Wind power bolt predictive maintenance method based on piezoelectric sensing

Abstract

The invention relates to the technical field of monitoring or testing of wind power engines, in particular to a wind power bolt predictive maintenance method based on piezoelectric sensing, which comprises the following steps: installing a piezoelectric sensing unit on the end face of a bolt, and collecting ultrasonic propagation time to establish a reference value; constructing a multi-dimensional data set based on the reference value and the risk level of the bolt connection part, and integrating sound time variation, wind speed, vibration spectrum and environment temperature data; fusing multi-source information through data processing to generate a bolt health index; comparing the health index with a preset threshold value, and matching a differentiated maintenance strategy according to a result, including prolonging a maintenance period, arranging recheck or executing retightening. High-precision measurement is achieved through piezoelectric sensing, the early warning capacity is improved in combination with a data driving model, unnecessary maintenance operation is reduced, the operation and maintenance cost is reduced, and the service life of the bolt is prolonged.

Description

Technical Field

[0001] This invention relates to the field of wind turbine monitoring or testing technology, and more particularly to a predictive maintenance method for wind turbine bolts based on piezoelectric sensing. Background Technology

[0002] Existing torque measurement technology has the following technical drawbacks: insufficient measurement accuracy makes it difficult to identify early loosening of high-strength wind turbine bolts, leading to over-maintenance and reduced service life; because the torque method relies on manual operation and mechanical principles, its error range is large and it cannot accurately reflect the state of bolt axial force decay. In the operation and maintenance of wind turbine units, such as during the inspection of tower flange connection bolts, maintenance personnel often perform uniform retightening on all bolts based on inaccurate measurement results, including those bolts whose axial force is still within the normal range. This kind of blind maintenance not only increases unnecessary workload, but also introduces additional stress due to repeated application of torque, accelerating bolt material fatigue, ultimately reducing the overall connection reliability and shortening service life. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a predictive maintenance method for wind turbine bolts based on piezoelectric sensing. This invention solves the technical problem that the insufficient accuracy of existing torque methods makes it difficult to identify early loosening of high-strength wind turbine bolts, leading to over-maintenance and reduced lifespan.

[0004] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows:

[0005] The predictive maintenance method for wind turbine bolts based on piezoelectric sensing provided by this invention includes: Step 1: Collect the propagation time of the ultrasonic wave in the bolt to establish a reference value; Step 2: Based on the baseline values ​​and the preset risk levels of the bolted connections, construct a multidimensional dataset including data on acoustic time variation, wind speed, vibration spectrum, and ambient temperature. Step 3: Generate bolt health indicators through data processing based on the multidimensional dataset; Step 4: Compare the health indicators with preset thresholds, and match differentiated maintenance strategies based on the comparison results. Differentiated maintenance strategies include extending the maintenance cycle, arranging re-examination, or performing re-tightening.

[0006] Furthermore, the present invention provides a method for maintenance and repair of wind turbine bolts based on piezoelectric sensing, wherein step 1 includes: A piezoelectric sensing unit is installed on the end face of the bolt. Ultrasonic waves are excited by the piezoelectric sensing unit, and the propagation time of the ultrasonic waves in the bolt is measured to establish a reference value. The piezoelectric sensing unit is a piezoelectric ceramic sheet with a fully flanged electrode structure. The central circular area serves as the positive electrode, and the edge annular area is connected to the back surface as the negative electrode. The piezoelectric ceramic sheet is connected using a magnetically attached probe. The magnetically attached probe has a built-in ring magnet and a spring pin. The spring pin contacts the positive electrode, and the outer ring metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time is collected as a reference value.

[0007] Furthermore, in the wind turbine bolt inspection-based maintenance method of the present invention based on piezoelectric sensing, step 2 includes: Bolts are classified according to the risk level of the bolted connection, and random inspections are performed based on the classification results. Acoustic time data is collected through random inspections, and the change in acoustic time is calculated in combination with the benchmark value. Wind speed data, vibration data and ambient temperature data are integrated from the preset monitoring system to construct a multidimensional dataset. Bolt classification includes defining leaf root bolts as Class A high-risk and tower bolts as Class B or Class C; The sampling ratio is set according to the bolt grade. The sampling ratio of Class A bolts is higher than that of Class B bolts, and the sampling ratio of Class B bolts is higher than that of Class C bolts. Wind speed, vibration spectrum, and ambient temperature data are acquired from the monitoring system, and characteristic frequency components of the vibration power spectrum are extracted through spectrum analysis. The sound time data is bound and stored with timestamps and ambient temperature data to form a time series dataset.

[0008] Furthermore, in the wind turbine bolt inspection-based maintenance method of the present invention based on piezoelectric sensing, step 3 includes: Based on the acoustic time variation and the multidimensional dataset, the axial force loss rate is determined, and based on the vibration data and ambient temperature data in the multidimensional dataset, vibration characteristics and temperature compensation parameters are generated respectively. Bolt health indicators are generated through data processing. The axial force loss rate is calculated by linear mapping of the acoustic time variation, and the mapping coefficient is calibrated according to the bolt material properties. The vibration power spectrum is extracted from the multidimensional dataset and compared with the reference power to obtain the relative vibration effect. The temperature drift compensation coefficient is calculated based on ambient temperature data, using an exponential function model. The health indicators are generated by weighted summation of axial force loss rate, relative vibration effect, and temperature drift compensation coefficient.

[0009] Furthermore, the present invention provides a method for maintenance and repair of wind turbine bolts based on piezoelectric sensing, wherein the weighted summation coefficient calibration includes: Collect historical operating data of wind turbines, including axial force attenuation data, vibration load data, and temperature change data. Establish a correlation model between historical data and bolt health status; The weighting coefficients are determined based on the correlation model, so that the weighting coefficients reflect the coupled effects of axial force attenuation, vibration load and temperature change.

[0010] Furthermore, in the wind turbine bolt inspection-based maintenance method of the present invention based on piezoelectric sensing, step 4 includes: The preset thresholds include a first threshold and a second threshold. The first threshold corresponds to a low-risk state, and the second threshold corresponds to a high-risk state. When the health indicators are less than or equal to the first threshold, the bolts are marked as Grade I, and the maintenance cycle is extended to more than 12 months. When the health indicators are greater than the first threshold and less than or equal to the second threshold, the bolt is marked as Grade II, triggering the re-inspection process and re-measuring the acoustic time data within 3 months. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated to perform re-tightening.

