CN116153333A - Wind turbine blade fault diagnosis method based on aerodynamic noise - Google Patents
Wind turbine blade fault diagnosis method based on aerodynamic noise Download PDFInfo
- Publication number
- CN116153333A CN116153333A CN202310073354.8A CN202310073354A CN116153333A CN 116153333 A CN116153333 A CN 116153333A CN 202310073354 A CN202310073354 A CN 202310073354A CN 116153333 A CN116153333 A CN 116153333A
- Authority
- CN
- China
- Prior art keywords
- data
- wind turbine
- noise
- turbine blade
- fault
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000003745 diagnosis Methods 0.000 title claims abstract description 20
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 23
- 230000005540 biological transmission Effects 0.000 claims abstract description 21
- 230000009467 reduction Effects 0.000 claims abstract description 14
- 238000001914 filtration Methods 0.000 claims abstract description 11
- 230000005236 sound signal Effects 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 14
- 238000001228 spectrum Methods 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 11
- 238000004088 simulation Methods 0.000 claims description 9
- 230000007547 defect Effects 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 6
- 230000003993 interaction Effects 0.000 claims description 5
- 239000004576 sand Substances 0.000 claims description 5
- 238000005299 abrasion Methods 0.000 claims description 4
- 238000012423 maintenance Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000005336 cracking Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 7
- 230000025518 detection of mechanical stimulus involved in sensory perception of wind Effects 0.000 abstract 1
- 238000007689 inspection Methods 0.000 description 9
- 238000009826 distribution Methods 0.000 description 8
- 239000000428 dust Substances 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D80/00—Details, components or accessories not provided for in groups F03D1/00 - F03D17/00
- F03D80/40—Ice detection; De-icing means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/06—Transformation of speech into a non-audible representation, e.g. speech visualisation or speech processing for tactile aids
- G10L21/10—Transforming into visible information
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/72—Wind turbines with rotation axis in wind direction
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Acoustics & Sound (AREA)
- Human Computer Interaction (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Sustainable Development (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a wind turbine blade fault diagnosis method based on aerodynamic noise in the field of blade fault detection of wind turbines, which comprises the steps of filtering and denoising sample historical data of the wind turbine blade, converting the sample historical data into image data, constructing a convolutional neural network model, establishing a training set, inputting the image data into the convolutional neural network model for pre-training, wherein the training set comprises the sample historical data and expansion data; acquiring pneumatic noise of the wind turbine blade to be detected as an acoustic signal, summarizing the acoustic signal to a server of a data acquisition and transmission center, and transmitting the data to a background server of a workstation through a network protocol; and (3) the data input into the background server are converted into image data after noise reduction, and then are input into a convolutional neural network model, and the fault type of the wind turbine blade is determined through fault characteristics. The invention has simple operation, can know the condition of the wind turbine blade in time, and can improve the failure detection efficiency of the wind turbine blade.
Description
Technical Field
The invention relates to the technical field of blade fault detection and machine learning of wind driven generators, in particular to a wind driven generator blade fault diagnosis method based on aerodynamic noise, which is used for judging faults such as blade tip breakage, blade icing, trailing edge cracking, leading edge abrasion, surface dirt and the like.
Background
Wind energy is becoming increasingly important worldwide as a clean renewable energy source. The installed capacity and the duty ratio of China are increased year by year, and the development potential of the wind power industry is huge. However, due to the influences of layout positions, environmental factors and long-time operation of the wind turbine, the wind turbine blade is easily damaged, and the problems of blade tip damage, blade icing, trailing edge cracking, surface dirt and the like can be caused. If the problem cannot be found in time, further damage to the wind turbine blade can be caused.
Currently, a wind farm generally adopts a manual periodic inspection mode, but the method has a certain problem. For example, with the continuous input of the wind power plant, the wind power plant is larger and larger in scale, and part of wind turbines are installed in remote positions. If a manual inspection mode is adopted, a great amount of time and energy are consumed, and the inspection efficiency is low; meanwhile, manual regular inspection can cause untimely problem discovery. At present, although unmanned aerial vehicle inspection exists, unmanned aerial vehicle inspection is adopted, and the unmanned aerial vehicle inspection is influenced by inspection distance and flight time and can only be used in a small range. How to improve inspection efficiency and discover problems in time is a technical problem to be solved in the management of the wind power generator at present.
