A fan blade diagnosis method and system based on voiceprint perception

Abstract

The application discloses a fan blade diagnosis method and system based on voiceprint perception, relates to the technical field of intelligent monitoring of wind power equipment, and comprises the following steps: performing frequency domain compensation processing based on a set of wind flow characteristic parameters, stripping environmental wind noise components, and generating a pure voiceprint signal; identifying a sensitive acoustic frequency band of a fan blade based on the pure voiceprint signal, exciting a sound wave phase coordination effect in the sensitive acoustic frequency band, and generating a strengthened voiceprint signal; quantifying the internal damage depth of the fan blade material according to the strengthened voiceprint signal, and calculating an acoustic energy flow disorder degree index; determining the health state of the fan blade according to the acoustic energy flow disorder degree index through a multi-dimensional acoustic characteristic state space mapping mechanism, and generating a three-dimensional health evaluation report; and through physical-level wind noise stripping technology, generating a transfer function matrix based on vortex interference modeling and sound wave-air flow phase offset correction, realizing frequency energy reweighting decoupling, and accurately stripping environmental wind noise in a strong turbulent flow environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for wind power equipment, and in particular to a method and system for diagnosing wind turbine blades based on voiceprint perception. Background Technology

[0002] With the development of acoustic sensing technology, passive acoustic monitoring based on microphone arrays has gradually become mainstream, achieving spatial localization of noise sources through beamforming algorithms. In recent years, acoustic emission detection technology has attracted attention due to its high sensitivity to microcracks, enabling early damage warning by capturing high-frequency transient signals released from stress waves within materials. At the signal processing level, adaptive filters and blind source separation methods are widely used in environmental noise suppression, particularly in separating wind noise from structural noise, resulting in various frequency domain processing schemes.

[0003] Traditional wind noise separation algorithms simplify wind flow as a steady noise source, neglecting the nonlinear modulation effect of vortex unsteady flow separation on the acoustic wave phase. Passive acoustic detection is limited by acoustic diffraction laws and cannot overcome the physical limitation of the sharp attenuation of acoustic energy flux density in the damaged area. Existing signal enhancement schemes generally adopt fixed resonant structures, which are difficult to achieve adaptive focusing of energy over a wide frequency band when facing the frequency-varying characteristics of composite materials and complex damage modes. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a wind turbine blade diagnostic method based on acoustic signature perception to solve the complex problems of wind noise suppression and micro-damage signal capture in the field of wind turbine blade fault diagnosis.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a wind turbine blade diagnostic method based on voiceprint perception, comprising,

[0008] Acquire acoustic fingerprint data and airflow parameters and preprocess them, analyze the mutual characteristics of the preprocessed airflow parameters and acoustic fingerprint data, and generate a set of airflow characteristic parameters.

[0009] Based on the set of wind flow characteristic parameters, frequency domain compensation processing is performed to remove environmental wind noise components and generate a pure acoustic signal.

[0010] Based on the pure acoustic signature signal, the sensitive acoustic frequency band of the wind turbine blade is identified, and the acoustic wave phase coordination effect is excited in the sensitive acoustic frequency band to generate an enhanced acoustic signature signal.

[0011] Based on the enhanced acoustic signature signal, the internal damage depth of the wind turbine blade material is quantified, and the acoustic energy flow turbulence index is calculated.

[0012] Based on the acoustic energy flow turbulence index, the health status of the wind turbine blades is determined through a multi-dimensional acoustic feature state space mapping mechanism, generating a three-dimensional health assessment report.

[0013] As a preferred embodiment of the wind turbine blade diagnosis method based on acoustic signature perception described in this invention, the acoustic signature data includes blade vibration acoustic characteristics and environmental wind noise components, and the wind flow parameters include three-dimensional wind speed vector, wind direction angle and turbulence intensity.

[0014] The preprocessing includes pre-emphasis filtering to suppress low-frequency noise, Hamming window framing processing, and calibration of airflow parameters, temperature, and altitude.

[0015] As a preferred embodiment of the wind turbine blade diagnostic method based on acoustic signature perception described in this invention, the method involves analyzing the mutual characteristics between the preprocessed airflow parameters and the acoustic signature data to generate an airflow characteristic parameter set, which is then generated through Doppler phase shift calculation and vortex interference modeling.

[0016] As a preferred embodiment of the wind turbine blade diagnostic method based on voiceprint perception described in this invention, the steps of performing frequency domain compensation processing based on the wind flow characteristic parameter set to remove environmental wind noise components and generate a pure voiceprint signal are as follows:

[0017] The three-dimensional motion parameters of the airflow around the fan are directly measured by an ultrasonic anemometer, and the vortex frequency is generated by calculating the periodic vortex shedding frequency based on the Strauhall law.

[0018] The vibration signal of the blade is captured by an acoustic sensor array, and the significant peak of the power spectrum is extracted and the fundamental frequency of the acoustic signature is generated by pre-emphasis filtering and Mel-frequency cepstral coefficient analysis.