[0011] Furthermore, in the wind turbine bolt inspection-based maintenance method based on piezoelectric sensing of the present invention, step 4 further includes: The distribution of health indicators of randomly inspected bolts was statistically analyzed, and the proportion of bolts with health indicators greater than the second threshold was calculated. When the proportion of bolts with health indicators exceeding the second threshold exceeds the preset limit, a full-circle bolt detection is initiated.

[0012] Furthermore, in the wind turbine bolt inspection-based maintenance method based on piezoelectric sensing of the present invention, step 2 further includes: Different data acquisition strategies are set according to the bolt grade: high-frequency acquisition mode is used for Class A bolts, medium-frequency acquisition mode is used for Class B bolts, and low-frequency acquisition mode is used for Class C bolts. During the data acquisition process, the portable measuring device automatically adjusts the ultrasonic emission power according to the current ambient temperature; The time-series dataset formed from the acoustic time data is used for temperature drift compensation and vibration analysis in step 3.

[0013] Furthermore, in the wind turbine bolt inspection-based maintenance method based on piezoelectric sensing of the present invention, step 3 further includes: The calculation of the temperature drift compensation coefficient incorporates the thermal expansion coefficient and the temperature coefficient of elastic modulus of the bolt material, and simultaneously compensates for the thermal expansion effect and the elastic change of the material through a coupled model. The vibration power spectrum was analyzed, and low-frequency vibration components and high-frequency vibration components were extracted respectively. The low-frequency vibration components were used to assess the risk of structural resonance, and the high-frequency vibration components were used to assess the micro-slippage of bolts. During weighted calculation, the weighting coefficients are dynamically adjusted according to the bolt grade. For Class A bolts, the vibration component weight is increased; for Class B bolts, the temperature compensation weight is increased; and for Class C bolts, the axial force loss rate weight is increased.

[0014] Furthermore, in the wind turbine bolt inspection-based maintenance method based on piezoelectric sensing of the present invention, step 4 further includes: After making the maintenance decision in step 4, record the actual maintenance effect data, which includes the change in the bolt health index after re-tightening. The maintenance effect data is compared and analyzed with the predicted health index to calculate the prediction accuracy. The parameters of the weighting coefficients are dynamically adjusted based on the prediction accuracy. When the prediction accuracy falls below a preset threshold, the weight coefficients are recalibrated to obtain updated weight coefficients, which are then used for the next round of health index calculation.

[0015] Beneficial effects of this invention; This invention achieves high-precision ultrasonic propagation time measurement using piezoelectric sensing technology, establishes a benchmark value, and constructs a multidimensional dataset including acoustic time variation, wind speed, vibration spectrum, and ambient temperature data. Based on the multidimensional dataset, bolt health indicators are generated. These health indicators are weighted and fused with axial force loss rate, relative vibration influence, and temperature compensation parameters. The weighting coefficients are calibrated based on historical data to reflect the coupled effects of multiple factors. After comparing the health indicators with preset thresholds, differentiated maintenance strategies are matched, including extending the maintenance cycle, arranging re-inspection, or performing re-tightening. This solves the problem of difficulty in early loosening identification caused by insufficient accuracy of existing torque methods. The data-driven model improves the accuracy of early warning, avoids excessive intervention caused by uniform maintenance, reduces unnecessary re-tightening operations, lowers maintenance costs, and extends the service life of bolts. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the predictive maintenance method for wind turbine bolts based on piezoelectric sensing provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0019] Please see Figure 1 The present invention provides a predictive maintenance method for wind turbine bolts based on piezoelectric sensing, comprising: Step 1: Collect the propagation time of the ultrasonic wave in the bolt to establish a reference value; Step 2: Based on the baseline values ​​and the preset risk levels of the bolted connections, construct a multidimensional dataset including data on acoustic time variation, wind speed, vibration spectrum, and ambient temperature. Step 3: Generate bolt health indicators through data processing based on the multidimensional dataset; Step 4: Compare the health indicators with preset thresholds, and match differentiated maintenance strategies based on the comparison results. Differentiated maintenance strategies include extending the maintenance cycle, arranging re-examination, or performing re-tightening.

[0020] The predictive maintenance method for wind turbine bolts based on piezoelectric sensing provided by this invention achieves accurate assessment and predictive maintenance of bolt health status through sequential steps. The method begins with the acquisition of ultrasonic reference values ​​and gradually progresses to multi-dimensional data integration, health index calculation and maintenance strategy matching. Each step is closely linked to form a complete technology chain.

[0021] Step 1 involves acquiring the propagation time of ultrasonic waves within the bolt to establish a baseline value. Specifically, this includes installing a piezoelectric sensing unit on the bolt end face. The piezoelectric sensing unit uses a piezoelectric ceramic sheet structure with a fully flanged electrode design. The central circular area serves as the positive electrode, while the outer annular area, connected to the back surface, serves as the negative electrode. A magnetic probe is used to connect to the piezoelectric ceramic sheet. The magnetic probe incorporates a ring magnet and a spring-loaded pin. The spring-loaded pin contacts the positive electrode, and the outer annular metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time is accurately measured as the baseline value. This step utilizes the piezoelectric effect to excite ultrasonic waves, and the propagation time measurement provides an initial reference value for subsequent comparisons. The electrode design of the piezoelectric ceramic sheet ensures stable signal transmission, and the magnetic probe structure adapts to the confined spaces and rapid installation requirements of wind power sites, achieving efficient and reliable testing operations.

