A stress wave based sensing method and system applied to wind turbine blades

Abstract

The application belongs to the technical field of equipment monitoring, and discloses a measurement method and system based on stress wave sensing applied to a wind power blade; the method comprises the following steps: collecting stress wave signal data and preprocessing the stress wave signal data, obtaining perfect stress wave signal data by successively performing bipolar collaborative decomposition, error correction and environmental interference suppression on the stress wave signal data; performing feature extraction on the perfect stress wave signal data, generating an intermediate feature matrix, further performing feature screening on the intermediate feature matrix, and outputting stress wave signal feature data; constructing a damage assessment model based on the stress wave signal feature data, using the damage assessment model to assess the damage of the wind power blade, and obtaining damage assessment records; generating a three-dimensional damage atlas based on the damage assessment records, and sending the three-dimensional damage atlas to a preset wind power equipment safety terminal; and the wind power blade is monitored by using stress waves.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment monitoring, more particularly, the present application relates to a stress wave sensing based measurement method and system applied to a wind turbine blade. BACKGROUND

[0002] The patent with the application publication number CN117074215A discloses a damage analysis method based on stress wave segmentation. By segmenting the stress wave, the stress wave obtained through one impact can analyze the damage inside the sample or structure, avoiding the need for additional damage testing after the structure is impacted, improving the timeliness of damage identification. It solves the problem of traditional damage testing, which requires other equipment to detect the damage of the sample, causing different degrees of disturbance to the sample during the process, resulting in errors in the experiment. It can test and analyze the dynamic impact damage evolution and its amplitude and frequency spectrum attenuation law of solid materials such as rock and concrete under in-situ pressure preservation. It fills the gap in the dynamic damage testing of materials under in-situ pressure preservation in the existing technical methods based on the Hopkinson bar.

[0003] With the development of wind turbine generators, the health status of wind turbine blades, as the core components, directly affects the operation of the entire wind turbine generator. In order to monitor the health status of wind turbine blades, stress waves are used to detect damage. The stress wave data collected by the sensor need to be finely preprocessed, otherwise errors may occur in the stress wave propagation process due to noise and environmental factors, resulting in inaccurate collected data. When extracting features from the collected data, the general method lacks joint analysis of time domain and frequency domain, and some features may be easily ignored. When constructing a model for damage assessment, the model cannot learn new features, resulting in reduced recognition accuracy and poor assessment effect. The assessment results output by the model are not intuitive, which is not conducive to technical personnel to view.

[0004] In view of this, the present application provides a stress wave sensing based measurement method and system applied to a wind turbine blade to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a stress wave sensing based measurement method applied to a wind turbine blade, comprising:

[0006] S1. Collecting stress wave signal data and preprocessing the stress wave signal data, the stress wave signal data is decomposed by bipolar cooperation, error correction and environmental disturbance suppression in sequence to obtain perfect stress wave signal data;

[0007] S2. Feature extraction is performed on the improved stress wave signal data to generate an intermediate feature matrix, and the intermediate feature matrix is further screened to output stress wave signal feature data;

[0008] S3. A damage assessment model is constructed based on the stress wave signal feature data, and the damage assessment model is used to assess the damage of the wind turbine blade to obtain a damage assessment record;

[0009] S4. A three-dimensional damage atlas is generated based on the damage assessment record, and the three-dimensional damage atlas is sent to a preset wind power equipment safety terminal.

[0010] Further, sensors are symmetrically arranged on the main beam, leading edge and trailing edge of each wind turbine blade, and the interval between any two sensors is less than or equal to a preset distance threshold. A stress wave generator is used to emit a stress wave, and a sensor is used to collect stress wave signal data, wherein the stress wave signal data includes time domain waveform signals and stress wave propagation time difference.

[0011] Further, the stress wave signal data is preprocessed in the following manner:

[0012] The time domain waveform signal is converted into a frequency domain spectrum signal using a fast Fourier transform algorithm. The frequency domain spectrum signal is subjected to bipolar collaborative decomposition, and the stress wave propagation speed is calculated based on the stress wave propagation time difference. The bipolar collaborative decomposition results, stress wave propagation time difference and stress wave propagation speed are integrated to obtain multi-modal decomposition data. A propagation dynamic compensation mechanism is designed to correct the multi-modal decomposition data error to generate propagation correction data. The propagation correction data is subjected to environmental interference suppression based on the collected multi-physical field parameters to obtain improved stress wave signal data.

[0013] The bipolar collaborative decomposition of the frequency domain spectrum signal includes:

[0014] The frequency domain spectrum signal is decomposed by N layers using a pre-selected wavelet packet basis function to generate multi-scale spectrum subbands and ensure that all generated spectrum subbands can cover the full frequency band; the energy proportion of each spectrum subband is calculated and a subband energy threshold is set; when the energy proportion of any spectrum subband is greater than or equal to the preset subband energy threshold, the spectrum subband is determined as an effective subband, otherwise the spectrum subband is determined as a noise subband; the effective subbands are filtered by an adaptive soft threshold to obtain effective high-frequency subbands; the noise subbands are decomposed by empirical mode decomposition to generate M intrinsic mode functions; the stability index of each intrinsic mode function is calculated, and the intrinsic mode component with a stability index less than a preset stability index threshold, i.e. an effective mode component, is retained; the energy of the intrinsic mode component with a stability index greater than or equal to the preset stability index threshold is extracted by using an energy feedback algorithm, and the energy is superimposed on the adjacent effective mode component; the effective high-frequency subbands and the effective mode components are synchronously extracted and transformed to construct a time-frequency feature matrix, and the time-frequency feature matrix, the stress wave propagation time difference and the stress wave propagation speed are integrated to obtain multi-modal decomposition data.

[0015] Further, the method for correcting the multi-modal decomposition data error comprises:

[0016] The basic propagation speed of the stress wave in the wind turbine blade and the material coefficient are obtained by querying a preset wind turbine blade parameter database; a wave speed function is constructed based on the basic propagation speed and the material coefficient; the actual propagation speed of the stress wave in the wind turbine blade is obtained by calculating the function value of the wave speed function;

[0017] The time delay sequence is obtained by sorting the stress wave propagation time difference in ascending order based on the timestamp, and the time delay mean value of the time delay sequence is calculated; the energy intensity distribution of each subband is extracted from the multi-modal decomposition data; the wave speed correction equation is constructed based on the time delay mean value and the energy intensity distribution of each subband; the coupling coefficient of the wave speed correction equation is updated using a pre-trained LSTM network model, and the actual propagation speed of the stress wave in the wind turbine blade is adjusted based on the updated wave speed correction equation; the attenuation compensation function is constructed and the subband energy attenuation in the multi-modal decomposition data is corrected based on the function; the multi-modal decomposition data is updated based on the adjusted actual propagation speed of the stress wave and the corrected subband energy;

[0018] The method for suppressing environmental interference on the propagation correction data comprises:

[0019] Physical parameter sensors are arranged on each wind power blade to collect temperature, stress and humidity data of the wind power blade; a multi-physical field coupling factor is constructed based on the temperature, stress and humidity data of the wind power blade; a self-adaptive band-stop filter is designed and used to suppress high-frequency noise caused by temperature drift in the frequency domain dimension; a metasurface beamforming method is used to generate beam nulls in the interference direction; and a multi-objective genetic algorithm is used to dynamically adjust the weights of each physical parameter in the multi-physical field coupling factor and the stop-band parameters of the self-adaptive band-stop filter.

