing wheel to move in the vertical plane, effectively controlling the operating space for grinding and polishing and ensuring the integrity of the process. This module also includes an encoder and meter counting wheel combination device, which, while counting the cable length, uses a defect detection module to accurately mark the defect location and feeds it back to other module controllers to further improve the defect processing of other modules.

[0017] 3. This invention effectively solves the problems of low efficiency and low accuracy of manual visual inspection in the current cable surface bump defect detection operation, high difficulty and poor effect of defect repair operation, and the inapplicability of traditional detection methods to small target sizes. It uses deep learning methods to analyze and automatically identify target defects, and combines the defect injection molding module to repair the specified defects, which effectively reduces the labor intensity of workers, while improving the integrity of the repair and ensuring the product quality of the cable. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the defect detection process of the present invention; Figure 2 This is a schematic diagram of the model framework of the CabelDefectNet algorithm of this invention; Figure 3 This is a loss transformation diagram during model training after the multi-scale feature fusion strategy is embedded in this invention; Figure 4 The image shows the real-time detection effect using the algorithm of this invention. Figure 5 For system assembly drawings; Figure 6 This is a structural diagram of the wire delivery module; Figure 7 This is a structural diagram of the upper injection molding device in the defect detection module; Figure 8 This is a structural diagram of the upper injection molding device in the defect detection module; Figure 9 This is a structural diagram of the lower-level clamping device in the defect detection module; Figure 10 The structural diagram of module 4 is modified; Figure 11 A diagram of the device for correcting module 5; Figure 12 A structural diagram of the encoder meter wheel device in the grinding and polishing module; Figure 13 This is a structural diagram of the grinding and polishing device; Figure 14 This is a structural diagram of the take-up module device; The components include: 1. Wire feeding module; 2. Defect detection module; 3. Injection molding module; 4. First trimming module; 5. Second trimming module; 6. Grinding and polishing module; 7. Wire take-up module; 8. Cable rack; 9. Support frame; 10. First motor; 11. Motor connecting wire; 12. Motor switch; 13. Inspection table; 14. Camera bracket; 15. Guide ring; 16. Power interface; 17. First support plate; 18. Camera mounting plate; 19. Industrial camera; 20. Light source fixing block; 21. Light source; 22. Guide ring upright plate; 23. Needle; 24. Nozzle sleeve fixing block; 25. Nozzle sleeve; 26. Glue tube knob; 27. Glue tube sleeve; 28. Glue tube; 29. ​​Buffer assembly knob; 30. Glue tube fixing block; 31. Buffer assembly; 32. Second support plate; 33. Fixing block; 34. Guide block; 35. Finger cylinder; 36. 37. Base; 38. First bracket; 39. First cylinder; 40. First cutter fixing block; 41. First arc cutter; 42. First lead wire fixing block; 43. Second arc cutter; 44. Second lead wire fixing block; 45. Second support block; 46. Second bracket; 47. Second cylinder; 48. Second cutter fixing block; 49. Encoder; 50. Third cylinder; 51. Truss; 52. Guide wheel; 53. Tension spring; 54. Meter counter wheel; 55. Ceramic eye; 56. Support block; 57. First synchronous pulley; 58. First baffle; 59. Second synchronous pulley; 60. Third synchronous pulley; 61. Polishing wheel; 62. Grinding wheel; 63. Cylinder liner; 64. Fourth cylinder; 65. Aluminum bracket; 66. Support; 67. Second motor; 68. Motor mounting plate; 69. Wire wheel; 70. Second baffle. Detailed Implementation

[0019] 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.

[0020] Please see Figures 1-14A cable surface manufacturing defect detection and repair system includes a cable feeding module 1, a defect detection module 2, an injection molding module 3, a first trimming module 4, a second trimming module 5, a polishing module 6, and a take-up module 7. The cable feeding module 1 is used to feed the manufactured cable to the defect detection module 2. The defect detection module 2 is used for real-time image acquisition and detection of concave and convex defects in the acquired images. The injection molding module 3 is used to inject mold the cable when the defect detection module 2 detects concave defects on the cable surface. The first trimming module 4 is used to trim convex defects on the cylindrical surface of the cable. The second trimming module 5 is used to trim convex defects on the cylindrical surface of the cable. The polishing module 6 is used to polish the surface of the injected or trimmed cable. The take-up module 7 is used to take up the detected and repaired cable.