[0022] Step 2 constructs a multidimensional dataset based on benchmark values ​​and preset risk levels of bolted connections. Specifically, this includes classifying bolts according to the risk level of the bolted connection; for example, leaf root bolts are defined as Class A high-risk bolts, and tower bolts as Class B or C risk bolts. Sampling inspections are performed based on the classification results, with the sampling ratio set according to the bolt level: a higher sampling ratio for Class A bolts, a medium sampling ratio for Class B bolts, and a lower sampling ratio for Class C bolts. Acoustic time data is collected through sampling, and the acoustic time variation is calculated in conjunction with the benchmark values. Simultaneously, wind speed data, vibration data, and ambient temperature data are integrated from a preset monitoring system. Vibration data undergoes spectral analysis to extract characteristic frequency components, forming a vibration power spectrum. Acoustic time data is bound to timestamps and ambient temperature data for storage, forming a time-series dataset. This step optimizes the allocation of detection resources through bolt classification, focuses the sampling strategy on high-risk areas, and integrates acoustic time variation, wind speed, vibration spectrum, and ambient temperature data into the multidimensional dataset. Acoustic time variation reflects the trend of bolt axial force changes, wind speed and vibration data characterize external dynamic loads, ambient temperature data provides a basis for temperature compensation, and the time-series dataset supports historical trend analysis and anomaly detection.

[0023] In the specific implementation of step 2, the sampling inspection mechanism is further refined: bolts are divided into three categories—A, B, and C—based on the criticality and risk level of the connection location. Category A bolts are defined as high-risk bolts, such as blade root bolts, as these locations bear high dynamic loads and failure would have serious consequences. Category B bolts are defined as medium-risk bolts, such as tower flange connection bolts. Category C bolts are defined as low-risk bolts, such as secondary connection bolts inside the engine room. The sampling inspection ratio is dynamically set based on the bolt level: 20% of Category A bolts are sampled annually, 10% of Category B bolts, and 5% of Category C bolts. During sampling inspection, Category A bolts are prioritized for high-frequency testing. All Category A bolts are 100% equipped with miniature piezoelectric ceramic sensor units, but actual measurements are only performed proportionally to balance testing efficiency and resource allocation. The sampling inspection operation uses a magnetic probe to quickly connect to the piezoelectric sensor unit, with a single bolt testing time not exceeding 5 seconds. Acoustic time data is collected and compared with a benchmark value to calculate the acoustic time change (ΔT / T0). Meanwhile, the sampling data is linked to timestamps and ambient temperature, and real-time wind speed, vibration spectrum, and other external load data are integrated from the SCADA system to form a time-series multidimensional dataset for subsequent health indicator calculations. The sampling ratio and frequency can be dynamically adjusted based on the actual operating data of the wind farm. For example, the sampling frequency of Class A bolts can be increased during high-wind-speed seasons to improve early warning response capabilities.

[0024] Furthermore, during the sampling inspection, the portable measuring device automatically adjusts the ultrasonic emission power according to the ambient temperature to ensure measurement stability. The sampling data undergoes lightweight processing through an edge computing layer, extracting feature values ​​before being uploaded to the platform layer. To ensure the representativeness of the sampling, the system employs a random sampling algorithm to avoid detection bias and periodically analyzes the distribution of health indicators for the sampled bolts. When the proportion of bolts with health indicators exceeding a threshold reaches a preset limit (e.g., 50%), a full-circle bolt inspection is automatically triggered to comprehensively assess the group risk. This tiered sampling inspection mechanism effectively optimizes inspection resources, realizing a shift from "periodic full inspection" to "on-demand sampling," improving operational economy and safety.

[0025] Step 3 generates bolt health indicators based on a multidimensional dataset through data processing. This includes determining the axial force loss rate, generating vibration characteristics, and calculating temperature compensation parameters. The axial force loss rate is calculated through a linear mapping of the acoustic time-varying quantities, with the mapping coefficient calibrated according to the bolt material properties. The vibration power spectrum is extracted from the multidimensional dataset and compared with a benchmark power to determine the relative vibration impact. The temperature drift compensation coefficient is calculated based on ambient temperature data using an exponential function model. The health indicators are generated by a weighted sum of the axial force loss rate, relative vibration impact, and temperature drift compensation coefficient. The weighting coefficients are calibrated using historical wind turbine operating data, including axial force attenuation data, vibration load data, and temperature change data. A correlation model between historical data and bolt health status is established, allowing the weighting coefficients to reflect the coupled effects of axial force attenuation, vibration load, and temperature change. This step integrates multi-source information. The axial force loss rate directly quantifies the degree of bolt preload attenuation; vibration characteristics assess structural resonance risk and microslippage; temperature compensation eliminates environmental drift effects; the weighted summation model integrates the contributions of various factors; and the health indicators serve as a quantitative output to provide a basis for maintenance decisions.

[0026] Step 3 generates bolt health indicators based on the multidimensional dataset through data processing, specifically including: determining the axial force loss rate based on the acoustic time variation and the multidimensional dataset, and generating vibration characteristics and temperature compensation parameters based on the vibration data and ambient temperature data in the multidimensional dataset, and generating bolt health indicators through data processing. The axial force loss rate is calculated by linear mapping of the acoustic time variation. The specific formula is as follows: ; Where K is the mapping coefficient calibrated according to the bolt material properties, ΔT is the acoustic time variation, and T0 is the reference value; The vibration power spectrum is extracted from the multidimensional dataset and compared with the reference power to obtain the relative vibration influence, which is expressed as Pvib / Pref, where Pvib is the vibration power and Pref is the reference power. The temperature drift compensation coefficient is calculated based on ambient temperature data, using an exponential function model, expressed as follows: ; Where T is the current ambient temperature, T0 is the reference temperature, and γ and δ are calibration parameters; The health indicator (denoted as N) is generated by a weighted sum of the axial force loss rate, relative vibration effect, and temperature drift compensation coefficient. The specific formula is as follows: ; α, β, γ, and δ are weighting coefficients, calibrated using historical operating data of the wind turbine to reflect the coupled effects of axial force attenuation, vibration load, and temperature changes.

[0027] Step 4 compares the health indicators with preset thresholds and matches differentiated maintenance strategies based on the comparison results. Specifically, this includes setting a first threshold and a second threshold, with the first threshold corresponding to a low-risk state and the second threshold corresponding to a high-risk state. When the health indicator is less than or equal to the first threshold, the bolt is marked as Grade I, and the maintenance cycle is extended to more than 12 months. When the health indicator is greater than the first threshold but less than or equal to the second threshold, the bolt is marked as Grade II, triggering a re-inspection process and re-measuring the acoustic time data within 3 months. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated for re-tightening. Furthermore, the distribution of health indicators for randomly selected bolts is statistically analyzed, and the proportion of bolts with health indicators greater than the second threshold is calculated. When this proportion exceeds a preset limit, a full-circle bolt inspection is initiated. This step achieves automated risk classification through threshold comparison. The differentiated maintenance strategy adopts extended cycles, re-inspections, or re-tightening measures for different risk levels, avoiding excessive intervention caused by uniform maintenance. The statistical monitoring mechanism improves the ability to respond to group risks and ensures the overall reliability of the connection.

