Wind turbine blade fatigue damage early warning method and system based on modal energy deviation

Abstract

The application provides a wind power blade fatigue damage early warning method and system based on modal energy offset, relates to the technical field of wind driven generators, and establishes a multi-dimensional health baseline library under the health state of a wind power blade, synchronously collects current total strain signals and temperature distribution in the operation process of the wind driven generator, and decouples the total strain by using the health baseline library to accurately extract real mechanical strain signals caused by mechanical load; on this basis, a working condition modal identification method is used to robustly obtain current natural frequency and strain modal vibration mode from the real mechanical strain; then, the current modal energy proportion is calculated and compared with the health baseline to generate an offset index reflecting the localization degree of modal energy; finally, the amplitude, duration and dynamic evolution trend of the offset index are combined, and the fatigue cumulative damage evaluation result is fused to perform graded early warning judgment, so that early perception, accurate positioning and risk grading output of blade fatigue damage are realized.

Description

Technical Field

[0001] This application relates to the field of wind turbine technology, and in particular to a method and system for early warning of fatigue damage in wind turbine blades based on modal energy shift. Background Technology

[0002] With the increasing size of wind turbines, blades endure complex loads under prolonged conditions, leading to accelerated fatigue damage accumulation. This can easily cause cracks, stiffness degradation, and even fracture, threatening operational safety and increasing maintenance costs. Current mainstream monitoring methods include acoustic emission, video recognition, and point sensors, but all have significant shortcomings: acoustic emission is insensitive to early microcracks and is easily affected by noise; image methods can only identify surface cracks and cannot detect internal damage; traditional point sensors have limited spatial coverage and cannot capture subtle modal responses caused by local stiffness changes. These methods largely rely on post-event identification or macroscopic judgment, lacking early detection, precise location, and evolution tracking capabilities for damage initiation, and thus cannot support high-precision online early warning. Overall, existing technologies largely remain at the level of post-event identification or macroscopic response discrimination, lacking early warning perception and dynamic evolution tracking capabilities for fatigue damage, making it difficult to meet the engineering requirements of high reliability, high precision, and all-time-domain online early warning for modern large wind turbine blades. Summary of the Invention

[0003] This application provides a method and system for early warning of fatigue damage in wind turbine blades based on modal energy shift, in order to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions to achieve high-sensitivity perception, high-precision positioning and dynamic evolution assessment of early fatigue damage inside wind turbine blades, thereby supporting intelligent operation and maintenance and safety early warning throughout the entire life cycle.

[0004] On the one hand, this application provides a method for early warning of fatigue damage in wind turbine blades based on modal energy migration, including the following steps:

[0005] When wind turbine blades are in a healthy and undamaged state, establish a multi-dimensional health baseline database;

[0006] The health baseline library includes initial natural frequencies, initial strain mode shapes, initial modal energy ratios, initial temperature distributions, temperature-strain sensitivity coefficients, and zero-drift terms.

[0007] During the operation of the wind turbine, the current total strain signal and the current temperature distribution are collected synchronously, and the temperature-strain decoupling of the current total strain signal is performed based on the health baseline library to obtain the real mechanical strain signal;

[0008] Based on the real mechanical strain signal, the current natural frequency and the current strain mode shape are extracted using the working condition modal identification method.

[0009] Based on the current strain mode shape, the current mode energy ratio is calculated, and combined with the initial mode energy ratio, a shift index characterizing the degree of modal energy localization is generated;

[0010] Based on the offset index and its dynamic trend, combined with the fatigue cumulative damage assessment results, a graded early warning judgment is performed, and a corresponding fatigue damage early warning signal is output.

[0011] Furthermore, distributed sensing optical fibers are continuously laid along the optical fiber path in the main beam, web, leading edge, trailing edge, and blade root transition area of ​​the wind turbine blade. The distributed sensing optical fibers include strain sensing optical fibers and temperature sensing optical fibers to collect the current total strain signal and current temperature distribution of the wind turbine blade. The current total strain signal includes mechanical strain signal and thermal strain signal.

[0012] The distributed sensing optical fiber is attached to the inner wall surface of the wind turbine blade in an S-shaped path, and at least one optical fiber loop is added in the middle region of the blade to enhance the monitoring sensitivity of this high-stress area.

[0013] Furthermore, before performing temperature-strain decoupling, the current total strain signal is subjected to noise reduction preprocessing. The noise reduction preprocessing adopts a combination of wavelet threshold denoising and sliding median filtering. The wavelet basis function and the number of decomposition layers are preset according to the frequency range of the dominant mode of the blade, so as to retain the key spatial distribution characteristics of the strain mode shape while suppressing high-frequency noise and spike interference.

[0014] Furthermore, the specific method of temperature-strain decoupling is as follows:

[0015] The thermally induced strain component is obtained by multiplying the difference between the current temperature distribution and the initial temperature distribution by the location-dependent temperature-strain sensitivity coefficient.

[0016] Subtract the thermal strain component from the current total strain signal, and then subtract the position-related zero drift term to obtain the true mechanical strain signal caused by the mechanical load.

[0017] Furthermore, the step of extracting the current natural frequency and current strain mode shape based on the actual mechanical strain signal using a working condition modal identification method includes the following steps:

[0018] Frequency domain decomposition analysis was performed on the real mechanical strain signal to preliminarily identify the frequency peaks of the dominant modes and the corresponding coarse strain mode shapes as the first analysis results.

[0019] The real mechanical strain signal is subjected to time-domain random subspace identification analysis to obtain the system state space model, and the frequency, damping ratio and strain mode shape of the stable mode are extracted as the second analysis result.

[0020] The first and second analysis results are cross-validated and fused, and modes with consistent frequencies are retained to extract the current natural frequency and the current strain mode shape.

[0021] Further, the step of calculating the current modal energy ratio based on the current strain mode shape and generating a shift index characterizing the degree of modal energy localization, combined with the initial modal energy ratio, includes the following steps:

[0022] The wind turbine blade is divided into multiple discrete segments along its length.

[0023] Based on the current strain mode shape, the local modal energy of each mode in each discrete segment is calculated and normalized to obtain the normalized current modal energy ratio.

[0024] The current modal energy ratio is compared with the initial modal energy ratio, and the relative energy offset of each discrete segment is calculated.

[0025] The relative energy offsets under different operating conditions or different modes are weighted and fused to generate a comprehensive offset index, which is used to characterize the degree of spatial localization of modal energy.

[0026] Furthermore, the weighting coefficients of the relative energy offset under different operating conditions or different modes are determined based on the energy proportion, mass distribution, or short-term stability of the corresponding mode under healthy conditions.

[0027] Furthermore, the step of performing a graded early warning judgment based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, and outputting a corresponding fatigue damage early warning signal includes the following steps:

[0028] When the maximum value of the offset index exceeds the first threshold, it is determined to be a first-level warning state, and a first-level warning signal is output.

[0029] When the duration of the first-level warning state exceeds a preset duration threshold, it is determined to be a second-level warning state, and a second-level warning signal is output.

[0030] In the first or second warning state, the real mechanical strain time history signal is extracted at the segment position with the largest offset index, the strain cycle spectrum is obtained by rain flow counting method, and the fatigue accumulation and its growth rate are calculated by combining the material SN curve and Miner linear cumulative damage theory.

[0031] When the maximum value of the offset index exceeds the second threshold and the growth rate of the accumulated fatigue exceeds the third threshold, it is determined to be a third-level warning state, and a third-level warning signal is output.

[0032] On the other hand, this application provides a wind turbine blade fatigue damage early warning system based on modal energy shift, including a sensing layer, a data acquisition layer, an edge computing layer and a cloud processing layer;

[0033] The sensing layer consists of distributed sensing optical fibers embedded in the main beam, web, leading edge, trailing edge and blade root transition area of ​​the wind turbine blade, including strain sensing optical fibers and temperature sensing optical fibers. It is attached to the inner wall surface of the wind turbine blade in an S-shaped path, and at least one optical fiber ring is added in the high stress area of ​​the blade.

[0034] The acquisition layer is configured to receive the optical signal output by the sensing layer, and use an optical fiber demodulator to restore the optical signal to the current total strain signal and the current temperature distribution, thereby generating a structural response data packet.

[0035] The edge computing layer is deployed inside the wind turbine nacelle or tower and connected to the acquisition layer. It is configured to perform data synchronization, noise reduction preprocessing, temperature-strain decoupling and operating condition mode identification on the structural response data packet, and extract the current natural frequency and current strain mode shape.

[0036] The cloud processing layer is configured to: store a pre-built multidimensional health baseline library; calculate the current modal energy ratio of each discrete segment based on the current strain mode shape; generate an offset index characterizing the degree of modal energy localization by combining the initial modal energy ratio; and further perform graded early warning judgment based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, and generate and issue corresponding fatigue damage early warning signals.