Disclosure of Invention
The invention aims at overcoming the technical defects in the background, and provides a wind turbine blade fault diagnosis method based on aerodynamic noise, which can be used for rapidly detecting the state of the existing wind turbine blade and timely finding out the blade fault and the fault type of the wind turbine.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a method of wind turbine blade fault diagnosis based on aerodynamic noise, the method comprising:
(1) Sample historical data and labels of pneumatic noise of a target wind turbine blade are converted into image data after noise reduction treatment, a convolutional neural network model is built, a training data set is built, and the image data is input into the convolutional neural network model for pre-training;
(2) Acquiring operation parameters, atmospheric parameters and aerodynamic noise of different wind turbine blades of the same wind farm in a fixed time scale as acquired acoustic signals; the sound signals are summarized to a server of a data acquisition and transmission center, and data are transmitted to a background server of a workstation through a network protocol; acquiring aerodynamic noise data of a wind turbine blade to be detected in a background server, converting the aerodynamic noise data into image data after noise reduction treatment, inputting a convolutional neural network model, outputting corresponding characteristic values, and displaying and judging the fault type of the wind turbine blade according to sample historical data of aerodynamic noise of the wind turbine blade and corresponding fault labels in a man-machine interaction operation interface;
the training data set of the convolutional neural network includes sample history data and expansion data, the sample history data is actually detected sample history data, but because the sample history data is often not enough and sufficient, more data is needed, and therefore the expansion data is needed, and the expansion data includes simulation data and data with increased data sets for the sample history data.
Further, the fault type corresponding to the fault label at least comprises one of icing, crack or defect, front edge abrasion and surface sand hole, the acquisition is to perform unsupervised learning on the existing faults in a local constraint sparse self-encoder mode, a plurality of categories are separated, and then 5 sections of audio in each category are randomly selected, so that the fault type of each category is judged, and the corresponding fault label is obtained.
The simulation data acquisition method in the invention comprises the following steps:
the method comprises the steps of adjusting wing profile data of a wind turbine of a wind power plant by sample historical data according to wing profile data of the wind turbine, CST class-shape functions and Latin hypercube sampling, and increasing or decreasing the wing profile at the position 80% away from a blade root to realize new modeling, so that new wing profile data and corresponding labels are obtained, and a data set under the wing profile is enlarged; and inputting the new wing profile data obtained after adjustment into a openfast, bladed or HAWC2 wind turbine generator simulation platform, and setting corresponding parameters to obtain aerodynamic noise data at the observation point. The method has the advantages that the data set is expanded on the basis of the limited data set, so that on one hand, the distribution type of the samples is wider, the overfitting of the data is avoided, on the other hand, the number of the samples which can be trained is increased due to the increase of the data, and the more information can be learned by a computer, so that the convergence is improved.
Further, the acquired pneumatic noise data is acquired in a designated acquisition time period of a sunny day through an acoustic signal acquisition and transmission device; the rainproof and dustproof cover is opened in a rainy day or a non-appointed time period, so that the safety of equipment is protected, and the power supply mode of the rainproof and dustproof cover can be powered by a solar panel at the top of the rainproof cover.
Further, the noise reduction is performed on the sample history data, and when the sample history data is converted into image data, the method sequentially comprises the following sub-steps:
(1-1) filtering the sample history data by means of a butterworth band-pass filter to remove background noise and noise generated during transmission and conversion;
(1-2) performing secondary filtering treatment by using empirical wavelet transformation, and filtering mechanical noise generated in the wind turbine to obtain noise-reduced pneumatic noise data;
(1-3) converting the noise-reduced aerodynamic noise data into image data by means of mel frequency spectrum.
In the invention, the data added to the data set of the sample historical data is the data obtained by carrying out inversion and image contrast adjustment on the image data obtained after noise reduction and Mel frequency spectrum, so as to obtain a new data expansion original training set and obtain the data by improving convergence.
