A fan blade crack monitoring method and system based on voiceprint recognition

Abstract

The application relates to the technical field of fan blade monitoring, and discloses a fan blade crack monitoring method and system based on voiceprint recognition, which comprises the following steps: collecting environmental parameters and establishing a sound velocity correction model; directionally tracking a noise source and dynamically adjusting microphone array parameters; emitting excitation through an excitation unit carried by a UAV according to blade attitude dynamic adjustment parameters, and synchronously collecting voiceprint signals; after noise reduction and correction processing, extracting multi-order modal characteristics and coupling degree coefficients; and comparing with a dynamic baseline to determine cracks. The system comprises environmental parameter collection, noise source tracking, sound velocity correction, UAV excitation, multi-channel microphone array, signal preprocessing, feature extraction and baseline management modules, and the modules are connected through wired / wireless communication. The application improves signal quality through environmental parameter correction and noise suppression, realizes non-stop monitoring by combining dynamic excitation and accurate collection, and effectively improves crack detection accuracy and monitoring efficiency.

Description

Technical Field

[0001] This invention relates to the field of wind turbine blade monitoring technology, specifically to a method and system for monitoring wind turbine blade cracks based on voiceprint recognition. Background Technology

[0002] Wind turbine blades are the core components of wind turbine generators, and their operating status directly determines the power generation efficiency and safety stability of the turbine. Operating in a complex outdoor environment for extended periods, blades are susceptible to cracks caused by wind impact and environmental corrosion. If not detected in time, these cracks will gradually expand, eventually leading to blade damage or even turbine shutdown. Therefore, wind turbine blade crack monitoring is a crucial aspect of wind farm operation and maintenance. Currently, methods and systems for monitoring wind turbine blade cracks based on acoustic signature recognition exist in the industry. By collecting and analyzing acoustic signature signals from the blades, crack identification and monitoring are achieved, and these methods are widely used in the operation and maintenance monitoring of various wind farms.

[0003] Existing methods and systems for monitoring wind turbine blade cracks based on voiceprint recognition can collect environmental parameters and establish voiceprint correction models, and also have noise suppression functions. However, they cannot accurately identify the specific information of various noise sources, making it difficult to suppress interference in a targeted manner. As a result, environmental factors and background noise can easily interfere with the monitoring signal, affecting the authenticity and reliability of the voiceprint signal and failing to provide high-quality data support for crack determination. The acoustic excitation units used are mostly fixed installations, which cannot adapt to real-time changes in blade attitude, making it difficult to achieve accurate excitation and voiceprint collection across the entire blade area. Moreover, the monitoring process often requires shutdown, which seriously affects the normal operation of the wind turbine and reduces the convenience and efficiency of monitoring, failing to meet the working requirements of wind turbine blade monitoring. Therefore, a method and system for monitoring wind turbine blade cracks based on voiceprint recognition is proposed. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for monitoring wind turbine blade cracks based on voiceprint recognition, thereby solving the technical problems of large monitoring interference, poor adaptability, low recognition accuracy, and high operation and maintenance costs.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring wind turbine blade cracks based on voiceprint recognition, comprising the following steps: S1 Environmental Parameter Acquisition and Sound Velocity Correction Model Establishment: First, the temperature, humidity, wind speed, and background noise of the wind farm are collected in real time, and a sound speed correction model is established. At the same time, the location, spectrum, and intensity change curve of the main noise sources are identified and marked through the noise source directional tracking module. S2 blade directional acoustic excitation and acoustic signature acquisition: Secondly, the fundamental frequency of the natural vibration of the blade is collected to generate an adaptive frequency sweep excitation signal covering at least two harmonics. At the same time, the excitation incident angle, amplitude and frequency sweep rate are dynamically adjusted according to the real-time torsional attitude and windward angle of the blade by the directional acoustic excitation unit carried by the UAV, and the excitation is transmitted to the blade. Simultaneously, the excitation signal of the blade under excitation resonance is collected by the multi-channel microphone array around the tower. S3 Voiceprint Signal Noise Reduction and Correction Preprocessing: Adaptive filtering and noise reduction are performed on the original voiceprint signal in step S2, and the microphone array beamforming parameters are switched in real time based on the dynamic information of the noise source. At the same time, dynamic nulls are generated to align with all major noise sources. Then, the signal delay and amplitude are corrected by the sound velocity correction model to obtain a high signal-to-noise ratio target voiceprint signal. S4 blade acoustic signature mode feature extraction: Perform FFT spectrum analysis on the target acoustic signature signal in step S3 to extract the first to fifth order modal frequencies, damping ratios and sub-band energy distribution entropy values ​​of the blade, and further calculate the coupling coefficient of adjacent modal features and modal energy transfer efficiency, which together form the core acoustic signature features. S5 Feature Baseline Comparison and Crack Location Determination: The dynamic acoustic signature baseline of the blade throughout its entire life cycle is retrieved and weighted by wind turbine operating time, cumulative power generation, environmental corrosion and blade dust accumulation. The current acoustic signature is then compared with the dynamic baseline, and if it exceeds the threshold, a crack is determined to exist.