[0037] Furthermore, the system also includes an operation and maintenance terminal; the operation and maintenance terminal is configured to receive fatigue damage early warning signals sent by the cloud processing layer, and realize the visualization of the wind turbine blade structural status, the presentation of early warning information, and the automatic generation of operation and maintenance reports.

[0038] The beneficial effects of this application are as follows: This application provides a method for early warning of fatigue damage in wind turbine blades based on modal energy migration. This method establishes a multi-dimensional health baseline library under the healthy state of the wind turbine blade, including initial natural frequency, initial strain mode shape, initial modal energy ratio, initial temperature distribution, temperature-strain sensitivity coefficient, and zero drift term, to provide accurate reference for subsequent damage identification. During wind turbine operation, the current total strain signal and temperature distribution are collected synchronously, and the temperature-strain decoupling of the total strain is performed using the health baseline library to accurately extract the real mechanical strain signal caused by mechanical load. On this basis, the operating condition modal identification method is used to robustly obtain the current natural frequency and strain mode shape from the real mechanical strain. Then, the current modal energy ratio is calculated and compared with the health baseline to generate a migration index reflecting the degree of localization of modal energy. Finally, the amplitude, duration, and dynamic evolution trend of the migration index are combined with the fatigue cumulative damage assessment results to perform graded early warning judgment, realizing early perception, accurate location, and risk classification output of blade fatigue damage, significantly improving the reliability, sensitivity, and engineering practicality of wind turbine blade health monitoring. This application also provides a corresponding system, the beneficial effects of which are similar to those of the method, and will not be described in detail here.

[0039] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0040] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0041] Figure 1 This is a flowchart of the wind turbine blade fatigue damage early warning method based on modal energy migration provided in this application;

[0042] Figure 2 This is a schematic diagram of the distributed optical fiber arrangement on the blades provided in this application;

[0043] Figure 3 This is a schematic diagram of the local modal strain energy distribution of the wind turbine blade provided in this application;

[0044] Figure 4 This is a structural diagram of the wind turbine blade fatigue damage early warning system based on modal energy migration provided in this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0046] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0047] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0049] As the global energy structure accelerates its transformation towards cleaner and lower-carbon energy, wind power, as one of the most mature and widely used forms of renewable energy, is rapidly developing towards larger scale, deep-sea applications, and high grid connection rates. Under this trend, the length of wind turbine blades continues to increase, and structural flexibility is significantly improved. Especially in wind farms with complex terrain such as mountains and coastal areas, blades are exposed to the coupled effects of multiple complex operating conditions, including non-uniform wind speed profiles, strong wind shear, high turbulence intensity, and yaw errors, bearing the combined loads of high-frequency small-amplitude vibrations and low-frequency large-amplitude oscillations. This complex dynamic loading environment greatly accelerates the fatigue damage accumulation process of blade materials, leading to frequent failure events such as crack initiation, stiffness degradation, and even sudden fracture. This seriously threatens the safe operation of the unit, reduces power generation efficiency, and significantly increases operation and maintenance costs and downtime risks.

[0050] To address these challenges, various blade structural health monitoring methods have been widely adopted in engineering practice. Currently, the traditional solutions commonly used by owners mainly fall into two categories: one is blade root video monitoring combined with point vibration sensors, and the other is acoustic emission monitoring systems. The former can only observe visible damage on the surface of the blade root and cannot detect early microcracks in critical load-bearing components such as the main beam and web. Furthermore, point vibration sensors have limited spatial coverage, typically only acquiring local single-point responses or overall low-order modal information, making it difficult to capture subtle modal changes caused by local stiffness degradation. While the latter can respond to transient elastic wave signals released during crack propagation, it is insensitive to the micro-damage stage before stable propagation and is highly susceptible to wind turbine operating noise, electromagnetic interference, and environmental vibrations, resulting in high false alarm and false negative rates. Therefore, current mainstream monitoring methods largely remain at the level of data collection and recording, making it difficult to achieve true early warning.

[0051] In recent years, some patents have attempted to introduce new technologies such as fiber Bragg gratings, aeroelastic simulation, and response surface modeling to improve monitoring capabilities. For example, patent CN104005917A uses Bayesian inference to predict wind turbine status and constructs a probabilistic model using historical operating data. However, this method is purely data-driven, lacks connection with the physical mechanisms of structural damage, cannot identify early fatigue characteristics such as microcracks or local stiffness degradation inside the blades, and cannot provide damage location information, making it difficult to support accurate operation and maintenance decisions. Patent CN108038320A proposes to calculate the vibration reliability of detuned bladed disks based on the multiple response surface method, simulating structural detuning by changing the elastic modulus and conducting modal and transient response analysis. However, this method is mainly used for reliability assessment in the design phase, involves a large amount of computation, and cannot achieve online continuous monitoring and real-time early warning.

[0052] Furthermore, patent CN113202701A discloses a fiber optic monitoring system for wind power plants, which uses distributed optical fibers to collect signals such as temperature, vibration, and stress and transmit them to a monitoring platform. However, its function is limited to the collection and transmission of raw signals and does not involve intelligent analysis or damage identification; essentially, it remains a data acquisition system rather than an early warning method. Other research focuses on the aeroelastic analysis of wind turbine blades. Patent CN119514260A calculates the response by coupling aerodynamic models with structural parameters to evaluate the aeroelastic stability of flexible blades. However, this also falls under the scope of design verification and lacks online monitoring and damage evolution tracking capabilities.

[0053] In summary, existing technologies generally suffer from the following shortcomings: First, most methods can only identify blade cracks when they have developed to the surface or entered a stable propagation stage, failing to detect early fatigue damage. Second, spatial resolution and coverage are severely insufficient; traditional point sensors or local imaging methods cannot continuously cover the entire blade length, resulting in numerous monitoring blind spots. Third, there is a lack of damage localization capabilities based on structural dynamics mechanisms. Existing solutions mostly rely on statistical anomaly detection or image feature discrimination, without incorporating physical mechanisms such as modal energy distribution, leading to weak generalization ability and poor early identification capability. Finally, anti-interference capabilities are weak, and data reliability is low. For example, acoustic emission technology is susceptible to operational noise interference, and most fiber optic or strain measurement schemes do not effectively decouple temperature and mechanical strain, resulting in strain data distortion and affecting the accuracy of subsequent analysis. These limitations make it difficult for current technologies to meet the urgent engineering requirements of modern large wind turbine blades for high reliability, high precision, and full-time online early warning.

[0054] To address the aforementioned problems, this application proposes a fatigue damage early warning method and system based on modal energy migration. The core of this application's technical solution lies in integrating distributed fiber optic sensing with modal energy localization theory. By continuously laying sensing fibers in the blade's main spar, web, leading edge, trailing edge, and blade root transition zone, it achieves synchronous strain and temperature sensing with full-scale, millimeter-level spatial resolution. Under healthy conditions, a multi-dimensional baseline library is established, including natural frequencies, strain mode shapes, modal energy ratios, temperature distribution, and temperature-strain coupling parameters. During operation, the total strain signal undergoes noise reduction preprocessing and temperature-strain decoupling to accurately extract the true mechanical strain. Combining frequency domain decomposition and random subspace identification algorithms, it achieves... The system robustly acquires local modal parameters without external excitation. Based on the difference between the current and initial modal energy ratios, an energy offset index is constructed to quantitatively characterize the modal energy localization phenomenon caused by stiffness degradation, achieving high-precision damage location. Furthermore, strain time histories are extracted in the located area, and the degree of fatigue accumulation and its growth trend are quantified using the rainflow counting method and Miner's linear cumulative damage theory. Finally, based on the offset index amplitude, duration, and fatigue growth rate, a three-level graded early warning logic is executed, and multi-dimensional information including damage location, dominant mode, fatigue state, and original data fragments is output. This enables mechanism-driven, highly sensitive, and highly reliable online early warning and intelligent operation and maintenance support for early fatigue damage inside wind turbine blades.

[0055] Modal energy localization theory posits that when a structure suffers damage, such as cracks or stiffness degradation, its vibration characteristics change, causing the energy of certain vibration modes to no longer be uniformly distributed throughout the structure, but rather to concentrate significantly near the damaged area. This anomalous spatial concentration of energy has a clear physical correlation with the location of the damage, thus serving as a key basis for identifying and locating structural damage. In this patent, this theory forms the core physical foundation for achieving early warning of fatigue damage in wind turbine blades. By monitoring changes in modal energy distribution, it can keenly detect signs of local stiffness degradation before macroscopic failure occurs, thereby achieving high-precision damage location and condition assessment.

[0056] First, the fatigue damage early warning method based on modal energy shift provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0057] Reference Figure 1 The implementation process of the fatigue damage early warning method based on modal energy migration provided in this application includes, but is not limited to, the following steps.