[0021] Furthermore, the cable delivery module 1 includes a cable rack 8, a support frame 9, a first motor 10, a motor connection line 11, and a motor switch 12. The support frame 9 is installed on the cable rack 8. The first motor 10 and the motor switch 12 are connected together through the motor connection line 11 and control the cable rack 8 to deliver the cable.

[0022] Furthermore, the defect detection module 2 includes a detection platform 13, a camera bracket 14, a guide ring 15, a power interface 16, a first support plate 17, a camera mounting plate 18, an industrial camera 19, a light source fixing block 20, a light source 21, and a guide ring upright plate 22. The camera bracket 14 and the guide ring upright plate 22 are mounted on the detection platform 13. The first support plate 17 is set along the cable travel direction on the camera bracket 14. The cable passes through the guide ring 15, passes through the first support plate 17, and then passes through the guide ring upright plate 22. This is to ensure that the cable maintains an accurate position when passing through the defect detection module and to reduce the friction between the cable and other surfaces caused by the vibration of the cable feeding module. The first support plate 17 is hexagonal, and a camera mounting plate 18 is installed every 120 degrees. Each camera mounting plate 18 is equipped with an industrial camera 19, a light source fixing block 20, and a light source 21 facing the cable travel direction. The light source 21 is mounted on the light source fixing block 20 and has a power interface 16, which facilitates the mounting of the industrial camera 19 and the light source 21 to detect the entire cylindrical surface of the cable, avoiding incomplete detection of the cable with a single camera. The industrial camera 19 is mounted on one side of the camera mounting plate 18, and three mounting slots are provided on the first support plate 17 to form a slider mechanism with the camera mounting plate 18. This mechanism is used to adjust the distance between the camera and the cable to ensure the optimal position for camera detection. The slider device can also absorb vibration and reduce the impact of system vibration. The light source 21 is mounted on the other side through the light source fixing block 20, and the light source 21 has a power interface 16 to control the light. This part provides illumination for detecting surface defects of the cable, reduces shadows, and suppresses reflections.

[0023] Furthermore, the injection molding module 3 includes an upper injection molding assembly and a lower clamping assembly. The upper injection molding assembly includes a needle 23, a nozzle sleeve fixing block 24, a nozzle sleeve 25, a glue tube knob 26, a glue tube sleeve 27, a glue tube 28, a buffer assembly knob 29, a glue tube fixing block 30, and a buffer assembly 31. The glue tube 28 stores molten plastic and is injected into the corresponding position of the defect on the cable surface through the needle 23. The needle 23 is fixed by the nozzle sleeve 25 installed in the nozzle sleeve fixing block 24. The glue tube 28 is fixed by the glue tube sleeve 27 installed in the glue tube fixing block 30. At the same time, the glue tube knob 26 is installed on the side to control the tightness of the glue tube 28. The buffer assembly 31 is equipped with a buffer assembly knob 29 on the top. The lower clamping assembly includes a second support plate 32, a fixing block 33, a guide block 34, a finger cylinder 35, and a base 36. The fixing block 33 is installed at the bottom of the second support plate 32. The fixing block 33 is controlled by adjusting the finger cylinder 35 to clamp and tighten the guide block 34. The guide block 34 is respectively set on the fixing block 33. The finger cylinder 35 is installed on the base 36. When the injection molding module 3 receives the defect location signal from the defect detection module, it starts the finger cylinder 35 of the lower clamping assembly and controls the guide block 34 to the cable defect location. At this time, the conveying system is stopped, the upper injection molding device is started, and the dripping amount of molten plastic is controlled to the defect location by the glue cylinder knob 26. After the defect location is repaired and cooled, the system switch is turned on to continue conveying the cable.

[0024] Furthermore, the first trimming module 4 includes a first bracket 37, a first cylinder 38, a first tool fixing block 39, a first arc tool 40, a first lead wire fixing block 41, and a first support block 42. The first support block 42 is installed at the end of the first bracket 37. The first support block 42 is used to support the first cylinder 38. The output end of the first cylinder 38 is connected to the first tool fixing block 39. One end of the first tool fixing block 39 is connected to the first arc tool 40, and trims the cylindrical surface of the cable supported in the first lead wire fixing block 41.