[0020] Further, the way of extracting features from the improved stress wave signal data comprises:

[0021] The improved stress wave signal data is classified based on data types, including sub-band energy intensity data, dynamic propagation data and time-frequency data; the sub-band energy intensity data and the time-frequency data are integrated into a time-frequency feature data set, and the dynamic propagation data is taken as a dynamic propagation feature data set; the time-frequency feature data set and the dynamic propagation feature data set from P sensors are stacked into a three-dimensional tensor, time sampling points C and feature dimensions D in any one preset time window are cross combined to obtain an initial feature tensor with a dimension of PxCxD in each time window; the initial feature tensor in each time window is subjected to a cooperative process of time domain synchronization calibration and frequency domain phase alignment to obtain a calibrated feature tensor; the position of a damage area of the wind power blade is calculated based on the time-frequency feature data set; the Euclidean distance between the position of the damage area of the wind power blade and any one sensor is calculated, and a signal-to-noise ratio index of each sensor is calculated, and each sensor is assigned a weight based on the signal-to-noise ratio index of the sensor and the Euclidean distance between the sensor and the position of the damage area of the wind power blade, i.e., the calibrated feature tensor is weighted to obtain an intermediate feature matrix.

[0022] Further, the way of further feature screening of the intermediate feature matrix comprises:

[0023] The local density of each feature point in the intermediate feature matrix is calculated, and all feature points in the intermediate feature matrix are clustered based on the local density of each feature point, so that the intermediate feature matrix is divided into two sub-matrices, i.e., a high-density matrix and a low-density matrix; an adaptive kernel function is constructed to map the high-density matrix and the low-density matrix to a high-dimensional space to obtain a high-dimensional space feature matrix.

[0024] The classification model is constructed, a random forest model is taken as a basic structure of the classification model, historical wind turbine blade damage data is collected, and the historical wind turbine blade damage data is converted into a historical feature matrix; key damage features involved in damage conditions corresponding to the historical feature matrix are taken as training labels of the classification model by querying a preset wind turbine blade parameter database; the classification model is trained by using the historical feature matrix until a function value of a loss function of the classification model no longer decreases, and a trained classification model is obtained; key features are screened from a high-dimensional space feature matrix by using the classification model, and a Gini index of each feature point in the high-dimensional space feature matrix is calculated; all feature points are sorted in descending order based on the Gini index, and feature points greater than a preset sorting threshold are determined as key feature points; redundant feature points in the key feature points are removed by using a mutual information analysis algorithm, and a damage sensitive feature set is output.

[0025] The damage sensitive feature set is standardized to obtain a standardized feature set; principal component analysis algorithm is used to reduce dimension compression on the standardized feature set to obtain stress wave signal feature data.

[0026] Further, the way of constructing the damage evaluation model comprises:

[0027] A damage coding system is constructed, which uses hierarchical coding rules; all specific damage conditions of the wind turbine blade and corresponding features of the damage conditions are determined by querying a preset wind turbine blade parameter database; all specific damage conditions of the wind turbine blade are coded based on the damage coding system, and the code of any specific damage condition is combined with the corresponding feature of the damage condition to form a feature-coding mapping matrix; all feature-coding mapping matrices are integrated to obtain a feature-coding mapping dataset;

[0028] A damage evaluation model is constructed, and a residual neural network model is taken as a basic structure of the damage evaluation model, including an input layer, a full connection layer, a feature processing layer, a damage quantization layer and an output layer; the stress wave signal feature data is taken as input data of the damage evaluation model; a double-channel parallel processing architecture is taken as a basic framework of the feature processing layer of the damage evaluation model, including a time-frequency feature channel and a spatial propagation channel; the input data is jointly processed by the time-frequency feature channel and the spatial propagation channel to generate an intermediate feature tensor; the intermediate feature tensor enters the damage quantization layer to calculate a damage index, and the damage index, a damage position coordinate and a damage type are output by the output layer; the damage index, the damage position coordinate and the damage type are matched to generate a damage code; the existing damage evaluation model is dynamically updated after each output of the damage code; real-time time stamps are added to all output damage codes, and the damage codes are sorted in chronological order to obtain damage evaluation records.

[0029] Further, the way of dynamically updating the existing damage evaluation model comprises:

[0030] The incremental learning architecture is constructed, a dynamic feature memory bank is established after the feature processing layer of the damage assessment model, the intermediate feature tensor is backed up, and feature extraction is performed on the backed-up intermediate feature tensor to obtain a historical damage feature dataset; a double-model collaborative distillation mechanism is introduced in the incremental learning architecture, including a first model and a second model; the feature processing layer of the damage assessment model is frozen using the first model, and a probability tensor of the historical damage feature dataset is output; the second model adjusts the parameters of the unfrozen layer based on the adjustment function constructed based on the probability tensor; the initial learning rate of the incremental learning architecture is set, and the initial learning rate is adjusted based on the cosine annealing algorithm; the parameter constraint is set for the damage quantization layer to ensure that the parameter offset is less than the preset offset threshold; the performance indicators of the damage assessment model after adding the incremental learning architecture are calculated, and if each performance indicator meets the expected value, the damage assessment model is updated, otherwise the damage assessment model rollback is triggered.

[0031] Further, the method for generating a three-dimensional damage atlas based on the damage assessment record comprises:

[0032] The damage position coordinates, damage index and damage type are extracted from the damage assessment record, and a damage spatial distribution matrix is constructed; the damage spatial distribution matrix is subjected to three-dimensional gridding processing by using the Kriging spatial interpolation algorithm to obtain a wind turbine blade damage global density field; a parameterized modeling of the wind turbine blade is performed based on the geometric parameters of the wind turbine blade to obtain a wind turbine blade three-dimensional model; the wind turbine blade damage global density field is mapped to the wind turbine blade three-dimensional model to obtain a three-dimensional damage model; the three-dimensional damage model is rendered by using color to obtain a damage rendering model; the damage rendering model corresponding to each timestamp is sorted in chronological order to obtain a three-dimensional damage atlas.

[0033] A stress wave sensing-based measurement system applied to a wind turbine blade is used to implement a stress wave sensing-based measurement method applied to a wind turbine blade, comprising:

[0034] A data acquisition module is configured to acquire stress wave signal data and pre-process the stress wave signal data to obtain improved stress wave signal data.

[0035] A feature extraction module is configured to perform feature extraction on the improved stress wave signal data and output stress wave signal feature data.

[0036] A damage assessment module is configured to construct a damage assessment model based on the stress wave signal feature data, perform damage assessment on the wind turbine blade by using the damage assessment model, and obtain a damage assessment record.

[0037] The visualization generation module generates a three-dimensional damage atlas based on the damage assessment record and sends the three-dimensional damage atlas to a preset wind power equipment safety terminal; and the modules are connected through wired and / or wireless modes.

[0038] The technical effect and advantage of the application of the stress wave sensing based measurement method and system to the wind power blade are as follows:

[0039] The stress wave signal data is collected, preprocessed and feature extracted to obtain stress wave signal feature data, and a model is constructed based on the data to assess the damage of the wind power blade, and finally a visual three-dimensional damage atlas is generated based on the damage assessment record, realizing the process of monitoring the wind power blade based on the stress wave; compared with the existing experience, the preprocessing of the stress wave signal data is more accurate, and the noise interference in the data, the error in the propagation process and the influence of the environmental factors are considered; the joint feature extraction is performed from the time domain and the frequency domain, the features related to the damage of the wind power blade are obtained, the classification model is constructed to further screen the features, the key features representing the damage condition are obtained and are compressed by dimensionality reduction, and the processing efficiency is improved; the damage assessment model is constructed and is dynamically updated based on the incremental learning architecture, and the model is used to identify and assess the damage condition, improving the adaptability and processing accuracy of the model; finally, the visual damage atlas is generated and is sent to the preset wind power equipment safety terminal, and the corresponding measures are taken by the technical personnel based on the three-dimensional damage atlas to repair the wind power blade. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The application of the stress wave sensing based measurement method to the wind power blade is shown in the schematic diagram.