[0058] Step S110: When the wind turbine blades are in a healthy and undamaged state, establish a multidimensional health baseline database.

[0059] The healthy baseline library includes initial natural frequencies, initial strain mode shapes, initial modal energy ratios, initial temperature distribution, temperature-strain sensitivity coefficients, and zero-drift terms.

[0060] In step S110, a comprehensive and reliable reference baseline is established for the entire damage early warning system. When the wind turbine blade is in a healthy and undamaged state, a series of key physical quantities are systematically collected and stored through specialized calibration and excitation tests, including initial natural frequencies, initial strain mode shapes, initial modal energy ratios, initial temperature distribution, temperature-strain sensitivity coefficients, and zero-drift terms. These parameters together constitute a multi-dimensional health baseline library, reflecting the structural dynamic characteristics and thermo-mechanical coupling response features of the blade in a damage-free state. This baseline library not only provides a quantitative comparison basis for anomaly detection under subsequent operating conditions but also serves as a prerequisite for achieving temperature compensation, modal identification, and energy shift calculation, ensuring that the entire early warning method has a clear physical reference and interpretability.

[0061] Step S120: During the operation of the fan, the current total strain signal and the current temperature distribution are collected synchronously, and the temperature-strain decoupling of the current total strain signal is performed based on the health baseline library to obtain the real mechanical strain signal.

[0062] It should be noted that temperature-strain decoupling refers to the process of separating the "true" mechanical strain caused by mechanical loads from the total strain signal measured by distributed optical fibers. Since optical fibers are simultaneously affected by structural stress and ambient temperature changes during actual operation, the measured total strain is a superposition of mechanical strain and thermally induced strain. To accurately reflect the true stress state of the blade, it is necessary to use synchronously acquired temperature distribution data and the position-dependent temperature-strain sensitivity coefficient calibrated under healthy conditions to compensate for the total strain, subtracting thermal expansion effects and system zero drift, thereby extracting the pure mechanical strain. This decoupling result is an indispensable prerequisite for subsequent operating condition modal identification, modal energy calculation, and damage early warning analysis, ensuring the accuracy and reliability of dynamic feature extraction.

[0063] In step S120, strain information caused by mechanical load is separated from the original sensing data. During the actual operation of the wind turbine, the distributed optical fiber synchronously outputs the current total strain signal and the current temperature distribution. The total strain signal simultaneously includes mechanical strain and thermal strain caused by temperature changes. If not distinguished, the thermal effect will seriously interfere with the judgment of the actual stress state of the structure. Therefore, this step uses the temperature-strain sensitivity coefficient and zero-drift term in the healthy baseline library established in step S110, combined with the difference between the current and initial temperature distributions, to perform temperature-strain decoupling processing on the total strain signal, thereby accurately extracting the true mechanical strain signal. This step effectively eliminates spurious strain caused by environmental temperature fluctuations, providing high-fidelity input data for subsequent modal analysis.

[0064] Step S130: Based on the real mechanical strain signal, the current natural frequency and the current strain mode shape are extracted using the working condition modal identification method.

[0065] It should be noted that operating condition modal identification is a technique that identifies the natural frequencies, damping ratios, and mode shapes of a structure under actual operating conditions, relying solely on environmental excitations such as wind and turbulence, without the need for artificial excitation. This method is suitable for large engineering structures such as wind turbines that are difficult to actively excite, and can accurately reflect their dynamic characteristics in the service environment.

[0066] In step S130, the dynamic characteristic parameters of the blade under actual operating conditions are identified from the real mechanical strain signal. Since wind turbine blades are only subjected to environmental excitations such as wind and turbulence during service and artificial excitation cannot be applied, a condition-based modal identification method is required. This step, based on the decoupled mechanical strain signal, uses algorithms such as frequency domain decomposition and random subspace identification to extract the current natural frequency and current strain mode shape under conditions without external excitation. These modal parameters directly reflect the current structural stiffness and mass distribution of the blade and are key indicators for determining whether there is local stiffness degradation or damage. This step realizes the transformation from a continuous strain field to structural dynamic characteristics, laying the foundation for subsequent energy calculations.

[0067] Step S140: Based on the current strain mode shape, calculate the current mode energy ratio, and combine it with the initial mode energy ratio to generate a offset index characterizing the degree of mode energy localization.

[0068] In step S140, the modal shape information is transformed into a quantifiable and comparable damage-sensitive index. This step first calculates the local modal energy of each discrete segment of the blade under each mode based on the current strain modal shape, and obtains the current modal energy ratio through normalization. Subsequently, this ratio is compared with the initial modal energy ratio established in step S110 to generate a shift index characterizing the degree of modal energy localization. This index can sensitively capture the abnormal spatial concentration of modal energy caused by local stiffness degradation; its magnitude and location distribution directly correspond to the presence and location of damage. This step transforms abstract modal changes into a damage indicator with clear physical meaning and spatial orientation.

[0069] Step S150: Based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, perform graded early warning judgment and output the corresponding fatigue damage early warning signal.

[0070] In step S150, the damage identification results are transformed into early warning decisions with engineering guidance value. This step comprehensively considers the amplitude, duration, and dynamic trend of the offset index over time, and combines the fatigue cumulative damage results assessed by the mechanical strain time history extracted in the high offset region to execute a set of graded early warning judgments. When the offset index exceeds a preset threshold and meets specific time or growth conditions, the system sequentially triggers early warning signals at the levels of suspicious, attention, or alarm. This step not only determines whether there is a damage risk, but also assesses its development trend and severity, and finally outputs structured early warning information, providing maintenance personnel with clear, graded, and actionable risk alerts, thereby achieving closed-loop management from state awareness to decision support.

[0071] In some embodiments of this application, distributed sensing fibers are continuously laid along the optical fiber path in the main beam, web, leading edge, trailing edge, and blade root transition zone of the wind turbine blade. These distributed sensing fibers include strain sensing fibers and temperature sensing fibers to collect the current total strain signal and current temperature distribution of the wind turbine blade. The current total strain signal includes both mechanical strain and thermal strain signals. By constructing a high-density sensing network covering key structural areas of the wind turbine blade, continuous and synchronous monitoring of structural strain and temperature status is achieved.

[0072] Specifically, by continuously laying distributed sensing fibers along the fiber optic path in areas bearing major loads or prone to damage, such as the main beam, web, leading edge, trailing edge, and blade root transition zone, the sensing system can comprehensively capture the mechanical response of the blade at each key location during operation. The distributed sensing fibers used include two types: strain sensing fibers and temperature sensing fibers, used to simultaneously acquire the current total strain signal and the current temperature distribution, respectively. The current total strain signal is essentially a superposition of the mechanical strain signal caused by the mechanical load and the thermal strain signal caused by changes in ambient temperature. This deployment method not only achieves full-scale, high spatial resolution data acquisition but also provides necessary temperature reference information for subsequently separating the true mechanical response from the mixed signal, serving as a fundamental prerequisite for the high-precision damage detection of the entire early warning method.

[0073] In some embodiments of this application, distributed sensing optical fibers are bonded to the inner surface of the wind turbine blade in an S-shaped path, and at least one optical fiber loop is added in the blade mid-area to enhance the monitoring sensitivity of this high-stress area. By optimizing the layout of the distributed sensing optical fibers inside the wind turbine blade, the monitoring capability of key high-stress areas is improved.

[0074] By attaching distributed sensing fibers to the inner surface of the blade in an S-shaped path, not only can the long structural path be effectively covered, but the density of fibers in specific areas can also be increased within a limited space, thereby improving the spatial resolution and signal stability of local strain sensing. Simultaneously, adding at least one fiber loop in the blade's mid-section is a strengthened monitoring measure taken to address the characteristics of this area, which typically bears the maximum bending moment and fatigue load during actual operation and is a high-stress concentration zone. The fiber loop, made of wound optical fibers, forms a locally closed-loop structure, significantly enhancing the sensitivity to minute strain changes in this area and helping to capture the initial signals of fatigue damage earlier and more accurately. This deployment strategy balances overall coverage with targeted reinforcement, providing a physical basis for achieving high-sensitivity and high-reliability damage early warning.

[0075] In some embodiments of this application, reference is made to Figure 2 , Figure 2This diagram illustrates one arrangement of distributed sensing optical fibers on a wind turbine blade. The fibers are attached to the inner surface of the blade in an S-shape to monitor strain signals and their distribution in real time. The optical fibers themselves are very lightweight and do not affect the load distribution on the blade. Furthermore, as a non-metallic material, they avoid issues such as attracting lightning strikes at high altitudes. Additionally, more sensitive fiber loops, made of wound optical fibers, can be placed in the vulnerable central areas for focused monitoring.