[0019] Based on the set of airflow characteristic parameters, a transfer function matrix is ​​generated by frequency domain energy overlap analysis of vortex frequency and acoustic fundamental frequency, combined with acoustic wave-airflow phase shift correction mechanism.

[0020] Frequency domain compensation is performed on the transfer function matrix and the preprocessed acoustic text data. Environmental wind noise components are removed by physical-level interference field decoupling to generate an optimized frequency domain acoustic text spectrum.

[0021] The optimized frequency domain acoustic spectrum is inversely transformed into a time domain waveform, and the purity is verified by the acoustic energy flow entropy value to generate a pure acoustic signal.

[0022] As a preferred embodiment of the wind turbine blade diagnostic method based on voiceprint perception described in this invention, the steps of identifying the sensitive acoustic frequency band of the wind turbine blade based on a pure voiceprint signal, and exciting a phase coordination effect of sound waves in the sensitive acoustic frequency band to generate an enhanced voiceprint signal are as follows.

[0023] A sweep frequency acoustic excitation experiment was performed on the blade composite material specimen to identify the characteristic curve of the propagation speed of sound waves in the material as a function of frequency. After temperature and humidity compensation, the intrinsic dispersion response of the blade material was generated.

[0024] Based on the pure acoustic signature signal, acoustic signature resonant peak features are extracted by Mel frequency cepstral coefficients, and damage feature matching is performed by combining the intrinsic dispersion response of the blade material to generate sensitive acoustic frequency bands.

[0025] Based on the sensitive acoustic frequency band, the frequency of the resonant unit of the acoustic metasurface array pre-installed on the surface of the wind turbine blade is dynamically adjusted and the phase spatial distribution of the sound wave front is reconstructed. The sound wave phase synergy effect is excited in the sensitive acoustic frequency band to generate a sound field focusing control signal.

[0026] By driving an acoustic metasurface array with a sound field focusing control signal, a sound field energy spatial focusing is formed in the sensitive acoustic frequency band, generating an enhanced acoustic pattern signal.

[0027] As a preferred embodiment of the wind turbine blade diagnostic method based on acoustic signature perception described in this invention, the steps of quantifying the internal damage depth of the wind turbine blade material and calculating the acoustic energy flow turbulence index based on the enhanced acoustic signature signal are as follows:

[0028] Based on the enhanced acoustic signature signal, the acoustic dispersion feature spectrum is extracted through the Mel frequency cepstral coefficient. A reference benchmark is established based on the intrinsic dispersion response of the blade material, and a layered damage feature vector is generated.

[0029] Based on the layered damage feature vector, the material degradation index is calculated and matched with the depth-dispersion mapping to drive the Bayesian algorithm to generate a quantitative value of damage depth.

[0030] Based on the enhanced acoustic signature signal and combined with the damage depth quantization value, the energy transfer efficiency of each frame of enhanced acoustic signature signal in the damaged area is calculated to generate an acoustic energy flow turbulence index.

[0031] As a preferred embodiment of the wind turbine blade diagnosis method based on acoustic signature perception described in this invention, the steps of determining the blade health status of the wind turbine based on the acoustic energy flow turbulence index through a multi-dimensional acoustic feature state space mapping mechanism and generating a three-dimensional health assessment report are as follows.

[0032] Based on the acoustic energy flow turbulence index and damage depth quantification value, combined with the three-dimensional coordinates of damage, a decision ratio is allocated through weight allocation rules to generate a scatter plot of the blade zone health status.

[0033] Based on the scatter plot of the health status of the blade partition, the cluster density of health status points in each partition is calculated, and health status labels are generated by combining the critical state of acoustic energy flow turbulence.

[0034] A spatial-temporal-state 3D fusion rendering engine is used to perform 3D fusion rendering of health status labels and scatter plots of regional situations, generating a 3D health assessment report.

[0035] Secondly, the present invention provides a wind turbine blade diagnostic system based on voiceprint perception, comprising,

[0036] The data preprocessing module is used to collect acoustic fingerprint data and airflow parameters and preprocess them, analyze the mutual characteristics between the preprocessed airflow parameters and acoustic fingerprint data, and generate a set of airflow characteristic parameters.

[0037] The environmental wind noise stripping module is used to perform frequency domain compensation processing based on the wind flow characteristic parameter set, strip away the environmental wind noise components, and generate a pure acoustic signal.

[0038] The acoustic signature enhancement module is used to identify the sensitive acoustic frequency bands of wind turbine blades based on pure acoustic signature signals, and to excite the acoustic wave phase synergy effect in the sensitive acoustic frequency bands to generate enhanced acoustic signature signals.

[0039] The turbulence calculation module is used to quantify the internal damage depth of the wind turbine blade material based on the enhanced acoustic signature signal, and to calculate the acoustic energy flow turbulence index.

[0040] The blade health assessment module is used to determine the blade health status of the wind turbine based on the acoustic energy flow turbulence index and through a multi-dimensional acoustic feature state space mapping mechanism, and generate a three-dimensional health assessment report.

[0041] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the wind turbine blade diagnostic method based on voiceprint perception as described in the first aspect of the present invention.

[0042] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the wind turbine blade diagnostic method based on voiceprint perception as described in the first aspect of the present invention.

[0043] The beneficial effects of this invention are as follows: by using physical-level wind noise stripping technology, based on vortex interference modeling and acoustic-airflow phase offset correction to generate a transfer function matrix, frequency domain energy reweighting decoupling is achieved, and environmental wind noise is accurately stripped in a strong turbulent environment. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart of a wind turbine blade diagnostic method based on voiceprint perception.

[0046] Figure 2 This is a schematic diagram of a wind turbine blade diagnostic system.

[0047] Figure 3 This is a flowchart of the environmental wind noise stripping process.

[0048] Figure 4 Flowchart for voiceprint signal enhancement. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] Reference Figures 1-4 This is one embodiment of the present invention, which provides a wind turbine blade diagnostic method based on voiceprint perception, including the following steps:

[0053] S1: Collect acoustic fingerprint data and airflow parameters and preprocess them, analyze the mutual characteristics of the preprocessed airflow parameters and acoustic fingerprint data, and generate a set of airflow characteristic parameters.

[0054] Acoustic data includes the acoustic characteristics of blade vibration and the components of environmental wind noise; wind flow parameters include three-dimensional wind speed vector, wind direction angle and turbulence intensity.

[0055] It should be noted that the acoustic signature data is collected using a high-sensitivity acoustic sensor array deployed on the leading edge surface of the wind turbine blades. The high-sensitivity acoustic sensor array is distributed at fixed intervals, and the sensor frequency response covers the infrasound to ultrasonic frequency bands. The acquisition process continuously records the waveform of sound pressure changing over time, and the acoustic characteristics include the complete sound pressure time-domain waveform and frequency-domain energy distribution. Airflow parameters are collected using an ultrasonic anemometer array installed at the top of the tower, employing an orthogonal ultrasonic probe group. Based on the ultrasonic time-of-flight propagation, the dynamic parameters of the spatial airflow are measured, capturing the three-dimensional wind speed vector, wind direction angle, and turbulence intensity during the acquisition process. The acoustic signature data and airflow parameters are aligned at the time of acquisition using a high-precision time synchronization device, ultimately forming a structured joint dataset of acoustic signature data and airflow parameters.

[0056] The preprocessing includes pre-emphasis filtering to suppress low-frequency noise, Hamming window framing processing, and calibration of wind flow parameters, temperature, and altitude.

[0057] It should be noted that the pre-emphasis filtering process for suppressing low-frequency noise involves applying a first-order high-pass filter to the acoustic data stream to enhance the acoustic feature energy in the high-frequency band and suppress low-frequency background noise in the environment. The Hamming window framing process involves slicing the acoustic data into segments of fixed duration, weighting them with a Hamming window to eliminate frame boundary effects, and forming a windowed framing sequence. The wind flow parameter temperature and altitude calibration operation involves correcting the sound speed measurement benchmark based on the ambient temperature sensor readings, converting the air density based on the barometer altitude data, and realizing three-dimensional wind speed calibration, turbulence intensity calibration, and wind direction angle calibration.

[0058] Based on the preprocessed airflow parameters and acoustic data, a set of airflow characteristic parameters is generated through Doppler phase shift calculation and vortex interference modeling.

[0059] Furthermore, based on the preprocessed airflow parameters and acoustic data, Doppler phase shift calculation generates the phase shift angle and the dominant frequency attenuation coefficient, and vortex interference modeling generates the vortex interference frequency and scattering intensity index. These four parameters constitute the airflow characteristic parameter set, and the entire process follows the constraints of fluid mechanics and Doppler acoustic equations.

[0060] S2: Based on the set of wind flow characteristic parameters, frequency domain compensation processing is performed to remove environmental wind noise components and generate a pure acoustic signal;

[0061] The three-dimensional motion parameters of the airflow around the fan are directly measured by an ultrasonic anemometer, and the vortex frequency is generated by calculating the periodic vortex shedding frequency based on the Strauhall law.

[0062] More specifically, the ultrasonic anemometer directly measures the three-dimensional motion parameters of the airflow around the fan. The process involves: the ultrasonic anemometer distributing ultrasonic probes in an orthogonal three-axis space, alternately emitting high-frequency ultrasonic pulses and receiving scattered echo signals; generating three-dimensional motion parameters by measuring the time difference of the pulse wave traversing a fixed acoustic path; the calculation process for the three-dimensional motion parameters: based on the time difference of the ultrasonic pulse propagation in the x, y, and z axes, the three-dimensional wind speed vector components are solved according to the sound speed temperature compensation value, while simultaneously recording the wind direction angle and turbulence intensity; vortex frequency generation: based on the Strouhal law formula, the magnitude of the three-dimensional wind speed vector is used as the input value, and a fixed Strouhal number is used to calculate the periodic vortex shedding frequency, generating the vortex frequency parameters.

[0063] Formula for calculating the frequency of periodic vortex shedding:

[0064]

[0065] Where f represents the vortex shedding frequency; S represents the Strouhal number, an empirical constant in the wind power industry; denoted as the three-dimensional wind speed vector magnitude, measured by an ultrasonic anemometer; 'a' represents the characteristic thickness of the blade.

[0066] The vibration signal of the blade is captured by an acoustic sensor array, and the significant peak of the power spectrum is extracted and the fundamental frequency of the acoustic signature is generated by pre-emphasis filtering and Mel-frequency cepstral coefficient analysis.

[0067] More specifically, the process of capturing blade vibration signals using an acoustic sensor array is as follows: An acoustic sensor array mounted on the blade surface acquires the sound pressure time-domain waveform at a fixed sampling rate, recording a mixed acoustic signal containing structural vibration and background noise; pre-emphasis filtering suppresses low-frequency noise from affecting the captured blade vibration signal, increasing the relative energy of the high-frequency vibration components; the Mel cepstral coefficient analysis process involves applying a Hamming window to the pre-emphasis filtered blade vibration signal, performing a fast Fourier transform to convert it to the frequency domain, generating a Mel spectrum through a Mel-scale filter bank, and outputting a Mel frequency cepstral coefficient sequence after logarithmic operations and discrete cosine transform; the significant peak location of the power spectrum is based on Mel spectrum analysis, identifying local maxima in the frequency-energy coordinate system, searching for the frequency corresponding to the maximum amplitude, and determining the fundamental frequency position of the acoustic signature using the energy significance criterion, ultimately generating the fundamental frequency parameter value of the acoustic signature.

[0068] Based on the set of airflow characteristic parameters, a transfer function matrix is ​​generated by frequency domain energy overlap analysis of vortex frequency and acoustic fundamental frequency, combined with acoustic wave-airflow phase shift correction mechanism.

[0069] More specifically, based on the wind flow characteristic parameter set, combined with the vortex frequency and acoustic signature fundamental frequency independently generated in the preprocess, frequency domain energy overlap analysis is performed: the cross-correlation of energy across the entire frequency band is calculated using the vortex frequency and acoustic signature fundamental frequency, and the normalized overlap coefficient is output after weighting by the dominant frequency attenuation coefficient; simultaneously, the phase offset angle of the wind flow characteristic parameter set is called to construct a complex conversion factor, which is multiplied by the scattering intensity index of the wind flow characteristic parameter set to generate a composite correction factor; the transfer function matrix generation process: a complex matrix covering the entire frequency band is constructed and initialized, the imaginary part of the overlap coefficient is superimposed at the frequency point corresponding to the vortex frequency, the real and imaginary parts of the correction factor are superimposed at the frequency point corresponding to the acoustic signature fundamental frequency, and all matrix elements are multiplied by the scattering intensity index of the wind flow characteristic parameter set, finally outputting a complex transfer function matrix.

[0070] Frequency domain compensation is performed on the transfer function matrix and the preprocessed acoustic text data. Environmental wind noise components are removed by physical-level interference field decoupling to generate an optimized frequency domain acoustic text spectrum.

[0071] More specifically, based on the transfer function matrix and the preprocessed acoustic signature data, frequency domain compensation processing is performed: a fast Fourier transform is performed on the preprocessed acoustic signature data to generate the original complex spectrum, and element-wise complex multiplication is performed between the original complex spectrum and the transfer function matrix to output the compensated complex spectrum; physical-level interference field decoupling process: the imaginary part of the compensated complex spectrum is extracted, and combined with the corresponding values ​​of the vortex interference frequency points of the transfer function matrix, the environmental wind noise component is removed from the spectrum through vortex-acoustic interference energy inverse operation; finally, an optimized frequency domain acoustic signature spectrum is generated, in which the real part spectrum characterizes the energy distribution of the acoustic characteristics of blade vibration, and the imaginary part spectrum records the phase information of pure structural vibration after phase correction.

[0072] The optimized frequency domain acoustic spectrum is inversely transformed into a time domain waveform, and the purity is verified by the acoustic energy flow entropy value to generate a pure acoustic signal.

[0073] More specifically, the frequency domain acoustic spectrum is optimized and an inverse fast Fourier transform is performed to generate a continuous time domain waveform; the purity of the acoustic energy flow is verified by calculating the acoustic energy flow probability density distribution of the continuous time domain waveform on the complete time axis and solving for the entropy value according to the Shannon entropy formula; based on the critical state judgment criterion of acoustic energy flow turbulence, if the state corresponding to the entropy value belongs to the healthy transmission range, a pure acoustic signal is output; otherwise, the process returns to the frequency domain compensation step to regenerate the transfer function matrix.

[0074] S3: Based on the pure acoustic signature signal, identify the sensitive acoustic frequency band of the wind turbine blade, and excite the acoustic wave phase coordination effect in the sensitive acoustic frequency band to generate an enhanced acoustic signature signal;

[0075] A sweep frequency acoustic excitation experiment was performed on the blade composite material specimen to identify the characteristic curve of the propagation speed of sound waves in the material as a function of frequency. After temperature and humidity compensation, the intrinsic dispersion response of the blade material was generated.

[0076] Furthermore, a swept-frequency acoustic excitation experiment was performed on the blade composite specimen: the blade composite specimen was symmetrically clamped between an acoustic transmitter and an acoustic receiver, and a linearly increasing frequency acoustic wave was generated by the acoustic transmitter in a temperature and humidity controlled environment, while the acoustic receiver recorded the propagation delay of the acoustic wave penetrating the specimen; acoustic wave propagation speed calculation: based on the basic time-distance equation, the distance between the acoustic receiver and the acoustic transmitter was compared with the acoustic wave propagation delay at each frequency point, and the original data sequence of sound speed-frequency was output; temperature and humidity compensation operation: based on the real-time ambient temperature sensor reading, linear compensation was performed on the reference sound speed according to the sound speed-temperature formula, and air density correction was performed on the relative humidity sensor value; finally, the intrinsic dispersion response of the blade material was generated.

[0077] Based on the pure acoustic signature signal, acoustic signature resonant peak features are extracted by Mel frequency cepstral coefficients, and damage feature matching is performed by combining the intrinsic dispersion response of the blade material to generate sensitive acoustic frequency bands.

[0078] Furthermore, based on the enhanced acoustic signature signal, Mel frequency cepstral coefficient extraction is performed: after applying pre-emphasis filtering to the enhanced acoustic signature signal, Hamming window framing processing is performed, and the signal is converted to the frequency domain by fast Fourier transform. The Mel energy spectrum is calculated through a Mel-scale triangular filter bank, and the Mel frequency cepstral coefficient sequence is output after logarithmic operation and discrete cosine transform. Local energy maxima in the Mel spectrum are identified as acoustic signature resonant peak features. Combined with the intrinsic dispersion response of the blade material: the characteristic frequency points of the acoustic signature resonant peak are compared with the corresponding sound velocity values ​​on the intrinsic dispersion response curve to calculate the sound velocity offset ratio. Damage feature matching process: when the sound velocity offset exceeds the preset tolerance range, the frequency band corresponding to the acoustic signature resonant peak is marked as a damage-sensitive area, and all marked frequency bands are integrated to generate sensitive acoustic frequency bands.

[0079] It should be noted that the preset tolerance range is determined based on the intrinsic sound velocity reference value calibrated by the frequency sweep experiment of healthy blades and its maximum allowable deviation under environmental fluctuations, and is set according to the importance level of the blade area.

[0080] Based on the sensitive acoustic frequency band, the frequency of the resonant unit of the acoustic metasurface array pre-installed on the surface of the wind turbine blade is dynamically adjusted and the phase spatial distribution of the sound wave front is reconstructed. The sound wave phase synergy effect is excited in the sensitive acoustic frequency band to generate a sound field focusing control signal.

[0081] Furthermore, based on the sensitive acoustic frequency band, the acoustic metasurface array pre-installed on the surface of the wind turbine blades is controlled as follows: the piezoelectric resonant unit driving the acoustic metasurface array performs dynamic frequency tuning, and the impedance matching parameters of the resonant unit are calculated in real time according to the boundary values ​​of the sensitive acoustic frequency band; wavefront reconstruction operation: based on the spatial geometric relationship between the sound wave propagation direction and the damage coordinates, the phase delay of each resonant unit is calculated, and the excitation voltage sequence of the resonant unit is generated; acoustic wave phase synergy effect excitation: in-phase step scanning voltage is applied within the sensitive acoustic frequency band, so that the output sound waves of each unit of the acoustic metasurface array form coherent superposition; finally, a sound field focusing control signal is generated.

[0082] By driving an acoustic metasurface array with a sound field focusing control signal, a sound field energy spatial focusing is formed in the sensitive acoustic frequency band, generating an enhanced acoustic pattern signal.

[0083] Furthermore, the acoustic metasurface array pre-installed on the surface of the wind turbine blades is driven by the acoustic field focusing control signal: the voltage component of the acoustic field focusing control signal is input to the power amplifier of the piezoelectric resonant unit, and the frequency component in the acoustic field focusing control signal is generated by extracting the center frequency through the boundary value of the sensitive acoustic frequency band. The frequency component is simultaneously written into the digital frequency synthesizer of the resonant unit, and the phase component is loaded to the phase shift controller. The acoustic metasurface array performs acoustic wave phase modulation response: each piezoelectric resonant unit outputs a controlled phase acoustic wave in the sensitive acoustic frequency band according to the driving signal, and the coherent superposition between the array units forms the spatial focusing of the acoustic field energy. The acoustic sensor array collects the enhanced acoustic field of the focusing area in real time, generates a time-domain sound pressure signal containing enhanced damage characteristics, and outputs an enhanced acoustic pattern signal.

[0084] S4: Based on the enhanced acoustic signature signal, quantify the internal damage depth of the wind turbine blade material and calculate the acoustic energy flow turbulence index.

[0085] Based on the enhanced acoustic signature signal, the acoustic dispersion feature spectrum is extracted through the Mel frequency cepstral coefficient. A reference benchmark is established based on the intrinsic dispersion response of the blade material, and a layered damage feature vector is generated.

[0086] Furthermore, based on the enhanced acoustic signature signal, a Mel frequency cepstral coefficient extraction of acoustic wave dispersion feature spectrum is performed: Mel frequency cepstral coefficient extraction is performed on the enhanced acoustic signature signal to construct a two-dimensional matrix of frequency-cepstral coefficient values ​​to generate an acoustic wave dispersion feature spectrum; a reference benchmark is established based on the intrinsic dispersion response of the blade material: the acoustic wave dispersion feature spectrum and the intrinsic dispersion response curve are compared to determine the sound velocity gradient distribution, and the sound velocity offset at each frequency point is calculated; a layered damage feature vector is generated: the absolute value of the sound velocity offset is statistically analyzed according to the preset frequency band, and a three-dimensional feature vector is synthesized by combining the Mel energy proportion of each frequency band.

[0087] Based on the layered damage feature vector, the material degradation index is calculated and matched with the depth-dispersion mapping to drive the Bayesian algorithm to generate a quantitative value of damage depth.

[0088] Furthermore, based on the layered damage feature vector, the material degradation index is calculated: the absolute values ​​of the sound velocity offsets in each frequency band are weighted and accumulated, and combined with the high-frequency energy ratio nonlinear correction factor, the material degradation index is output; the depth-dispersion mapping matching process is implemented: the material degradation index is input into the pre-stored depth-dispersion relationship database (by performing artificial damage calibration at different depths on standard composite material specimens through laboratory frequency-sweeping acoustic excitation experiments, and collecting dispersion response curves at each damage depth), the depth gradient curve corresponding to the closest material degradation mode is matched, and the depth value of the dispersion feature point corresponding to the current feature vector is read; the Bayesian algorithm is driven to generate a composite parameter package: a Bayesian conditional probability model is constructed, the layered damage feature vector and the depth gradient curve are used as observation evidence, the most likely damage depth value and spatial coordinates are solved by maximizing the posterior probability, and a composite parameter package containing the quantified damage depth value and the three-dimensional coordinates of the damage is encapsulated and generated.

[0089] Based on the enhanced acoustic signature signal and combined with the damage depth quantization value, the energy transfer efficiency of each frame of enhanced acoustic signature signal in the damaged area is calculated to generate an acoustic energy flow turbulence index.

[0090] Furthermore, based on the enhanced acoustic signature signal and the quantized damage depth value, the energy transfer efficiency of the damaged area is calculated: the enhanced acoustic signature signal is processed in frames, and the frequency domain energy spectrum is extracted by fast Fourier transform; the focal point of the acoustic energy flow in the damaged area is located: the frame energy value of the focal area is extracted according to the position index of the three-dimensional coordinates of the damage in the acoustic sensor array; the energy transfer efficiency is calculated: the frame energy of the focal area is compared with the overall average energy of the same frame, and the single frame efficiency value is output; the standard deviation of efficiency fluctuation is calculated by integrating the frame sequence of the whole time period, and combined with the nonlinear amplification effect of the quantized damage depth value, an acoustic energy flow turbulence index is generated.

[0091] Formula for calculating the acoustic energy flow turbulence index of wind turbine blades:

[0092]

[0093] Where Γ represents the acoustic energy flow turbulence index; N represents the total number of frames; k represents the depth nonlinear amplification factor; d represents the damage depth quantization value; η represents the single-frame energy transfer efficiency; and μ represents the average efficiency.

[0094] S5: Based on the acoustic energy flow turbulence index, the health status of the wind turbine blades is determined through a multi-dimensional acoustic characteristic state space mapping mechanism, and a three-dimensional health assessment report is generated.

[0095] Based on the acoustic energy flow turbulence index and damage depth quantification value, combined with the three-dimensional coordinates of damage, a decision ratio is allocated through weight allocation rules to generate a scatter plot of the blade zone health status.

[0096] More specifically, based on the acoustic energy flow turbulence index, the damage depth quantification value, and the three-dimensional coordinates of the damage, the blade surface is divided into grid zones; the location risk weight coefficient is determined as follows: if the three-dimensional coordinates of the damage are located in the blade root flange connection area, the location risk weight is adjusted to 15%; if it is located in the leading edge aerodynamic sensitive area, the location risk weight is adjusted to 20%; and the location risk weight remains at 10% in other areas; the weight allocation rule is executed as follows: the acoustic energy flow turbulence index is allocated a weight of 60%, the damage depth quantification value a weight of 30%, and the location risk weight a weight of 10%, and the weighted sum is used to generate the zone health index; all damage points are labeled with health index values ​​according to the zone coordinates, and a scatter plot of the blade zone health status is generated.

[0097] Based on the scatter plot of the health status of the blade partition, the cluster density of health status points in each partition is calculated, and health status labels are generated by combining the critical state of acoustic energy flow turbulence.

[0098] More specifically, a kernel density estimation algorithm is used to count the number of grid points per unit area and output a normalized density index. Combined with the critical state of acoustic energy flow turbulence: when the density index exceeds the characteristic boundary value (determined based on the statistical distribution of spatial point density of healthy blades, representing the critical density index for transitioning from a random discrete distribution to a damage aggregation state) and the weighted mean of the health index crosses the critical characteristic of damage intensity (quasi-calibrated based on the yield strength experiment of blade materials, representing the value of the weighted mean of the health index exceeding the critical point of elastic deformation of the material), a banded aggregation judgment is triggered, matching the critical state classification rules to generate a health state label; finally, the health state label is output.

[0099] Formula for calculating the cluster density of healthy state points:

[0100]

[0101] in, The cluster density index of the m-th partition is represented by F; F represents the smooth decay function; x m p represents the coordinates of the center of the m-th partition; j The coordinates of the j-th healthy state point are represented by h; the kernel density bandwidth is represented by h; and the number of points in the current partition is represented by n.

[0102] A spatial-temporal-state 3D fusion rendering engine is used to perform 3D fusion rendering of health status labels and scatter plots of regional situations, generating a 3D health assessment report.

[0103] More specifically, the spatial dimension maps health status labels onto the surface of the 3D model of the wind turbine blades, generating a damage thermal topology map; the temporal dimension calls upon a historical acoustic energy flow turbulence index database, fits an acoustic energy flow turbulence decay curve, and labels the predicted remaining lifespan of the material; the state dimension sorts the health index values ​​according to the scatter plot of the regional situation, generating a maintenance priority list; the 3D engine synchronously drives the spatial thermal map, time lifespan curve, and state priority table to encapsulate and generate a 3D health assessment report, ultimately outputting a 3D health assessment package.

[0104] This embodiment also provides a wind turbine blade diagnostic system based on voiceprint perception, including:

[0105] The data preprocessing module is used to collect acoustic fingerprint data and airflow parameters and preprocess them, analyze the mutual characteristics between the preprocessed airflow parameters and acoustic fingerprint data, and generate a set of airflow characteristic parameters.

[0106] The environmental wind noise stripping module is used to perform frequency domain compensation processing based on the wind flow characteristic parameter set, strip away the environmental wind noise components, and generate a pure acoustic signal.

[0107] The acoustic signature enhancement module is used to identify the sensitive acoustic frequency bands of wind turbine blades based on pure acoustic signature signals, and to excite the acoustic wave phase synergy effect in the sensitive acoustic frequency bands to generate enhanced acoustic signature signals.

[0108] The turbulence calculation module is used to quantify the internal damage depth of the wind turbine blade material based on the enhanced acoustic signature signal, and to calculate the acoustic energy flow turbulence index.

[0109] The blade health assessment module is used to determine the blade health status of the wind turbine based on the acoustic energy flow turbulence index and through a multi-dimensional acoustic feature state space mapping mechanism, and generate a three-dimensional health assessment report.

[0110] This embodiment also provides a computer device applicable to the wind turbine blade diagnosis method based on voiceprint perception, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the wind turbine blade diagnosis method based on voiceprint perception as proposed in the above embodiment.

[0111] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0112] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the wind turbine blade diagnostic method based on voiceprint perception as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0113] In summary, this invention utilizes physical-level wind noise stripping technology, based on vortex interference modeling and acoustic-airflow phase offset correction to generate a transfer function matrix, thereby achieving frequency domain energy reweighting decoupling and accurately stripping environmental wind noise in strong turbulent environments.

[0114] 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 it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for fan blade diagnostics based on voiceprint perception, characterized in that: The method comprises the following steps: Collecting and preprocessing the voiceprint data and wind flow parameters, analyzing the mutual characteristics of the preprocessed wind flow parameters and voiceprint data, and generating a wind flow characteristic parameter set; The voiceprint data includes blade vibration acoustic characteristics and environmental wind noise components, and the wind flow parameters include three-dimensional wind speed vector, wind direction angle and turbulence intensity; The preprocessing includes pre-emphasis filtering to suppress low-frequency noise, Hamming window framing processing and wind flow parameter temperature and altitude calibration; Based on the wind flow characteristic parameter set, frequency domain compensation processing is performed to strip the environmental wind noise components and generate a pure voiceprint signal, which comprises the following steps: Directly measure the three-dimensional motion parameters of the airflow around the fan by using an ultrasonic anemometer, calculate the periodic vortex shedding frequency based on the Strouhal law to generate the vortex frequency; Capture the blade vibration signal by using an acoustic sensor array, perform pre-emphasis filtering and Mel cepstrum coefficient analysis to extract the power spectrum significant peak value positioning to generate the voiceprint fundamental frequency; Based on the wind flow characteristic parameter set, the frequency energy overlap analysis of the vortex frequency and the voiceprint fundamental frequency is performed, and the sound wave-airflow phase offset correction mechanism is combined to generate a transfer function matrix; Perform frequency domain compensation processing on the transfer function matrix and the preprocessed voiceprint data, strip the environmental wind noise components by using physical level interference field decoupling to generate an optimized frequency domain voiceprint spectrum; Inverse transform the optimized frequency domain voiceprint spectrum into a time domain waveform, verify the purity by using the acoustic energy flow entropy value, and generate a pure voiceprint signal; Based on the pure voiceprint signal, identify the sensitive acoustic frequency band of the fan blade, excite the sound wave phase synergy effect in the sensitive acoustic frequency band, and generate a strengthened voiceprint signal, which comprises the following steps: Perform a sweep frequency sound wave excitation experiment on the blade composite material specimen, identify the characteristic curve of the sound wave propagation speed in the material with the change of frequency, and generate the intrinsic dispersion response of the blade material after temperature and humidity compensation; Based on the pure voiceprint signal, extract the voiceprint resonance peak characteristics by using the Mel frequency cepstrum coefficient, and perform damage feature matching combined with the intrinsic dispersion response of the blade material to generate the sensitive acoustic frequency band; Based on the sensitive acoustic frequency band, control the dynamic adjustment of the resonance unit frequency of the acoustic metasurface array preinstalled on the surface of the fan blade and the reconstruction of the phase space distribution of the sound wave front to excite the sound wave phase synergy effect in the sensitive acoustic frequency band, and generate a sound field focusing control signal; Drive the acoustic metasurface array by using the sound field focusing control signal to form a sound field energy space focus in the sensitive acoustic frequency band, and generate a strengthened voiceprint signal; According to the strengthened voiceprint signal, quantify the internal damage depth of the fan blade material, and calculate the acoustic energy flow disorder degree index; According to the acoustic energy flow disorder degree index, perform blade health state judgment of the fan through a multi-dimensional acoustic feature state space mapping mechanism to generate a three-dimensional health evaluation report.

2. The acoustic profile aware fan blade diagnostic method of claim 1, wherein: The analysis of the mutual characteristics of the preprocessed wind flow parameters and voiceprint data to generate the wind flow characteristic parameter set is specifically performed by Doppler phase shift calculation and vortex interference modeling to generate the wind flow characteristic parameter set.

3. The acoustic signature aware fan blade diagnostic method of claim 1, wherein: According to the strengthened voiceprint signal, the internal damage depth of the fan blade material is quantified, and the acoustic energy flow disorder degree index is calculated, which comprises the following steps: According to the strengthened voiceprint signal, extract the sound wave dispersion characteristic spectrum by using the Mel frequency cepstrum coefficient, establish a reference benchmark according to the intrinsic dispersion response of the blade material, and generate a hierarchical damage feature vector. Based on the hierarchical damage feature vector, the material degradation index is calculated and matched through the depth-dispersion mapping to drive the Bayesian algorithm to generate the damage depth quantization value; Based on the enhanced voiceprint signal, combined with the damage depth quantization value, the energy transfer efficiency of each frame of enhanced voiceprint signal in the damage area is calculated to generate the acoustic energy flow disorder degree index.

4. The acoustic signature aware fan blade diagnostic method of claim 3, wherein: According to the acoustic energy flow disorder degree index, the fan blade health state is determined through a multi-dimensional acoustic feature state space mapping mechanism to generate a three-dimensional health evaluation report, and the steps are as follows: Based on the acoustic energy flow disorder degree index and the damage depth quantization value, combined with the damage three-dimensional coordinates, the decision proportion is allocated through the weight allocation rule to generate a blade partition health situation scatter plot; Based on the blade partition health situation scatter plot, the aggregation density of the health state points in each partition is calculated, combined with the critical state of the acoustic energy flow disorder degree, to generate a health state label; A space-time-state three-dimensional fusion rendering engine is used to perform three-dimensional fusion rendering on the health state label and the partition situation scatter plot to generate a three-dimensional health evaluation report.

5. A system for fan blade diagnostics based on voiceprint perception, based on the method for fan blade diagnostics based on voiceprint perception according to any one of claims 1-4, characterized in that: It includes: A data preprocessing module for collecting and preprocessing voiceprint data and wind flow parameters, analyzing the mutual characteristics of the preprocessed wind flow parameters and voiceprint data, and generating a wind flow characteristic parameter set; An environmental wind noise stripping module for performing frequency domain compensation processing based on the wind flow characteristic parameter set to strip the environmental wind noise component and generate a pure voiceprint signal; A voiceprint signal enhancement module for identifying the sensitive acoustic frequency band of the fan blade based on the pure voiceprint signal, exciting the acoustic phase synergy effect in the sensitive acoustic frequency band, and generating an enhanced voiceprint signal; A disorder degree calculation module for quantizing the internal damage depth of the fan blade material according to the enhanced voiceprint signal and calculating the acoustic energy flow disorder degree index; A blade health determination module for determining the health state of the fan blade according to the acoustic energy flow disorder degree index through a multi-dimensional acoustic feature state space mapping mechanism to generate a three-dimensional health evaluation report. 6.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the fan blade diagnosis method based on voiceprint sensing according to any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the fan blade diagnosis method based on voiceprint sensing according to any one of claims 1-4.

Images