Wind blade crack acoustic emission early warning method, system, equipment and medium

Abstract

The invention discloses a wind power blade crack acoustic emission early warning method, system, equipment and medium, and belongs to the technical field of structural health monitoring, and the method comprises the following steps: receiving acoustic emission signals and working condition data collected by a multi-sensor array; distinguishing ice layer noise and crack signal shielding ice layer noise interference for the acoustic emission signal to obtain an effective acoustic emission signal; extracting features of the effective acoustic emission signals, and constructing feature vectors; inputting the feature vector and the working condition data into a prediction model, and outputting a crack propagation amount prediction value in a future period of time; and matching a corresponding reinforcement scheme in the knowledge base according to the predicted value of the crack propagation amount, and outputting a crack coordinate and the reinforcement scheme. According to the method, the acoustic emission signals are collected through the multi-sensor array, preprocessing and noise distinguishing are combined, the accuracy of crack recognition in the cold environment is remarkably improved, the crack propagation trend is predicted through the LSTM model, the early warning threshold value is calculated in combination with the dynamic working condition, and more accurate and reliable crack early warning is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of structural health monitoring, and in particular to a method, system, equipment and medium for early warning of acoustic emission of cracks in wind blades. Background Art

[0002] With the rapid development of wind energy technology, wind power generation has become a vital component of renewable energy. As a crucial component of wind turbine generators, the structural integrity of wind blades plays a critical role in the safe and stable operation of wind power systems. However, over long-term operation, wind blades are susceptible to factors such as wind speed fluctuations, climate, and load fluctuations, leading to material fatigue and the initiation and propagation of microcracks, which in turn affect the safety and service life of the overall structure.

[0003] Currently, monitoring technologies for wind turbine blade cracks primarily include visual inspection, ultrasonic testing, and acoustic emission testing. Acoustic emission testing, as a highly efficient nondestructive testing method, can capture high-frequency acoustic signals generated by crack propagation within the material in real time, offering the advantages of rapid response and high detection sensitivity. However, traditional acoustic emission monitoring systems, in cold environments, suffer from insufficient recognition accuracy due to low temperatures and ice interference. Furthermore, they lack effective crack development trend prediction and dynamic alarm mechanisms, making it difficult to provide timely warnings and address crack issues. Summary of the Invention

[0004] To solve the above technical problems, a wind blade crack acoustic emission early warning method is proposed, including receiving acoustic emission signals and operating condition data collected by a multi-sensor array, preprocessing the acoustic emission signals according to the operating conditions to obtain preprocessed acoustic emission signals; distinguishing ice layer noise from crack signals and shielding ice layer noise interference from the preprocessed acoustic emission signals to obtain effective acoustic emission signals; extracting features from the effective acoustic emission signals to construct a feature vector; inputting the feature vector and operating condition data into a prediction model to output a predicted value of crack extension in the future; calculating a dynamic early warning threshold based on real-time operating conditions, and if the predicted value of crack extension exceeds the dynamic early warning threshold in multiple consecutive time windows, triggering an early warning and locating the crack coordinates; matching the corresponding reinforcement scheme in a knowledge base according to the predicted value of crack extension, and outputting the crack coordinates and reinforcement scheme.

[0005] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the pre-processing step includes performing frequency selective filtering on the acoustic emission signal.

[0006] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the frequency selective filtering frequency band is adjusted according to environmental parameters.

[0007] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the crack signal extraction step includes extracting time domain features and frequency domain features, wherein the time domain features include rise time, decay time and ringing count, and the frequency domain features include energy entropy and frequency band energy distribution.

[0008] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the time domain and frequency domain features are extracted by wavelet transform, and the wavelet transform frequency band is 80kHz to 350kHz.

[0009] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the prediction model adopts a long short-term memory network, the long short-term memory network inputs the crack feature vector and operating condition data for training, and outputs a predicted value of the crack extension amount; the long short-term memory network updates the training data through a sliding time window.

