Power transmission tower inspection path planning method and device, equipment and storage medium

Abstract

The application discloses a power transmission tower inspection path planning method and device, equipment and a storage medium, relates to the technical field of electric power, and comprises the following steps: based on the environment data collected by a unmanned aerial vehicle, driving a digital twin model to analyze the structural response of the environment data, and generating a structural risk heat map; based on the structural risk heat map, a path is constructed through a weighted coverage path planning algorithm to generate an initial inspection path; the unmanned aerial vehicle is controlled to detect structural defects based on the initial inspection path, and when a target anomaly is detected, a local path re-planning is triggered to generate a target inspection path corresponding to the target anomaly; based on the target inspection path, inspection is performed to obtain inspection data, and the digital twin model is updated based on the inspection data to obtain an updated digital twin model. Through the construction of the digital twin model and the updating of the model based on the inspection data, the updated model generates the next round of inspection route, and the accuracy of the power transmission tower structural safety inspection is improved.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus, equipment and storage medium for planning inspection routes of transmission towers. Background Technology

[0002] As the core load-bearing components of high-voltage transmission lines, transmission towers are subjected to the combined effects of wind loads, icing loads, and temperature loads over long periods, making them prone to structural defects such as corrosion, node loosening, and localized deformation. Regular inspections are crucial for ensuring the safe operation of the power grid. With the development of unmanned aerial vehicle (UAV) technology, its mobility, flexibility, and low cost have led to its widespread application in transmission line inspection. Currently, UAV inspection path planning methods are mainly based on geometric coverage or terrain perception principles. Existing methods achieve inspection coverage by generating flight paths parallel and equidistant from the cables, or by using directed bounding box calculations on laser point clouds to extract suspension point coordinates and generate inspection tracks. These methods treat the generation of inspection paths as a purely geometric problem, considering only the uniformity of spatial coverage and obstacle avoidance safety, without incorporating the structural mechanical state of the transmission tower into the path planning decision-making process. The shortcomings of existing technologies are that the generation of inspection paths does not include structural mechanics information, which causes the flight time and inspection resources of drones to be evenly distributed across all tower components, rather than tilted towards high-risk components with high stress concentration and excessive deformation in actual service. This results in the risk of missing inspections of high-risk components within a limited inspection time, making it difficult to meet the needs of refined management and control of transmission tower structural safety.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a method, device, equipment and storage medium for planning inspection routes of transmission towers, aiming to solve the technical problem of insufficient accuracy in the safety management of transmission tower structures.

[0005] To achieve the above objectives, this application proposes a method for planning inspection routes for transmission towers, the method comprising: Based on environmental data collected by drones, a digital twin model is driven to perform structural response analysis on the environmental data, generating a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. Based on the structural risk heat map, a weighted coverage path planning algorithm is used to construct the initial inspection path for the transmission tower. The drone is controlled to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, a local path replanning corresponding to the target anomaly is triggered to generate a target inspection path. The drone is controlled to perform inspections based on the target inspection path to obtain inspection data of the transmission tower. The digital twin model is then updated based on the inspection data to obtain an updated digital twin model.

[0006] In one embodiment, the step of using environmental data collected by a drone to drive a digital twin model to perform structural response analysis on the environmental data and generate a structural risk heat map of the transmission tower includes: The drone acquires real-time environmental data of the power transmission tower, and constructs a combined load vector based on the environmental data. The combined load vector is input into the digital twin model for nonlinear solution, and the stress ratio of each component is extracted from the solution results. The stress ratios are classified according to preset hazard level thresholds to generate a structural risk heat map covering the entire tower.

[0007] In one embodiment, the environmental data includes wind speed data, wind direction data, and temperature data, and the step of constructing a combined load vector based on the environmental data includes: Based on the shape coefficient and wind-receiving area of ​​each component in the digital twin model, the wind speed data and the wind direction data are converted to obtain the wind load vector; Based on the temperature data, the axial temperature deformation of each component in the digital twin model caused by the temperature difference is calculated to obtain the temperature load vector; The combined load vector is obtained by superimposing the wind load vector, the temperature load vector, and the structural self-weight load vector.

[0008] In one embodiment, the step of generating the initial inspection path for the transmission tower by constructing a path using a weighted coverage path planning algorithm based on the structural risk heatmap includes: The risk level of each component is extracted from the structural risk heat map. Based on the risk level, a corresponding path planning weight coefficient is determined for each component. The risk level is divided into high-risk, medium-risk and low-risk levels according to the relationship between the component stress ratio and a preset threshold. The initial inspection path of the transmission tower is obtained by weighting the path planning weight coefficients using the weighted shortest coverage path algorithm.

[0009] In one embodiment, the step of triggering local path replanning corresponding to the target anomaly and generating a target inspection path includes: Obtain the defect location information of the target anomaly, and generate a local reshoot path based on the defect location information within a preset range around the target anomaly; The initial inspection path that has not yet been executed is updated based on the local reshoot path to obtain the target inspection path.

[0010] In one embodiment, the step of updating the digital twin model based on the inspection data to obtain the updated digital twin model includes: Point cloud data and image data are extracted from the inspection data. The point cloud data is registered and fused to extract the measured displacement of key nodes. The image data is reconstructed in three dimensions to extract the cross-sectional change characteristics of the components. Using the measured displacement of the key node and the characteristic quantity of the component cross-section change as observation values, the Bayesian data assimilation method is used to perform posterior estimation and parameter correction on the current state vector of the digital twin model, resulting in a digital twin model with corrected model parameters. The model prediction value is calculated based on the digital twin model after the model parameters are corrected. The residual between the model prediction value and the measured value is calculated. When the residual is less than a preset threshold, the model parameters are corrected to obtain the updated digital twin model.

[0011] In one embodiment, the step of generating the initial inspection path of the transmission tower by constructing a path using a weighted coverage path planning algorithm based on the structural risk heat map further includes: When the number of transmission towers is greater than one, the comprehensive health parameters of each transmission tower are obtained. The comprehensive health parameters include the maximum stress ratio of components, the ratio of the maximum displacement of nodes to the design limit, the normalized value of service life, and the normalized value of the historical cumulative number of defects. A weighted calculation is performed based on the comprehensive health parameters to obtain a comprehensive health score for each transmission tower. The comprehensive health score is used to determine the inspection priority of each transmission tower.

[0012] Furthermore, to achieve the above objectives, this application also proposes a transmission tower inspection path planning device, which includes: The risk heat map generation module is used to drive a digital twin model to perform structural response analysis on environmental data collected by UAVs and generate a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. The initial path construction module is used to construct the initial inspection path of the transmission tower based on the structural risk heat map using a weighted coverage path planning algorithm. The target path generation module is used to control the UAV to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, it triggers local path replanning corresponding to the target anomaly to generate a target inspection path. The model parameter update module is used to control the UAV to perform inspections based on the target inspection path, obtain inspection data of the transmission tower, and update the digital twin model based on the inspection data to obtain the updated digital twin model.

[0013] In addition, to achieve the above objectives, this application also proposes a transmission tower inspection path planning device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the transmission tower inspection path planning method as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the transmission tower inspection path planning method described above.

[0015] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the transmission tower inspection path planning method described above.

[0016] One or more technical solutions proposed in this application have at least the following technical effects: This application proposes a method, apparatus, equipment, and storage medium for planning inspection paths for power transmission towers. Based on environmental data collected by a drone, a digital twin model is driven to perform structural response analysis on the environmental data, generating a structural risk heatmap of the power transmission tower. The digital twin model is generated based on the structural information of the power transmission tower. Based on the structural risk heatmap, a weighted coverage path planning algorithm is used to construct a path, generating an initial inspection path for the power transmission tower. The drone is controlled to detect structural defects based on the initial inspection path. When a target anomaly is detected, local path replanning corresponding to the target anomaly is triggered, generating a target inspection path. The drone is controlled to inspect the power transmission tower based on the target inspection path, obtaining inspection data. The digital twin model is updated based on the inspection data, resulting in an updated digital twin model. This application improves the accuracy of power transmission tower structural safety inspections by constructing a digital twin model and updating the model based on inspection data. The updated model generates the next round of inspection routes. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the first embodiment of the transmission tower inspection path planning method of this application; Figure 2 This is a schematic diagram of the module structure of the transmission tower inspection path planning device according to an embodiment of this application; Figure 3 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the transmission tower inspection path planning method in the embodiments of this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] The main solution of this application embodiment is as follows: Based on environmental data collected by a UAV, a digital twin model is driven to perform structural response analysis on the environmental data to generate a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. Based on the structural risk heat map, a weighted coverage path planning algorithm is used to construct a path and generate an initial inspection path for the transmission tower. The UAV is controlled to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, a local path replanning corresponding to the target anomaly is triggered to generate a target inspection path. The UAV is controlled to perform inspection based on the target inspection path to obtain inspection data of the transmission tower. The digital twin model is updated based on the inspection data to obtain an updated digital twin model.

[0024] In this embodiment, for ease of description, the following description will focus on the inspection path planning device of the transmission tower.

[0025] As the core load-bearing components of high-voltage transmission lines, transmission towers are subjected to the combined effects of wind loads, icing loads, and temperature loads over long periods, making them prone to structural defects such as corrosion, node loosening, and localized deformation. Regular inspections are crucial for ensuring the safe operation of the power grid. With the development of unmanned aerial vehicle (UAV) technology, its mobility, flexibility, and low cost have led to its widespread application in transmission line inspection. Currently, UAV inspection path planning methods are mainly based on geometric coverage or terrain perception principles. Existing methods achieve inspection coverage by generating flight paths parallel and equidistant from the cables, or by using directed bounding box calculations on laser point clouds to extract suspension point coordinates and generate inspection tracks. These methods treat the generation of inspection paths as a purely geometric problem, considering only the uniformity of spatial coverage and obstacle avoidance safety, without incorporating the structural mechanical state of the transmission tower into the path planning decision-making process. The shortcomings of existing technologies are that the generation of inspection paths does not include structural mechanics information, which causes the flight time and inspection resources of drones to be evenly distributed across all tower components, rather than tilted towards high-risk components with high stress concentration and excessive deformation in actual service. This results in the risk of missing inspections of high-risk components within a limited inspection time, making it difficult to meet the needs of refined management and control of transmission tower structural safety.

[0026] This application provides a solution that improves the accuracy of safety inspections of transmission tower structures by constructing a digital twin model and updating the model based on inspection data. The updated model generates the next round of inspection routes.

[0027] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a transmission tower inspection path planning device. The following description uses a transmission tower inspection path planning device as an example to illustrate this embodiment and the subsequent embodiments.

[0028] Based on this, embodiments of this application provide a method for planning inspection paths for transmission towers, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the transmission tower inspection path planning method of this application.

[0029] In this embodiment, the inspection path planning method for the transmission tower includes steps S11 to S14: Step S11: Based on the environmental data collected by the UAV, drive the digital twin model to perform structural response analysis on the environmental data and generate a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower.