[0006] A wind turbine blade crack monitoring system based on voiceprint recognition, employing the aforementioned wind turbine blade crack monitoring method based on voiceprint recognition, is characterized by comprising an environmental parameter acquisition module, a noise source directional tracking module, a sound velocity correction module, a UAV excitation module, a multi-channel microphone array module, a signal preprocessing module, a feature extraction module, and a baseline management module. The environmental parameter acquisition module is connected to the sound velocity correction module via wired communication. It is used to collect temperature, humidity, and wind speed parameters at different heights of the wind turbine tower and transmit them to the sound velocity correction module. The noise source directional tracking module is connected to the multi-channel microphone array module and the signal preprocessing module via wireless communication. It is used to identify and mark the location, spectrum, and intensity changes of the main noise sources through beam scanning technology and transmit this information to the signal preprocessing module. The sound velocity correction module is connected to the signal preprocessing module via wired communication. It is used to train the sound velocity correction model based on historical data and output real-time correction coefficients. The UAV excitation module is connected to the environmental parameter acquisition module and the multi-channel microphone array module via wireless communication. It is used to acquire the fundamental frequency of the natural vibration of the blade and generate an adaptive frequency sweep excitation signal. At the same time, it dynamically adjusts the excitation parameters according to the blade attitude. The multi-channel microphone array module is connected to the signal preprocessing module via wired communication. It is used to synchronously acquire the resonance acoustic wave signal generated by the excitation of the blade. The signal preprocessing module is connected to the feature extraction module via wired communication and is used to perform adaptive filtering and noise reduction and sound velocity correction on the original acoustic signature signal. The feature extraction module is connected to the crack detection module via wired communication and is used to extract multi-mode parameters and coupling features of the blade and fuse them into core acoustic signature features. The baseline management module is connected to the crack detection module via wired communication.

[0007] Preferably, in step S1, multiple sets of environmental sensors are arranged at different heights on the wind turbine tower to collect temperature, humidity, and wind speed parameters at different heights. The average environmental parameters of the wind farm are calculated comprehensively. At the same time, the sound velocity correction model is trained and established based on historical environmental data and corresponding measured sound velocity values, and the correction coefficient is dynamically output according to the real-time collected environmental parameters. Meanwhile, the noise source directional tracking module uses the beam scanning technology of the microphone array to perform an omnidirectional scan of the background noise on site, distinguishing between the wind turbine's own operating noise and external environmental noise, and marking the dynamic change information of various noise sources. This can obtain more representative wind farm environmental parameters, improve the adaptability and accuracy of the sound velocity correction model, and accurately distinguish different types of noise sources, providing targeted support for subsequent signal noise reduction and ensuring the reliability of acoustic signal acquisition.

[0008] Preferably, in step S2, the acquisition of the blade's natural vibration fundamental frequency is carried out when the wind turbine is in an unloaded and stopped state. During the generation of the adaptive frequency sweep excitation signal, the background noise spectrum information obtained in step S1 is combined synchronously to automatically avoid the frequency band where noise energy is concentrated. At the same time, when acquiring the fundamental frequency signal, at least three sets of data are continuously acquired and averaged to obtain a purer blade natural vibration fundamental frequency signal, avoid interference of noise frequency bands on the excitation signal, improve the effectiveness of the excitation signal, ensure that the blade can generate a stable and identifiable resonance sound pattern after being excited, and improve the quality of monitoring basic data.