[0010] As a preferred solution of the wind blade crack acoustic emission early warning method described in the present invention, the step of locating the crack coordinates includes establishing a set of hyperbolic equations based on at least three sensor coordinates and the time difference of the received signals; solving the crack coordinates in combination with the sound velocity of the blade material; and converting the blade local coordinate system into WGS84 geographic coordinates based on the GPS coordinates of the blade root installation point and the blade azimuth angle.

[0011] As a preferred solution of the wind blade crack acoustic emission warning system described in the present invention, it is characterized by including: a sensor array, which is deployed in the key areas of the blade main beam, web, and blade root, and collects acoustic emission signals and operating condition data; a preprocessing module, which preprocesses the acoustic emission signals collected by the sensor array according to the operating condition data; an anti-ice interference module, which eliminates ice layer noise interference from the acoustic emission signals processed by the preprocessing module; a feature extraction module, which extracts crack characteristic frequency band energy entropy and time domain characteristics from the acoustic emission signals after eliminating ice layer noise interference; a crack extension amount prediction module, which predicts the crack extension amount in a future period of time according to the crack characteristic frequency band energy entropy and time domain characteristics and operating condition data; an early warning module, which calculates a dynamic early warning threshold based on real-time operating conditions, and compares the crack extension amount prediction value with the dynamic early warning threshold to determine whether to issue an early warning; a crack location module, which locates the crack location; a reinforcement plan generation module, which matches the reinforcement plan according to the crack extension amount prediction value, and outputs the crack location and reinforcement plan.

[0012] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the wind blade crack acoustic emission early warning method when executing the computer program.

[0013] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the wind blade crack acoustic emission early warning method.

[0014] The present invention has the following beneficial effects: It uses a multi-sensor array to collect acoustic emission signals and operating condition data, and then pre-processes them to remove noise interference, ensuring clear and reliable signals. By distinguishing ice noise from crack signals, it effectively shields against ice interference and ensures the accuracy of crack signals.

[0015] Wavelet transform is used to extract crack characteristics, including time domain characteristics and frequency domain characteristics, and combined with the LSTM deep learning model to predict crack extension, realizing dynamic monitoring and early warning of crack development.

[0016] By calculating dynamic warning thresholds based on real-time operating conditions and adjusting them based on historical data, the accuracy and real-time nature of warnings are improved.

[0017] In addition, crack location positioning technology is used to accurately determine the crack location through sensor coordinates and signal time difference, providing support for subsequent reinforcement plans.

[0018] It effectively improves the accuracy and response speed of wind blade crack monitoring, and is particularly suitable for low temperature and ice interference environments, providing a guarantee for the safe operation of wind power generation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 This is an overall flow chart of a wind blade crack acoustic emission early warning method provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0021] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0022] Example 1, with reference to Figure 1 , which is the first embodiment of the present invention, provides a wind blade crack acoustic emission early warning method, comprising:

[0023] Step 1: receiving acoustic emission signals and operating condition data collected by a multi-sensor array, and preprocessing the acoustic emission signals according to the operating conditions to obtain preprocessed acoustic emission signals;

[0024] In step 1, the operating condition data includes ambient temperature, ice thickness, ambient wind speed and blade load data.

[0025] In a preferred embodiment of the present invention, preprocessing the acoustic emission signal includes: performing FIR band-stop filtering on the acoustic emission signal to eliminate blade swing noise below 0.1kHz. When a wind blade is subjected to wind load, it will generate low-frequency swing vibrations of the order of 0.05-0.1Hz. This mechanical vibration will be transmitted to the acoustic emission sensor through the blade structure, forming periodic noise with an amplitude far exceeding the crack signal. The time domain amplitude of the blade swing noise can reach 10-100mV, while the amplitude of the crack acoustic emission signal is usually only 0.1-5mV. If the low-frequency noise is not filtered out, the high-frequency crack signal will be completely submerged.

