A method and system for linking edge intelligent decision-making and live-line defect elimination tools

Abstract

The application belongs to the technical field of power transmission line operation and maintenance, and discloses a linkage method and system combining edge intelligent decision and live-line defect elimination tools. The application can realize on-site identification and edge side analysis of the fastener of the power transmission tower by collecting image data, vibration data and bolt models of the fastener of the power transmission tower, inputting the standardized preprocessed data into the bolt defect fine-grained identification network deployed on the edge side, distinguishing different defect types such as bolt loosening, thread corrosion and missing lock washer, and obtaining defect grades, attitude parameters and confidence information. The application can improve the fine degree of fastener defect identification and reduce manual secondary verification.

Description

Technical Field

[0001] The present invention belongs to the technical field of transmission line operation and maintenance, and specifically relates to a linkage method and system integrating edge intelligent decision-making and live defect elimination tools. Background Art

[0002] Transmission lines are in an outdoor complex environment for a long time. The connection parts and fasteners of the tower are easily affected by factors such as wind vibration, humidity, corrosion, and mechanical load, resulting in defects such as bolt loosening, thread corrosion, missing lock washers, and foreign object entanglement. If not discovered and processed in time, it may affect the structural stability of the tower and the safe operation of the transmission line. In the prior art, there is already an intelligent risk control platform for the input, review, treatment, reminder, and evaluation of potential hazards in transmission lines, which can improve the standardization and closed-loop management of potential hazard management. However, it mainly focuses on the information flow and business management of potential hazards, still relying on manual inspections, manual judgments, and manual defect elimination, and cannot achieve on-site automatic identification, precise positioning, intelligent decision-making, and linkage execution of live defect elimination tools for tower fastener defects.

[0003] Currently, during live defect elimination operations, maintenance personnel usually discover abnormalities through visual inspection, drone inspection, or ordinary video monitoring, and then use tools such as insulated long rods for processing. This method relies strongly on personnel experience, and the defect recognition granularity is low. It can often only judge the existence of bolt-like defects and is difficult to accurately distinguish specific types such as bolt loosening, thread corrosion, and missing lock washers, resulting in a need for secondary on-site verification. At the same time, in complex environments such as occlusion, sudden changes in light, and wind vibration interference, small target defects are prone to misjudgment or missed judgment; there are also problems such as insufficient alignment accuracy, unstable operation posture, high risk of accidental contact with live bodies, and the need for re-inspection and confirmation after defect elimination during manual operations. Therefore, there is an urgent need for a live defect elimination system and method that integrates multi-modal perception, edge intelligent analysis, fine-grained defect recognition, dynamic defect elimination decision-making, safe linkage execution, and effect closed-loop verification. Summary of the Invention

[0004] The purpose of the present invention is to overcome the problems existing in the prior art of live defect elimination for transmission towers, such as low defect recognition granularity, insufficient accuracy and safety of manual operations, lack of intelligent linkage between recognition and defect elimination tools, and difficulty in closed-loop verification of defect elimination effects, and provide a linkage method and system integrating edge intelligent decision-making and live defect elimination tools.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: In the first aspect, the present invention provides a linkage method integrating edge intelligent decision-making and live defect elimination tools, including the following steps: Collect image data, vibration data, and bolt models of transmission tower fasteners, and perform standardized preprocessing on the collected data; The preprocessed data is input into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, attitude parameters and confidence information of the fastener; Based on the fastener's defect type, defect level, attitude parameters, and confidence information, and combined with defect-action mapping data, a defect removal control command is generated to control the live defect removal tool. Safety checks were performed on the voltage at the work site and the distance between the live fault-finding tool and the live conductor. After the safety verification is passed, the live-line troubleshooting tool is controlled to perform live-line troubleshooting on the fastener according to the troubleshooting control command; Image and vibration data of the fasteners were re-acquired, and the defect elimination effect was verified based on a fine-grained bolt defect identification network. If the defect elimination effect fails to meet the standard, the defect elimination control command will be regenerated based on the reason for failure, and the live defect elimination tool will be controlled to perform the defect elimination action again until the defect elimination effect meets the standard.

[0006] A further improvement of this invention lies in the following method for standardizing and preprocessing the collected image data, vibration data, and bolt type of the transmission tower fasteners: Collect image data, vibration data, environmental data, and conductor motion data of transmission tower fasteners; Adaptive denoising and exposure compensation are performed on the image data, the area required for the fastener is cropped, and the image data is compressed to the required resolution through bilinear interpolation. Data enhancement processing with random flipping and required brightness perturbation is then applied to the compressed image data. Kalman filtering is applied to the vibration data to eliminate noise, and the vibration data is standardized to the required range. The bolt type is structured and coded to obtain bolt type code data; The preprocessed image data, vibration data, and bolt type code data are stored in the edge database.

[0007] A further improvement of this invention lies in inputting the preprocessed data into a fine-grained bolt defect recognition network deployed on the edge side. The specific method for recognizing the defect information, attitude parameters, and confidence information of the fastener is as follows: The fine-grained identification network for bolt defects is activated by inputting the preprocessed data into the network. The global structural features of bolts are extracted through the backbone feature extraction layer of the fine-grained bolt defect identification network. The thread grayscale gradient features are extracted by the thread texture branch of the fine-grained bolt defect recognition network, and the corrosion probability and corrosion level are output. The annular edge features of the washer are extracted by the washer contour branch of the bolt defect fine-grained identification network, and the probability of the existence and integrity of the washer are output. The bolt posture branch of the bolt defect fine-grained identification network is used to extract the gap features, bolt tilt angle and vibration features between the bolt head and the tower contact surface, and output the loosening probability and loosening level. The features of each branch are dynamically weighted and fused. The final confidence level is determined by visual recognition confidence level and vibration correction coefficient. The output includes the defect type (normal, loose bolt, rusted thread or missing anti-loosening washer), the defect level (minor, moderate or severe), the bolt tilt angle, and the final confidence level, which serve as the defect information, attitude parameters and confidence level information of the fastener.

[0008] A further improvement of this invention lies in the following method for generating defect removal control commands for controlling live defect removal tools, based on the defect type, defect level, attitude parameters, and confidence information of the fastener, combined with defect-motion mapping data: Obtain the defect type, defect level, attitude parameters and confidence information of the fastener, call the historical defect database and the standard parameter library of the same type of tower for cross-validation, supplement the standard torque and past defect elimination case information of the corresponding bolt type, and generate defect feature file; Based on the defect feature file, the defect and action mapping data is called. When the defect type is loose bolt, an electric fastening head is matched; when the defect type is thread corrosion, an insulating derusting head is matched; and when the defect type is missing anti-loosening washer, a washer installation head is matched. Match the number of rotations, torque, and range of motion according to the defect level and bolt type; The initial execution angle of the end effector is determined based on the bolt tilt angle, and the action parameters are corrected by combining wind speed and humidity data, which serve as the defect elimination control command.