[0009] Preferably, in step S2, before operation, the directional acoustic excitation unit carried by the UAV first acquires real-time attitude images of the blades through the onboard vision module, calculates the blade torsion angle and windward angle by combining the wind turbine's operating data, and automatically adjusts its flight position and attitude according to the calculation results, so that the emission direction of the directional acoustic excitation unit is always perpendicular to the excited surface of the blade. This can achieve precise alignment between the excitation direction and the excited surface of the blade, ensuring that the excitation energy is efficiently applied to the blade, improving the strength and clarity of the acoustic signal, and adapting to changes in blade attitude to ensure comprehensive monitoring of the entire area.

[0010] Preferably, in step S3, the adaptive filtering noise reduction first performs frame-by-frame processing on the original voiceprint signal, calculates the energy distribution of the signal frame by frame, and initially filters out background noise with too low energy. Then, based on the dynamic information of the noise sources marked in step S1, the beamforming weight of the microphone array is adjusted in real time to accurately focus the main beam of the array on the stimulated area of ​​the blade. At the same time, a deep dynamic null is generated in all directions of the main noise sources identified. The filtered signal is then subjected to time delay and amplitude correction through the sound velocity correction model to eliminate the influence of environmental factors on signal propagation. This can optimize the voiceprint signal noise reduction effect step by step, accurately focus on the stimulated area of ​​the blade and suppress various noise interferences. Combined with sound velocity correction, the environmental influence is completely eliminated, and the signal-to-noise ratio of the target voiceprint signal is greatly improved, laying the foundation for feature extraction.

[0011] Preferably, in step S4, before performing FFT spectrum analysis on the target acoustic signature signal, the signal is first windowed. When extracting the modal frequency and damping ratio, a combination of peak search and curve fitting is used to identify the characteristic parameters of each modality. After calculating all characteristic parameters, each parameter is normalized, and different weights are assigned according to the sensitivity of each parameter to cracks. These weights are then fused to form a core acoustic signature feature vector that reflects the structural state of the blade. This reduces spectrum analysis errors, accurately identifies the characteristic parameters of each modality, and highlights crack-sensitive features through weight allocation, making the core acoustic signature features more comprehensive and accurate, thereby improving the accuracy and sensitivity of subsequent crack identification.

[0012] Preferably, in step S5, the initial value of the dynamic acoustic signature baseline for the entire life cycle is established by collecting acoustic signature signals of the blades at least four times under standard environmental conditions after the wind turbine blades are installed and accepted, and taking the average value. The self-updating process of the baseline is automatically triggered after each blade monitoring is completed. The baseline is weighted and adjusted by combining the wind turbine's operating data and environmental data, which can establish a baseline that is more in line with the initial health state of the blades, ensuring the scientific and timely nature of the baseline self-updating, adapting to the long-term operating status changes of the blades, effectively reducing the probability of misjudgment and missed judgment in crack detection, and ensuring the reliability of monitoring.

[0013] Preferably, the UAV excitation module integrates an airborne vision submodule, an attitude control submodule, and a directional acoustic wave emission submodule. The airborne vision submodule is connected to the attitude control submodule via an internal bus and is used to acquire image information of the wind turbine blades in real time and identify the blade's torsional attitude and windward angle. The attitude control submodule is connected to the ground control terminal via wireless communication and is used to automatically adjust the UAV's flight position and hovering attitude according to the blade attitude information, ensuring that the emission direction of the directional acoustic wave emission submodule is always perpendicular to the excited surface of the blade. The directional acoustic wave emission submodule is connected to the airborne vision submodule via an internal bus and is used to generate an adaptive sweep frequency excitation signal by combining the blade's fundamental frequency signal and environmental noise information, and dynamically adjust the excitation parameters according to the structural characteristics of different regions of the blade.

[0014] Preferably, the baseline management module is connected to the wind farm remote operation and maintenance platform via an industrial Ethernet to acquire data on wind turbine operating time, cumulative power generation, environmental corrosion level, and blade dust accumulation. The baseline management module is internally equipped with a baseline storage unit and a baseline update calculation unit. The baseline storage unit is used to store the initial health baseline and historical monitoring baseline data of the blades. The baseline update calculation unit is connected to the baseline storage unit via an internal bus and is used to perform weighted adjustment of the dynamic baseline after each blade monitoring is completed, combining the latest operating data and acoustic signature data.