[0026] FIR band-stop filtering is a preferred solution compared to high-pass filtering, wavelet denoising and other methods in the prior art. Conventional high-pass filtering can easily cause the edge features of the crack signal to be blurred. Although the wavelet denoising method has certain adaptability, the parameter setting depends on experience, and the filtering effect on low-frequency strong interference is unstable. The present invention adopts a fixed-parameter FIR band-stop filter to carry out directional suppression of swing noise below 0.1kHz, which can more effectively maintain the high-frequency crack signal characteristics and is particularly suitable for low-frequency disturbance scenarios of wind turbine blades. In addition, optional but less effective methods include: STFT frequency domain energy filtering, empirical mode decomposition (EMD) denoising, spectral subtraction denoising, etc.

[0027] It should be noted that in a low-temperature environment, gain compensation is performed on the signal amplitude. Low temperature will cause the piezoelectric constant of the piezoelectric ceramic material to decrease, about 0.3%-0.5% per degree Celsius. When the ambient temperature is below 20°C, the electromechanical conversion efficiency of the acoustic emission sensor is significantly reduced, resulting in an approximately linear attenuation of the output signal amplitude as the temperature decreases.

[0028] Therefore, the optimization formula for gain compensation proposed in this embodiment is:

[0029] K=1+0.02×(25+T)

[0030] Where K is the gain compensation coefficient, T is the real-time ambient temperature, and the slope of 0.02 / °C is obtained through experimental calibration of the temperature-sensitivity attenuation curve.

[0031] The gain compensation formula is derived from a linear fit of the measured temperature-sensitivity relationship of the piezoelectric sensor. The principle is that the piezoelectric constant of the acoustic emission sensor decreases at low temperatures, and the signal output amplitude decreases with decreasing temperature. Therefore, compensation is required using the coefficient γ = 1 + k (T0 - T)γ = 1 + k (T0 - T). Here, k is the empirical slope (approximately 0.004 / °C), T0 is the calibration reference temperature (20°C), T is the current temperature, and γ is the gain compensation coefficient.

[0032] For example, when T = -10°C, the gain factor γ = 1 + 0.004 × (20 - (-10)) = 1.12, meaning the signal amplitude needs to be increased by 12%. This slope k is derived from actual sensitivity measurements and regression of piezoelectric materials (such as PZT-5H) at -40°C to 20°C.

[0033] The beneficial effect of this preferred technical solution is that, by combining FIR band-stop filtering with gain compensation technology, it effectively addresses the low-frequency noise interference and low-temperature effects encountered during wind blade crack detection. FIR band-stop filtering accurately eliminates the low-frequency periodic noise caused by blade vibration. This low-frequency noise has an amplitude far higher than the crack signal. If not filtered out, the crack signal would be drowned out, affecting the sensitivity and accuracy of the monitoring system. By filtering out low-frequency noise below 0.1kHz, the high-frequency crack signal is clearly extracted, improving the accuracy and reliability of the monitoring system.

[0034] Furthermore, gain compensation in low-temperature environments effectively overcomes the attenuation effect of ambient temperature changes on the acoustic emission signal amplitude. At low temperatures, the piezoelectric constant of piezoelectric materials decreases, resulting in a decrease in the output signal amplitude. Traditional sensors may not be able to accurately capture crack signals. This solution compensates for the signal attenuation caused by temperature changes through gain compensation, allowing the system to maintain efficient and stable monitoring capabilities in low-temperature environments. The application of this technology significantly improves the ability to monitor wind blade cracks in extreme environments, especially in cold climates and high wind speeds. It ensures real-time and accurate crack warnings, enhancing the safety and reliability of wind power generation systems.

