A method and system for detecting the state of a wind power bolt

Abstract

The present invention discloses a method and system for detecting the state of wind power bolts, belonging to the technical field of mechanical structure testing and health monitoring. It includes obtaining real-time strain data of each monitoring point; obtaining real-time operating condition data of the unit; dynamically generating a finite element model of the tower barrel structure according to the operating condition data; combining the finite element model and the real-time strain data to generate a stress distribution map of all bolts in the tower barrel; inputting the stress distribution map into a preset fatigue damage model to calculate the bolt life attenuation rate and locate the external load abnormal source; when it is detected that the life attenuation rate exceeds the preset threshold, generating a hierarchical warning signal and outputting a maintenance strategy according to the positioning result. The present invention adopts a wireless strain sensor network, dynamic finite element model correction and multi-dimensional fatigue damage analysis technology to generate a stress distribution map of the entire tower barrel in real time and associate it with external condition data, and can accurately evaluate the bolt life attenuation rate, locate the load abnormal source and output a hierarchical maintenance strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical structure testing and health monitoring, and in particular to a method and system for detecting the state of wind power bolts. Background Art

[0002] Wind power tower bolts are subjected to alternating loads and complex environmental effects for a long time, and are prone to fatigue fracture, resulting in structural failure. Traditional detection relies on regular manual inspections and local sensor networks, and can only obtain static data at discrete time points, making it difficult to detect sudden stress anomalies under dynamic loads in a timely manner.

[0003] In the prior art, a finite element model with fixed parameters is used for stress prediction, but the influence of real-time working conditions such as impeller speed and pitch action on the structural boundary conditions is not considered, resulting in a significant deviation between the model calculation results and the actual stress distribution. At the same time, the abnormal warning relies on a single strain threshold determination and cannot distinguish different failure modes such as mechanical resonance and temperature deformation.

[0004] Due to the neglect of the dynamic coupling relationship of the working condition stress, such methods often have false alarms or missed detections. Moreover, the source of abnormal load cannot be located, resulting in lagging or insufficient targeted maintenance measures, making it difficult to effectively extend the service life of bolts and reduce maintenance costs. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method and system for detecting the state of wind power bolts, which adopt a wireless strain sensor network, dynamic finite element model correction, and multi-dimensional fatigue damage analysis technology to generate a stress distribution map of the entire tower in real time and associate external working condition data, and can accurately evaluate the bolt life attenuation rate, locate the source of abnormal load, and output a hierarchical maintenance strategy.

[0006] The above object can be achieved by the following solutions:

[0007] A method for detecting the state of wind power bolts includes setting monitoring points at preset circumferential positions on each layer of the tower barrel to obtain real-time strain data of each monitoring point; obtaining the operating condition data of the unit in real time from the wind farm telecontrol communication system, including wind speed, pitch and yaw data, generator torque and impeller speed; dynamically generating a finite element model of the tower barrel structure according to the operating condition data, and correcting the finite element model by using the real-time strain data; combining the finite element model and the real-time strain data, and using the stress field gradient compensation algorithm to calculate the strain of unmonitored points, generating a stress distribution map of all bolts in the tower barrel; inputting the stress distribution map into a preset fatigue damage model, calculating the bolt life attenuation rate, and performing time series correlation analysis on the real-time strain data and the pitch and yaw data in the operating condition data to locate the abnormal source of external load and obtain a positioning result; when it is detected that the life attenuation rate exceeds a preset threshold, generating a hierarchical warning signal and outputting a maintenance strategy according to the positioning result.

[0008] Optionally, the setting of monitoring points at preset circumferential positions on each layer of the tower barrel includes: symmetrically arranging a first sensor subset on both sides of the main wind direction impact surface of the tower barrel, and setting the circumferential spacing between adjacent sensors at a preset angle; arranging a second sensor subset in the high bending moment area of the tower barrel flange connection, and collecting three-dimensional strain components in an orthogonal arrangement manner.

[0009] Optionally, the dynamically generating a finite element model of the tower barrel structure according to the operating condition data includes: generating an equivalent centrifugal force load according to the impeller speed, and calculating the torque fluctuation value of the transmission chain according to the generator torque; performing vector synthesis on the equivalent centrifugal force load and the torque fluctuation value of the transmission chain to generate a dynamic load matrix that changes with time; inputting the dynamic load matrix into a reference model established based on the original design parameters of the tower barrel to generate a finite element model including time-varying boundary conditions.

[0010] Optionally, the calculating the strain of unmonitored points by using the stress field gradient compensation algorithm to generate a stress distribution map of all bolts in the tower barrel includes: extracting the strain difference between adjacent monitoring points from the real-time strain data, and calculating the curvature change characteristic quantity of the local area; comparing the curvature change characteristic quantity with the theoretical gradient of the finite element model, and calculating the residual stress deviation coefficient; iteratively adjusting the elastic modulus of the unmonitored points according to the residual stress deviation coefficient until the error between the predicted stress calculated by the finite element model and the measured stress converges to a set interval, and outputting the predicted stress to construct a stress distribution map.

[0011] Optionally, input the stress distribution map into a preset fatigue damage model, calculate the bolt life attenuation rate, and perform time series correlation analysis on the real-time strain data and the pitch and yaw data in the operating condition data to locate the external load anomaly source. The obtained location result includes: obtaining the S-N curve of the bolt material and historical load spectrum data, establishing the mapping relationship between the strain amplitude and the fatigue damage degree, and calculating the bolt life attenuation rate; performing rain flow counting method processing on the stress distribution map to extract the equivalent alternating stress amplitude sequence; performing convolution operation on the equivalent alternating stress amplitude sequence and the pitch and yaw data to generate a fatigue cumulative factor coupled with the working condition, and locating the external load anomaly source to obtain the location result.

[0012] Optionally, the method further includes: when the deviation between the measured stress and the predicted stress continuously exceeds the set number of times, extracting the stress phase delay characteristics of each monitoring point; adjusting the weight ratio of the damping coefficient and the material yield strength in the finite element model according to the phase delay characteristics; storing the corrected model parameters in the historical database as the reference value for the next model initialization.

[0013] Optionally, the generation basis of the hierarchical warning signal is: constructing a multi-dimensional warning parameter set including the stress fluctuation entropy value, the temperature load sensitivity, and the environmental corrosion rate; triggering a status prompt when a single parameter exceeds the first-level threshold, and triggering an emergency shutdown instruction when the combined effect of at least two parameters exceeds the second-level threshold.