[0030] It should be noted that environmental data specifically refers to meteorological data such as wind speed, wind direction, and temperature collected in real time by drones flying around the transmission tower. The digital twin model is a virtual mapping of the transmission tower entity; in this embodiment, it specifically refers to a parametric finite element model constructed based on the NIDA (Nonlinear Integrated Design and Analysis) method. Its initial state is generated based on the transmission tower's design drawings and initial measurement data (such as node coordinates, component cross-sectional properties, material parameters, and initial defects). The structural risk heatmap is a visual representation that uses color coding (e.g., red for high risk, yellow for medium risk, and blue for low risk) to intuitively show the degree of danger of each component on the transmission tower under the current environmental loads (wind, temperature). Its generation is based on the component stress ratio and node displacement calculated by the digital twin model.

[0031] Understandably, the purpose of this step is to combine the environmental loads of the physical world with the computational capabilities of the virtual model to achieve real-time perception of the structural state. The drone, acting as a mobile sensor platform, collects real-time meteorological data, which is synchronously input into a pre-built digital twin model. The model uses the NIDA method (an advanced finite element analysis method that simultaneously considers geometric and material nonlinearities) to perform high-precision mechanical solutions, calculating the stress level of each component and the displacement of each node. By comparing and classifying the calculation results with preset thresholds, a heat map reflecting the structural health and risk distribution is ultimately generated. This step realizes the transformation from physical environment to structural state perception, providing a scientific basis for subsequent intelligent path decision-making.

[0032] Specifically, firstly, a drone equipped with meteorological sensors (such as anemometers, wind vanes, and temperature sensors) collects wind speed, wind direction, and temperature data in real time around the tower along a prescribed route (such as circling flight), and transmits this data back to the ground control station via a wireless link. Upon receiving the data, the ground control station converts the wind speed and direction data into nodal wind load vectors based on the component shape coefficient and windward area, calculates the temperature data as an equivalent load vector for temperature deformation, and then superimposes this vector with the constant structural self-weight load vector to form a combined load vector for the current operating condition. This combined load vector is then input into the pre-built digital twin model (NIDA model) of the transmission tower. The model is solved using a nonlinear solver (such as the Newton-Raphson iterative method) to obtain the internal forces (axial forces, bending moments) and nodal displacements of each component. Finally, all components are classified according to preset hazard level thresholds, and a structural risk heat map covering all components of the entire tower is generated.

[0033] For example, during an inspection mission, the UAV measured an instantaneous wind speed of 15 m / s at the top of the tower, with the wind direction perpendicular to the guide wire, and an ambient temperature of 5°C. The ground system processed the wind speed and direction data into wind loads acting on various nodes of the tower, and combined them with temperature loads, inputting them into a digital twin model for calculation. The solution results showed that the stress ratio of the tower's inclined member A reached 0.85, far higher than other components. Based on this, the system marked component A in red (high risk) on the heat map, while components with stress ratios less than 0.5 were marked in blue (low risk), thus generating a clear risk distribution map.

[0034] Step S12: Based on the structural risk heat map, a weighted coverage path planning algorithm is used to construct the path and generate the initial inspection path for the transmission tower.

[0035] It should be noted that the weighted coverage path planning algorithm is an optimization algorithm designed to plan a flight path for a drone that covers (i.e., flies over and photographs) all target points that need to be detected (in this case, detection points on the transmission tower components). Unlike conventional coverage path planning (such as equal-spacing coverage), the weighted characteristic of this algorithm is reflected in assigning different weights to target points. These weights are derived from the hazard level corresponding to each component in the structural risk heatmap. Detection points of high-risk components are assigned higher weights, meaning that when planning the path, the algorithm will tend to prioritize visiting these high-risk points, or ensure that these high-risk points are more fully covered within the limited flight time.

[0036] Understandably, the purpose of this step is to address the problem that traditional equidistant path planning fails to prioritize high-risk components, leading to a high risk of missed inspections of these components within a limited inspection time. The principle is to transform the path planning problem into a weighted coverage path planning problem. The structural risk heatmap assigns a weight (hazard level) representing the importance of each detection point. The weighted shortest coverage path algorithm, when planning the path, aims to minimize the total flight distance (or time) while satisfying the weighted coverage constraints for all detection points. The algorithm optimizes the UAV's flight sequence, ensuring that high-weighted points (high-risk components) are visited earlier or at better locations along the path. This maximizes the detection priority and coverage quality of high-risk components under hard constraints such as battery life and safe distance. Essentially, this achieves on-demand allocation and intelligent prioritization of inspection resources.

[0037] Specifically, the structural risk heatmap is read, and the hazard level (e.g., high-risk, medium-risk, low-risk) of each component in the map is mapped to different path planning weight coefficients (e.g., high-risk weight = 3, medium-risk weight = 2, low-risk weight = 1). Next, the coordinates and corresponding weights of all the inspection points on the transmission tower that need to be photographed (usually located at component connection nodes or in the middle of the component) are input into the algorithm. When planning the path, the algorithm uses minimizing the total flight cost (usually distance and time) as the objective function, while being constrained by the UAV's physical constraints (e.g., minimum turning radius, maximum flight speed) and environmental constraints (e.g., electromagnetic safety distance from the conductor, minimum obstacle avoidance distance from the tower). A feasible implementation is to use an improved genetic algorithm or ant colony algorithm to solve this constrained optimization problem. The planning output is an ordered sequence containing a series of spatial waypoints, i.e., the initial inspection path, which the UAV will follow sequentially to each inspection point for photographing.

[0038] For example, suppose there are 100 detection points on the tower, of which 5 are high-risk components (red), 20 are medium-risk components (yellow), and the rest are low-risk. During planning, the algorithm will prioritize placing these 5 high-risk points at the beginning of the path and may plan closer shooting distances or more shooting angles for these points to ensure that even if the inspection is interrupted for any reason, the high-risk components have been inspected. The final generated path will show a clear tendency to cluster towards the red area in three-dimensional space.

[0039] Step S13: Control the UAV to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, trigger local path replanning corresponding to the target anomaly to generate a target inspection path.

[0040] It should be noted that structural defect detection refers to the process by which an UAV, during its inspection flight, uses its onboard computing unit or ground system to process images or video streams captured by its onboard camera in real time. This process automatically identifies typical defects on the surface of transmission tower components by running a pre-trained defect detection model (such as a deep learning model based on convolutional neural networks). Target anomalies are the output of the defect detection model, specifically referring to structural problems such as cracks, corrosion, loose bolts, and missing components identified by the model with high confidence. Local path replanning is a dynamic response mechanism. When the UAV is executing a preset initial inspection path, if a target anomaly is detected online, the system will interrupt the inspection of subsequent preset points and immediately plan a new, local supplementary flight path for the anomaly location and its surrounding area. This path is used for more detailed, multi-angle, and intensive imaging of the anomaly area. Upon triggering, a local supplementary imaging path with a preset radius is generated around the detection point of the target anomaly.

[0041] Understandably, the purpose of this step is to address the issues of untimely response and insufficient data collection when unexpected defects are discovered during inspections. The principle is to introduce a closed loop of online perception and dynamic decision-making. The drone flies strictly along a path based on structural risk prediction and performs online defect detection. When the detection model detects an anomaly, it means a real defect has appeared that the initial risk assessment (heatmap) may not have foreseen or that requires close attention. The system responds immediately, triggering local path replanning. The new local path (re-shooting path) aims to generate a small coverage area around the anomaly point, guiding the drone to take high-resolution re-shoots of the defect from multiple perspectives to obtain richer information on the defect's size, shape, and location. This is equivalent to dynamically inserting micro-level, actual-discovery-based emergency detection tasks into the execution of a macro-level, prediction-based inspection plan, thereby improving the completeness and accuracy of defect data collection.

[0042] Specifically, the UAV flies along the initial inspection path, its onboard camera continuously acquiring images. Image data can be downloaded to the ground system in real time, or the onboard computing unit can run a lightweight defect detection model for real-time inference. Once the detection model identifies a target defect (such as corrosion) in a single frame or multiple consecutive frames, and the confidence level exceeds a preset threshold (e.g., 0.9), an anomaly event is triggered. After triggering, the system first records the 3D coordinates of the anomaly point (obtained through the fusion of UAV positioning and visual positioning). Then, the local replanning unit in the path planning module is activated. This unit generates a spherical or cylindrical local space with a radius of R (e.g., 3 meters) centered on the anomaly point. Within this space, a new set of waypoints is planned to ensure that the UAV can perform intensive imaging of the anomaly component from different angles (e.g., frontal, side, and overhead views). After planning, the UAV pauses execution of subsequent waypoints on the original path and instead executes this local re-imaging path. After re-imaging, the UAV automatically flies back to the waypoint interrupted before the trigger, or directly flies to the next unexecuted waypoint on the original path to continue completing the remaining inspection tasks. Images and anomaly information (location, type, time) collected during the reshoot process are simultaneously pushed to the model update module.

[0043] For example, when a drone was inspecting a crossarm, the onboard model detected a suspected crack on the edge of a connecting plate in real time. This triggered a local replanning process; the drone immediately hovered and automatically planned a small, figure-eight-shaped flight path around the crack, taking three high-resolution close-up photos from directly in front of the crack, at a 45-degree angle to the left, and at a 45-degree angle to the right. Afterward, the drone returned to its original path and continued inspecting the next component.

[0044] Step S14: Control the UAV to perform inspections based on the target inspection path to obtain inspection data of the transmission tower, and update the digital twin model based on the inspection data to obtain the updated digital twin model.

[0045] It should be noted that the target inspection path is the generated final flight path, which integrates the initial global path and all triggered local re-shooting paths. Inspection data refers to the collection of all data gathered by the UAV after completing its flight along the target inspection path. It mainly includes two parts: first, high-precision point cloud data of the tower body obtained through LiDAR or visual SLAM (Simultaneous Localization and Mapping) technology, used to describe the structure's geometry; and second, image / video data captured by visible light, infrared, or high-definition cameras, used to record surface conditions and identified defects. Updating the digital twin model is a data assimilation process. Its core is to use the measured data (observations) obtained from the inspection to correct parameters in the digital twin model that may have deviated from the true state, such as the elastic modulus of the material (reflecting aging), the initial geometric defects of components (reflecting deformation), and the stiffness of node connections (reflecting loosening), so that the model can more accurately reflect the current true mechanical state of the transmission tower.

[0046] Understandably, the purpose of this step is to address the problem of digital twin models gradually deviating from their actual state due to a lack of real-world data feedback after long-term operation, thus enabling model self-calibration and evolution. A feedback loop is established to drive model updates based on inspection data. UAVs, as mobile measurement platforms, collect point cloud and image data containing information about the actual geometric shape and apparent state of the transmission tower. Through data processing (such as point cloud registration and 3D reconstruction), quantifiable features (such as the actual displacement of key nodes and the actual dimensions of component cross-sections) can be extracted from this data. These features are considered as observations of the digital twin model's output. Using a Bayesian data assimilation method, the uncertainty of model prediction (prior distribution) is combined with the uncertainty of observation data (observation error) to calculate the optimal estimate of model parameters (posterior distribution), thereby correcting the model parameters. This process enables the digital twin model itself to be optimized and updated using the results of a single inspection, allowing it to form a continuous closed loop of perception-planning-detection-updating in the next inspection prediction, based on a more accurate model state.

