Full-process intelligent analysis system for evaluating damage of ultrahigh-voltage power transmission tower structure

Abstract

The invention belongs to the field of power transmission lines, and provides a whole-process intelligent analysis system for ultrahigh-voltage power transmission tower structure damage assessment. Comprising an unmanned aerial vehicle data acquisition module, a point cloud preprocessing and component identification module, a parametric modeling and damage calibration module, a finite element mechanical simulation analysis module, an AI intelligent defect judgment and trend prediction module and an automatic report generation module. According to the method, the unmanned aerial vehicle and the laser radar are combined with the self-adaptive spiral route, so that no-blind-area scanning of the whole structure of the iron tower is achieved, the point cloud data precision is superior to that of traditional manual measurement, micro damage such as angle steel bending and fine cracks can be captured, and millimeter-level precision input is provided for modeling; meanwhile, modeling can be rapidly completed through a point cloud automatic processing flow, the efficiency is improved compared with manual drawing modeling, and the evaluation period is greatly shortened.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power transmission lines, and particularly relates to a full-process intelligent analysis system for damaged evaluation of an ultrahigh-voltage power transmission tower structure. BACKGROUND

[0002] With the rapid development of ultrahigh-voltage AC and DC power transmission projects in China, a large number of 500kV and above ultrahigh-voltage power transmission lines are deployed in complex environments such as plateaus, mountainous areas, snow and wind. As a key load-bearing structure, the stability and strength of the tower are crucial to the safe operation of the line.

[0003] Currently, tower structure mechanics analysis mainly relies on manual inspection combined with design drawings for statics calculation, but the following technical problems exist: 1. Manual inspection is low in efficiency and difficult to cover wide-area towers, especially in post-disaster investigation, which is slow and has the risk of subjective misjudgment; 2. The structure evaluation model is strong in statics, and most of them are designed and checked based on the "perfect state", which is difficult to evaluate the structural stress changes under local damage (such as angle steel bending and loose connection); 3. There is a lack of data-driven modeling methods, and current unmanned aerial vehicle point cloud and three-dimensional modeling have not been widely used in structural simulation analysis; 4. The existing simulation analysis method lacks universality and cannot quickly adapt to multiple tower types and different damage modes.

[0004] For example, the document "Simulation Analysis of Mechanical Properties of Power Transmission Tower" (Power System Protection and Control, 2019) proposes an evaluation method based on ANSYS static simulation, but does not involve unmanned aerial vehicle modeling, AI analysis or emergency suggestions; and another example is "Research on Modeling Method of Power Transmission Tower Based on Unmanned Aerial Vehicle Point Cloud Data" (Power Grid Technology, 2021), which focuses on geometric modeling and does not deeply analyze the structural response.

[0005] Therefore, there is no complete system integrating "unmanned aerial vehicle modeling-structure simulation-AI analysis-emergency suggestions", and it is urgent to propose an analysis model and application framework with better adaptability and intelligence. SUMMARY

[0006] To solve the above technical problems, the application provides a full-process intelligent analysis system for damaged evaluation of an ultrahigh-voltage power transmission tower structure to solve the problems in the prior art.

[0007] A full-process intelligent analysis system for damaged evaluation of an ultrahigh-voltage power transmission tower structure, comprising: An unmanned aerial vehicle data acquisition module, which surrounds the ultra-high voltage transmission tower through a preset flight path to acquire multi-view data of the tower, surface three-dimensional point cloud data and corresponding high-resolution two-dimensional image data; A point cloud preprocessing and component identification module, which is in communication connection with the unmanned aerial vehicle data acquisition module, is configured to preprocess the acquired three-dimensional point cloud data, and perform component clustering to identify each component of the tower based on the preprocessed point cloud data. A parameterized modeling and damage calibration module, which is in communication connection with the point cloud processing module, is configured to automatically construct a finite element structure model of the tower according to the component geometric parameters and damage characteristics output by the point cloud processing module, automatically set material parameters of each component, and automatically calculate and apply a stiffness reduction factor of a damaged component to achieve fine adjustment of the properties of the damaged part. A finite element mechanics simulation analysis module, which is in communication connection with the parameterized modeling and damage calibration module, is configured to import the finite element structure model into a preset finite element solver for static analysis. An AI intelligent defect judgment and trend prediction module, which is in communication connection with the finite element mechanics simulation analysis module, is configured to receive structural response parameters output by the finite element mechanics simulation analysis module. A report automatic generation module, which is in communication connection with the AI intelligent defect judgment and trend prediction module, is configured to automatically generate a structure safety analysis report including a three-dimensional tower body damage visualization diagram, a simulation stress nephogram, a component-level risk grade table, and specific emergency operation suggestions according to the judgment and prediction results of the AI intelligent defect judgment and trend prediction module, and efficiently link with a power grid dispatching platform and an emergency repair system through a preset interface.

[0008] Preferably, the unmanned aerial vehicle data acquisition module includes at least one laser radar sensor and at least one image sensor, and the three-dimensional point cloud data and two-dimensional image data are used to represent the tower component morphology, identify local damage characteristics, and quantify structural deformation information.