[0014] Optionally, the calculation of the stress fluctuation entropy value includes: performing Fourier transform on the strain data within a specified time window, and extracting the energy distribution characteristics of a preset frequency band; calculating the stress fluctuation entropy value according to the energy distribution characteristics, and determining it as the resonance risk mode when the energy ratio in the low-frequency band exceeds the preset ratio.

[0015] Optionally, the method further includes: when the resonance risk mode is recognized, synchronously collecting the axial acceleration data of the unit vibration sensor; performing amplitude-frequency correlation analysis on the acceleration data and the low-frequency band energy, and adding the debugging task of the vibration suppression device to the maintenance strategy if the correlation coefficient is greater than the preset correlation value.

[0016] Based on the same inventive concept, the present invention also provides a detection system for the state of wind power bolts. The system includes: a wireless strain sensor group deployed at preset circumferential positions on each layer of the tower barrel to obtain real-time strain data of each monitoring point; a working condition data acquisition module for obtaining the operating condition data of the unit in real time from the wind farm telecontrol communication system, including wind speed, pitch and yaw data, generator torque and impeller speed; a finite element model construction module for dynamically generating a finite element model of the tower barrel structure according to the operating condition data and correcting the finite element model by using the real-time strain data; a stress distribution generation module for combining the finite element model and the real-time strain data and using a stress field gradient compensation algorithm to estimate the strain of unmonitored points and generate a stress distribution map of all the bolts on the tower barrel; an abnormal analysis and positioning module for inputting the stress distribution map into a preset fatigue damage model, calculating the bolt life attenuation rate and performing a time series correlation analysis on the real-time strain data and the pitch and yaw data in the operating condition data to locate the external load abnormal source and obtain a positioning result; and a warning decision module for generating a hierarchical warning signal and outputting a maintenance strategy according to the positioning result when it is detected that the life attenuation rate exceeds a preset threshold.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. By deploying a wireless strain sensor group and integrating dynamic finite element model correction technology, the present invention realizes high-precision real-time estimation of the stress distribution of all the bolts on the tower barrel. Combining with the stress field gradient compensation algorithm, it breaks through the limitations of traditional single-point monitoring, can accurately predict the strain values at key positions where sensors are not arranged, and significantly improves the monitoring coverage rate and data reliability.

[0019] 2. By using the time series correlation analysis of the operating condition data and the strain signal, the present invention can accurately locate the external load abnormal source. By means of convolution operation, it reveals the causal relationship between the pitch and yaw actions and the stress fluctuations, solves the problem of fuzzy correlation analysis of abnormal events in the traditional method, and provides a direct basis for targeted maintenance.

[0020] 3. The present invention establishes a multi-dimensional warning parameter set and a hierarchical decision-making mechanism, comprehensively evaluates the coupling effects of multiple factors such as stress fluctuation entropy value, temperature sensitivity and corrosion rate, and realizes the dynamic response from early warning to emergency shutdown. Compared with the single-threshold alarm mechanism, it effectively reduces the false alarm rate and prevents malignant failures.

[0021] 4. The present invention constructs a life attenuation rate calculation model based on the fatigue damage model and historical data, which can dynamically update the predicted value of the remaining life of the bolts. Combining with the rain flow counting method and the material S-N curve, it quantifies the influence of alternating stress on the bolt aging and provides a scientific basis for formulating a preventive maintenance plan.

[0022] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention may be realized and attained by the structure particularly pointed out in the specification, claims as well as the drawings. Brief Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 It is a schematic flowchart of a method for detecting the state of a wind power bolt according to an embodiment of the present invention.

[0025] Figure 2 It is a schematic diagram of the layout of circumferential sensors on a tower barrel according to an embodiment of the present invention.

[0026] Figure 3 It is a circumferential stress distribution map of a tower barrel according to an embodiment of the present invention.

[0027] Figure 4 It is an S-N curve of the material of a wind power bolt according to an embodiment of the present invention.

[0028] Figure 5 It is a bolt life attenuation curve according to an embodiment of the present invention.

[0029] Figure 6 It is a schematic structural diagram of a system for detecting the state of a wind power bolt according to an embodiment of the present invention. Detailed Embodiments

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0031] Referring to Figure 1 , an embodiment of the present invention provides a method for detecting the state of a wind power bolt. By using a wireless strain sensor network, dynamic finite element model updating, and multi-dimensional fatigue damage analysis techniques, it can generate a circumferential stress distribution map of the entire tower barrel in real time and correlate with external operating condition data, and can accurately evaluate the bolt life attenuation rate, locate the source of abnormal load, and output a hierarchical maintenance strategy.

[0032] The method of this embodiment specifically includes:

[0033] Monitoring points are set at preset circumferential positions on each layer of the tower barrel to obtain real-time strain data of each monitoring point;

[0034] Specifically, according to the structural mechanical characteristics of the tower barrel, wireless strain sensors are distributed in the flange connection area and the windward surface in the main wind direction of each layer of the tower barrel, and the sensing data is uploaded to the edge computing node through the LoRa communication protocol. Among them, the wireless strain sensor is of passive design, which converts mechanical strain into a charge signal through the piezoelectric effect, and generates a digital strain waveform through a signal conditioning circuit. Among them, the preset circumferential position is the sensor layout point preset according to the stress simulation result of the tower barrel, covering the high-stress concentration area of the flange bolt group; the real-time strain data is the micro-strain value collected by the sensor, which characterizes the deformation amount of the bolt connection part; the wireless strain sensor group is a distributed monitoring network composed of multiple sensors, supporting synchronous sampling and data fusion.

[0035] Exemplarily, at the second-layer flange of the tower barrel of a 2MW wind turbine, 16 wireless strain sensors are arranged at intervals of 50 degrees and compared with a reference resistance strain gauge. The sensor group successfully captures the asymmetric distribution characteristics of the flange surface strain when the wind speed suddenly changes, and the data delay is less than 200ms. Through a reasonable layout strategy, dynamic monitoring of key areas is achieved, and the passive design avoids interference from the power supply line to the tower barrel structure, ensuring long-term reliability.

[0036] Obtain the operating condition data of the unit in real time from the wind farm telecontrol communication system, including wind speed, pitch and yaw data, generator torque and impeller speed;

[0037] Specifically, it is docked with the wind farm monitoring system (SCADA) through the power system IEC 60870-5-104 communication protocol, and the real-time operating parameters of the unit are read at a frequency of once per second. The original message is converted into a structured condition data stream through a data parsing module and aligned with the strain data timestamp. Among them, the telecontrol communication system is a data transmission system based on a dedicated power network, ensuring the secure isolation of sensitive data; the generator torque is the rotational torque transmitted from the wind wheel to the generator, reflecting the mechanical load intensity; the impeller speed is the number of revolutions per minute (RPM) of the wind wheel, characterizing the dynamic characteristics of the aerodynamic load.