Further, the step (2) may include the following sub-steps:
(2-1) acquiring aerodynamic noise of a wind turbine blade to be detected, transmitting an acoustic signal to a background server through a network protocol through a data acquisition and transmission center, reducing noise through a Butterworth band-pass filter and empirical wavelet variation, and converting the noise into image data through a Mel frequency spectrum;
(2-2) inputting the image data into a convolutional neural network model for classification, and outputting corresponding characteristic values by the convolutional neural network model according to fault classification in pre-training;
and (2-3) judging whether the characteristic value is within a preset threshold range, classifying according to the characteristic value if the characteristic value is within the preset threshold range, and judging that the characteristic value cannot be determined if the characteristic value exceeds the preset range.
And when the characteristic value exceeds the preset range and the judgment result is not confirmed, manually confirming the characteristic value, calling and displaying an audio signal and a time-frequency diagram which cannot confirm the fault through a computer operation interface of the workstation, judging the type of fault sound at the workstation by an operation and maintenance personnel according to the sound signal and the time-frequency diagram of the fault, checking on site when the type of the fault sound cannot be judged, and storing the result and correcting the original classification result after the fault result is obtained. The fault identification is more accurate by adopting an unsupervised and manual identification combined mode.
The picture data converted by the aerodynamic noise and the wind turbine blade fault type classification result are stored in a background server. Further, the stored picture data and the corrected classification result are input into a convolutional neural network to be used as a training set, and then the convolutional neural network model is updated periodically.
Compared with the prior art, the invention has the beneficial effects that: according to the invention, pneumatic noise of the wind turbine blade is taken as a signal source, a training set of the convolutional neural network is established through sample historical data of samples and expanded sample data, and then noise data of the wind turbine blade to be judged is acquired and is input into the convolutional neural network, so that the type of fault is judged. For example, in the case of ice-making in an icing state, the energy of the acoustic signal is mainly concentrated at 605Hz and 190Hz, and in the case of frosted ice, the energy of the acoustic signal is mainly concentrated at 108Hz and 830Hz; in a crack or defect fault, the probability of damage to a single blade is the greatest, a high-frequency signal is generated, the center of gravity of a frequency spectrum becomes large, and the larger energy in a sound signal shifts to a high frequency band, namely the measured frequency is larger than the frequency under a normal working condition. The energy of the acoustic signal is mainly distributed at 1-2kHz when the leading edge wears. When the sand holes are defective, the energy of the sound signals is mainly distributed at 6-8kHz; the distribution frequencies corresponding to different fault conditions are different, so that the distribution frequencies can be distinguished by combining the time domain, the frequency domain and the energy of the acoustic signals. Unnecessary loss caused by sudden events and workload of human judgment can be reduced, and the method has the advantages of high judgment speed, simplicity in operation, low time cost and labor cost and the like.
Drawings
FIG. 1 is a schematic diagram of a wind turbine blade fault diagnosis method based on aerodynamic noise according to an embodiment of the present invention.
Fig. 2 is a schematic top view of a microphone according to an embodiment of the present invention.
Fig. 3 is a schematic view of a rain cover for acoustic signal acquisition according to an embodiment of the present invention.
FIG. 4 is a block diagram of a wind turbine and acoustic signal acquisition provided by an embodiment of the present invention.
Fig. 5 is a training flowchart of blade state detection according to an embodiment of the present invention.
In the figure, 1-wind turbine blade; 2-a tower of the wind turbine; 3-a tower foundation of the wind turbine; 4-a data centralizing module and a transmission assembly; 5-acquisition and transmission module: a 101-microphone; 102-a data transmission module; 103-a standby capacity expansion module; 6, wall surface; 7-solar panel.
Description of the embodiments
In order to make the objects, technical solutions and effects of the present invention clearer and more specific, the present invention will be described in further detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The wind turbine blade fault diagnosis method based on aerodynamic noise provided by the invention can be applied to a terminal with computing capability and interaction capability, and the terminal can execute the wind turbine blade fault diagnosis method based on aerodynamic noise provided by the invention to check the blade state.
According to the wind turbine blade state detection system based on the pneumatic noise convolutional neural network, as shown in the figure, a wind turbine tower drum 2 is arranged on a wind turbine tower foundation 3, a wind turbine blade 1 and a generator body are arranged at the top of the wind turbine tower drum 2, a microphone 101 is used for collecting sound signals, the sound signals are transmitted to a collecting and transmitting module 5, and then the sound signals are connected to a background server through a data concentration module and a transmitting assembly 4.