[0009] Preferably, the point cloud preprocessing and data preprocessing of the component recognition module include voxel filtering and statistical outlier rejection; the point cloud preprocessing and data preprocessing of the component recognition module further perform component clustering to identify each truss component of the tower based on the preprocessed point cloud data, perform RANSAC algorithm on the identified each truss component to accurately fit the component main axis and cross-section geometric parameters; and further based on the geometric parameters and the local damage features determined with the aid of the two-dimensional image data, extract the axis, start and end nodes, length, direction vector, and damage type and degree of each component, and perform parameterized three-dimensional reconstruction of the damaged tower structure.

[0010] Preferably, the static analysis accurately calculates the node displacement, member internal force, and overall structure safety factor of the tower under the condition of considering the composite load combination, and outputs the stress concentration information of the key dangerous component.

[0011] Preferably, the structure response parameters include displacement peak value, stress distribution, member internal force, damage type, and stiffness reduction rate; the AI intelligent defect judgment and trend prediction module integrates an AI defect judgment model based on an XGBoost classifier, which is used for intelligent judgment of structure safety according to the structure response parameters, and fine division of risk levels, the risk levels including safe, early warning, and dangerous; the AI intelligent defect judgment and trend prediction module integrates a trend prediction model based on an LSTM model, which is used for accurately predicting the stress trend of the tower structure in a future period of time based on historical stress sequence data, and comprehensively evaluating the remaining life of the tower.

[0012] Preferably, the flight path preset strategy of the preset flight path in the unmanned aerial vehicle data acquisition module includes: intelligently generating an adaptive spiral surrounding flight path according to the actual height of the tower, which is used to comprehensively cover the lateral structure surface of the tower body; and performing horizontal supplementary shooting path at a specific height and angle to capture detailed image and point cloud information of complex local components such as cross arms and stay wires; the final data output format of the unmanned aerial vehicle data acquisition module includes three-dimensional point cloud data in.las or.pcd format, and digital image files containing high-precision geographic coordinates and unmanned aerial vehicle attitude information; Preferably, in the preprocessing and component identification module, the voxel size of the voxel filtering is set to 0.05 meters; the number of neighborhood points of the statistical outlier removal is set to 50, and the standard deviation ratio is set to 1.0; in the execution of the component region clustering algorithm, the DBSCAN clustering algorithm is adopted, the clustering radius eps is set to 0.2 meters, and the minimum point number min_points is set to 20; for each identified truss component region, the RANSAC algorithm is used for straight line or plane fitting, the distance threshold of the RANSAC algorithm is set to 0.02 meters, the number of random sampling points ransac_n is set to 3, and the maximum number of iterations num_iterations is set to 1000.

[0013] Preferably, the model data generated by the automatic construction of the truss finite element structure model of the power transmission tower is in the format of ANSYS.cdb or OpenSees.tcl grid file; the pre-set finite element solver can be OpenSees, ANSYS or ABAQUS; the material model of the truss finite element structure model adopts an elastic-plastic constitutive model, and the mechanical performance indicators of Q345 steel are specifically set as the material parameters of the steel, including a Poisson's ratio of 0.3 and an elastic modulus of 2.1 x 10 11 Pa; the load combination considered in the static analysis further includes applying a base hinged boundary condition at the bottom of the power transmission tower, and superimposing wind load, icing load and conductor tension.

[0014] Preferably, in the AI intelligent missing judgment and trend prediction module, the input feature vector of the AI missing judgment model constructed based on the XGBoost classifier includes node displacement peak value, rod stress distribution characteristic value, component damage type and its stiffness reduction rate, rod geometry, material grade and environmental load parameters; the XGBoost classifier intelligently judges the structural safety of the power transmission tower and divides the structural risk level, according to the judged risk level, the AI missing judgment model further outputs the corresponding emergency operation suggestion, and the operation suggestion includes structural reinforcement of local components, temporary reduction of line load, or immediate power-off alarm instruction.

[0015] Preferably, the output content of the automatically generated structural safety analysis report includes: a three-dimensional damaged state visualization diagram of the ultra-high voltage power transmission tower, which clearly shows the location and type of damaged components; a stress nephogram generated by the finite element mechanics simulation analysis module, which intuitively displays the stress concentration area and dangerous components; a detailed component-level risk level list, which lists the identification code, damage condition, risk level and corresponding suggestion of each key component; and specific emergency operation suggestions generated based on the AI analysis results.

[0016] Compared with the prior art, the present application has the following beneficial effects: 1、The present application realizes full-structure blind area scanning of the tower by adopting M300RTK unmanned aerial vehicle + DJI L2 laser radar (precision ±2cm, point density >300pts / m 2 ), combined with adaptive spiral flight path, the point cloud data precision is better than that of traditional manual measurement (the manual measurement error is usually >5cm), and micro-damage such as angle steel bending and slight crack can be captured, thereby providing millimeter-level precision input for modeling; at the same time, through the point cloud automatic processing flow (filtering, clustering, fitting), the modeling of a 500kV strain tower can be completed in 8 minutes, which is more than 200 times higher in efficiency than manual drawing modeling (traditional manual modeling needs 2-3 days), and the evaluation period is greatly shortened.