[0041] Figure 2 The application of the stress wave sensing based measurement system to the wind power blade is shown in the schematic diagram. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0043] Embodiment 1

[0044] Please refer to Figure 1 The application of the stress wave sensing based measurement method to the wind power blade is shown in the schematic diagram.

[0045] S1. Collecting stress wave signal data and preprocessing the stress wave signal data, obtaining perfect stress wave signal data by successively performing bipolar collaborative decomposition, error correction and environmental interference suppression on the stress wave signal data;

[0046] S2. Feature extraction is performed on the perfect stress wave signal data to generate an intermediate feature matrix, and the intermediate feature matrix is further screened for features to output stress wave signal feature data;

[0047] S3. Constructing a damage assessment model based on the stress wave signal feature data, and using the damage assessment model to assess the damage of the wind turbine blade to obtain a damage assessment record;

[0048] S4. Generating a three-dimensional damage map based on the damage assessment record, and sending the three-dimensional damage map to a pre-set wind power equipment safety terminal.

[0049] In each wind turbine blade, sensors are symmetrically arranged on the main beam, leading edge and trailing edge, and the distance between any two sensors is less than or equal to a pre-set distance threshold (in this embodiment, the distance threshold is 30 cm); a stress wave generator (a device for generating controllable stress pulses. In this embodiment, electromagnetic force is used to excite stress waves) is used to emit stress waves, and stress wave signal data is collected by the sensors, the stress wave signal data including time domain waveform signals and stress wave propagation time differences (each sampling point in the time domain waveform signal contains a time stamp corresponding to the collection time of each sampling point; the stress wave propagation time difference is calculated based on the time stamp difference of multiple sensors, for example, the time stamp of sensor A1 receiving a time domain waveform signal is time1, and the time stamp of sensor A2 receiving the same time domain waveform signal is time2; the absolute value of the difference between time1 and time2 is the stress wave propagation time difference).

[0050] Even if high-precision instruments are used to collect stress wave signal data during the data collection stage, preprocessing is an indispensable step in order to obtain more accurate data in subsequent processing of stress wave data; the sensor used in this embodiment is a piezoelectric ceramic sensor, and for piezoelectric ceramic sensors, the dielectric constant of the piezoelectric material is easily affected by environmental factors, resulting in noise in the received data or a decrease in accuracy, for example, temperature and electromagnetic interference in the data collection site; Therefore, it is necessary to perform denoising and environmental interference suppression on the collected data during preprocessing, so as to maximize the accuracy of the stress wave data used in this embodiment, and indirectly improve the accuracy and efficiency of subsequent operations.

[0051] The preprocessing method for stress wave signal data includes:

[0052] The time-domain waveform signal is converted into a frequency-domain spectrum signal by using a fast Fourier transform algorithm (the Fourier transform algorithm is a commonly used algorithm for converting a waveform signal from the time domain to the frequency domain; compared with other types of Fourier transform algorithms, the fast Fourier transform algorithm has lower complexity, occupies less computing resources, and has extremely high precision and conversion efficiency; the time characteristics of the frequency-domain spectrum signal converted from the time-domain waveform signal strictly correspond to the original waveform, that is, the time stamps of the two signals completely correspond); the frequency-domain spectrum signal is subjected to bipolar collaborative decomposition, and the stress wave propagation velocity is calculated based on the stress wave propagation time difference (the distance between any two sensors is measured, and the stress wave propagation velocity is calculated in combination with the stress wave propagation time difference of the two sensors at a certain time stamp), and the bipolar collaborative decomposition result, the stress wave propagation time difference and the stress wave propagation velocity are integrated to obtain multi-modal decomposition data; a propagation dynamic compensation mechanism is designed to correct the multi-modal decomposition data error, and propagation correction data is generated; the propagation correction data is subjected to environmental interference suppression based on the collected multi-physical field parameters, and perfect stress wave signal data is obtained (the bipolar collaborative decomposition is used for denoising the original data, the propagation dynamic compensation mechanism is used for processing the error of the denoised data, and the environmental interference suppression is used for reducing the influence of environmental factors).

[0053] The bipolar collaborative decomposition refers to collaborative signal decomposition by using a wavelet packet basis function and empirical mode decomposition, wherein the wavelet packet basis function is used for processing high-frequency signals, and the empirical mode decomposition is used for processing low-frequency signals, and the combination of the two ensures complete processing of the frequency-domain spectrum signal, and the energy dispersion problem is solved by using a synchronous extraction transform algorithm, and finally high-quality multi-modal decomposition data is obtained.

[0054] The bipolar collaborative decomposition of the frequency-domain spectrum signal includes:

[0055] The frequency-domain spectrum signal is subjected to N-layer decomposition (N=5 in this embodiment) by using a pre-selected wavelet packet basis function (for example, a Daubechies5 wavelet basis function) to generate multi-scale spectrum subbands and ensure that all generated spectrum subbands can cover the full frequency band (in this embodiment, the full frequency band refers to the frequency interval [1kHz, 50MHz]; when a stress wave is used for device monitoring, the sensitive frequency band of the device damage belongs to the above frequency interval, and therefore full-band coverage is implemented in order to achieve comprehensive perception); the energy proportion of each spectrum subband is calculated, and a subband energy threshold is set; the calculation formula of the energy proportion of each spectrum subband is as follows: wherein E represents the energy proportion of any spectrum subband; H j,k(t) represents the wavelet packet coefficient of the kth spectral sub-band in the jth layer with a timestamp of t; when the energy proportion of any spectral sub-band is greater than or equal to a preset sub-band energy threshold, the spectral sub-band is determined as an effective sub-band, otherwise, the spectral sub-band is determined as a noise sub-band; the effective sub-band is screened based on the sub-band energy threshold, and the noise sub-band is separated (since the energy proportion of the sub-band with a low frequency is small, and noise is generally distributed in the low-frequency sub-band, the sub-band with an energy proportion less than the preset sub-band energy threshold is determined as a noise sub-band); the effective sub-band is subjected to adaptive soft threshold filtering to obtain an effective high-frequency sub-band; the calculation formula of the adaptive soft threshold filtering is: wherein, represents the wavelet packet coefficient corresponding to the effective high-frequency sub-band; λ j represents a threshold coefficient that dynamically attenuates with the decomposition layer number j, and in the embodiment, λ = 0.8; σ j represents the noise standard deviation of the jth layer sub-band; the noise sub-band is subjected to empirical mode decomposition to generate M intrinsic mode functions (in the embodiment, M = 10); the stability index of each intrinsic mode function is calculated, and the calculation formula of the stability index is: wherein, N0 represents the number of sampling points in a window (all sampling points in each intrinsic mode function are analyzed through a sliding window); f n0 represents the instantaneous frequency of the nth0 sampling point in the window; the intrinsic mode component with a stability index less than a preset stability index threshold is reserved, that is, an effective mode component; the energy of the intrinsic mode component with a stability index greater than or equal to the preset stability index threshold is extracted by using an energy back injection algorithm, and the energy is superimposed on the adjacent effective mode component (the energy in the intrinsic mode component that is not reserved is superimposed on the adjacent effective mode component, avoiding energy dispersion and information loss); the effective high-frequency sub-band and the effective mode component are subjected to synchronous extraction transformation and construction of a time-frequency feature matrix (synchronous extraction transformation is a signal processing method combining time-frequency analysis and mode decomposition, which is used to extract high-resolution instantaneous frequency and construct a time-frequency feature matrix; the data processing accuracy is extremely high and the mode aliasing is avoided); the time-frequency feature matrix, the stress wave propagation time difference and the stress wave propagation speed are integrated to obtain the multi-modal decomposition data (wherein the time-frequency feature matrix refers to a feature matrix including the three-dimensional information of the timestamp, frequency and energy intensity distribution of any sub-band in the multi-modal decomposition data).