[0076] In some embodiments of this application, distributed sensing fibers are continuously laid along the fiber optic path in the main beam, web, leading edge, trailing edge, and blade root transition zone of the wind turbine blade, achieving high-density, full-scale sensing of key structural areas of the blade. In subsequent measurement processes, using... Represents the arc length coordinates of the optical fiber, that is, the physical position of the optical fiber along the surface of the blade (unit: meters). The time is expressed in seconds. The total strain signal output by the distributed fiber optic demodulation system is denoted as... The signal comprises two parts: mechanical strain caused by mechanical load and thermal strain caused by temperature change. Simultaneously, the system synchronously acquires the current temperature distribution. To achieve temperature compensation, it is necessary to record the initial temperature distribution of the blades when they are in a healthy and undamaged state. This serves as a reference for subsequent temperature difference calculations. During the healthy period, the first [condition / condition] is identified through stimulation tests. initial natural frequency of the order (Unit: Hz), and record the number of Hz. Initial strain mode shape of the first order ,in , The modal order used. Indicates the first First mode at position The normalized initial amplitude response at the point. For subsequent modal energy comparison analysis, the blade structure is discretized along its length into [a specific region / section]. Each segment (which can be considered a finite element element or a measurement point segment) is numbered as follows: And calculate the first during the healthy period The first mode in the 1st order Normalized initial modal energy ratios on each segment This refers to the proportion of the energy of that mode in that segment. Additionally, the temperature-strain sensitivity coefficient is calibrated. This represents the change in apparent strain caused by a unit increase in temperature; it also calibrates the zero drift term. , representing the initial offset of the system under no-load and no-temperature-change conditions.

[0077] Through the above process, a multidimensional reference baseline library of healthy, non-destructive leaves can be constructed. These correspond to the initial natural frequency, initial strain mode shape, initial modal energy ratio, initial temperature distribution, temperature-strain sensitivity coefficient, and zero drift term under healthy conditions, respectively. This baseline library provides a physical reference for damage identification under subsequent operating conditions.

[0078] In some embodiments of this application, before performing temperature-strain decoupling, the current total strain signal is subjected to noise reduction preprocessing. The noise reduction preprocessing adopts a combination of wavelet threshold denoising and sliding median filtering. The wavelet basis function and the number of decomposition layers are preset according to the frequency range of the dominant mode of the blade, so as to retain the key spatial distribution characteristics of the strain mode shape while suppressing high-frequency noise and spike interference, thereby improving the quality of the current total strain signal and providing high-fidelity input data for subsequent temperature-strain decoupling and mode identification.

[0079] In real-world operating environments, the total strain signal acquired by distributed optical fibers inevitably contains non-structural response components such as high-frequency electronic noise, environmental vibration interference, and occasional spike pulses. Without proper processing, these components will severely impact the accuracy of subsequent analyses. Therefore, before temperature-strain decoupling, the original strain signal undergoes noise reduction preprocessing, specifically employing a combination of wavelet threshold denoising and sliding median filtering. Wavelet threshold denoising effectively filters out high-frequency noise unrelated to structural modes, while sliding median filtering excels at suppressing sudden spike interference. Crucially, the selection of wavelet basis functions and the setting of the decomposition level are not arbitrary but pre-determined based on the frequency range of the dominant modes of the wind turbine blade. This ensures that while removing noise, the strain mode shapes and their key spatial distribution characteristics, reflecting the true structural dynamics, are not destroyed or smoothed out. This preprocessing strategy achieves a balance between noise suppression and feature preservation, a necessary step in ensuring the reliability of the entire damage warning process.

[0080] In some embodiments of this application, during wind turbine operation, the fiber optic demodulator continuously outputs the total strain signal along the fiber optic path. and the current temperature distribution collected simultaneously. To achieve multi-source data fusion, and On the same timeline The operating parameters, such as power and rotor speed, are aligned with those provided by the SCADA system to ensure time synchronization. To suppress the impact of random noise on subsequent analysis, the total strain signal is... Perform robust noise reduction preprocessing: in length of Within the sliding time window, smoothing is performed by combining wavelet thresholding (for suppressing high-frequency noise) and sliding median filtering (for suppressing spike interference). Other noise reduction algorithms can also be selected according to actual needs. Time window length The selection of the sampling rate needs to comprehensively consider the sampling rate and the bandwidth of the main modal frequencies of the structure to ensure that noise is effectively reduced without destroying the modal peak values ​​and their phase characteristics. The total strain signal after this preprocessing... As a key input for subsequent temperature compensation and operating condition mode identification, it provides a high-quality data foundation for accurately extracting real mechanical strain and structural dynamic characteristics.

[0081] In some embodiments of this application, step S120 involves performing a temperature-strain solution on the current total strain signal based on a healthy baseline library, including the following steps.

[0082] Step S210: Calculate the thermally induced strain component obtained by multiplying the difference between the current temperature distribution and the initial temperature distribution by the position-related temperature-strain sensitivity coefficient.

[0083] In step S210, the spurious strain component caused by temperature changes is precisely quantified. During wind turbine operation, fluctuations in ambient temperature cause thermal expansion and contraction of the optical fiber, introducing non-mechanical thermal strain into the total strain signal. To eliminate this interference, this step utilizes the location-dependent temperature-strain sensitivity coefficients calibrated in the healthy baseline library, combined with the difference between the current temperature distribution and the initial temperature distribution, to calculate the corresponding thermal strain component at each spatial location. This calculation fully considers the non-uniformity of thermal response in different regions of the blade due to differences in materials, structure, or installation conditions, giving the thermal strain estimation spatial resolution and laying the foundation for accurate separation of the true mechanical strain in subsequent steps.

[0084] Step S220: Subtract the thermal strain component from the current total strain signal, and then subtract the position-related zero drift term to obtain the true mechanical strain signal caused by the mechanical load.

[0085] In step S220, the influence of non-mechanical factors is completely removed from the original measurement signal, extracting the true mechanical strain signal that purely reflects the stress state of the structure. After obtaining the thermally induced strain component, this step subtracts the thermally induced strain component and the pre-calibrated position-related zero-drift term from the healthy baseline library from the current total strain signal. The zero-drift term represents the output offset of the fiber optic system itself under conditions without any load or temperature changes, and is usually caused by factors such as installation residual stress, fiber aging, or demodulator drift. By simultaneously subtracting the thermally induced strain and the zero-drift term, the final signal is caused by external mechanical loads, completely eliminating the interference of temperature effects and system deviations. This decoupling result is the core input upon which subsequent operating condition modal identification and damage assessment rely, and its accuracy directly determines the reliability and sensitivity of the entire early warning method.

[0086] In some embodiments of this application, in order to obtain the mechanical strain under actual stress, it is necessary to measure the total strain of the distributed optical fiber. Thermal-mechanical decoupling is performed. This process is achieved by subtracting thermally induced strain and system zero-drift terms caused by temperature changes, resulting in the true mechanical strain signal after decoupling. The expression is:

[0087] ;

[0088] in, This represents the actual mechanical strain signal caused by the mechanical load. The total strain signal output by the demodulator includes both mechanical and thermal strain. Current temperature distribution Relative to the initial temperature distribution The temperature rise reflects the local temperature change of the structure. The location-dependent temperature-strain sensitivity coefficient represents the amount of apparent strain change caused by a unit temperature rise. Its value is calibrated during the healthy period and can be slowly updated over time to adapt to environmental or aging effects. The zero-drift term represents the initial offset of the system under no-load and no-temperature-change conditions, and is also calibrated during the health period. This model assumes that the temperature-strain relationship is approximately linear within a short operating window, and that the coupling between the optical fiber and the substrate material remains stable.

[0089] In some embodiments of this application, during the long-term operation of the wind turbine, the recursive least squares method can be used based on the reference values ​​under the lossless state. and Slow online updates are performed to correct drift caused by seasonal temperature variations or fiber optic aging. The resulting signal is the true mechanical strain signal. It is a key input signal for subsequent working condition modal identification and fatigue cumulative damage assessment, ensuring the accuracy and reliability of dynamic analysis.

[0090] In some embodiments of this application, step S130 involves extracting the current natural frequency and current strain mode shape based on the actual mechanical strain signal using a working condition mode identification method, including the following steps.

[0091] Step S310: Perform frequency domain decomposition analysis on the real mechanical strain signal to initially identify the frequency peak of the dominant mode and the corresponding coarse strain mode shape as the first analysis result.

[0092] It should be noted that Frequency Domain Decomposition (FDD) is a modal identification method based on output response. Its core idea is to construct a power spectral density matrix from vibration signals at multiple measurement points, perform singular value decomposition at each frequency, identify the natural frequencies of the structure using the peak positions of the principal singular values, and simultaneously use the corresponding principal singular vectors as estimates of the mode shapes at those frequencies. This method requires no input excitation information, relying only on response data under environmental excitation. It is computationally simple, highly visual, and particularly suitable for practical operating scenarios such as wind turbine blades subjected only to random excitations such as wind and turbulence. In this application, FDD is used to quickly and initially extract the frequencies and coarse strain mode shapes of the dominant modes of the blade, providing an initial reference for subsequent detailed analysis.