[0038] Dynamically generate a finite element model of the tower barrel structure according to the operating condition data, and correct the finite element model by using the real-time strain data;

[0039] Specifically, a baseline finite element model is established based on the tower design drawings, including shell units to simulate the tower wall and beam units to simulate the connecting bolts. The equivalent wind pressure distribution load is calculated according to the real-time wind speed, the centrifugal force load vector is generated according to the impeller speed, and the model boundary conditions are updated through the dynamic load matrix. The real-time strain data is substituted into the corresponding nodes to correct the displacement of the finite element model. Among them, the finite element model is a numerical model that discretizes the tower into grid units for mechanical calculations; the dynamic load matrix is ​​the load distribution data containing time variables, reflecting the transient characteristics of external excitation.

[0040] Combining the finite element model and the real-time strain data, a stress field gradient compensation algorithm is used to calculate the strain of unmonitored points to generate a stress distribution map of the entire tower bolts;

[0041] Specifically, the measured strain difference between adjacent monitoring points is extracted to calculate the curvature change of the local area. Combined with the stress gradient theoretical value of the finite element model, the residual stress correction coefficient is fitted by the least squares method. The strain distribution of the unmonitored points is iteratively calculated using the corrected material constitutive equation until the convergence condition is met. Among them, the stress field gradient compensation algorithm is a dynamic correction method based on the difference between the measured data and the model prediction; the residual stress correction coefficient is a weighting factor that characterizes the model prediction error.

[0042] Input the stress distribution map into a preset fatigue damage model, calculate the bolt life decay rate, and perform time series correlation analysis on the real-time strain data and the pitch yaw data in the operating condition data to locate the abnormal source of the external load and obtain a positioning result;

[0043] Performing time series correlation analysis on the real-time strain data corresponding to the high-risk position and the pitch yaw data in the operating condition data to locate the abnormal source of the external load and obtain a positioning result;

[0044] Specifically, the strain fluctuation data of high-risk bolts is subjected to short-time Fourier transform to extract the dominant frequency component. A cross-correlation analysis is performed with the time domain signals of the pitch angle and yaw error angle to identify the phase synchronization between the two. If a frequency component matches the pitch action cycle, it is determined to be a resonance caused by abnormal aerodynamic load.

[0045] When it is detected that the life decay rate exceeds a preset threshold, a graded warning signal is generated and a maintenance strategy is output according to the positioning result.

[0046] Specifically, a three-level early warning mechanism is defined based on the predicted value of the remaining life of the bolts: the first level is an observation reminder, the second level is a planned maintenance suggestion, and the third level is an emergency shutdown instruction. The location results of the associated abnormal source are used to generate targeted maintenance plans, such as calibration of the variable pitch system or re-tightening of the loose flange surface.

[0047] Specifically, this method establishes a closed-loop feedback mechanism for bolt strain monitoring and external loads by integrating passive sensor networks, real-time operating condition data, and dynamic finite element modeling techniques. By dynamically correcting model parameters to compensate for theoretical calculation deviations and using multi-source data correlation analysis to locate the source of abnormal loads, it solves the problem of insufficient accuracy caused by the static model and isolated data in traditional monitoring.

[0048] Exemplarily, this method was implemented on a 3MW unit at a wind farm in Inner Mongolia, successfully capturing the abnormal stress fluctuations of the bolt group on the third layer of the tower during the passage of a typhoon. By comparing SCADA data, the source of the abnormality was located as the asymmetric wind load caused by the delayed response of the pitch system. A secondary warning was triggered and it was recommended to adjust the pitch control parameters. The detection after maintenance showed that the peak bolt stress in this area decreased by at least 30%. Through data fusion and dynamic modeling, a rapid response to complex operating conditions was achieved, significantly improving the accuracy of early warning and the pertinence of maintenance measures, and avoiding the losses caused by unplanned outages.

[0049] Optionally, as Figure 2 shown, the setting of monitoring points at preset circumferential positions on each layer of the tower includes:

[0050] Symmetrically arrange the first sensor subset on both sides of the impact surface of the main wind direction of the tower, and set the circumferential spacing between adjacent sensors at a preset angle;

[0051] Specifically, according to the analysis of the wind vibration response characteristics of the tower, determine the central axis of the impact surface of the main wind direction, and use the range within 10 degrees on both sides of this axis as the deployment area of the first sensor subset. Each subset contains 8 - 12 wireless strain sensors, i.e., main wind direction sensors, which are mirror-symmetrically distributed with the tower center line as the axis of symmetry. The circumferential spacing between adjacent sensors satisfies 45 ≤ ≤ 60 degrees, and the calculation formula for the total arc length covered is:

[0052] ;

[0053] In the formula, is the number of sensors in a single subset, when taking 50 degrees , 8 sensors are taken as integer values in actual configuration. The installation position is calibrated for circumferential distribution by a laser locator to ensure the continuity of strain gradient measurement between adjacent sensors. The impact surface of the main wind direction is an arc-shaped area of 15 degrees on the left and right of the windward surface of the tower, which is most significantly affected by aerodynamic loads; the first sensor subset is a cluster of symmetrically distributed strain monitoring nodes, which monitors the uniformity of strain distribution in the main load direction through cross-validation; the circumferential spacing of 45 - 60 degrees is an optimized interval value considering both monitoring density and redundancy to prevent missed detection in key areas when wind loads change suddenly.

[0054] A second sensor subset is arranged in the bending moment area at the flange connection of the tower barrel, and three-dimensional strain components are collected in an orthogonal arrangement.

[0055] Specifically, an orthogonal strain gauge group, that is, a flange surface sensor, is arranged in an annular area extending 20 cm above and below the flange connection surface. Each group contains three mutually orthogonal strain detection directions. The strain gauge in the horizontal direction is parallel to the tangent direction of the flange surface, the vertical direction coincides with the axial direction of the bolt, and the third strain gauge is set at 45 degrees obliquely. Three groups of orthogonal units are configured for a single flange surface to form a three-dimensional monitoring grid. Calculate the three-dimensional principal stress direction through the orthogonal strain data:

[0056] ;

[0057] ;

[0058] In the formula, is the maximum value of the in-plane principal stress, is the minimum value of the in-plane principal stress, , are the strains in the orthogonal axis directions, is the shear strain component. Among them, the bending moment area is the area where the bending moment value of the flange connection surface exceeds 300% of the average bending moment of the tower barrel wall; the second sensor subset is a high-density layout group for multi-directional strain measurement; the orthogonal arrangement is a strain gauge group configuration with three-direction staggered distribution, which is used to decouple each stress component in the complex stress state.