Fig. 2 shows a noise signal acquisition device, in which a microphone 101 is used for acquiring a sound signal and transmitting the sound signal to an acquisition and transmission module 5 through a data transmission module 102; the spare capacity expansion module 103 is used for expanding capacity; the noise signal acquisition device is arranged on the wall surface 6 of the wind turbine tower 2 and can supply power through the solar panel 7.
In this embodiment, the microphone 101 is installed at a height of 850-950 mm from the ground, so as to avoid that the reflection of sound is large when the microphone is too close to the ground, and the microphone shakes much when the microphone is too high, so that the data error is enlarged; the distance from the tower drum to the tower drum is 4m, so that the sound wave reflection caused by the fact that the tower drum is too close to the wind turbine is avoided; when the data acquisition and transmission module rises and the power supply to the acquisition device is interrupted when the specified acquisition time is not met in a sunny day, the safety of equipment is ensured, the rain and dust covers firstly rise the baffle plate from the periphery and then pass through the rotating structure at the upper end of the baffle plate farthest from the tower barrel, the top surface baffle plate is rotated to seal, the baffle plate is a solar panel, and meanwhile, the whole data acquisition and transmission module and the rain and dust cover adopt inclined structures, so that the problem of accumulation of rainwater and dust at the top is avoided on the basis of effectively preventing the damage of the data acquisition device caused by rainwater and dust, and the rain and dust covers can be collected when the specified acquisition time is in a sunny day, and the top surface baffle plate can convert solar energy into electric energy to be stored in a storage battery for the device to use, thereby realizing self-production and self-use, and improving the utilization rate of resources, as shown in fig. 3; simultaneously, three microphones are adopted, so that inaccurate experimental data obtained by measurement is avoided because a yaw single microphone is possibly located at the position where pneumatic noise of the wind turbine is minimum, the three microphones are circularly distributed about a tower of the wind turbine, an included angle of 120 degrees is formed between every two microphones, and the microphones are aligned to a fan blade, the data of the pneumatic noise of the wind turbine can be more comprehensively obtained, and meanwhile, the sampling rate of the microphones is 2.5kHz.
The three microphones are set to be a circle in collection period, the collection time is set to be 8 points, 14 points and 20 points on the same day respectively, collection is set to be 3 minutes, then the three microphones intercept noise data of 30 seconds in each section of collection time randomly, and meanwhile, operation parameters and atmosphere parameters of the wind turbine blade in the collection time are recorded.
The collected data are collected to a data collection and transmission center in a unified way, and when the collection period reaches the designated time, the collected data are transmitted to a background server in a unified way through a network protocol.
The wind turbine blade fault diagnosis method based on aerodynamic noise by the device comprises the following steps:
(1) Sample historical data and labels of pneumatic noise of a target wind turbine blade are converted into image data after noise reduction treatment, a convolutional neural network model is built, a training data set is built, and the image data is input into the convolutional neural network model for pre-training;
(2) Acquiring operation parameters, atmospheric parameters and aerodynamic noise of different wind turbine blades of the same wind farm in a fixed time scale as acquired acoustic signals; the sound signals are summarized to a server of a data acquisition and transmission center, and data are transmitted to a background server of a workstation through a network protocol; acquiring aerodynamic noise data of a wind turbine blade to be detected in a background server, converting the aerodynamic noise data into image data after noise reduction treatment, inputting a convolutional neural network model, outputting corresponding characteristic values, and displaying and judging the fault type of the wind turbine blade according to sample historical data of aerodynamic noise of the wind turbine blade and corresponding fault labels in a man-machine interaction operation interface;
the training data set of the convolutional neural network includes sample history data and expansion data, the sample history data is actually detected sample history data, but because the sample history data is often not enough and sufficient, more data is needed, and therefore the expansion data is needed, and the expansion data includes simulation data and data with increased data sets for the sample history data.
Further, the fault type corresponding to the fault label at least comprises one of icing, crack or defect, front edge abrasion and surface sand hole, the acquisition is to perform unsupervised learning on the existing faults in a local constraint sparse self-encoder mode, a plurality of categories are separated, and then 5 sections of audio in each category are randomly selected, so that the fault type of each category is judged, and the corresponding fault label is obtained.