[0015] Compared with the prior art, the present invention provides a method and system for monitoring wind turbine blade cracks based on voiceprint recognition, which has the following beneficial effects: 1. This wind turbine blade crack monitoring method and system based on acoustic signature recognition, by collecting wind farm environmental parameters in real time and establishing an acoustic signature correction model, combined with noise source directional tracking function, can accurately identify and suppress various noise interferences, avoid the influence of environmental factors and background noise on the monitoring signal, ensure the authenticity and reliability of the acoustic signature signal, and provide high-quality data support for subsequent crack judgment. Furthermore, through the design of a UAV equipped with a directional acoustic excitation unit, it can flexibly adapt to the real-time attitude changes of the blade and dynamically adjust the excitation parameters to achieve accurate excitation and acoustic signature collection of the entire blade area. Monitoring can be completed without stopping the machine, which greatly reduces the impact of the monitoring process on the normal operation of the wind turbine and improves the convenience and efficiency of monitoring. 2. This wind turbine blade crack monitoring method and system based on voiceprint recognition, through the extraction and fusion of multi-dimensional voiceprint features, can comprehensively reflect the structural state of the blade. Compared with single feature monitoring, it can more accurately capture subtle changes caused by crack formation, improving the sensitivity of crack identification. The self-updating design of the dynamic baseline throughout the entire life cycle can adapt to changes in the blade's condition such as aging and wear during long-term operation, avoiding misjudgments and missed judgments caused by a fixed baseline, extending the service life of the monitoring system, reducing operation and maintenance costs, providing reliable protection for the safe and stable operation of wind turbine blades, and indirectly improving the operational efficiency of wind farms. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system architecture block diagram of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides a technical solution: a method for monitoring wind turbine blade cracks based on voiceprint recognition, comprising the following steps: S1 Environmental Parameter Acquisition and Sound Velocity Correction Model Establishment: First, the temperature, humidity, wind speed, and background noise of the wind farm are collected in real time, and a sound speed correction model is established. At the same time, the location, spectrum, and intensity change curve of the main noise sources are identified and marked through the noise source directional tracking module. Meanwhile, multiple sets of environmental sensors are deployed at different heights on the wind turbine tower to collect temperature, humidity and wind speed parameters at different heights. The average environmental parameters of the wind farm site are calculated by combining the data. At the same time, the sound speed correction model is trained and established based on historical environmental data and corresponding measured sound speed values. The correction coefficient is dynamically output according to the real-time collected environmental parameters. Meanwhile, the noise source directional tracking module uses the beam scanning technology of the microphone array to perform an omnidirectional scan of the background noise on site, distinguishing the wind turbine's own operating noise from the external environmental noise, and marking the dynamic change information of various noise sources. S2 blade directional acoustic excitation and acoustic signature acquisition: Secondly, the fundamental frequency of the natural vibration of the blade is collected to generate an adaptive frequency sweep excitation signal covering at least two harmonics. At the same time, the excitation incident angle, amplitude and frequency sweep rate are dynamically adjusted according to the real-time torsional attitude and windward angle of the blade by the directional acoustic excitation unit carried by the UAV, and the excitation is transmitted to the blade. Simultaneously, the excitation signal of the blade under excitation resonance is collected by the multi-channel microphone array around the tower. Furthermore, the acquisition of the fundamental frequency of the blade's natural vibration is carried out when the wind turbine is in an unloaded and stopped state. In the process of generating the adaptive frequency sweep excitation signal, the background noise spectrum information obtained in step S1 is combined synchronously to automatically avoid the frequency band where noise energy is concentrated. At the same time, when acquiring the fundamental frequency signal, at least three sets of data are continuously acquired and averaged. Before operation, the directional acoustic excitation unit on the drone first acquires real-time attitude images of the blades through the onboard vision module, and calculates the twist angle and windward angle of the blades by combining the wind turbine's operating data. At the same time, based on the calculation results, the drone automatically adjusts its flight position and attitude so that the emission direction of the directional acoustic excitation unit is always perpendicular to the excited surface of the blades. S3 Voiceprint Signal Noise Reduction and Correction Preprocessing: Adaptive filtering and noise reduction are performed on the original voiceprint signal in step S2, and the microphone array beamforming parameters are switched in real time based on the dynamic information of the noise source. At the same time, dynamic nulls are generated to align with all major noise sources. Then, the signal delay and amplitude are corrected by the sound velocity correction model to obtain a high signal-to-noise ratio target voiceprint signal. The adaptive filtering noise reduction first processes the original voiceprint signal into frames, calculates the energy distribution of the signal frame by frame, and initially filters out background noise with too low energy. Then, based on the dynamic information of the noise source marked in step S1, the beamforming weight of the microphone array is adjusted in real time to accurately focus the main beam of the array on the excited area of ​​the blade. At the same time, a deep dynamic null is generated in all directions of the main noise source identified. The filtered signal is then corrected for time delay and amplitude through the sound speed correction model to eliminate the influence of environmental factors on signal propagation. S4 blade acoustic signature mode feature extraction: Perform FFT spectrum analysis on the target acoustic signature signal in step S3 to extract the first to fifth order modal frequencies, damping ratios and sub-band energy distribution entropy values ​​of the blade, and further calculate the coupling coefficient of adjacent modal features and modal energy transfer efficiency, which together form the core acoustic signature features. Before performing FFT spectrum analysis on the target acoustic signature signal, the signal is first windowed. When extracting the modal frequency and damping ratio, a combination of peak search and curve fitting is used to identify the characteristic parameters of each mode. After calculating all the characteristic parameters, each parameter is normalized and different weights are assigned according to the sensitivity of each parameter to the crack. The result is fused to form the core acoustic signature feature vector that reflects the structural state of the blade. S5 Feature Baseline Comparison and Crack Location Determination: The dynamic acoustic signature baseline of the blade throughout its entire life cycle is retrieved and weighted by wind turbine running time, cumulative power generation, environmental corrosion and blade dust accumulation. The current acoustic signature is then compared with the dynamic baseline. If the threshold is exceeded, a crack is determined to exist. The initial value of the dynamic acoustic signature baseline throughout the entire life cycle is established by collecting acoustic signature signals of the blades at least four times under standard environmental conditions after the wind turbine blades are installed and accepted, and taking the average value. The baseline self-update process is automatically triggered after each blade monitoring is completed, and the baseline is adjusted by weighting the wind turbine's operating data and environmental data.