[0009] A further improvement of this invention is that the specific method for safety verification of the voltage at the work point and the distance between the live fault-finding tool and the live conductor is as follows: After receiving the troubleshooting control command, the live fault elimination tool detects the voltage at the work point through a voltage sensing probe and determines whether there is a risk of voltage exceeding the standard based on the safety threshold corresponding to the voltage level at the work point. Images of the work site are collected, and the distance between the end effector and the live conductor is determined to be greater than or equal to the safety threshold based on the images. When the voltage at the work site does not exceed the corresponding safety threshold, and the distance between the end effector and the live conductor is greater than or equal to the safety threshold, the safety verification is deemed to have passed.

[0010] A further improvement of this invention is that, after the safety verification is passed, the specific method for controlling the live-line fault-clearing tool to perform live-line fault-clearing actions on the fastener according to the fault-clearing control command is as follows: After the safety verification is passed, the insulating rod of the live defect elimination tool is extended and retracted to the target position, and the end effector is switched to the type of effector corresponding to the defect type and adjusted to the execution angle determined by the defect elimination control command. During the defect elimination process, the attitude data of the live defect elimination tool is transmitted back in real time. When the position error exceeds the accuracy threshold, the live defect elimination tool is corrected. The real-time live fault removal tool monitors the fit between the end effector and the fastener, and dynamically adjusts the action force. The tool's posture, motion progress, and real-time image data during execution are archived.

[0011] A further improvement of this invention lies in the method of re-acquiring image and vibration data of the fasteners and verifying the defect elimination effect based on a fine-grained bolt defect identification network, as follows: After the live fault elimination operation is completed, image and vibration data of the fasteners are reacquired. The fine-grained bolt defect recognition network verifies the effectiveness of the re-acquired image and vibration data, compares the defect-removed image data with the stored standard template of the same bolt model to determine whether the visual defects have been eliminated, and judges whether the fastener connection status and heating status are qualified based on the defect-removed vibration and image data. When visual defects are eliminated and the fastener connection status and heating status are qualified, the defect elimination effect is deemed to have passed the verification.

[0012] A further improvement of this invention is that if the defect elimination effect verification fails to meet the standard, the defect elimination control command is regenerated based on the reason for failure, and the live defect elimination tool is controlled to perform the defect elimination action again until the defect elimination effect verification meets the standard. The specific method is as follows: When the defect elimination effect verification result is not up to standard, feedback information is generated based on the reason for not meeting the standard; Based on the feedback information, the defect elimination control instructions were re-optimized, and the safety verification, live defect elimination actions, and defect elimination effect verification were re-executed. When the defect elimination effect is verified to be up to standard, a defect elimination work report containing defect information, execution parameters and effect data is generated and archived.

[0013] Secondly, the present invention provides a linkage system integrating edge intelligent decision-making and live fault elimination tools, comprising: The preprocessing module is used to collect image data, vibration data, and bolt type of fasteners on transmission towers, and to perform standardized preprocessing on the collected data; The network recognition module is used to input preprocessed data into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, posture parameters and confidence information of the fastener; The instruction generation module is used to generate defect removal control instructions for controlling live defect removal tools based on the defect type, defect level, attitude parameters and confidence information of the fastener, combined with defect and motion mapping data. The safety verification module is used to perform safety verification on the voltage at the work point and the distance between the live fault removal tool and the live conductor; The execution module is used to control the live-line repair tool to perform live-line repair actions on the fasteners according to the repair control instructions after the safety verification is passed; The effect verification module is used to re-acquire image and vibration data of fasteners and verify the defect elimination effect based on the fine-grained bolt defect recognition network. The instruction optimization module is used to regenerate the defect elimination control instruction based on the reason for failure if the defect elimination effect verification fails, and then control the live defect elimination tool to perform the defect elimination action again until the defect elimination effect verification meets the standard.

[0014] A further improvement of this invention is that the function of the preprocessing module is implemented through the following method: Collect image data, vibration data, environmental data, and conductor motion data of transmission tower fasteners; Adaptive denoising and exposure compensation are performed on the image data, the area required for the fastener is cropped, and the image data is compressed to the required resolution through bilinear interpolation. Data enhancement processing with random flipping and required brightness perturbation is then applied to the compressed image data. Kalman filtering is applied to the vibration data to eliminate noise, and the vibration data is standardized to the required range. The bolt type is structured and coded to obtain bolt type code data; The preprocessed image data, vibration data, and bolt type code data are stored in the edge database.

[0015] A further improvement of this invention is that the function of the network identification module is implemented through the following method: The fine-grained identification network for bolt defects is activated by inputting the preprocessed data into the network. The global structural features of bolts are extracted through the backbone feature extraction layer of the fine-grained bolt defect identification network. The thread grayscale gradient features are extracted by the thread texture branch of the fine-grained bolt defect recognition network, and the corrosion probability and corrosion level are output. The annular edge features of the washer are extracted by the washer contour branch of the bolt defect fine-grained identification network, and the probability of the existence and integrity of the washer are output. The bolt posture branch of the bolt defect fine-grained identification network is used to extract the gap features, bolt tilt angle and vibration features between the bolt head and the tower contact surface, and output the loosening probability and loosening level. The features of each branch are dynamically weighted and fused. The final confidence level is determined by visual recognition confidence level and vibration correction coefficient. The output includes the defect type (normal, loose bolt, rusted thread or missing anti-loosening washer), the defect level (minor, moderate or severe), the bolt tilt angle, and the final confidence level, which serve as the defect information, attitude parameters and confidence level information of the fastener.

[0016] A further improvement of this invention is that the function of the instruction generation module is implemented through the following method: Obtain the defect type, defect level, attitude parameters and confidence information of the fastener, call the historical defect database and the standard parameter library of the same type of tower for cross-validation, supplement the standard torque and past defect elimination case information of the corresponding bolt type, and generate defect feature file; Based on the defect feature file, the defect and action mapping data is called. When the defect type is loose bolt, an electric fastening head is matched; when the defect type is thread corrosion, an insulating derusting head is matched; and when the defect type is missing anti-loosening washer, a washer installation head is matched. Match the number of rotations, torque, and range of motion according to the defect level and bolt type; The initial execution angle of the end effector is determined based on the bolt tilt angle, and the action parameters are corrected by combining wind speed and humidity data, which serve as the defect elimination control command.

[0017] A further improvement of this invention is that the function of the security verification module is implemented through the following method: After receiving the troubleshooting control command, the live fault elimination tool detects the voltage at the work point through a voltage sensing probe and determines whether there is a risk of voltage exceeding the standard based on the safety threshold corresponding to the voltage level at the work point. Images of the work site are collected, and the distance between the end effector and the live conductor is determined to be greater than or equal to the safety threshold based on the images. When the voltage at the work site does not exceed the corresponding safety threshold, and the distance between the end effector and the live conductor is greater than or equal to the safety threshold, the safety verification is deemed to have passed.

[0018] A further improvement of this invention is that the function of the execution module is implemented through the following method: After the safety verification is passed, the insulating rod of the live defect elimination tool is extended and retracted to the target position, and the end effector is switched to the type of effector corresponding to the defect type and adjusted to the execution angle determined by the defect elimination control command. During the defect elimination process, the attitude data of the live defect elimination tool is transmitted back in real time. When the position error exceeds the accuracy threshold, the live defect elimination tool is corrected. The real-time live fault removal tool monitors the fit between the end effector and the fastener, and dynamically adjusts the action force. The tool's posture, motion progress, and real-time image data during execution are archived.

[0019] A further improvement of this invention is that the function of the effect verification module is implemented through the following method: After the live fault elimination operation is completed, image and vibration data of the fasteners are reacquired. The fine-grained bolt defect recognition network verifies the effectiveness of the re-acquired image and vibration data, compares the defect-removed image data with the stored standard template of the same bolt model to determine whether the visual defects have been eliminated, and judges whether the fastener connection status and heating status are qualified based on the defect-removed vibration and image data. When visual defects are eliminated and the fastener connection status and heating status are qualified, the defect elimination effect is deemed to have passed the verification.

[0020] A further improvement of this invention is that the function of the instruction optimization module is implemented through the following method: When the defect elimination effect verification result is not up to standard, feedback information is generated based on the reason for not meeting the standard; Based on the feedback information, the defect elimination control instructions were re-optimized, and the safety verification, live defect elimination actions, and defect elimination effect verification were re-executed. When the defect elimination effect is verified to be up to standard, a defect elimination work report containing defect information, execution parameters and effect data is generated and archived.

[0021] Thirdly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a method for linking edge intelligent decision-making with live fault elimination tools.

[0022] Fourthly, the present invention provides a storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, it implements the steps of a method for linking edge intelligent decision-making with live fault elimination tools.

[0023] Compared with the prior art, the present invention has the following beneficial effects: This invention collects image data, vibration data, and bolt type information of fasteners on power transmission towers. The standardized, pre-processed data is then input into a fine-grained bolt defect identification network deployed on the edge side. This enables on-site identification and edge-side analysis of fasteners on power transmission towers, distinguishing between different defect types such as bolt loosening, thread corrosion, and missing anti-loosening washers. It also obtains defect level, attitude parameters, and confidence information. This invention improves the precision of fastener defect identification and reduces the need for secondary manual checks. Based on the fastener defect type, defect level, attitude parameters, and confidence information, this invention generates defect elimination control commands by combining defect-action mapping data. This allows live-line defect elimination tools to match corresponding defect elimination actions for different defect types and determines the corresponding execution parameters based on the defect level and attitude parameters. This reduces the reliance on personnel experience and judgment in defect elimination operations, improving the matching between the defect elimination plan and the actual defect state, and enhancing the consistency of operation control. This invention performs safety checks on the voltage at the work point and the distance between the live-line defect elimination tool and the live conductor before the tool performs the defect elimination action. Only after the safety check is passed is the tool controlled to perform the action. This allows for the assessment of voltage over-limit risks and accidental contact with live conductors before live-line work, reducing safety risks during live-line defect elimination and improving the safety of the operation. After the live-line defect elimination action is completed, the invention re-collects image and vibration data of the fasteners and verifies the defect elimination effect based on a fine-grained bolt defect identification network. If the defect elimination effect verification fails, a new defect elimination control command is generated based on the reason for the failure, and the live-line defect elimination tool is controlled to perform the action again until the defect elimination effect verification is successful. This forms a closed-loop processing flow from defect identification, control decision-making, safe execution to effect verification, avoiding the separation of defect identification, defect elimination execution, and effect confirmation, reducing repetitive inspections and operations, and improving the automation level and maintenance efficiency of live-line defect elimination for transmission tower fasteners. Attached Figure Description

[0024] Figure 1 This is a flowchart of the present invention; Figure 2 This is a system diagram of the present invention; Figure 3 This is a system architecture diagram for Example 3; Figure 4 A schematic diagram of the overall structure of the Bolt Defect Fine-Grained Recognition Network (BFR-Net); Figure 5 This is a flowchart of multimodal data acquisition and preprocessing. Figure 6 A schematic diagram of dynamic weighted fusion of multi-branch features; Figure 7 A schematic diagram of the confidence level correction logic; Figure 8 This is a schematic diagram of Example 6. Detailed Implementation

[0025] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0026] Example 1: See Figure 1 A method for integrating edge intelligent decision-making with live fault elimination tools includes the following steps: S1 collects image data, vibration data, and bolt type of fasteners on transmission towers, and performs standardized preprocessing on the collected data.

[0027] S2 inputs the preprocessed data into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, attitude parameters and confidence information of the fastener.

[0028] S3 generates defect removal control commands for controlling live defect removal tools based on the defect type, defect level, attitude parameters, and confidence information of the fastener, combined with defect-action mapping data.

[0029] S4 performs safety checks on the voltage at the work point and the distance between the live fault-finding tool and the live conductor.

[0030] S5, after the safety verification is passed, controls the live fault removal tool to perform live fault removal on the fastener according to the fault removal control command.

[0031] S6, re-acquire image and vibration data of the fasteners, and verify the defect elimination effect based on the fine-grained bolt defect identification network.

[0032] S7. If the defect elimination effect verification fails to meet the standard, the defect elimination control command will be regenerated according to the reason for failure, and the live defect elimination tool will be controlled to perform the defect elimination action again until the defect elimination effect verification meets the standard.

[0033] Example 2: See Figure 2 A linkage system integrating edge intelligent decision-making and live fault elimination tools, comprising: The preprocessing module is used to collect image data, vibration data, and bolt type of fasteners on power transmission towers, and to perform standardized preprocessing on the collected data.

[0034] The network recognition module is used to input preprocessed data into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, attitude parameters and confidence information of the fastener.

[0035] The instruction generation module is used to generate defect removal control instructions for controlling live defect removal tools based on the defect type, defect level, attitude parameters, and confidence information of the fastener, combined with defect-action mapping data.

[0036] The safety verification module is used to perform safety verification on the voltage at the work point and the distance between the live fault removal tool and the live conductor.

[0037] The execution module is used to control the live fault removal tool to perform live fault removal actions on the fasteners according to the fault removal control instructions after the safety verification is passed.

[0038] The effect verification module is used to re-acquire image and vibration data of fasteners and verify the defect elimination effect based on a fine-grained bolt defect identification network.

[0039] The instruction optimization module is used to regenerate the defect elimination control instruction based on the reason for failure if the defect elimination effect verification fails, and then control the live defect elimination tool to perform the defect elimination action again until the defect elimination effect verification meets the standard.

[0040] Example 3: See Figure 3 This embodiment is applicable to the identification, elimination, linkage execution, and effect verification of live defects in transmission tower fasteners. The fasteners in this embodiment include bolts, nuts, and anti-loosening washers on transmission towers. The identified defect types include normal, loose bolts, thread corrosion, and missing anti-loosening washers.

[0041] This embodiment is based on a system that integrates edge intelligent decision-making and live-line defect elimination tools. The system includes a sensing and acquisition module, an intelligent identification and decision-making module, an execution and defect elimination module, a cloud platform management center, and an energy supply module. The sensing and acquisition module collects image data, vibration data, and bolt type of transmission tower fasteners, and can also collect environmental data such as wind speed, humidity, and temperature, as well as conductor displacement data. The intelligent identification and decision-making module is deployed in the edge intelligent gateway or edge IoT agent of the transmission line, and includes a data access unit, an AI analysis unit, and a control decision-making unit. The execution and defect elimination module is the live-line defect elimination tool, including an insulating rod, an end effector, and tool-side sensing components. The cloud platform management center is deployed in the operation and maintenance monitoring center to store historical defect databases, standard template libraries, defect and action mapping libraries, and defect elimination process data, and provides algorithm iteration training, operation early warning, and data visualization functions. The energy supply module supplies power to the sensing and acquisition module and the intelligent identification and decision-making module.

[0042] The sensing and acquisition module includes an infrared bullet camera, a binocular camera with a motion pan-tilt head, a tower vibration sensor, a micro-weather sensor, and a conductor galloping ball. The infrared bullet camera is used to acquire infrared images of the tower and conductor in real time, especially for acquiring infrared images of the fastener area to monitor for overheating defects; the binocular camera is used to acquire visible light images of the tower and fastener area in real time and to determine the location of the work point; the tower vibration sensor is used to acquire vibration data of fasteners or tower connections; the micro-weather sensor is used to acquire environmental parameters such as wind speed, humidity, and temperature; and the conductor galloping ball is used to monitor conductor displacement and prevent interference between the live fault removal tool and the conductor galloping.

[0043] The intelligent identification and decision-making module has two states: data monitoring and defect elimination. In data monitoring mode, the module receives data from infrared bullet cameras, binocular cameras, and sensors, and routinely detects bolt defects using a fine-grained bolt defect identification network. In defect elimination mode, the module receives data from live defect elimination tools and binocular camera data, activates target detection and distance judgment algorithms, and collaboratively completes the defect elimination operation.

[0044] The execution and defect elimination module includes an actuator, a sensing module, and an insulating rod. After receiving defect elimination control commands, the actuator controls the torque motor. The motor's end can be fitted with different tools, including an electric fastening head, an insulated cutting head, and a foreign object gripper. It can also be matched with an insulated rust removal head or a washer mounting head depending on the defect type. The tool-side sensing module includes an IMU attitude sensor and a small camera. The IMU attitude sensor has an accuracy of ±0.5°, and the small camera has at least 2 megapixels. The insulating rod is made of epoxy resin, with a length of at least 5m and a withstand voltage rating of at least 500kV, used to achieve voltage isolation between the operator and the workpiece. The energy supply module uses a 200W photovoltaic panel and a 12V / 100Ah lithium battery pack, supporting photovoltaic-energy storage complementary power supply, with a runtime of at least 72 hours in the absence of sunlight.

[0045] Example 4: See Figure 5This embodiment collects image data, vibration data, and bolt type data of transmission tower fasteners, and performs standardized preprocessing on the collected data. A data acquisition mechanism is simultaneously activated: infrared images of tower bolts are acquired using an infrared bullet camera to monitor for heating defects; visible light images are acquired by focusing on the bolt area using a binocular pan-tilt camera to capture structural defects; bolt vibration data is collected using a vibration sensor at a sampling frequency of 100Hz; environmental parameters such as wind speed, humidity, and temperature are recorded simultaneously using a micro-meteorological sensor; and conductor displacement is monitored using a conductor swivel ball to avoid interference with defect elimination operations. The image data, vibration data, and bolt type code data serve as input data for a fine-grained bolt defect identification network; environmental data and conductor motion data serve as auxiliary data for motion parameter correction, safety assessment, and operational interference assessment.

[0046] During the standardized preprocessing of the collected data, adaptive denoising and exposure compensation were performed on the image data, the region of interest for the bolts was cropped, and the image was compressed to a resolution of 640×480 using bilinear interpolation. Data augmentation processing with random flipping and ±15% brightness perturbation was then applied to the compressed image data. Kalman filtering was used to eliminate noise in the vibration data, which was then standardized to the 0-100 range. Gaussian noise with a variance no greater than 0.05 was added to improve the model's generalization ability. Bolt models were structured and encoded, ranging from M12 to M24. The preprocessed image data, vibration data, and bolt model encoding data were stored in a unified format in the edge database and backed up in real time to the cloud platform management center.

[0047] The preprocessed data is input into a fine-grained bolt defect recognition network deployed at the edge to identify the defect type, defect level, attitude parameters, and confidence information of the fasteners. The AI ​​analysis unit of the edge smart gateway activates the fine-grained bolt defect recognition network by inputting the preprocessed image data, vibration data, and bolt model code data into the network, and initiates parallel feature extraction and defect determination.