The simulation data acquisition method comprises the following steps:
the method comprises the steps of adjusting wing profile data of a wind turbine of a wind power plant by sample historical data according to wing profile data of the wind turbine, CST class-shape functions and Latin hypercube sampling, and increasing or decreasing the wing profile at the position 80% away from a blade root to realize new modeling, so that new wing profile data and corresponding labels are obtained, and a data set under the wing profile is enlarged; and inputting the new wing profile data obtained after adjustment into a openfast, bladed or HAWC2 wind turbine generator simulation platform, and setting corresponding parameters to obtain aerodynamic noise data at the observation point.
The CST class function uses a class function and a shape function to describe the geometry of the airfoil, which gives a highly accurate fitted curve with few parameters. X is the abscissa, y is the ordinate, y TE The CST class function, expressed as the ordinate of the trailing edge, is given by:
wherein C (x) is a class function, N1, N2 take different values to construct different geometric shapes, which is specified by the national aviation consultation committee of america, and n1=0.5 and n2=1.0 are taken as the class functions of the tip-tip airfoil; a double-arc airfoil profile, n1=1.0, n2=1.0; s (x) is a shape function, ai represents an introduced weight factor, called shape function coefficient, i=0, 1, ·n; s (x) is a shape function, is a weighted sum of Bernstein polynomials of n orders, and can obtain CST parametric curves with different precision by adjusting the orders of the polynomials.
Latin hypercube sampling is a hierarchical sampling method, which first splits the design space of the variables into N equidistant intervals, then selects a random data point in each interval, then there are N data points for each variable, and then combines the data points randomly. The advantages are that: the sample points can be uniformly distributed in the design space regardless of the variable dimension of the optimization problem.
The method has the advantages that the data set is expanded on the basis of the limited data set, so that on one hand, the distribution type of the samples is wider, the overfitting of the data is avoided, on the other hand, the number of the samples which can be trained is increased due to the increase of the data, and the more information can be learned by a computer, so that the convergence is improved.
Further, the acquired pneumatic noise data is acquired in a designated acquisition time period of a sunny day through an acoustic signal acquisition and transmission device; the rainproof and dustproof cover is opened in a rainy day or a non-appointed time period, so that the safety of equipment is protected, and the power supply mode of the rainproof and dustproof cover can be powered by a solar panel at the top of the rainproof cover.
Further, the noise reduction is performed on the sample history data, and when the sample history data is converted into image data, the method sequentially comprises the following sub-steps:
(1-1) filtering the sample history data by means of a butterworth band-pass filter to remove background noise and noise generated during transmission and conversion;
(1-2) performing secondary filtering treatment by using empirical wavelet transformation, and filtering mechanical noise generated in the wind turbine to obtain noise-reduced pneumatic noise data;
(1-3) converting the noise-reduced aerodynamic noise data into image data by means of mel frequency spectrum.
Wherein the butterworth band-pass filter is adopted in consideration of that the frequency response of the butterworth filter shows stable characteristics both inside and outside the passband, the main purpose is to filter out low-frequency wind noise, ensure that the signal does not attenuate in the fault frequency range as much as possible, and simultaneously, in consideration of the fault type, the lower limit cut-off frequency is set to 100Hz and the upper limit cut-off frequency is set to 20kHz.
The empirical wavelet transformation is to divide the frequency spectrum into N frequency bands according to a division algorithm, construct a filter bank according to the boundary coordinates of the frequency bands, filter the frequency bands to obtain component signals through wavelet basis functions, and reconstruct the signals to obtain filtered aerodynamic noise data.
And the data obtained after filtering and the simulated capacity expansion data are introduced into a Mel spectrogram, so that the audio signal is converted into a time-frequency diagram of an image, and the distribution of aerodynamic noise in a time domain, a frequency domain and energy can be better observed in the time-frequency diagram, thereby being beneficial to the manual judgment of faults. The Mel spectrogram is to pre-emphasis and frame the input signal, divide the original signal into several small blocks according to time, each block is a frame; adding a window function to each frame to obtain a better sidelobe reduction amplitude; then, carrying out Fourier transform on each frame to obtain an energy spectrum; and finally, applying the Mel filter to the energy spectrum obtained in the last step to obtain a time-frequency diagram based on the Mel spectrogram.