[0093] In step S310, the main vibration characteristics of the structure in the frequency domain are rapidly extracted from the real mechanical strain signals. By constructing a power spectral density matrix from the real mechanical strain signals acquired synchronously at multiple points and performing FDD analysis, the peak positions of the singular values ​​at each frequency point can be identified. These peaks correspond to the dominant modal frequencies of the structure. Simultaneously, the principal singular vectors corresponding to the peaks provide the relative amplitude and phase relationships of each measurement point at that frequency, thus forming a rough strain mode shape. This step efficiently acquires preliminary modal information in a non-parametric manner, providing an initial reference and candidate mode set for subsequent more refined time-domain analysis.

[0094] Step S320: Perform time-domain random subspace identification analysis on the real mechanical strain signal to obtain the system state space model, and extract the frequency, damping ratio and strain mode shape of the stable mode as the second analysis result.

[0095] It should be noted that Stochastic Subspace Identification (SSI) is a system identification technique based on a state-space model. It organizes the multi-channel output response data of a structure under environmental excitation into a Hankel matrix, estimates the system's state sequence using subspace projection methods, and then constructs the system matrix from which modal frequencies, damping ratios, and mode shapes are extracted. SSI exhibits good noise robustness and the ability to resolve dense modes, making it particularly suitable for handling large flexible structures with low signal-to-noise ratios and dense modes.

[0096] In step S320, a time-domain system identification method is used to deeply explore the structural dynamic characteristics contained in the real mechanical strain signal. This step uses the real mechanical strain signal as the system output, employs a random subspace identification method to construct a state-space model, estimates the system matrix and observation matrix through a data-driven approach, and then extracts the modal frequencies and damping ratios from the eigenvalues ​​of the system matrix. A more robust and accurate strain mode shape is reconstructed from the column space of the observation matrix. Because this method is based on time-domain response and has good noise robustness, it is particularly suitable for actual operating conditions where the system is only excited by the environment and the signal-to-noise ratio is low, effectively identifying physically meaningful stable modes.

[0097] Step S330: Cross-validate and fuse the first analysis results and the second analysis results, retain the modes with consistent frequencies, and extract the current natural frequency and the current strain mode shape.

[0098] Step S330 aims to improve the reliability and accuracy of modal parameter identification, avoiding spurious modes or omissions of true modes that may be introduced by a single algorithm. This step cross-validates the first analysis result obtained from frequency domain decomposition with the second analysis result obtained from time domain random subspace identification, focusing on comparing the overlap of the two in modal frequencies. Only when a mode is stably identified and has consistent frequencies in both independent methods is it determined to be a real structural mode. Through this fusion strategy, the advantages of frequency domain methods in terms of peak sensitivity are utilized, while the rigor of time domain methods in judging system stability is also taken advantage of. Finally, a set of high-confidence current natural frequencies and current strain mode shapes are output, providing a solid foundation for subsequent modal energy calculations.

[0099] In some embodiments of this application, real mechanical strain signals are used. Discretize along the fiber optic path There are discrete measurement points, denoted as... ,in , For the total number of segments, each Corresponding to the The goal is to identify the local modal characteristics at each discrete point under operating conditions subject only to random environmental excitations (such as wind load and turbulence), including the location of the first discrete point. First natural frequency and the First strain mode vibration .

[0100] First, a preliminary identification is performed using frequency domain decomposition (FDD): a multi-channel power spectral density matrix is ​​constructed. The matrix is The cross-spectral matrix, whose elements represent the frequency differences between each measurement point. The coherent response at the location; for Singular value decomposition is performed at each frequency point, and the peak position of the principal singular value corresponds to the first singular value of the structure. First natural frequency The corresponding principal singular vectors at each discrete point The components on the order constitute the preliminary estimate of that mode. .

[0101] Subsequently, random subspace identification (SSI) was employed in the time domain to further improve identification accuracy: using mechanical strain sequences Composition of output vector A state-space model is constructed using a system identification algorithm to identify the system matrix. With observation matrix ;Depend on The eigenvalues ​​can be used to extract more robust modal frequencies. ,Depend on The column space can obtain more accurate mode shapes. .

[0102] Will , , and Cross-validation and fusion are performed, modes with consistent frequencies are retained, and the current natural frequencies and current strain mode shapes are extracted. The final output is a set of local modal features, denoted as... These parameters reflect the dynamic behavior of the blade under the current operating conditions and are the basis for subsequent calculations of modal energy ratio and generation offset index.

[0103] In some embodiments of this application, step S140 involves calculating the current modal energy ratio based on the current strain mode shape, and combining it with the initial modal energy ratio to generate a shift index characterizing the degree of modal energy localization, including the following steps.

[0104] Step S410: Divide the wind turbine blade into multiple discrete segments along its length.

[0105] In step S410, a spatially discretized computational framework is established for the localized analysis of modal energy. By dividing the wind turbine blade along its length into multiple discrete segments, the originally continuous structure is transformed into several analysis units with clear boundaries and physical meaning. This division not only facilitates the independent calculation and comparison of energy in each region but also provides a basic mesh for achieving high spatial resolution damage localization. Each segment corresponds to a key structural part of the blade, such as the main beam section, the web connection area, or the tip region, thereby ensuring that the energy analysis can reflect the details of local stiffness changes.

[0106] Step S420: Based on the current strain mode shape, calculate the local modal energy of each mode in each discrete segment and normalize it to obtain the normalized current modal energy ratio.

[0107] In step S420, the energy contribution of each segment under different modes in the current state is quantified to eliminate the influence of differences in dimensions and total amount. Based on the current strain mode shape and combined with the equivalent stiffness information of each segment, the local modal energy of each mode in each discrete segment is calculated. This energy is then normalized by dividing the energy of each segment by the total energy of that mode on the entire blade, thus obtaining the normalized current modal energy ratio. This ratio reflects the relative contribution of each spatial region to the specific modal energy under the current operating state and is a direct basis for measuring whether local energy concentration has occurred.

[0108] Step S430: Compare the current modal energy ratio with the initial modal energy ratio, and calculate the relative energy offset of each discrete segment.

[0109] In step S430, by comparing with the healthy state, the abnormal energy distribution caused by the local stiffness degradation of the structure is revealed. This step compares the current modal energy ratio obtained in step S420 with the initial modal energy ratio established in step S110 segment by segment, calculates the relative difference between the two, and forms the relative energy offset of each discrete segment. This offset directly characterizes the increase or decrease of the energy ratio of a certain segment relative to the healthy state under the same modal excitation; when a local stiffness decreases, modal energy tends to concentrate in that region, resulting in a significantly positive offset, thus providing a sensitive and interpretable indicator for damage identification.

[0110] Step S440: Weighted fusion of relative energy offsets under different operating conditions or different modes is performed to generate a comprehensive offset index, which is used to characterize the degree of spatial localization of modal energy.

[0111] In step S440, a robust and comprehensive damage indicator is formed by integrating offset information from multiple modes or operating conditions. Since a single mode may be limited by specific operating conditions or have identification uncertainties, this step assigns corresponding weights to the relative energy offsets of different orders of modes or under different operating conditions and performs weighted fusion to generate a unified comprehensive offset index. This index comprehensively reflects the overall localization trend of modal energy in space, enhancing the sensitivity to real damage while suppressing spurious fluctuations caused by noise or random factors, ultimately providing a reliable and stable basis for subsequent graded early warning.

[0112] In some embodiments of this application, in step S440, the weighting coefficient of the relative energy offset under different operating conditions or different modes is determined based on the energy proportion, mass distribution or short-term stability of the corresponding mode in a healthy state.

[0113] By assigning reasonable weights to the relative energy offsets under different modes or operating conditions, the physical meaning and reliability of the comprehensive offset index are improved. The weighting coefficients are not arbitrarily set, but rather quantitatively allocated based on the inherent characteristics of each mode in a healthy state. Specifically, if a certain mode accounts for a high proportion of energy in a healthy state, it indicates a significant contribution to the overall dynamic behavior of the structure, and should be given a higher weight accordingly. If the mass distribution corresponding to a certain mode is more concentrated in the critical load-bearing area, its energy changes are more sensitive to damage, and its weight should also be appropriately increased. Furthermore, if a mode exhibits good identification stability within a short time window, i.e., it is less affected by noise or operating condition fluctuations, its offset is more reliable and can also be assigned a larger weight. Through this weighting strategy based on the physical characteristics of the healthy baseline, the fused comprehensive offset index can more accurately and robustly reflect the modal energy localization phenomenon caused by actual damage, avoiding interference introduced by low-contribution or unstable modes, thereby enhancing the accuracy and robustness of damage identification.