[0028] After execution, the method of this invention records actual maintenance effect data, such as the change in the health index of bolts after re-tightening. The maintenance effect data is compared and analyzed with the predicted health index to calculate the prediction accuracy. The weighting coefficient calibration parameters are dynamically adjusted based on the prediction accuracy. When the prediction accuracy falls below a preset threshold, the weighting coefficients are recalibrated, and the updated weighting coefficients are used for the next round of health index calculation, forming a closed-loop optimization mechanism. Non-destructive detection is achieved through piezoelectric sensing technology, and data-driven models improve early warning accuracy. Each step of this invention progresses sequentially: benchmark establishment lays the foundation for data construction, multidimensional datasets support the generation of health indicators, indicator comparison directly guides maintenance actions, and the feedback mechanism continuously optimizes model parameters, ensuring the adaptability and long-term effectiveness of the method.

[0029] This invention includes mounting a piezoelectric sensing unit on the bolt end face. The piezoelectric sensing unit is a piezoelectric ceramic sheet with a fully flanged electrode structure. The central circular area serves as the positive electrode, and the outer annular area, which is fully connected to the back surface, serves as the negative electrode. A magnetic probe is used to connect the piezoelectric ceramic sheet. The magnetic probe contains a built-in ring magnet and a spring-loaded pin. The spring-loaded pin contacts the positive electrode, and the outer annular metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time is collected as a reference value. This implementation uses the piezoelectric effect to excite ultrasonic waves, and the propagation time measurement provides the basis for establishing a reference value. The electrode structure design supports a stable electrical connection, and the magnetic probe facilitates quick installation and removal, adapting to the wind power site environment.

[0030] This invention classifies bolts based on the risk level of the bolted connection location. For example, leaf root bolts are defined as Class A (high risk), and tower bolts are defined as Class B or C. Sampling inspections are performed based on the classification results, with the sampling ratio set according to the bolt class: a higher sampling ratio for Class A bolts, a medium sampling ratio for Class B bolts, and a lower sampling ratio for Class C bolts. Acoustic time data is collected through sampling, and the change in acoustic time is calculated by combining it with benchmark values. Wind speed data, vibration data, and ambient temperature data are integrated from the monitoring system. Vibration data is analyzed to extract characteristic frequency components, forming a vibration power spectrum. Acoustic time data is bound and stored with timestamps and ambient temperature data to form a time-series dataset. This process optimizes detection resources through risk classification, and the multidimensional dataset integrates various parameters to support subsequent health indicator calculations.

[0031] This invention calculates the axial force loss rate based on the temporal variation of sound through a linear mapping, with the mapping coefficient calibrated according to the bolt material properties. Vibration power spectra are extracted from multidimensional datasets and compared with reference power to determine the relative vibration impact. Temperature drift compensation coefficients are calculated using an exponential function model based on ambient temperature data. Health indicators are generated through a weighted sum of the axial force loss rate, relative vibration impact, and temperature drift compensation coefficients. This data processing integrates multi-source information; the axial force loss rate directly reflects preload changes, vibration characteristics assess the impact of external loads, temperature compensation eliminates environmental interference, and the weighted model synthesizes the contributions of various factors.

[0032] The weighting coefficient calibration described in this invention includes collecting historical operating data of the wind turbine, including axial force attenuation data, vibration load data, and temperature change data. A correlation model is established between the historical data and the bolt health status. Based on this correlation model, weighting coefficients are determined so that they reflect the coupled effects of axial force attenuation, vibration load, and temperature change. This calibration process optimizes model parameters through historical data analysis, improving the accuracy of health indicators.

[0033] In step 4 of this invention, the preset thresholds are determined by statistical analysis based on historical operating data of wind turbine units. The first threshold corresponds to a low-risk state for the bolts, and the second threshold corresponds to a high-risk state. When the health index is less than or equal to the first threshold, the system automatically marks the bolt as Grade I, extends the maintenance cycle to more than 12 months, and updates the maintenance plan through the database. When the health index is greater than the first threshold and less than or equal to the second threshold, the bolt is marked as Grade II, triggers a re-inspection process, and remeasures the acoustic time data using a portable measuring device within 3 months. When the health index is greater than the second threshold, the bolt is marked as Grade III, the system generates a maintenance work order, and automatically assigns it to maintenance personnel to perform a re-tightening operation. This step achieves bolt risk classification through threshold comparison, and the differentiated maintenance strategy avoids excessive intervention caused by uniform maintenance. After the health index is calculated, the invention directly compares and makes decisions, and the maintenance action is triggered based on quantitative indicators.

[0034] In step 4 of this invention, the system statistically analyzes the distribution of health indicators of the sampled bolts and calculates the proportion of bolts with health indicators greater than a second threshold through a data query module. When the proportion of bolts with health indicators greater than the second threshold exceeds a preset limit (set based on the wind turbine safe operation specifications), the system automatically initiates the whole-circle bolt detection process. The whole-circle bolt detection involves using portable measuring equipment to collect acoustic time data of all bolts at the connection point, and the detection data is uploaded to the platform layer for centralized analysis. This step improves risk response capabilities through group health indicator monitoring. The statistical analysis of the sampled data in this invention serves as the trigger condition for the whole-circle detection, completing the verification of high-risk bolt groups.

[0035] In step 2 of this invention, differentiated acquisition strategies are set according to bolt level. Type A bolts use a high-frequency acquisition mode, with the acquisition frequency set based on the wind turbine operating cycle; Type B bolts use a medium-frequency acquisition mode; and Type C bolts use a low-frequency acquisition mode. During acquisition, the portable measuring device integrates a temperature sensor, automatically adjusting the ultrasonic transmission power according to the current ambient temperature to ensure signal stability. The acoustic time data is bound and stored with timestamps and ambient temperature data to form a time-series dataset. This dataset is used for temperature drift compensation and vibration analysis in step 3. This step optimizes resource allocation through tiered acquisition. The acquisition strategy of this invention is based on bolt risk level, and the time-series dataset provides the foundational data for subsequent health indicator calculations.

[0036] In step 3 of this invention, the calculation of the temperature drift compensation coefficient incorporates the thermal expansion coefficient and the temperature coefficient of elastic modulus of the bolt material. A coupled model simultaneously compensates for thermal expansion effects and material elasticity changes. The vibration power spectrum is analyzed, and low-frequency and high-frequency vibration components are extracted. The low-frequency vibration components are used to assess the structural resonance risk, while the high-frequency vibration components are used to assess bolt microslippage. During weighted calculation, the weighting coefficients are dynamically adjusted according to the bolt level. For Class A bolts, the vibration component weight is increased; for Class B bolts, the temperature compensation weight is increased; and for Class C bolts, the axial force loss rate weight is increased. This step improves the accuracy of health indicators through refined parameter processing. Temperature compensation and vibration analysis are data processing steps, and the dynamic weight adjustment enhances the model's adaptability based on the bolt level.

[0037] After making a maintenance decision in step 4, the system records the actual maintenance effect data, including the change in bolt health indicators after re-tightening. The maintenance effect data is compared and analyzed with the predicted health indicators, and the prediction accuracy is calculated through the error calculation module. The parameters of the weight coefficient calibration are dynamically adjusted according to the prediction accuracy. When the prediction accuracy is lower than the preset threshold, the system triggers the weight coefficient recalibration process, and the updated weight coefficient is used for the next round of health indicator calculation. This step forms a closed-loop optimization mechanism, using the maintenance effect feedback for model calibration to achieve continuous improvement in health indicator calculation.

[0038] Existing torque measurement techniques rely on manual operation and mechanical principles, resulting in a large measurement error range. They cannot accurately reflect the axial force decay state of high-strength wind turbine bolts, making early loosening difficult to identify. Maintenance personnel, based on inaccurate measurement results, often perform uniform retightening on all bolts, including those with axial force still within the normal range. This indiscriminate maintenance not only increases unnecessary workload but also introduces additional stress due to repeated torque application, accelerating bolt material fatigue, ultimately reducing connection reliability and shortening service life. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing provided by this invention solves these problems through a sequential technical process.

[0039] The method of this invention first installs a piezoelectric sensing unit on the end face of the bolt. The piezoelectric sensing unit is a piezoelectric ceramic sheet with a fully flanged electrode structure. The central circular area serves as the positive electrode, and the outer annular area, which is fully connected to the back surface, serves as the negative electrode. A magnetic probe is used to connect the piezoelectric ceramic sheet. The magnetic probe contains a ring magnet and a spring-loaded pin. The spring-loaded pin contacts the positive electrode, and the outer annular metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time of the ultrasonic waves in the bolt is measured to establish a reference value. This step achieves non-destructive, high-precision measurement through the piezoelectric effect. The propagation time reference value serves as the basis for subsequent comparisons, overcoming the large error of the torque method.

[0040] Based on benchmark values ​​and preset risk levels for bolted connections, a multidimensional dataset is constructed, including data on acoustic time variation, wind speed, vibration spectrum, and ambient temperature. Bolts are classified according to the criticality of their connection location; for example, leaf root bolts are defined as Class A (high risk), while tower bolts are defined as Class B or C. Sampling inspections are performed based on these classifications. The sampling ratio is set according to the bolt class: Class A bolts have a higher sampling ratio, Class B bolts a medium ratio, and Class C bolts a low ratio. Acoustic time data is collected through sampling, and the acoustic time variation is calculated in conjunction with benchmark values. Simultaneously, wind speed, vibration, and ambient temperature data are integrated from the monitoring system. Vibration data undergoes spectral analysis to extract characteristic frequency components, forming a vibration power spectrum. The acoustic time data is bound to timestamps and ambient temperature data for storage, forming a time-series dataset. This process optimizes the allocation of detection resources through risk classification, integrates multi-source parameters in the multidimensional dataset, allows the acoustic time variation to directly reflect the axial force attenuation trend, characterizes external dynamic loads with wind speed and vibration data, and supports temperature compensation with ambient temperature data, thus improving the accuracy of early loosening identification.

[0041] Bolt health indices are generated through data processing based on a multidimensional dataset. The axial force loss rate is calculated using a linear mapping of acoustic time-varying quantities, with the mapping coefficient calibrated according to bolt material properties. Vibration power spectra are extracted from the multidimensional dataset and compared with baseline power to determine the relative vibration impact. Temperature drift compensation coefficients are calculated based on ambient temperature data using an exponential function model. The health indices are generated by a weighted sum of the axial force loss rate, relative vibration impact, and temperature drift compensation coefficients. The weighting coefficients are calibrated using historical wind turbine operating data, including axial force attenuation data, vibration load data, and temperature change data. A correlation model between historical data and bolt health status is established, allowing the weighting coefficients to reflect the coupled effects of axial force attenuation, vibration load, and temperature changes. This step integrates multi-source information: axial force loss rate quantifies preload attenuation, vibration characteristics assess structural risk, temperature compensation eliminates environmental drift, and the health indices serve as a comprehensive output to improve the accuracy of the condition assessment.

[0042] Health indicators are compared with preset thresholds, and differentiated maintenance strategies are matched based on the comparison results. The preset thresholds include a first threshold and a second threshold; the first threshold corresponds to a low-risk state, and the second threshold corresponds to a high-risk state. When the health indicator is less than or equal to the first threshold, the bolt is marked as Grade I, and the maintenance cycle is extended to more than 12 months. When the health indicator is greater than the first threshold but less than or equal to the second threshold, the bolt is marked as Grade II, triggering a re-inspection process and re-measuring acoustic time data within 3 months. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated for re-tightening. Furthermore, the distribution of health indicators for randomly sampled bolts is statistically analyzed, and the proportion of bolts with health indicators greater than the second threshold is calculated. When this proportion exceeds a preset limit, a full-circle bolt inspection is initiated. This step achieves automated risk classification through threshold comparison, and the differentiated maintenance strategy only performs re-tightening on high-risk bolts, avoiding excessive intervention caused by uniform maintenance and extending the life of normal bolts.