[0035] Step 2: differentiating ice layer noise from crack signals on the preprocessed acoustic emission signals, shielding ice layer noise interference, and obtaining effective acoustic emission signals;

[0036] In step 2, the zero-crossing rate and kurtosis of the preprocessed acoustic emission signal are calculated. If the zero-crossing rate is greater than 1200 times / second and the kurtosis is less than 2.5, it is determined to be ice noise, and the band-stop filter is activated for filtering. The filtering frequency band is 50-200kHz. Otherwise, the crack feature extraction process is entered. Ice layer vibration usually manifests as broadband random vibration with a frequency band of 50-200kHz. The waveform is continuous and the amplitude fluctuation is gentle, resulting in high-frequency zero crossing. The amplitude distribution of ice layer noise is close to Gaussian distribution, and the kurtosis value is low, with a typical value of less than 2.5. The crack extension acoustic emission signal is sudden, with a frequency band of 80-350kHz, and the waveform manifests as a short-time pulse cluster with a low number of zero-point crossings. The crack signal amplitude distribution presents a "peaked and thick-tailed" characteristic, and the kurtosis value is significantly increased, with a typical value greater than 3.5. This can be used to distinguish ice layer noise from crack signals.

[0037] This method offers an improved approach for distinguishing ice noise from crack signals. Traditional methods typically employ frequency domain energy thresholds or image template matching, but suffer from low accuracy and poor adaptability. This invention, for the first time, combines two statistical characteristic indicators: "zero-crossing rate + kurtosis" for differentiation: ice noise has a high zero-crossing rate (>1200 times / s) and low kurtosis (<2.5), while crack signals exhibit the opposite. This differentiation method offers the advantages of strong real-time performance, wide applicability, and high classification accuracy.

[0038] It should be noted that the filtering frequency band of the band-stop filter is dynamically adjusted according to the thickness of the ice layer. When the ice layer thickness is less than 10 mm, the stiffness is low and it is easily stimulated by external stress to produce high-frequency vibrations. It is suitable to monitor its dynamic response through a wide frequency band. At this time, the filtering frequency band is 50-200 kHz. When the ice layer thickness is greater than 10 mm, due to the increase in mass and structural stiffness, it mainly transmits medium-frequency sound waves. At this time, shrinking the frequency band can avoid high-frequency noise interference and match the inherent vibration characteristics of the ice layer. The filtering frequency band is shrunk to 80-150 kHz. After the band-stop filter band is shrunk, the high-frequency components of the ice layer noise, such as ice crystal collisions and wind vibrations, may penetrate the suppression range of the filter. Therefore, the SVM classifier is activated for secondary filtering of the ice layer noise.

[0039] In a preferred embodiment of the present invention, activating an SVM classifier to perform secondary filtering of ice noise includes: extracting the short-time energy ratio and peak-to-average ratio of the pre-processed acoustic emission signal in the 50-200kHz frequency band to generate an input vector, wherein ice noise is mostly continuous vibration or low-frequency fluctuation, and the peak power is close to the average power. Wide-band noise is generated due to ice crystal friction, wind vibration, etc., and the energy is dispersed, so the short-time energy ratio and peak-to-average ratio are low, while crack propagation is accompanied by brittle fracture of the material, generating high-amplitude transient pulses, and the energy released by the expansion is concentrated in a high-frequency narrow band, so the short-time energy ratio and peak-to-average ratio are high, thereby distinguishing ice noise from crack signals; inputting the input vector into the pre-trained SVM classifier and outputting the classification result; if the classification result is ice noise, shielding the current signal and marking it as invalid data, otherwise entering the crack feature extraction process.

[0040] This preferred technical solution has the beneficial effect of accurately distinguishing ice noise from crack signals by calculating the zero-crossing rate and kurtosis of the acoustic emission signal. Ice noise typically manifests as broadband random vibrations with low kurtosis and a high zero-crossing rate, while crack signals exhibit abrupt characteristics with high kurtosis and a low zero-crossing rate. This technical solution effectively filters out ice noise through this determination method, preventing the influence of ice interference on crack signal identification, thereby improving the accuracy and reliability of the system.

[0041] Furthermore, the dynamic adjustment of the band-stop filter, which adaptively adjusts the filter frequency band according to changes in ice thickness, effectively addresses noise interference in varying ice conditions. A wider frequency band is used for thinner ice, while a narrower frequency band is used for thicker ice, thereby adapting to the noise characteristics of varying ice conditions. This dynamic adjustment method better matches the inherent vibration characteristics of the ice layer, improving the adaptability of acoustic emission signal processing.

[0042] When ice noise cannot be completely removed by the band-stop filter, the SVM classifier is activated for secondary filtering, further improving the system's noise suppression capabilities. By analyzing the short-term energy ratio and peak-to-average ratio, the SVM classifier can accurately distinguish ice noise from crack signals, avoiding misjudgments caused by high-frequency ice noise interference.

[0043] Step 3: Extract features from the effective acoustic emission signal and construct a feature vector;

[0044] This method improves upon the previous one by integrating frequency domain (energy entropy) and time domain (rise time, ring count) features, enhancing robustness in complex noise environments. Conventional feature extraction often focuses on frequency domain energy and spectral center of gravity, but ignores temporal characteristics, making it difficult to effectively identify transient burst cracks. Rise time reflects the velocity of the pulse front edge, while ring count reflects the number of decay cycles, both of which are highly correlated with the mechanical properties of crack propagation. This multi-dimensional combination improves the ability to suppress false positives.

[0045] In step 3, a preferred embodiment of the present invention is to perform five-layer db6 wavelet packet decomposition on the effective acoustic emission signal and calculate the energy entropy of the 80-350kHz frequency band. The energy entropy is a quantitative indicator of the energy distribution of the crack characteristic frequency band, reflecting the frequency domain energy concentration characteristics of the crack propagation.

[0046] Extract time domain features, including rise time and ring count;

[0047] Construct the feature vector F = [E, rise time, ring count].

[0048] It should be noted that rise time and ring count solve the misjudgment problem of pure frequency domain methods in complex noise scenarios by quantifying the transient characteristics and oscillation attenuation characteristics of acoustic emission signals. The combined use of the two and energy entropy is essentially to approximate the physical nature of crack propagation from the time domain and frequency domain dimensions, and is a key design to improve the robustness of the early warning system. Energy entropy can quantify the energy distribution of the crack characteristic frequency band, but cannot distinguish between transient cracks and high-frequency narrowband noise. Through rise time and ring count, non-transient or long oscillation signals can be excluded from the waveform morphology, making up for the blind spots of frequency domain characteristics.

[0049] The beneficial effect of this preferred technical solution is that it comprehensively enhances the crack detection capability of acoustic emission signals by combining time-domain and frequency-domain characteristics. First, wavelet packet decomposition effectively extracts energy entropy within the 80-350kHz frequency band, which serves as a quantitative indicator of the energy distribution in the crack's characteristic frequency band. Energy entropy reflects the frequency-domain energy concentration during crack propagation, enabling accurate quantification of the energy characteristics of crack signals, thereby improving crack detection sensitivity.

[0050] At the same time, combining time-domain features such as rise time and ring count can quantify the transient and oscillation decay characteristics of the acoustic emission signal, effectively resolving the misjudgment problem that can occur with purely frequency-domain methods in complex noise environments. Because rise time and ring count reflect the dynamic changes in crack growth, they can exclude non-transient or long oscillation signals, addressing the blind spot where frequency-domain features cannot distinguish transient cracks from high-frequency, narrowband noise.

[0051] By combining time domain features with frequency domain features, not only is the crack signal extraction capability enhanced, but the system’s robustness to noise in complex environments is also improved, enabling the system to more accurately detect and predict crack propagation in adverse environments, effectively improving the accuracy and reliability of the crack warning system.

[0052] In an optional embodiment of the present invention, feature extraction can be achieved using a conventional Fourier transform method. This method converts the acoustic emission signal from the time domain to the frequency domain, extracts the signal's frequency components, and then analyzes the characteristic frequency and energy distribution of the crack signal. In most cases, this method effectively reflects the frequency domain characteristics of the crack signal and provides basic data for subsequent crack detection and analysis. By performing a Fourier transform on the signal, spectral information can be obtained, further identifying the frequency characteristics of the crack signal, thereby improving the reliability of crack monitoring.

[0053] As a conventional signal processing method, Fourier transform can extract the basic features of the signal in the frequency domain, but it has certain limitations when processing wind blade crack signals. First, the Fourier transform assumes that the signal is stationary and cannot effectively capture the non-stationary characteristics of the crack signal, especially the changes in frequency components during crack propagation. Secondly, the frequency resolution of the Fourier transform is low, and it lacks sufficient time-frequency localization analysis capabilities for transient signals (such as sudden crack propagation). In contrast, the preferred embodiment of the present invention combines wavelet transform with time domain features (such as rise time and ringing count), which can not only process non-stationary signals, but also provide higher time-frequency resolution, effectively improve noise suppression capabilities, and avoid the problem of Fourier transform misjudgment in complex noise environments. Therefore, the method of the present invention has stronger robustness and accuracy, and can accurately monitor crack propagation in complex environments.

[0054] Step 4: Input the characteristic vector and working condition data into the prediction model, and output the predicted value of crack extension in the future period;

[0055] Compared with BP neural network, SVR, RF and other models, LSTM can retain time dependency and avoid the problem of vanishing gradient in long sequences. It is particularly suitable for trend modeling scenarios where crack propagation has "historical dependency + nonlinear change".

[0056] In step 4, an LSTM network model is constructed. The input layer receives the feature vector F and the working condition data. The hidden layer contains 128 neurons. The output layer is the predicted value of the crack extension in the next three weeks.

[0057] The LSTM model is trained based on historical crack growth data, with an input data time span of twelve weeks and a sliding time window step size of one week.

[0058] In an optional embodiment of the present invention, the prediction model can also select the support vector regression method as an alternative prediction model. This method predicts the crack extension by finding a hyperplane to maximize the interval of prediction error. This method is suitable for nonlinear regression problems and can effectively handle complex, nonlinear data relationships. SVM regression can use operating condition data and crack characteristics to output a prediction result of crack extension. In particular, when crack development has more complex dynamic changes, SVR can provide more robust prediction performance.

[0059] Decision trees are also a common prediction method. They use a tree-like structure to divide input data into multiple levels, thereby predicting crack growth. Decision trees can handle both categorical and continuous variables and generate rule sets through training to predict crack growth trends. They are suitable for situations with clear data structures and strong nonlinear relationships.

[0060] Step 5: Calculate the dynamic warning threshold based on the real-time working conditions. If the predicted value of the crack extension exceeds the dynamic warning threshold in multiple consecutive time windows, trigger the warning and locate the crack coordinates.

[0061] For example: define the time window as one day. If the predicted values ​​for three days are 2.1mm, 2.3mm, and 1.9mm respectively, and the dynamic thresholds are 1.9mm, 1.9mm, and 1.8mm respectively, an alarm will be triggered.

[0062] In a preferred embodiment of the present invention, the calculation formula of the dynamic warning threshold is:

[0063]

[0064] Among them, C is a constant, representing the basic threshold, V and V max Indicates the real-time wind speed and the maximum wind speed that the blade can withstand, L and L max They represent the real-time load and the maximum load that the blade can withstand, respectively. α and β are the wind speed sensitivity coefficient and load sensitivity coefficient, respectively, which are determined by the regression analysis of historical crack growth data and corresponding working conditions.

[0065] The dynamic threshold construction mechanism is a significant improvement over the traditional static early warning mechanism. Conventional methods set alarm thresholds based on fixed crack size or growth rate, ignoring actual variations in wind speed and load conditions, and thus presenting a risk of false alarms. This invention constructs a normalized function model of wind speed and load, and introduces a sensitivity coefficient for dynamic correction. For example, when the wind speed reaches 80% of the maximum wind speed and the load reaches 90% of the maximum load, the threshold increases according to the formula, avoiding unnecessary false alarms and effectively improving the practicality and reliability of the early warning system.

[0066] The beneficial effect of this preferred technical solution is that, by calculating the dynamic warning threshold, combined with the real-time wind speed, blade load and its maximum load-bearing capacity, the warning standard can be adaptively adjusted according to changes in working conditions. This dynamic adjustment mechanism can optimize the warning threshold in real time according to the actual changes in wind speed and load, avoiding the problem of false alarms or missed alarms caused by fixed thresholds. When the predicted value of the crack extension exceeds the dynamic warning threshold, an alarm is triggered, and the crack is further located and processed. By introducing the wind speed sensitivity coefficient and the load sensitivity coefficient, the method performs regression analysis based on historical data, thereby improving the accuracy of the warning and being able to better cope with complex and changing working conditions. The introduction of dynamic thresholds makes the monitoring system of wind blades more intelligent and flexible, and effectively improves the reliability of prediction and warning.

[0067] In a preferred embodiment of the present invention, the crack coordinates are located, and the specific steps include:

[0068] Establishing a set of hyperbolic equations based on at least three sensor coordinates and time differences of received signals;

[0069] Combined blade material sound velocity The crack coordinates are solved by the least square method, where E is the elastic modulus and ρ is the density;

[0070] The blade local coordinate system is converted into WGS84 geographic coordinates based on the GPS coordinates of the blade root installation point and the blade azimuth.

[0071] The beneficial effect of this preferred technical solution is that a set of hyperbolic equations is established by using the time difference of at least three sensor coordinates and received signals to achieve high-precision positioning of the crack position. This method improves the accuracy of crack position estimation by collecting the signal time difference of multiple sensors and using geometric positioning technology, avoiding the error or positioning deviation that may be caused by a single sensor. Combined with the sound velocity, elastic modulus and density parameters of the blade material, the solution of the crack position is further optimized by the least squares method, thereby improving the accuracy of the algorithm. Using GPS coordinates and azimuth angles to convert the crack position from a local coordinate system to WGS84 geographic coordinates can provide maintenance personnel with accurate geographic location information, making subsequent crack reinforcement and repair work more accurate and efficient. This step not only reduces the time and cost of maintenance and reinforcement, but also effectively improves the overall safety and stability of the wind power generation system.

[0072] Step 6: Match the corresponding reinforcement scheme in the knowledge base according to the predicted value of crack extension, and output the crack coordinates and reinforcement scheme.

[0073] Traditional monitoring systems mostly remain at the anomaly identification and alarm stage, lacking subsequent response strategies. This invention innovatively introduces a knowledge-based automatic reinforcement recommendation module that automatically matches the appropriate reinforcement method (such as fiber coating, resin injection, etc.) based on the predicted crack growth value and location, and generates a geographic coordinate recommendation report. This module significantly improves the system's practicality and intelligence, enabling maintenance personnel to quickly respond and develop optimal intervention plans.

[0074] Example 2 is the second embodiment of the present invention, which is different from the previous embodiment in that:

[0075] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or the part of the technical solution, can be implemented in the form of software functional units and sold or used as independent products.

[0076] The computer software product is embodied in the form of a software product stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0077] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0078] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0079] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, it can be implemented using a combination of any of the following technologies known in the art: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0080] Embodiment 3, the third embodiment of the present invention, provides a wind blade crack acoustic emission early warning system, comprising:

[0081] Sensor arrays are deployed in key areas of the blade's main beam, web, and root to collect acoustic emission signals and operating condition data;

[0082] A preprocessing module preprocesses the sound emission signals collected by the sensor array according to the working condition data;

[0083] The anti-ice interference module eliminates ice noise interference on the acoustic emission signal processed by the pre-processing module;

[0084] Feature extraction module, which extracts the crack characteristic frequency band energy entropy and time domain characteristics from the acoustic emission signal after eliminating ice layer noise interference;

[0085] The crack extension prediction module predicts the crack extension in the future based on the crack characteristic frequency band energy entropy, time domain characteristics, and working condition data;