[0056] The stress wave propagation speed changes due to different materials of the wind turbine blade and different energy distribution of the sub-bands of each frequency in the multi-modal decomposition data; the energy of the stress wave is easily attenuated due to medium absorption, scattering and path deformation and other factors during the propagation of the stress wave, and in order to ensure the accuracy of subsequent operations, the energy attenuation needs to be corrected; therefore, the wave speed correction equation and the attenuation compensation function are respectively constructed to correct the errors in the multi-modal decomposition data.

[0057] The method for correcting the errors in the multi-modal decomposition data includes:

[0058] The basic propagation speed of the stress wave in the wind turbine blade and the material coefficient (the material coefficient is different for different materials of the wind turbine blade, and the propagation speed of the stress wave in the wind turbine blade with different materials is also different) are obtained by querying the preset wind turbine blade parameter database; the wave speed function is constructed based on the basic propagation speed and the material coefficient; the calculation formula of the wave speed function is: Wherein, V(f) represents the actual propagation speed of the stress wave frequency band with frequency f in the wind turbine blade with material coefficient a; V0 represents the basic propagation speed of the stress wave in the wind turbine blade with material coefficient a; f1 represents the average frequency of the stress wave; the value of the material coefficient is determined by the specific condition of the wind turbine blade material; the actual propagation speed of the stress wave in the wind turbine blade is obtained by calculating the function value of the wave speed function.

[0059] The time delay sequence is obtained by sorting the time delay of the stress wave in ascending order based on the time stamp, and the time delay mean of the time delay sequence is calculated; the energy intensity distribution of each sub-band is extracted from the multi-modal decomposition data; the wave speed correction equation is constructed based on the time delay mean and the energy intensity distribution of each sub-band; the calculation formula of the wave speed correction equation is: Wherein, represents the adjusted actual propagation speed; Δt represents the time delay mean; E1 represents the energy proportion of any sub-band; β represents the coupling coefficient; the coupling coefficient of the wave speed correction equation is updated by using the pre-trained LSTM network model (the LSTM network model is trained by combining the historical data collected and the enhanced samples generated by online simulation, and the energy intensity distribution of the stress wave frequency band with frequency f and the time delay sequence are used as input data to dynamically update the coupling coefficient of the wave speed correction equation, so as to ensure the timeliness of the wave speed correction equation); and the actual propagation speed of the stress wave in the wind turbine blade is adjusted based on the updated wave speed correction equation (the original actual propagation speed is updated to the adjusted actual speed calculated by the wave speed correction equation); the attenuation compensation function is constructed, and the sub-band energy attenuation in the multi-modal decomposition data is corrected based on the function; the calculation formula of the attenuation compensation function is: Wherein R(f2) represents the compensated frequency attenuation value of the stress wave frequency band with a frequency of f2; γ0(f2) represents the initial attenuation compensation coefficient of the stress wave frequency band with a frequency of f2 (the coefficient reflects the inherent attenuation characteristics of the medium, which is related to the medium material and environmental factors); a0 represents an adjustment factor (in this embodiment, a0 = 0.2); E2(f2) represents the energy intensity of the stress wave frequency band with a frequency of f2; Ew(f2) represents the standard energy intensity of the stress wave frequency band with a frequency of f2 obtained by querying the preset wind turbine blade parameter database; and the multi-modal decomposition data is updated based on the adjusted actual propagation speed of the stress wave and the corrected sub-band energy.

[0060] In the environment in which the wind turbine blade operates, the most common influences are generally temperature, stress, and humidity, so a multi-field coupled factor of multiple physical fields is designed to quantify the influence degree; meanwhile, a self-adaptive band-stop filter is designed and a metasurface beamforming method is used to jointly suppress the interference generated by environmental influences from the frequency domain dimension and the spatial dimension; finally, the parameters of the multi-field coupled factor and the self-adaptive band-stop filter are dynamically adjusted by a multi-objective genetic algorithm, so that the two methods can maximize their effects.

[0061] The way of suppressing environmental interference on the propagation correction data includes:

[0062] Physical parameter sensors are arranged on each wind turbine blade to collect temperature, stress, and humidity data of the wind turbine blade; a multi-field coupled factor is constructed based on the temperature, stress, and humidity data of the wind turbine blade (the multi-field coupled factor is used to quantify the contribution of each physical parameter to the signal attenuation of the propagation correction data, which is manifested as weighted summation of each physical parameter); a self-adaptive band-stop filter is designed and used to suppress high-frequency noise generated by temperature drift in the frequency domain dimension (the stopband depth DE of the self-adaptive band-stop filter is defined as DE = -40 x log 10 (1+μ); wherein μ represents a stopband parameter); a metasurface beamforming method is used to generate a beam null in the interference direction (the beam null is a very low gain area generated by the metasurface beamforming method in the direction of the interference source, which is used to suppress electromagnetic wave interference); and the weights of each physical parameter in the multi-field coupled factor and the stopband parameter of the self-adaptive band-stop filter are dynamically adjusted by a multi-objective genetic algorithm (the weights of each physical parameter in the multi-field coupled factor and the stopband parameter of the self-adaptive band-stop filter are converted into a parameter vector; a genetic algorithm population is defined, and any individual in the population represents any parameter vector; the population is subjected to crossover and mutation operations and the process is repeatedly performed until a maximum number of iterations is reached, and the optimal individual, i.e., the optimal parameter vector, is selected based on the Pareto front).

[0063] In order to more accurately determine the damage of the wind turbine blade, feature extraction is performed on the improved stress wave signal data; in order to make the feature extraction effect more comprehensive, the data of all sensors are fused for processing, a feature tensor is constructed to accommodate damage-related features, the feature tensor is calibrated to remove errors, and then feature screening is performed to obtain accurate and improved stress wave signal feature data.

[0064] The feature extraction method for the improved stress wave signal data includes:

[0065] The improved stress wave signal data is classified based on data types (the data type classification standard is the difference in data units, for example, the units of sub-band energy intensity data are different from the units of dynamic propagation data); the data types include sub-band energy intensity data, dynamic propagation data, and time-frequency data (the time-frequency data include the corresponding waveband frequency and time delay sequence collected by any one sensor at each time stamp within the collection time interval; the dynamic propagation data include actual stress wave propagation speed, speed change rate, etc.); the sub-band energy intensity data and the time-frequency data are integrated into a time-frequency feature data set, and the dynamic propagation data are taken as a dynamic propagation feature data set; the time-frequency feature data set and the dynamic propagation feature data set from P sensors (P represents the number of installed sensors) are stacked into a three-dimensional tensor (three-dimensional tensor stacking means stacking three-dimensional feature matrices to form a tensor); time sampling points C and feature dimensions D within any one preset time window are cross combined (one time window size can be changed based on actual conditions; in this embodiment, one time window size is 1S; the time sampling points C refer to the number of sampling points within one time window, which is C; the feature dimensions D refer to the sum of the feature dimensions of each sub-band within the time window, for example, a certain sub-band contains 3 feature dimensions including sub-band energy intensity distribution, propagation speed, and speed change rate, another sub-band contains 5 feature dimensions, and only these two sub-bands are contained in one time window; therefore, the feature dimensions D represent a total of 8 feature dimensions), to obtain an initial feature tensor with dimensions of PxCxD within each time window; the initial feature tensor within each time window is subjected to cooperative processing of time domain synchronization calibration and frequency domain phase alignment, to obtain a calibrated feature tensor (the time domain synchronization calibration method is as follows: one sensor is selected as a master sensor, the time shift amount of other sensor signals relative to the master sensor is calculated through mutual correlation dominant peak tracking algorithm, and each sensor signal is time-shift compensated based on the time shift amount to obtain a compensated sensor signal; the frequency domain phase alignment method is as follows: the compensated sensor signal is subjected to short-time Fourier transform to obtain a compensated time-frequency matrix, and the phase of the compensated time-frequency matrix is corrected by using instantaneous frequency gradient estimation); and the position of the wind turbine blade damage area is calculated based on the time-frequency feature data set (the calculation formula of the position of the wind turbine blade damage area is as follows: Wherein, the coordinates of the position of the wind turbine blade damage area are (x, y); the coordinates of the sensor a are (x a ,y a ), and the coordinates of the sensor b are (x b ,y b ); represents the actual propagation speed of the stress wave with the frequency f; t ab represents the stress wave propagation time difference between the sensor a and the sensor b); the Euclidean distance between the position of the wind turbine blade damage area and any one sensor is calculated, and the signal-to-noise ratio index of each sensor is calculated, and each sensor is given a weight based on the signal-to-noise ratio index of each sensor and the Euclidean distance between the sensor and the position of the wind turbine blade damage area, that is, the calibration feature tensor is weighted (the calculation formula of the weight of each sensor is: Wherein, ω p0 represents the weight of the p0th sensor; SNR(p0) represents the signal-to-noise ratio index of the p0th sensor; -γ0(f) represents the initial attenuation compensation coefficient of the p0th sensor in the frequency f band; disp0 represents the Euclidean distance between the p0th sensor and the position of the damage area; the weight takes into account the signal-to-noise ratio index of each sensor and the Euclidean distance from the damage area, so as to facilitate the control of the contribution proportion of the global and local features, for example, the higher the sensor signal-to-noise ratio, the greater the energy of the collected stress wave, and the smaller the Euclidean distance between the sensor and the damage area, the closer the sensor to the damage area, so the data of the sensor is easier to extract the damage feature, and the sensor should be given a greater weight), to obtain an intermediate feature matrix.

[0066] The intermediate feature matrix contains features of all dimensions, but because the feature quantity is large and contains repeated or similar redundant features, the intermediate feature matrix needs to be further feature screened; first, the feature points are clustered based on the local density and mapped to a high-dimensional space through a clustering algorithm, and then a classification model with a random forest model as a basic structure is used for accurate feature screening, and finally a principal component analysis algorithm is used to reduce and compress the screened features, improving the feature screening accuracy and processing efficiency.

[0067] The way of further feature screening of the intermediate feature matrix includes:

[0068] The local density of each feature point in the intermediate feature matrix is calculated (the calculation formula of the local density is: Wherein, ρ represents the local density of any one feature point pt; ki(pt) represents the K-neighbor set of the feature point pt, and in this embodiment, the K-neighbor parameter ki of the K-neighbor set is 8; dis(pt-pt sum) represents the Euclidean distance between the feature point pt and the sumth feature point) and clustering all feature points in the intermediate feature matrix based on the local density of each feature point, the intermediate feature matrix is divided into two sub-matrixes, high-density matrix and low-density matrix (the distance between adjacent feature points in the high-density matrix is small, and this kind of matrix is beneficial to enhance local details; the distance between adjacent feature points in the low-density matrix is large, and this kind of matrix is beneficial to suppress overfitting); an adaptive kernel function is constructed to map the high-density matrix and the low-density matrix to a high-dimensional space respectively to obtain a high-dimensional space feature matrix (the calculation formula of the adaptive kernel function is: HE(pt-pt sum )=β1×HE RB (pt-pt sum )+β2×HE RF (pt-pt sum ); wherein, HE represents the adaptive kernel function; HE RB represents the radial basis kernel function, β1 represents the weight coefficient of HE RB ; HE RF represents the linear kernel function, β2 represents the weight coefficient of HE RF ; in this embodiment, β1>β2 when processing the high-density matrix, which is used to capture local details; β1<β2 when processing the low-density matrix, which is used to capture overall trends; β1+β2=1; using this function, the features originally showing nonlinear distribution are mapped to a high-dimensional space, making them linearly separable in the high-dimensional space, and enhancing the separability of the features).

[0069] A classification model is constructed, and a random forest model is used as the basic structure of the classification model (the random forest model is an excellent classification model with strong robustness and fault tolerance, and is suitable for processing high-dimensional data); historical wind turbine blade damage data is collected, and the data is converted into a historical feature matrix; by querying a pre-set wind turbine blade parameter database, the key damage features involved in the damage of the historical feature matrix are taken as the training labels of the classification model; the classification model is trained using the historical feature matrix until the function value of the loss function (such as the cross-entropy loss function) of the classification model no longer decreases, and a trained classification model is obtained; the classification model is used to screen key features from the high-dimensional space feature matrix, and the Gini index of each feature point in the high-dimensional space feature matrix is calculated (the Gini index is an index for measuring the importance of features in the random forest model); all feature points are sorted in descending order based on the Gini index, and the feature points whose Gini index is greater than a pre-set sorting threshold (the sorting threshold is 30%, that is, the top 30%) are determined as key feature points; the mutual information analysis algorithm is used to remove redundant feature points in the key feature points, and a damage-sensitive feature set is output (by calculating the mutual information between any two feature points, if it is greater than a pre-set mutual information threshold, one of the feature points is removed, which reduces the data dimension and improves the efficiency of subsequent processing).

[0070] The damage sensitive feature set is standardized to obtain a standardized feature set; and a principal component analysis algorithm is used to reduce dimension and compress the standardized feature set to obtain stress wave signal feature data (the principal components in the standardized feature set with a cumulative variance contribution rate greater than or equal to 95% are retained, wherein the cumulative variance contribution rate is an index used to measure the amount of information contained in the principal components in the principal component analysis algorithm, and the greater the value, the greater the amount of information contained in the principal components, and the more important the corresponding principal components are).

[0071] In order to be able to evaluate the damage condition of the wind turbine blade based on the feature data, a damage evaluation model is constructed to identify and evaluate the damage condition of the wind turbine blade; and a damage coding system is designed to express the evaluation results of the model.

[0072] The method for constructing the damage evaluation model comprises:

[0073] The damage coding system uses hierarchical coding rules (each damage code is divided into three parts, including damage type, damage location and damage degree; wherein the damage degree is represented by the damage index calculated in subsequent processing; for example, the damage code G-01-L1 represents that a slight degree of crack is generated at the tip of the wind turbine blade, wherein G represents that a crack is generated; 01 represents that the crack position is at the 01 coordinate position, i.e. the tip of the wind turbine blade, and the detailed coordinates of the 01 coordinate position can be calculated; L1 represents the damage degree, i.e. slight; all parameters in the above damage code can be queried or calculated through the preset wind turbine blade parameter database); the preset wind turbine blade parameter database is queried to determine all specific damage conditions of the wind turbine blade and the corresponding features of the damage conditions; based on the damage coding system, all specific damage conditions of the wind turbine blade are coded, and the code of any specific damage condition is combined with the corresponding feature of the damage condition to form a feature-coding mapping matrix (for example, the damage code G-01-L1 corresponds to the feature that the energy intensity at the position exceeds the preset energy threshold and the stress wave propagation speed and path are mutated); and all feature-coding mapping matrices are integrated to obtain a feature-coding mapping dataset.