In the invention, the data added to the data set of the sample historical data is the data obtained by carrying out inversion and image contrast adjustment on the image data obtained after noise reduction and Mel frequency spectrum, so as to obtain a new data expansion original training set and obtain the data by improving convergence.
Further, the step (2) may include the following sub-steps:
(2-1) acquiring aerodynamic noise of a wind turbine blade to be detected, transmitting an acoustic signal to a background server through a network protocol through a data acquisition and transmission center, reducing noise through a Butterworth band-pass filter and empirical wavelet variation, and converting the noise into image data through a Mel frequency spectrum;
(2-2) inputting the image data into a convolutional neural network model for classification, and outputting corresponding characteristic values by the convolutional neural network model according to fault classification in pre-training;
and (2-3) judging whether the characteristic value is within a preset threshold range, classifying according to the characteristic value if the characteristic value is within the preset threshold range, and judging that the characteristic value cannot be determined if the characteristic value exceeds the preset range.
The acoustic signals obtained by different fault types have larger difference, and the energy of the acoustic signals is mainly concentrated at 605Hz and 190Hz when the ice is frozen, and the energy of the acoustic signals is mainly concentrated at 108Hz and 830Hz when the ice is frosted; in a crack or defect fault, the probability of damage to a single blade is maximum, a high-frequency signal can be generated, the frequency spectrum center of gravity is enlarged, and the larger energy in the sound signal is shifted to a high frequency band, namely the measured frequency is larger than that under a normal working condition; when the front edge is worn, the energy of the acoustic signal is mainly distributed at 1-2kHz; in sand holes, the energy of the acoustic signals is mainly distributed at 6-8kHz; the distribution frequencies corresponding to different fault conditions are different, so that the distribution frequencies can be distinguished by combining the time domain, the frequency domain and the energy of the acoustic signals.
The frequency distribution of the various faults is shown in table 1:
the cracks or defects include various lines of cracks, blade breakage, and the like generated in the blade.
The obtained image data of the aerodynamic noise of the wind turbine blade to be tested is input into a fault type judging module in an acoustic signal processing module, so that a characteristic value is obtained through a convolutional neural network model, whether the characteristic value is in a preset threshold range or not is judged, if the characteristic value is in the preset range, classification is carried out according to the characteristic value, if the characteristic value exceeds the preset range, the characteristic value is considered to be indeterminate, further manual confirmation is needed, an audio signal and a time-frequency diagram which cannot be used for determining faults can be retrieved and displayed through a computer operation interface of a workstation, an operation staff judges the type of fault sound in the workstation according to the fault acoustic signal and the time-frequency diagram, and when the operation staff still cannot judge, the operation staff needs to check on site, and after the fault result is obtained, the result is stored and the original classification result is corrected;
inputting the judgment result to a man-machine interaction operation interface;
if 5 fault types exist, the generated characteristic value is a range value, and is corresponding to 0-5 with 6 numbers, 0 is normal, the fault with the number 1 is the detected characteristic value (0.6,1.4), the fault with the number 2 is the detected characteristic value (1.6,2.4), and the like; if the detection feature values are [1.4,1.6], [2.4,2.6], the detection feature values cannot be judged, that is, the detection feature values exceed the threshold range, and therefore the classification result needs to be manually confirmed and corrected at the operation interface.
The pneumatic noise audio frequency, the picture data and the classification result are stored.
The stored picture data and the corrected classification result are also input into a convolutional neural network as a training set, a database is supplemented and perfected, the convolutional neural network model is updated, the database tends to be complete after the wind field operates for one year, most of noise information can be judged, and full-automatic fault monitoring is realized.
The invention is not limited to the above embodiments, and based on the technical solution disclosed in the invention, a person skilled in the art may make some substitutions and modifications to some technical features thereof without creative effort according to the technical content disclosed, and all the substitutions and modifications are within the protection scope of the invention.