[0019] By collecting wind farm environmental parameters in real time and establishing an acoustic signature correction model, combined with noise source directional tracking, various noise interferences can be accurately identified and suppressed, avoiding the influence of environmental factors and background noise on the monitoring signal. This ensures the authenticity and reliability of the acoustic signature signal, providing high-quality data support for subsequent crack detection. Furthermore, the design of a UAV equipped with a directional acoustic excitation unit allows for flexible adaptation to real-time blade attitude changes and dynamic adjustment of excitation parameters, achieving precise excitation and acoustic signature acquisition across the entire blade area. Monitoring can be completed without downtime, significantly reducing the impact of monitoring on normal wind turbine operation and improving the convenience and efficiency of monitoring. Simultaneously, the fusion of multi-dimensional acoustic signature features provides a comprehensive reflection of the blade's structural state. Compared to single-feature monitoring, it can more accurately capture subtle changes caused by crack formation, improving the sensitivity of crack identification. The self-updating design of the dynamic baseline throughout the entire lifecycle adapts to changes in the blade's condition, such as aging and wear, during long-term operation, avoiding misjudgments and omissions caused by a fixed baseline. This extends the service life of the monitoring system, reduces maintenance costs, provides reliable assurance for the safe and stable operation of wind turbine blades, and indirectly improves the operational efficiency of the wind farm.

[0020] Please see Figure 2 A wind turbine blade crack monitoring system based on voiceprint recognition, employing the aforementioned wind turbine blade crack monitoring method based on voiceprint recognition, is characterized by comprising an environmental parameter acquisition module, a noise source directional tracking module, a sound velocity correction module, a UAV excitation module, a multi-channel microphone array module, a signal preprocessing module, a feature extraction module, and a baseline management module.