[0074] The residual neural network model is taken as a basic structure of the damage assessment model, including an input layer, a full connection layer, a feature processing layer, a damage quantification layer and an output layer; stress wave signal feature data is taken as input data of the damage assessment model; a double-channel parallel processing architecture is taken as a basic framework of the feature processing layer in the damage assessment model, including a time-frequency feature channel and a spatial propagation channel; the time-frequency feature channel extracts time-frequency features of the input data through a convolution kernel integrated in a residual block of the residual neural network, and simultaneously introduces a cross-channel attention mechanism (CBAM) to enhance a damage-sensitive frequency band; the spatial propagation channel obtains stress wave propagation relationships between various sensors through a graph attention network (GAT) to capture a propagation mode of the stress wave at a damage position (the attention coefficients of the GAT reflect energy transmission between the sensors, and meanwhile, the damage position changes a stress wave propagation path, so the GAT can also capture such a propagation mode); the input data is jointly processed through the time-frequency feature channel and the spatial propagation channel to generate an intermediate feature tensor; the intermediate feature tensor enters the damage quantification layer to calculate a damage index (the calculation formula of the damage index is: LI = μ1 x ΔV + μ2 x ΔE + μ3 x S dis ; wherein, LI represents the damage index; ΔV represents a stress wave propagation speed attenuation rate at the damage position; ΔE represents an energy attenuation rate at the damage position; S dis represents spatial dispersion at the damage position, which is calculated based on a variance of detailed coordinates of the damage position; μ1, μ2 and μ3 are weights of ΔV, ΔE and S dis respectively, and the particle swarm algorithm is used to optimize the three weights to realize dynamic adjustment of the weights; damage safety, first and second thresholds are set, and the damage degree is divided into four categories based on the three damage thresholds; when the damage index is less than or equal to the damage safety threshold, it is determined as a safe degree; when the damage index is greater than the damage safety threshold and less than or equal to the damage first threshold, it is determined as a slight damage degree; when the damage index is greater than the damage first threshold and less than or equal to the damage second threshold, it is determined as a moderate damage degree; when the damage index is greater than the damage second threshold, it is determined as a severe damage degree; in this embodiment, the damage safety threshold is 0.1, the damage first threshold is 0.3, the damage second threshold is 0.7, and the damage index is in a value range of (0, 1); and the output layer outputs the damage index, damage position coordinates and damage type; the damage index, damage position coordinates and damage type are matched to generate a damage code; the existing damage assessment model is dynamically updated after each output of the damage code (in order to ensure the evaluation accuracy and real-time performance of the damage assessment model); a real-time time stamp is added to all output damage codes, and the damage codes are sorted in chronological order to obtain a damage assessment record (sorting each damage code in chronological order is conducive to observing changes in the damage condition and facilitating subsequent operation to generate a more accurate three-dimensional damage atlas).

[0075] Since new damage features may be introduced each time the damage condition of a wind turbine blade is evaluated, the efficiency of the model is quite low if the model has to relearn all the features each time an evaluation is performed, so incremental learning is selected to dynamically update the damage evaluation model, so that the model can adapt to new features more quickly; at the same time, in order to avoid the performance of the model from being reduced after the model is updated, a model parameter rollback mechanism is introduced to dynamically update the model while ensuring the stability of the model.

[0076] The way to dynamically update the existing damage evaluation model includes:

[0077] An incremental learning architecture is constructed, a dynamic feature memory bank is established after the feature processing layer of the damage evaluation model, the intermediate feature tensor is backed up, and the backup intermediate feature tensor is feature extracted to obtain a historical damage feature dataset (the historical damage feature dataset includes damage sensitive features and high confidence damage features, i.e., features with a damage index greater than or equal to a preset confidence threshold, to ensure the reliability of the memory features; in this embodiment, the confidence threshold is 0.8); a double model collaborative distillation mechanism is introduced in the incremental learning architecture, including a first model and a second model (wherein the first model is used to retain and pass old features, and the second model is used to receive old features and learn new features); the feature processing layer of the damage evaluation model is frozen using the first model (the feature processing layer is frozen to retain the identified damage features, ensuring the basic feature extraction capability), and the probability tensor of the historical damage feature dataset is output (the probability tensor refers to the probability weight of the old features and similar features, improving the efficiency of the first model in passing old features); the second model copies the parameters of the frozen layer from the first model (the parameters of the first model are copied to ensure that the second model has the same extraction capability as the first model), and adjusts the parameters of the unfrozen layer based on the probability tensor (the calculation formula of the adjustment function is: Wherein, CA represents the adjustment function; KL represents the KL divergence loss function, which is a loss function for measuring the difference between probability distributions; CE represents the cross-entropy loss function; And respectively represent the weights of the KL divergence loss function and the cross-entropy loss function, and in this embodiment ); setting an initial learning rate of the incremental learning architecture and adjusting the initial learning rate based on a cosine annealing algorithm (setting the initial learning rate to 30% of the learning rate of the damage assessment model before the update; gradually decaying based on the cosine annealing algorithm, balancing the learning speed of new and old features, and preventing the situation of forgetting old features due to the dramatic oscillation of parameters); setting a parameter constraint for the damage quantification layer to ensure that the parameter offset is less than a preset offset threshold (in this embodiment, the offset threshold is 0.15, that is, when the parameter offset is less than 0.15, no adjustment is made, otherwise the parameter is restored to the value before the offset, ensuring the stability of the damage quantification layer); calculating the performance indicators (such as evaluation accuracy, false positive rate and evaluation time consumption) of the damage assessment model after adding the incremental learning architecture, if each performance indicator meets the expected value, the damage assessment model update is completed, otherwise the damage assessment model rollback is triggered (since each time the damage assessment model is dynamically updated, the relevant parameters of the damage assessment model before the update are retained in a piece of data buffer, if the performance indicators do not meet the expected value, the parameters of the damage assessment model are rolled back to the state before the update, ensuring the stability of the damage assessment model).

[0078] In order to visualize the damage assessment records generated by the damage assessment model, it is convenient for technicians to view; therefore, the damage assessment records are converted into three-dimensional damage maps, wherein a three-dimensional model of the wind turbine blade is established and the damage area and its damage degree are labeled by color, which is convenient for technicians to repair the wind turbine blade based on the map.

[0079] The way of generating a three-dimensional damage map based on the damage assessment records includes:

[0080] extract the damage location coordinates, damage index and damage type from the damage assessment record, and construct a damage spatial distribution matrix (the matrix is used to store the spatial distribution characteristics of the damage); utilize the Kriging spatial interpolation algorithm to perform three-dimensional gridding processing on the damage spatial distribution matrix, to obtain a wind turbine blade damage global density field (the Kriging spatial interpolation algorithm is used to analyze spatial correlation to construct the wind turbine blade damage global density field, which is used to represent the global damage distribution state of the wind turbine blade); perform parameterized modeling on the wind turbine blade based on the geometric parameters of the wind turbine blade (the above-mentioned geometric parameters can be obtained by querying a preset wind turbine blade parameter database, such as material parameters and blade specifications and the like; the parameterized modeling method is, for example, modeling by using CAD), to obtain a wind turbine blade three-dimensional model; map the wind turbine blade damage global density field to the wind turbine blade three-dimensional model, to obtain a three-dimensional damage model (wherein the mapping area includes the model surface and internal structural layers); utilize color to render the three-dimensional damage model, to obtain a damage rendering model (adopting HSL color gamut to map the damage degree, classifying the damage degree corresponding to the color based on the damage index size, the safety degree being represented as green; the slight damage degree being represented as gradually changing from green to blue, and when completely changing to blue, the damage index is at the first damage threshold; the moderate damage degree being represented as gradually changing from blue to yellow, and when completely changing to yellow, the damage index is at the second damage threshold; the severe damage degree being represented as gradually changing from yellow to red); sort the damage rendering model corresponding to each timestamp in chronological order, to obtain a three-dimensional damage atlas.

[0081] The embodiment realizes the process of monitoring the wind turbine blade based on the stress wave, by collecting stress wave signal data, pre-processing and feature extraction, obtaining stress wave signal feature data, constructing a model based on the data to evaluate the damage of the wind turbine blade, and finally generating a visual three-dimensional damage atlas based on the damage assessment record. Compared with existing experience, the pre-processing of the stress wave signal data is more accurate, while considering the influence of noise interference in the data, errors in the propagation process and environmental factors. Joint feature extraction is performed from the time domain and the frequency domain, to obtain features related to the damage of the wind turbine blade, construct a classification model to further filter the features, obtain key features representing the damage condition, and perform dimensionality reduction compression, to improve the processing efficiency. A damage assessment model is constructed and dynamically updated based on the incremental learning architecture, while the model is used to identify and evaluate the damage condition, to improve the adaptability and processing accuracy of the model. Finally, the visual damage atlas is generated and sent to the preset wind turbine equipment safety terminal, and subsequent technical personnel take corresponding measures to repair the wind turbine blade based on the three-dimensional damage atlas.

[0082] Embodiment 2

[0083] Please refer to Figure 2As shown, the embodiment does not describe part of the embodiment 1, and provides a stress wave sensing based measurement system applied to a wind turbine blade, comprising:

[0084] A data acquisition module is configured to acquire stress wave signal data and pre-process the stress wave signal data, and obtain perfect stress wave signal data by successively performing bipolar collaborative decomposition, error correction and environmental interference suppression on the stress wave signal data.

[0085] A feature extraction module is configured to perform feature extraction on the perfect stress wave signal data and output stress wave signal feature data.

[0086] A damage assessment module is configured to construct a damage assessment model based on the stress wave signal feature data, and perform damage assessment on the wind turbine blade by using the damage assessment model to obtain damage assessment records.

[0087] A visualization generation module is configured to generate a three-dimensional damage atlas based on the damage assessment records and send the three-dimensional damage atlas to a preset wind power equipment safety terminal. The modules are connected through wired and / or wireless means.

[0088] Embodiment 3

[0089] The embodiment discloses an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned application of a stress wave sensing based measurement method applied to a wind turbine blade is realized.

[0090] Since the electronic device introduced in the embodiment is the electronic device used to implement the application of a stress wave sensing based measurement method applied to a wind turbine blade in the embodiment, the specific implementation of the electronic device and its various forms can be understood by those skilled in the art based on the application of a stress wave sensing based measurement method applied to a wind turbine blade in the embodiment. Therefore, the implementation of the method in the embodiment will not be described in detail. As long as the electronic device used to implement the application of a stress wave sensing based measurement method applied to a wind turbine blade in the embodiment is implemented by those skilled in the art, it belongs to the scope of protection of the present application.

[0091] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation. The preset parameters and threshold values in the formula are set by those skilled in the art according to the actual situation.

[0092] The above merely describes the preferred embodiments of the present application, and the protection scope of the present application is not limited to the above-described embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary technical users in the technical field, several improvements and refinements without departing from the principle of the present application shall also be considered as falling within the protection scope of the present application.

Claims

1. A stress wave based sensing method of measurement applied to a wind turbine blade, characterized in that, Comprise: S1. Collecting stress wave signal data and preprocessing stress wave signal data, by bipolar collaborative decomposition, error correction and environmental interference suppression on stress wave signal data in turn, perfect stress wave signal data is obtained; S2. Feature extraction is performed on the perfect stress wave signal data, an intermediate feature matrix is generated, and further feature screening is performed on the intermediate feature matrix, and stress wave signal feature data is output; S3. Constructing a damage assessment model based on stress wave signal feature data, and using the damage assessment model to assess the damage of the wind turbine blade, and obtaining damage assessment records; S4. Generating a three-dimensional damage atlas based on the damage assessment records, and sending the three-dimensional damage atlas to a preset wind power equipment safety terminal.

2. A stress wave based sensing method for measuring applied to a wind turbine blade as claimed in claim 1, wherein, The sensors are symmetrically arranged on the main beam, leading edge and trailing edge of each wind turbine blade, and the interval between any two sensors is less than or equal to a preset distance threshold; stress waves are emitted by a stress wave generator, and stress wave signal data is collected by sensors, the stress wave signal data includes time domain waveform signal and stress wave propagation time difference.

3. A stress wave based sensing method for measuring applied to a wind turbine blade as claimed in claim 2, wherein, The preprocessing method of stress wave signal data comprises: The time domain waveform signal is converted into a frequency domain spectrum signal by using a fast Fourier transform algorithm; the frequency domain spectrum signal is subjected to bipolar collaborative decomposition, and the stress wave propagation speed is calculated based on the stress wave propagation time difference; the bipolar collaborative decomposition results, the stress wave propagation time difference and the stress wave propagation speed are integrated to obtain multi-modal decomposition data; a propagation dynamic compensation mechanism is designed to correct the error of the multi-modal decomposition data, and propagation correction data is generated; the propagation correction data is subjected to environmental interference suppression based on the collected multi-physical field parameters, and perfect stress wave signal data is obtained; The bipolar collaborative decomposition method of the frequency domain spectrum signal comprises: A preselected wavelet packet basis function is used to perform N-layer decomposition on the frequency domain spectrum signal to generate multi-scale spectrum subbands and ensure that all generated spectrum subbands can cover the full frequency band; the energy proportion of each spectrum subband is calculated and a subband energy threshold is set; when the energy proportion of any spectrum subband is greater than or equal to the preset subband energy threshold, the spectrum subband is determined as an effective subband, otherwise the spectrum subband is determined as a noise subband; the effective subbands are subjected to adaptive soft threshold filtering to obtain effective high-frequency subbands; the noise subbands are subjected to empirical mode decomposition to generate M intrinsic mode functions; the stability index of each intrinsic mode function is calculated, and the intrinsic mode component with a stability index less than a preset stability index threshold is retained, i.e. the effective mode component; the energy of the intrinsic mode component with a stability index greater than or equal to the preset stability index threshold is extracted by using an energy back injection algorithm, and the energy is superimposed on the adjacent effective mode component; the effective high-frequency subbands and the effective mode components are subjected to synchronous extraction transformation and construction of a time-frequency feature matrix, and the time-frequency feature matrix, the stress wave propagation time difference and the stress wave propagation speed are integrated to obtain the multi-modal decomposition data.

4. A stress wave based sensing method for measuring the application of wind turbine blades as claimed in claim 3, wherein, The error correction method of the multi-modal decomposition data comprises: The basic propagation speed of the stress wave in the wind turbine blade and a material coefficient are obtained by querying a preset wind turbine blade parameter database; a wave speed function is constructed based on the basic propagation speed and the material coefficient; and the actual propagation speed of the stress wave in the wind turbine blade is obtained by calculating the function value of the wave speed function; The time delay sequence is obtained by sorting the stress wave propagation time difference in ascending order based on the time stamp, and the time delay mean value of the time delay sequence is calculated; the energy intensity distribution of each subband is extracted from the multi-modal decomposition data; the wave speed correction equation is constructed based on the time delay mean value and the energy intensity distribution of each subband; the coupling coefficient of the wave speed correction equation is updated using the pre-trained LSTM network model, and the actual propagation speed of the stress wave in the wind turbine blade is adjusted based on the updated wave speed correction equation; the attenuation compensation function is constructed, and the subband energy attenuation in the multi-modal decomposition data is corrected based on the function; and the multi-modal decomposition data is updated based on the adjusted actual propagation speed of the stress wave and the corrected subband energy. The method for suppressing environmental interference on the propagation correction data comprises: Physical parameter sensors are arranged on each wind turbine blade to collect temperature, stress and humidity data of the wind turbine blade; a multi-physical field coupling factor is constructed based on the temperature, stress and humidity data of the wind turbine blade; an adaptive band-stop filter is designed, and the filter is used to suppress high-frequency noise generated by temperature drift in the frequency domain dimension; a wave beam null is generated in the interference direction using a metasurface beam forming method; and the weight of each physical parameter in the multi-physical field coupling factor and the stop band parameter of the adaptive band-stop filter are dynamically adjusted through a multi-objective genetic algorithm.

5. A stress wave based sensing method for measuring the application of wind turbine blades as claimed in claim 4, wherein, The method for extracting features from the improved stress wave signal data comprises: The improved stress wave signal data is classified based on data types, including subband energy intensity data, dynamic propagation data and time-frequency data; the subband energy intensity data and the time-frequency data are integrated into a time-frequency feature data set, and the dynamic propagation data is taken as a dynamic propagation feature data set; time sampling points C and feature dimensions D in any one preset time window are cross combined to obtain an initial feature tensor with a dimension of PxCxD in each time window through three-dimensional tensor stacking of the time-frequency feature data set and the dynamic propagation feature data set from P sensors; the initial feature tensor in each time window is subjected to collaborative processing of time domain synchronization calibration and frequency domain phase alignment to obtain a calibrated feature tensor; the position of the wind turbine blade damage area is calculated based on the time-frequency feature data set; the Euclidean distance between the position of the wind turbine blade damage area and any one sensor is calculated, and the signal-to-noise ratio index of each sensor is calculated, and each sensor is assigned a weight based on the signal-to-noise ratio index of the sensor and the Euclidean distance between the sensor and the position of the wind turbine blade damage area, i.e., the calibrated feature tensor is weighted to obtain an intermediate feature matrix.

6. A stress wave based sensing measurement method applied to a wind turbine blade according to claim 5, characterized in that, The method for further feature screening of the intermediate feature matrix comprises: The local density of each feature point in the intermediate feature matrix is calculated, and all feature points in the intermediate feature matrix are clustered based on the local density of each feature point, so as to divide the intermediate feature matrix into two sub-matrices, i.e., a high-density matrix and a low-density matrix; and an adaptive kernel function is constructed to map the high-density matrix and the low-density matrix to a high-dimensional space, so as to obtain a high-dimensional space feature matrix; A classification model is constructed, and a random forest model is used as the basic structure of the classification model; historical wind turbine blade damage data is collected, and the data is converted into a historical feature matrix; by querying a preset wind turbine blade parameter database, key damage features related to the damage of the historical feature matrix are used as training labels of the classification model; the classification model is trained using the historical feature matrix until the function value of the loss function of the classification model no longer decreases, and a trained classification model is obtained; the classification model is used to screen key features from the high-dimensional space feature matrix, and the Gini index of each feature point in the high-dimensional space feature matrix is calculated; all feature points are sorted in descending order based on the Gini index, and feature points greater than a preset sorting threshold are determined as key feature points; redundant feature points in the key feature points are removed using a mutual information analysis algorithm, and a damage sensitive feature set is output; The damage sensitive feature set is standardized to obtain a standardized feature set; principal component analysis is used to reduce the dimension of the standardized feature set to obtain stress wave signal feature data.

7. A stress wave based sensing method for measuring applied to a wind turbine blade as claimed in claim 6, wherein, The method for constructing the damage assessment model comprises: A damage coding system is constructed, which uses hierarchical coding rules; a preset wind turbine blade parameter database is queried to determine all specific damage conditions of the wind turbine blade and corresponding features of the damage conditions; the damage coding system is used to code all specific damage conditions of the wind turbine blade, and the code of any specific damage condition is combined with the corresponding features of the damage condition to form a feature-code mapping matrix; all feature-code mapping matrices are integrated to obtain a feature-code mapping dataset; A damage assessment model is constructed, and a residual neural network model is used as the basic structure of the damage assessment model, including an input layer, a fully connected layer, a feature processing layer, a damage quantification layer, and an output layer; the stress wave signal feature data is used as the input data of the damage assessment model; a double-channel parallel processing architecture is used as the basic framework of the feature processing layer of the damage assessment model, including a time-frequency feature channel and a spatial propagation channel; the input data is processed jointly by the time-frequency feature channel and the spatial propagation channel to generate an intermediate feature tensor; the intermediate feature tensor enters the damage quantification layer to calculate a damage index, which is output by the output layer together with damage location coordinates and damage types; the damage index, damage location coordinates, and damage types are matched to generate a damage code; the existing damage assessment model is dynamically updated after each output of the damage code; a real-time timestamp is added to all output damage codes, and the damage codes are sorted in chronological order to obtain a damage assessment record.

8. A stress wave based sensing measurement method applied to a wind turbine blade according to claim 7, characterized in that, The method for dynamically updating the existing damage assessment model comprises: The incremental learning architecture is constructed, a dynamic feature memory bank is established after the feature processing layer of the damage assessment model, the intermediate feature tensor is backed up, and feature extraction is performed on the backed-up intermediate feature tensor to obtain a historical damage feature dataset; a double-model collaborative distillation mechanism is introduced into the incremental learning architecture, including a first model and a second model; the feature processing layer of the damage assessment model is frozen by using the first model, and a probability tensor of the historical damage feature dataset is output; the second model adjusts the parameters of the unfrozen layer based on the adjustment function constructed by the probability tensor; the initial learning rate of the incremental learning architecture is set, and the initial learning rate is adjusted based on the cosine annealing algorithm; the parameter constraint is set for the damage quantization layer to ensure that the parameter offset is less than the preset offset threshold; the performance indicators of the damage assessment model after adding the incremental learning architecture are calculated, and if each performance indicator meets the expected value, the damage assessment model is updated, otherwise the damage assessment model is rolled back.

9. A stress wave based sensing measurement method applied to a wind turbine blade according to claim 8, characterized in that, The method for generating a three-dimensional damage atlas based on the damage assessment record comprises: Damage position coordinates, damage indexes, and damage types are extracted from the damage assessment record, and a damage spatial distribution matrix is constructed; a Krige spatial interpolation algorithm is used to perform three-dimensional gridding processing on the damage spatial distribution matrix to obtain a wind turbine blade damage global density field; a parameterized modeling of the wind turbine blade is performed based on the geometric parameters of the wind turbine blade to obtain a wind turbine blade three-dimensional model; the wind turbine blade damage global density field is mapped to the wind turbine blade three-dimensional model to obtain a three-dimensional damage model; the three-dimensional damage model is rendered using color to obtain a damage rendering model; the damage rendering models corresponding to each timestamp are sorted in chronological order to obtain a three-dimensional damage atlas.

10. A stress wave based sensing measurement system applied to a wind turbine blade for implementing a stress wave based sensing measurement method according to any one of claims 1 to 9, characterized in that, The method comprises: The data acquisition module is configured to acquire stress wave signal data and pre-process the stress wave signal data to obtain improved stress wave signal data. The feature extraction module is configured to extract features from the improved stress wave signal data and output stress wave signal feature data. The damage assessment module is configured to construct a damage assessment model based on the stress wave signal feature data, perform damage assessment on the wind turbine blade using the damage assessment model, and obtain a damage assessment record. The visualization generation module is configured to generate a three-dimensional damage atlas based on the damage assessment record and send the three-dimensional damage atlas to a preset wind power equipment safety terminal. The modules are connected through wired and / or wireless means.

Images