Claims (10)
1. A wind turbine blade fault diagnosis method based on aerodynamic noise is characterized by comprising the following steps:
(1) Sample historical data and labels of pneumatic noise of a target wind turbine blade are converted into image data after noise reduction treatment, a convolutional neural network model is built, a training data set is built, and the image data is input into the convolutional neural network model for pre-training;
(2) Acquiring operation parameters, atmospheric parameters and aerodynamic noise of different wind turbine blades of the same wind farm in a fixed time scale as acquired acoustic signals; the sound signals are summarized to a server of a data acquisition and transmission center, and data are transmitted to a background server of a workstation through a network protocol; acquiring aerodynamic noise data of a wind turbine blade to be detected in a background server, converting the aerodynamic noise data into image data after noise reduction treatment, inputting a convolutional neural network model, outputting corresponding characteristic values, and displaying and judging the fault type of the wind turbine blade according to sample historical data of aerodynamic noise of the wind turbine blade and corresponding fault labels in a man-machine interaction operation interface;
the training data set of the convolutional neural network comprises sample historical data and capacity expansion data, wherein the sample historical data are truly detected data, and the capacity expansion data comprise simulation data and data for increasing the data set of the sample historical data.
2. The wind turbine blade fault diagnosis method based on aerodynamic noise according to claim 1, wherein the fault type corresponding to the fault label at least comprises one of icing, cracking or defect, front edge abrasion and surface sand hole, the acquisition is to perform unsupervised learning on the existing fault by means of local constraint sparse self-encoder, a plurality of categories are separated, then 5 sections of audio in each category are randomly selected, and the fault type of each category is judged to obtain the corresponding fault label.
3. The aerodynamic noise-based wind turbine blade fault diagnosis method according to claim 1, characterized in that the simulation data acquisition method is as follows:
acquiring sample historical data of wing profile data of a wind turbine of a wind power plant, adjusting the wing profile according to the wing profile data of the wind turbine, a CST class-shape function and Latin hypercube sampling, and increasing or decreasing the wing profile at the position 80% away from a blade root to realize new modeling, thereby obtaining new wing profile data and corresponding labels, and further expanding a data set under the wing profile;
and inputting the new wing profile data obtained after adjustment into a openfast, bladed or HAWC2 wind turbine generator simulation platform, and setting corresponding parameters to obtain aerodynamic noise data at the observation point.
4. The aerodynamic noise-based wind turbine blade failure diagnosis method according to claim 1, characterized in that the obtained aerodynamic noise data is obtained by means of an acoustic signal acquisition and transmission device during a specified acquisition time period on a sunny day.
5. The aerodynamic noise-based wind turbine blade fault diagnosis method according to claim 1, wherein the noise reduction of the sample history data into image data comprises the following sub-steps in sequence:
(1-1) filtering the sample history data by means of a butterworth band-pass filter to remove background noise and noise generated during transmission and conversion;
(1-2) performing secondary filtering treatment by using empirical wavelet transformation, and filtering mechanical noise generated in the wind turbine to obtain noise-reduced pneumatic noise data;
(1-3) converting the noise-reduced aerodynamic noise data into image data by means of mel frequency spectrum.
6. The aerodynamic noise-based wind turbine blade fault diagnosis method according to claim 5, wherein the data added to the data set of the sample history data is data obtained by inverting the image data obtained by noise reduction and mel frequency spectrum, and adjusting the image contrast, thereby obtaining new data expansion original training set, and obtaining data by improving convergence.
7. The aerodynamic noise based wind turbine blade failure diagnosis method according to claim 1, characterized in that step (2) comprises the sub-steps of:
(2-1) acquiring aerodynamic noise of a wind turbine blade to be detected, transmitting an acoustic signal to a background server through a network protocol through a data acquisition and transmission center, reducing noise through a Butterworth band-pass filter and empirical wavelet variation, and converting the noise into image data through a Mel frequency spectrum;
(2-2) inputting the image data into a convolutional neural network model for classification, and outputting corresponding characteristic values by the convolutional neural network model according to fault classification in pre-training;
and (2-3) judging whether the characteristic value is within a preset threshold range, classifying according to the characteristic value if the characteristic value is within the preset threshold range, and judging that the characteristic value cannot be determined if the characteristic value exceeds the preset range.