[0059] Exemplarily, at the flange of the fourth layer of the tower barrel of a certain 5MW offshore unit, 9 sensors are arranged on both sides of the main wind direction impact surface, with a circumferential spacing of 55 degrees. Three groups of orthogonal sensors are deployed in the high bending moment area of the flange surface, with a spacing of 120 degrees between each group. During on-site testing, it was monitored that the horizontal strain at the 8th bolt position increased suddenly, and the synchronous orthogonal sensor showed that the shear strain increased by up to 120 μɛ. Through model inversion, it was found that there was a moment redistribution caused by bolt loosening at this point. The strain asymmetry rate in the abnormal area was verified to exceed 50% through the data of the symmetrically arranged first subset, and the data of the orthogonally arranged second subset accurately identified the failure mode dominated by shear strain, supporting the maintenance personnel to implement positioning and tightening operations. The symmetric layout strategy successfully captures the spatial asymmetry characteristics of the load distribution, and the three-dimensional strain detection with orthogonal arrangement effectively distinguishes the coupling effect of bending and shear loads. Compared with the single-direction layout scheme, this method accurately identifies the change of the complex stress state caused by bolt loosening through multi-directional data fusion, avoids false alarms caused by misjudging the excessive pure tensile stress, and improves the discrimination accuracy of defect types.

[0060] Optionally, the dynamically generating a finite element model of the tower barrel structure according to the operating condition data includes:

[0061] Generate an equivalent centrifugal force load based on the impeller rotational speed, and calculate the torque fluctuation value of the transmission chain according to the generator torque;

[0062] Specifically, the actual impeller rotational speed value is used to calculate the equivalent centrifugal force load based on the rigid body dynamics principle. Set up an impeller mass distribution data storage module, which contains the centroid coordinates and mass data of each blade. When calculating the equivalent centrifugal force, the impeller is discretized into multiple mass micro-elements, and the total centrifugal force is calculated to obtain the equivalent centrifugal force load. For the equivalent centrifugal force load , there is:

[0063] ;

[0064] In the formula, is the number of mass micro-elements, is the mass of the th mass micro-element, is the angular velocity of the impeller, is the th radial distance of the mass micro-element from the rotation axis. After the generator torque data is compensated by the slip frequency, the torque fluctuation value of the transmission chain is obtained using the torque pulsation rate calculation formula. For the torque fluctuation value of the transmission chain , there is:

[0065] ;

[0066] Among them, , are constant coefficients, is the average torque within the statistical period, is the gearbox meshing fundamental frequency, is the random disturbance component. The impeller rotational speed is a quantization characterization parameter of the wind turbine rotor angular velocity; the equivalent centrifugal force load is the equivalent value of the inertial force generated by the impeller rotary motion; the torque fluctuation value of the transmission chain is the torque oscillation amplitude under the combined action of gear meshing and electrical control.

[0067] Perform vector synthesis on the equivalent centrifugal force load and the torque fluctuation value of the transmission chain to generate a dynamic load matrix that changes with time;

[0068] Specifically, set the dimension of the dynamic load matrix to Q×3, where Q is the number of time nodes, and the three-dimensional space coordinates. Align the equivalent centrifugal force load and the torque fluctuation value of the transmission chain according to the time series, and use the spatial vector superposition method to calculate the load components. For the dynamic load vector , there is:

[0069] ;

[0070] In the formula, , are load coupling weight coefficients, is the direction vector of the centrifugal force (radial direction), is the direction vector of the torque (tangential direction). The dynamic load vector is updated every 5 ms time step to generate a dynamic load matrix. The vector sum of the equivalent centrifugal force load and the torque fluctuation value of the transmission chain is a spatial superposition operation considering the orthogonality of the acting directions; the dynamic load matrix is a set of dynamic load vectors recorded in time series and is used for setting the time-varying boundary conditions of the finite element model.

[0071] Input the dynamic load matrix into the benchmark model established based on the original design parameters of the tower barrel to generate a finite element model with time-varying boundary conditions.

[0072] Specifically, a benchmark finite element model is established based on the elastic modulus and Poisson's ratio in the tower barrel material certificate, and the mesh is divided using tetrahedral elements. After importing the dynamic load matrix, a fixed constraint is applied to the model base, and dynamic load parameters are applied to the top. The displacement field at each time step is updated through an implicit dynamics solver, and the formula is expressed as:

[0073] ;

[0074] In the formula, is the mass matrix, is the damping coefficient matrix, is the yield strength matrix, is the nodal displacement vector, is the dynamic load vector, is the damping coefficient weight, is the yield strength weight. The finite element model is a numerical simulation model based on the equations of continuum mechanics; the time-varying boundary conditions are the load and constraint settings that evolve over time.

[0075] Exemplarily, during the commissioning stage of a 6.2 MW unit in a certain offshore wind power project, the measured impeller speed is 11.2 rpm, and the generator torque is 2450 kN·m. The calculated centrifugal force load , and the torque fluctuation value of the transmission chain . A dynamic load matrix is generated through vector synthesis. After inputting it into the tower barrel finite element model, the predicted stress fluctuation range of the bolts at the 30-degree position of the third-layer flange is ±127 MPa. Verification sensors are installed at the corresponding positions on-site, and the measured stress fluctuation is ±139 MPa, with a relative error of 8.6%. During modeling, the stress fluctuation law under gust impact is successfully captured through dynamic load update. The dynamic load synthesis method accurately characterizes the superposition effect of the rotating machinery load and the transmission chain disturbance, and the dynamic update of the time-varying boundary conditions enables the finite element model to reflect the impact response of the instantaneous wind speed change on the structure. Compared with the static load model, this method improves the ability to capture the stress peak under extreme conditions and provides accurate input for subsequent life prediction.

[0076] Optionally, as shown in Figure 3 , the method of using the stress field gradient compensation algorithm to calculate the strain of unmonitored points and generate the stress distribution map of all tower bolts includes:

[0077] Extract the strain difference between adjacent monitoring points from the real-time strain data and calculate the curvature change characteristic quantity of the local area;

[0078] Specifically, select adjacent monitoring points A and B with a circumferential spacing of within the same tower layer, and read their measured strain values and , and calculate the first-order strain gradient . According to the tower radius R and the circumferential arc length , calculate the strain curvature characteristic quantity :

[0079] ;

[0080] This curvature change characteristic quantity characterizes the strain change rate per unit length and is used to describe the severity of the bending deformation in the local area. The adjacent monitoring points are two sensor nodes with a circumferential interval of 45 - 60 degrees in the same tower layer; the curvature change characteristic quantity is a geometric bending parameter deduced from the measured strain gradient and reflects the degree of local plastic deformation of the structure.

[0081] Compare the curvature change characteristic quantity with the theoretical gradient of the finite element model and calculate the residual stress deviation coefficient;

[0082] Specifically, derive the theoretical strain gradient at the corresponding position from the finite element model, and calculate the theoretical curvature characteristic quantity . Define the residual stress deviation coefficient as the ratio of the measured curvature to the theoretical curvature. For the residual stress deviation coefficient , there is:

[0083] ;

[0084] When the absolute value of the residual stress deviation coefficient exceeds a preset value such as 0.15, the correction process is triggered. The residual stress deviation coefficient is an evaluation index for the model prediction accuracy. A negative value indicates that the model overestimates the structural stiffness, and a positive value indicates that the actual stress concentration exceeds the expectation.

[0085] Iteratively adjust the elastic modulus of the unmonitored points according to the residual stress deviation coefficient until the error between the predicted stress calculated by the finite element model and the measured stress converges to the set interval, and output the predicted stress to construct the stress distribution map.

[0086] Specifically, calculate the predicted stress based on the linear elastic constitutive equation. For the predicted stress , there is:

[0087] ;

[0088] In the formula, is the elastic modulus of the unmonitored point, is the strain of the unmonitored point. The strain of the unmonitored point is obtained by multiplying the node displacement of the unmonitored point in the finite element model by the strain displacement matrix, and the strain displacement matrix is ​​obtained by the input data of the finite element model. For the modified elastic modulus ,have:

[0089] ;

[0090] In the formula, is the elastic modulus before correction, is the material sensitivity factor, for The sign function of . The elastic modulus of the grid cells in the currently unmonitored area is adjusted by spatial interpolation, and the step size of each correction is limited to ±5%. After 3 to 5 iterations, check whether the maximum relative error is less than 10%. If it meets the standard, the iteration is terminated and the corrected global stress distribution map is output; the error convergence is the state where the maximum deviation rate between the calculated predicted stress and the measured stress is lower than the threshold for three consecutive iterations.

[0091] This method reversely calibrates the material parameter deviation by comparing the measured strain gradient with the theoretical model, and eliminates the residual stress calculation error by using an iterative correction mechanism. Its technical effect is reflected in: achieving high-precision calculation of stress distribution in unmeasurable areas in complex assembly structures, overcoming the engineering prediction deviation caused by the solidification of material parameters in traditional finite element analysis, and providing a reliable dynamic stress data basis for bolt life assessment.

[0092] Optionally, the step of inputting the stress distribution map into a preset fatigue damage model, calculating the bolt life decay rate, and performing time series correlation analysis on the real-time strain data and the pitch yaw data in the operating condition data to locate the abnormal external load source, and obtaining the positioning result includes:

[0093] Obtain the SN curve and historical load spectrum data of the bolt material, establish the mapping relationship between strain amplitude and fatigue damage degree, and calculate the bolt life attenuation rate;

[0094] Specifically, Figure 4 As shown in the figure, the SN curve is constructed based on the material fatigue test data. Each data point corresponds to the number of cycles of bolt failure under a specific stress amplitude. The load time series in the historical operation of the unit is integrated, and the number of cycles of each stress amplitude is counted using the rain flow counting method. The total damage is calculated according to the Miner linear cumulative damage law. , the formula is:

[0095] ;

[0096] wherein, is the number of stress amplitudes, is the th actual cycle number corresponding to the stress amplitude, is the th failure cycle number corresponding to the stress amplitude in the S-N curve. The life attenuation rate is defined as the percentage of fatigue damage degree to the critical damage (D = 1). When D = 0.7, the attenuation rate is 70%. Among them, the S-N curve is the material fatigue characteristic curve obtained through experiments, with the abscissa representing the stress amplitude and the ordinate representing the corresponding cycle number; the historical load spectrum is the load time history recorded during the actual service of the bolt, and the bolt life attenuation curve is as shown in Figure 5 ; the fatigue damage degree is a fatigue damage quantification index calculated based on the cumulative damage theory.

[0097] Perform rainflow counting method processing on the stress distribution map to extract the equivalent alternating stress amplitude sequence;

[0098] Specifically, divide the time-stress data at each bolt position in the stress distribution map by a one-minute window, and use the four-peak detection method to identify the closed stress cycle. Extract the amplitude and mean value for each cycle, and correct the influence of the mean stress according to the Goodman formula to obtain the equivalent alternating stress amplitude sequence. The correction formula is:

[0099] ;

[0100] wherein, is the actual stress amplitude, is the cyclic mean stress, is the material ultimate strength. The rainflow counting method is a statistical method for decomposing irregular stress waveforms into independent cycles; the equivalent alternating stress amplitude is the equivalent constant amplitude stress value considering the influence of the mean stress.

[0101] Perform convolution operation on the equivalent alternating stress amplitude sequence and the pitch-yaw data to generate a fatigue cumulative factor coupled with the working conditions, locate the external load abnormal source, and obtain the location result.

[0102] Specifically, after adopting sliding average filtering for the pitch angle data, perform time-domain convolution with the equivalent alternating stress amplitude sequence, and calculate the convolution value , there is:

[0103] ;

[0104] wherein, is the equivalent alternating stress amplitude sequence, which is a function of the time variation, is the integration duration, is the function of the pitch angle change rate varying with time changing, is the current time, is the time variable in the integration. Set a matching threshold. When the convolution peak exceeds the threshold, determine the correlation between the pitch action and the stress fluctuation. Traverse all operating condition parameters and select the parameter corresponding to the maximum convolution value as the abnormal source. Convolution operation is a mathematical operation to measure the temporal correlation of two signals; the fatigue accumulation factor is an index characterizing the coupling strength between the external operating conditions and the stress response.

[0105] Exemplarily, in a certain 2MW unit, the above method is applied to the M24 bolt. Its S-N curve shows that the number of failure cycles corresponding to a stress amplitude of 150MPa is 10,000 cycles. The equivalent amplitude of 135MPa is accumulated 1500 times by the measured rain flow counting, and the fatigue damage degree D = 1500 / 10000 = 0.15, that is, the life is attenuated by 15%. The convolution analysis of the pitch data finds that there is a 0.8-second delay between the yaw angle jitter and the stress peak, and the correlation reaches 0.92, positioning the abnormal yaw bearing clearance. By quantifying the association between damage and operating conditions, accurately identify the external incentives leading to the accelerated aging of the bolt.