[0114] In some embodiments of this application, the wind turbine blade is divided along its length into... Each segment is a discrete piece, and the equivalent stiffness of each segment is denoted as . This reflects the local stiffness characteristics of the structure. Based on the first [unclear] identified in the preceding steps... First strain mode vibration As input, define the first The section in The piecewise submodal energy under the first mode is: ,in, This represents the contribution of the modal energy to that segment. To eliminate the influence of differences in total energy between different modes, for each mode... After normalization, the energy proportions in each segment are obtained, satisfying the following formula:

[0115] ;

[0116] in, Indicates the first The first modal energy is at the first The proportion of segments reflects the spatial distribution characteristics of this mode. The current operating state... Ratio of initial modal energy to healthy baseline state By comparison, a local energy shift index is defined. It satisfies the following formula:

[0117] ;

[0118] in, For a very small stability constant (e.g.) This is used to prevent numerical divergence caused by an excessively small denominator, ensuring computational stability. Considering the complementarity among multiple modes, to obtain a more robust comprehensive judgment, the offsets of each mode are weighted and fused according to their weights to obtain a unique energy localization index for each segment. It satisfies the following formula:

[0119] ;

[0120] in, Weight According to the first Factors such as the energy proportion, mass distribution, or short-term identification stability of each key mode under healthy conditions are determined to enhance the contribution of the key modes. To avoid interference from accidental noise or transient fluctuations, a time window can be used. right Robust statistics can be employed, such as using robust mean and median absolute deviation to estimate point estimates and their confidence ranges, or smoothing methods like interquartile ranges can be used. When local stiffness decreases due to fatigue damage, modal energy tends to concentrate near the damage area, exhibiting... A continuous occurrence of positive peak values ​​indicates the potential location of structural deterioration, providing a basis for subsequent damage localization.

[0121] In some embodiments of this application, step S150 involves performing a graded early warning judgment based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, and outputting a corresponding fatigue damage early warning signal, including the following steps.

[0122] Step S510: When the maximum value of the offset index exceeds the first threshold, it is determined to be a first-level warning state, and a first-level warning signal is output.

[0123] In step S510, preliminary identification and early warning of structural anomalies in wind turbine blades are achieved. When the maximum value of the offset index calculated by the system in all segments exceeds a preset first threshold, it indicates that the modal energy distribution of a certain part has significantly deviated from the healthy baseline, and there may be signs of local stiffness degradation or micro-damage initiation. At this time, the system automatically determines to enter the first-level warning state and outputs a first-level warning signal to mark the area as "suspicious," prompting operation and maintenance personnel to start paying attention to structural response changes at this location, thereby initiating the monitoring and tracking mechanism in the early stage of damage.

[0124] Step S520: When the duration of the first-level warning state exceeds the preset duration threshold, it is determined to be a second-level warning state, and a second-level warning signal is output.

[0125] In step S520, the initial warning is verified over time to eliminate false alarms caused by transient interference or short-term fluctuations. If the first-level warning status persists for more than a preset duration threshold, it indicates that the anomaly is not accidental but has a certain degree of persistence and development trend. At this time, the system upgrades the warning level to the second level, determines it as a "pay attention" state, and outputs a second-level warning signal. This step, by introducing a time stability judgment, effectively enhances the reliability of the warning results, ensuring that only areas with persistent anomaly characteristics will enter the more in-depth evaluation process.

[0126] In step S530, under the first or second warning state, the real mechanical strain time history signal is extracted at the segment position with the largest offset index. The strain cycle spectrum is obtained by rainflow counting method, and the fatigue accumulation and its growth rate are calculated by combining the material SN curve and Miner linear cumulative damage theory.

[0127] It should be noted that the rainflow counting method is a standard method for extracting equivalent stress or strain cycles from the time history of complex variable amplitude loads, and is widely used in fatigue life analysis of engineering structures. This method identifies peak-valley inflection points in the strain signal and decomposes the irregular load history into a series of semi-cycles or full cycles with definite amplitudes, mean values, and cycle numbers according to specific flow rules, thereby transforming the randomly fluctuating strain history into a discrete cyclic spectrum that can be used for fatigue calculation. In this application, the rainflow counting method is used to process the real mechanical strain time history signal extracted from the damaged area, generating a strain cyclic spectrum as input to Miner's linear cumulative damage theory, and then quantitatively assessing fatigue cumulative damage and its growth trend, providing a key basis for graded early warning.

[0128] In step S530, a refined quantitative analysis of fatigue damage is conducted in the area where persistent anomalies have been confirmed. When the system is in a first or second level warning state, the segment with the largest offset index is automatically located, and the true mechanical strain time history signal is extracted from that location. Subsequently, the rainflow counting method is used to process this time history, decomposing it into a series of strain cycles with definite amplitudes and cycle numbers, forming a strain cycle spectrum. Based on this, the strain amplitude is converted into the corresponding stress amplitude using the material's SN curve, and the current fatigue accumulation and its growth rate over time are calculated according to Miner's linear cumulative damage theory. This step achieves a quantitative connection from structural dynamic anomalies to the degree of material-level fatigue damage.

[0129] Step S540: When the maximum value of the offset index exceeds the second threshold and the growth rate of the fatigue accumulation exceeds the third threshold, it is determined to be a third-level warning state, and a third-level warning signal is output.

[0130] In step S540, the highest-level risk alarm is triggered to provide a basis for critical operation and maintenance decisions. When the maximum value of the offset indicator not only exceeds the higher second threshold, indicating that the local energy concentration phenomenon is very significant, but also that the growth rate of fatigue accumulation exceeds the third threshold, it indicates that the damage is in an accelerated expansion phase and the structural safety risk is rising sharply. Under this dual high-risk condition, the system determines the state to be in a third-level warning state and outputs a third-level warning signal, marked as "alarm". This judgment integrates the severity indicators of two dimensions: abnormal structural response and fatigue evolution trend, ensuring that the highest-level alarm is issued only when the damage is severe and rapidly deteriorating, thereby supporting the timely implementation of high-priority operation and maintenance measures such as load limiting, shutdown, or emergency maintenance.

[0131] In some embodiments of this application, the identified energy localization index Near the peak value, the potential fatigue damage area is located, and the corresponding mechanical strain time history signal is extracted from that location. Spatial coordinates omitted here This indicates that a local target point has been locked. Rainflow counting was used to process the data, resulting in a series of strain cycle spectra. ,in For the first Cyclic strain amplitude, This represents the number of occurrences of this type of cycle. The equivalent elastic modulus is used. strain amplitude Converted to the corresponding stress amplitude : ,in, Used to achieve the conversion of strain into stress.

[0132] Introducing the SN curve relationship of the material, we describe the power-law relationship between stress amplitude and fatigue life: ,in, For the first The fatigue life that a cycle can withstand (unit: number of cycles). and These are material constants, obtained through experimental calibration, and reflect the fatigue characteristics of the material.

[0133] Based on Miner's linear cumulative damage theory, the current time is defined as follows: The cumulative fatigue damage is: , of which each Indicates the first The proportion of damage contributed by loop-like structures, in total This indicates the cumulative degree of fatigue damage to the structure during the current operating phase.

[0134] To assess the development trend of damage, using time steps Define the rate of increase in fatigue accumulation, satisfying the following formula:

[0135] ;

[0136] in, This represents the rate of increase in fatigue damage per unit time, reflecting the speed of damage evolution. Combined with... The indicated location of the damage, This indicates the possible location of fatigue damage. It characterizes the degree of cumulative fatigue, while This quantifies the dynamic trend of fatigue accumulation. These three elements work together to form a complete damage evolution assessment system, providing crucial evidence for graded early warning.

[0137] In some embodiments of this application, an energy localization index is set. Threshold parameters: First threshold Second threshold , and duration threshold and growth rate threshold This is used to construct a three-tiered early warning mechanism, corresponding to "Suspicious," "Attention," and "Alarm" respectively. Based on the current segments... The values ​​and their dynamic changing trends are used to perform the following hierarchical judgments:

[0138] (1) When the maximum value of the offset index When the first threshold is exceeded, i.e. The system is classified as "suspicious" under Level 1 warning and a Level 1 warning signal is output. "Suspicious" indicates that a local modal energy anomaly has occurred in a certain area, which may indicate early signs of damage.

[0139] (2) When the duration of the first-level warning status Exceeding the preset duration threshold At that time, that is and The system is classified as a "watchful" state, triggering a Level 2 warning, and a Level 2 warning signal is output. "Watchful" indicates that the anomaly has persisted for more than a preset time. This indicates that it has potential for development and requires stronger monitoring.

