Offshore wind turbine inspection monitoring system based on digital twinning

Abstract

The application relates to the technical field of offshore wind turbine inspection, and discloses an offshore wind turbine inspection and monitoring system based on digital twinning. A twin data synchronization module of the system collects wind turbine structure state data and environmental hydrological data in real time through a multi-source sensor cluster, and constructs an offshore wind turbine digital twin. An inspection path dynamic generation module analyzes a wind turbine fragile area and risk distribution based on the digital twin, generates a multi-unmanned aerial vehicle (UAV) cooperative inspection path in combination with real-time marine weather forecasts, and dispatches the UAV fleet to execute tasks according to the cooperative inspection path, and dynamically responds to real-time data updated by the twin data synchronization module. A real-time anomaly identification module fuses and analyzes UAV return data, identifies wind turbine defects, outputs abnormal information, calculates relative spatial deviation, generates attitude control instructions, and sends the attitude control instructions to the UAV cooperative control module to realize real-time adjustment of the UAV attitude.

Description

Technical Field

[0001] This invention relates to the field of offshore wind turbine inspection technology, specifically to an offshore wind turbine inspection and monitoring system based on digital twins. Background Technology

[0002] Offshore wind turbines, as crucial equipment for marine renewable energy development, operate in harsh environments characterized by high salt spray, strong winds and waves, and complex hydrology. Their structural integrity and operational stability directly impact energy supply efficiency and operational safety. Currently, offshore wind turbine inspections primarily rely on manual tower climbing and fixed sensor monitoring. However, both methods have significant limitations. Manual tower climbing requires personnel to work at high altitudes, making it highly susceptible to marine weather conditions. Inspections are often interrupted by strong winds, heavy rain, or dense fog, resulting in low efficiency and high safety risks. Furthermore, manual inspections depend on personnel experience and judgment, making it prone to overlooking or misjudging hidden defects such as minute cracks in turbine blades or stress concentrations within the tower, thus failing to meet the monitoring requirements for long-term stable operation of wind turbines.

[0003] While fixed sensor monitoring can achieve continuous monitoring of key components of wind turbines, it suffers from limited monitoring range and single data dimension. Existing fixed sensors are mostly deployed in easily accessible locations such as the nacelle and the bottom of the tower, making it difficult to cover critical and vulnerable areas such as the distal ends of the blades and the top of the tower, resulting in blind spots in the monitoring data. Furthermore, most sensors can only collect single types of data such as vibration and temperature, failing to achieve collaborative analysis of structural status and environmental hydrological data, thus failing to comprehensively reflect the wind turbine's operating status and potential risks. In addition, traditional inspection methods lack dynamic planning and collaborative control capabilities for the inspection process. When using drones for inspection, tasks are often executed according to preset paths, unable to adjust inspection strategies based on the real-time structural status of the wind turbine and changes in marine weather, easily leading to unreasonable inspection paths and wasted drone resources. Simultaneously, the data transmitted by drones lacks effective fusion and analysis methods, making it difficult to quickly and accurately identify wind turbine defects and locate abnormal locations. This results in inspection results failing to guide subsequent maintenance operations in a timely manner, affecting the efficiency of wind turbine fault handling and increasing operating costs.

[0004] With the continuous expansion of offshore wind power installed capacity and the continuous increase in the capacity of a single wind turbine, traditional inspection and monitoring methods are no longer able to meet the needs of large-scale and high-reliability operation and maintenance. There is an urgent need for an integrated inspection and monitoring system that can realize comprehensive monitoring of wind turbines, dynamic inspection planning, efficient collaborative control and accurate anomaly identification, so as to solve the problems of low inspection efficiency, limited monitoring range and insufficient risk identification capability in existing technologies.

[0005] In existing offshore wind turbine drone inspection technologies, the drone's attitude adjustment relies on its own GPS and IMU integrated navigation system, which lacks precise spatial correlation with the defective area of ​​the wind turbine. Offshore winds and waves can easily cause relative spatial deviations between the drone and the defective area. Traditional methods cannot calculate this deviation in real time and dynamically generate attitude control commands, requiring manual remote control adjustments. Manual control suffers from response delays, low efficiency, and low adjustment accuracy. Summary of the Invention

[0006] The purpose of this invention is to provide a digital twin-based offshore wind turbine inspection and monitoring system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a digital twin-based offshore wind turbine inspection and monitoring system, the system comprising:

[0008] Twin data synchronization module: Real-time acquisition of structural status data and environmental hydrological data through a multi-source sensor cluster deployed on the surface of the offshore wind turbine, and construction of a digital twin of the offshore wind turbine by combining the structural status data and the environmental hydrological data;

[0009] Inspection path dynamic generation module: Based on the digital twin of the offshore wind turbine, analyze the distribution of vulnerable areas and potential risks in the wind turbine structure, and combine real-time marine meteorological forecast information to generate a collaborative inspection path for multiple UAVs;

[0010] Unmanned Aerial Vehicle (UAV) Collaborative Control Module: Dispatch a UAV swarm to perform inspection tasks according to the collaborative inspection path, and dynamically receive and respond to real-time data updated by the twin data synchronization module during the inspection process;

[0011] Real-time anomaly identification module: It fuses and analyzes the high-definition images and vibration spectrum data transmitted back by the UAV, identifies surface defects and abnormal vibration modes of the wind turbine blades, outputs anomaly location information and anomaly level signals, calculates relative spatial deviation, generates attitude control commands, and sends them to the UAV collaborative control module to achieve real-time adjustment of the UAV's attitude.

[0012] Preferably, the inspection path dynamic generation module generates a collaborative inspection path for multiple UAVs, including:

[0013] Obtain the set of coordinates of stress concentration areas of the wind turbine tower and blades output by the digital twin of the offshore wind turbine;

[0014] Extract current wind turbine operating power and ocean wind speed and direction data, and calculate the inspection priority weight for each area;

[0015] An initial inspection waypoint sequence is generated based on the inspection priority weight, and conflict resolution and path smoothing are performed on the initial inspection waypoint sequence.

[0016] Based on the drone's endurance and inspection load constraints, the smoothed waypoint sequence is assigned tasks and the flight segment duration is optimized to output the final collaborative inspection path for multiple drones.

[0017] Preferably, the UAV collaborative control module dynamically receives and responds to real-time data updated by the twin data synchronization module, including:

[0018] Real-time acquisition of sudden wind and wave change data and wind turbine emergency shutdown signals pushed by the twin data synchronization module;

[0019] If the sudden change in wind and waves exceeds the wind resistance threshold of the drone, a path replanning command will be triggered and the drone will be controlled to avoid high-risk areas.

[0020] If an emergency shutdown signal for the wind turbine is received, the current inspection task will be immediately terminated and the drone will be controlled to return to a safe hovering coordinate.

[0021] Preferably, the real-time anomaly identification module performs fusion analysis on the high-definition images and vibration spectrum data transmitted back by the UAV, including:

[0022] Multi-scale texture segmentation and defect contour extraction are performed on high-definition images to generate masks of cracks and corrosion areas on the blade surface;

[0023] Wavelet packet decomposition and feature energy extraction are performed on vibration spectrum data to identify abnormal frequency components and modal parameter shifts;

[0024] The surface cracks and corrosion areas of the blade are spatially mapped and fused with the abnormal frequency components to output abnormal location information and abnormal level signals.

[0025] Preferably, the system further includes a twin model update module: correcting the model parameters and degradation prediction curve of the offshore wind turbine digital twin based on the anomaly location information and the anomaly level signal;

[0026] The twin model update module corrects the model parameters and degradation prediction curves of the offshore wind turbine digital twin, including:

[0027] Based on the abnormal location information, locate the geometric model and material property nodes of the corresponding components in the digital twin;

[0028] The anomaly level signal is converted into a model parameter correction value using a Bayesian update algorithm, and the fatigue cumulative damage coefficient of the component is updated.

[0029] The overall life prediction curve and reliability index of the wind turbine are reconstructed based on the updated fatigue cumulative damage coefficient.

[0030] Preferably, the inspection path dynamic generation module also dynamically adjusts the collaborative inspection path based on historical inspection results, including:

[0031] Obtain the historical anomaly distribution heatmap and defect growth trend vector output by the real-time anomaly identification module;

[0032] Extract the coordinates of abnormally dense areas from the heat map and calculate the area re-inspection priority by combining the defect growth trend vector;

[0033] Integrating re-inspection priority into the waypoint sequence generation process increases the inspection frequency and data collection density of high-risk areas.

[0034] Preferably, the UAV collaborative control module also realizes perception data sharing and collaborative positioning among multiple UAVs, including:

[0035] Local point cloud data and image frames collected by each UAV are exchanged in real time through an inter-UAV communication network.

[0036] A distributed fusion algorithm is used to generate a 3D point cloud model of the wind turbine as a whole, and the relative positioning deviation of each UAV is corrected accordingly.

[0037] The waypoint coordinates and flight attitude in the collaborative inspection path are dynamically adjusted based on the corrected positioning results.

[0038] Preferably, the twin model update module also supports human expert intervention and model verification, including:

[0039] Provides a manual parameter correction interface for twin models and a visual editing tool for degradation curves;

[0040] It receives manually input verification data and empirical correction coefficients, and performs consistency verification on the automatically updated model;

[0041] When the automatically updated model conflicts with the results of manual verification, the model arbitration mechanism is triggered and an anomaly report to be confirmed is generated.

[0042] Preferably, the system further includes an extreme weather adaptive module, which performs the following operations:

[0043] Real-time access to regional typhoon warning information and wave forecast data;

[0044] When the forecast level exceeds the system design threshold, the drone is scheduled to return to base in advance and the wind turbine twin is locked and put into read-only mode.

[0045] Once the weather returns to normal, a compensatory inspection path is generated based on the estimated cumulative load of the wind turbines during the shutdown period.

[0046] Preferably, the system further includes an automatic inspection report generation module, which performs the following operations:

[0047] Integrate all anomaly location information and anomaly level signals output by the real-time anomaly identification module;

[0048] By associating the lifetime prediction curve and reliability index generated by the twin model update module, a structured inspection report containing defect distribution maps, risk level assessments, and maintenance recommendations is automatically generated.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] The dynamic inspection path generation module analyzes the distribution of vulnerable areas and potential risks in offshore wind turbine structures based on digital twins of the turbines. Combined with real-time marine weather forecasts, it generates collaborative inspection paths for multiple drones, changing the traditional reliance on preset paths and lack of flexibility in drone inspections. This module can develop differentiated inspection plans based on the vulnerability and risk level of different areas of the wind turbine, as well as real-time changing marine weather conditions. It ensures that drones prioritize inspections of high-risk areas while fully considering the impact of weather factors on drone flight, avoiding inspection interruptions or safety hazards due to unsuitable weather conditions. This improves the rationality of inspection paths and the continuity of inspection operations, reduces ineffective drone flights, and increases the efficiency of inspection resource utilization.

[0051] The UAV collaborative control module schedules a swarm of UAVs to perform inspection tasks according to the collaborative inspection path. During the inspection process, it dynamically receives and responds to real-time data updates from the twin data synchronization module, achieving efficient collaboration and dynamic adjustment of the UAV swarm. Compared to traditional single-UAV inspections or simple formation inspections, this module can coordinate multiple UAVs to operate simultaneously, expanding the inspection coverage and shortening the inspection time. Furthermore, during the inspection process, when the twin data synchronization module pushes new wind turbine status data or environmental data, the UAVs can respond promptly, adjusting their flight paths and inspection priorities based on the new data. This avoids mismatches between the inspection content and the actual state of the wind turbine due to data lag, ensuring that the inspection task always revolves around the wind turbine's current critical needs.

[0052] The real-time anomaly identification module fuses and analyzes high-definition images and vibration spectrum data transmitted from the UAV to identify surface defects and abnormal vibration patterns on the wind turbine blades. It outputs anomaly location information and anomaly level signals. By calculating relative spatial deviations, generating attitude control commands, and sending them to the UAV collaborative control module, it enables real-time adjustment of the UAV's attitude. This function ensures that the UAV always operates in an attitude adapted to the inspection needs of the defect area, avoiding the impact of attitude deviations caused by sea waves on data acquisition quality. High-definition images clearly reveal details of defects such as blade cracks and corrosion, and vibration spectrum data accurately reflects the characteristics of abnormal structural vibrations, providing high-quality data support for anomaly identification. Simultaneously, the attitude adjustment process eliminates the need for manual intervention, reducing response delays and accuracy errors from manual control, improving the attitude control coordination during multi-UAV collaborative inspections, enhancing the system's adaptability to complex marine environments, and ensuring a stable and efficient inspection process. Attached Figure Description

[0053] Figure 1 This is a timing diagram of the offshore wind turbine inspection and monitoring system based on digital twins as described in this invention.

[0054] Figure 2 A flowchart for generating multi-UAV collaborative inspection paths for the inspection path dynamic generation module;

[0055] Figure 3 This is a flowchart for the real-time anomaly identification module to integrate and analyze data. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1This invention provides a digital twin-based offshore wind turbine inspection and monitoring system. The system integrates multi-source sensor data, UAV-assisted inspection, and real-time data analysis to achieve dynamic monitoring and intelligent early warning of the offshore wind turbine's structural health. The core of the system comprises four modules: twin data synchronization, dynamic inspection path generation, UAV-assisted control, and real-time anomaly identification. These modules work together to form a closed-loop digital twin-driven inspection process. The twin data synchronization module is deployed on the surface of the offshore wind turbine, using a multi-source sensor cluster to collect structural status and environmental hydrological data. After preprocessing and fusion, a high-fidelity digital twin of the offshore wind turbine is constructed based on a three-dimensional geometric model, integrating physical properties and real-time operating parameters to dynamically reflect the actual state of the wind turbine. Compared to traditional fixed sensors that can only collect single types of data, this method achieves comprehensive acquisition of key data on wind turbine operation. Furthermore, based on this data, a digital twin of the offshore wind turbine is constructed, accurately mapping the physical turbine's operating status to the virtual space. This breaks the limitation of traditional monitoring where physical entities and data are separated, allowing staff to intuitively grasp the overall operating status of the turbine in a virtual environment, avoiding biased judgments due to incomplete data or fragmented information. The dynamic inspection path generation module, based on the digital twin, analyzes the distribution of vulnerable areas and potential risks in the turbine structure, while simultaneously integrating real-time marine weather forecasts to generate multi-UAV collaborative inspection paths covering high-risk areas. The UAV collaborative control module receives path scheduling data from a fleet of drones to execute tasks. The drones are equipped with high-definition cameras and vibration sensors, receiving real-time data during flight to achieve adaptive trajectory adjustments.

[0058] Example 1: See Figure 2 The digital twin of the offshore wind turbine serves as the data foundation for the entire inspection mission. Its output set of coordinates for stress concentration areas in the turbine tower and blades is the starting point for path planning. This coordinate set originates from the mechanical simulation model operating within the twin. This model dynamically calculates the stress distribution of structural components in a complex marine environment by integrating load data collected from multiple sources in real time. The calculation results are marked on the surface of the wind turbine model in the form of a three-dimensional point cloud. Each point not only contains spatial coordinate information but also metadata such as stress magnitude and variation trend. After obtaining this coordinate set, the dynamic inspection path generation module needs to conduct a comprehensive analysis combining the real-time operating status of the wind turbine and external environmental factors. It reads the current operating power parameters from the wind turbine monitoring system and simultaneously obtains real-time ocean wind speed and direction data from the environmental hydrological sensor cluster. The operating power directly affects the vibration modes of the tower and nacelle, while wind speed and direction determine the magnitude and direction of the aerodynamic loads on the blades. These dynamic variables collectively influence the evolution of structural weak points.

[0059] The module uses a multi-factor weighted evaluation algorithm to calculate the priority weight of each potential inspection area. Variables such as peak stress in stress concentration areas, historical failure rate, current operating power deviation from rated values, and wind speed excitation effects are quantified as weighting factors. These factors are then linearly weighted and summed using a preset weight matrix to calculate the quantified priority score for each area to be inspected. Higher scores indicate greater risk or uncertainty, and these areas should be prioritized for coverage in subsequent path planning. After calculating the weights, the module generates an initial inspection waypoint sequence, following spatial coverage and priority ranking principles. Areas with weights higher than a threshold are included first, and then ordered from highest to lowest weight to form a rough list of flight path points. However, the initial waypoint sequence has issues, such as dense waypoints causing frequent acceleration and deceleration of UAVs, spatiotemporal conflicts in multiple paths, and sharp path turns that are detrimental to flight. Conflict resolution and path smoothing are required. For multi-UAV collaborative scenarios, a spatiotemporal window-based conflict detection algorithm is used to simulate flight trajectories, identify conflict spatiotemporal points, and eliminate conflicts by inserting waiting waypoints, adjusting arrival times, or reassigning waypoints. Path smoothing uses B-spline curve fitting technology to transform the polyline path into a smooth curve, enabling the UAV to fly stably and at a constant speed, which is beneficial for collecting high-quality data.

[0060] After conflict resolution and path smoothing, the module needs to assign the optimized waypoint sequence to each individual drone in the swarm and optimize the segment duration. This step is strictly limited by the endurance and inspection load constraints of each drone. The task allocation problem is a typical constrained optimization problem. The module usually uses heuristic algorithms such as genetic algorithms or particle swarm optimization to solve it. The objective function is to balance the workload of each drone as much as possible and minimize the total task completion time while satisfying the battery constraints of each drone. The allocation result clarifies which drone is responsible for visiting which waypoints and the order of execution. Segment duration optimization focuses on fine control of the flight speed of individual segments. The cruising speed is appropriately increased in longer segments to save time, while the speed is reduced in critical segments approaching the inspection target to ensure the clarity and stability of sensor data acquisition. Finally, a detailed multi-drone collaborative inspection path plan is output. This plan is issued to the drone collaborative control module in the form of a task list, which includes the waypoint coordinates, expected arrival time, suggested flight speed, and inspection action instructions to be executed at each waypoint for each drone.

[0061] After receiving and parsing the final collaborative inspection path plan, the UAV collaborative control module begins scheduling the UAV swarm to perform the inspection task. However, the maritime environment is complex and changeable, and the preset path may become unsafe or no longer optimal due to unforeseen circumstances. Therefore, this module has the ability to dynamically receive and respond to real-time data updated by the twin data synchronization module. The module obtains real-time data on sudden environmental changes pushed by the twin data synchronization module through a dedicated data subscription mechanism. Examples include sudden wind and wave changes monitored by radar or weather stations, and emergency shutdown signals from wind turbine control systems. This data typically carries high-priority tags, triggering the module's immediate response process. When sudden wind and wave changes, especially instantaneous wind speed data, exceed the preset UAV wind resistance threshold, it indicates that the current weather conditions pose a serious threat to UAV flight safety, and the module immediately triggers a path replanning command. The path replanning process does not start entirely from scratch. Instead, it takes the current location and status of the drone as the starting point, combines the latest wind and wave field forecast information, and calls a fast replanning algorithm. This algorithm prioritizes the flight safety of the drone, and its core objective is to control the drone to quickly fly away from high-risk areas, such as strong wind shear zones or wave splash zones. The new temporary path will guide the drone to fly to a safe airspace with relatively small wind and waves to hover and wait, or directly plan the safest path back to the base.

[0062] When the module receives an emergency shutdown signal from the wind turbine control system, it indicates a major malfunction in the wind turbine itself or an impending emergency operation, making the surrounding airspace extremely dangerous. In this situation, the module's response strategy is more decisive: it immediately sends a mission abort command to all drones operating near the wind turbine, regardless of their current stage of the inspection mission. After the abort command is issued, the module uniformly schedules these drones to execute a predefined emergency return-to-home procedure. The core of this procedure is to control all drones to return along the optimal safe path to pre-calculated safe hovering coordinates. These safe coordinates are typically located upwind or crosswind outside the wind turbine's danger zone, maintaining a sufficient safe distance from the wind turbine. Once the drones reach the safe coordinates, they enter a hovering state, awaiting further instructions.

[0063] Example 2: See Figure 3The core task of the real-time anomaly identification module is to fuse and analyze the multimodal sensor data transmitted back by the UAV. Its processing begins with deep analysis of high-definition image data. High-definition cameras mounted on the UAV capture high-resolution image data of the wind turbine blade surface during close-range inspections. These images are transmitted back to the ground control station or edge computing node in real time via a wireless data transmission link. By fusing image and spectrum data, this module can simultaneously capture visible defects on the wind turbine surface and abnormal vibrations in the internal structure, achieving multi-dimensional identification of wind turbine defects and improving the comprehensiveness and accuracy of anomaly identification. Simultaneously, it clearly outputs anomaly location information and anomaly level signals, enabling personnel to quickly pinpoint the fault location and assess its severity, providing clear guidance for subsequent maintenance operations. This avoids maintenance delays or resource misallocation caused by ambiguous fault location or inaccurate level judgment, ensuring that the wind turbine receives timely and targeted maintenance and maintains a stable operating state. The image analysis subsystem first preprocesses the received raw image, including grayscale correction, noise suppression, and image enhancement, to improve image quality and highlight defect features of interest. The preprocessed image is then fed into a multi-scale texture segmentation network based on a deep learning architecture. This network can simultaneously identify the macroscopic contours and microscopic texture changes of the blade at different spatial scales. The texture segmentation results divide the blade image region into different material and state categories, such as normal coating areas, suspected crack areas, corrosion spot areas, and background interference areas. Based on the completed region segmentation, the defect contour extraction algorithm begins its work. It accurately delineates the boundary contours of defects from the segmented regions using an active contour model or edge connection algorithm, ultimately generating a binary mask of crack and corrosion areas on the blade surface. In this mask image, white pixels represent defect areas, and black pixels represent normal areas, while also recording the geometric feature descriptors of each defect area.

[0064] Another data stream in parallel processing is for vibration spectrum analysis. Vibration sensors installed on key parts of the wind turbine and on drones continuously collect dynamic response signals of the structure. These time-domain signals, after anti-aliasing filtering and digitization, are sent to the vibration analysis subsystem for wavelet packet decomposition. Wavelet packet decomposition is a signal processing method with higher resolution than traditional wavelet transform. It can uniformly and completely decompose the vibration signal across the entire frequency band, generating a set of subdivided frequency band signals covering low to high frequencies. Each subdivided frequency band signal carries information about the vibration characteristics of the structure in different frequency ranges. After decomposition, the system calculates the characteristic energy values ​​of each frequency band signal and constructs a characteristic energy vector representing the current structural state. By comparing this real-time characteristic vector with the normal state baseline characteristic vector stored in the digital twin, the system can identify frequency components with abnormally high or low energy. These abnormal frequency components are often closely related to fault modes such as local damage, loose connections, or material degradation of the structure. At the same time, the system also monitors the offset of structural modal parameters (such as frequency, damping ratio, and mode shape) relative to the baseline model. Modal parameter offset is a sensitive indicator of changes in the overall stiffness or mass distribution of the structure.

[0065] After the aforementioned image and vibration analyses produce results, the data fusion stage becomes crucial. The purpose of fusion is to combine the advantages of visual inspection and dynamic response analysis to improve the accuracy and reliability of defect identification. The fusion process begins with spatial mapping. Since the image mask and vibration sensor data originate from different spatial coordinate systems, a unified coordinate system needs to be established. This coordinate system is typically based on the three-dimensional digital model of the wind turbine. Using known camera parameters and sensor installation positions, the defect contours extracted from the images are accurately mapped to corresponding positions on the surface of the wind turbine's three-dimensional model. Simultaneously, the abnormal frequency components identified by the vibration signals also need to be associated with specific nodes on the model based on the physical location of the sensors. Based on spatial alignment, the system performs confidence fusion, assigning a confidence score to the defect identification results from image analysis and the anomaly diagnosis results from vibration analysis. This score is calculated based on the historical accuracy of the detection method, the current signal quality, and the significance of the anomaly features. Then, information fusion algorithms such as DS evidence theory are used to combine evidence from different sources. Finally, a comprehensive judgment is made on whether an anomaly exists in each detected area and the type of anomaly. The output includes anomaly location information containing precise three-dimensional coordinates and a quantified anomaly level signal. The anomaly level is usually divided into multiple levels according to the severity of the defect, the development speed, and the potential risk.

[0066] The digital twin update module begins operation after receiving the anomaly location information and anomaly level signal from the real-time anomaly identification module. Its primary task is to accurately locate the corresponding component geometric model and material property nodes in the vast model database of the digital twin based on the three-dimensional coordinates contained in the anomaly location information. Digital twins are typically organized using a hierarchical data structure, such as whole machine-subsystem-component-subcomponent-finite element mesh node. The location algorithm quickly queries the smallest geometric unit containing the anomaly point using spatial indexing techniques (such as octrees or KD trees) and associates it with the corresponding material properties, current stress state, historical damage accumulation records, and other model parameters. After location is completed, the module uses the Bayesian update algorithm, a probabilistic statistical tool, to process the anomaly level signal and convert it into a quantitative correction of the model parameters. The core idea of ​​Bayesian update is to use newly acquired observation data (i.e., the anomaly level signal) to update the probability distribution of prior knowledge of model parameters, thereby obtaining the posterior probability distribution. Specifically, the system treats anomaly level signals as observational evidence of the current health status of a component. This evidence is probabilistically correlated with certain key model parameters reflecting the degree of component degradation (such as elastic modulus and fatigue damage coefficient). Using a pre-established observation likelihood function, the Bayesian update algorithm calculates the posterior distribution of the model parameters, and its mean or maximum posterior probability value is used as the parameter correction amount for the model. This process is particularly used to update the component's cumulative fatigue damage coefficient, a key parameter for predicting the material's lifespan under cyclic loading. The magnitude of the correction is closely related to the anomaly level; higher-level anomalies typically imply a larger correction magnitude for the damage coefficient.

[0067] Based on the updated fatigue cumulative damage coefficient, the module initiates the reconstruction process of the overall wind turbine life prediction curve and reliability indicators. The fatigue cumulative damage coefficient is input into a physics-based life prediction model, which typically considers the material's SN curve (stress-life curve) and the actual load spectrum experienced by the wind turbine. It calculates the probability that a component will reach a critical damage state at a future point in time through integration. The reconstructed life prediction curve graphically displays the remaining life of the wind turbine or a specific component over time. Simultaneously, the system calculates a set of reliability indicators, such as failure probability, reliability function, and mean time between failures (MTBF). These indicators provide quantitative data support for wind turbine maintenance decisions and risk management.

[0068] After outputting anomaly location information and anomaly level signals, the real-time anomaly identification module uses the three-dimensional coordinate system of the offshore wind turbine digital twin as a reference, combined with the real-time position and attitude data transmitted by the UAV and the three-dimensional coordinates of the defect, to establish a spatial positional relationship model between the UAV and the defect area. It calculates the relative spatial deviation between the two, encompassing lateral distance, longitudinal distance, and angular deviation. Subsequently, based on this deviation and the UAV's flight performance parameters, it generates attitude control commands according to a preset attitude control algorithm. These commands include specific yaw, pitch, and roll angle adjustment parameters. The attitude control commands are sent to the UAV collaborative control module via a dedicated data transmission link. This module parses the commands and converts them into control signals recognizable by the UAV's actuators, driving the actuators to adjust the UAV's attitude. During the adjustment process, the UAV collaborative control module receives UAV attitude feedback data in real time and synchronizes it to the real-time anomaly identification module. The real-time anomaly identification module recalculates the relative spatial deviation; if the deviation exceeds a preset threshold, it regenerates commands until the UAV's attitude meets the inspection requirements.

[0069] Example 3: The inspection path dynamic generation module has the ability to self-optimize based on historical inspection results. Its optimization process begins with in-depth mining and characterization of historical anomaly data. The module extracts all anomaly events recorded in each inspection from the output log of the real-time anomaly identification module. Each event includes anomaly type, precise three-dimensional spatial coordinates, severity level, and timestamp. These discrete anomaly events are imported into a spatial analysis engine, which generates a historical anomaly distribution heatmap on the surface of the wind turbine's three-dimensional model using a kernel density estimation algorithm. The color depth or numerical value of each pixel on the heatmap represents the combined density of the frequency and severity of anomalies occurring at that location and in its surrounding area throughout history. High-frequency, high-severity areas are presented in warm colors (such as red), while low-frequency areas are presented in cool colors (such as blue). The module also tracks the evolution of specific anomalies. For the same type of anomaly (such as a crack) that is detected in multiple consecutive inspections, the system records the sequence of changes in its size and severity level over time, and uses linear regression or exponential smoothing to fit a defect growth trend line. The slope vector of this trend line is the defect growth trend vector, the direction of the vector indicates the direction of defect development (such as along the blade span or chord), and the modulus represents the rate of defect expansion.

[0070] After obtaining the historical anomaly distribution heatmap and defect growth trend vector, the module needs to calculate the re-inspection priority for each region. This is a crucial step in transforming historical information into a guide for future action. Re-inspection priority ( The calculation of anomalies comprehensively considers both the static accumulation and dynamic development of anomalies, and its calculation relationship can be expressed as:

[0071] ,

[0072] in: Representing the The re-examination priority score for each assessment area. It is the average thermal value of the corresponding location on the historical anomaly distribution heat map of the region, which reflects the density and severity of historical anomalies in the region. It is the defect growth trend index of the region, which is obtained by the magnitude-weighted average of all defect growth trend vectors passing through the region, and reflects the activity level and development speed of anomalies in the region. and These are the thermal value weight and the growth trend weight. These two weight coefficients need to be set using domain knowledge or optimization algorithms to balance the impact of historical accumulation and current development on the urgency of re-inspection.

[0073] After calculating the re-inspection priority score for each region, these scores are incorporated into the generation process of the new round of collaborative inspection paths. When generating the initial inspection waypoint sequence, the module no longer relies solely on the real-time stress analysis results provided by the digital twin, but instead uses the re-inspection priority as an important weighting factor. For regions with high re-inspection priority, the system allocates more inspection waypoints to them in the waypoint sequence, meaning that the UAV will conduct more detailed scanning flights in that region, collecting denser images and vibration data. Simultaneously, the system may instruct the UAV to reduce its flight speed and switch sensor operating modes (such as enabling a higher resolution camera mode) in that region, thereby improving the density and quality of data acquisition. Furthermore, the inspection frequency of high-priority regions is also increased. When planning periodic inspection tasks, these regions may be included in the mandatory list for each inspection, while low-priority regions may adopt a rotational inspection strategy, covering them only once every one or more inspections. To achieve accurate collaborative inspection, the UAV collaborative control module establishes a mechanism for sharing perception data and collaborative positioning among multiple UAVs. During the execution of inspection tasks, the UAVs do not operate in isolation. Through an established inter-drone communication network (which can employ adhoc networks or 5G-based slicing technology), drones exchange their collected local perception data in near real-time. This data includes local 3D point cloud data blocks generated by LiDAR or photogrammetry, as well as key image frames captured by high-definition cameras. The data exchange is accompanied by precise timestamps and the drones' initial positioning information (from the GPS / IMU integrated navigation system).

[0074] The exchanged local point cloud data is fed into a distributed point cloud fusion algorithm for processing. This algorithm does not rely on a central processing node but is completed collaboratively within the drone swarm through negotiation. The algorithm first registers point cloud data from different perspectives with overlapping areas, using algorithms such as iterative nearest point to accurately locate their spatial transformation relationships. Then, all point cloud data is transformed to the same global coordinate system, and denoising, thinning, and fusion are performed to ultimately generate a more complete 3D point cloud model covering a larger area or even the entire structure of the wind turbine. This collaboratively generated overall 3D point cloud model serves as a shared, high-precision spatial reference frame, with accuracy superior to the position information obtained by a single drone relying solely on its own GPS. Based on this shared high-precision point cloud model, each drone can correct its relative positioning deviation. The drone matches its own real-time sensor-collected local point cloud with the shared overall point cloud to calculate a positioning correction. This correction significantly eliminates the impact of GPS positioning errors, IMU drift errors, and wind-induced trajectory deviations on positioning accuracy. The corrected positioning results enable each drone to have a more accurate understanding of its own and its teammates' positions in global space. Based on this corrected positioning information, the drone collaborative control module dynamically adjusts the waypoint coordinates and flight attitude in the pre-planned collaborative inspection path. For example, when a drone is found to be laterally deviating from the planned path, the module will send control commands to fine-tune its waypoint coordinates to return to the correct track. When multiple drones are required to perform stereoscopic photography of the same area, the module will precisely control the flight attitude of each drone (such as yaw angle and pitch angle) to ensure optimal coordination of sensor field of view, thereby completing a high-quality data acquisition task.

[0075] Example 4: The system provides an interactive digital twin management interface. Engineers can directly access the underlying parameters of the digital twin through this interface. The interface includes a manual correction interface for model parameters, allowing experts to directly adjust key model parameters, such as the material fatigue limit of a specific blade, the coating degradation rate coefficient, or the wear rate constant of a gearbox, by filling in a form or dragging a slider. Simultaneously, the system provides a degradation curve visualization and editing tool, presenting the component's remaining life curve predicted by the automatically updated algorithm in the form of an interactive chart. Experts can intuitively adjust the shape of the entire degradation trajectory by dragging key data points (such as the predicted failure time point) on the chart based on their experience. The system backend quantifies these graphical adjustments as changes to the corresponding mathematical model parameters.

[0076] A specific application scenario occurred at a coastal wind farm. Its digital twin system, based on UAV inspection data and vibration analysis, automatically identified an abnormal vibration signal at the root of a blade on a certain wind turbine. The real-time anomaly identification module classified it as a "Level 2 anomaly." The twin model update module then automatically ran a Bayesian update algorithm, increasing the blade's cumulative fatigue damage coefficient from 0.15 to 0.28 and revising its predicted remaining life from 96 months to 74 months. However, a senior engineer on-site, reviewing the report, based on past maintenance records for this blade model and his own experience, believed that the probability of severe root cracks in this type of blade at this operational age was extremely low. The current abnormal vibration was more likely due to a slight deviation in the counterweight installed during the previous maintenance, a problem of aerodynamic imbalance rather than structural damage. Therefore, the engineer manually corrected the system's automatically updated cumulative fatigue damage coefficient from 0.28 back to 0.20, entering "Historical data shows that fatigue damage development for this type of blade is slow at this stage; the current vibration characteristics are more consistent with mass imbalance" in the correction reason field. He accessed the degradation curve editing tool and smoothly adjusted the predicted remaining life curve from 74 months to 84 months, which better aligned with his experience in handling such cases. After the manual correction, the system triggered a consistency verification process. This process logically compared the manually entered correction data with the original sensor data and historical fault database records used for automatic updates to check for obvious contradictions, such as whether the manually set life value far exceeded the known maximum life limit of this type of blade. In this case, the verification result was "passed" because the 84-month life prediction was still within a reasonable range. The system also included an arbitration mechanism to handle conflicting opinions. Suppose in another case, the system automatically detected a high gearbox temperature accompanied by a specific frequency sideband. The automatic update model significantly lowered the gearbox's reliability index, predicting that it urgently needed maintenance. However, one expert, based on the oil analysis report, believed the condition was acceptable and only slightly increased the risk level. Another expert might completely agree with the system's automatic judgment, thus creating a conflict. The system will trigger a model arbitration mechanism, which typically employs a weighted voting algorithm or a confidence-based weighted average method. The historical accuracy of each expert participating in the arbitration is converted into a weight value, while the certainty of the current judgment is used as the confidence level. The system integrates this information to arrive at an arbitration result and generates an anomaly report that clearly lists the automatic judgment result, the opinions of each expert, the arbitration result and its reasons. This report is then submitted to a higher-level technical leader for final decision-making, and the entire decision-making chain is recorded for subsequent model optimization.

[0077] The extreme weather adaptive module operates independently and proactively, prioritizing the inspection mission. Its core responsibility is to ensure the safety of system assets (primarily UAVs and digital twin data) under severe sea conditions. The module accesses typhoon warnings, wave forecasts, and thunderstorm monitoring information from authoritative regional meteorological agencies in real time via secure data interfaces. This data typically includes key parameters such as the typhoon's center location, path, speed, central pressure, maximum wind speed radius, and radius of the 7-level wind circle, as well as wave elements such as significant wave height and spectral peak period at the forecast point. When the module analyzes the forecast data and determines that a certain forecast indicator (such as maximum wind speed or significant wave height) exceeds the system's preset design threshold (e.g., wind speed exceeding the wind resistance level or wave height exceeding the operational limits of the UAV recovery platform), it immediately initiates an emergency response procedure. The first step of the procedure is to pre-arrange for all UAVs performing inspection missions to return immediately. The system sends the highest priority instruction to the UAV collaborative control module, requiring all UAVs, regardless of their current mission status, to unconditionally execute the emergency return procedure and choose the fastest and safest path back to the mother ship or shore-based platform. The second step of the program is to lock the digital twins of all affected wind turbines in the typhoon field and switch their status to "read-only mode." The twins can still receive and display real-time monitoring data from the turbine's own sensors, but any automatic or manual write operations such as updating model parameters or correcting lifetime prediction curves are prohibited. This is to prevent incorrect updates to the model in the event that the data may be distorted due to extreme weather interference, and also to maintain a "data snapshot" of the system state before the extreme event, for post-event comparative analysis.

[0078] Following the passage of extreme weather, the module needs to assess the impact of the shutdown and schedule restorative inspections. Since the wind turbines are shut down during the typhoon, but the structure actually bears enormous dynamic loads, the module estimates the accumulated fatigue damage to key components of the turbine during this period based on historical meteorological data recorded during the shutdown (such as wind speed time-history curves and wave time-history curves) and the turbine's load calculation model. Subsequently, the system generates a compensatory inspection route plan, which specifically focuses on areas most vulnerable to damage under extreme loads, such as specific chordal positions of the blades, welded areas of the tower, and foundation connection sections. The level of detail and priority of this inspection is higher than that of routine inspections, aiming to quickly identify potential damage that may have been caused by this extreme event (see Table 1).

[0079] Table 1: Comparison of key model parameters before and after manual correction:

[0080]

[0081] Example 5: The core function of the automatic inspection report generation module is to integrate scattered, multi-source data streams into a clear and comprehensive structured document. This module automatically starts after each collaborative inspection task cycle or immediately upon receiving a manual trigger command. The module first extracts all data records related to the current inspection task from the system database. Its primary data source is all anomaly location information and anomaly level signals output by the real-time anomaly identification module. This information is stored in the form of structured data entries. Each record contains a unique anomaly identifier, anomaly type description (e.g., "surface crack," "corrosion pit," "loose bolt"), precise three-dimensional spatial coordinates (relative to the wind turbine foundation coordinate system), anomaly size measurement (length, area), severity level (e.g., represented by L1 to L4, with L4 being the most severe), and a timestamp indicating that the anomaly was identified. The module's data integration engine then categorizes and summarizes this anomaly information according to the wind turbine's physical structure (e.g., blade number, tower segment, nacelle components), forming a preliminary anomaly list.

[0082] After initial integration of anomalous data, the module initiates a data correlation process. The goal is to establish a direct link between anomalous information and long-term health status predictions for the wind turbine. The module accesses the latest lifetime prediction curves and reliability indices generated by the twin model update module through a data interface. This data typically exists in the form of time-series datasets or parametric models, such as the remaining lifetime probability distribution curve of a blade, a function of gearbox reliability over time, or fatigue damage accumulation curves for critical connections. The correlation process is achieved through unique identifiers for wind turbine components. For example, an anomaly identified as a surface crack on blade "B-03" is correlated with the remaining lifetime prediction curve and current fatigue damage coefficient of "B-03" in the twin model. This ensures that the report's description of the crack not only describes its current state but also references the model's predictions for its future development.

[0083] Based on the integrated and correlated data, the module automatically generates a structured inspection report. The report content typically follows the standard industrial equipment inspection report format and is divided into several logical sections. The first important component of the report is a defect distribution map. This map is not a simple schematic diagram, but an interactive visualization view generated based on a high-precision 3D model of the wind turbine. The model surface uses color coding or icon markings to show the location and type of anomalies. For example, red highlighted areas represent cracks, and yellow spots represent corrosion. Users can rotate and zoom the model to examine the distribution of defects from different angles, and clicking on a specific marker will bring up a detailed information card for that anomaly.

[0084] The core of the report is the risk level assessment section. This section doesn't simply list anomalies; instead, it comprehensively analyzes anomaly signals and associated reliability indicators. The system has a pre-built risk matrix model that rates each identified anomaly based on two dimensions: "anomaly severity" and "impact on overall reliability." The rating results are typically presented as "high risk," "medium risk," "low risk," or more granular levels. For example, a blade crack rated L3 might be rated "high risk" if the twin model predicts a significantly shortened remaining life and a rapid decline in reliability. Conversely, minor surface corrosion at L1 might be rated "low risk" if the model shows minimal impact on overall structural reliability. This assessment provides a direct basis for prioritizing maintenance resource allocation. The final part of the report generates maintenance recommendations based on all the above analyses. These recommendations rely on a built-in rule base that maps different anomaly types, risk levels, and model predictions to a series of standardized maintenance action recommendations. The recommendations are specific and actionable. For example, for blade cracks classified as "high-risk," the recommendation might include "arrange a close-up inspection within the next week to confirm the crack depth using penetrant testing." For gearbox vibration anomalies classified as "medium-risk," the recommendation might be "increase vibration monitoring frequency to twice a week and schedule oil analysis next month." For "low-risk" anomalies, the recommendation might be "add to the routine observation list and conduct a focused re-inspection during the next scheduled inspection." After the report is generated, the system will output it in a predefined format (such as PDF or Word) and automatically send it to the designated maintenance engineer or management system via the messaging system, completing the closed loop from data collection to decision support.

[0085] Taking the generation of a specific inspection report for wind turbine No. 08 in a wind farm as an example, this inspection was completed collaboratively by three drones. The real-time anomaly identification module identified a total of 7 anomalies. The automatic report generation module first summarized these anomalies: a surface crack of approximately 15 mm in length was found in the middle of blade A-02 (Level 3); a coating corrosion area of ​​approximately 200 square centimeters was found at the root of blade B-01 (Level 2); an abnormal vibration mode related to bolt loosening was identified at the flange connection in the middle of the tower (Level 2); and the remaining 4 were minor anomalies of Level 1. The module then correlated the data with the twin model update module and found that the model had revised the predicted remaining life of blade A-02 from 120 months to 90 months, with a reliability index of 0.85; the life prediction of blade B-01 did not change significantly. When generating the defect distribution map, the location of the crack in blade A-02 on the 3D model was clearly marked as an orange flashing area. During the risk assessment, the crack in blade A-02 was classified as "high-risk" due to its high severity and the predicted lifespan reduction from the model; the corrosion of blade B-01 and the loosening of tower bolts were classified as "medium-risk." Finally, the maintenance recommendations for the "high-risk" crack in blade A-02 were: "Immediately request a temporary shutdown window, dispatch professional personnel on a maintenance vessel within 48 hours to conduct a re-inspection using a close-range high-definition zoom camera mounted on a drone, and prepare blade repair materials. It is recommended to shorten the next inspection interval for this blade to one month." This report was automatically generated and sent to the wind farm maintenance supervisor's terminal within 15 minutes of the inspection's completion.

[0086] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital twin-based offshore wind turbine inspection and monitoring system, characterized in that, include: Twin data synchronization module: Real-time acquisition of structural status data and environmental hydrological data through a multi-source sensor cluster deployed on the surface of the offshore wind turbine, and construction of a digital twin of the offshore wind turbine by combining the structural status data and the environmental hydrological data; Inspection path dynamic generation module: Based on the digital twin of the offshore wind turbine, analyze the distribution of vulnerable areas and potential risks in the wind turbine structure, and combine real-time marine meteorological forecast information to generate a collaborative inspection path for multiple UAVs; Unmanned Aerial Vehicle (UAV) Collaborative Control Module: Dispatch a UAV swarm to perform inspection tasks according to the collaborative inspection path, and dynamically receive and respond to real-time data updated by the twin data synchronization module during the inspection process; Real-time anomaly identification module: It performs fusion analysis on the high-definition images and vibration spectrum data transmitted back by the UAV, identifies surface defects and abnormal vibration modes of the wind turbine blades, outputs anomaly location information and anomaly level signal, calculates relative spatial deviation, generates attitude control commands, and sends them to the UAV collaborative control module to realize real-time adjustment of UAV attitude. The inspection path dynamic generation module generates collaborative inspection paths for multiple UAVs, including: Obtain the set of coordinates of stress concentration areas of the wind turbine tower and blades output by the digital twin of the offshore wind turbine; Extract current wind turbine operating power and ocean wind speed and direction data, and calculate the inspection priority weight for each area; An initial inspection waypoint sequence is generated based on the inspection priority weight, and conflict resolution and path smoothing are performed on the initial inspection waypoint sequence. Based on the drone's endurance and inspection load constraints, the smoothed waypoint sequence is assigned tasks and the flight segment duration is optimized to output the final collaborative inspection path for multiple drones. The real-time anomaly identification module performs fusion analysis on the high-definition images and vibration spectrum data transmitted back by the UAV, including: Multi-scale texture segmentation and defect contour extraction are performed on high-definition images to generate masks of cracks and corrosion areas on the blade surface; Wavelet packet decomposition and feature energy extraction are performed on vibration spectrum data to identify abnormal frequency components and modal parameter shifts; The surface crack and corrosion area mask of the blade is spatially mapped and fused with the abnormal frequency component to output the abnormal location information and abnormal level signal. It also includes a twin model update module: correcting the model parameters and degradation prediction curve of the offshore wind turbine digital twin based on the anomaly location information and the anomaly level signal; The twin model update module corrects the model parameters and degradation prediction curves of the offshore wind turbine digital twin, including: Based on the abnormal location information, locate the geometric model and material property nodes of the corresponding components in the digital twin; The anomaly level signal is converted into a model parameter correction value using a Bayesian update algorithm, and the fatigue cumulative damage coefficient of the component is updated. The overall life prediction curve and reliability index of the wind turbine are reconstructed based on the updated fatigue cumulative damage coefficient.

2. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 1, characterized in that, The UAV collaborative control module dynamically receives and responds to real-time data updated by the twin data synchronization module, including: Real-time acquisition of sudden wind and wave change data and wind turbine emergency shutdown signals pushed by the twin data synchronization module; If the sudden change in wind and waves exceeds the wind resistance threshold of the drone, a path replanning command will be triggered and the drone will be controlled to fly away from high-risk areas. If an emergency shutdown signal for the wind turbine is received, the current inspection task will be immediately terminated and the drone will be controlled to return to a safe hovering coordinate.

3. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 1, characterized in that, The inspection path dynamic generation module also dynamically adjusts the collaborative inspection path based on historical inspection results, including: Obtain the historical anomaly distribution heatmap and defect growth trend vector output by the real-time anomaly identification module; Extract the coordinates of abnormally dense areas from the heat map and calculate the area re-inspection priority by combining the defect growth trend vector; Integrating re-inspection priority into the waypoint sequence generation process increases the inspection frequency and data collection density of high-risk areas.

4. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 2, characterized in that, The UAV collaborative control module also enables perception data sharing and collaborative positioning among multiple UAVs, including: Local point cloud data and image frames collected by each UAV are exchanged in real time through an inter-UAV communication network. A distributed fusion algorithm is used to generate a 3D point cloud model of the wind turbine as a whole, and the relative positioning deviation of each UAV is corrected accordingly. The waypoint coordinates and flight attitude in the collaborative inspection path are dynamically adjusted based on the corrected positioning results.

5. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 1, characterized in that, The twin model update module also supports human expert intervention and model verification, including: Provides a manual parameter correction interface for twin models and a visual editing tool for degradation curves; It receives manually input verification data and empirical correction coefficients, and performs consistency verification on the automatically updated model; When the automatically updated model conflicts with the results of manual verification, the model arbitration mechanism is triggered and an anomaly report to be confirmed is generated.

6. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 1, characterized in that, The system also includes an extreme weather adaptive module, which performs the following operations: Real-time access to regional typhoon warning information and wave forecast data; When the forecast level exceeds the system design threshold, the drone is scheduled to return to base in advance and the wind turbine twin is locked and put into read-only mode. Once the weather returns to normal, a compensatory inspection path is generated based on the estimated cumulative load of the wind turbines during the shutdown period.

7. The offshore wind turbine inspection and monitoring system based on digital twin as described in claim 5, characterized in that, The system also includes an automatic inspection report generation module, which performs the following operations: Integrate all anomaly location information and anomaly level signals output by the real-time anomaly identification module; By associating the lifetime prediction curve and reliability index generated by the twin model update module, a structured inspection report containing defect distribution maps, risk level assessments, and maintenance recommendations is automatically generated.

Images