[0043] After the method is implemented, the actual maintenance effect data is recorded, including the change in the bolt health index after re-tightening. The maintenance effect data is compared and analyzed with the predicted health index to calculate the prediction accuracy. Based on the prediction accuracy, the weight coefficient calibration parameters are dynamically adjusted to form a closed-loop optimization system. Through high-precision measurement by piezoelectric sensing, data-driven modeling, and dynamic decision-making, the method effectively solves the problem of insufficient accuracy of the torque method, achieves accurate early loosening identification, reduces over-maintenance, and extends the service life of bolts.

[0044] This invention addresses the technical problem of insufficient accuracy in existing torque measurement methods, leading to difficulty in identifying early loosening of high-strength wind turbine bolts, resulting in over-maintenance and shortened lifespan. It provides a predictive maintenance method based on piezoelectric sensing. The specific implementation includes installing a piezoelectric sensing unit on the bolt end face. The piezoelectric sensing unit is a piezoelectric ceramic sheet with a fully flanged electrode structure. The central circular area serves as the positive electrode, while the outer annular area, connected to the back surface, serves as the negative electrode. A magnetic probe is used to connect the piezoelectric ceramic sheet. The magnetic probe contains a ring magnet and a spring-loaded pin. The spring-loaded pin contacts the positive electrode, and the outer annular metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time of the ultrasonic waves in the bolt is measured to establish a reference value. This step achieves high-precision measurement through the piezoelectric effect, and the propagation time reference value provides a basis for subsequent comparisons.

[0045] Based on baseline values ​​and preset risk levels for bolted connections, a multidimensional dataset is constructed, including data on acoustic time variation, wind speed, vibration spectrum, and ambient temperature. Bolts are classified according to the criticality of their connection location; for example, leaf root bolts are defined as Class A (high risk), while tower bolts are defined as Class B or C. Sampling inspections are performed based on the classification results, with the sampling ratio set according to the bolt class: a higher sampling ratio for Class A bolts, a medium sampling ratio for Class B bolts, and a lower sampling ratio for Class C bolts. Acoustic time data is collected through sampling, and the acoustic time variation is calculated in conjunction with the baseline values. Wind speed, vibration, and ambient temperature data are integrated from the monitoring system. Vibration data undergoes spectral analysis to extract characteristic frequency components, forming a vibration power spectrum. Acoustic time data is bound and stored with timestamps and ambient temperature data to form a time-series dataset. This process optimizes detection resources through risk classification, and the multidimensional dataset integrates multi-source parameters to support the calculation of health indicators.

[0046] Bolt health indices are generated through data processing based on a multidimensional dataset. The axial force loss rate is calculated using a linear mapping of acoustic time variation, with the mapping coefficient calibrated according to the bolt material properties. Vibration power spectra are extracted from the multidimensional dataset and compared with baseline power to determine the relative vibration impact. Temperature drift compensation coefficients are calculated based on ambient temperature data using an exponential function model. The health indices are generated by a weighted sum of the axial force loss rate, relative vibration impact, and temperature drift compensation coefficients. The weighting coefficients are calibrated using historical wind turbine operating data, including axial force attenuation data, vibration load data, and temperature change data. A correlation model between historical data and bolt health status is established, ensuring that the weighting coefficients reflect the coupled effects of axial force attenuation, vibration load, and temperature changes.

[0047] Health indicators are compared with preset thresholds, and differentiated maintenance strategies are matched based on the comparison results. The preset thresholds include a first threshold and a second threshold; the first threshold corresponds to a low-risk state, and the second threshold corresponds to a high-risk state. When the health indicator is less than or equal to the first threshold, the bolt is marked as Grade I, and the maintenance cycle is extended. When the health indicator is greater than the first threshold but less than or equal to the second threshold, the bolt is marked as Grade II, triggering a re-inspection process and re-measuring the acoustic time data within a specified time. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated for re-tightening. The distribution of health indicators of randomly selected bolts is statistically analyzed, and the proportion of bolts with health indicators greater than the second threshold is calculated. When this proportion exceeds a preset limit, a full-circle bolt inspection is initiated.

[0048] After the method is executed, the actual maintenance effect data is recorded, including the changes in bolt health indicators after retightening. The maintenance effect data is compared and analyzed with the predicted health indicators to calculate the prediction accuracy. The weight coefficient calibration parameters are dynamically adjusted based on the prediction accuracy. When the prediction accuracy falls below a preset threshold, the weight coefficients are recalibrated. The updated weight coefficients are used for the next round of health indicator calculation, forming a closed-loop optimization system. Non-destructive detection is achieved through piezoelectric sensing technology, and data-driven models improve early warning accuracy.

[0049] Embodiment 1 of this invention relates to a predictive maintenance application for wind turbine blade root bolts. The blade root bolt connection is classified as a Class A bolt with a high risk level. A piezoelectric sensing unit is installed on the bolt end face. The piezoelectric sensing unit uses a piezoelectric ceramic sheet structure, with the central circular area serving as the positive electrode and the outer annular area connected to the back surface as the negative electrode. A magnetic probe is used to connect to the piezoelectric ceramic sheet. The magnetic probe contains a ring magnet and a spring-loaded pin, with the spring-loaded pin contacting the positive electrode and the outer annular metal sheet contacting the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time of the ultrasonic waves in the bolt is measured to establish a baseline value. Based on the baseline value and the risk level of the blade root bolt, a high sampling rate is implemented. Acoustic time data is collected through these samplings and combined with the baseline value to calculate the change in acoustic time. Wind speed data, vibration data, and ambient temperature data are integrated from the monitoring system. Vibration data is analyzed to extract characteristic frequency components, forming a vibration power spectrum. The acoustic time data is bound and stored with timestamps and ambient temperature data to form a time-series dataset. Based on the acoustic time-varying data and a multidimensional dataset, the axial force loss rate is determined. The axial force loss rate is calculated through a linear mapping of the acoustic time-varying data, with the mapping coefficient calibrated according to the bolt material properties. The vibration power spectrum is extracted from the multidimensional dataset and compared with a benchmark power to determine the relative vibration impact. The temperature drift compensation coefficient is calculated based on ambient temperature data using an exponential function model. Health indicators are generated by a weighted sum of the axial force loss rate, relative vibration impact, and temperature drift compensation coefficient. The weighting coefficients are calibrated using historical wind turbine operating data, including axial force attenuation data, vibration load data, and temperature change data. Health indicators are compared with preset thresholds, including a first threshold and a second threshold. The first threshold corresponds to a low-risk state, and the second threshold corresponds to a high-risk state. When the health indicator is less than or equal to the first threshold, the bolt is marked as Grade I, and the maintenance cycle is extended. When the health indicator is greater than the first threshold but less than or equal to the second threshold, the bolt is marked as Grade II, triggering a re-inspection process. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated for re-tightening. The distribution of health indicators of randomly sampled bolts is statistically analyzed, and the proportion of bolts with health indicators exceeding a second threshold is calculated. When this proportion exceeds a preset limit, a full-circle bolt inspection is initiated. This invention solves the problem of early loosening identification of leaf root bolts through high-precision measurement using piezoelectric sensing and data fusion, thus avoiding excessive maintenance.

[0050] Embodiment 2 of this invention relates to a predictive maintenance application for wind turbine tower bolts. The risk level of the tower bolt connection is defined as either Class B or Class C. A piezoelectric sensing unit, consisting of a piezoelectric ceramic plate with a fully flanged electrode structure, is installed on the bolt end face. A magnetically attached probe is used to connect the piezoelectric ceramic plate, and the probe incorporates a ring magnet and a spring-loaded pin. Ultrasonic waves are excited using a portable measuring device, and the propagation time is measured to establish a baseline value. Based on the baseline value and the bolt risk level, random sampling is performed, with the sampling ratio set according to the level: a medium sampling ratio for Class B bolts and a low sampling ratio for Class C bolts. Acoustic time data is collected through random sampling, and the acoustic time variation is calculated in conjunction with the baseline value. Wind speed data, vibration data, and ambient temperature data are integrated from the monitoring system. Vibration data undergoes spectral analysis to extract characteristic frequency components. The acoustic time data forms a time-series dataset. Based on the multidimensional dataset, the axial force loss rate is determined, calculated through a linear mapping of the acoustic time variation. The vibration power spectrum is extracted from the multidimensional dataset and compared with the baseline power to obtain the relative vibration impact. A temperature drift compensation coefficient is calculated based on the ambient temperature data. Health indicators are generated through weighted summation, with weighting coefficients reflecting the coupled effects of axial force attenuation, vibration load, and temperature changes. The health indicators are compared to preset thresholds, and differentiated maintenance strategies are matched based on the comparison results. When the health indicator is less than or equal to the first threshold, the maintenance cycle is extended. When the health indicator is greater than the first threshold but less than or equal to the second threshold, a re-inspection process is triggered. When the health indicator is greater than the second threshold, re-tightening is performed. Differentiated acquisition strategies are set according to bolt grade: Class B bolts use a medium-frequency acquisition mode, and Class C bolts use a low-frequency acquisition mode. During acquisition, the portable measuring device automatically adjusts the ultrasonic emission power according to the ambient temperature. This invention optimizes tower bolt maintenance resources, improves inspection efficiency, and reduces unnecessary intervention through graded acquisition and dynamic adjustment.

[0051] The specific embodiments of the present invention are further described in detail below. The system architecture of the present invention achieves predictive maintenance of high-strength wind turbine bolts through a layered architecture, specifically including a sensing layer, an edge computing layer, and a platform layer. Each layer works collaboratively to execute the steps of the present invention and the predictive maintenance method for wind turbine bolts based on piezoelectric sensing. The sensing layer consists of miniature piezoelectric ceramic sheets permanently installed on the end face of the bolt. The piezoelectric ceramic sheets adopt a fully flanged electrode structure, with the central circular area serving as the positive electrode and the edge annular area connected to the back surface as the negative electrode. They are connected via a magnetically attached probe, which incorporates a ring magnet and a spring-loaded pin, enabling rapid and stable electrode contact within a confined space.

[0052] The edge computing layer integrates a portable axial force gauge. Its ultrasonic excitation module has a frequency of 5 MHz, and the error of the high-precision timer is less than 0.1 μs. It can measure the relative change in acoustic time ΔT / T0 within the single-bolt detection time ≤ 5 seconds and compress and transmit it to the platform layer through the data lightweight module. The platform layer includes a bolt health portrait engine that calculates the health index N based on multi-dimensional data sets (such as the change in acoustic time, real-time wind speed provided by the SCADA system, vibration spectrum, and environmental temperature data). The calculation formula is ; where the weight coefficients α, β, γ, and δ are calibrated through historical operation data to reflect the coupled effects of axial force attenuation, vibration load, and temperature change.

[0053] First, in the implementation mode of the present invention, when the bolt is initially installed, a piezoelectric ceramic piece is bonded to the end face and the initial acoustic time T0 is measured as the reference value. Secondly, according to the bolt criticality classification (such as the blade root bolt being a class A high risk), differential sampling inspection is performed (the annual sampling ratio for class A is 20%). Acoustic time data is collected through the edge layer, and a multi-dimensional data set is constructed by combining wind speed, vibration, and environmental temperature. Then, the platform layer calculates the axial force loss rate ΔF / F0 based on the linear mapping of the change in acoustic time ΔT / T0 (the mapping coefficient K is calibrated according to the material characteristics). At the same time, the relative vibration influence is obtained by extracting the vibration power spectrum and comparing it with the reference power, and temperature drift compensation is performed using an exponential function model. Finally, the health index N is generated through weighted summation. Finally, N is compared with the preset thresholds (the first threshold is 0.8, and the second threshold is 0.9): when N ≤ 0.8, it is marked as level I, and the maintenance period is extended to 12 - 18 months; when 0.8 < N ≤ 0.9, it is marked as level II, and a re-inspection is triggered within 3 months; when N > 0.9, it is marked as level III, and a maintenance work order is immediately generated to perform re-tightening. And if the proportion of bolts with N > 0.9 in the sampled bolts exceeds 50%, a full-circle bolt inspection is started.