[0048] The fine-grained bolt defect recognition network consists of an input layer, a backbone feature extraction layer, a multi-branch recognition layer, a feature fusion layer, and an output and confidence correction layer. The backbone feature extraction layer is based on the MobileNetV4 architecture and extracts 60×45×128-dimensional global structural features of bolts through a 3×3 initial convolutional kernel and four concatenated depthwise separable convolutional blocks.

[0049] The multi-branch recognition layer includes a thread texture branch, a washer contour branch, and a bolt posture branch. The thread texture branch is adapted for thread corrosion recognition, and consists of two 3×3 small convolutional kernels, a BatchNorm layer, a global average pooling layer, and a fully connected layer. The small convolutional kernels extract the grayscale gradient features of the thread, which are then activated by Softmax to output a 1×1×2 feature vector containing the corrosion probability and corrosion level (slight, moderate, and severe). The washer contour branch is adapted for anti-loosening washer missingness recognition, and consists of a HOG feature extraction layer, a circular fitting layer, an IoU calculation layer, and a fully connected layer. The HOG feature extraction layer extracts the annular edge features of the washer, the circular fitting layer calculates the center deviation of the contour, and a valid washer is determined when the center deviation is no greater than 2 pixels. The IoU calculation layer compares the washer contour with a standard washer template, and a washer is determined to exist when the IoU is no less than 0.7. A 1×1×2 feature vector is output after Sigmoid activation, containing the washer presence probability and integrity probability. Bolt attitude branch adaptation for bolt loosening identification includes an adaptive threshold Canny edge detection layer, an attitude regression layer, and a physical parameter fusion layer. The adaptive threshold Canny edge detection layer extracts the gap features between the bolt head and the tower contact surface, and determines loosening when the gap is greater than 0.5mm. The attitude regression layer outputs the bolt tilt angle with an accuracy of ±0.1°. The physical parameter fusion layer fuses vibration data, and when the vibration amplitude is greater than 30, it corresponds to severe loosening, and outputs a 1×1×3 feature vector containing the loosening probability, loosening level, and tilt angle.

[0050] See Figure 6 The feature fusion layer employs a dynamic weighted summation method to fuse features from the thread texture branch, washer profile branch, and bolt attitude branch to eliminate ambiguity from single features. The weights of each branch are adaptively adjusted based on the confidence level of the vibration data and satisfy normalization constraints. After fusion, a BatchNorm layer and a Dropout layer are added. The BatchNorm layer has a momentum of 0.9 and an epsilon of 1e-5, while the Dropout layer has a dropout rate of 0.2 to prevent overfitting. The fusion output is a 1×1×7 fused feature vector, which integrates seven core features from the three branches.

[0051] See Figure 7 The output and confidence correction layer performs two-dimensional confidence correction using visual features and physical parameters, ultimately satisfying the following confidence level:

[0052] in, For the final confidence level, For visual recognition confidence, This is the vibration correction factor. When the vibration amplitude is greater than 30, When the vibration amplitude is greater than or equal to 10 and less than or equal to 30, When the vibration amplitude is less than 10, The final output includes four defect types: normal, loose bolt, thread corrosion, or missing anti-loosening washer; the defect level is output as minor, moderate, or severe; the bolt tilt angle is output; and the final confidence level is output. When the final confidence level is not less than 95%, the identification result is determined as a valid defect identification result.

[0053] Based on the fastener's defect type, defect level, attitude parameters, and confidence level information, and combined with defect-action mapping data, a defect elimination control command is generated to control the live-line defect elimination tool. The edge side uploads the valid defect identification results to the cloud platform management center. The cloud platform management center cross-validates the results using a historical defect database and a standard parameter library for the same type of tower, eliminating misidentifications caused by shadows or occlusions, and supplementing relevant information such as the standard torque of the bolt model and past defect elimination cases to generate a defect feature file. The control decision unit, based on the defect feature file, calls the pre-stored defect-action mapping library, matches the end effector according to the defect type, and matches the number of rotations, torque, and action amplitude according to the defect level and bolt model. It also presets the initial execution angle of the end effector according to the bolt tilt angle. When the defect type is bolt loosening, an electric tightening head is matched; when the defect type is thread corrosion, an insulated rust removal head is matched; when the defect type is missing anti-loosening washer, a washer installation head is matched. For example, for an M16 bolt, slight loosening corresponds to 1 turn clockwise and a torque of 40 N·m, while severe loosening corresponds to 2.5 turns clockwise and a torque of 55 N·m.

[0054] The control decision unit also adjusts the action parameters by combining wind speed data, humidity data, and IMU attitude data and voltage sensing data returned by the execution and troubleshooting modules. When the wind speed is greater than 5 m / s, the action amplitude is reduced by 30% and the torque is increased by 10%; when the humidity is greater than 85%, the metal contact actuator is disabled; and the operating angle of the insulating rod is adjusted according to the tool attitude and voltage sensing data to form the final troubleshooting control command.

[0055] Safety checks are performed on the voltage at the work point and the distance between the live-line troubleshooting tool and the live conductor. After receiving the troubleshooting control command, the execution and troubleshooting module detects the voltage at the work point using a voltage sensing probe and confirms whether there is a risk of voltage exceeding the limit based on the safety thresholds corresponding to voltage levels of 10kV-500kV. A small camera on the tool side captures close-up images of the work point, and the AI ​​analysis unit determines whether the distance between the end effector and the live conductor is greater than or equal to the safety threshold. If the voltage at the work point does not exceed the corresponding safety threshold, and the distance between the end effector and the live conductor is greater than or equal to the safety threshold, the safety check is considered passed.

[0056] After the safety verification is passed, the live-line fault-finding tool performs live fault-finding actions on the fastener according to the fault-finding control instructions. The transmission control module drives the insulating rod to extend and retract to the target position according to the instructions, and the end effector switches to the specified type and adjusts the angle to the preset value. During the execution, a real-time correction mechanism is activated, and the IMU attitude sensor transmits tool attitude data back every 10ms. When the position error is greater than 2cm or the angle deviation is greater than 0.5°, the control decision unit immediately issues a correction command. A small camera on the tool side monitors the contact status between the actuator and the bolt in real time and dynamically adjusts the action force. The tool attitude, action progress, and real-time image data during the execution process are transmitted back to the edge smart gateway in real time through the built-in communication module of the insulating rod, and are simultaneously uploaded to the cloud platform management center for archiving.