[0047] Specifically, after the inspection, the system processes all inspection data. First, it registers and fuses multi-station laser point cloud data to reconstruct a high-precision 3D model of the transmission tower. By comparing this model with the baseline geometry of the digital twin model, it extracts the measured displacements of key nodes under load. Simultaneously, it performs 3D reconstruction or measurement on high-resolution images to extract the cross-sectional dimensional change features of key components (such as main members and diagonal members). These measured displacements and dimensional changes are used as observations. Then, a Bayesian filtering method (such as Ensemble Kalman Filtering, EnKF) is used to update the state vector of the digital twin model. The state vector contains all parameters to be corrected, such as the degradation coefficient of the elastic modulus of each component, the initial defect vector, and the node connection stiffness. The filtering algorithm uses the model's predicted values ​​as priors and the measured values ​​as observations to calculate the posterior estimates after parameter correction. Finally, the system calculates the predicted residuals of the updated model to the measured values. If the residual is less than the preset confidence threshold, the parameter update is automatically accepted and the digital twin model is refreshed; if the residual is too large, it is considered that the observed data may be abnormal or the model has been severely distorted, and an alarm is generated to prompt manual review. The update record will be stored in the historical version repository.

[0048] For example, after an inspection, point cloud data extraction showed that the measured displacement of a certain node was 10% larger than the predicted value of the digital twin model. Based on this, the Bayesian data assimilation process adjusted the elastic modulus parameter (slightly lowering it to simulate material aging) and connection stiffness parameter of the components near that node. After the update, the model's displacement prediction for that node was closer to the measured value. Simultaneously, the image identified corrosion at a certain location, and the system may fine-tune the cross-sectional area parameter of that component in the model (simulating cross-sectional loss). After this update, the digital twin model can more realistically reflect the current state of the transmission tower.

[0049] This embodiment, through the aforementioned scheme, firstly, uses real-time environmental data collected by drones to drive the analysis of a digital twin model, generating a heatmap reflecting the real-time risk level of components, thus realizing the mapping from environmental physical quantities to structural risk perception. Next, based on this heatmap, weighted path planning is performed, ensuring that the generated inspection path prioritizes high-risk areas. This significantly improves the detection coverage and priority of high-risk components with limited inspection resources, effectively reducing the risk of safety accidents caused by missed inspections. Then, an online defect detection and dynamic replanning mechanism is introduced during inspection execution, enabling timely and detailed re-photographing of suddenly discovered defects, ensuring the completeness of defect data collection. Finally, the complete inspection data is used to update the digital twin model itself, allowing the model to evolve synchronously with the actual state degradation of the transmission tower, maintaining the fidelity of the digital twin. This series of steps forms a two-way closed loop of state perception-driven path planning and inspection data-driven model updates, solving the core problems of lack of mechanical basis in path planning and lagging digital twin model updates in traditional methods, thereby improving the intelligence level and safety of transmission tower operation and maintenance.

[0050] Based on the above implementation scheme, in one feasible implementation, the step of driving a digital twin model to perform structural response analysis on the environmental data collected by the UAV and generating a structural risk heat map of the transmission tower includes S21~S23: Step S21: The UAV acquires environmental data of the transmission tower in real time, and constructs a combined load vector based on the environmental data.

[0051] It should be noted that the combined load vector is a mathematical vector that represents the resultant effect of all external loads acting simultaneously on the transmission tower structure. In structural finite element analysis, loads are usually represented as vectors, where each element corresponds to a structural degree of freedom (such as force or bending moment in the x, y, and z directions of a node). Constructing the combined load vector involves superimposing loads of different types and locations into a total load input according to mechanical principles.

[0052] Understandably, the purpose of this step is to convert the discrete, multi-dimensional environmental perception data collected by the UAV into standard load inputs that can be used for structural mechanics calculations. This is based on the principles of load equivalence and superposition in structural mechanics. Transmission towers, in actual service, are subjected to the combined effects of multiple loads, including wind, temperature, and self-weight. The wind speed, wind direction, and temperature collected in real-time by the UAV are point-like or small-scale measurements. First, these measurements are converted, according to structural design specifications (such as wind load calculation specifications) and thermodynamic principles, into equivalent concentrated forces (equivalent forces caused by wind load and temperature deformation) acting on each node of the structural finite element model. These forces are then algebraically added to the constant structural self-weight load vector.

[0053] Specifically, the drone flies around the tower (e.g., at three different altitude levels: upper, middle, and lower), and the meteorological sensors on it collect and transmit wind speed, wind direction, and temperature data at a certain frequency (e.g., 1Hz).

[0054] After receiving these data streams, the ground system first performs spatial interpolation or averaging on the wind speed and direction data to obtain a representative environmental wind speed V and wind direction angle α for the tower. Based on the shape coefficient μs of each component of the transmission tower (obtained from design drawings or specifications) and the wind-receiving area A (the projected area of ​​the component on a plane perpendicular to the wind direction), the equivalent concentrated wind load acting on the nodes at both ends of the component is calculated according to the wind load calculation formula (e.g., Fw=0.5*ρ*V²*μs*A, where ρ is the air density). The wind loads of all components are assembled into a global wind load vector {Fw} by nodes. Simultaneously, based on the difference ΔT between the real-time temperature T and the reference temperature T0, as well as the linear expansion coefficient α and length L of the component, the temperature deformation ΔL=α*ΔT*L is calculated, and this deformation is then converted into an equivalent temperature load vector {Ft} according to finite element theory. Finally, the wind load vector {Fw}, the temperature load vector {Ft}, and the structural self-weight vector {Fg} (which is a constant) are superimposed: {Ftotal} = {Fw} + {Ft} + {Fg}, which gives the combined load vector {Ftotal} under the current working condition.

[0055] Step S22: Input the combined load vector into the digital twin model for nonlinear solution, and extract the stress ratio of each component in the solution result.

[0056] It should be noted that nonlinear solution refers to applying a combined load vector to a digital twin model (NIDA finite element model) and using numerical calculation methods to solve for the equilibrium state of the structure under this load, obtaining the nodal displacements and internal forces of the components. Its nonlinearity manifests in two aspects: first, geometric nonlinearity, meaning that large deformations of the structure lead to changes in stiffness (e.g., P-Δ effect – geometric nonlinearity at the structural level, P-δ effect – geometric nonlinearity at the component level); second, material nonlinearity, meaning that when the material stress reaches yield, the stress-strain relationship is no longer linear (using a bilinear elastoplastic model). This process involves continuously updating the tangent stiffness matrix of the structure through iterative algorithms (such as the Newton-Raphson method) until the force equilibrium condition is met. The stress ratio is a key indicator for measuring the stress state of a component, defined as the ratio of the maximum combined stress on the cross-section of the component under load (e.g., the maximum normal stress under the combined action of axial force and bending moment) to the allowable stress of the component material (e.g., the yield strength of steel). The closer the stress ratio is to 1, the closer the component is to its bearing limit, and the higher the risk.

[0057] Understandably, the purpose of this step is to calculate the most realistic mechanical response of the transmission tower using a high-fidelity digital twin model under accurate load input, and to quantify the risk level of each component using the stress ratio as a standardized indicator. The principle is based on the application of nonlinear finite element theory. The digital twin model (NIDA model) can accurately simulate the complex behavior of the transmission tower under load, including axial compression / tension and bending of members, as well as semi-rigid node connections. The combined load vector is applied to the model as external forces, and the nonlinear solver iteratively calculates the node displacements {U} that bring the structure to force equilibrium. Based on the displacements {U}, the internal forces (axial force N, bending moment M) of each component can be further calculated. Then, based on the component's cross-sectional properties (such as area A, section modulus W), its maximum combined stress σ_max = |N / A| + |M / W| is calculated. Finally, the stress ratio R = σ_max / [σ] is calculated, where [σ] is the allowable stress of the material. This ratio eliminates the influence of different component sizes and materials, providing a unified and comparable risk quantification indicator for all components.

[0058] Specifically, the NIDA nonlinear solver is invoked, employing an incremental iterative method. First, the global tangent stiffness matrix [K_T] is assembled based on the model's initial configuration (i.e., the undeformed state). ]0 Solve the linear equation system [K_T]0*{ΔU}0={Ftotal} to obtain the displacement increment {ΔU}0, and update the nodal coordinates. Based on the new configuration, recalculate the element strain and stress to obtain the internal nodal force vector {Fint}. Calculate the residual force {R}={Ftotal}-{Fint}. Determine if the norm of {R} is less than the convergence tolerance. If it does not converge, recalculate [K_T]1 based on the new configuration, solve for the new displacement increment, and repeat the iteration until {R} satisfies the convergence condition. After convergence, extract the axial force N for each beam-column element from the results. i and bending moment M i For the i-th component, its stress ratio R i =(|N i | / A i +|M i | / W i ) / fy, where A i W is the cross-sectional area. i Let fy be the section modulus for bending, and fy be the yield strength of the steel. By iterating through all components, the stress ratio vector {R1, R2, ..., Ry} for the entire tower is obtained. n}

[0059] Step S23: Classify the stress ratios according to preset hazard level thresholds to generate a structural risk heat map covering the entire tower.

[0060] It should be noted that the preset hazard level thresholds are a predefined set of stress ratio numerical ranges used to discretize the continuous risk index (stress ratio) into a finite number of risk levels. For example, components with a stress ratio greater than 0.8 can be defined as high-risk, those between 0.5 and 0.8 as medium-risk, and those less than 0.5 as low-risk. These thresholds are set based on engineering experience, design safety factors, and regulatory requirements. The structural risk heatmap is a data visualization product that maps the stress ratio value of each component to different colors (such as red, yellow, and blue) according to its hazard level, and renders these colors on the corresponding components of the transmission tower's 3D model, thus forming a color map that intuitively displays the spatial distribution of risk across the entire tower.

[0061] Understandably, the purpose of this step is to transform abstract numerical calculation results into an intuitive and actionable risk distribution map, providing direct input for subsequent path planning. The principle is data classification and visualization mapping. While the stress ratio is a precise numerical value, for path planning algorithms, the relative level and category of risk are more important. By setting thresholds for classification, all components can be categorized into a few risk levels. Different levels are assigned distinct colors (e.g., red - high risk, yellow - medium risk, blue - low risk) and rendered on a 3D model, allowing operators or subsequent automated algorithms to easily identify which areas require priority attention and monitoring.