8. The wind turbine blade fault diagnosis method based on aerodynamic noise according to claim 7, wherein the characteristic value exceeds a preset range, when the judging result is that the characteristic value cannot be confirmed, the characteristic value is manually confirmed, an audio signal and a time-frequency diagram which cannot confirm the fault are called and displayed through a computer operation interface of a workstation, an operation and maintenance person judges the type of fault sound in the workstation according to the fault sound signal and the time-frequency diagram, when the fault sound cannot be judged, the operation and maintenance person checks the operation and maintenance person on site again, and after the fault result is obtained, the result is stored and the original classification result is corrected.
9. The aerodynamic noise-based wind turbine blade fault diagnosis method according to claim 8, wherein the aerodynamic noise converted picture data and wind turbine blade fault type classification results are stored in a background server.
10. The wind turbine blade fault diagnosis method based on aerodynamic noise according to claim 9, wherein the stored picture data and the corrected classification result are input into a convolutional neural network as a training set, and the convolutional neural network model is updated periodically.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073354.8A CN116153333A (en) | 2023-02-07 | 2023-02-07 | Wind turbine blade fault diagnosis method based on aerodynamic noise |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073354.8A CN116153333A (en) | 2023-02-07 | 2023-02-07 | Wind turbine blade fault diagnosis method based on aerodynamic noise |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116153333A true CN116153333A (en) | 2023-05-23 |
Family
ID=86356655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310073354.8A Pending CN116153333A (en) | 2023-02-07 | 2023-02-07 | Wind turbine blade fault diagnosis method based on aerodynamic noise |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116153333A (en) |
-
2023
- 2023-02-07 CN CN202310073354.8A patent/CN116153333A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112233091B (en) | Wind turbine blade image damage detection and positioning method | |
| CN109763944B (en) | Non-contact monitoring system and monitoring method for blade faults of offshore wind turbine | |
| Chen et al. | Acoustical damage detection of wind turbine blade using the improved incremental support vector data description | |
| CN104677623B (en) | A kind of blade of wind-driven generator fault acoustics in place diagnostic method and monitoring system | |
| US20230123117A1 (en) | Method and Apparatus for Inspecting Wind Turbine Blade, And Device And Storage Medium Thereof | |
| CN112727704B (en) | Method and system for monitoring corrosion of leading edge of blade | |
| CN112324629A (en) | Wind power blade early damage monitoring system and method based on vibration and sound | |
| CN104101652B (en) | Audio signal based wind power blade damage monitoring method and system | |
| CN110985310B (en) | Wind driven generator blade fault monitoring method and device based on acoustic sensor array | |
| CN113806893A (en) | Fan state monitoring and fault diagnosis method and system based on industrial internet | |
| CN103926075B (en) | Method for improving fault analysis precision of wind turbine generating set gear box | |
| Sánchez et al. | Wind Turbines Acoustic Inspections performed with UAV and sound frequency domain analysis | |
| CN111400959B (en) | Blade fault diagnosis method and device for wind generating set | |
| Dhanraj et al. | Statistical data mining through credal decision tree classifiers for fault prediction on wind turbine blades using vibration signals | |
| CN117782570A (en) | Mesh ad hoc network-based life prediction system and method for offshore wind turbine | |
| CN112465776B (en) | Crack intelligent detection method based on wind turbine surface blurred image | |
| CN117419010A (en) | Deep learning-based fan blade health monitoring method and fault detection method thereof | |
| CN116153333A (en) | Wind turbine blade fault diagnosis method based on aerodynamic noise | |
| CN111833336A (en) | Hyperspectrum-based wind power blade surface sand hole fault detection system and method | |
| Li | Few-shot wind turbine blade damage early warning system based on sound signal fusion | |
| CN219081774U (en) | Wind turbine blade state monitoring system | |
| Munteanu et al. | Intelligent Condition Monitoring of Wind Turbine Blades: A Preliminary Approach | |
| CN115855461B (en) | Wind turbine generator fan blade noise abnormity system and method | |
| CN112780503B (en) | Fan blade protective paint damage monitoring method and system based on audio signals | |
| CN118148851B (en) | Yun Bianduan-coordinated wind turbine hub key component sound vibration fusion on-line monitoring system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |