Non-destructive testing analysis system for wind turbine towers

Abstract

The present application relates to the technical field of wind power equipment state monitoring and nondestructive testing analysis, in particular to a nondestructive testing analysis system for a wind power tower, which comprises: a multi-modal data acquisition module for receiving nondestructive testing feature data of preset components such as blades, main shafts and tower cylinders, and preprocessing and extracting defect feature data with local spatial coordinates; a spatial mapping and fusion module for mapping the defect feature data to a global three-dimensional coordinate system to generate digital twin model data containing global defect distribution; a load coupling analysis module for calculating dynamic load transmission links and stress coupling data in combination with real-time operating condition data; a risk assessment and prediction module for determining structural failure risk levels and outputting comprehensive maintenance strategies; the present application realizes the transition from single-point defect discovery to system failure risk identification.

Description

Technical Field

[0001] This invention relates to the field of wind power equipment condition monitoring and non-destructive testing analysis technology, specifically to a non-destructive testing analysis system for wind turbine towers. Background Technology

[0002] Currently, non-destructive testing and analysis mechanisms for wind turbine towers and related force-transmitting components are usually based on independent testing of single components such as blades, main shafts, flange connection areas, or tower welds. When the defect data received by different testing terminals differ in format, coordinate representation, and acquisition timing, existing solutions cannot perform unified mapping, fusion analysis, and load coupling judgment of multi-source defect data in combination with real-time operating conditions. Each assessment may only focus on whether a single defect exceeds the standard, reducing the accuracy of identifying overall structural failure risks and making maintenance decisions. Summary of the Invention

[0003] The purpose of this invention is to provide a non-destructive testing and analysis system for wind turbine towers, addressing the following technical problems: existing technologies typically evaluate the qualification of individual wind turbine components in isolation, lacking a system-level failure risk identification and continuous force transmission path coupling analysis mechanism across components; this invention aims to transform dispersed multi-source non-destructive testing results into a unified analysis object oriented towards the overall safety of the turbine, combining dynamic operating loads to identify the superimposed consequences of multi-source minor defects on the continuous force transmission path, and can find maintenance entry points that meet preset evaluation conditions through source intervention calculations, thereby achieving comprehensive maintenance decisions across components.

[0004] The objective of this invention can be achieved through the following technical solutions: A non-destructive testing and analysis system for wind turbine towers, comprising: A multimodal data acquisition module is used to receive non-destructive testing feature data of preset components constituting the load transmission link of a wind turbine, and to preprocess the non-destructive testing feature data to extract defect feature data with local spatial coordinates. The preset components include at least blades, main shafts and towers. The spatial mapping and fusion module is used to map the defect feature data to the global three-dimensional coordinate system based on the local spatial coordinates according to the preset three-dimensional structural model, so as to generate digital twin model data containing global defect distribution. The load coupling analysis module is used to acquire real-time operating condition data of the wind turbine through the configured external device communication interface, calculate the dynamic load transmission link in combination with the digital twin model data, and calculate the stress coupling data of the defect feature data under dynamic load based on the dynamic load transmission link. The risk assessment and prediction module is used to determine the structural failure risk level based on the stress coupling data, and output a comprehensive maintenance strategy according to the structural failure risk level.

[0005] In one possible implementation, the multimodal data acquisition module includes: The receiving service unit is used to receive non-destructive testing data of blades, non-destructive testing data of main shaft, non-destructive testing data of flange bolts, and non-destructive testing data of tower welds as the non-destructive testing feature data. The format conversion unit is used to convert the nondestructive testing feature data into a unified coordinate format to obtain the defect feature data with the local spatial coordinates.

[0006] In one possible implementation, the spatial mapping and fusion module includes: The coordinate transformation unit is used to obtain the local spatial coordinates of the defect feature data, and convert the local spatial coordinates into global coordinates in the global three-dimensional coordinate system based on the coordinate system transformation matrix jointly generated by the static assembly geometry of each preset component and the instantaneous attitude parameters collected by the sensor. The model generation unit is used to mark the defect feature data with the global coordinates on the preset three-dimensional structural model to generate the digital twin model data.

[0007] In one possible implementation, the load coupling analysis module includes: The aerodynamic load simulation unit is used to calculate asymmetric aerodynamic loads based on the blade non-destructive testing data and the real-time operating condition data in the digital twin model data. The stress field coupling calculation unit is used to simulate the transmission process of the asymmetric aerodynamic load along the dynamic load transmission link based on a preset finite element analysis model, so as to calculate the stress coupling data. The transmission path of the dynamic load transmission link includes blades, main shaft, flange bolts and tower welds.

[0008] In one possible implementation, the stress field coupling calculation unit includes: The frequency comparison sub-unit is used to extract the natural frequency of the main shaft based on the modal analysis of the finite element analysis model, and to calculate the abnormal vibration frequency of the blade based on the periodic fluctuation of the asymmetric aerodynamic load. If the absolute value of the difference between the abnormal vibration frequency of the blade and the natural frequency of the main shaft is lower than the preset frequency threshold, it is determined that a resonance amplification effect has occurred. If the absolute value of the difference is not lower than the preset frequency threshold, it is determined that no resonance amplification effect has occurred. The stress superposition calculation subunit is used to calculate the amplified shear force when the asymmetric aerodynamic load is transmitted to the flange bolts when the resonance amplification effect is determined to occur, and to calculate the local stress concentration peak value as the stress coupling data in combination with the non-destructive testing data of the tower weld; and to calculate the basic shear force when the asymmetric aerodynamic load is transmitted to the flange bolts when the resonance amplification effect is determined not to occur, and to calculate the local stress concentration peak value as the stress coupling data in combination with the non-destructive testing data of the tower weld.

[0009] In one possible implementation, the risk assessment and prediction module includes: The life prediction unit is used to take the stress coupling data as input to a preset life prediction model, and to determine the remaining service life of the whole machine based on the output of the preset life prediction model. The strategy generation unit is used to classify the structural failure risk level according to the remaining service life of the whole machine, and generate the corresponding comprehensive maintenance strategy.

[0010] In one possible implementation, the policy generation unit includes: An intervention calculation subunit is used to select the target component with the largest stress concentration peak as the first component on the dynamic load transfer link when the structural failure risk level is higher than a preset risk threshold, and to simulate and eliminate the defect feature data corresponding to the first component in the digital twin model data. The life re-estimation subunit is used to determine the second component located downstream of the first component in the dynamic load transfer link, and to calculate the extended fatigue life of the second component after eliminating the defect feature data corresponding to the first component. The solution output subunit is used to output the operation of eliminating the defect feature data corresponding to the first component as the comprehensive maintenance strategy; and, if the structural failure risk level is not higher than the preset risk threshold, the above-mentioned operation of simulating the elimination of the defect feature data is not performed, and the comprehensive maintenance strategy of routine inspection is directly output.

[0011] In one possible implementation, the policy generation unit includes: The risk classification unit is used to determine the structural failure risk level as high risk level when the remaining service life of the whole machine is lower than the first service life threshold. If the remaining service life of the entire machine is not less than the first service life threshold and is less than the second service life threshold, the structural failure risk level is determined to be a medium risk level. If the remaining service life of the entire machine is not less than the second service life threshold, the structural failure risk level is determined to be a low risk level; wherein, the first service life threshold is less than the second service life threshold.

[0012] In one possible implementation, the system also includes: The feedback correction module is used to receive the actual execution results of the comprehensive maintenance strategy and calculate the prediction deviation based on the actual execution results; The parameter update module is used to update the weight parameters of the preset life prediction model based on the prediction deviation.

[0013] In one possible implementation, the receiving service unit includes: A standardized interface subunit is used to configure a preset communication protocol bus to receive non-destructive testing feature data from different testing terminals via the communication protocol bus. The testing terminals include UAV testing equipment, portable phased array equipment, and magnetic adsorption wall-climbing robot equipment.

[0014] The beneficial effects of this invention are: 1) This invention receives non-destructive testing data from different terminals through a multimodal data acquisition module and extracts defect features with local spatial coordinates by performing format conversion; this method eliminates the differences in format and coordinate expression of multi-source data, solves the problem that data is difficult to summarize uniformly due to independent testing of each component, and provides homogeneous structural input for subsequent spatial fusion analysis; 2) This invention maps local defects to a global three-dimensional coordinate system through a spatial mapping and fusion module to generate digital twin model data. This method breaks the limitation of isolated evaluation of single-point defects and can reconstruct the relative spatial position and upstream and downstream correlation of multi-source micro defects in the whole system in a unified three-dimensional space, realizing the transformation from single-point defect discovery to system-level analysis. 3) This invention combines real-time operating conditions with digital twin models to calculate dynamic load transmission links and judges resonance amplification effects by frequency comparison; this method effectively combines static defects with dynamic alternating loads, accurately identifies the stress coupling superposition consequences caused by minor defects on the force transmission path, and avoids the deviation of ignoring the real amplified operating conditions or uniformly treating them as the most dangerous situation. 4) This invention inputs stress coupling data into the life prediction model to determine the remaining service life of the entire unit and classifies the structural failure risk level accordingly. This mechanism intuitively transforms the complex underlying physical calculation results into standardized risk levels, providing a unified and objective basis for the scheduling of operation and maintenance resources at the wind farm level, and effectively improving the accuracy and timeliness of maintenance decisions. 5) This invention simulates and eliminates the defects of the component with the largest stress concentration peak on the dynamic load link and re-estimates the extended fatigue life of downstream components. This method can accurately locate the source defects that cause load amplification and find the optimal maintenance entry point through cross-component intervention calculations, avoiding unnecessary intervention caused by non-targeted repairs based solely on surface phenomena, and effectively optimizing the comprehensive maintenance strategy. 6) This invention receives the actual execution results of the maintenance strategy and calculates the prediction deviation through the feedback correction module, and updates the weight parameters of the life prediction model accordingly. This mechanism enables the system to continuously absorb actual on-site maintenance experience and real operation feedback, overcomes the prediction drift caused by long-term operation of equipment in different service environments, and ensures the reliability of the system's long-term risk assessment. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 This is a schematic diagram of the modules of the non-destructive testing and analysis system for wind turbine towers provided in the embodiments of this application. Detailed Implementation

[0017] 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.

[0018] Please see Figure 1 A non-destructive testing and analysis system for wind turbine towers includes: a multi-modal data acquisition module, used to receive non-destructive testing feature data of preset components constituting the load transmission link of the wind turbine, and to preprocess the non-destructive testing feature data to extract defect feature data with local spatial coordinates, wherein the preset components include at least blades, main shafts and towers; The spatial mapping and fusion module is used to map the defect feature data to the global three-dimensional coordinate system based on the local spatial coordinates according to the preset three-dimensional structural model, so as to generate digital twin model data containing global defect distribution. The load coupling analysis module is used to acquire real-time operating condition data of the wind turbine through the configured external device communication interface, calculate the dynamic load transmission link in combination with the digital twin model data, and calculate the stress coupling data of the defect feature data under dynamic load based on the dynamic load transmission link. The risk assessment and prediction module is used to determine the structural failure risk level based on the stress coupling data, and output a comprehensive maintenance strategy according to the structural failure risk level.

[0019] This embodiment provides a non-destructive testing and analysis mechanism for wind turbine towers; specifically, the system is deployed on the operation and maintenance center server of a coastal wind farm to perform whole-machine-level defect fusion analysis on a 3.X MW horizontal axis wind turbine before the typhoon season. The unit operates under conditions of high salt spray, high gusts of wind, and frequent start-stop. The blades, main shaft, flange connection and tower weld are not subjected to independent forces, but form a continuous load transmission link. Therefore, the system does not judge the qualification of a single part in isolation, but focuses on identifying the superposition of multiple small defects in the force transmission path. Specifically, the multimodal data acquisition module receives data uploaded by each detection terminal and performs unified preprocessing; this preprocessing includes at least data timestamp alignment, defect fragment extraction, detection noise removal, and local coordinate attachment. The local spatial coordinates refer to the position of the defect in the coordinate system of its component, such as the distance of the blade root along the spanwise direction, the circumferential angle position of the main shaft, the circumferential angle of the tower weld, and the axial level, etc. The reason for retaining local coordinates is to solve the problem that the original output data of different detection terminals have multiple heterogeneous formats, but physically they all correspond to the specific defect location on a real component. The beneficial effect of the above steps is that the spatial mapping and fusion module further calls the preset three-dimensional structural model to project the local defects of each component into the unified coordinate system of the whole machine. The technical advantage of this approach is that defects originally collected by different teams and different equipment can be reconstructed in the same three-dimensional space to determine their upstream and downstream relationships. For example, whether a certain blade defect is located in the direction that causes a specific main shaft to be stressed, or whether a certain bolt abnormality corresponds to a weak area of ​​a certain tower weld below. The load coupling analysis module then receives real-time operating condition data, which may include wind speed, wind direction, engine speed, pitch angle, yaw angle, power generation, and start / stop status. This module does not simply add up the number of defects, but reconstructs the dynamic load transmission link based on the current working conditions. Its technical mechanism is that the same defect will present different hazards beyond the preset safety tolerance under low wind speed and strong gust conditions. In particular, abnormal blade aerodynamic shape will first change the aerodynamic force distribution, and then transmit it to the tower through the main shaft and flange connection, causing the local weld to enter a fatigue sensitive state under alternating load. The risk assessment and prediction module outputs the structural failure risk level and comprehensive maintenance strategy based on stress coupling data. The comprehensive approach here is not limited to direct repair at the location of the defect, but can also be to reduce the downstream danger by dealing with upstream components, such as repairing local debonding of blades to reduce fatigue of tower welds. It should be further noted that, in order to ensure the consistency of terminology in the description of the embodiments, in the subsequent embodiments of this specification, the flange connection part, flange connection area and bolt connection area all refer to the flange connection area between the main shaft and the tower unless otherwise specified. When it is necessary to emphasize the specific stress-bearing components in the area, the flange bolts are used as the general term. Correspondingly, the tower weld, the weak weld area and the weld and heat-affected zone below it all refer to the welded connection area and its adjacent heat-affected range on the tower structure. They will only be expanded at different granularities when discussing the fine level of the inspection object or the life reassessment object, without changing the technical object to which they belong. As a supplementary implementation, when a component lacks detection data, the system does not directly interrupt the process, but instead marks the component as incompletely observed and reduces the confidence level of the conclusions in subsequent analyses. If a critical component lacks both detection data and real-time operating data, the system only outputs a conservative warning and does not output a detailed maintenance sequence to avoid being misled by incomplete information. If the 3D model version is inconsistent with the equipment modification status, such as using the old model after the blades are replaced, the system will first trigger the model verification process. If the verification fails, the fusion analysis will be paused and only the original detection results will be archived. During the annual major component inspection of the coastal wind farm, a drone discovered a sign of debonding near the root of blade No. 2, a portable phased array detected a microcrack at a certain circumferential position of the main shaft, an abnormal pre-tightening condition of bolt No. 12 was detected in the flange connection area, and a wall-climbing robot detected a small slag inclusion near the circumferential weld of the lower section of the tower. If judged according to the standards for individual components, these results do not exceed the individual scrapping index; however, after the system maps them to the coordinates of the whole machine, it is found that the above defects are located in the same main load transmission direction and show a clear coupling trend under the recent high wind speed operation records. Therefore, a higher structural risk level is given and it is recommended to prioritize the shutdown window. The purpose of this step is to transform the scattered non-destructive testing results into a unified analysis object for the overall safety of the machine, thereby realizing the transformation from single-point defect discovery to system failure risk identification.

[0020] In a preferred embodiment of the present invention, the multimodal data acquisition module includes: a receiving service unit, used to receive non-destructive testing data of blades, non-destructive testing data of main shafts, non-destructive testing data of flange bolts, and non-destructive testing data of tower welds as the non-destructive testing feature data; and a format conversion unit, used to convert the non-destructive testing feature data from the original format to a unified coordinate format to obtain the defect feature data with the local spatial coordinates.

[0021] This embodiment provides a multimodal data acquisition mechanism; specifically, on the same target unit in the aforementioned coastal wind farm, different detection objects are usually completed by different terminals, and the original output formats are significantly different. If only basic acquisition functions are relied upon without format unification, the subsequent fusion analysis may obtain several isolated defect points, but it is difficult to clarify the order and correspondence of these defects in the same payload chain. Therefore, this embodiment adds a receiving service unit and a format conversion unit in the receiving stage. Specifically, the receiving service unit receives four types of data: blade non-destructive testing data, main shaft non-destructive testing data, flange bolt non-destructive testing data, and tower weld non-destructive testing data; blade data can come from infrared thermography, ultrasonic or visual surface mapping, and is usually recorded as image frames, hot spot areas or spanwise segments; The spindle data can be represented as A-scan, B-scan, or defect echo segments; flange bolt data can include bolt number, circumferential position, preload anomaly, or crack characteristics; tower weld data often includes weld number, layer, circumferential position, and defect echo characteristics; the format conversion unit converts these original formats into a unified coordinate format. The uniformity of format here does not mean forcing all data to be the same image size, but rather unifying it into a structured expression of component identification + local location + defect category + defect degree + acquisition time; for example, the thermal anomaly area in the original thermal image of the blade can be converted to blade No. 2, 1.2 meters from the root, chordally towards the middle, suspected of debonding; The abnormal spindle echo can be converted to the spindle, 45 degrees circumferentially, a certain range in the axial direction, suspected to be a microcrack; the echo point of the tower weld can be converted to the circumferential weld of the first tower section, 210 degrees circumferentially, suspected to be slag inclusion. To facilitate understanding, an example can be given: Suppose that the blade end, the main shaft end, and the weld end upload original record segments A, B, and C from different terminals, respectively. After conversion, they form corresponding structured conversion results, namely unified record R1, unified record R2, and unified record R3. Specifically, unified record R1 includes component = blade and local coordinate = L1, where L1 represents the local position identifier of the blade defect in the blade's own coordinate system. The unified record R2 contains component = principal axis and local coordinate = L2, where L2 represents the local position identifier of the principal axis defect in its own coordinate system; the unified record R3 contains component = tower weld and local coordinate = L3, where L3 represents the local position identifier of the tower weld defect in the coordinate system to which the tower weld belongs; subsequent modules do not care about the original file extension format, but only call the structured fields in the unified records R1, R2, and R3; As a supplementary implementation, if the data uploaded by a terminal lacks local coordinates but has the original measurement trajectory, the format conversion unit can infer the local coordinates based on the detection trajectory and the component reference point; if there are neither coordinates nor traceable trajectory, the data is marked as only for manual review and will not be included in the automatic coupling analysis. If the same part is detected by two different inspections and the defect descriptions are contradictory, the receiving service unit will retain both records and mark the source of collection, and will not directly overwrite them, so as to avoid accidental deletion of risk information. In the aforementioned unit inspection, the drone uploads the thermal image sequence of the blade surface, the portable phased array uploads the scanning result files of the main shaft and bolts, and the wall-climbing robot uploads the weld seam scanning trajectory and defect echo. After format conversion, the system generates unified records for the debonding area at the root of blade No. 2, the 45-degree micro-crack area in the circumferential direction of the main shaft, the abnormal area of ​​bolt No. 12, and the slag inclusion area at 210 degrees of the circumferential weld of the first section of the tower, each with its own local location. The purpose of this step is to eliminate the differences in storage format and coordinate representation of multi-source detection data, thereby achieving the data homogenization input required for subsequent spatial fusion analysis.

[0022] In a preferred embodiment of the present invention, the spatial mapping and fusion module includes: a coordinate transformation unit, used to acquire the local spatial coordinates of the defect feature data, and convert the local spatial coordinates into global coordinates in the global three-dimensional coordinate system based on a coordinate system transformation matrix jointly generated by the static assembly geometry of each preset component and the instantaneous attitude parameters collected by the sensor; and a model generation unit, used to mark the defect feature data with the global coordinates on the preset three-dimensional structural model to generate the digital twin model data.

[0023] This embodiment provides a spatial mapping and fusion mechanism. Specifically, if we only stay at the local coordinate level inside the component, although we can know which blade or weld the defect is located on, we still cannot determine its spatial correspondence in the force transmission direction of the whole machine. Especially under the combined effect of yaw attitude, impeller azimuth angle and tower segment structure, local coordinates are difficult to directly support system-level risk identification. Therefore, this embodiment sets up a coordinate transformation unit and a model generation unit. Specifically, the coordinate transformation unit completes multi-level coordinate transformation based on the preset three-dimensional structural model; the transformation relationship here is not a simple mathematical deformation, but reflects the real assembly relationship: the local coordinates of the blade need to be transformed into the hub coordinates first, then into the nacelle coordinates, and then into the global coordinates of the whole machine in combination with the tower top installation attitude. The underlying implementation of the aforementioned multi-level coordinate transformation relies on a coordinate system transformation matrix generated jointly by the static assembly geometry of each preset component and the instantaneous attitude parameters collected by sensors. Specifically, this matrix is ​​a composite transformation matrix derived by multiplying the static translation and rotation geometry matrices determined by the assembly drawings of each component system with the dynamic homogeneous transformation matrix generated jointly by the instantaneous attitude parameters such as the impeller azimuth angle and nacelle yaw angle collected in real time by sensors. The spatial mapping calculation formula based on the preset three-dimensional structural model is as follows: ,in, Represents the transformed global coordinates. The local spatial coordinates of the defect feature data are represented. Represents the composite transformation matrix; The coordinates of the main shaft need to be converted by combining its axial direction and the nacelle reference position; the coordinates of the tower weld need to be determined based on the tower section number, flange elevation, and circumferential reference direction. To further clarify, for components such as blades and hubs that rotate with the rotor, the system does not permanently fix a defect as a single unchanging spatial point, but expresses it using a combination of reference global coordinates and time-attitude parameters. The reference global coordinates are used to describe the installation position of the defect in the reference attitude or unified zero-position attitude during shutdown maintenance, and the time-attitude parameters include at least the impeller azimuth angle, yaw angle and nacelle attitude at the time of acquisition. When performing spatial alignment or load chain analysis, the reference global coordinates can be restored to instantaneous global coordinates according to the specific analysis time. This avoids mistaking rotating blade defects for fixed and stationary locations on a certain wind side and ensures that defect records collected at different times can be correctly correlated under a unified reference system. After completing the coordinate transformation, the model generation unit overlays the defects onto the preset 3D structural model as a label layer to form digital twin model data. This label layer can be represented by points, lines, surfaces, or voxel blocks: slender cracks can be represented by line segments, sheet-like debonding can be represented by surface patches, and volumetric defects in welds can be represented by local voxel elements. In this way, the subsequent analysis engine not only knows that there are defects, but also knows the spatial orientation of the defects relative to the tower, flange, and spindle. To facilitate the explanation of the processing mechanism of the present invention, the following is a specific example: Suppose that the local point P1 of the blade defect is located near the blade root, the local point P2 of the main shaft defect is located at 45 degrees in the circumferential direction, and the local point P3 of the tower weld defect is located at 210 degrees in the circumferential direction of the first section circumferential weld; After conversion, the local points P1, P2, and P3 correspond to the global points G1, G2, and G3 of the whole machine, respectively. For defects in rotating components such as local point P1, global point G1 can be first recorded as the reference point under the zero position attitude, and the azimuth angle corresponding to the acquisition time is added; when analyzing a certain running time, it is restored to the instantaneous position at that time by the azimuth angle; the model generation unit superimposes global points G1, G2, and G3 into a unified model and records the relative azimuth relationship of the three under the prevailing wind direction. As an alternative implementation, if the equipment has undergone component replacement or reinforcement addition, but the three-dimensional structural model has not been updated, the coordinate transformation unit will first read the most recent equipment configuration file for offset correction; if obvious geometric conflicts still occur after correction, such as defect points being mapped outside the component entity, the record will be returned for manual verification. If a component's attitude sensor malfunctions and the current azimuth angle cannot be determined, the system can first use the static reference attitude under shutdown and maintenance conditions to generate a basic twin map, but the final release of the dynamic coupling results will be suspended. Furthermore, if the moment of blade defect acquisition is missing but the blade number and installation section can be confirmed, the model generation unit will at least retain its baseline global coordinates and will not mistakenly write it as a certain instantaneous wind side position. Instead, the record will be marked as attitude and will be used in static correlation analysis after being supplemented. In the aforementioned target unit, the system maps the debonding area at the root of blade No. 2 as a reference planar anomaly area under zero-position attitude and records the impeller azimuth angle at the time of acquisition; when it is necessary to analyze a certain high wind speed moment, the reference anomaly area is then converted to the corresponding windward spatial position of the tower top at that moment. At the same time, the system maps the microcrack at the 45-degree position of the main shaft as a linear abnormal area near the transmission chain, and maps the abnormality of bolt No. 12 and the slag inclusion in the weld below it as a weak connection area near the tower portal. After the model is generated, the operation and maintenance personnel can see intuitively that these abnormalities are not scattered and isolated, but are approximately aligned along the same off-center load direction under a unified reference attitude, and a consistent instantaneous spatial correspondence can be recovered at the corresponding operating time. The purpose of this step is to transform local defects of multiple components into associative objects in the unified space of the whole machine, thereby realizing defect correspondence and load path identification across components.

[0024] In a preferred embodiment of the present invention, the load coupling analysis module includes: an aerodynamic load simulation unit, used to calculate the asymmetric aerodynamic load based on the blade non-destructive testing data and the real-time operating condition data in the digital twin model data; and a stress field coupling calculation unit, used to simulate the transmission process of the asymmetric aerodynamic load along the dynamic load transmission link based on a preset finite element analysis model, so as to calculate the stress coupling data, wherein the transmission path of the dynamic load transmission link includes the blade, the main shaft, the flange bolts, and the tower weld.

[0025] This embodiment provides a load coupling analysis mechanism; specifically, relying solely on spatial mapping can reveal the correlation of defects in geometric location, but it still cannot answer whether these defects will actually pose a danger under current operation. The technical mechanism is that wind turbines are a typical aerodynamic-mechanical-structural coupled system. The structural stress caused by the initial aerodynamic anomaly of the upstream blades under low wind speed conditions is usually lower than the safety limit, but it may be amplified downstream under gust and yaw mismatch conditions. Therefore, this embodiment further sets up an aerodynamic load simulation unit and a stress field coupling calculation unit. Specifically, the aerodynamic load simulation unit calculates asymmetric aerodynamic loads based on blade defect information and real-time operating condition information in the digital twin model. The asymmetric aerodynamic load refers to the lift, drag, or bending moment generated by wind on each blade no longer maintaining an ideal uniform state; debonding of the blade surface, local cracks, or abnormal stiffness will change the local shape and deformation response, causing the impeller to experience periodic off-center loading during rotation. This off-center load physically manifests as fluctuations in hub input torque, increased combined bending and torsional stress on the main shaft, and exacerbated lateral swaying at the tower top. In the specific calculation process, the aerodynamic load simulation unit first corrects the airfoil aerodynamic coefficient at the corresponding blade section based on the specific type and geometric size of the defects in the blade non-destructive testing feature data, including the changes in lift coefficient and drag coefficient. The corrected aerodynamic coefficients are substituted into the preset blade element momentum theory calculation model or aeroelastic model, and combined with real-time wind speed, wind direction and rotational speed data, the actual aerodynamic forces of the defective blades at each azimuth angle are solved, and compared with the forces of a uniform impeller under normal conditions, and the specific values ​​of the asymmetric aerodynamic loads are extracted. The stress field coupling calculation unit simulates load transfer along the path of blade-main shaft-flange bolt-tower weld based on a preset finite element analysis model; The finite element analysis model is not limited to full-machine fine mesh analysis; a partitioned refinement strategy can also be adopted: the upstream uses aerodynamic load boundaries, and the downstream uses local mesh refinement in the main shaft support, flange connection and weld heat-affected zone. The system focuses on the evolution of stress concentration rather than simply providing the nominal load of a component. In order to realistically reflect the physical effects of the aforementioned minute defects in the preset finite element analysis model, the stress field coupling calculation unit is also equipped with mesh mapping transformation logic, which is used to convert the spatial coordinates of defects, defect categories, and size characteristics in the digital twin model data into reduction coefficients of mechanical properties such as material stiffness and yield strength in the corresponding finite element mesh node regions, or directly apply discontinuous boundary conditions in the corresponding local mesh to simulate the real physical crack and material debonding state. It should be further noted here that, in order to maintain consistency with the object names in the embodiments, all descriptions involving the force transmission in the flange area shall use the flange bolt as the explicit node name in the transmission path; The flange connection area mentioned in the text is a description of the installation or stress environment around the node and does not replace the flange bolts as the object of inspection and calculation. Accordingly, the end of the load chain is uniformly named as the tower weld. If slag inclusions, weld toe geometric transition, or heat-affected zone sensitivity are discussed, they are all regarded as an expansion of the local state of the tower weld, rather than adding a new independent node. Through this unified naming, the load transfer path in the text is always maintained as four key objects: blades, main shaft, flange bolts, and tower weld. To make it easier to understand, an example can be given: Suppose there are four key nodes N1, N2, N3, and N4 in the load path, corresponding to the blade, main shaft, flange bolt, and tower weld, respectively. The aerodynamic load simulation unit first forms a set of off-center load inputs at blade node N1. The stress field coupling calculation unit observes whether this input is transformed into increased main shaft vibration at main shaft node N2, whether it is transformed into increased shear of bolts on one side at flange bolt node N3, and whether it is transformed into increased local peak stress of the weld at tower weld node N4. If the response from node N1 to N4 shows a continuous increasing trend, it is determined that there is a significant coupling effect in the load chain. As a supplementary implementation, if real-time operating data is only partially available, such as wind speed being available but yaw angle being missing, the aerodynamic load simulation unit will use the attitude parameters of the most recently stable operating segment as a substitute, but the analysis results will be marked as approximate results. If the quality of the blade defect data is insufficient and it is impossible to determine the direction of its influence on the aerodynamic shape, the system will not perform fine off-center load simulation, but will degenerate into a joint analysis of conservative uniform load and local defect sensitivity; if the finite element model has abnormal local convergence or unstable boundary conditions, the automatic release of stress peak values ​​will be suspended, and only trend-level warnings will be retained. In the aforementioned target unit, the system judged based on the debonding zone at the root of blade No. 2 and the high wind speed conditions over the past week that the blade would experience stronger local aerodynamic imbalance when rotating to a certain quadrant, causing the main shaft to experience off-center load fluctuations in the corresponding phase. After this fluctuation is transmitted through the connection area where the flange bolts are located, the area near the No. 12 flange bolts is subjected to higher alternating shear, and there is slag inclusion in the tower weld in this location in terms of spatial physical position. Therefore, the local stress peak of the tower weld is significantly higher than that in other locations. The purpose of this step is to link the spatial distribution of static defects with dynamic operating loads, thereby enabling a physical judgment on whether a defect truly poses a danger.

[0026] In a preferred embodiment of the present invention, the stress field coupling calculation unit includes: a frequency comparison subunit, used to extract the natural frequency of the main shaft based on the modal analysis of the finite element analysis model, and to calculate the abnormal vibration frequency of the blade based on the periodic fluctuation of the asymmetric aerodynamic load; if the absolute value of the difference between the abnormal vibration frequency of the blade and the natural frequency of the main shaft is lower than a preset frequency threshold, it is determined that a resonance amplification effect has occurred; and if the absolute value of the difference is not lower than the preset frequency threshold, it is determined that no resonance amplification effect has occurred. The stress superposition calculation subunit is used to calculate the amplified shear force when the asymmetric aerodynamic load is transmitted to the flange bolts, and to calculate the local stress concentration peak as the stress coupling data in combination with the non-destructive testing data of the tower weld. In addition, if the resonance amplification effect is determined not to occur, the basic shear force when the asymmetric aerodynamic load is transmitted to the flange bolt is calculated, and the local stress concentration peak value is calculated as the stress coupling data in combination with the non-destructive testing data of the tower weld.

[0027] This embodiment provides a resonance identification and stress superposition mechanism; specifically, in the previous embodiment, even if the load transfer process has been established, if the difference between ordinary off-center load and near-resonance off-center load is not further distinguished, the destructiveness of some tiny defects with sizes below the preset detection threshold may still be underestimated. The main shaft in the wind turbine drive train has inherent vibration characteristics. When the periodic excitation frequency caused by blade abnormalities approaches this natural frequency, an amplified response will occur. At this time, the fatigue cumulative damage rate of the downstream flange and weld will be in a nonlinear accelerated accumulation state. Therefore, this embodiment further adds a frequency comparison sub-unit and a stress superposition calculation sub-unit. Specifically, the frequency comparison sub-unit first extracts the natural frequency range of the main shaft through modal analysis, and then extracts the abnormal vibration frequency of the blade based on the periodic changes of the asymmetric aerodynamic load. The frequency proximity is not simply a random coincidence of data, but rather indicates a dynamic state in which the external excitation rhythm and the inherent modal frequency of the structure tend to be consistent, thereby significantly increasing the amplitude of the structural response. When the difference between the two falls within a preset frequency threshold, the system determines that there is a resonance amplification effect; when the difference exceeds the threshold, it is determined that there is no obvious resonance. The stress superposition calculation subunit handles the two cases separately. If resonance is identified, the actual shear input at the flange bolts will no longer use the base value, but will instead use the amplified shear response, and the local stress concentration peak value will be calculated in combination with the existing defect morphology of the tower weld, the geometric discontinuity of the weld toe, and the sensitivity of the heat-affected zone; if resonance is identified, the stress concentration peak value will still be calculated according to the base shear transfer; this can avoid treating all off-center loads as the most dangerous case and also avoid ignoring the real amplified working conditions. To facilitate the explanation of the processing mechanism of this invention, the following is illustrated with specific examples: Let the natural frequency obtained from the spindle modal analysis be F1, and the dominant excitation frequency caused by the blade anomaly be F2; if the excitation frequency F2 is close to the natural frequency F1, then the amplification path is entered, and the subsequent bolt force record switches from the basic shear response B1, which represents the transmission to the flange bolt before resonance amplification, to the amplified shear response B2, which represents the transmission to the flange bolt after resonance amplification; if the two are not close, then the processing follows the basic path, and the bolt force remains at the basic shear response B1; Regardless of whether it is the basic shear response B1 or the amplified shear response B2, the system combines it with the tower weld defect state record W1 that participates in the stress concentration calculation, and outputs the local stress concentration peak value S1 calculated by combining record W1 under the basic path, or the local stress concentration peak value S2 calculated by combining record W1 under the amplified path. As a supplementary implementation, if the main shaft modal information is insufficient, for example, if effective main shaft state modeling has not been completed recently, the system can retrieve the calibration modal range of the same model unit as a priori reference, but this will reduce the certainty level of resonance judgment; if the abnormal vibration frequency of the blade is too unstable due to excessive transient gust disturbance, the system will prioritize the comparison of the dominant frequency that recurs in multiple operating cycles to avoid false alarms triggered by instantaneous spikes; if the frequency difference is near the threshold boundary, the system can mark the result as a suspected amplification area and suggest adding on-site retesting or short-term online vibration monitoring. In the aforementioned unit, the debonding of blade No. 2 caused a stable abnormal excitation rhythm in the impeller rotation, while the microcracks in the main shaft slightly reduced its local stiffness and caused a shift in its inherent vibration characteristics. After system comparison, it was found that the abnormal excitation was close to the vibration range of the main shaft. Therefore, it was processed according to the amplification path and the shear force borne by bolt No. 12 under specific wind conditions was calculated to be significantly higher than the normal value. Combined with the slag inclusion in the weld below the bolt and the poor geometric transition of the weld toe, a higher local stress concentration peak value was finally obtained. The purpose of this step is to identify structural response amplification caused by similar frequencies, thereby enabling a more precise distinction of high-risk coupling states.

[0028] In a preferred embodiment of the present invention, the risk assessment and prediction module includes: a life prediction unit, used to take the stress coupling data as input to a preset life prediction model, so as to determine the remaining service life of the whole machine based on the output of the preset life prediction model; and a strategy generation unit, used to classify the structural failure risk level according to the remaining service life of the whole machine, and generate the corresponding comprehensive maintenance strategy.

[0029] This embodiment provides a risk assessment and prediction mechanism. Specifically, if the analysis process stops at the peak stress concentration, although it can output the specific location of the high-risk stress concentration area, it still cannot directly support the wind farm operation and maintenance scheduling. In actual operation and maintenance, what matters more is the remaining safe operating time, the priority of the shutdown window, and whether immediate shutdown is required. Therefore, this embodiment further sets up a life prediction unit and a strategy generation unit to convert the stress coupling results into the remaining lifespan and maintenance action suggestions. Specifically, the life prediction unit inputs stress coupling data into a preset life prediction model; this model comprehensively considers local stress amplitude, alternation frequency, defect type, material fatigue sensitivity, operating history, and environmental corrosion background. The preset life prediction model is specifically based on Minor's linear fatigue cumulative damage theory and SN curves; its internal operating logic is as follows: The stress amplitude limit is corrected for the SN curve of the basic component based on the material fatigue sensitivity, operating history and environmental corrosion background. The corresponding limit cycle number is extracted according to the local stress concentration peak in the stress coupling data. The fatigue damage accumulation rate under the current operating condition is calculated by combining the load alternation frequency. Finally, the remaining service life of the whole machine is deduced based on the set damage tolerance threshold of the whole machine. Its core is not to abstractly output a score, but to estimate how long the whole machine can maintain safe operation under the current combination of defects and the current operating trend. The technical mechanism of using the remaining service life of the whole machine as the evaluation index is that the failure of a critical connection of the wind turbine can lead to the shutdown of the whole machine or even a structural accident. The weakest link downstream will in turn limit the usable life of the whole machine. The strategy generation unit classifies the structural failure risk level based on the remaining service life and generates corresponding comprehensive maintenance strategies. Comprehensive maintenance is not limited to replacing the worst-performing component, but may also include power-limited operation, adjusting the maintenance sequence, preparing hoisting windows in advance, or arranging upstream component repairs. This can avoid only maintaining a single high-altitude weld and ignoring more suitable solutions to reduce the impact of the load source. For ease of understanding, a simplified explanation can be provided: Assume that the life prediction unit outputs three status values, representing the need for handling in the short term, the handling within the planning period, and the continuation of routine tracking; the strategy generation unit outputs three types of plans based on these values: emergency shutdown maintenance, near-term overhaul maintenance, and routine re-inspection maintenance. As a supplementary implementation, if some key factors are missing in the life prediction input, such as the lack of long-term operating history, the system will prioritize using the current coupling stress and the empirical boundary of similar units to generate conservative life results; if the differences in the remaining life judgments of different model branches are too large, the system will output a life range instead of a single point value, and suggest supplementing online monitoring and re-evaluation; if the current unit is in an extreme weather warning period, the strategy generation unit can add temporary risk tightening rules on the basis of the existing life conclusions. In the aforementioned target unit, based on the stress coupling state formed by blade debonding, main shaft microcracks, flange bolt abnormalities, and slag inclusions in tower welds, the system predicts that if the unit maintains its current high wind speed operation mode, the remaining safe operating time will be lower than the preset lower limit of the safe operating cycle. Therefore, the strategy generation unit does not simply suggest observation, but combines the unit's operation control strategy to provide a comprehensive solution: first limit power operation, and then address the critical defects during the nearest shutdown window. The purpose of this step is to transform the complex coupled stress results into actionable operation and maintenance decision information, thereby enabling the risk analysis results to directly support on-site maintenance.

[0030] In a preferred embodiment of the present invention, the strategy generation unit includes: an intervention calculation subunit, used to select the target component with the largest stress concentration peak on the dynamic load transfer link as the first component when the structural failure risk level is higher than a preset risk threshold, and to simulate and eliminate the defect feature data corresponding to the first component in the digital twin model data; The life reassessment subunit is used to determine the second component located downstream of the first component in the dynamic load transfer link, and calculate the extended fatigue life of the second component after eliminating the defect feature data corresponding to the first component; the scheme output subunit is used to output the operation of eliminating the defect feature data corresponding to the first component as the comprehensive maintenance strategy; and, if the structural failure risk level is not higher than the preset risk threshold, the above-mentioned operation of simulating the elimination of the defect feature data is not performed, and the comprehensive maintenance strategy of routine inspection is directly output.

[0031] This embodiment provides a cross-component intervention calculation mechanism; specifically, it only gives the risk level based on the remaining service life. Although it can indicate when danger occurs, there is still a practical problem in high-risk conditions: determining the priority maintenance location; if it is determined only based on the defect size or maintenance convenience, maintenance resources are often preferentially allocated to shallow defect locations where the appearance features are easily identifiable, while ignoring the source that actually causes the load amplification. Therefore, this embodiment introduces an intervention calculation subunit, a lifetime re-estimation subunit, and a scheme output subunit. Specifically, when the risk level is higher than a preset threshold, the intervention calculation subunit selects the target component with the largest stress concentration peak along the dynamic load transfer link as the first component and simulates the elimination of its defect characteristics in the digital twin model. The so-called simulation elimination refers to temporarily replacing the defect in the model with the repaired state. For example, restoring the debonding area of ​​the blade to a complete aerodynamic shape, restoring the abnormal bolt to normal pre-tightening, and restoring the weld defect to a state without significant discontinuities. Here, it is further specified that the determination of the first component follows the main rule in the embodiment, that is, sorting the components according to the peak value of stress concentration on the current dynamic load transfer link and selecting the target component with the largest peak value. Based on this, if there is a downstream structure in the load transfer direction of the first component that can be independently assessed for life changes, the life recalculation subunit directly uses the downstream structure as the second component to calculate the extended fatigue life; in other words, whether there is a downstream second component that belongs to the life recalculation stage does not change the premise that the first component is determined according to the principle of maximum peak value. The fatigue life recalculation sub-unit recalculates the fatigue life changes of the downstream second component based on the repaired model. The technical mechanism behind this is that some upstream defects, although not at the final failure location, are the root cause of off-center loading and amplification. Once the upstream problems are eliminated, the fatigue life of the downstream weld or connection may recover significantly. The output subunit of the scheme will output the repair action when the fatigue life extension rate is greater than the preset expected threshold as a comprehensive maintenance strategy. The output content can include not only the repair object, but also the expected effect after the repair, such as the degree of improvement of the fatigue life of downstream components, the time node of the re-evaluation that can be postponed, and whether it can replace the direct emergency repair of some high-risk parts. For example: Suppose that there are components C1, C2 and C3 in the load chain. The system first determines the first component based on the peak stress concentration. If the peak value of component C2 is the largest, then component C2 is determined as the first component, and component C3, which is downstream of it, is determined as the second component. The system simulates the elimination of defects in component C2 and calculates the extension of the life of component C3. Based on this, it decides whether to output a maintenance strategy that prioritizes the treatment of component C2. If the first component with the largest peak value is located at the end of the link, such as the tower weld seam being a high-risk point at the end, the system will still retain the first component and its simulation elimination results as the basis for the main strategy. At the same time, the life re-estimation object will be switched to the life improvement amount before and after the end component is repaired, or the load reduction effect of the upstream component after defect elimination on the high-risk point at the end will be shown as supplementary comparison information. In this way, the implementation logic always remains the same: first, lock the first component with the largest peak value, and then determine the second component or the equivalent lifetime re-evaluation object based on the link location. As a supplementary implementation, if the stress concentration peak values ​​of multiple components are similar, the intervention calculation subunit can be jointly ranked according to repair feasibility, downtime arrangement constraints, and the extent of improvement in downstream lifespan; if the improvement in downstream lifespan is not significant after simulation elimination, it indicates that the component is not the dominant source, and the system continues to try the next candidate component. If the peak value is highest at the end of the link, the system will still retain it as the first component for display and output. When cross-component effect comparison is required, the load reduction calculation results of the upstream candidate components will be added to avoid the high-risk results at the end being ignored. If all single component treatments cannot significantly improve the lifespan, a joint intervention strategy will be output, such as simultaneously repairing blade defects and replacing specific bolts. If the risk has exceeded the safe operating limit, the system will prioritize outputting shutdown protection suggestions without waiting for complex scheme comparisons. In the aforementioned target unit, the system identified that the bolt connection area of ​​flange No. 12 showed the largest stress concentration peak in the current dynamic load transmission link. Therefore, the abnormal part of the flange bolt was taken as the first component, and the tower weld and heat-affected zone below it were taken as the second component. The system first simulates the elimination of the abnormality of flange bolt No. 12 and obtains the first set of extended fatigue life results of the downstream tower weld. Then, combined with the feasibility of maintenance, it outputs the basic strategy of prioritizing the handling of the abnormality of flange bolt No. 12 and reviewing the condition of the tower weld below it. Furthermore, in the evaluation of the extended scheme, the system can also add a comparison of the overall load reduction effect after the debonding repair of the root of the upstream No. 2 blade. If the additional action brings more significant life improvement to the flange bolts and the downstream tower weld, it will be presented as a joint or subsequent optimization suggestion, but it does not replace the main strategy output obtained based on the first component. The purpose of this step is to find maintenance entry points that meet preset evaluation conditions through source intervention calculations, thereby enabling cross-component maintenance decisions.

[0032] In a preferred embodiment of the present invention, the strategy generation unit includes: a risk classification unit, configured to determine the structural failure risk level as a high-risk level when the remaining service life of the whole machine is lower than a first service life threshold; and to determine the structural failure risk level as a medium-risk level when the remaining service life of the whole machine is not lower than the first service life threshold and is lower than a second service life threshold. If the remaining service life of the entire machine is not less than the second service life threshold, the structural failure risk level is determined to be a low risk level; wherein, the first service life threshold is less than the second service life threshold.

[0033] This embodiment provides a risk classification mechanism. Specifically, the life prediction result alone is not enough for the system to standardize the assessment of the status of different units, because different units, different maintenance teams, and different spare parts plans require standardized classification criteria. If the original life value is read directly each time, it is easy to cause inconsistencies in the handling standards. Therefore, this embodiment adds a risk classification unit to classify the remaining service life into intervals. Specifically, the risk classification unit sets a first lifespan threshold and a second lifespan threshold, with the first lifespan threshold being lower than the second lifespan threshold; its physical and operational significance lies in distinguishing equipment from three levels: those requiring emergency intervention, those requiring planned intervention, and those requiring sustainable tracking. When the remaining service life is lower than the first service life threshold, it indicates that the weak part of the structure will reach an unacceptable risk zone within the first preset time period during subsequent operation. At this time, it is classified as a high-risk level, which usually corresponds to rapid shutdown assessment, operation restriction or emergency maintenance. When the remaining service life is between two thresholds, it means that the equipment still has a certain safety margin, but it is no longer suitable for long-term operation without intervention. At this time, it is classified as a medium-risk level and is appropriate to be included in the recent maintenance plan. When the remaining service life is not less than the second service life threshold, it indicates that the current defect coupling has not yet formed an imminent risk and can be classified as a low-risk level, and routine inspections should be maintained. The threshold here can be set according to the aircraft type, wind field environment, and maintenance organization capabilities, and is not limited to a fixed absolute value; for highly corrosive marine environments, the threshold can be appropriately tightened, while for inland low-turbulence environments, a more conventional range can be adopted; As a supplementary implementation method, if the life prediction results are presented as an interval rather than a single point, the risk classification unit can be conservatively graded according to the principle of prioritizing the lower limit of the interval. If the unit is on the eve of seasonal extreme weather, even if the lifespan result is slightly higher than the first lifespan threshold, the risk level can be temporarily raised by one level according to the wind farm strategy. If the same unit is evaluated across multiple levels in a series of consecutive assessments, the system can trigger manual review to avoid unnecessary maintenance disruptions caused by occasional data fluctuations. In the aforementioned unit, the system comprehensively considered the recent high wind speed season, the existing defect coupling status, and the life prediction results, and determined that the remaining safe operating time of the unit was close to the emergency response boundary, and therefore it was classified as a high-risk level. Although the adjacent unit also has minor abnormalities in the tower welds, there are no obvious signs of blade overload or main shaft resonance. The remaining lifespan is between two thresholds, so it is classified as a medium-risk unit and will be handled in the next planned maintenance cycle. The purpose of this step is to transform continuous life results into standardized risk levels, thereby enabling unified prioritization and processing of maintenance tasks.

[0034] In a preferred embodiment of the present invention, the system further includes: a feedback correction module, configured to receive the actual execution results of the comprehensive maintenance strategy and calculate the prediction deviation based on the actual execution results; and a parameter update module, configured to update the weight parameters of the preset lifetime prediction model based on the prediction deviation.

[0035] This embodiment provides a feedback correction mechanism; specifically, even though the aforementioned life prediction and strategy generation can output relatively reasonable maintenance suggestions, the operating environment of wind turbines is complex, and the life evolution of the same model under different wind fields, different corrosion levels and different maintenance quality is not completely consistent. If the model does not receive feedback from the field for a long time, the prediction results will gradually deviate from the actual state; therefore, this embodiment further sets up a feedback correction module and a parameter update module. Specifically, the feedback correction module receives the actual execution results of the comprehensive maintenance strategy; the execution results may include: whether the blade defects were repaired as recommended, whether the abnormal bolts were replaced, whether the vibration decreased after repair, whether the weld re-inspection was stable, and whether new alarms or shutdown events occurred in the subsequent period of time. This module compares these results with the original predictions to identify prediction biases. For example, the system predicted that prioritizing the repair of blade defects could significantly alleviate the risk of downstream welds. If the unit's vibration response does indeed decrease after on-site repair and the weld re-inspection shows stability, it indicates that the model has accurately identified the load source. If downstream risks still appear rapidly after repair, it indicates that the model may have underestimated the dominant role of defects in the bolts or welds themselves. The parameter update module updates the weight parameters of the life prediction model based on the prediction deviation. Specifically, the parameter update module uses the backpropagation algorithm or the proportional-integral-derivative adjustment mechanism, takes the minimization of the prediction deviation value as the objective function, calculates the gradient direction or compensation step size of each weight parameter with respect to the deviation, and performs quantitative iteration and numerical optimization on the original blade influence weight, bolt influence weight and weld influence weight, until the life prediction deviation in the recent running samples converges to the preset error allowable range. The specific mechanism of the update is to gradually adjust the weight coefficients of different characteristic variables in the model. For example, increase the weight of the impact of marine corrosion background on weld life, or increase the weight of the impact of abnormal blade stiffness on the main shaft eccentric load. The updated model can provide more realistic evaluation results on similar units in the future. For example: Suppose that the model originally assigned corresponding blade influence weights W1, bolt influence weights W2, and weld influence weights W3 to the influence of blade defects, bolt defects, and weld defects, respectively. If multiple field results indicate that the downstream life improvement after blade repair is more significant than predicted, the blade influence weight W1 can be appropriately increased in subsequent updates; if the actual improvement is weak, the blade influence weight W1 should be reduced accordingly and the bolt influence weight W2 and weld influence weight W3 should be rebalanced. As a supplementary implementation method, if the on-site execution results are incomplete, for example, only whether maintenance was performed is recorded but no post-maintenance operation feedback is provided, the feedback correction module will only perform low-intensity updates and will not make significant parameter adjustments. If there is significant construction deviation or secondary damage during a maintenance process, the result can be marked as an anomalous sample and not directly used for model learning. If the environmental differences between different wind farms are too large, the parameter update module can adopt a local update strategy based on different wind farms and different turbine models to avoid introducing inapplicable prior data across wind farms, which could lead to prediction bias. In the aforementioned target unit, the system recommended prioritizing the repair of the debonding at the root of blade No. 2 and rechecking bolt No. 12; after actual implementation, the unit was reconnected to the grid and the online vibration trend decreased, and the welds in the corresponding positions of the tower did not show further deterioration in subsequent re-inspections; Based on this, the feedback correction module determined that the previous judgment that the blade source defect caused downstream stress amplification was consistent with the actual situation, and the parameter update module increased the weight of this type of coupling mode in the same model. The purpose of this step is to allow the model to continuously absorb the results of on-site maintenance, thereby achieving gradual correction and adaptation of its life prediction capabilities.

[0036] In a preferred embodiment of the present invention, the receiving service unit includes: a standardized interface subunit, used to configure a preset communication protocol bus to receive the non-destructive testing feature data from different testing terminals through the communication protocol bus, wherein the testing terminals include UAV testing equipment, portable phased array equipment and magnetic adsorption wall-climbing robot equipment.

[0037] This embodiment provides a standardized access mechanism; specifically, in the aforementioned multimodal acquisition process, if different detection terminals each use isolated data export methods, common problems include disordered upload timing, inconsistent field naming, inconsistent units, and even different ways of writing the same component number in different devices; this will directly undermine the reliability of subsequent format conversion and spatial mapping. Therefore, this embodiment further includes a standardized interface subunit in the receiving service unit; specifically, the standardized interface subunit is configured with a preset communication protocol bus for uniformly receiving data from different detection terminals; the terminals include at least a UAV detection device, a portable phased array device, and a magnetic adsorption wall-climbing robot device; Unmanned aerial vehicle (UAV) inspection equipment is typically used for rapid inspection of anomalies on the outer surface and local internal parts of blades. Its output often includes flight path information, image frame number, and blade number. Portable phased array (PADA) equipment is often used for the inspection of spindles and flange bolts. Its output is more focused on scanning channels, echo characteristics, and inspection location markings. The magnetic adsorption wall-climbing robot is mainly used for tower weld inspection. Its output includes the robot's trajectory, weld segment number, and echo timing. The standardized interface subunit integrates the outputs of the three types of terminals into the same access bus through unified field conventions, unified component numbering rules, unified timestamp format, and unified coordinate reference. To make it easier to understand, an example can be given: the UAV uploads data packet D1 containing the blade number, waypoint number, and image frame; the phased array uploads data packet D2 containing the component number, scan segment, and echo peak. The wall-climbing robot uploads data packet D3 containing tower section number, weld number, and trajectory point; after the interface sub-unit specification, data packets D1, D2, and D3 are transformed into unified data record U1, unified data record U2, and unified data record U3, respectively, and these unified data records all have basic fields such as equipment source, component number, acquisition time, location identifier, and original data index. As a supplementary implementation, if a terminal does not yet support the bus protocol, it can be converted and connected through an adapter program; if the terminals upload data in batches after offline operation, the interface subunit will rearrange the data entry order according to the collection time to prevent later data from overwriting earlier data; if the data field uploaded by a terminal is missing the component number, the interface subunit can make preliminary completion based on the task order and work order, but will simultaneously mark it as pending verification. If communication is interrupted or data packets are corrupted, the system retains the received portion and sends a retransmission request to the corresponding terminal. During the annual inspection of the aforementioned coastal wind farm, drones completed blade inspections during the day, portable phased array radars completed main shaft and bolt inspections inside the nacelle, and wall-climbing robots completed weld inspections on the outer wall of the tower. The three types of equipment are connected to the operation and maintenance center through a unified protocol bus. The system aggregates them according to the same unit number and the same maintenance task number, so that subsequent processes can directly link the four types of records: defect of blade No. 2, micro-crack at 45 degrees of main shaft, abnormality of bolt No. 12, and slag inclusion in the circumferential weld of section 1. The purpose of this step is to provide a consistent and traceable data access foundation for heterogeneous testing terminals, thereby achieving stable aggregation and unified processing of multi-source nondestructive testing data.

[0038] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A non-destructive testing and analysis system for wind turbine towers, characterized in that, The system includes: A multimodal data acquisition module is used to receive non-destructive testing feature data of preset components constituting the load transmission link of a wind turbine, and to preprocess the non-destructive testing feature data to extract defect feature data with local spatial coordinates. The preset components include at least blades, main shafts and towers. The spatial mapping and fusion module is used to map the defect feature data to a global three-dimensional coordinate system based on a preset three-dimensional structural model, thereby generating digital twin model data containing global defect distribution. The structural parameters of the preset three-dimensional structural model include the static assembly geometric parameters and node coordinates of each preset component, and the mapping calculation formula is as follows: ,in, This represents the global coordinates in the global three-dimensional coordinate system. Represents the local spatial coordinates, Represents the coordinate system transformation matrix; The load coupling analysis module is used to acquire real-time operating condition data of the wind turbine through the configured external device communication interface, calculate the dynamic load transmission link in combination with the digital twin model data, and calculate the stress coupling data of the defect feature data under dynamic load based on the dynamic load transmission link. The risk assessment and prediction module is used to determine the structural failure risk level based on the stress coupling data, and output a comprehensive maintenance strategy according to the structural failure risk level.

2. The non-destructive testing and analysis system for wind turbine towers according to claim 1, characterized in that, The multimodal data acquisition module includes: The receiving service unit is used to receive non-destructive testing data of blades, non-destructive testing data of main shaft, non-destructive testing data of flange bolts, and non-destructive testing data of tower welds as the non-destructive testing feature data. The format conversion unit is used to convert the nondestructive testing feature data into a unified coordinate format to obtain the defect feature data with the local spatial coordinates.

3. The non-destructive testing and analysis system for wind turbine towers according to claim 1, characterized in that, The spatial mapping and fusion module includes: The coordinate transformation unit is used to obtain the local spatial coordinates of the defect feature data, and convert the local spatial coordinates into global coordinates in the global three-dimensional coordinate system based on the coordinate system transformation matrix jointly generated by the static assembly geometry of each preset component and the instantaneous attitude parameters collected by the sensor. The model generation unit is used to mark the defect feature data with the global coordinates on the preset three-dimensional structural model to generate the digital twin model data.

4. The non-destructive testing and analysis system for wind turbine towers according to claim 2, characterized in that, The load coupling analysis module includes: The aerodynamic load simulation unit is used to calculate asymmetric aerodynamic loads based on the blade non-destructive testing data and the real-time operating condition data in the digital twin model data. The stress field coupling calculation unit is used to simulate the transmission process of the asymmetric aerodynamic load along the dynamic load transmission link based on a preset finite element analysis model, so as to calculate the stress coupling data. The transmission path of the dynamic load transmission link includes blades, main shaft, flange bolts and tower welds.

5. The non-destructive testing and analysis system for wind turbine towers according to claim 4, characterized in that, The stress field coupling calculation unit includes: The frequency comparison sub-unit is used to extract the natural frequency of the main shaft based on the modal analysis of the finite element analysis model, and to calculate the abnormal vibration frequency of the blade based on the periodic fluctuation of the asymmetric aerodynamic load. If the absolute value of the difference between the abnormal vibration frequency of the blade and the natural frequency of the main shaft is lower than the preset frequency threshold, it is determined that a resonance amplification effect has occurred. If the absolute value of the difference is not lower than the preset frequency threshold, it is determined that no resonance amplification effect has occurred. The stress superposition calculation subunit is used to calculate the amplified shear force when the asymmetric aerodynamic load is transmitted to the flange bolts when the resonance amplification effect is determined to occur, and to calculate the local stress concentration peak value as the stress coupling data in combination with the non-destructive testing data of the tower weld; and to calculate the basic shear force when the asymmetric aerodynamic load is transmitted to the flange bolts when the resonance amplification effect is determined not to occur, and to calculate the local stress concentration peak value as the stress coupling data in combination with the non-destructive testing data of the tower weld.

6. The non-destructive testing and analysis system for wind turbine towers according to claim 1, characterized in that, The risk assessment and prediction module includes: The life prediction unit is used to take the stress coupling data as input to a preset life prediction model, and to determine the remaining service life of the whole machine based on the output of the preset life prediction model. The strategy generation unit is used to classify the structural failure risk level according to the remaining service life of the whole machine, and generate the corresponding comprehensive maintenance strategy.

7. The non-destructive testing and analysis system for wind turbine towers according to claim 6, characterized in that, The strategy generation unit includes: An intervention calculation subunit is used to select the target component with the largest stress concentration peak as the first component on the dynamic load transfer link when the structural failure risk level is higher than a preset risk threshold, and to simulate and eliminate the defect feature data corresponding to the first component in the digital twin model data. The life re-estimation subunit is used to determine the second component located downstream of the first component in the dynamic load transfer link, and to calculate the extended fatigue life of the second component after eliminating the defect feature data corresponding to the first component. The solution output subunit is used to output the operation of eliminating the defect feature data corresponding to the first component as the comprehensive maintenance strategy; and, if the structural failure risk level is not higher than the preset risk threshold, the above-mentioned operation of simulating the elimination of the defect feature data is not performed, and the comprehensive maintenance strategy of routine inspection is directly output.

8. The non-destructive testing and analysis system for wind turbine towers according to claim 6, characterized in that, The strategy generation unit includes: The risk classification unit is used to determine the structural failure risk level as high risk level when the remaining service life of the whole machine is lower than the first service life threshold. If the remaining service life of the entire machine is not less than the first service life threshold and is less than the second service life threshold, the structural failure risk level is determined to be a medium risk level. If the remaining service life of the entire machine is not less than the second service life threshold, the structural failure risk level is determined to be a low risk level; wherein, the first service life threshold is less than the second service life threshold.

9. The non-destructive testing and analysis system for wind turbine towers according to claim 6, characterized in that, The system also includes: The feedback correction module is used to receive the actual execution results of the comprehensive maintenance strategy and calculate the prediction deviation based on the actual execution results; The parameter update module is used to update the weight parameters of the preset life prediction model based on the prediction deviation.

10. The non-destructive testing and analysis system for wind turbine towers according to claim 2, characterized in that, The receiving service unit includes: A standardized interface subunit is used to configure a preset communication protocol bus to receive non-destructive testing feature data from different testing terminals via the communication protocol bus. The testing terminals include UAV testing equipment, portable phased array equipment, and magnetic adsorption wall-climbing robot equipment.

Images