[0025] Furthermore, the second trimming module includes a second arc cutter 43, a second lead wire fixing block 44, a second support block 45, a second bracket 46, a second cylinder 47, and a second cutter fixing block 48. The second support block 45 is installed at the end of the second bracket 46. The second arc cutter 43 is connected to the second cylinder 47 through the second cutter fixing block 48 to trim the cylindrical surface of the cable supported in the second lead wire fixing block 44. These two trimming modules work together to trim defects. When the first trimming module 4 receives a signal of a bump defect from the defect detection module 2, it simultaneously determines the location of the defect by using the camera position. This includes upper and lower semicircular bump defects. When an upper semicircular bump defect is detected, the first cylinder 38 of the first trimming module 4 is activated, controlling the first arc cutter 40 to move to the defect position and perform a reciprocating linear motion to trim the defect. When the detected defect is a lower semicircular bump defect, the second cylinder 47 of the second trimming module 5 is activated, controlling the second arc cutter 43 to move to the defect position. The trimming method is the same as that of the first trimming module 4, and the degree of trimming of the bump is controlled by controlling the start and stop time of the cylinder.

[0026] Furthermore, the grinding and polishing module 6 includes an encoding and measuring wheel assembly and a grinding and polishing assembly. The encoding and measuring wheel assembly includes an encoder 49, a third cylinder 50, a truss 51, a guide wheel 52, a tension spring 53, and a measuring wheel 54. The third cylinder 50 is mounted on one side of the bottom of the truss 51, and the output end of the third cylinder 50 is connected to the guide wheel 52. The measuring wheel 54 is mounted on the side of the truss 51 corresponding to the guide wheel 52. The cable passes between the measuring wheel 54 and the guide wheel 52, driving the measuring wheel 54 to rotate and transmitting the rotation data to the encoder 49. The tension spring 53 connects the guide wheel 52 and the measuring wheel 54. The measuring wheel 54 is mounted on the truss 51 and arranged opposite to the guide wheel 52. The cable passes between the two wheels. When the cable moves, the measuring wheel 54 rotates accordingly, and the rotation signal is collected by the encoder 49 to realize real-time monitoring of the cable running length. The guide wheel 52 is rotatably mounted on the carriage and can move up and down vertically under the drive of the cylinder 50.

[0027] The tension spring 53 is used to maintain the continuous pressing contact between the guide wheel 52 and the cable. One end of the spring is connected to the guide wheel carriage, and the other end is fixed to the truss 51, forming a pulling force opposite to the direction of the third cylinder's movement. This ensures that the guide wheel can still press the cable in the non-drive state without affecting the free rotation of the wheel. The grinding and polishing assembly includes a ceramic eye 55, a support block 56, a first synchronous wheel 57, a first baffle 58, a second synchronous wheel 59, a third synchronous wheel 60, a polishing wheel 61, a grinding wheel 62, a cylinder liner 63, and a fourth cylinder 64. The ceramic eye 55 is used to guide the cable and is located on one side of the support block 56. The ceramic eye 55 is positioned in front of the grinding wheel 62 and the polishing wheel 61. The grinding wheel 62 and the polishing wheel 61, which perform grinding and polishing sequentially, are installed on the same side of the support block 56. There are two grinding wheels 62 and two polishing wheels 61, arranged longitudinally and correspondingly divided into upper and lower sections. The first synchronous wheel 57 is connected to the second synchronous wheel 59 via a belt to work together. The second synchronous wheel 59 rotates via an externally added motor, driving the lower wheel of the grinding wheel 62 and polishing wheel 61 to rotate. The third synchronous wheel 60 is mounted on the back of the support block 56 and rotates via the movement of the fourth cylinder 64, causing the upper wheel of the grinding wheel 62 and polishing wheel 61 connected to the third synchronous wheel 60 to rotate. The first baffle 58 is connected to the back of the support block 56. The working area of ​​the synchronous pulley is isolated to prevent external impurities from interfering with its movement. When the wire feeding module 1 starts, this module activates the cylinder in the encoder device, driving the cable while simultaneously recording the cable delivery length and receiving defect location signals from the defect detection module, thus precisely controlling the defect location. When the grinding and polishing device receives signals from the defect detection module and the encoder meter wheel, it activates the motors and cylinders of the grinding and polishing wheels, controlling them to process the defect repair location. The ceramic eye 55 is located on one side of the support block 56 and is used to guide the cable and align it with the subsequent grinding and polishing unit. Two grinding wheels 62 and two polishing wheels 61 are installed on the support block 56, which are the upper wheel and the lower wheel, respectively. They are arranged longitudinally along the direction of the cable and clamp the cable from the top and bottom for grinding and polishing. The ceramic eye 55 is located on one side of the support block 56 and is used to guide the cable and align it with the subsequent grinding and polishing unit. Two grinding wheels 62 and two polishing wheels 61 are installed on the support block 56, which are the upper and lower wheels respectively. They are arranged longitudinally along the direction of the cable and clamp the cable from the top and bottom for grinding and polishing.