Claims

1. A predictive maintenance method for wind turbine bolts based on piezoelectric sensing, characterized in that, include: Step 1: Collect the propagation time of the ultrasonic wave in the bolt to establish a reference value; Step 2: Based on the baseline values ​​and the preset risk levels of the bolted connections, construct a multidimensional dataset including data on acoustic time variation, wind speed, vibration spectrum, and ambient temperature. Step 3: Generate bolt health indicators through data processing based on the multidimensional dataset; Step 4: Compare the health indicators with preset thresholds, and match differentiated maintenance strategies based on the comparison results. Differentiated maintenance strategies include extending the maintenance cycle, arranging re-examination, or performing re-tightening.

2. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 1, characterized in that, Step 1 includes: A piezoelectric sensing unit is installed on the end face of the bolt. Ultrasonic waves are excited by the piezoelectric sensing unit, and the propagation time of the ultrasonic waves in the bolt is measured to establish a reference value. The piezoelectric sensing unit is a piezoelectric ceramic sheet with a fully flanged electrode structure. The central circular area serves as the positive electrode, and the edge annular area is connected to the back surface as the negative electrode. The piezoelectric ceramic sheet is connected using a magnetically attached probe. The magnetically attached probe has a built-in ring magnet and a spring pin. The spring pin contacts the positive electrode, and the outer ring metal sheet contacts the negative electrode. Megahertz-level ultrasonic waves are excited using a portable measuring device, and the propagation time is collected as a reference value.

3. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 1, characterized in that, Step 2 includes: Bolts are classified according to the risk level of the bolted connection, and random inspections are performed based on the classification results. Acoustic time data is collected through random inspections, and the change in acoustic time is calculated in combination with the benchmark value. Wind speed data, vibration data and ambient temperature data are integrated from the preset monitoring system to construct a multidimensional dataset. Bolt classification includes defining leaf root bolts as Class A high-risk and tower bolts as Class B or Class C; The sampling ratio is set according to the bolt grade. The sampling ratio of Class A bolts is higher than that of Class B bolts, and the sampling ratio of Class B bolts is higher than that of Class C bolts. Wind speed, vibration spectrum, and ambient temperature data are acquired from the monitoring system, and characteristic frequency components of the vibration power spectrum are extracted through spectrum analysis. The sound time data is bound and stored with timestamps and ambient temperature data to form a time series dataset.

4. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 1, characterized in that, Step 3 includes: Based on the acoustic time variation and the multidimensional dataset, the axial force loss rate is determined, and based on the vibration data and ambient temperature data in the multidimensional dataset, vibration characteristics and temperature compensation parameters are generated respectively. Bolt health indicators are generated through data processing. The axial force loss rate is calculated by linear mapping of the acoustic time variation, and the mapping coefficient is calibrated according to the bolt material properties. The vibration power spectrum is extracted from the multidimensional dataset and compared with the reference power to obtain the relative vibration effect. The temperature drift compensation coefficient is calculated based on ambient temperature data, using an exponential function model. The health indicators are generated by weighted summation of axial force loss rate, relative vibration effect, and temperature drift compensation coefficient.

5. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 4, characterized in that, Weighted summation weight coefficient calibration includes: Collect historical operating data of wind turbines, including axial force attenuation data, vibration load data, and temperature change data. Establish a correlation model between historical data and bolt health status; The weighting coefficients are determined based on the correlation model, so that the weighting coefficients reflect the coupled effects of axial force attenuation, vibration load and temperature change.

6. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 1, characterized in that, Step 4 includes: The preset thresholds include a first threshold and a second threshold. The first threshold corresponds to a low-risk state, and the second threshold corresponds to a high-risk state. When the health indicators are less than or equal to the first threshold, the bolts are marked as Grade I, and the maintenance cycle is extended to more than 12 months. When the health indicators are greater than the first threshold and less than or equal to the second threshold, the bolt is marked as Grade II, triggering the re-inspection process and re-measuring the acoustic time data within 3 months. When the health indicator is greater than the second threshold, the bolt is marked as Grade III, and a maintenance work order is generated to perform re-tightening.

7. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 6, characterized in that, Step 4 also includes: The distribution of health indicators of randomly inspected bolts was statistically analyzed, and the proportion of bolts with health indicators greater than the second threshold was calculated. When the proportion of bolts with health indicators exceeding the second threshold exceeds the preset limit, a full-circle bolt detection is initiated.

8. The predictive maintenance method for wind turbine bolts based on piezoelectric sensing as described in claim 3, characterized in that, Step 2 also includes: Different data acquisition strategies are set according to the bolt grade: high-frequency acquisition mode is used for Class A bolts, medium-frequency acquisition mode is used for Class B bolts, and low-frequency acquisition mode is used for Class C bolts. During the data acquisition process, the portable measuring device automatically adjusts the ultrasonic emission power according to the current ambient temperature; The time-series dataset formed from the acoustic time data is used for temperature drift compensation and vibration analysis in step 3.

9. The method for wind turbine bolt inspection-based maintenance and repair as described in claim 4, characterized in that, Step 3 also includes: The calculation of the temperature drift compensation coefficient incorporates the thermal expansion coefficient and the temperature coefficient of elastic modulus of the bolt material, and simultaneously compensates for the thermal expansion effect and the elastic change of the material through a coupled model. The vibration power spectrum was analyzed, and low-frequency vibration components and high-frequency vibration components were extracted respectively. The low-frequency vibration components were used to assess the risk of structural resonance, and the high-frequency vibration components were used to assess the micro-slippage of bolts. During weighted calculation, the weighting coefficients are dynamically adjusted according to the bolt grade. For Class A bolts, the vibration component weight is increased; for Class B bolts, the temperature compensation weight is increased; and for Class C bolts, the axial force loss rate weight is increased.

10. The method for wind turbine bolt inspection-based maintenance and repair as described in claim 7, characterized in that, Step 4 also includes: After making the maintenance decision in step 4, record the actual maintenance effect data, which includes the change in the bolt health index after re-tightening. The maintenance effect data is compared and analyzed with the predicted health index to calculate the prediction accuracy. The parameters of the weighting coefficients are dynamically adjusted based on the prediction accuracy. When the prediction accuracy falls below a preset threshold, the weight coefficients are recalibrated to obtain updated weight coefficients, which are then used for the next round of health index calculation.

Images