[0017] 2、The present application avoids subjective misjudgment of artificial inspection (the artificial misjudgment rate is usually >15%) by fusing XGBoost missing judgment model and finite element simulation, taking displacement, stress, damage and other multi-dimensional characteristics as input, and the risk level judgment accuracy is >92%; for example, in the evaluation of the wind-damaged tower, hidden risks such as “bending moment of top main rod exceeding limit” can be accurately identified, compared with the ability of artificial to only identify obvious deformation (such as inclination and fracture), the comprehensive evaluation from surface to deep layer is realized, the stress trend in the next 5 days is predicted through the historical stress sequence, the remaining life is evaluated combined with fatigue theory, and the limitation of traditional evaluation that only focuses on the current state and has no long-term prediction is solved. For example, in the case of ±800kV strain tower, the system accurately predicts that the remaining life is less than 90 days, which provides sufficient time for power grid dispatching to develop load limiting and replacement plan, and avoids sudden outage accidents; the actual measurement prediction deviation is better than ±8%, which meets the demand of engineering practice for prediction accuracy.

[0018] 3、The present application supports multiple terrain environments such as plateau, mountainous area and seaside by built-in typical ultra-high voltage tower model library of NGS (including 500kV and above straight tower and strain tower), can automatically match modeling parameters and load combination according to the type of tower (such as adding corrosion load correction for seaside tower and adjusting wind load coefficient for plateau tower), solves the problem of traditional method that is single-tower type and difficult to adapt to different scenes, the application range covers main ultra-high voltage tower types in China, the finite element module supports mainstream solvers such as OpenSees, ANSYS and ABAQUS, and the report module outputs multiple format files such as PDF / Word / GeoJSON / DXF, which meets the use habits of different users (such as ANSYS for design units and PDF for dispatching departments); at the same time, through the linkage of RESTfulAPI and dispatching and emergency system, the “data island” is broken, the evaluation results are quickly transformed into dispatching instructions and repair actions, and the response time is <30s, which is much faster than artificial information transmission (usually needs several hours).

[0019] 4. This invention achieves 5-8 times the inspection efficiency of manual labor through drone inspections (manual inspections average 5-8 towers per day, while drones average 30-40 towers per day), and reduces the transportation and manpower input required for manual inspections in remote areas. Simultaneously, AI-powered automatic analysis and report generation eliminates the time required for manual report compilation (manual reporting for one tower takes 2-3 hours, while automatic generation takes less than 30 seconds), significantly reducing maintenance labor costs. After disasters (such as typhoons and blizzards), it can quickly complete wide-area tower inspections, avoiding the omission of potential hazards due to slow manual inspections. Through "risk warning + remaining lifespan prediction," it achieves a shift from "post-disaster repair" to "pre-disaster prevention," reducing large-scale power outages caused by sudden tower failures. Attached Figure Description

[0020] Figure 1 A system module structure diagram of a full-process intelligent analysis system for assessing structural damage to ultra-high voltage transmission towers; Figure 2 This is a system module flowchart of a full-process intelligent analysis system for assessing structural damage to ultra-high voltage transmission towers. Detailed Implementation

[0021] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0022] As attached Figure 1 To be continued Figure 2 As shown: This invention provides a full-process intelligent analysis system for assessing structural damage to ultra-high voltage transmission towers, including a UAV data acquisition module. The UAV data acquisition module circles the ultra-high voltage transmission tower via a preset flight path to collect data from multiple perspectives, thereby obtaining three-dimensional point cloud data of the tower's surface and corresponding high-resolution two-dimensional image data. The UAV data acquisition module includes at least one lidar sensor and at least one image sensor, and the three-dimensional point cloud data and two-dimensional image data are used to characterize the shape of the tower components, identify local damage features, and quantify structural deformation information. The flight path preset strategy includes: intelligently generating an adaptive spiral-shaped flight path based on the actual height of the tower to fully cover the lateral structural surfaces of the tower; and performing horizontal supplementary shooting at specific heights and angles to capture detailed images and point cloud information of complex local components such as crossarms and guy wires; the final data output format of the UAV data acquisition module includes 3D point cloud data in .las or .pcd format, as well as digital image files containing high-precision geographic coordinates and UAV attitude information; The point cloud preprocessing and component identification module is in communication connection with the unmanned aerial vehicle data acquisition module, and is configured to preprocess the acquired three-dimensional point cloud data; and based on the preprocessed point cloud data, each truss component of the tower is identified through component clustering; the data preprocessing of the point cloud preprocessing and component identification module includes voxel filtering and statistical outlier elimination; the point cloud preprocessing and component identification module further performs component clustering to identify each truss component of the tower based on the preprocessed point cloud data, performs RANSAC algorithm on the identified each truss component to accurately fit the component main axis and cross-section geometric parameters; and further based on the geometric parameters and the local damage features determined with the aid of the two-dimensional image data, the axis, start and end nodes, length, direction vector, damage type and degree of each component are extracted, and the parametric three-dimensional reconstruction of the damaged tower structure is performed; The point cloud preprocessing and component identification module is further configured to, when preprocessing the acquired three-dimensional point cloud data, the voxel size of the voxel filtering is preferably set to 0.05 meters; the number of neighborhood points of the statistical outlier elimination is set to 50, and the standard deviation ratio is set to 1.0; when performing the component region clustering algorithm, the DBSCAN clustering algorithm is adopted, the clustering radius eps is set to 0.2 meters, and the minimum point number min_points is set to 20; for the identified each truss component region, the RANSAC algorithm is adopted for straight line or plane fitting, the distance threshold of the RANSAC algorithm is set to 0.02 meters, the number of randomly sampled points ransac_n is set to 3, and the maximum iteration number num_iterations is set to 1000; The parametric modeling and damage calibration module is in communication connection with the point cloud processing module of the parametric modeling and damage calibration module, and is configured to automatically build the truss finite element structure model of the tower according to the component geometric parameters and damage features output by the point cloud processing module, and automatically set the material parameters (elastic modulus, Poisson's ratio and cross-section characteristics) of each component, and automatically calculate and apply the stiffness reduction factor of the damaged component according to the extracted damage type and degree, so as to realize the fine adjustment of the properties of the damaged parts; The model data format generated by automatically constructing the finite element structure model of the truss of the power transmission tower is ANSYS.cdb format or OpenSees.tcl grid file, and the fine adjustment of the damage site attribute is specifically realized in the following manner: when it is identified that the component has a damage type, the system automatically calculates and applies a stiffness reduction factor K_damage according to a preset damage evaluation rule, the calculation manner of which is: K_damage=K_original*α, wherein K_original is the original stiffness of the component, and α is a damage reduction coefficient; α is preset according to the damage type, when the damage type is slight bending deformation, α is preferably set to 0.85; when the damage type is slight crack, α is preferably set to 0.5; when the damage type is partial fracture or severe deformation of the component, α is preferably set to 0.1; the preset finite element solver can be selected from OpenSees, ANSYS or ABAQUS; the material model of the finite element structure model of the truss adopts an elastic-plastic constitutive model, and the mechanical performance indexes of Q345 steel are specifically set as the material parameters of the steel, including a Poisson's ratio of 0.3 and an elastic modulus of 2.1×10 11 Pa; the load combination considered in the static force analysis further includes: applying a base hinged boundary condition at the bottom of the power transmission tower, and superimposedly considering wind load (according to local wind speed and wind pressure standards), icing load (according to regional icing level) and conductor tension (considering conductor tension under operating conditions).