[0028] The lower wheels of the grinding wheel 62 and the polishing wheel 61 are connected to the second synchronous wheel 59 via the first synchronous wheel 57. The second synchronous wheel 59 is driven to rotate by an external motor, and the first synchronous wheel 57 and the connected lower wheel are driven to rotate by a synchronous belt.

[0029] The upper wheel is supported by an independent carriage, which is driven by a fourth cylinder 64 to move up and down. The third synchronous pulley 60 is mounted on the back of the carriage and is coaxially connected to the upper grinding wheel and polishing wheel. To achieve the rotation function, a rack and pinion mechanism is set in the carriage. The fourth cylinder pushes the rack to move linearly back and forth, which drives the third synchronous pulley 60 to rotate, thereby driving the upper wheel to rotate.

[0030] The first baffle 58 is located on the back of the support block 56 and is used for auxiliary limiting and protection.

[0031] The lower wheels of the grinding wheel 62 and the polishing wheel 61 are connected to the second synchronous wheel 59 via the first synchronous wheel 57. The second synchronous wheel 59 is driven to rotate by an external motor, and the first synchronous wheel 57 and the connected lower wheel are driven to rotate by a synchronous belt.

[0032] The upper wheel is supported by an independent carriage, which is driven by a fourth cylinder 64 to move up and down. The third synchronous pulley 60 is mounted on the back of the carriage and is coaxially connected to the upper grinding wheel and polishing wheel. To achieve the rotation function, a synchronous belt drive structure is set in the support block 56. The fourth cylinder 64 drives the drive wheel to move, which drives the third synchronous pulley 60 to rotate via the synchronous belt, thereby driving the upper wheel to rotate.

[0033] Furthermore, the take-up module 7 includes an aluminum bracket 65, a support 66, a second motor 67, a motor mounting plate 68, a reel 69, and a second baffle 70. The aluminum bracket 65 has four supports 66 installed at its bottom. The second motor 67 is mounted on the aluminum bracket 65 via the motor mounting plate 68. The reel 69 is mounted on the aluminum bracket 65. The second baffle 70 is mounted on the center edge of the cable.