[0140] (3) When the maximum value of the offset index The growth rate of accumulated fatigue exceeds the second threshold. When the third threshold is exceeded, that is and The system is classified as an "alarm" state, triggering a Level 3 warning, and a Level 3 warning signal is output. An "alarm" indicates a significant concentration of local modal energy and a rapid increase in fatigue accumulation rate. Exceeding the critical value This indicates that the structure is in a state of accelerated damage and poses a high safety risk.

[0141] When any level of warning is triggered, the system output includes: first, determining the location where energy localization is most significant, i.e., the current location. Largest segment number This refers to the physical location where fatigue damage is most likely to occur. Simultaneously, the system uses modal weights... Filter out the The set of dominant modes that contribute the most indicates which modes play a major role in fatigue damage accumulation. Additionally, the cumulative fatigue value at the current moment is output. and its growth rate This is used to quantify the severity and evolution trend of damage. Finally, the system provides operation and maintenance suggestions matching the warning level, such as "monitor and track," "re-check and inspect," "limited operation," or "maintenance shutdown." To support on-site verification, the system includes pre- and post-triggered parameters. The original total strain per second With temperature Data fragment.

[0142] In some embodiments of this application, dynamic regression between the three states is achieved through a conditional fallback mechanism. For example, if the judgment of the "suspicious" or "concerned" state is no longer met, the system can conditionally return to the previous safe state. This logic enables hierarchical early warning and closed-loop management of the structural damage evolution process, supporting real-time monitoring and risk response.

[0143] In some embodiments of this application, the energy localization index The three threshold parameters: the first threshold Second threshold Duration threshold and growth rate threshold Based on long-term operational data during the healthy period, adaptive calibration can be performed using methods such as quantile analysis or Bayesian updates to ensure that the early warning strategy is dynamically optimized according to environmental and aging conditions.

[0144] In some embodiments of this application, reference is made to Figure 3 , Figure 3This diagram illustrates the local modal strain energy distribution on a wind turbine blade under extreme wind conditions: a yaw error of 15°, rated wind speed coupled with extreme gusts, and high wind speed of 20 m / s coupled with extreme turbulent winds. The local modal energy information of the blade under different operating conditions can accurately characterize the location of fatigue damage. Combined with distributed fiber optic measurement, online monitoring and early warning can be performed. Specifically, a color gradient visually reflects the local energy concentration areas of the structure. The color scale on the right indicates the strain energy magnitude, from blue (low value) to red (high value), where the red area represents the high strain energy region, which is a key area of ​​concern for potential fatigue damage or structural weaknesses. The three blades in the diagram correspond to the strain energy distribution under different vibration modes or loading conditions: the upper blade shows a significant high strain energy accumulation area near the root at the leading edge; the middle and lower blades show locally densely distributed high-energy regions near the blade tip, appearing as dots or stripes, indicating that this area has undergone significant deformation under specific modal excitation. The red dashed boxes in the diagram represent "high modal strain energy regions," emphasizing that these areas require key monitoring and life assessment. Such visualization results can be used to guide structural optimization design, damage early warning, and health status assessment.

[0145] Secondly, refer to Figure 4 This application provides a wind turbine blade fatigue damage early warning system based on modal energy shift, including a sensing layer, a data acquisition layer, an edge computing layer, and a cloud processing layer.

[0146] The sensing layer consists of distributed sensing optical fibers embedded in the main spar, web, leading edge, trailing edge, and root transition zone of the wind turbine blade. These fibers include strain sensing fibers and temperature sensing fibers, which are attached to the inner surface of the wind turbine blade in an S-shaped path. At least one fiber loop is added in the high-stress area of ​​the blade to ensure full coverage along the blade length and improve local spatial resolution. At the same time, at least one fiber loop made of optical fiber is added in this high-stress concentration area in the blade to enhance the monitoring sensitivity of this vulnerable part, thereby providing high-quality raw sensing data for the entire system.

[0147] The acquisition layer is configured to receive the optical signal output from the sensing layer and restore the optical signal to the current total strain signal and current temperature distribution through an optical fiber demodulator, generating a structural response data packet as input for subsequent processing steps. This ensures that the data is synchronized in time and uniform in format, providing a reliable data source for subsequent noise reduction, decoupling, and modal analysis of the system.

[0148] The edge computing layer is deployed inside the wind turbine nacelle or tower and connected to the acquisition layer. It is configured to perform data synchronization, noise reduction preprocessing, temperature-strain decoupling, and operating condition mode identification on the structural response data packets, and extract the current natural frequency and current strain mode shape.

[0149] Specifically, the edge computing layer performs preliminary data processing and feature extraction near the wind turbine site, and then uploads the processed data to the server cluster of the cloud processing layer via 5G or 4G communication modules to reduce transmission load and improve response speed. This layer is deployed inside the nacelle or tower of the wind turbine and is directly connected to the acquisition layer. It can perform multi-step processing on the structural response data packets: first, time synchronization is performed to ensure consistency of multi-channel data; then, noise reduction preprocessing is implemented to suppress high-frequency noise and spike interference; next, temperature-strain decoupling is performed based on a healthy baseline library to separate the true mechanical strain signal; finally, a condition modal identification method is used to extract the current natural frequency and current strain mode shape from the signal. This series of operations is completed at the edge, effectively improving the system's real-time performance and anti-interference capability.

[0150] The cloud processing layer is configured as follows: it stores a pre-built multi-dimensional health baseline library, calculates the current modal energy ratio of each discrete segment based on the current strain mode shape, generates an offset index characterizing the degree of modal energy localization by combining the initial modal energy ratio, and further performs graded early warning judgment based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, generates and issues corresponding fatigue damage early warning signals, and completes the closed loop from data to decision.

[0151] In some embodiments of this application, the system further includes an operation and maintenance terminal. The operation and maintenance terminal is configured to receive fatigue damage early warning signals sent from the cloud processing layer, and to realize the visualization of the wind turbine blade structural status, the presentation of early warning information, and the automatic generation of operation and maintenance reports.

[0152] The role of the operation and maintenance terminal is to provide wind farm operation and maintenance personnel with an intuitive and efficient human-machine interface and decision support tools. This terminal receives fatigue damage early warning signals from the cloud processing layer, transforming complex structural health status information into easily understandable visualizations, including the damage location of different blade regions, the distribution of offset indicators, dominant mode information, and key parameters such as fatigue accumulation and growth rate. Simultaneously, the terminal presents tiered early warning information in real time, such as suspicious, watchful, or alarm status, and automatically associates corresponding handling suggestions based on the warning level. Furthermore, the operation and maintenance terminal can automatically generate structured operation and maintenance reports based on historical monitoring data and current early warning results, covering event summaries, data analysis, risk assessments, and subsequent action recommendations, thereby improving the timeliness, standardization, and intelligence of operation and maintenance responses.

[0153] In summary, the wind turbine blade fatigue damage early warning method and system based on modal energy shift provided in this application have the following technical effects.

[0154] This application achieves full-scale, high spatial resolution synchronous strain and temperature sensing by continuously laying distributed sensing optical fibers in the critical load-bearing areas of the blade, effectively covering internal structures that traditional point sensors cannot reach. Based on a healthy, undamaged state, a multi-dimensional baseline library is established, including natural frequencies, strain mode shapes, modal energy ratios, and thermo-mechanical coupling parameters, providing a physically interpretable quantitative benchmark for damage identification. By performing noise reduction preprocessing and temperature-strain decoupling on the total strain signal, the true mechanical strain is accurately extracted, significantly improving data reliability. By integrating frequency domain decomposition and time-domain random subspace identification methods, operating modal parameters are robustly acquired under conditions without external excitation, enhancing the robustness of modal identification under complex operating conditions. Utilizing the modal energy localization mechanism, the current modal energy is compared with the initial modal energy. The system generates offset indices proportionally, enabling highly sensitive capture and precise positioning of local stiffness degradation. Furthermore, by combining rainflow counting and Miner's cumulative damage theory, it quantifies the cumulative fatigue amount and its growth rate at the damage location, ensuring that the early warning not only reflects the presence of anomalies but also the trend of damage evolution. Through a three-level hierarchical early warning mechanism, it combines structural response anomalies with fatigue acceleration states, outputting multi-dimensional early warning information including location, dominant mode, fatigue state, and original data fragments, supporting differentiated operation and maintenance decisions. The entire system adopts a collaborative architecture of perception layer, acquisition layer, edge computing layer, and cloud processing layer, balancing real-time performance, accuracy, and intelligence. Ultimately, it achieves mechanism-driven, online, and high-precision early warning of early fatigue damage in wind turbine blades, overcoming the limitations of existing technologies that can only identify or qualitatively judge damage after the fact.