[0057] Image and vibration data of the fasteners are re-acquired, and the defect elimination effect is verified based on the fine-grained bolt defect recognition network. After the defect elimination action is completed, the execution and defect elimination module reverts to a safe position. The binocular pan-tilt camera re-acquires visible light images of the defect area, the infrared bullet camera re-acquires infrared images of the defect area, the vibration sensor re-acquires bolt vibration data, and the micro-meteorological sensor records the current environmental parameters for comparison with the data before defect elimination. The AI ​​analysis unit reuses the fine-grained bolt defect recognition network for effect verification. At the visual level, the image after defect elimination is compared pixel-level with the standard template of the same type of bolt stored in the cloud platform management center to determine whether the visual defect has been eliminated; at the physical parameter level, it is verified whether the vibration amplitude is no greater than 10 to determine whether the fastening is qualified, and combined with the infrared image to determine whether there are any abnormal heat points to determine whether the electrical connection is qualified. When the visual defect is eliminated, the vibration amplitude is no greater than 10, and there are no abnormal heat points in the infrared image, the defect elimination effect is determined to be qualified.

[0058] If the defect elimination effect verification fails to meet the standards, the defect elimination control command is regenerated based on the reasons for failure, and the live-line defect elimination tool is controlled to perform the defect elimination action again until the defect elimination effect verification meets the standards. The cloud platform management center receives the verification results and performs a second review. If the defect elimination effect verification meets the standards, a defect elimination operation report containing defect information, execution parameters, and effect data is generated and archived in the database. If the defect elimination effect verification fails to meet the standards, a second defect elimination process is triggered, returning to the defect elimination control command generation stage, and the control command is optimized based on the reasons for failure. Reasons for failure include insufficient action force and execution angle deviation. The control decision unit re-optimizes the defect elimination control command and repeats the safety verification, linkage execution, and effect verification process until the defect elimination effect verification meets the standards.

[0059] Example 5: This embodiment discloses a specific application for live-line troubleshooting of bolts on 110kV high-voltage transmission towers. The working environment includes high humidity and windy conditions, with humidity ranging from 75% to 90% and wind speeds ranging from 1 to 6 m / s. The tower bolts range from M16 to M20. The edge gateway is deployed on the middle crossarm of the tower, and the power supply uses a combination of photovoltaic and battery power. The sensing module is installed on the tower crossarm, enabling the camera to capture images of the transmission line and connectors. The edge gateway is fixed to the tower platform and enters daily monitoring mode.

[0060] After the system is started, the binocular camera captures images of the M16 bolt, the micro-weather sensor records wind speed of 4.2 m / s and humidity of 82%, and the vibration sensor collects vibration data. The AI ​​analysis unit crops the region of interest of the bolt in the image and compresses it to a resolution of 640×480. The vibration data is then subjected to Kalman filtering and standardization, resulting in a standardized vibration data value of 28 and a bolt model code of [0,1,0].

[0061] See Figure 4 BFR-Net receives preprocessed image data, vibration data, and bolt type code data. After processing through a main feature extraction layer, a multi-branch recognition layer, a feature fusion layer, and an output and confidence correction layer, the output recognition result is bolt looseness, a severe defect level, a bolt tilt angle of 0.7°, and a final confidence level of 96.8%. The cloud platform management center performs a secondary verification of the recognition result. After the verification passes, the binocular camera generates the three-dimensional coordinates of the bolt: X: 12.3m, Y: 5.6m, Z: 8.9m.

[0062] The execution and troubleshooting tools are carried to the work site by maintenance personnel and paired with the edge intelligent gateway for communication, activating the system into troubleshooting mode. Based on the defect characteristic file indicating loose bolts of a severe level, the control decision unit matches an electric tightening head with basic parameters of 2.5 clockwise rotations and a torque of 55 N·m. Considering a wind speed of 4.2 m / s, the adjustment reduces the amplitude of the movement by 25% and increases the torque by 8%. On-site personnel confirm the operation strategy via the control screen.

[0063] Before performing the defect elimination operation, a binocular camera monitors the distance between the tool's end and other tower components, and adjusts the execution angle of the live defect elimination tool. After the safety check is passed, the tightening operation is initiated. During the operation, an IMU attitude sensor corrects the attitude in real time, a small camera monitors the insertion status of the tightening head, and dynamically adjusts the torque according to the fit.

[0064] After the tightening operation is completed, the execution and defect elimination module reverts to a safe position, and the system re-acquires bolt images and vibration data. BFR-Net verifies the effect of the re-acquired data and determines the result to be normal. After a second observation and confirmation by on-site personnel, the cloud platform management center generates a defect elimination report and archives the defect information, execution parameters, and effect data. The system then re-enters routine inspection mode.

[0065] After this embodiment is implemented, it can promptly detect defects in transmission tower bolts and accurately identify the defect type and severity. It can also provide defect elimination solutions based on defect generation specifications, reducing the incompleteness of defect elimination due to reliance on personnel experience. At the same time, it reduces the number of times inspection personnel need to climb the pole for confirmation, shortens the time spent on each defect elimination operation, reduces the rate of repetitive work, and improves the efficiency of transmission line operation and maintenance and the safety of live defect elimination.

[0066] Example 6: See Figure 8 The present invention also provides an electronic device 100 that integrates edge intelligent decision-making and live fault elimination tools; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0067] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the linkage method for integrating edge intelligent decision-making and live fault elimination tools described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0068] The at least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor. The processor 102 is the control center of the electronic device 100, connecting various parts of the electronic device 100 via various interfaces and lines.

[0069] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for linking edge intelligent decision-making with live fault elimination tools, and the processor 102 can execute the multiple instructions to achieve the following: Collect image data, vibration data, and bolt type of fasteners on power transmission towers, and perform standardized preprocessing on the collected data; The preprocessed data is input into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, attitude parameters and confidence information of the fastener; Based on the fastener's defect type, defect level, attitude parameters, and confidence information, and combined with defect-action mapping data, a defect removal control command is generated to control the live defect removal tool. Safety checks were performed on the voltage at the work site and the distance between the live fault-finding tool and the live conductor. After the safety verification is passed, the live-line troubleshooting tool is controlled to perform live-line troubleshooting on the fastener according to the troubleshooting control command; Image and vibration data of the fasteners were re-acquired, and the defect elimination effect was verified based on a fine-grained bolt defect identification network. If the defect elimination effect fails to meet the standard, the defect elimination control command will be regenerated based on the reason for failure, and the live defect elimination tool will be controlled to perform the defect elimination action again until the defect elimination effect meets the standard.