[0034] Furthermore, after acquiring images of the cable surface, defect detection module 2 performs defect detection, with the following specific steps: S1. Construct an image preprocessing and ROI intelligent localization module, such as... Figure 2As shown. In the image input and normalization stage, RGB images from an industrial camera with a resolution of 224×224 pixels are accepted; the input image is normalized, and the pixel range is adjusted to [0, 1]; ImageNet normalization parameters are applied, with a mean of [0.485, 0.456, 0.406] and a standard deviation of [0.229, 0.224, 0.225]. In the cable region intelligent localization stage, a preset 3×3 Sobel edge detection convolution is used to convolve the input image to calculate the edge intensity map. The cable region T is determined by threshold segmentation, and connected regions satisfying E(x, y)>T are extracted. The minimum bounding matrix is ​​calculated, and the detection region is expanded by 20 pixels to ensure complete coverage of the cable region. In the ROI region verification and optimization stage, the area ratio of the ROI region is verified to be within the range of 10%-80%. If the ROI region is too small or too large, the entire image is used as the detection region, and the ROI coordinate information is recorded for subsequent result mapping. S2. Construct a lightweight multi-scale feature extraction module and a multi-scale feature fusion module. In the initial feature extraction stage, a 3×3 convolutional layer is applied to the ROI region, with 3 input channels, 16 output channels, and a stride of 2; batch normalization and ReLU6 activation function are applied; the feature map size is reduced from 224×224 to 112×112. In the depthwise separable convolutional feature extraction stage, four separable convolutional blocks are applied sequentially as follows: the first convolutional block outputs a feature map of 56×56×32, the second convolutional block outputs a feature map of 28×28×128, the third convolutional block outputs a feature map of 28×28×128, and the fourth convolutional block outputs a feature map of 28×28×128. In the feature alignment and dimensionality reduction stage, features F_shallow (28×28×64) and deep features F_deep (28×28×128) are extracted. Shallow features are reduced to 32 channels using 1×1 convolutions, and deep features are also reduced to 32 channels using 1×1 convolutions, ensuring consistent spatial dimensions between the two feature maps. In the channel attention calculation stage, global average pooling and global max pooling are applied to the fused feature F_concat (28×28×64). Channel weights are calculated using a two-layer fully connected network, and attention weights are applied. A depthwise separable convolution is applied to the attention-enhancing feature, outputting the fused feature F_fused (28×28×128). S3. Construct an adaptive segmentation and direction-aware module and a defect prior knowledge injection module. In the cable direction detection stage, the horizontal gradient G_h and vertical gradient G_v of the fused features are calculated, along with the direction energies E_h and E_v. The cable direction is determined: if E_h > E_v, it is a horizontal cable; otherwise, it is a vertical cable. In the adaptive segmentation strategy stage, segmentation parameters are selected based on the detected cable direction: for horizontal cables, pacth_size = (3, 5), stride = 1, padding = (1, 2); for vertical cables, patch_size = (5, 3), stride = 1, padding = (2, 1). In the feature segmentation operation stage, the unfold operation is used to convert the feature map into patches, the number of patches N is calculated, and the pacth feature matrix P (N×128) is output. In the defect pattern matching stage, four learnable defect pattern parameters are maintained: M1, M2, M3, and M4, each a 128-dimensional vector. The similarity between each patch and each pattern is calculated; that is, the channels are reduced from 128 to 4 using a two-layer fully connected network, and the probability distribution S is obtained by applying the Softmax function. In the prior knowledge fusion stage, the weighted defect pattern M_weighted is calculated. The input concat(P, S) is processed through a feature enhancement network with a dimension of 132, and the enhanced feature P_enchanted is calculated. S4. Construct a miniature discriminator anomaly detection module. In the anomaly score stage, a miniature discriminator is applied to each enhancement patch. The number of channels in the first layer is reduced from 128 to 64, with batch normalization and LeakyReLU (0.2) activation. The number of channels in the second layer is reduced from 64 to 1, with no activation. Anomaly scores (Score) are output simultaneously. In the spatial anomaly map reconstruction stage, pace-level anomaly scores are rearranged into a spatial grid. The network size (grid_h) is calculated, and bilinear interpolation is used to upsample the anomaly score map to the original ROI size, generating a pixel-level anomaly heatmap. In the image-level anomaly evaluation stage, the image-level anomaly score (Image_score) is calculated. A threshold is applied to determine the output detection results. All thresholds used are set to the default value of 0.5. Model testing also shows that a threshold of 0.5 yields the best results. Inspired by the SimpleNet model, CableDefectNet constructs a five-layer progressive technical route: intelligent ROI localization module, lightweight backbone network redesign module, adaptive block segmentation strategy module, and multi-scale feature module. First, the system uses the Sobel edge detection operator in the ROI intelligent localization module to quickly identify cable regions in industrial images, reducing the computational range by 60-80%. The lightweight backbone network replaces traditional convolution with depthwise separable convolution, progressively expanding through 16→32→64→128 channels and decreasing spatial resolution. Feature extraction hooks are set in layers 3 and 5 to achieve multi-level feature representation from fine-grained texture to high-level semantics. Then, the multi-scale feature fusion module aligns features