[0021] The environmental parameter acquisition module is connected to the sound velocity correction module via wired communication. It is used to collect temperature, humidity, and wind speed parameters at different heights of the wind turbine tower and transmit them to the sound velocity correction module. The noise source directional tracking module is connected to the multi-channel microphone array module and the signal preprocessing module via wireless communication. It is used to identify and mark the location, spectrum, and intensity changes of the main noise sources through beam scanning technology and transmit this information to the signal preprocessing module.

[0022] The sound velocity correction module is connected to the signal preprocessing module via wired communication. It is used to train the sound velocity correction model based on historical data and output real-time correction coefficients. The UAV excitation module is connected to the environmental parameter acquisition module and the multi-channel microphone array module via wireless communication. It is used to acquire the fundamental frequency of the natural vibration of the blade and generate an adaptive frequency sweep excitation signal. At the same time, it dynamically adjusts the excitation parameters according to the blade attitude. The multi-channel microphone array module is connected to the signal preprocessing module via wired communication. It is used to synchronously acquire the resonance acoustic wave signal generated by the excitation of the blade.

[0023] The signal preprocessing module is connected to the feature extraction module via wired communication and is used to perform adaptive filtering and noise reduction and sound velocity correction on the original acoustic signature signal. The feature extraction module is connected to the crack detection module via wired communication and is used to extract multi-mode parameters and coupling features of the blade and fuse them into core acoustic signature features. The baseline management module is connected to the crack detection module via wired communication.

[0024] The UAV excitation module integrates an airborne vision submodule, an attitude control submodule, and a directional acoustic wave emission submodule. The airborne vision submodule is connected to the attitude control submodule via an internal bus to acquire real-time image information of the wind turbine blades and identify the blade's torsional attitude and angle of attack. The attitude control submodule is connected to the ground control terminal via wireless communication to automatically adjust the UAV's flight position and hovering attitude based on the blade attitude information, ensuring that the emission direction of the directional acoustic wave emission submodule is always perpendicular to the excited surface of the blade. The directional acoustic wave emission submodule is connected to the airborne vision submodule via an internal bus to generate an adaptive sweep frequency excitation signal by combining the blade's fundamental frequency signal and environmental noise information, and dynamically adjusts the excitation parameters according to the structural characteristics of different areas of the blade.

[0025] The baseline management module is connected to the wind farm's remote operation and maintenance platform via industrial Ethernet to acquire data on wind turbine runtime, cumulative power generation, environmental corrosion level, and blade dust accumulation. The baseline management module is equipped with a baseline storage unit and a baseline update calculation unit. The baseline storage unit stores the initial health baseline and historical monitoring baseline data of the blades. The baseline update calculation unit is connected to the baseline storage unit via an internal bus and is used to perform weighted adjustments to the dynamic baseline after each blade monitoring is completed, combining the latest operating data and acoustic signature data.