[0062] Specifically, read the stress ratio R of all components. i Preset risk level thresholds, such as: High Risk (R) i ≥0.8), medium risk (0.5≤R) i <0.8, low risk (R i <0.5). Traverse each component, based on its R... i The value determines its risk level. Then, each level is assigned a color and a path planning weight coefficient (e.g., high risk - red - weight 3, medium risk - yellow - weight 2, low risk - blue - weight 1). Finally, each component is highlighted with its corresponding color on the 3D digital model of the transmission tower (which can be a simplified wireframe model or a more detailed model). The resulting heatmap can be a static image or an interactive 3D visualization interface that supports rotation, zoom, and click-to-query.

[0063] This embodiment employs the above-described scheme. First, it uses a drone to collect environmental data in real time and construct a combined load vector, enabling the load input of the digital twin model to dynamically reflect the actual environment of the transmission tower, providing a foundation for accurate analysis. Second, it solves the load input using a high-fidelity nonlinear NIDA model and extracts the stress ratio, fully utilizing the advantages of the digital twin model in complex stress analysis. The resulting component risk index (stress ratio) is scientific and accurate, reflecting geometric and material nonlinear effects, and is closer to engineering reality than traditional linear analysis. Finally, the stress ratio is classified and visualized as a heat map using preset thresholds, transforming the complex mechanical calculation results into an intuitive and easy-to-understand risk spatial distribution map. This series of sub-steps ensures that the final generated structural risk heat map is not only real-time but also highly accurate and intuitive, providing a reliable and quantitative decision-making basis for subsequent targeted adaptive path planning.

[0064] Based on the above implementation scheme, in one feasible implementation, the environmental data includes wind speed data, wind direction data, and temperature data, and the step of constructing a combined load vector based on the environmental data includes S31~S33: Step S31: Based on the shape coefficient and wind-receiving area of ​​each component in the digital twin model, the wind speed data and the wind direction data are converted to obtain the wind load vector.

[0065] It should be noted that the shape factor is a dimensionless coefficient used to reflect the influence of the component's cross-sectional shape on wind pressure. Different cross-sectional shapes (such as angle steel, steel pipe, and composite sections) experience different wind forces at the same wind speed. The shape factor is given by wind tunnel tests or design specifications (such as building structure load codes). The windward area refers to the projected area of ​​the component on a plane perpendicular to the wind direction. Converting wind speed and direction data into a wind load vector is a physical calculation process based on Bernoulli's principle in aerodynamics. The basic formula is wind pressure w = 0.5 * ρ * v², where ρ is air density and v is wind speed. Then, based on the component's shape factor μs and the windward area A, the wind load F = w * μs * A is calculated. Finally, the wind load acting on all components is distributed to each node according to the node connection relationship of the finite element model, forming a global wind load vector.

[0066] Understandably, the purpose of this step is to convert the point-like wind speed and direction information measured by the UAV into distributed nodal forces acting on the structural digital model. The principle lies in discretizing and equivalencing the continuous wind field effect. The wind speed and direction measured by the UAV represent the main characteristics of the wind field around the tower. By converting the wind speed into wind pressure and then multiplying it by the shape coefficient and wind-receiving area of ​​each component, the total wind force acting on that component is obtained. Since the finite element model uses nodal displacements as the basic unknowns, the distributed or concentrated wind loads on the components need to be converted (distributed) to the nodes at both ends of the components using the static equivalence principle, forming nodal loads. Performing this operation on all components and arranging all nodal loads in order of degrees of freedom constitutes the wind load vector. This process ensures that the way the wind loads are applied to the digital model is mechanically equivalent to the actual action.

[0067] Specifically, firstly, the wind speed v and wind direction angle α transmitted back by the drone are preprocessed, such as by taking the average value over a period of time to eliminate the influence of fluctuations. The air density ρ is then queried or calculated based on the altitude of the area where the transmission tower is located. Then, for each component unit in the digital twin model, the following calculations are performed: First, based on the component's orientation in the global coordinate system and the wind direction angle α, the angle θ between the wind direction and the component's axis normal is calculated. Then, based on the angle θ and the component's cross-sectional shape, its shape coefficient μs(θ) is queried from a pre-stored database. Next, the projected area of ​​the component on the plane perpendicular to the wind direction, A_proj = A*|sin(θ)|, is calculated, where A is the actual surface area of ​​the component. Then, the total wind load acting on the component, F_wind = 0.5*ρ*v²*μs(θ)*A_proj, is calculated. Finally, based on the node coordinates at both ends of the component, F_wind is distributed to the two nodes according to the static equivalence principle, typically decomposed into two components perpendicular to the component's axis. After completing the above calculations for all components, the wind load components distributed from different components at each node are superimposed to obtain the equivalent wind load for that node. Finally, the equivalent wind loads of all nodes are arranged in order of node number to form the global wind load vector {F_wind}.

[0068] Step S32: Calculate the axial temperature deformation of each component in the digital twin model caused by the temperature difference based on the temperature data to obtain the temperature load vector.

[0069] It should be noted that temperature difference refers to the difference between the current ambient temperature measured by the drone and a reference temperature (usually the temperature at the time of structural installation or the temperature under stress-free conditions). Axial temperature deformation refers to the elongation or shortening of a component along its length due to thermal expansion and contraction. For statically determinate structures, temperature deformation is free and does not generate internal forces; however, for statically indeterminate structures (such as multi-dimensional statically indeterminate truss structures like transmission towers), temperature deformation is constrained by adjacent components, thus generating temperature stress within the components. The temperature load vector is the equivalent nodal force vector applied in finite element analysis to simulate this constraint effect. To prevent free deformation of components due to temperature differences, forces need to be applied to both ends of the component. In finite element analysis, temperature stress and deformation can be automatically calculated by directly applying temperature changes, and the calculation of the equivalent load vector is usually built into the element matrix.

[0070] Understandably, the purpose of this step is to consider the impact of ambient temperature changes on the internal forces of the transmission tower structure. Temperature changes cause thermal expansion and contraction of component materials. In truss structures like transmission towers, the components are interconnected and mutually constrained. The free thermal deformation of a single component is restricted by other components, thus generating additional internal forces (temperature stress) within the structure. This temperature stress, combined with the stress generated by wind loads and self-weight, may exacerbate the stress conditions of certain components. By calculating the temperature load vector and applying it to the finite element model, the temperature effect can be incorporated into the structural analysis, making the calculation results of the digital twin model more comprehensive and accurate. The principle is to use thermoelasticity and the finite element method to introduce the temperature change ΔT as the initial strain ε0=α*ΔT (α is the coefficient of linear expansion) into the constitutive relation of the element, thereby generating equivalent nodal force terms in the element stiffness equation.

[0071] Specifically, first, a reference temperature T_ref (e.g., 20℃) is set. The temperature T_current, collected in real-time by the drone, is read. The temperature difference ΔT = T_current - T_ref is calculated. Then, for each component element in the digital twin model, its coefficient of linear expansion α (typically 1.2e-5 / ℃ for steel) is obtained. In finite element analysis, for rod or beam elements, the equivalent nodal force vector caused by the temperature difference ΔT can be directly calculated from the element stiffness matrix and temperature strain. A common calculation method is: for rod elements considering only axial deformation, the equivalent nodal force generated by the temperature change ΔT in the element's local coordinate system is {F_T} = EA * α * ΔT * [-1,1]^T, where E is the elastic modulus and A is the cross-sectional area. For beam elements considering bending, the calculation is more complex and is usually handled internally by the finite element software. In practical system implementation, the preprocessing or solution interface of the finite element software is typically called to directly apply a temperature load ΔT to each element. The solver automatically incorporates this contribution when assembling the overall stiffness matrix and overall load vector. Ultimately, all equivalent nodal forces caused by temperature changes are assembled into a global temperature load vector {F_temp}.

[0072] Step S33: Superimpose the wind load vector, the temperature load vector, and the structural self-weight load vector to obtain the combined load vector.

[0073] It should be noted that superposition is a vector addition operation. In structural finite element analysis, if the structure is within the linear elastic range and the deformation is small enough not to affect the load (small deformation assumption), then the effects (internal forces, displacements) produced by different load cases can be superimposed. For the nonlinear NIDA method used, although the final equilibrium equations require iterative solution, when constructing the total load vector, the nodal force vectors corresponding to various load cases are still algebraically added to obtain a total nodal load vector, which serves as the right-hand side term of the nonlinear equations. The structural self-weight load vector is constant and is calculated from the density of the structural materials, gravitational acceleration, and the volume (or element mass matrix) of each component, and is determined during model initialization.

[0074] Understandably, the purpose of this step is to integrate the three main load effects—wind, temperature, and self-weight—to form a complete load case, simulating the actual stress state of the transmission tower in a real environment. This is based on the superposition principle (within its applicable range) and equilibrium equations in mechanics. In actual service, the stress state of a transmission tower is the result of the combined effects of its self-weight, wind pressure, and temperature changes. To accurately analyze its structural response, these loads must be considered simultaneously. The calculated wind load vector {F_wind}, temperature load vector {F_temp}, and the pre-calculated self-weight load vector {F_gravity} are added together: {F_total} = {F_wind} + {F_temp} + {F_gravity}. This {F_total} vector completely represents all external forces acting on the transmission tower under the current environmental conditions. Substituting it into the equilibrium equations and solving for it yields the structural response under this integrated load case.

[0075] Specifically, the system receives wind load vectors, temperature load vectors, and retrieves pre-calculated and stored structural self-weight load vectors from the digital twin model parameter library. These three vectors have the same dimension (equal to the total number of degrees of freedom in the model). Vector addition is performed for each degree of freedom i (corresponding to a node in a certain direction). By traversing all degrees of freedom, the final combined load vector is generated. This combined load vector is then passed to the nonlinear solver as the load input for the current time step or load condition. To ensure load synchronization, the system typically employs a timestamp alignment mechanism to ensure that the superimposed wind and temperature data were collected at the same time or within the same time period.

[0076] This embodiment, through the aforementioned scheme, calculates wind load vectors and temperature load vectors separately, and then superimposes them with the self-weight load vector. This method comprehensively and accurately characterizes the main load types borne by transmission towers in the field environment. Converting wind speed and direction into wind load vectors considers the dynamic effects of wind and the influence of component shape; calculating the temperature load vector incorporates the thermal stress effects caused by seasonality and even diurnal temperature variations. This refined load modeling approach makes the input conditions driving the digital twin model for structural response analysis closer to engineering reality. This ensures that the subsequently calculated structural risk heatmap more realistically reflects the actual safety status of the transmission tower under the current wind-temperature coupling environment, laying a solid data foundation for risk-based path planning.

[0077] Based on the above implementation scheme, in one feasible implementation, the step of generating the initial inspection path of the transmission tower by constructing a path using a weighted coverage path planning algorithm based on the structural risk heat map includes S41~S42: Step S41: Extract the hazard level of each component from the structural risk heat map, and determine the corresponding path planning weight coefficient for each component based on the hazard level. The hazard level is divided into high-risk, medium-risk and low-risk levels according to the relationship between the component stress ratio and the preset threshold.

[0078] It's important to note that the path planning weight coefficient is a positive real number used to quantify the importance of each detection point (usually associated with a component) in the path planning problem. A higher weight indicates a higher priority for that point in path optimization, and the algorithm will tend to plan shorter or earlier paths to that point. The weight coefficient is directly mapped from the component's hazard level. Hazard levels are discrete classifications, such as high-risk, medium-risk, and low-risk, determined by the component's stress ratio falling within a preset threshold range. For example, a stress ratio ≥ 0.8 is high-risk, 0.5 ≤ stress ratio < 0.8 is medium-risk, and a stress ratio < 0.5 is low-risk.

[0079] Understandably, the purpose of this step is to quantify the results of the structural risk assessment into optimization objective weights for path planning, thereby establishing a direct link between structural status and inspection behavior. The principle is importance weighting. The structural risk heatmap has already identified which components are high-risk (red). In path planning, we want the drone to prioritize and reliably inspect these high-risk points. By assigning different weight coefficients to components of different risk levels (e.g., high-risk weight = 5, medium-risk weight = 2, low-risk weight = 1), the algorithm perceives the differences in importance between different points during the optimization process when constructing the optimization objective function (e.g., minimizing the total weighted access cost). The benefits or reduced risk costs of accessing a high-risk point are far greater than those of accessing a low-risk point. Therefore, under the premise of satisfying all constraints (e.g., endurance, obstacle avoidance), the algorithm will automatically generate a path that tends to access high-risk points first or in a more favorable location, thus achieving intelligent tilting of inspection resources towards high-risk areas.

[0080] Specifically, the system reads the data structure of the structural risk heatmap. This data structure contains the ID of each component, the 3D coordinates of its center point or preset detection point, and its corresponding hazard level label (e.g., high-risk, medium-risk, low-risk). An internally pre-configured or user-configurable weight mapping table is provided, for example: {high-risk: 5, medium-risk: 2, low-risk: 1}. For each component (or each detection point associated with a component) in the heatmap data, the system looks up and assigns its corresponding weight coefficient Wi from the weight mapping table based on its hazard level label. If a component has multiple detection points, these detection points typically inherit the component's weight. Finally, a set P = {p1, p2, ..., pN} containing all points to be accessed (total number N) is obtained, where each point pi has its 3D coordinates (xi, yi, zi) and corresponding weight coefficient Wi.

[0081] For example, a transmission tower has 200 components to be inspected, of which 10 are marked as high-risk (stress ratio > 0.8), 40 as medium-risk, and 150 as low-risk. The system assigns a weight of 5 to the 10 high-risk components, a weight of 2 to the 40 medium-risk components, and a weight of 1 to the 150 low-risk components, according to a pre-defined mapping table. Therefore, in subsequent path planning, the algorithm recognizes that visiting these 10 points with a weight of 5 is far more valuable or urgent than visiting other points.

[0082] Step S42: The initial inspection path of the transmission tower is obtained by weighting the path planning weight coefficients using the weighted shortest coverage path algorithm.

[0083] It should be noted that the weighted shortest coverage path algorithm is a combinatorial optimization algorithm used to solve the weighted coverage path planning problem. The goal of this problem is to plan a continuous flight path for the UAV, starting from a starting point (e.g., take-off and landing point), visiting (covering) all target points (or at least once), and finally returning to the starting point (or reaching the destination), while satisfying a series of constraints (such as UAV endurance, minimum turning radius, and safe distance from obstacles), thus optimizing a certain objective function. Common optimization objectives include: minimizing the total flight distance (or time), minimizing the weighted access delay (i.e., higher-weighted points should be visited earlier), or maximizing the total weighted coverage gain within a given flight time. The output of this step is the initial inspection path, which is a sequence of ordered three-dimensional spatial coordinate points (waypoints), and the UAV will fly strictly according to this sequence.

[0084] Understandably, the purpose of this step is to solve a complex, constrained combinatorial optimization problem to generate an optimal or near-optimal inspection path. The principle is to abstract the real-world engineering problem into a mathematical model and solve it. The input consists of all inspection points and their weights, as well as the UAV's kinematic and dynamic constraints and the environment's geometric and electromagnetic constraints. Algorithms (such as improved genetic algorithms, ant colony algorithms, simulated annealing algorithms, or sampling-based motion planning algorithms) search this high-dimensional, non-convex solution space. During the search, the algorithm evaluates the quality of each candidate path, with its evaluation criteria (objective function) including weight coefficients. For example, a common objective function is to minimize the total weighted access distance, which is the sum of the flight distance from one point to another on the path multiplied by the weight of the destination point (or a weight-related function). Thus, a path that forces the UAV to detour to visit a low-weight point will have a high weighted distance cost; while a path, although slightly longer in total distance, may have a lower weighted distance cost by prioritizing visits to all high-weight points, and is therefore selected as the better solution by the algorithm. By solving this optimization problem, the final path satisfies all safety and physical constraints while achieving key coverage of high-risk components.

[0085] Specifically, the solver receives the following inputs: a set of all detection points P, a weight set W, the coordinates of the UAV's take-off and landing points, UAV performance parameters (maximum speed, endurance, minimum turning radius), and environmental constraints (such as the safe distance d_safe from power lines and the minimum observation distance d_obs from tower components). The algorithm first constructs a mathematical model of the problem. A feasible model is: suppose the UAV needs to start from the starting point S, visit each point in the point set P at least once, and finally return to the destination G (which can be the same as S). The total path cost C is defined as: C = summation(f(Wj) of the flight distance from point i to point j), where f(Wj) is an increasing function of the weight of the target point j (e.g., f(Wj) = Wj). Constraints include: total flight distance / time ≤ UAV endurance; the distance between any point on the path and any obstacle (power line, tower) ≥ safe distance; the turning angle between adjacent waypoints ≤ the UAV's maximum turning angle. This is a constrained variant of the Traveling Salesman Problem. The solver employs a heuristic algorithm (such as ant colony optimization) for solving the problem: the algorithm maintains an ant colony, with each ant representing a candidate path. Ants select the next visit point based on the path length and the weight of the next target point (pheromone and heuristic information), ultimately generating a complete path. All ants' paths are evaluated based on the total weighted cost, and the pheromone is updated accordingly. After multiple iterations, a relatively optimal path converges, which is the initial inspection path. This path is output as a list of waypoint sequences.

[0086] This embodiment, through the aforementioned scheme, deeply integrates structural risk information based on mechanical analysis into the decision-making logic of path planning. First, by extracting hazard levels from heatmaps and mapping them to weighting coefficients, a quantitative transformation from risk assessment to planning priority is achieved. Then, through a weighted shortest coverage path algorithm, these weights are explicitly introduced into the optimization objective, ensuring that the generated inspection path is no longer a blind geometric traversal but an intelligent plan aimed at minimizing risk-weighted costs. This ensures that, within the limited battery life and flight time, the UAV's flight trajectory spontaneously clusters towards high-risk components, significantly improving the detection frequency and reliability of these critical parts. This method fundamentally solves the problem that traditional equidistant or terrain-following path planning cannot distinguish the importance of components, potentially leading to insufficient inspection of high-risk areas, thus optimizing inspection efficiency and safety benefits.

[0087] Based on the above implementation scheme, in one feasible implementation, the step of triggering local path replanning corresponding to the target anomaly and generating the target inspection path includes S51~S52: Step S51: Obtain the defect location information of the target anomaly, and generate a local reshoot path based on the defect location information within a preset range around the target anomaly.

[0088] It should be noted that defect location information refers to the precise coordinates of the target anomaly in three-dimensional space, determined simultaneously when an anomaly is identified by an airborne or online defect detection model. This is typically obtained by fusing real-time positioning of the UAV with measurements from airborne visual sensors (such as binocular ranging and visual SLAM). The preset range is a pre-defined spatial region parameter, such as a spherical space with a radius of R (e.g., 2 meters) centered on the defect point, or a cylindrical space with a length extension L and a radial distance D along the defective component. This range defines the detailed inspection area that the local supplementary imaging needs to cover. The local supplementary imaging path is a newly planned, shorter flight trajectory within the preset range, designed to guide the UAV to acquire more detailed, higher-resolution images or point cloud data of the defect target from multiple angles and distances. This path is temporarily generated and is only used for enhanced inspection of the anomaly area.

[0089] Understandably, the purpose of this step is to dynamically respond to unexpected structural defects discovered during the inspection process and generate targeted, detailed inspection plans. The principle is based on motion planning in local space. When an online defect detection system identifies a target anomaly (such as a crack), relying solely on a single image or perspective at the time of discovery often fails to provide sufficient information to assess the severity of the defect (such as crack length, depth, and orientation). First, the precise three-dimensional location of the defect is obtained. Then, a local area of ​​interest is defined centered on this location. Within this area, a new flight path (a local re-shooting path) is planned, guiding the drone to fly around the defect point and capture images from multiple preset, optimal observation angles, including the front, side, and top. This multi-angle, close-range shooting acquires richer two-dimensional images and three-dimensional point cloud data, providing high-quality data input for subsequent defect quantification (such as dimensional measurement) and model updates. This is equivalent to temporarily inserting a local, detailed sub-task into a global, general inspection task.

[0090] Specifically, when an abnormal trigger signal (containing defect type, confidence level, and 3D position P(x,y,z)) is received from the defect detection module, local path replanning is triggered. First, it determines the local reshoot range R based on the defect type and preset rules. Then, within a sphere (or a cylinder determined according to the component orientation) centered at point P with radius R in 3D space, a set of predefined observation viewpoints is generated. Each viewpoint includes a 3D coordinate and the camera's orientation (looking towards the defect point). The viewpoint generation strategy can be: taking several points at equal intervals on a plane circle perpendicular to the defect surface normal, and always pointing the camera towards the defect center; or setting one viewpoint above, below, to the left, and to the right of the defect point. Then, the local replanning unit needs to plan a flight path for the UAV starting from its current position, efficiently visiting these new viewpoints in sequence, and finally returning to the original path. This is essentially a small-scale traveling salesman problem, but due to the small number of points and limited space, it can be solved using a fast algorithm (such as the nearest neighbor method). The ordered waypoint sequence that visits these new viewpoints is the local reshoot path.

[0091] For example, during flight, the onboard camera of the drone identifies a bolt loosening anomaly with a 95% confidence level at the crossarm connecting plate and obtains its three-dimensional coordinates P by fusing positioning information. A local replanning unit is triggered, with a preset range R = 1 meter. The system generates 5 supplementary shooting viewpoints: V1 (1 meter directly in front of P), V2 (1 meter 45 degrees to the left and front of P), V3 (1 meter 45 degrees to the right and front of P), V4 (1 meter directly above P), and V5 (1 meter directly below P). Then, a path is planned from the drone's current position to V1, then sequentially visiting V2, V3, V4, and V5, and finally flying back to the original trigger point. This path is the local supplementary shooting path.

[0092] Step S52: Update the currently unexecuted initial inspection path based on the local reshoot path to obtain the target inspection path.

[0093] It should be noted that the currently unexecuted initial inspection path refers to the sequence of remaining waypoints that the UAV has not yet flown when it detects an anomaly and triggers a local replanning during the initial inspection path execution. Updating refers to modifying this remaining waypoint sequence, inserting the newly generated local reshoot path into the appropriate position, and may require adjusting the order of subsequent waypoint visits. The target inspection path is the final generated complete flight plan that the UAV will actually execute; it integrates the executed portion, the newly inserted local reshoot path, and the adjusted remaining initial inspection path.

[0094] Understandably, the purpose of this step is to seamlessly integrate urgent local reshooting tasks into the overall inspection task flow, forming a coherent and executable final flight plan. The principle is the dynamic scheduling and path splicing of task sequences. A UAV cannot execute two paths simultaneously; it requires a complete, collision-free flight trajectory from the starting point to the ending point (or back to the starting point). Once the local reshooting path is generated, a decision needs to be made on how to insert it into the original plan. In one embodiment of this application, the execution of the current initial path is paused, the local reshooting path is executed immediately, and after the reshooting is completed, the UAV returns to the trigger point (or the next waypoint closest to the trigger point in the original path) to continue executing the remaining initial path. The UAV's flight task queue needs to be modified: the remaining initial path is replaced with the local reshooting path + a connecting segment from the reshooting end point to the remaining initial path + the remaining initial path. This new, integrated waypoint sequence is the target inspection path. It ensures that anomalies can be checked immediately and in detail, while the overall inspection task can be resumed and completed after an interruption.

[0095] Specifically, after generating the local reshoot path, the UAV's flight task queue is modified. Assume the UAV detects an anomaly at the i-th waypoint W_i on the initial inspection path (or en route to W_i). The currently unexecuted initial path is a sequence from W_i to the endpoint (or return point): Remain_Path=[W_i,W{i+1},...,W_end]. The local reshoot path is Local_Path=[L_start,L_1,L_2,...,L_m,L_end], where L_start is usually the UAV's current position or W_i when triggered. First, a return segment Return_Segment is planned from the endpoint L_end of the local reshoot path to a suitable waypoint (usually W_i or W{i+1}) on the original remaining path. Then, a new target inspection path is constructed: Target_Path=[already flown path]+[Local_Path]+[Return_Segment]+[Remain_Path (starting from the connection point)]. Here, [the paths already flown] are historical records, and [Local_Path] + [Return_Segment] are the newly inserted task blocks. The new path is checked to ensure it meets all constraints (e.g., total length does not exceed flight range). Finally, this new Target_Path is sent to the drone flight control system, replacing the original flight plan. The drone will immediately begin executing this updated path.

[0096] For example, a drone is flying from waypoint A to waypoint B when an anomaly is triggered en route. The initial remaining path is B->C->D->E (return point). The partial reshoot path is current position->X1->X2->X3->return point R. The system plans a return link from X3 to waypoint B. Therefore, the new target inspection path is: A->[Partial reshoot path: X1->X2->X3]->[Return link: X3->B]->[Remaining initial path: B->C->D->E]. After completing the reshoots for X1, X2, and X3, the drone flies back to point B and continues to C, D, and E.

[0097] This embodiment, through the aforementioned scheme, obtains the precise location of defects and plans local re-shooting paths. This allows the system to guide the UAV to conduct multi-angle, close-range specialized inspections of abnormal areas, acquiring richer and higher-quality defect data. This is crucial for subsequent defect classification, maintenance decisions, and digital twin model updates. Furthermore, by dynamically updating the local re-shooting paths to the overall inspection path, the global inspection task and local emergency inspection tasks are seamlessly integrated. This ensures the timeliness and sufficiency of sudden defect inspections without affecting the completion of the overall inspection task. This mechanism greatly enhances the autonomy and intelligence of the UAV inspection system, enabling it to identify suspicious points and immediately conduct focused re-inspections, much like an experienced inspector. This effectively avoids misjudgments or missing information caused by poor single-shot angles or insufficient resolution.

[0098] Based on the above implementation scheme, in one feasible implementation, the step of updating the digital twin model based on the inspection data to obtain the updated digital twin model includes S61~S63: Step S61: Extract point cloud data and image data from the inspection data; register and fuse the point cloud data to extract the measured displacement of key nodes; and perform three-dimensional reconstruction of the image data to extract the cross-sectional change characteristics of the components.

[0099] It should be noted that point cloud data is a collection of three-dimensional spatial points generated by a lidar system mounted on a drone or through visual SLAM technology. Each point contains (X, Y, Z) coordinates and may also include reflection intensity information. It characterizes the geometric shape of the transmission tower surface. Image data consists of two-dimensional images or video frames captured by a visible light, infrared, or high-resolution camera mounted on a drone. Registration and fusion refer to the process of aligning point cloud data acquired from multiple scans or different perspectives to the same global coordinate system using algorithms such as ICP (Iterative Closest Point) and merging them into a complete and consistent three-dimensional point cloud model. The measured displacement of key nodes is obtained by comparing the actual three-dimensional positions of key structural nodes of the transmission tower (such as the intersection of main and diagonal members, crossarm endpoints, etc.) under load from the high-precision point cloud model after registration and fusion, through feature matching or direct measurement, with the predicted positions of the digital twin model under the same load, thus obtaining the displacement difference. The component cross-sectional change characteristic refers to the change in the geometric dimensions (such as the side length of angle steel, the diameter of steel pipe) or appearance features (such as the area of ​​rusted area) of the component cross-section, which are reconstructed from image data through three-dimensional reconstruction (such as structure for motion recovery, SfM) or stereoscopic vision measurement technology, compared with the design value or the previous inspection value, such as rust depth and cross-sectional loss rate.

[0100] Understandably, the purpose of this step is to extract observational evidence from the raw, multimodal inspection data collected by UAVs, which can be used for quantitative evaluation and correction of the digital twin model. The principle is to use 3D vision and photogrammetry techniques for reverse engineering and precision measurement. The digital twin model predicts nodal displacements and component internal forces (stresses), while the UAV directly collects point clouds and images. By registering and fusing multiple point clouds, a complete and accurate 3D scan model of the transmission tower can be obtained. By comparing this scan model with the geometric model of the digital twin under no-load conditions, the actual displacements (observed values) of key nodes under wind, temperature, and other loads can be calculated. Through 3D reconstruction of multi-view images, more refined geometric and texture information of the component surface can be obtained, thereby measuring the changes in cross-sectional dimensions (observed values) caused by corrosion and damage. These measured displacements and cross-sectional changes are the most direct evidence reflecting the true state of the structure and are the data foundation driving model updates.

[0101] Specifically, the system receives raw point cloud frames and image sequences transmitted from the UAV. Preprocessing of the point cloud data, including denoising and downsampling, is performed. Then, algorithms such as ICP are used to register the point clouds of all stations in this inspection to the same global coordinate system. The registered point cloud is coarsely aligned with the reference geometric model of the digital twin model (usually the design model or the model updated in the last update) using invariant features such as the tower base. Near a predefined set of key nodes (usually stress-critical or easily deformable nodes) in the digital twin model, the 3D point set of the node region is extracted from the registered point cloud, and its centroid is calculated or the precise 3D coordinates P_measured of the node are obtained through fitting. This coordinate is subtracted from the node coordinates P_predicted predicted by the digital twin model under the same load conditions to obtain the measured displacement observation value δ_obs = P_measured - P_initial (initial coordinates) of the node. For high-resolution images of the same component taken from multiple angles, SfM (Structure of Motion) technology is used for sparse point cloud reconstruction and dense reconstruction to generate a local 3D mesh model of the component. From the reconstructed 3D model, the cross-sectional dimensions of the target component are measured, such as the length of the angle steel edge l_measured, and compared with the design dimension l_design or the dimension from the last inspection l_previous to obtain cross-sectional change characteristics, such as the dimension change Δl = l_measured - l_design, or an estimate of the corrosion depth.

[0102] Step S62: Using the measured displacement of the key node and the characteristic quantity of the component cross-section change as observation values, the Bayesian data assimilation method is used to perform posterior estimation and parameter correction on the current state vector of the digital twin model, so as to obtain the digital twin model after correcting the model parameters.

[0103] It should be noted that the current state vector of the digital twin model is a set containing all the parameters of the model to be corrected, such as: the elastic modulus degradation coefficient of each component (characterizing material aging), the initial geometric defect vector (characterizing installation error or cumulative deformation), and the semi-rigid stiffness coefficient of the node connection (characterizing bolt preload loss or loosening), etc. It is a multidimensional vector θ. The observed values ​​are the extracted measured displacements and cross-sectional changes, denoted as vector y. Bayesian data assimilation methods are probabilistic methods that combine prior knowledge with observed data to update state estimates, such as ensemble Kalman filtering, unscented Kalman filtering, or particle filtering. Its core is Bayes' theorem: P(θ|y)∝P(y|θ)*P(θ), where P(θ) is the prior probability distribution of the state vector (based on design values ​​or the previous update result), P(y|θ) is the likelihood function of observed y given parameter θ (characterizing observation error), and P(θ|y) is the posterior probability distribution of the state vector after fusing the observed data. By estimating the posterior distribution (e.g., by taking the mean), the corrected state vector estimate θ_updated can be obtained.

[0104] Understandably, the purpose of this step is to use measured data to calibrate and correct parameters in the digital twin model that may have deviated from their true values, making the model predictions more consistent with physical reality. Its principle is based on the quantification and updating of uncertainty within a probabilistic framework. The parameters of the digital twin model (such as material properties and initial defects) have uncertainties (prior uncertainties), and the UAV measurement data also contains errors (observational uncertainties). The Bayesian data assimilation method provides a rigorous mathematical framework to handle these two types of uncertainties. It treats the model parameters as random variables with a prior distribution P(θ). When new observational data y is obtained, the posterior distribution P(θ|y) is calculated using Bayes' theorem. This posterior distribution integrates prior knowledge and new observational evidence, representing the latest and most accurate understanding of the model parameters after obtaining new data. The expected value of the posterior distribution is taken as the optimal estimate of the parameters, and this estimate is used to update the corresponding parameters of the digital twin model. This process is equivalent to training or calibrating the simulation model with measured data, making it continuously approximate the real structure.

[0105] Specifically, taking Ensemble Kalman Filtering (EnKF) as an example: First, initialization is performed. Based on the prior uncertainty of the model parameters, a parameter set {θ_i, i=1,...,N} with N members is generated, where each θ_i is a random sample of the state vector. Then, in the prediction step, for each member θ_i, a digital twin model is run (i.e., a finite element analysis is performed), with the current environmental load as input, to obtain the model-predicted observations (i.e., predicted nodal displacements and cross-sectional dimensions) y_i^f=H(θ_i), where H is the observation operator (a function that maps model parameters to predicted observations, i.e., the finite element solver). The set mean y^f and covariance P_yy of the predicted observations are calculated. Then, in the analysis step (update step), the actual observation y and its error covariance R are obtained. The Kalman gain K=P_θy*(P_yy+R)^{-1} is calculated, where P_θy is the covariance between the parameters and the predicted observations. Then, update each member: θ_i^a = θ_i^f + K*(y + ε_i - y_i^f), where ε_i is the observation perturbation following an N(0,R) distribution. Finally, output the mean of the updated parameter set {θ_i^a}, which is the corrected state vector estimate θ_updated. Replace the corresponding parameters in the digital twin model with θ_updated (e.g., write the new elastic modulus value and the new initial defect vector into the model file) to obtain the digital twin model with corrected model parameters.

[0106] Step S63: Calculate the model prediction value based on the digital twin model after correcting the model parameters, calculate the residual between the model prediction value and the measured value, and when the residual is less than a preset threshold, accept the correction of the model parameters to obtain the updated digital twin model.

[0107] It should be noted that the model prediction value refers to the predicted values ​​y_pred of the key node displacements and component cross-sectional dimensions obtained by rerunning the digital twin model (with the same input environmental loads) after obtaining the corrected model parameters θ_updated. The measured values ​​are the extracted observation values ​​y_obs. The residual is the difference between the predicted and measured values, usually measured using the norm of the vector (such as the L2 norm, i.e., Euclidean distance), i.e., residual e = ||y_pred - y_obs||. The preset threshold is a pre-set tolerance threshold used to determine whether the quality of this model update is acceptable. If the residual is below the threshold, it means that the updated model prediction matches the measured data very well, and the update is reliable; if the residual is above the threshold, it means that the update may have problems (such as gross errors in the observation data, or the model itself is severely distorted), and the update result is unreliable.

[0108] Understandably, the purpose of this step is to perform quality control and confidence assessment on the model correction results, preventing unreliable observational data or abnormal update processes from contaminating the digital twin model, acting as a safety valve. Its principle is hypothesis testing and model validation. By calculating the residuals and comparing them with a preset threshold, a validity test can be performed. Smaller residuals indicate that the assimilation process has successfully incorporated observational information into the model, the model parameters have been reasonably corrected, and the model's predictive ability has been improved. Larger residuals, on the other hand, raise a red flag, indicating the need for manual intervention to check the data or model.

[0109] Specifically, after obtaining the corrected model parameters θ_updated, the digital twin model is rerun once using the corrected parameters (forward computation), with the same environmental load conditions as when the observed values ​​y_obs were generated. From the results of this model run, the predicted values ​​y_pred (i.e., displacements of the same nodes and dimensions of the same components) corresponding to the observed values ​​y_obs are extracted. The residual e = ||y_pred - y_obs|| is calculated. Typically, displacement residuals and dimensional residuals are calculated separately and then weighted and combined. The residual e is compared with a preset threshold e_threshold. This threshold can be a fixed value or dynamically set based on the statistical values ​​of historically updated residuals (e.g., twice the average of historical residuals). Finally, if e ≤ e_threshold, the update is considered reliable. The system will officially accept this parameter correction, permanently replacing the main version parameters of the digital twin model with θ_updated, completing the model update. At this point, the updated digital twin model is obtained. If e > e_threshold, the update is considered unreliable. Instead of updating the model, a manual review alert log is generated, recording the parameters, observations, predictions, and residuals of the attempted update, for the operations engineer to check. Simultaneously, the model state is rolled back to the version before the update.

[0110] This embodiment constructs a rigorous and reliable closed-loop automatic update system for the digital twin model through the above-described scheme. First, it accurately extracts physical observations (displacement, dimensional changes) directly comparable to model predictions from multimodal inspection data, providing high-quality input for data assimilation. Second, it employs a Bayesian data assimilation method for parameter correction. This method elegantly handles model and observation uncertainties within a probabilistic framework, scientifically integrating new and old information to obtain optimal parameter estimates. This allows the model to gradually approximate the true physical state, eliminating dependence on fixed sensors. Finally, a confidence gating mechanism based on residual calculation and threshold comparison verifies the validity of each update. Only updates that pass the verification are adopted, effectively preventing erroneous or abnormal observation data from contaminating the digital twin model and ensuring the reliability of the model's long-term evolution. This series of steps collectively achieves a secure and automated closed-loop update of the digital twin model based on mobile inspection data, which is crucial for the vitality and sustained fidelity of the digital twin.

[0111] Based on the above implementation scheme, in one feasible implementation, the step of generating the initial inspection path of the transmission tower by constructing the path using a weighted coverage path planning algorithm based on the structural risk heat map further includes S71~S72: Step S71: When the number of transmission towers is greater than one, obtain the comprehensive health parameters of each transmission tower. The comprehensive health parameters include the maximum stress ratio of components, the ratio of the maximum displacement of nodes to the design limit, the normalized value of service life, and the normalized value of the historical cumulative number of defects.

[0112] It should be noted that the comprehensive health parameters are a set of indicators that quantify the overall structural health of a single transmission tower from different dimensions. The maximum stress ratio of a component is the maximum value among all the stress ratios (the ratio of combined axial force and bending moment stress to allowable stress) of all components of the tower within the current inspection cycle, reflecting the current ultimate stress state of the structure. The ratio of maximum node displacement to design limit is the ratio of the maximum displacement response of all nodes of the tower to the displacement limit allowed by the design code (e.g., 1 / 50 of the tower height), reflecting the overall stiffness and deformation safety margin of the structure. The normalized service life value is the ratio within the range [0,1] obtained by dividing the actual service life of the tower by the design service life of this type of tower (e.g., 50 years); a larger ratio indicates an older tower. The normalized value of the historical cumulative defect count is calculated from the historical inspection record database. It represents the total number of various defects (such as rust, cracks, loose bolts, etc.) found and recorded for the tower within a preset time period (e.g., the past 5 years). This number is then normalized by dividing the total number of defects by a baseline value (the average number of defects across all towers on the same line). This value reflects the severity of the tower's health condition. These parameters together constitute a multi-dimensional indicator system for assessing the health status of a tower.

[0113] Understandably, the purpose of this step is to establish a comprehensive and quantifiable health profile for each transmission tower to be inspected in a multi-tower inspection scenario, providing a data foundation for line-level global resource optimization. The principle is multi-source information fusion and feature extraction. A single structural risk heatmap reflects real-time risk under specific environmental loads, but the long-term health status of a tower also needs to be considered in conjunction with its historical performance and aging patterns. Four defined parameters characterize the tower's health from four key dimensions: current instantaneous stress state, current deformation state, natural aging process, and historical damage accumulation. Obtaining these parameters means not only processing the real-time data from the current inspection but also accessing and correlating historical databases and design files, thereby forming a more comprehensive and long-term understanding of each tower.

[0114] Specifically, when the system plans an inspection task involving multiple transmission towers (such as N towers for a line), for each transmission tower T_i in the task list, the following operations are performed to obtain its four comprehensive health parameters: First, obtain the maximum stress ratio R_max_i of the components: Obtain the structural risk heat map data generated for tower T_i under the current (or most recent) environmental data driving force, and find the maximum value R_max_i from the stress ratios of all components. Then, obtain the ratio D_ratio_i of the maximum displacement of the nodes to the design limit: Similarly, extract the displacement response of all nodes from the finite element solution results for tower T_i, and find the displacement D_max_i with the largest absolute value. Read the design displacement limit [D]i (usually related to the tower height) from the design database of tower T_i. Calculate the ratio D_ratio_i of the maximum displacement of the nodes to the design limit = D_max_i / [D]i. Next, obtain the normalized service life value A_norm_i: Obtain the commissioning year Y_start_i and current year Y_now of tower T_i from the asset management system (or manually input), and calculate the service life A_i = Y_now - Y_start_i. Obtain the design service life A_design of towers of the same type (e.g., 50 years). Calculate the normalized value A_norm_i = A_i / A_design. If A_norm_i > 1, then set it to 1. Finally, obtain the normalized value F_norm_i of the historical cumulative defect count: Query the historical defect database and count the total number of valid defects C_i recorded for tower T_i in the past M years (e.g., M=5). Calculate the average number of defects C_avg for the entire line or all towers in the same area over the past M years. Calculate the normalized value F_norm_i = C_i / C_avg. Finally, a parameter vector Para_i=[R_max_i,D_ratio_i,A_norm_i,F_norm_i] is generated for each tower.

[0115] Step S72: Based on the comprehensive health parameters, a weighted calculation is performed to obtain the comprehensive health score of each transmission tower. The comprehensive health score is used to determine the inspection priority of each transmission tower.

[0116] It's important to note that weighted calculation involves assigning different weight coefficients to the four acquired health parameters according to their importance to the overall health status of the tower. A single scalar score is then calculated using a linear or non-linear aggregation function (such as a weighted sum). The overall health score is this calculated scalar value; it's a relative value used for comparisons among multiple towers in the same inspection mission. A higher score indicates a worse overall health status and a higher potential risk of failure. Therefore, the tower should be given a higher inspection priority in this mission, meaning it should be visited and inspected earlier or with more time allocated.

[0117] Understandably, the purpose of this step is to integrate multi-dimensional health parameters into an intuitive, ranking priority indicator, thereby solving the core resource scheduling problem of which tower to inspect first in multi-tower line inspections. The principle is a simple weighted method in multi-criteria decision analysis. The four health parameters reflect the tower's condition from different perspectives, but their importance may vary (for example, the current ultimate stress ratio may be a better predictor of immediate risk than service life). By assigning weights (w1, w2, w3, w4) to each parameter and calculating the weighted sum Score = w1*R_max + w2*D_ratio + w3*A_norm + w4*F_norm, a comprehensive score is obtained. This score considers immediate risk, aging, and historical issues. When planning the multi-tower access sequence, the approximate access priority can be determined according to the comprehensive health score from high to low (i.e., from least healthy to healthiest). This ensures that, within the limited overall inspection time window, the drone fleet can prioritize inspecting the towers that are most likely to require attention and pose the greatest safety hazards, thereby achieving optimal allocation of inspection resources and maximizing risk control benefits at the line level.

[0118] Specifically, after obtaining the comprehensive health parameter vector Para_i of all towers to be inspected, the unit performs the following calculations: First, weights are configured using a preset set of weight coefficients W=[w1,w2,w3,w4]. These weights can be determined based on domain expert experience or historical data analysis, and satisfy w1+w2+w3+w4=1. For example, they can be set to w1=0.4 (emphasizing current stress), w2=0.3 (emphasizing current deformation), w3=0.2 (considering aging), and w4=0.1 (considering historical defects). Then, a score is calculated: For each tower T_i, its comprehensive health score Score_i is calculated using the following formula: Score_i=w1*R_max_i+w2*D_ratio_i+w3*A_norm_i+w4*F_norm_i.

[0119] Since larger values ​​for all parameters indicate a worse state, a larger Score_i value indicates a worse overall health status for that tower, and thus a higher priority. Finally, the towers are sorted and output. After calculating the scores for all towers, they are sorted from largest to smallest Score_i to obtain a priority sequence. For example, suppose there are towers A, B, and C with scores of 0.72, 0.65, and 0.50 respectively. The access priority order would be: A > B > C. This priority sequence will serve as a key input for subsequent line-level joint path optimization, guiding the planning of multi-tower access order. For instance, when modeling a constrained traveling salesman problem, higher-priority towers will be assigned higher access rewards or tighter time window constraints, driving the algorithm to generate paths that prioritize accessing them.

[0120] This embodiment, through the aforementioned scheme, acquires and calculates the comprehensive health score of each tower, enabling the system to scientifically and quantitatively rank the health status of all towers on the power line. Based on this ranking and the determined inspection priority, when planning inspection tasks involving multiple towers, the drone's access sequence is no longer random or simply based on geographical order, but rather a risk-oriented intelligent sequence driven by structural health status. This ensures that, given limited total inspection resources (such as daily flight time and battery capacity), the towers with the worst health and highest risk are prioritized for inspection, thereby maximizing risk control efficiency within the larger operational unit of the power line. This method overcomes the shortcomings of traditional inspection plans that rely on fixed cycles or manual experience and cannot dynamically respond to the differentiated health status of each tower, thus improving the refined and intelligent management level of power grid operation and maintenance.

[0121] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the transmission tower inspection path planning method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0122] This application also provides a transmission tower inspection route planning device; please refer to... Figure 2 The transmission tower inspection route planning device includes: The risk heat map generation module 201 is used to drive a digital twin model to perform structural response analysis on the environmental data collected by the UAV and generate a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. The initial path construction module 202 is used to construct a path based on the structural risk heat map using a weighted coverage path planning algorithm, and generate the initial inspection path of the transmission tower. The target path generation module 203 is used to control the UAV to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, it triggers local path replanning corresponding to the target anomaly to generate a target inspection path. The model parameter update module 204 is used to control the UAV to perform inspections based on the target inspection path, obtain inspection data of the transmission tower, and update the digital twin model based on the inspection data to obtain the updated digital twin model.

[0123] The transmission tower inspection path planning device provided in this application, employing the transmission tower inspection path planning method in the above embodiments, can solve the technical problem of insufficient accuracy in the safety management of transmission tower structures. Compared with the prior art, the beneficial effects of the transmission tower inspection path planning device provided in this application are the same as those of the transmission tower inspection path planning method provided in the above embodiments, and other technical features in the transmission tower inspection path planning device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0124] This application provides a transmission tower inspection path planning device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the transmission tower inspection path planning method in the above embodiment 1.

[0125] The following is for reference. Figure 3 The diagram illustrates a structural schematic of a transmission tower inspection path planning device suitable for implementing embodiments of this application. The transmission tower inspection path planning device in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The transmission tower inspection route planning device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0126] like Figure 3As shown, the transmission tower inspection route planning device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the transmission tower inspection route planning device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the transmission tower inspection route planning equipment to communicate wirelessly or wiredly with other equipment to exchange data. Although the figure shows transmission tower inspection route planning equipment with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0127] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0128] The transmission tower inspection path planning device provided in this application, employing the transmission tower inspection path planning method described in the above embodiments, can solve the technical problem of insufficient accuracy in the safety management of transmission tower structures. Compared with the prior art, the beneficial effects of the transmission tower inspection path planning device provided in this application are the same as those of the transmission tower inspection path planning method provided in the above embodiments, and other technical features of this transmission tower inspection path planning device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0129] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0130] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0131] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the transmission tower inspection path planning method in the above embodiments.

[0132] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0133] The aforementioned computer-readable storage medium may be included in the inspection path planning equipment of the transmission tower; or it may exist independently and not be installed in the inspection path planning equipment of the transmission tower.

[0134] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the transmission tower inspection path planning device, the transmission tower inspection path planning device: based on environmental data collected by a UAV, drives a digital twin model to perform structural response analysis on the environmental data, generating a structural risk heat map of the transmission tower, wherein the digital twin model is generated based on the structural information of the transmission tower; based on the structural risk heat map, constructs a path using a weighted coverage path planning algorithm to generate an initial inspection path for the transmission tower; controls the UAV to perform structural defect detection based on the initial inspection path, and when a target anomaly is detected, triggers local path replanning corresponding to the target anomaly to generate a target inspection path; controls the UAV to perform inspection based on the target inspection path to obtain inspection data of the transmission tower, and updates the digital twin model based on the inspection data to obtain an updated digital twin model.

[0135] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0137] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0138] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described transmission tower inspection path planning method, which can solve the technical problem of insufficient accuracy in the safety management of transmission tower structures. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the transmission tower inspection path planning method provided in the above embodiments, and will not be repeated here.

[0139] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the transmission tower inspection path planning method described above.

[0140] The computer program product provided in this application can solve the technical problem of insufficient accuracy in the safety management of transmission tower structures. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the transmission tower inspection path planning method provided in the above embodiments, and will not be repeated here.

[0141] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for planning inspection routes for transmission towers, characterized in that, The method for planning inspection routes for the transmission towers includes: Based on environmental data collected by drones, a digital twin model is driven to perform structural response analysis on the environmental data, generating a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. Based on the structural risk heat map, a weighted coverage path planning algorithm is used to construct the initial inspection path for the transmission tower. The drone is controlled to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, a local path replanning corresponding to the target anomaly is triggered to generate a target inspection path. The drone is controlled to perform inspections based on the target inspection path to obtain inspection data of the transmission tower. The digital twin model is then updated based on the inspection data to obtain an updated digital twin model. The step of updating the digital twin model based on the inspection data to obtain the updated digital twin model includes: Point cloud data and image data are extracted from the inspection data. The point cloud data is registered and fused to extract the measured displacement of key nodes. The image data is reconstructed in three dimensions to extract the cross-sectional change characteristics of the components. The measured displacement of key nodes refers to the displacement difference between the actual three-dimensional position of the key node of the transmission tower under load and the predicted three-dimensional position of the digital twin model under the same load. The cross-sectional change characteristics of the components refer to the change between the geometric dimensions or apparent features of the component cross-section and the previous inspection value. Using the measured displacement of the key node and the characteristic quantity of the component cross-section change as observation values, the Bayesian data assimilation method is used to perform posterior estimation and parameter correction on the current state vector of the digital twin model, resulting in a digital twin model with corrected model parameters. The model prediction value is calculated based on the digital twin model after the model parameters are corrected. The residual between the model prediction value and the measured value is calculated. When the residual is less than a preset threshold, the model parameters are corrected to obtain the updated digital twin model.

2. The method for planning inspection routes for transmission towers as described in claim 1, characterized in that, The step of using environmental data collected by UAVs to drive a digital twin model to perform structural response analysis on the environmental data and generate a structural risk heat map of the transmission tower includes: The drone acquires real-time environmental data of the power transmission tower, and constructs a combined load vector based on the environmental data. The combined load vector is input into the digital twin model for nonlinear solution, and the stress ratio of each component is extracted from the solution results. The stress ratios are classified according to preset hazard level thresholds to generate a structural risk heat map covering the entire tower.

3. The method for planning inspection routes for transmission towers as described in claim 2, characterized in that, The environmental data includes wind speed data, wind direction data, and temperature data. The step of constructing a combined load vector based on the environmental data includes: Based on the shape coefficient and wind-receiving area of ​​each component in the digital twin model, the wind speed data and the wind direction data are converted to obtain the wind load vector; Based on the temperature data, the axial temperature deformation of each component in the digital twin model caused by the temperature difference is calculated to obtain the temperature load vector; The combined load vector is obtained by superimposing the wind load vector, the temperature load vector, and the structural self-weight load vector.

4. The method for planning inspection routes for transmission towers as described in claim 1, characterized in that, The step of constructing the initial inspection path for the transmission tower based on the structural risk heat map using a weighted coverage path planning algorithm includes: The risk level of each component is extracted from the structural risk heat map. Based on the risk level, a corresponding path planning weight coefficient is determined for each component. The risk level is divided into high-risk, medium-risk and low-risk levels according to the relationship between the component stress ratio and a preset threshold. The initial inspection path of the transmission tower is obtained by weighting the path planning weight coefficients using the weighted shortest coverage path algorithm.

5. The method for planning inspection routes for transmission towers as described in claim 1, characterized in that, The step of triggering local path replanning corresponding to the target anomaly and generating a target inspection path includes: Obtain the defect location information of the target anomaly, and generate a local reshoot path based on the defect location information within a preset range around the target anomaly; The initial inspection path that has not yet been executed is updated based on the local reshoot path to obtain the target inspection path.

6. The method for planning inspection routes for transmission towers as described in claim 4, characterized in that, The step of constructing the initial inspection path for the transmission tower based on the structural risk heat map using a weighted coverage path planning algorithm further includes: When the number of transmission towers is greater than one, the comprehensive health parameters of each transmission tower are obtained. The comprehensive health parameters include the maximum stress ratio of components, the ratio of the maximum displacement of nodes to the design limit, the normalized value of service life, and the normalized value of the historical cumulative number of defects. A weighted calculation is performed based on the comprehensive health parameters to obtain a comprehensive health score for each transmission tower. The comprehensive health score is used to determine the inspection priority of each transmission tower.

7. A transmission tower inspection route planning device, characterized in that, The inspection route planning device for the transmission tower includes: The risk heat map generation module is used to drive a digital twin model to perform structural response analysis on environmental data collected by UAVs and generate a structural risk heat map of the transmission tower. The digital twin model is generated based on the structural information of the transmission tower. The initial path construction module is used to construct the initial inspection path of the transmission tower based on the structural risk heat map using a weighted coverage path planning algorithm. The target path generation module is used to control the UAV to perform structural defect detection based on the initial inspection path. When a target anomaly is detected, it triggers local path replanning corresponding to the target anomaly to generate a target inspection path. The model parameter update module is used to control the UAV to perform inspections based on the target inspection path, obtain inspection data of the transmission tower, and update the digital twin model based on the inspection data to obtain the updated digital twin model. The model parameter update module is further configured to extract point cloud data and image data from the inspection data, register and fuse the point cloud data to extract the measured displacement of key nodes, and perform three-dimensional reconstruction of the image data to extract the component cross-sectional change characteristics. The measured displacement of key nodes refers to the displacement difference between the actual three-dimensional position of the key node of the transmission tower under load and the predicted three-dimensional position of the digital twin model under the same load. The component cross-sectional change characteristics refer to the change in the geometric dimensions or apparent features of the component cross-section compared to the previous inspection value. Using the measured displacement of the key node and the characteristic quantity of the component cross-section change as observation values, the Bayesian data assimilation method is used to perform posterior estimation and parameter correction on the current state vector of the digital twin model, resulting in a digital twin model with corrected model parameters. The model prediction value is calculated based on the digital twin model after the model parameters are corrected. The residual between the model prediction value and the measured value is calculated. When the residual is less than a preset threshold, the model parameters are corrected to obtain the updated digital twin model.

8. A transmission tower inspection route planning device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the transmission tower inspection path planning method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the transmission tower inspection path planning method as described in any one of claims 1 to 6.

Images