at different scales using bilinear interpolation, highlights important feature channels using a lightweight channel attention mechanism, and completes feature fusion using depthwise separable convolution to form a comprehensive feature representation that combines detail and semantic information. Finally, the adaptive block segmentation and orientation-aware module... The block automatically determines the cable direction by calculating horizontal and vertical gradient energy and dynamically adjusts the patch shape (3×5 for horizontal cables and 5×3 for vertical cables) to perfectly match the detection strategy with the cable geometry. Next, the defect prior knowledge injection module maintains four learnable defect pattern parameters and calculates the similarity between each patch and typical defects through a pattern matching network. Expert knowledge is injected into the feature representation in a weighted fusion manner, realizing the organic combination of domain knowledge and deep learning. Finally, the mini discriminator adopts a minimalist two-layer fully connected architecture, learns the discrimination boundary of normal features through adversarial training, outputs patch-level anomaly scores during inference, and generates pixel-level anomaly heatmaps and image-level anomaly assessments through spatial reconstruction and interpolation. A large number of normal cable images are used as the training set, and the validation set consists of normal test samples and defect test samples. Together, they form a cable surface bump defect image dataset composed of the training set and the validation set. During pre-training, the hyperparameters in the configuration file are modified. The main hyperparameters in the configuration file are the learning rate, number of images extracted, number of iterations, and weight decay coefficient; the remaining hyperparameters are left at their default values. In this example, the learning rate is 0.01, the number of iterations is 100, the weight decay coefficient is 0.0005, and the remaining hyperparameters are left at their default values. The experimental equipment is a desktop computer with an NVIDIA GeForce RTX 2080 GPU, 16GB of RAM, and an Intel(R) Core(TM) i5-13600KF CPU. Input a validation set, and use the PyTorch deep learning network framework to evaluate the performance of the pre-trained weight parameters obtained in the above steps, selecting the optimal pre-trained weight parameters. Specifically, quantitative performance evaluation of the model is performed using the validation set of a cable surface defect bump dataset, with optimization based on average precision, frame rate, false negative rate, and accuracy. The optimal weight parameters are then loaded into the object detection algorithm to generate a deep learning detection model. The equipment begins operation after supplying the cable to be inspected, activating the defect detection module, and adjusting the camera parameters. Camera parameters include image acquisition speed, pixel size, and memory space. Specifically, the camera acquires RGB images of 3246×1026 resolution, and the image acquisition card has 5000Mb of memory. The industrial camera is a Basler color camera with a resolution of 2.3MP and a frame rate of 51fps. The camera lens is a Basler Standard C-mount lens with a maximum image circle of 1 inch, an 8.0mm fixed focal length, an aperture range of F1.4-F16, and a standard resolution of 2 megapixels. A light hood is used to overcome ambient light interference, ensuring stable lighting and ideal imaging results during image acquisition. This invention focuses on the mechanical detection of defects in cables produced in factories, specifically lugs and caves. Since lugs and caves occur frequently in actual production, their detection is highly significant. Furthermore, given image data of other types of defects used to train the network, the defect detection module of this invention can also effectively detect the corresponding defects. This invention uses two different network models trained to detect lugs and caves, and the specific results are shown in Table 1. The detection accuracy table shows that the AUC values ​​for all configurations exceed 0.90, with three configurations reaching over 0.93. Precision, recall, and F1 score maintain good consistency across configurations, demonstrating the model's stability and reliability. The high-precision configuration achieves optimal performance across all metrics, with an AUC of 0.97, precision of 0.95, recall of 0.96, and F1 score of 0.95. These figures demonstrate that this configuration meets the stringent requirements of industrial applications in terms of both accuracy and completeness of defect detection, making it particularly suitable for critical applications demanding extremely high detection precision. The Lightweight configuration, while the most lightweight version, still achieves an AUC of 0.92, precision of 0.89, recall of 0.91, and F1 score of 0.90. The Default configuration, as the standard version, performs excellently across all metrics (AUC 0.95, precision 0.93, recall 0.94, F1 score 0.93), representing the optimal balance between accuracy and efficiency. The Realtime configuration is optimized for real-time performance; although it slightly compromises on precision, it still achieves an AUC of 0.93, precision of 0.90, recall of 0.92, and F1 score of 0.91, fully meeting the performance requirements of real-time detection. Table 1. Results of Experimental Detection Accuracy and Inference Performance By applying the above technical solution, the problems of low efficiency, low accuracy, difficult and poor results of manual visual inspection, and unsuitability of traditional detection algorithms for small target sizes in current cable surface bump defect detection operations are effectively solved. This invention utilizes deep learning for analysis, automatically identifying target defects using a trained model, and then repairing the specified defects using a defect injection molding module. This effectively reduces the labor intensity of workers while improving the integrity of the repair and ensuring the product quality of the cable.