[0026] This solution, by collecting wind farm environmental parameters in real time and establishing an acoustic signature correction model, combined with noise source directional tracking, can accurately identify and suppress various noise interferences, avoiding the influence of environmental factors and background noise on the monitoring signal, ensuring the authenticity and reliability of the acoustic signature signal, and providing high-quality data support for subsequent crack detection. Furthermore, the design of a UAV equipped with a directional acoustic excitation unit can flexibly adapt to real-time blade attitude changes and dynamically adjust excitation parameters, achieving accurate excitation and acoustic signature acquisition of the entire blade area. Monitoring can be completed without downtime, significantly reducing the impact of the monitoring process on the normal operation of the wind turbine, and improving the convenience and efficiency of monitoring. Simultaneously, through the extraction and fusion of multi-dimensional acoustic signature features, the structural state of the blade can be comprehensively reflected. Compared with single-feature monitoring, it can more accurately capture subtle changes caused by crack formation, improving the sensitivity of crack identification. The self-updating design of the dynamic baseline throughout the entire life cycle can adapt to changes in the aging and wear states of the blades during long-term operation, avoiding misjudgments and omissions caused by a fixed baseline, extending the service life of the monitoring system, reducing operation and maintenance costs, providing reliable protection for the safe and stable operation of wind turbine blades, and indirectly improving the operational efficiency of the wind farm.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0028] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring wind turbine blade cracks based on voiceprint recognition, characterized in that, Includes the following steps: S1 Environmental Parameter Acquisition and Sound Velocity Correction Model Establishment: First, the temperature, humidity, wind speed, and background noise of the wind farm are collected in real time, and a sound speed correction model is established. At the same time, the location, spectrum, and intensity change curve of the main noise sources are identified and marked through the noise source directional tracking module. S2 blade directional acoustic excitation and acoustic signature acquisition: Secondly, the fundamental frequency of the natural vibration of the blade is collected to generate an adaptive frequency sweep excitation signal covering at least two harmonics. At the same time, the excitation incident angle, amplitude and frequency sweep rate are dynamically adjusted according to the real-time torsional attitude and windward angle of the blade by the directional acoustic excitation unit carried by the UAV, and the excitation is transmitted to the blade. Simultaneously, the excitation signal of the blade under excitation resonance is collected by the multi-channel microphone array around the tower. S3 Voiceprint Signal Noise Reduction and Correction Preprocessing: Adaptive filtering and noise reduction are performed on the original voiceprint signal in step S2, and the microphone array beamforming parameters are switched in real time based on the dynamic information of the noise source. At the same time, dynamic nulls are generated to align with all major noise sources. Then, the signal delay and amplitude are corrected by the sound velocity correction model to obtain a high signal-to-noise ratio target voiceprint signal. S4 blade acoustic signature mode feature extraction: Perform FFT spectrum analysis on the target acoustic signature signal in step S3 to extract the first to fifth order modal frequencies, damping ratios and sub-band energy distribution entropy values ​​of the blade, and further calculate the coupling coefficient of adjacent modal features and modal energy transfer efficiency, which together form the core acoustic signature features. S5 Feature Baseline Comparison and Crack Location Determination: The dynamic acoustic signature baseline of the blade throughout its entire life cycle is retrieved and weighted by wind turbine operating time, cumulative power generation, environmental corrosion and blade dust accumulation. The current acoustic signature is then compared with the dynamic baseline, and if it exceeds the threshold, a crack is determined to exist.

2. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S1, multiple sets of environmental sensors are arranged at different heights of the wind turbine tower to collect temperature, humidity and wind speed parameters at different heights. The average environmental parameters of the wind farm site are calculated. At the same time, the sound velocity correction model is trained and established based on historical environmental data and corresponding measured sound velocity values. The correction coefficient is dynamically output according to the real-time collected environmental parameters. Meanwhile, the noise source directional tracking module uses the beam scanning technology of the microphone array to perform an omnidirectional scan of the background noise on site, distinguishes the wind turbine's own operating noise from the external environmental noise, and marks the dynamic change information of various noise sources.

3. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S2, the acquisition of the fundamental frequency of the blade's natural vibration is carried out when the wind turbine is in an unloaded and stopped state. During the generation of the adaptive frequency sweep excitation signal, the background noise spectrum information obtained in step S1 is combined synchronously to automatically avoid the frequency band where noise energy is concentrated. At the same time, when acquiring the fundamental frequency signal, at least three sets of data are continuously acquired and averaged.

4. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S2, before the directional acoustic excitation unit on the UAV starts working, it first acquires the real-time attitude image of the blade through the airborne vision module, and calculates the twist angle and windward angle of the blade by combining the wind turbine's operating data. At the same time, based on the calculation results, the UAV automatically adjusts its own flight position and attitude so that the emission direction of the directional acoustic excitation unit is always perpendicular to the excited surface of the blade.

5. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S3, the adaptive filtering noise reduction first performs frame-by-frame processing on the original acoustic signal, calculates the energy distribution of the signal frame by frame, and initially filters out background noise with too low energy. Then, based on the dynamic information of the noise sources marked in step S1, the beamforming weight of the microphone array is adjusted in real time to accurately focus the main beam of the array on the excited area of ​​the blade. At the same time, a deep dynamic null is generated in all directions of the main noise sources. The filtered signal is then corrected for time delay and amplitude through the sound speed correction model to eliminate the influence of environmental factors on signal propagation.

6. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S4, before performing FFT spectrum analysis on the target acoustic signature signal, the signal is first windowed. When extracting the modal frequency and damping ratio, a combination of peak search and curve fitting is used to identify the characteristic parameters of each mode. After calculating all the characteristic parameters, each parameter is normalized and different weights are assigned according to the sensitivity of each parameter to the crack, and the core acoustic signature feature vector reflecting the structural state of the blade is formed.

7. The method for monitoring wind turbine blade cracks based on voiceprint recognition according to claim 1, characterized in that: In step S5, the initial value of the dynamic acoustic signature baseline for the entire life cycle is established by collecting acoustic signature signals of the blades at least four times under standard environmental conditions after the wind turbine blades are installed and accepted, and taking the average value. The self-updating process of the baseline is automatically triggered after each blade monitoring is completed, and the baseline is adjusted by weighting in combination with the wind turbine's operating data and environmental data.

8. A wind turbine blade crack monitoring system based on voiceprint recognition, employing the wind turbine blade crack monitoring method based on voiceprint recognition as described in any one of claims 1-7, characterized in that, It includes an environmental parameter acquisition module, a noise source directional tracking module, a sound velocity correction module, a UAV excitation module, a multi-channel microphone array module, a signal preprocessing module, a feature extraction module, and a baseline management module. The environmental parameter acquisition module is connected to the sound velocity correction module via wired communication. It is used to collect temperature, humidity, and wind speed parameters at different heights of the wind turbine tower and transmit them to the sound velocity correction module. The noise source directional tracking module is connected to the multi-channel microphone array module and the signal preprocessing module via wireless communication. It is used to identify and mark the location, spectrum, and intensity changes of the main noise sources through beam scanning technology and transmit this information to the signal preprocessing module. The sound velocity correction module is connected to the signal preprocessing module via wired communication. It is used to train the sound velocity correction model based on historical data and output real-time correction coefficients. The UAV excitation module is connected to the environmental parameter acquisition module and the multi-channel microphone array module via wireless communication. It is used to acquire the fundamental frequency of the natural vibration of the blade and generate an adaptive frequency sweep excitation signal. At the same time, it dynamically adjusts the excitation parameters according to the blade attitude. The multi-channel microphone array module is connected to the signal preprocessing module via wired communication. It is used to synchronously acquire the resonance acoustic wave signal generated by the excitation of the blade. The signal preprocessing module is connected to the feature extraction module via wired communication and is used to perform adaptive filtering and noise reduction and sound velocity correction on the original acoustic signature signal. The feature extraction module is connected to the crack detection module via wired communication and is used to extract multi-mode parameters and coupling features of the blade and fuse them into core acoustic signature features. The baseline management module is connected to the crack detection module via wired communication.

9. A wind turbine blade crack monitoring system based on voiceprint recognition according to claim 8, characterized in that: The UAV excitation module integrates an airborne vision submodule, an attitude control submodule, and a directional acoustic wave emission submodule. The airborne vision submodule is connected to the attitude control submodule via an internal bus and is used to acquire image information of the wind turbine blades in real time and identify the blade's torsional attitude and windward angle. The attitude control submodule is connected to the ground control terminal via wireless communication and is used to automatically adjust the UAV's flight position and hovering attitude based on the blade attitude information, ensuring that the emission direction of the directional acoustic wave emission submodule is always perpendicular to the excited surface of the blade. The directional acoustic wave emission submodule is connected to the airborne vision submodule via an internal bus and is used to generate an adaptive sweep frequency excitation signal by combining the blade's fundamental frequency signal and environmental noise information, and dynamically adjust the excitation parameters according to the structural characteristics of different regions of the blade.

10. A wind turbine blade crack monitoring system based on voiceprint recognition according to claim 8, characterized in that: The baseline management module is connected to the wind farm's remote operation and maintenance platform via an industrial Ethernet network. It is used to acquire data on wind turbine operating time, cumulative power generation, environmental corrosion level, and blade dust accumulation. The baseline management module is equipped with a baseline storage unit and a baseline update calculation unit. The baseline storage unit is used to store the initial health baseline and historical monitoring baseline data of the blades. The baseline update calculation unit is connected to the baseline storage unit via an internal bus and is used to perform weighted adjustments to the dynamic baseline after each blade monitoring is completed, combining the latest operating data and acoustic signature data.

Images