[0106] Optionally, the method further includes:

[0107] When the deviation between the measured stress and the predicted stress continuously exceeds the set number of times, extract the stress phase delay characteristics of each monitoring point;

[0108] Specifically, when the stress error of the same measuring point is detected to be greater than 15% continuously for 5 times, collect the time series data of the measured stress and the predicted stress. Conduct frequency domain phase analysis, and obtain the phase difference of the main frequency component through Fourier transform , there is:

[0109] ;

[0110] In the formula, is the measured stress, is the predicted stress, indicates that the measurement lags behind the theoretical response. The phase delay characteristic is a frequency domain parameter characterizing the lag of the structural dynamic response.

[0111] Adjust the weight ratio of the damping coefficient and the material yield strength in the finite element model according to the phase delay characteristic;

[0112] Specifically, establish the mapping relationship between the phase difference and the material parameters. For the damping coefficient weight and the yield strength weight , its update rule is that for the updated damping coefficient weight , there is:

[0113] ;

[0114] In the formula, is the fitting coefficient of the damping coefficient. For the updated yield strength weight , there is:

[0115] ;

[0116] In the formula, is the fitting coefficient of the yield strength. When , increase the damping coefficient weight to compensate for the dynamic response delay and reduce the yield strength weight to reflect the accumulation of plastic deformation.

[0117] Store the corrected model parameters in the historical database as the benchmark value for the next model initialization.

[0118] Specifically, establish a parameter version chain in the database to save the adjusted damping coefficient, yield strength, and the corresponding time series tags. When modeling next time, preferentially load the average value of the results of the last three corrections as the initial parameters. The historical database is a time-series storage system for storing the iterative parameters of the model.

[0119] This method compensates for the deviation caused by the deterioration of the environment and material properties by reversely correcting the finite element model parameters. It breaks through the limitation of the traditional model being static; realizes the causal traceability of the abnormal source; improves the engineering practicability of the remaining life prediction; and provides a direct decision-making basis for preventive maintenance.

[0120] Optionally, the generation basis of the hierarchical warning signal is:

[0121] Construct a multi-dimensional warning parameter set including stress fluctuation entropy value, temperature load sensitivity, and environmental corrosion rate;

[0122] Specifically, by collecting the strain time series data of the bolt monitoring points in real time, using the sliding window analysis method to extract the stress fluctuation sequence within 20 minutes, and using the information entropy theory to calculate its statistical dispersion. At the same time, obtain the bolt surface temperature value and its change rate through a temperature sensor, establish a temperature stress response curve in combination with the material thermal expansion coefficient, and calculate the stress offset per unit temperature change as the temperature load sensitivity. The environmental corrosion rate calculates the corrosion equivalent value according to the salt spray sensor data and the material corrosion resistance coefficient, and integrates the above three parameters to form a multi-dimensional vector.

[0123] Among them, the stress fluctuation entropy value is an index of the structural stress disorder degree based on the frequency-domain energy distribution, reflecting the randomness of load impact; the temperature load sensitivity is the offset response coefficient of the bolt stress caused by temperature fluctuation; the environmental corrosion rate is the corrosion thickness per unit time of the metal material under specific temperature and humidity conditions; the multi-dimensional early warning parameter set is a comprehensive evaluation vector formed by the fusion of multiple physical quantities, characterizing the degradation state of the composite material.

[0124] When a single parameter exceeds the first-level threshold, a status prompt is triggered. When the combined effect of at least two parameters exceeds the second-level threshold, an emergency shutdown instruction is triggered.

[0125] Specifically, the first-level threshold is set as the 95% quantile value of the historical data of each parameter, and the second-level threshold is the composite critical value obtained through support vector machine training. For example, when the stress fluctuation entropy value H > 4.2 or the temperature load sensitivity > 0.15 MPa / °C, a yellow early warning is issued; if the entropy value is exceeded and the environmental corrosion rate > 0.1 mm / month at the same time, a shutdown signal is triggered. The combined effect adopts a weighted superposition model:

[0126] ;

[0127] In the formula, is the weight of the stress fluctuation entropy value, is the weight of the temperature load sensitivity, is the weight of the environmental corrosion rate. When it is determined to be in an emergency state. The first-level threshold is the early warning trigger condition for a single parameter, defined based on statistical methods; the second-level threshold is the failure boundary under the non-linear coupling effect of multiple physical quantities, and the decision surface is divided through machine learning; the combined effect is the weighted evaluation result of the failure rate under the interaction of multiple parameters.

[0128] Exemplarily, during the detection period, the environmental corrosion rate rises to 0.13 mm / month (exceeding the first-level threshold of 0.1 mm / month) due to a typhoon passing by, triggering a yellow prompt. At the same time, the stress fluctuation entropy value H = 4.8, and the combined weight value S = 0.92, triggering an emergency state shutdown instruction. The on-site inspection by the operation and maintenance team found that 40% of the bolt surface was rusted and there were stress corrosion cracks. The single corrosion rate exceeding the standard gives an early warning, and the over-limit of the composite parameters accurately identifies the potential fracture risk, avoiding the damage of the gearbox caused by the overall failure of the flange surface. The collaborative determination of multi-dimensional parameters enhances the reliability of early warning and effectively distinguishes the boundary between normal abrasion and malignant corrosion.

[0129] Optionally, the calculation of the stress fluctuation entropy value includes:

[0130] Performing Fourier transform on the strain data within a specified time window, and extracting the energy distribution characteristics of a preset frequency band;

[0131] Specifically, strain signals are collected with a 60 - second time window, and the frequency accuracy after performing the Fast Fourier Transform (FFT) is 0.1 Hz. When calculating the frequency - domain energy distribution, three characteristic segments are divided; the low - frequency segment is 0.5 - 5 Hz, corresponding to the overall swing frequency of the tower barrel; the middle - frequency segment is 5 - 20 Hz, corresponding to the torsional vibration frequency of the drive chain; the high - frequency segment is 20 - 50 Hz, corresponding to the abnormal noise of local bolt loosening. After the total energy is normalized, the proportion of low - frequency energy The formula is:

[0132] ;

[0133] In the formula, is the integral value of the low - frequency segment energy, is the total energy of the full frequency band. The Fourier transform is an analysis method for converting time - domain signals into frequency - domain energy spectra; the energy distribution characteristic is the proportion value of the energy in a specific frequency band to the entire frequency band, which is used to identify vibration modes.

[0134] According to the energy distribution characteristic, the stress fluctuation entropy value is calculated. When the proportion of low - frequency segment energy exceeds the preset proportion, it is determined as the resonance risk mode.

[0135] Specifically, the stress fluctuation entropy value The calculation method is:

[0136] ;

[0137] Among them, is the normalized energy proportion of the low - frequency segment, is the normalized energy proportion of the middle - frequency segment, is the normalized energy proportion of the high - frequency segment, is the low - frequency segment energy, is the middle - frequency segment energy, is the high - frequency segment energy, is the total energy. The preset proportion can be 70%. If and , it is determined as the low - frequency resonance - dominated mode, and the characteristic is that the energy of a single frequency is highly concentrated. The resonance risk mode is the resonance state caused by the narrow - band excitation of the structure, resulting in a doubling of the stress amplitude.

[0138] Exemplarily, on - site monitoring shows that the proportion of low - frequency energy of the bolt stress reaches 83%, and the stress fluctuation entropy value . After triggering the resonance warning, inspection finds that the impeller mass block has fallen off, resulting in speed fluctuations and inducing a 2.4 - Hz swing of the tower barrel. After maintenance, the proportion of low - frequency energy drops to 52%, returning to the normal distribution. The hidden resonance source is identified through the energy focusing characteristic of the frequency band, avoiding bolt fatigue fracture caused by continuous resonance. The entropy value quantization significantly improves the sensitivity to capture abnormal vibration modes.

[0139] Optionally, the method further includes:

[0140] When the resonance risk mode is recognized, synchronously collect the axial acceleration data of the vibration sensors of the unit.

[0141] Specifically, after triggering the resonance warning, synchronously obtain the axial vibration acceleration signal of the gearbox at a sampling rate of 2000 Hz. Retain the effective frequency bandwidth of 0 - 1000 Hz through anti-aliasing filtering, and use a third-order Butterworth filter to eliminate high-frequency noise. The data alignment method uses the IEEE 1588 Precision Time Protocol (PTP), and the time scale error between the strain and acceleration signals is less than 1 ms. The axial acceleration data is a quantization index of the vibration intensity along the axis of the transmission chain; synchronous acquisition is a time-axis alignment technology for multi-sensor data to ensure event correlation analysis.

[0142] Perform amplitude-frequency correlation analysis on the acceleration data and the low-frequency band energy. If the correlation coefficient is greater than the preset correlation value, add the commissioning task of the vibration suppression device to the maintenance strategy.

[0143] Specifically, use the cross-spectral density method to calculate the acceleration amplitude and the low-frequency band energy of the stress of the correlation coefficient :

[0144] ;

[0145] In the formula, is the covariance function, is the standard deviation of the acceleration amplitude, is the standard deviation of the low-frequency band energy of the stress. When , it is determined that the two are strongly correlated, and the task of leveling the dynamic vibration absorber or replacing the damper needs to be added to the maintenance list. Amplitude-frequency correlation analysis is a mathematical method for evaluating the matching degree of the frequency components of vibration and stress signals; the vibration suppression device is an additional device used to absorb or dissipate mechanical vibration energy.

[0146] Exemplarily, after the resonance warning is triggered, the synchronous vibration data shows a strong acceleration component of 3.7 Hz, and the correlation coefficient with the low-frequency energy of the strain signal . After the maintenance team installed a tuned mass damper (TMD), the vibration amplitude decreased by 56%, and the stress fluctuation entropy value returned to the normal range. Accurately locate the vibration-stress coupling path through cross-sensor data analysis, guide the deployment of targeted vibration reduction measures, and solve the problem of bolt array failure caused by combined excitation.

[0147] This method constructs a dynamic early warning system through multi-dimensional parameter coupling analysis (stress fluctuation, temperature sensitivity, environmental corrosion), combines frequency domain energy entropy evaluation with vibration signal collaborative verification, and realizes refined diagnosis of bolt health status. The resonance risk detection module identifies structural resonance through frequency band energy focusing, and accurately locates the vibration source through synchronous vibration correlation analysis. Its beneficial effects include: breaking through the misjudgment limitation of single parameter monitoring, distinguishing normal strain from malignant damage through multi-physical field data fusion; real-time matching of resonance mode and vibration source improves fault tracing efficiency; early warning strategy is hierarchical and progressive, which not only avoids frequent false alarms but also ensures timely containment of failure risks. In the exemplary verification, after applying this method to a certain intertidal zone wind power project, hydrogen embrittlement of the bolt group was discovered through a combined early warning of corrosion rate and entropy value before the typhoon season, avoiding the accident of the whole machine tower collapse. Through entropy monitoring and damper adjustment, the average service life of the bolts is extended.

[0148] Based on the same inventive concept, Figure 6 As shown, the present invention also provides a wind power bolt status detection system, the system comprising:

[0149] Wireless strain sensor groups are deployed at preset circumferential positions on each floor of the tower to obtain real-time strain data at each monitoring point;

[0150] The operating data acquisition module is used to obtain the operating data of the unit in real time from the wind farm telecontrol communication system, including wind speed, pitch and pitch data, generator torque and impeller speed;

[0151] A finite element model building module, used to dynamically generate a finite element model of the tower structure according to the operating condition data, and to modify the finite element model using the real-time strain data;

[0152] A stress distribution generation module, for combining the finite element model and the real-time strain data, using a stress field gradient compensation algorithm to calculate the strain of unmonitored points, and generating a stress distribution map of the entire tower bolts;

[0153] An abnormality analysis and positioning module is used to input the stress distribution map into a preset fatigue damage model, calculate the bolt life decay rate and perform time-series correlation analysis on the real-time strain data and the pitch yaw data in the operating condition data, locate the abnormal source of the external load, and obtain a positioning result;

[0154] The early warning decision module is used to generate a graded early warning signal and output a maintenance strategy according to the positioning result when it is detected that the life decay rate exceeds a preset threshold.

[0155] It should be noted that the electrical connections between the above-mentioned units do not necessarily represent direct connections of the circuits. Indirect connection methods, as long as they can achieve the purpose of the present invention, are applicable to the embodiments of the present invention. The above are only exemplary embodiments of the present invention and should not be used to limit the scope of the present invention.

[0156] That is, any equivalent changes and modifications made in accordance with the teachings of the present invention still fall within the scope covered by the present invention. Those skilled in the art will easily think of other implementation schemes of the present invention after considering the specification and the disclosure of the practical truth. This application aims to cover any variations, uses or adaptation changes of the present invention, and these variations, uses or adaptation changes follow the general principles of the present invention and include the common general knowledge or conventional technical means in the technical field not recorded in the present invention.

Claims

1. A method for detecting the state of a wind power bolt, characterized in that, The method comprises: Monitoring points are set at preset circumferential positions on each floor of the tower to obtain real-time strain data of each monitoring point; Obtain the unit's operating data in real time from the wind farm telecontrol communication system, including wind speed, pitch data, generator torque and rotor speed; Dynamically generate a finite element model of the tower structure according to the operating condition data, and modify the finite element model using the real-time strain data; Combining the finite element model and the real-time strain data, a stress field gradient compensation algorithm is used to calculate the strain of unmonitored points to generate a stress distribution map of the entire tower bolts; Input the stress distribution map into a preset fatigue damage model, calculate the bolt life decay rate, and perform time series correlation analysis on the real-time strain data and the pitch yaw data in the operating condition data to locate the abnormal source of the external load and obtain a positioning result; When it is detected that the life decay rate exceeds a preset threshold, a graded warning signal is generated and a maintenance strategy is output according to the positioning result.

2. The detection method for the state of a wind power bolt according to claim 1, characterized in that, The monitoring points are set at preset circumferential positions on each floor of the tower, including: A first subset of sensors is symmetrically arranged on both sides of the main wind direction impact surface of the tower, and the circumferential spacing between adjacent sensors is set according to a preset angle; The second subset of sensors is arranged in the high bending moment area at the tower flange connection, and the three-dimensional strain components are collected in an orthogonal arrangement.

3. The detection method for the state of a wind power bolt according to claim 1, characterized in that, The dynamically generating a finite element model of the tower structure according to the operating condition data comprises: Generate an equivalent centrifugal force load according to the impeller speed, and calculate a transmission chain torque fluctuation value according to the generator torque; Performing vector synthesis of the equivalent centrifugal force load and the transmission chain torque fluctuation value to generate a dynamic load matrix that varies with time; The dynamic load matrix is ​​input into a benchmark model established based on the original design parameters of the tower to generate a finite element model including time-varying boundary conditions.

4. The detection method for the state of a wind power bolt according to claim 1, characterized in that, The method of using the stress field gradient compensation algorithm to calculate the strain of the unmonitored points and generate the stress distribution map of the whole tower bolts includes: Extracting strain differences between adjacent monitoring points from the real-time strain data and calculating curvature change characteristic quantities of the local area; Comparing the curvature change characteristic quantity with the theoretical gradient of the finite element model to calculate the residual stress deviation coefficient; The elastic modulus of the unmonitored point is iteratively adjusted according to the residual stress deviation coefficient until the error between the predicted stress calculated by the finite element model and the measured stress converges to a set interval, and the predicted stress is output to construct a stress distribution map.

5. The detection method for the state of a wind power bolt according to claim 1, characterized in that The stress distribution map is input into a preset fatigue damage model, the bolt life decay rate is calculated, and the real-time strain data is analyzed in time series correlation with the pitch yaw data in the operating condition data to locate the abnormal external load source, and the positioning result is obtained, including: Obtain the SN curve and historical load spectrum data of the bolt material, establish the mapping relationship between strain amplitude and fatigue damage degree, and calculate the bolt life attenuation rate; Processing the stress distribution spectrum by rain flow counting method to extract equivalent alternating stress amplitude sequence; The equivalent alternating stress amplitude sequence is convolved with the pitch yaw data to generate a fatigue accumulation factor coupled with working conditions, locate the abnormal source of the external load, and obtain a positioning result.

6. The detection method for the state of a wind power bolt according to claim 1, characterized in that The method further comprises: When the deviation between the measured stress and the predicted stress continues to exceed the set number of times, the stress phase delay characteristics of each monitoring point are extracted; Adjusting the weight ratio of the damping coefficient and the material yield strength in the finite element model according to the phase delay characteristics; The modified model parameters are stored in the historical database as the reference values ​​for the next model initialization.

7. The detection method for the state of a wind power bolt according to claim 1, characterized in that The generation basis of the graded warning signal is: Construct a multidimensional early warning parameter set including stress fluctuation entropy value, temperature load sensitivity and environmental corrosion rate; When a single parameter exceeds the first-level threshold, a status prompt is triggered, and when the combined effect of at least two parameters exceeds the second-level threshold, an emergency shutdown command is triggered.

8. The detection method for the state of a wind power bolt according to claim 7, wherein The calculation of the stress fluctuation entropy value includes: Perform Fourier transform on the strain data within the specified time window to extract the energy distribution characteristics of the preset frequency band; The stress fluctuation entropy value is calculated according to the energy distribution characteristics, and when the proportion of low-frequency energy exceeds a preset proportion, it is determined to be a resonance risk mode.

9. The detection method of the state of a wind power bolt according to claim 8, characterized in that, The method further comprises: When a resonance risk mode is identified, the axial acceleration data of the unit vibration sensor is collected synchronously; An amplitude-frequency correlation analysis is performed on the acceleration data and the low-frequency band energy, and if the correlation coefficient is greater than a preset correlation value, a debugging task of the vibration suppression device is added to the maintenance strategy.

10. A detection system for the state of a wind power bolt, which is applied to a method for detecting the state of a wind power bolt according to any one of claims 1-9, characterized in that, The system comprises: Wireless strain sensor groups are deployed at preset circumferential positions on each floor of the tower to obtain real-time strain data at each monitoring point; The operating data acquisition module is used to obtain the operating data of the unit in real time from the wind farm telecontrol communication system, including wind speed, pitch and pitch data, generator torque and impeller speed; A finite element model building module, used to dynamically generate a finite element model of the tower structure according to the operating condition data, and to modify the finite element model using the real-time strain data; A stress distribution generation module, for combining the finite element model and the real-time strain data, using a stress field gradient compensation algorithm to calculate the strain of unmonitored points, and generating a stress distribution map of the entire tower bolts; An abnormality analysis and positioning module is used to input the stress distribution map into a preset fatigue damage model, calculate the bolt life decay rate and perform time series correlation analysis on the real-time strain data and the pitch yaw data in the operating condition data, locate the abnormal source of the external load, and obtain a positioning result; The early warning decision module is used to generate a graded early warning signal and output a maintenance strategy according to the positioning result when it is detected that the life decay rate exceeds a preset threshold.

Images