[0035] This invention also provides a method for detecting and repairing manufacturing defects on the surface of cables, comprising the following steps: Step 1: Turn on the wire feeding module, and use the defect detection module 2 to acquire images in real time and detect concave and convex defects in the acquired images; Step 2: When the defect detection module 2 detects a concave defect on the cable surface, the cable is injection molded by the injection molding module 3; When the defect is determined to be a dent, the injection molding module (3) is started to the designated defect location. After the molten plastic is injected and cooled, the trimming module is started to trim the repaired area on the cable surface. Step 3: Repair the protrusion defects on the upper and lower cylindrical surfaces of the cable using the first trimming module 4 and the second trimming module 5; When the defect is determined to be an upper bump defect, the first trimming module (4) is started and the defect is trimmed after moving to the designated position; When the defect is determined to be a lower bump defect, the second trimming module (5) is activated and trimmed at the specified position; Step 4: Start the polishing module 6 to polish the surface of the injection-molded or repaired cable, and then use the cable take-up module 7 to take up the inspected and repaired cable.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cable surface manufacturing defect detection and repair system, characterized in that: The system includes a wire feeding module (1), a defect detection module (2), an injection molding module (3), a first trimming module (4), a second trimming module (5), a polishing module (6), and a take-up module (7). The wire feeding module (1) is used to feed the produced cable to the defect detection module (2). The defect detection module (2) is used to acquire images in real time and detect concave and convex defects in the acquired images. The injection molding module (3) is used to inject the cable when the defect detection module (2) detects concave defects on the cable surface. The first trimming module (4) is used to trim convex defects on the cylindrical surface of the cable. The second trimming module (5) is used to trim convex defects on the cylindrical surface of the cable. The polishing module (6) is used to polish the surface of the injected or trimmed cable. The take-up module (7) is used to take up the detected and repaired cable. The grinding and polishing module (6) includes an encoding meter wheel assembly and a grinding and polishing assembly; The grinding and polishing assembly includes a ceramic eye (55), a support block (56), a first synchronous wheel (57), a first baffle (58), a second synchronous wheel (59), a third synchronous wheel (60), a polishing wheel (61), a grinding wheel (62), a cylinder liner (63), and a fourth cylinder (64). The ceramic eye (55) is used to guide the cable and is located on one side of the support block (56). The ceramic eye (55) is located in front of the grinding wheel (62) and the polishing wheel (61). The support block (56) has two grinding wheels (62) and two polishing wheels (61) installed on the same side for grinding and polishing, and they are arranged longitudinally. It should be divided into upper and lower wheels. The fourth cylinder (64) is assembled with the cylinder liner (63) and connected to the support block (56). The first synchronous wheel (57) is connected to the second synchronous wheel (59) through a belt to work together. The second synchronous wheel (59) rotates through an externally added motor, driving the lower wheel of the grinding wheel (62) and polishing wheel (61) to rotate. The third synchronous wheel (60) is installed on the back of the support block (56) and rotates through the movement of the fourth cylinder (64), so that the upper wheel of the grinding wheel (62) and polishing wheel (61) connected to the third synchronous wheel (60) rotates. The first baffle (58) is connected to the back of the support block (56).

2. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The cable delivery module (1) includes a cable rack (8), a support frame (9), a first motor (10), a motor connection line (11), and a motor switch (12). The support frame (9) is installed on the cable rack (8). The first motor (10) and the motor switch (12) are connected by the motor connection line (11) and control the cable rack (8) to deliver the cable.

3. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The defect detection module (2) includes a detection platform (13), a camera bracket (14), a guide ring (15), a power interface (16), a first support plate (17), a camera mounting plate (18), an industrial camera (19), a light source fixing block (20), a light source (21), and a guide ring upright plate (22). The camera bracket (14) and the guide ring upright plate (22) are installed on the detection platform (13). The camera bracket (14) is provided with a first support plate (17) along the cable travel direction. The first support plate (17) is hexagonal, and a camera mounting plate (18) is installed every 120 degrees. Each camera mounting plate (18) is provided with an industrial camera (19), a light source fixing block (20), and a light source (21) facing the cable travel direction. The light source (21) is set on the light source fixing block (20). The light source (21) has a power interface (16). The cable passes through the guide ring (15), passes through the first support plate (17), and then passes through the guide ring upright plate (22).

4. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The injection molding module (3) includes an upper injection molding assembly and a lower clamping assembly. The upper injection molding assembly includes a needle (23), a nozzle sleeve fixing block (24), a nozzle sleeve (25), a glue tube knob (26), a glue tube sleeve (27), a glue tube (28), a buffer assembly knob (29), a glue tube fixing block (30), and a buffer assembly (31). The glue tube (28) stores molten plastic and is injected into the corresponding position of the defect on the cable surface through the needle (23). The needle (23) is fixed by the nozzle sleeve (25) installed in the nozzle sleeve fixing block (24). The glue tube (28) is fixed by the glue tube sleeve (27) installed in the glue tube fixing block (30). At the same time, the glue tube knob (26) is installed on the side to control the tightness of the glue tube (28). The buffer assembly (31) is equipped with a buffer assembly knob (29) on the top. The lower clamping assembly includes a second support plate (32), a fixing block (33), a guide block (34), a finger cylinder (35), and a base (36). The fixing block (33) is installed at the bottom of the second support plate (32). The fixing block (33) is controlled by adjusting the finger cylinder (35) to clamp and tighten the guide block (34). The finger cylinder (35) is installed on the base (36).

5. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The first trimming module (4) includes a first bracket (37), a first cylinder (38), a first tool fixing block (39), a first arc tool (40), a first lead wire fixing block (41), and a first support block (42). The first support block (42) is installed at the end of the first bracket (37). The first support block (42) is used to support the first cylinder (38). The output end of the first cylinder (38) is connected to the first tool fixing block (39). One end of the first tool fixing block (39) is connected to the first arc tool (40) and trims the cylindrical surface of the cable supported in the first lead wire fixing block (41).

6. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The second trimming module includes a second arc cutter (43), a second lead wire fixing block (44), a second support block (45), a second bracket (46), a second cylinder (47), and a second cutter fixing block (48). The second support block (45) is installed at the end of the second bracket (46). The second arc cutter (43) is connected to the second cylinder (47) through the second cutter fixing block (48) to trim the cylindrical surface of the cable supported in the second lead wire fixing block (44).

7. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The take-up module (7) includes an aluminum bracket (65), a support (66), a second motor (67), a motor mounting plate (68), a reel (69), and a second baffle (70). The aluminum bracket (65) has four supports (66) installed at its bottom. The second motor (67) is mounted on the aluminum bracket (65) via the motor mounting plate (68). The reel (69) is mounted on the aluminum bracket (65). The second baffle (70) is mounted on the center edge of the cable.

8. The cable surface manufacturing defect detection and repair system according to claim 2, characterized in that: After acquiring the cable surface image, the defect detection module (2) performs defect detection. The specific steps are as follows: S1: Accepts RGB images from industrial camera (19); S2: Defects are identified using the constructed CableDefectNet, and pixel-level anomaly heatmaps and image-level anomaly assessments are generated, with the defect type output.

9. The cable surface manufacturing defect detection and repair system according to claim 1, characterized in that: The encoder counting wheel assembly includes an encoder (49), a third cylinder (50), a truss (51), a guide wheel (52), a tension spring (53), and a counting wheel (54). The third cylinder (50) is installed on one side of the bottom of the truss (51). The output end of the third cylinder (50) is connected to the guide wheel (52). The counting wheel (54) is installed on the side of the truss (51) corresponding to the guide wheel (52). The cable passes between the counting wheel (54) and the guide wheel (52) and drives the counting wheel (54) to rotate, and transmits the rotation data to the encoder (49). The tension spring (53) connects the guide wheel (52) and the counting wheel (54).

10. A method for detecting and repairing manufacturing defects on the surface of cables, based on the cable surface manufacturing defect detection and repair system described in any one of claims 1-9, characterized in that: Includes the following steps: Step 1: Turn on the wire feeding module, and use the defect detection module (2) to collect images in real time and detect concave and convex defects in the collected images; Step 2: When the defect detection module (2) detects a concave defect on the surface of the cable, the cable is injection molded by the injection molding module (3); Step 3: Repair the protrusion defects on the upper and lower cylindrical surfaces of the cable using the first trimming module (4) and the second trimming module (5); Step 4: Start the polishing module (6) to polish the surface of the injection-molded or repaired cable, and use the cable take-up module (7) to take up the inspected and repaired cable.

Images