[0023] The finite element mechanics simulation analysis module is in communication connection with the finite element modeling module, and is used for importing the damage finite element structure model into a preset finite element solver for static force analysis; the static force analysis accurately calculates the node displacement, the internal force of the rod and the overall structural safety factor of the tower under the condition of considering the composite load combination (such as self-weight load, wind load, icing load and conductor tension), and outputs the stress concentration information of the key dangerous component; The AI intelligent missing judgment and trend prediction module is in communication connection with the simulation analysis module and is configured to receive the structural response parameters output by the finite element mechanics simulation analysis module; the structural response parameters include displacement peak value, stress distribution, internal force of a rod, damage type, and stiffness reduction rate; the AI intelligent missing judgment and trend prediction module integrates an AI missing judgment model constructed based on an XGBoost classifier, is configured to intelligently judge the safety of a structure according to the structural response parameters, and finely divide a risk level, the risk level including safety, early warning, and danger; the AI intelligent missing judgment and trend prediction module integrates a trend prediction model constructed based on an LSTM model, is configured to accurately predict the stress trend of a tower structure in a future period of time based on historical stress sequence data, and comprehensively evaluate the residual life of the tower; in the AI intelligent missing judgment and trend prediction module, the input feature vector of the AI missing judgment model constructed based on the XGBoost classifier includes node displacement peak value, rod stress distribution characteristic value, component damage type and stiffness reduction rate, rod geometry, material grade, and environmental load parameters; the XGBoost classifier intelligently judges the safety of a transmission tower structure and divides a structural risk level, the risk level including “safety”, “early warning”, and “danger”; according to the judged risk level, the AI missing judgment model further outputs corresponding emergency operation suggestions, the operation suggestions including structural reinforcement of local components, temporary reduction adjustment of line load, or issuing an immediate power-off alarm instruction; in the AI intelligent missing judgment and trend prediction module, the trend prediction model constructed based on the LSTM model, the input sequence being stress peak value data of key components of a transmission tower recorded at the past 10 consecutive time points in historical monitoring data; the LSTM model can predict the stress change trend curve of the transmission tower components in the future 5 days based on the historical sequence, to provide forward-looking information for preventive maintenance; the AI intelligent missing judgment and trend prediction module further utilizes the stress prediction trend curve, combines material fatigue cumulative damage theory, performs structural fatigue trend analysis, and quantitatively evaluates the residual safe operation life of the transmission tower, to provide data support for equipment replacement plan and operation and maintenance strategy; The report automatic generation module is in communication connection with the AI intelligent missing judgment and trend prediction module, and is used for automatically generating a structural safety analysis report containing a three-dimensional tower body damage visualization graph, a simulation stress cloud map, a component level risk grade table, and specific emergency operation suggestions such as reinforcement suggestions, load adjustment suggestions or immediate power-off alarm) according to the judgment result and the prediction result of the AI intelligent missing judgment and trend prediction module, and efficiently linking with the power grid dispatching platform and the emergency repair system through a preset interface; the output content of the automatically generated structural safety analysis report includes: a three-dimensional damaged state visualization graph of the extra-high voltage power transmission tower, which clearly shows the position and type of the damaged component; a stress cloud map generated by the simulation analysis module, which directly displays the stress concentration area and the dangerous component; a detailed component level risk grade list, which lists the identification code, damage condition, risk grade and corresponding suggestions of each key component; and specific emergency operation suggestions generated based on the AI analysis result, the output format of the report includes PDF format, Word format, GeoJSON format or DXF format, and the report generation and response suggestion module links with the power grid dispatching platform and the emergency repair system through a preset interface, aiming to realize the real-time sharing and quick response mechanism of the power transmission tower structure safety information; specifically, the linkage mechanism is realized by constructing a RESTful API interface, ensuring that the evaluation results and emergency suggestions can be exchanged and transmitted in a standardized and programmable manner with the existing power grid dispatching system and emergency command system; the data exchange content includes tower ID, damage position, risk level, suggestion operation code and geographic location information; the instruction transmission includes dispatching instructions, repair task allocation or emergency power-off request, thereby improving the power grid operation safety and repair efficiency.