[0070] Example 7: If the modules / units integrated in the electronic device 100 are 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, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0071] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for linking edge intelligence decision and live defect removal tool, characterized in that, Includes the following steps: Collect image data, vibration data, and bolt type of fasteners on power transmission towers, and perform standardized preprocessing on the collected data; The preprocessed data is input into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, attitude parameters and confidence information of the fastener; Based on the fastener's defect type, defect level, attitude parameters, and confidence information, and combined with defect-action mapping data, a defect removal control command is generated to control the live defect removal tool. Safety checks were performed on the voltage at the work site and the distance between the live fault-finding tool and the live conductor. After the safety verification is passed, the live-line troubleshooting tool is controlled to perform live-line troubleshooting on the fastener according to the troubleshooting control command; Image and vibration data of the fasteners were re-acquired, and the defect elimination effect was verified based on a fine-grained bolt defect identification network. If the defect elimination effect fails to meet the standard, the defect elimination control command will be regenerated based on the reason for failure, and the live defect elimination tool will be controlled to perform the defect elimination action again until the defect elimination effect meets the standard.

2. The method of claim 1, wherein the method further comprises: The specific method for collecting image data, vibration data, and bolt type of transmission tower fasteners, and performing standardized preprocessing on the collected data, is as follows: Collect image data, vibration data, environmental data, and conductor motion data of transmission tower fasteners; Adaptive denoising and exposure compensation are performed on the image data, the area required for the fastener is cropped, and the image data is compressed to the required resolution through bilinear interpolation. Data enhancement processing with random flipping and required brightness perturbation is then applied to the compressed image data. Kalman filtering is applied to the vibration data to eliminate noise, and the vibration data is standardized to the required range. The bolt type is structured and coded to obtain bolt type code data; The preprocessed image data, vibration data, and bolt type code data are stored in the edge database.

3. The method of claim 1, wherein the method further comprises: The preprocessed data is input into a fine-grained bolt defect recognition network deployed on the edge side. The specific method for recognizing fastener defect information, attitude parameters, and confidence information is as follows: The fine-grained identification network for bolt defects is activated by inputting the preprocessed data into the network. The global structural features of bolts are extracted through the backbone feature extraction layer of the fine-grained bolt defect identification network. The thread grayscale gradient features are extracted by the thread texture branch of the fine-grained bolt defect recognition network, and the corrosion probability and corrosion level are output. The annular edge features of the washer are extracted by the washer contour branch of the bolt defect fine-grained identification network, and the probability of the existence and integrity of the washer are output. The bolt posture branch of the bolt defect fine-grained identification network is used to extract the gap features, bolt tilt angle and vibration features between the bolt head and the tower contact surface, and output the loosening probability and loosening level. The features of each branch are dynamically weighted and fused. The final confidence level is determined by visual recognition confidence level and vibration correction coefficient. The output includes the defect type (normal, loose bolt, rusted thread or missing anti-loosening washer), the defect level (minor, moderate or severe), the bolt tilt angle, and the final confidence level, which serve as the defect information, attitude parameters and confidence level information of the fastener.

4. The method for linking edge intelligent decision-making with live fault elimination tools according to claim 1, characterized in that, Based on the fastener's defect type, defect level, attitude parameters, and confidence information, and combined with defect-motion mapping data, the specific method for generating defect removal control commands for controlling live-line defect removal tools is as follows: Obtain the defect type, defect level, attitude parameters and confidence information of the fastener, call the historical defect database and the standard parameter library of the same type of tower for cross-validation, supplement the standard torque and past defect elimination case information of the corresponding bolt type, and generate defect feature file; Based on the defect feature file, the defect and action mapping data is called. When the defect type is loose bolt, an electric fastening head is matched; when the defect type is thread corrosion, an insulating derusting head is matched; and when the defect type is missing anti-loosening washer, a washer installation head is matched. Match the number of rotations, torque, and range of motion according to the defect level and bolt type; The initial execution angle of the end effector is determined based on the bolt tilt angle, and the action parameters are corrected by combining wind speed and humidity data, which serve as the defect elimination control command.

5. The method for linking edge intelligent decision-making with live fault elimination tools according to claim 1, characterized in that, The specific methods for safety verification of the voltage at the work point and the distance between the live fault-finding tool and the live conductor are as follows: After receiving the troubleshooting control command, the live fault elimination tool detects the voltage at the work point through a voltage sensing probe and determines whether there is a risk of voltage exceeding the standard based on the safety threshold corresponding to the voltage level at the work point. Images of the work site are collected, and the distance between the end effector and the live conductor is determined to be greater than or equal to the safety threshold based on the images. When the voltage at the work site does not exceed the corresponding safety threshold, and the distance between the end effector and the live conductor is greater than or equal to the safety threshold, the safety verification is deemed to have passed.

6. The method for linking edge intelligent decision-making with live fault elimination tools according to claim 1, characterized in that, After the safety verification is passed, the specific method for controlling the live fault removal tool to perform live fault removal actions on the fastener according to the fault removal control command is as follows: After the safety verification is passed, the insulating rod of the live defect elimination tool is extended and retracted to the target position, and the end effector is switched to the type of effector corresponding to the defect type and adjusted to the execution angle determined by the defect elimination control command. During the defect elimination process, the attitude data of the live defect elimination tool is transmitted back in real time. When the position error exceeds the accuracy threshold, the live defect elimination tool is corrected. The real-time live fault removal tool monitors the fit between the end effector and the fastener, and dynamically adjusts the action force. The tool's posture, motion progress, and real-time image data during execution are archived.

7. The method for linking edge intelligent decision-making with live fault elimination tools according to claim 1, characterized in that, The specific method for verifying the defect elimination effect based on a fine-grained bolt defect identification network, after re-acquiring image and vibration data of the fasteners, is as follows: After the live fault elimination operation is completed, image and vibration data of the fasteners are reacquired. The fine-grained bolt defect recognition network verifies the effectiveness of the re-acquired image and vibration data, compares the defect-removed image data with the stored standard template of the same bolt model to determine whether the visual defects have been eliminated, and judges whether the fastener connection status and heating status are qualified based on the defect-removed vibration and image data. When visual defects are eliminated and the fastener connection status and heating status are qualified, the defect elimination effect is deemed to have passed the verification.