[0155] It should be noted that in all specific embodiments of this application, all data processing activities related to user identity or personal characteristics, such as user information, user behavior data, historical data, and location information, will be conducted in accordance with the principles of legality, legitimacy, and necessity. All data collection, use, storage, and processing will be subject to compliance with applicable national and regional laws, regulations, and industry standards, and informed consent from users will be obtained in a clear and explicit manner before processing. For the processing of sensitive personal information, separate consent from users will be obtained through prominent means such as pop-up prompts and independent confirmation pages. If any processing conflicts with laws and regulations, the laws and regulations will prevail, and necessary data processing will only be carried out within the scope permitted by laws and regulations, ensuring that all data-based applications, analyses, and technical implementations are conducted within the scope permitted by laws and regulations.

[0156] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0157] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary skill. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0158] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0160] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Additionally, computer-readable media can even be paper or other suitable media on which programs can be printed, for example, by optically scanning the paper or other media, then editing, interpreting, or, if necessary, processing it in a suitable manner to obtain the program electronically, and then storing it in computer memory.

[0161] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0162] In the foregoing description of this specification, the reference to terms such as "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations," etc., indicates that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in an embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0163] Although embodiments of the invention have been shown and described, those skilled in the art will understand 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 claims and their equivalents.

[0164] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A method for early warning of fatigue damage in wind turbine blades based on modal energy migration, characterized in that, Includes the following steps: When wind turbine blades are in a healthy and undamaged state, establish a multi-dimensional health baseline database; The health baseline library includes initial natural frequencies, initial strain mode shapes, initial modal energy ratios, initial temperature distributions, temperature-strain sensitivity coefficients, and zero-drift terms. During the operation of the wind turbine, the current total strain signal and the current temperature distribution are collected synchronously, and the temperature-strain decoupling of the current total strain signal is performed based on the health baseline library to obtain the real mechanical strain signal; The specific method of temperature-strain decoupling is as follows: calculate the thermal strain component obtained by multiplying the difference between the current temperature distribution and the initial temperature distribution by the position-related temperature-strain sensitivity coefficient; subtract the thermal strain component from the current total strain signal, and then subtract the position-related zero drift term to obtain the real mechanical strain signal caused by mechanical load. Based on the actual mechanical strain signal, the current natural frequency and current strain mode shape are extracted using the working condition modal identification method, including the following steps: Frequency domain decomposition analysis was performed on the real mechanical strain signal to preliminarily identify the frequency peaks of the dominant modes and the corresponding coarse strain mode shapes as the first analysis results. The real mechanical strain signal is subjected to time-domain random subspace identification analysis to obtain the system state space model, and the frequency, damping ratio and strain mode shape of the stable mode are extracted as the second analysis result. The first and second analysis results are cross-validated and fused, and modes with consistent frequencies are retained to extract the current natural frequency and current strain mode shape. Based on the current strain mode shape, the current mode energy ratio is calculated, and combined with the initial mode energy ratio, a shift index characterizing the degree of modal energy localization is generated; Based on the offset index and its dynamic trend, combined with the fatigue cumulative damage assessment results, a graded early warning judgment is performed, and a corresponding fatigue damage early warning signal is output.

2. The wind turbine blade fatigue damage early warning method based on modal energy migration according to claim 1, characterized in that, Distributed sensing optical fibers are continuously laid along the optical fiber path in the main beam, web, leading edge, trailing edge and blade root transition area of ​​the wind turbine blade. The distributed sensing optical fibers include strain sensing optical fibers and temperature sensing optical fibers to collect the current total strain signal and current temperature distribution of the wind turbine blade. The current total strain signal includes mechanical strain signal and thermal strain signal. The distributed sensing optical fiber is attached to the inner wall surface of the wind turbine blade in an S-shaped path, and at least one optical fiber loop is added in the middle region of the blade to enhance the monitoring sensitivity of high stress areas.

3. The wind turbine blade fatigue damage early warning method based on modal energy migration according to claim 1, characterized in that, Before performing temperature-strain decoupling, the current total strain signal is subjected to noise reduction preprocessing. The noise reduction preprocessing adopts a combination of wavelet threshold denoising and sliding median filtering. The wavelet basis function and the number of decomposition layers are preset according to the frequency range of the dominant mode of the blade, so as to retain the key spatial distribution characteristics of the strain mode shape while suppressing high-frequency noise and spike interference.

4. The wind turbine blade fatigue damage early warning method based on modal energy migration according to claim 1, characterized in that, The process of calculating the current modal energy ratio based on the current strain mode shape and generating a shift index characterizing the degree of modal energy localization, combined with the initial modal energy ratio, includes the following steps: The wind turbine blade is divided into multiple discrete segments along its length. Based on the current strain mode shape, the local modal energy of each mode in each discrete segment is calculated and normalized to obtain the normalized current modal energy ratio. The current modal energy ratio is compared with the initial modal energy ratio, and the relative energy offset of each discrete segment is calculated. The relative energy offsets under different operating conditions or different modes are weighted and fused to generate a comprehensive offset index, which is used to characterize the degree of spatial localization of modal energy.

5. The wind turbine blade fatigue damage early warning method based on modal energy migration according to claim 4, characterized in that, The weighting coefficients of the relative energy offset under different operating conditions or different modes are determined based on the energy proportion, mass distribution or short-term stability of the corresponding mode under healthy conditions.

6. The wind turbine blade fatigue damage early warning method based on modal energy migration according to claim 1, characterized in that, The step of performing a graded early warning judgment and outputting a corresponding fatigue damage early warning signal based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, includes the following steps: When the maximum value of the offset index exceeds the first threshold, it is determined to be a first-level warning state, and a first-level warning signal is output. When the duration of the first-level warning state exceeds a preset duration threshold, it is determined to be a second-level warning state, and a second-level warning signal is output. In the first-level warning state or the second-level warning state, the real mechanical strain time history signal is extracted at the segment position with the largest offset index, the strain cycle spectrum is obtained by rain flow counting method, and the fatigue accumulation and its growth rate are calculated by combining the material SN curve and Miner linear cumulative damage theory. When the maximum value of the offset index exceeds the second threshold and the growth rate of the accumulated fatigue exceeds the third threshold, it is determined to be a third-level warning state, and a third-level warning signal is output.

7. A wind turbine blade fatigue damage early warning system based on modal energy migration, characterized in that, It includes a perception layer, a data acquisition layer, an edge computing layer, and a cloud processing layer; The sensing layer consists of distributed sensing optical fibers embedded in the main beam, web, leading edge, trailing edge and blade root transition area of ​​the wind turbine blade, including strain sensing optical fibers and temperature sensing optical fibers. It is attached to the inner wall surface of the wind turbine blade in an S-shaped path, and at least one optical fiber ring is added in the high stress area of ​​the blade. The acquisition layer is configured to receive the optical signal output by the sensing layer, and use an optical fiber demodulator to restore the optical signal to the current total strain signal and the current temperature distribution, thereby generating a structural response data packet. The edge computing layer is deployed inside the wind turbine nacelle or tower and connected to the acquisition layer. It is configured to perform data synchronization, noise reduction preprocessing, temperature-strain decoupling and operating condition mode identification on the structural response data packet, and extract the current natural frequency and current strain mode shape. The specific method of temperature-strain decoupling is as follows: calculate the thermal strain component obtained by multiplying the difference between the current temperature distribution and the initial temperature distribution by the position-related temperature-strain sensitivity coefficient; subtract the thermal strain component from the current total strain signal, and then subtract the position-related zero drift term to obtain the real mechanical strain signal caused by mechanical load. Based on the actual mechanical strain signal, the current natural frequency and current strain mode shape are extracted using the working condition modal identification method, including the following steps: Frequency domain decomposition analysis was performed on the real mechanical strain signal to preliminarily identify the frequency peaks of the dominant modes and the corresponding coarse strain mode shapes as the first analysis results. The real mechanical strain signal is subjected to time-domain random subspace identification analysis to obtain the system state space model, and the frequency, damping ratio and strain mode shape of the stable mode are extracted as the second analysis result. The first and second analysis results are cross-validated and fused, and modes with consistent frequencies are retained to extract the current natural frequency and current strain mode shape. The cloud processing layer is configured to: store a pre-built multidimensional health baseline library; calculate the current modal energy ratio of each discrete segment based on the current strain mode shape; generate an offset index characterizing the degree of modal energy localization by combining the initial modal energy ratio; and further perform graded early warning judgment based on the offset index and its dynamic change trend, combined with the fatigue cumulative damage assessment results, and generate and issue corresponding fatigue damage early warning signals.

8. The wind turbine blade fatigue damage early warning system based on modal energy migration according to claim 7, characterized in that, The system also includes an operation and maintenance terminal; the operation and maintenance terminal is configured to receive fatigue damage early warning signals sent by the cloud processing layer, and realize the visualization of the wind turbine blade structural status, the presentation of early warning information, and the automatic generation of operation and maintenance reports.

Images