[0024] The specific operation process is: The unmanned aerial vehicle data acquisition module: M300RTK unmanned aerial vehicle is selected as the flight platform, which has the characteristics of long endurance time, high flight stability and strong anti-interference ability, and can adapt to flight tasks in complex environments such as highlands, mountains and snowstorms; equipped with DJI L2 laser radar, which has an accuracy of ±2cm and a point density of more than 300pts / m 2, which can clearly capture the subtle morphology and damage characteristics of the tower components, and is equipped with a Sony 7R camera for obtaining high-resolution images to assist in texture mapping and damage identification of the point cloud data; The spiral wrap + horizontal fill flight path based on the height of the tower body is adopted. Before flight, the system automatically calculates the radius of spiral wrap, height interval, and the number and angle of horizontal fill according to the design height of the tower, actual geographic location information, to ensure that the UAV can scan the tower in all directions and avoid data collection blind area; For example, for a tower with a height of 80m, the radius of the spiral wrap flight path gradually increases from 10m at the bottom to 15m at the top, the height interval is set to 5m, and the horizontal fill is performed every 30° to ensure that the point cloud data can completely cover each component of the tower; The collected data is output in the form of point cloud files in.las / .pcd format and image files containing EXIF GPS records. The point cloud file records the three-dimensional coordinate information of each point of the tower, and the EXIF GPS record contains the geographic location, shooting angle, time, etc. Information when each image is taken, which is convenient for subsequent geographic positioning and coordinate system calibration of point cloud data.

[0025] Point cloud preprocessing and component recognition module: Processing flow: Python language is used to call open3d library and numpy library to realize the processing and analysis of point cloud data, and the specific process is as follows: Point cloud reading: read the point cloud file in.las / .pcd format through the o3d.io.read_point_cloud(“tower.pcd”) function, and load the point cloud data into the system; Voxel filtering: In order to reduce the amount of point cloud data and improve the efficiency of subsequent processing, voxel filtering algorithm is adopted, through pcd.voxel_down_sample(voxel_size=0.05) function, the point cloud data is down-sampled according to the voxel size of 0.05 meters, and the redundant data is removed under the premise of retaining the main features of the point cloud; Outlier rejection: Due to the influence of environmental noise (such as dust, flying birds, light reflection, etc.) during point cloud collection, there are some outliers, through the pcd_down.remove_statistical_outlier(nb_neighbors=50, std_ratio=1.0) function, the distance mean and standard deviation of each point and its 50 neighborhood points are calculated, the points with distance mean greater than 1.0 times the standard deviation are judged as outliers and removed, to improve the purity of the point cloud data; Clustering extraction component area: using the DBSCAN (density-based spatial clustering of applications with noise) algorithm, through the np.array(pcd_clean.cluster_dbscan(eps=0.2, min_points=20)) function, set the neighborhood radius to 0.2 meters and the minimum point number to 20, cluster the points with similar density into different areas, and each area corresponds to a component of the tower, realizing the preliminary separation of the components; RANSAC fitting main component: using the RANSAC (random sample consensus) algorithm, through the pcd_clean.segment_plane(distance_threshold=0.02, ransac_n=3, num_iterations=1000) function, set the distance threshold to 0.02 meters, the random sampling point number to 3, and the iteration number to 1000, perform plane or straight line fitting on the clustered component area, extract the key parameters of the component such as axis, start and end nodes, length, and direction vector, and provide accurate geometric information for subsequent parameterized modeling; Parameter output: this module finally outputs the axis equation of each component, the three-dimensional coordinates of the start and end nodes, the length of the component, and the direction vector, which will be directly used in the parameterized modeling and damage modeling module to build the finite element model of the tower; Parameterized modeling and damage calibration module: According to the structural characteristics of the ultra-high voltage transmission tower, generate an ANSYS.cdb or OpenSees.tcl format finite element grid model; during modeling, the system will automatically select the corresponding element type according to the type of the component (such as main material, auxiliary material, inclined material, node, etc.), for example, for main material and auxiliary material, beam elements are used for simulation; for nodes, rigid or hinged elements are used for simulation to ensure that the model can accurately reflect the actual structure form and stress characteristics of the tower; Damage modeling logic: the system predefines the correspondence between damage types and stiffness reduction coefficients, and realizes automatic assignment of stiffness reduction coefficients through a user-defined function, which is as follows: def compute_stiffness_reduction(damage_type): if damage_type == "none": return 1.0 elif damage_type == "bend_light": return 0.85 elif damage_type == "crack_minor": return 0.5 elifdamage_type=="fracture": return 0.1 else: return0.7#default Stiffness reduction formula: in For the original stiffness of the member, This is the damage reduction factor.

[0026] The software interface for the finite element mechanics simulation analysis module can use solvers such as OpenSees, ANSYS, and ABAQUS; Material model: Elastic-plastic constitutive model, steel Load combinations: Wind load: Referring to the "Code for Design of Overhead Transmission Lines" (GB50545-2010), based on parameters such as the basic wind pressure, terrain type, and tower height of the area where the tower is located, the wind load calculation formula W=μ s μ z W0 (where W is the standard value of wind load, μ) s Let μ be the wind load shape coefficient. z The wind load is calculated using the wind pressure height variation coefficient (W0 is the basic wind pressure) and applied to the finite element model according to the distribution along the tower height to simulate the wind conditions of the tower under different wind speeds.

[0027] Icing load: Based on the icing thickness data provided by the local meteorological department, the icing load is calculated according to the formula for calculating the icing load of a circular cross-section member: qb=27.727×10−3b(d+b) (where qb is the icing load per unit length, b is the icing thickness, and d is the member diameter). This icing load is then applied as a uniformly distributed load to the crossarms, conductor suspension points, and other easily iced parts of the tower to reflect the impact of icing on the stress on the tower.

[0028] Conductor tension: Based on parameters such as conductor type, sag, and span, the tension value of the conductor under different working conditions (such as maximum wind, icing, high temperature, etc.) is obtained through conductor mechanics calculations. This value is then applied as a concentrated force to the conductor suspension point of the tower to simulate the tension of the conductor on the tower.

[0029] Simulation Output: After the finite element mechanics simulation analysis is completed, the system outputs the following key results data: Node displacement: This includes the displacement of each node of the tower in the X, Y, and Z directions. Displacement data can be used to determine whether the tower is excessively deformed and to identify dangerous areas with large deformations.

[0030] Member internal forces: Output the axial force, bending moment, shear force and other internal force parameters of each member, clarify the magnitude and distribution of the force on each member, and identify members whose internal forces exceed the limit.

[0031] Safety Factor: Based on the internal forces and allowable stresses of the materials, calculate the strength safety factor and stability safety factor for each member. Members with a safety factor less than 1 are considered unsafe and require special attention and handling. The formula for the safety factor is: ; AI-powered intelligent analysis and prediction module: AI-powered defect detection model: XGBoost Python import xgboostasxgb model = xgb.XGBClassifier() model.fit(X_train, y_train) # Features include displacement, internal force, damage reduction rate, etc. y_pred = model.predict(X_test) Input features: peak displacement, stress distribution, damage type, member length, material grade, etc. Output: Risk level (safe / warning / dangerous), and recommended action plan (power outage, reinforcement, continue operation, etc.).

[0032] LSTM trend prediction model Python fromkeras.modelsimportSequential fromkeras.layersimportLSTM,Dense,Dropout model=Sequential() The model uses peak stress data from the past 10 point cloud acquisitions as its input sequence. This input sequence contains the stress changes of the tower over a period of time, reflecting the stress evolution trend of the tower. By learning from the input sequence, the model outputs a stress prediction trend curve for the next 5 days (the prediction period can be adjusted according to actual needs). The curve visually shows the direction and magnitude of stress changes in the tower over the next few days. Simultaneously, combining the fatigue damage accumulation theory of materials, the model calculates the degree of fatigue damage to the tower components based on the stress prediction trend curve, thereby assessing the remaining safe operating life of the tower and providing a basis for power grid dispatching departments to formulate long-term maintenance plans. To improve the prediction accuracy of the model, new stress monitoring data is periodically added to the training set, and the LSTM model is iteratively trained to update the model parameters, so that the model can adapt to the changes in the stress state of the tower and ensure the accuracy of the prediction results. Through actual measurement verification, the prediction deviation of the model is better than ±8%, which can meet the actual engineering needs; The report generated by the report automatic generation module contains four core contents: three-dimensional tower body damage map, simulated stress cloud map, component-level risk grade table and recommended operation. The three-dimensional tower body damage map visually displays the overall shape and damaged parts of the tower through three-dimensional modeling technology. Users can rotate and zoom in on the view to see the damage in detail. The simulated stress cloud map uses different colors to identify the stress levels in different areas of the tower. The redder the color, the higher the stress, making it easy to quickly locate high-stress areas. The component-level risk grade table lists the number, damage type, safety factor, risk level and recommended operation of each component in a table format, clearly presenting the safety status of each component. The recommended operation section provides systematic maintenance and repair recommendations based on the overall risk level and specific component conditions, with strong operability.

[0033] Output format: Supports PDF, Word, GeoJSON, and DXF output formats to meet different usage needs. PDF format reports have good cross-platform compatibility and stable format, suitable for archiving and sharing. Word format reports are easy for users to edit and modify content according to actual needs. GeoJSON format can import spatial information such as tower location and damaged areas into a geographic information system (GIS), enabling overlay analysis with other spatial data. DXF format is the standard file format for AutoCAD, allowing three-dimensional models of towers to be imported into AutoCAD software for subsequent reinforcement scheme design by designers.

[0034] Dispatch linkage: Through RESTful API (Representational State Transfer Application Programming Interface) interface with power grid control platform and emergency system. When the system generates an evaluation report, it automatically pushes key information (such as tower number, risk level, recommended operation, remaining life, etc.) in the report to the control platform and emergency system through the API interface. The control platform adjusts the power grid operation mode based on the report information, such as limiting load or power off for lines with towers at risk levels of warning or danger. The emergency system generates a repair task list based on the recommended operation scheme, automatically assigns repair personnel, equipment and materials, and improves repair efficiency. Through actual measurement, the report generation and information pushing time is less than 30s, which is much better than the current manual report preparation and information transmission process, and saves valuable time for quickly dealing with tower safety hazards.

[0035] Working principle: Data acquisition stage: The UAV three-dimensional modeling module serves as the data entry point, equipped with high-precision laser radar and cameras, and performs full-range scanning of the ultra-high voltage transmission tower according to the preset flight path of adaptive spiral wrapping + horizontal supplementary shooting. During flight, the laser radar collects real-time three-dimensional point cloud data of the tower, records the spatial coordinates, morphological characteristics, and damage information (such as bending and cracking) of each component of the tower, and the camera simultaneously captures high-resolution images to assist in texture mapping of the point cloud data and verification of damage characteristics. The collected data is stored in the form of.las / .pcd point cloud and image files containing EXIF GPS records, providing raw data support for subsequent processing. The core of this stage is to ensure data coverage completeness and accuracy, laying the foundation for subsequent analysis. Data processing stage: After receiving the collected point cloud data, the point cloud processing and component extraction module executes a standardized processing procedure through Python calling professional libraries. First, voxel filtering is used for downsampling to reduce data volume and improve efficiency, and then statistical outlier rejection is used to remove environmental noise (such as dust and bird interference points), obtaining pure point cloud. Subsequently, the DBSCAN density clustering algorithm is used to separate point clouds of different components based on density differences, achieving preliminary partitioning of components. Finally, the RANSAC random sample consensus algorithm is used to fit component axes, extracting the axis equation, start and end node coordinates, length, and direction vector of each component, converting unordered point cloud into ordered component parameters, and providing accurate geometric input for model construction. Model construction stage: The parametric modeling and damage modeling module automatically generates finite element mesh models in ANSYS.cdb or OpenSees.tcl format based on component parameters, matches built-in tower type library templates according to tower types (such as strain towers and straight towers), and quickly adapts to different tower structures. At the same time, combined with image recognition or manually labeled damage types (such as slight bending, slight cracking, and fracture), the stiffness reduction function is called to assign corresponding reduction coefficients (such as a fracture component coefficient of 0.1), adjust the stiffness of damaged components according to the formula, and reconstruct the nodes in the damaged area, achieving precise modeling in two dimensions of "geometric shape + damage state" and ensuring that the model is highly consistent with the actual state of the tower. Mechanical simulation stage: The finite element mechanical simulation module loads the preset load combination (self-weight, wind load, ice load, and conductor tension), and calls solvers such as OpenSees and ANSYS for mechanical calculation. During simulation, based on the elastic-plastic steel constitutive model, the stress response of the tower under different working conditions is simulated, and node displacement (reflecting overall deformation), internal force of members (axial force, bending moment, etc., reflecting local stress), and safety factor (judging the bearing capacity of components) are output, converting the mechanical state of the tower structure into quantitative data and providing objective basis for intelligent analysis. The core of this stage is to accurately assess the safety performance of damaged towers through professional mechanical calculation. Intelligent analysis stage: After receiving the simulation data, the AI intelligent analysis and prediction module works collaboratively: the XGBoost risk judgment model takes displacement peak, stress distribution, damage type, etc. as features, judges the overall risk level of the tower (safe, warning, dangerous) through ensemble learning algorithm, and generates targeted operation suggestions; the LSTM trend prediction model takes the historical 10 stress peaks as sequence input, captures the stress change law through the gating mechanism, predicts the stress trend curve in the next 5 days, evaluates the remaining life based on the fatigue damage theory, realizes the dual analysis of "current safety judgment + future trend prediction", and improves the systematicness and forward-looking of the evaluation; Reporting and linkage stage: The reporting and dispatching linkage module converts the intelligent analysis results into multi-format reports (PDF / Word / GeoJSON / DXF), visually presents the evaluation conclusions through three-dimensional damage map, stress cloud map and risk table; at the same time, it pushes the key information (risk level, suggestion operation, remaining life) to the power grid control platform and emergency system through RESTful API, and the control platform adjusts the line operation mode (such as load limiting, power off) according to the key information, and the emergency system automatically generates the repair task list, realizes the seamless connection of "evaluation results - dispatching decision - repair action", forms a whole-process closed loop, and finally achieves the goal of quickly identifying hidden dangers, scientifically guiding repair, and ensuring the safety of power grid.

[0036] The embodiments of the present application are given for example and description, although the embodiments of the present application have been shown and described above, it can be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and the person skilled in the art can make changes, modifications, replacements and modifications to the above-mentioned embodiments within the scope of the present application, which should be included in the protection scope of the present application.

Claims

1. A full-process intelligent analysis system for assessing structural damage to ultra-high voltage transmission towers, characterized in that: include: The UAV data acquisition module circles the ultra-high voltage transmission tower via a preset flight path to collect data from multiple perspectives, thereby obtaining three-dimensional point cloud data of the tower's surface and corresponding high-resolution two-dimensional image data. The point cloud preprocessing and component recognition module is communicatively connected to the UAV data acquisition module and is used to preprocess the acquired 3D point cloud data. Based on the preprocessed point cloud data, component clustering is performed to identify each pole component of the tower; The parametric modeling and damage calibration module is communicatively connected to the point cloud processing module. It is used to automatically construct the finite element structural model of the tower's pole system based on the component geometric parameters and damage characteristics output by the point cloud processing module, automatically set the material parameters of each component, and automatically calculate and apply the stiffness reduction factor of the damaged component based on the extracted damage type and degree, so as to achieve fine adjustment of the attributes of the damaged part. The finite element mechanical simulation analysis module is communicatively connected to the parametric modeling and damage calibration module, and is used to import the finite element structural model into a preset finite element solver for static analysis. The AI ​​intelligent defect detection and trend prediction module is communicatively connected to the finite element mechanical simulation analysis module and is used to receive the structural response parameters output by the finite element mechanical simulation analysis module. The automatic report generation module is connected to the AI ​​intelligent defect detection and trend prediction module. It is used to automatically generate a structural safety analysis report based on the judgment and prediction results of the AI ​​intelligent defect detection and trend prediction module. The report includes a three-dimensional visualization of tower damage, a simulated stress cloud map, a component-level risk level table, and specific emergency operation suggestions. It can also be efficiently linked with the power grid dispatching platform and emergency repair system through a preset interface.

2. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The UAV data acquisition module includes at least one lidar sensor and at least one image sensor, and the three-dimensional point cloud data and two-dimensional image data are used to characterize the shape of the tower components, identify local damage features, and quantify structural deformation information.

3. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The data preprocessing of the point cloud preprocessing and component identification module includes voxel filtering and statistical outlier removal. Based on the preprocessed point cloud data, the module also performs component clustering to identify each pole component of the tower. For each identified pole component, the RANSAC algorithm is applied to accurately fit the main axis and cross-sectional geometric parameters. Furthermore, based on the geometric parameters and the local damage features determined by the two-dimensional image data, the module extracts the axis, start and end nodes, length, direction vector, damage type, and degree of each component, and performs parametric three-dimensional reconstruction of the damaged tower structure.

4. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The static analysis, taking into account the combination of composite loads, accurately calculates the nodal displacements, member internal forces, and overall structural safety factor of the tower, and outputs stress concentration information of key dangerous components.

5. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The structural response parameters include peak displacement, stress distribution, member internal forces, damage type, and stiffness reduction rate. The AI ​​intelligent defect detection and trend prediction module integrates an AI defect detection model built based on an XGBoost classifier, which is used to intelligently determine the structural safety based on the structural response parameters and finely classify the risk levels, including safe, warning, and dangerous. The AI ​​intelligent defect detection and trend prediction module also integrates a trend prediction model built based on an LSTM model, which is used to accurately predict the stress trend of the tower structure over a future period based on historical stress sequence data and comprehensively assess the remaining lifespan of the tower.

6. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The flight path preset strategy in the UAV data acquisition module includes: intelligently generating an adaptive spiral-shaped flight path based on the actual height of the tower to fully cover the lateral structural surfaces of the tower; and performing horizontal supplementary shooting at specific heights and angles to capture detailed images and point cloud information of complex local components such as crossarms and guy wires; the final data output format of the UAV data acquisition module includes 3D point cloud data in .las or .pcd format, as well as digital image files containing high-precision geographic coordinates and UAV attitude information.

7. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 3, characterized in that: The point cloud preprocessing and component identification module is further configured to, when preprocessing the acquired 3D point cloud data, preferably set the voxel size of the voxel filter to 0.05 meters; set the number of neighborhood points for statistical outlier removal to 50 and the standard deviation ratio to 1.0; when executing the component region clustering algorithm, the DBSCAN clustering algorithm is used, the clustering radius eps is set to 0.2 meters, and the minimum number of points min_points is set to 20; for each identified rod component region, the RANSAC algorithm is used for line or plane fitting, the distance threshold of the RANSAC algorithm is set to 0.02 meters, the number of random sampling points ransac_n is set to 3, and the maximum number of iterations num_iterations is set to 1000.

8. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The model data format generated by automatically constructing the finite element structural model of the tower's pole system is ANSYS.cdb or OpenSees.tcl mesh file; the preset finite element solver can be OpenSees, ANSYS, or ABAQUS; the material model of the pole system finite element structural model adopts an elastic-plastic constitutive model, and the steel material parameters are specifically set to the mechanical properties of Q345 steel, including Poisson's ratio of 0.3 and elastic modulus of 2.1 × 10⁻⁶. 11 Pa; the load combination considered in the static analysis further includes: applying a base hinge boundary condition at the bottom of the transmission tower, and superimposing wind load, icing load and conductor tension.

9. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 5, characterized in that: In the AI ​​intelligent defect detection and trend prediction module, the AI ​​defect detection model built based on the XGBoost classifier has input feature vectors including peak node displacement, characteristic values ​​of member stress distribution, member damage type and its stiffness reduction rate, member geometry, material grade, and environmental load parameters. The XGBoost classifier intelligently determines the structural safety of the transmission tower and classifies the structural risk level. Based on the determined risk level, the AI ​​defect detection model further outputs corresponding emergency operation suggestions, which include structural reinforcement of local members, temporary reduction and adjustment of line load, or issuing an immediate power outage alarm command.

10. The intelligent analysis system for the entire process of damage assessment of ultra-high voltage transmission tower structures as described in claim 1, characterized in that: The automatically generated structural safety analysis report outputs the following: a three-dimensional visualization of the damage status of the ultra-high voltage transmission tower, clearly showing the location and type of damaged components; a stress cloud map generated by the finite element mechanical simulation analysis module, intuitively displaying stress concentration areas and dangerous components; a detailed list of component-level risk levels, specifying the identification code, damage status, risk level, and corresponding recommendations for each key component; and specific emergency operation recommendations generated based on the AI ​​analysis results.

Images