[0086] The early warning module calculates the dynamic early warning threshold based on the real-time working conditions, and compares the predicted value of crack extension with the dynamic early warning threshold to determine whether to issue an early warning;

[0087] Crack location module, locates the crack position;

[0088] The reinforcement scheme generation module matches the reinforcement scheme according to the predicted value of crack extension and outputs the crack location and reinforcement scheme.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A wind blade crack acoustic emission early warning method, characterized by: include, Receiving acoustic emission signals and operating condition data collected by a multi-sensor array, and preprocessing the acoustic emission signals according to the operating conditions to obtain preprocessed acoustic emission signals; Distinguishing ice layer noise from crack signals and shielding ice layer noise interference from the preprocessed acoustic emission signals to obtain effective acoustic emission signals; Extracting features from the effective acoustic emission signal and constructing a feature vector; Inputting the characteristic vector and working condition data into a prediction model, and outputting a predicted value of crack extension in a future period of time; The dynamic warning threshold is calculated based on the real-time working conditions. If the predicted value of the crack extension exceeds the dynamic warning threshold in multiple consecutive time windows, an early warning is triggered and the crack coordinates are located. According to the predicted value of crack extension, the corresponding reinforcement scheme in the knowledge base is matched and the crack coordinates and reinforcement scheme are output.

2. The wind blade crack acoustic emission early warning method according to claim 1, characterized in that: The pre-processing step includes performing frequency selective filtering on the acoustic emission signal.

3. The wind blade crack acoustic emission early warning method according to claim 2, characterized in that: The frequency selective filtering frequency band is adjusted according to environmental parameters.

4. The wind blade crack acoustic emission early warning method according to claim 3, characterized in that: The crack signal extraction step includes extracting time domain features and frequency domain features, wherein the time domain features include rise time, decay time and ringing count, and the frequency domain features include energy entropy and frequency band energy distribution.

5. The wind blade crack acoustic emission early warning method according to claim 4, characterized in that: The time domain and frequency domain features are extracted by wavelet transform, and the wavelet transform frequency band is 80kHz to 350kHz.

6. The wind blade crack acoustic emission early warning method according to claim 5, characterized in that: The prediction model adopts a long short-term memory network, which inputs crack feature vectors and working condition data for training and outputs a predicted value of crack extension; The long short-term memory network updates training data through a sliding time window.

7. The wind blade crack acoustic emission early warning method according to claim 6, characterized in that: The step of locating the crack coordinates includes: Establishing a set of hyperbolic equations based on at least three sensor coordinates and time differences of received signals; Solve the crack coordinates by combining the blade material sound velocity; The blade local coordinate system is converted into WGS84 geographic coordinates based on the GPS coordinates of the blade root installation point and the blade azimuth.

8. A wind blade crack acoustic emission early warning system, using the wind blade crack acoustic emission early warning method according to any one of claims 1 to 7, characterized in that: include: Sensor arrays are deployed in key areas of the blade's main beam, web, and root to collect acoustic emission signals and operating condition data; A preprocessing module preprocesses the sound emission signals collected by the sensor array according to the working condition data; The anti-ice interference module eliminates ice noise interference on the acoustic emission signal processed by the pre-processing module; Feature extraction module, which extracts the crack characteristic frequency band energy entropy and time domain characteristics from the acoustic emission signal after eliminating ice layer noise interference; The crack extension prediction module predicts the crack extension in the future based on the crack characteristic frequency band energy entropy, time domain characteristics, and working condition data; The early warning module calculates the dynamic early warning threshold based on the real-time working conditions, and compares the predicted value of crack extension with the dynamic early warning threshold to determine whether to issue an early warning; Crack location module, locates the crack position; The reinforcement scheme generation module matches the reinforcement scheme according to the predicted value of crack extension and outputs the crack location and reinforcement scheme.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the wind blade crack acoustic emission early warning method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind blade crack acoustic emission early warning method according to any one of claims 1 to 7 are implemented.

Images