8. The method for linking edge intelligent decision-making with live fault elimination tools according to claim 1, characterized in that, If the defect elimination effect verification fails to meet the standard, the defect elimination control command will be regenerated based on the reason for failure, and the live defect elimination tool will be controlled to perform the defect elimination action again until the defect elimination effect verification meets the standard. The specific method is as follows: When the defect elimination effect verification result is not up to standard, feedback information is generated based on the reason for not meeting the standard; Based on the feedback information, the defect elimination control instructions were re-optimized, and the safety verification, live defect elimination actions, and defect elimination effect verification were re-executed. When the defect elimination effect is verified to be up to standard, a defect elimination work report containing defect information, execution parameters and effect data is generated and archived.

9. A linkage system integrating edge intelligent decision-making and live fault elimination tools, characterized in that, include: The preprocessing module is used to collect image data, vibration data, and bolt type of fasteners on transmission towers, and to perform standardized preprocessing on the collected data; The network recognition module is used to input preprocessed data into a fine-grained bolt defect recognition network deployed on the edge side to identify the defect type, defect level, posture parameters and confidence information of the fastener; The instruction generation module is used to generate defect removal control instructions for controlling live defect removal tools based on the defect type, defect level, attitude parameters and confidence information of the fastener, combined with defect and motion mapping data. The safety verification module is used to perform safety verification on the voltage at the work point and the distance between the live fault removal tool and the live conductor; The execution module is used to control the live-line repair tool to perform live-line repair actions on the fasteners according to the repair control instructions after the safety verification is passed; The effect verification module is used to re-acquire image and vibration data of fasteners and verify the defect elimination effect based on the fine-grained bolt defect recognition network. The instruction optimization module is used to regenerate the defect elimination control instruction based on the reason for failure if the defect elimination effect verification fails, and then control the live defect elimination tool to perform the defect elimination action again until the defect elimination effect verification meets the standard.

10. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The preprocessing module's functionality is implemented using the following methods: Collect image data, vibration data, environmental data, and conductor motion data of transmission tower fasteners; Adaptive denoising and exposure compensation are performed on the image data, the area required for the fastener is cropped, and the image data is compressed to the required resolution through bilinear interpolation. Data enhancement processing with random flipping and required brightness perturbation is then applied to the compressed image data. Kalman filtering is applied to the vibration data to eliminate noise, and the vibration data is standardized to the required range. The bolt type is structured and coded to obtain bolt type code data; The preprocessed image data, vibration data, and bolt type code data are stored in the edge database.

11. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The functionality of the network identification module is achieved through the following methods: The fine-grained identification network for bolt defects is activated by inputting the preprocessed data into the network. The global structural features of bolts are extracted through the backbone feature extraction layer of the fine-grained bolt defect identification network. The thread grayscale gradient features are extracted by the thread texture branch of the fine-grained bolt defect recognition network, and the corrosion probability and corrosion level are output. The annular edge features of the washer are extracted by the washer contour branch of the bolt defect fine-grained identification network, and the probability of the existence and integrity of the washer are output. The bolt posture branch of the bolt defect fine-grained identification network is used to extract the gap features, bolt tilt angle and vibration features between the bolt head and the tower contact surface, and output the loosening probability and loosening level. The features of each branch are dynamically weighted and fused. The final confidence level is determined by visual recognition confidence level and vibration correction coefficient. The output includes the defect type (normal, loose bolt, rusted thread or missing anti-loosening washer), the defect level (minor, moderate or severe), the bolt tilt angle, and the final confidence level, which serve as the defect information, attitude parameters and confidence level information of the fastener.

12. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The functionality of the instruction generation module is implemented through the following methods: Obtain the defect type, defect level, attitude parameters and confidence information of the fastener, call the historical defect database and the standard parameter library of the same type of tower for cross-validation, supplement the standard torque and past defect elimination case information of the corresponding bolt type, and generate defect feature file; Based on the defect feature file, the defect and action mapping data is called. When the defect type is loose bolt, an electric fastening head is matched; when the defect type is thread corrosion, an insulating derusting head is matched; and when the defect type is missing anti-loosening washer, a washer installation head is matched. Match the number of rotations, torque, and range of motion according to the defect level and bolt type; The initial execution angle of the end effector is determined based on the bolt tilt angle, and the action parameters are corrected by combining wind speed and humidity data, which serve as the defect elimination control command.

13. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The security verification module's functionality is implemented through the following methods: After receiving the troubleshooting control command, the live fault elimination tool detects the voltage at the work point through a voltage sensing probe and determines whether there is a risk of voltage exceeding the standard based on the safety threshold corresponding to the voltage level at the work point. Images of the work site are collected, and the distance between the end effector and the live conductor is determined to be greater than or equal to the safety threshold based on the images. When the voltage at the work site does not exceed the corresponding safety threshold, and the distance between the end effector and the live conductor is greater than or equal to the safety threshold, the safety verification is deemed to have passed.

14. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The functionality of the execution module is implemented through the following methods: After the safety verification is passed, the insulating rod of the live defect elimination tool is extended and retracted to the target position, and the end effector is switched to the type of effector corresponding to the defect type and adjusted to the execution angle determined by the defect elimination control command. During the defect elimination process, the attitude data of the live defect elimination tool is transmitted back in real time. When the position error exceeds the accuracy threshold, the live defect elimination tool is corrected. The real-time live fault removal tool monitors the fit between the end effector and the fastener, and dynamically adjusts the action force. The tool's posture, motion progress, and real-time image data during execution are archived.

15. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The functionality of the effect verification module is implemented through the following methods: After the live fault elimination operation is completed, image and vibration data of the fasteners are reacquired. The fine-grained bolt defect recognition network verifies the effectiveness of the re-acquired image and vibration data, compares the defect-removed image data with the stored standard template of the same bolt model to determine whether the visual defects have been eliminated, and judges whether the fastener connection status and heating status are qualified based on the defect-removed vibration and image data. When visual defects are eliminated and the fastener connection status and heating status are qualified, the defect elimination effect is deemed to have passed the verification.

16. The linkage system for integrating edge intelligent decision-making and live fault elimination tools according to claim 9, characterized in that, The instruction optimization module's functionality is implemented through the following methods: When the defect elimination effect verification result is not up to standard, feedback information is generated based on the reason for not meeting the standard; Based on the feedback information, the defect elimination control instructions were re-optimized, and the safety verification, live defect elimination actions, and defect elimination effect verification were re-executed. When the defect elimination effect is verified to be up to standard, a defect elimination work report containing defect information, execution parameters and effect data is generated and archived.

17. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the linkage method for integrating edge intelligent decision-making and live fault elimination tools as described in any one of claims 1 to 8.

18. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the linkage method for integrating edge intelligent decision-making and live fault elimination tools as described in any one of claims 1 to 8.

Images