CN112727704B - Method and system for monitoring corrosion of leading edge of blade - Google Patents
Method and system for monitoring corrosion of leading edge of blade Download PDFInfo
- Publication number
- CN112727704B CN112727704B CN202011478576.0A CN202011478576A CN112727704B CN 112727704 B CN112727704 B CN 112727704B CN 202011478576 A CN202011478576 A CN 202011478576A CN 112727704 B CN112727704 B CN 112727704B
- Authority
- CN
- China
- Prior art keywords
- blade
- data
- monitoring
- damage
- corrosion
- 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.)
- Active
Links
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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
- F05B2260/00—Function
- F05B2260/82—Forecasts
- F05B2260/821—Parameter estimation or prediction
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
- F05B2260/00—Function
- F05B2260/83—Testing, e.g. methods, components or tools therefor
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
- F05B2260/00—Function
- F05B2260/84—Modelling or simulation
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
- F05B2270/00—Control
- F05B2270/70—Type of control algorithm
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
- F05B2270/00—Control
- F05B2270/80—Devices generating input signals, e.g. transducers, sensors, cameras or strain gauges
- F05B2270/806—Sonars
Abstract
The application relates to a method for monitoring corrosion of a leading edge of a blade, which comprises the following steps: collecting a wind sweeping signal of the blade through a sound sensor; acquiring sound data meeting the conditions in the running process of the fan at fixed intervals by a condition filtering module; performing quality detection on the voice data through a data quality processing module, and screening out the voice data with poor data quality; and identifying whether the corrosion of the leading edge of the blade exists or not and evaluating the severity of the corrosion, and giving early warning information when the corrosion exceeds a set threshold value. The monitoring method and the monitoring system for the corrosion of the front edge of the blade can identify the corrosion of the front edge of the blade based on the machine position level in time, provide alarm information before the corrosion damage of the front edge of the blade to a structural layer and support targeted timely maintenance.
Description
Technical Field
The application relates to a monitoring method and a monitoring system for corrosion of a leading edge of a blade, which are applicable to the technical field of damage monitoring.
Background
The blade is a key part for catching wind, the aerodynamic shape of a specific design is a key part for ensuring the wind catching capacity, and the aerodynamic shape of the blade tip area, particularly the length area of the blade tip 1/3, is the most critical area. The damage of aerodynamic shape of the blade tip area, especially the damage of the front edge part of the airfoil shape, can greatly reduce the power generation output of the fan. Therefore, ensuring the integrity of the aerodynamic profile over the operational life cycle of a wind turbine is a key operational goal in ensuring the power generation performance of a wind turbine.
However, the fans operate under the condition of complicated or even severe environment all the year round, the fans in some areas operate in cloud all the year round, and some fans are located in areas with abundant annual rainfall. Such a high rain erosion environment presents challenges to the operation of the fan blades. In order to continuously pursue lower power consumption cost, the length of a fan blade is increased, and the linear speed of a blade tip area generally reaches 80-120 m/s. In addition, due to the superposition of wind speeds, the blade tip area, particularly the front edge of the airfoil, can collide with water molecules in the air at a very high speed, and chemical components such as sunshine, salt mist in the air and the like erode a blade front edge protection system all the year round, so that the front edge of the airfoil is continuously accumulated to generate damage, namely front edge corrosion. Leading edge erosion involves damage to the aerodynamic profile of the blade tip and is located in the leading edge region and is distributed along the length primarily in the vicinity of the distal end 1/3 of the blade, with the damage itself and the degree of damage being directly reflected in the aerodynamic noise of the blade.
The early stage of corrosion is mainly in a coating and a putty layer, has limited influence on the generated energy, has strong maintainability and short maintenance time, and is easy to restore to the original pneumatic appearance. In the later stage, damage is deeply applied to a structural layer, such as glass fiber reinforced plastic and even adhesive, so that on one hand, the maintenance difficulty is high, the time is long, and the original structural layer damages the outline of the supporting appearance, so that the maintenance is difficult to recover to the original pneumatic appearance, the influence on the pneumatic performance is large, and the irreversibility is provided to a certain degree; on the other hand, rainwater, salt fog and the like invade the interior of the blade cavity, and can influence the exposed glass fiber reinforced plastic and the bonding interface in the interior, even invade the sandwich structure, and influence the structural strength and the service life of the blade. Therefore, the early identification of the corrosion of the front edge of the blade based on the machine position ensures that the identification and the maintenance are completed before the corrosion penetrates into the structural layer, and is the key of the operation and the maintenance of the blade. The mode that artifical patrolling and examining was adopted in the discernment of current blade damage more, wastes time and energy, and once a year usually, has the hysteresis quality. Along with the rise of the accumulated loading amount of the fan, the annual inspection of some wind fields is difficult to guarantee due to the inspection resource limitation.
The existing blade online monitoring solutions have no damage to the blade, lack of monitoring for the corrosion damage of the front edge of the blade and have no monitoring system and solution for identifying the corrosion damage degree of the front edge.
The patent with the application number of CN201710287370.1 discloses a fan blade protective film damage detection method based on edge segmentation, and mainly relates to a blade protective film damage identification method of a blade image signal. The acquisition of blade appearance image usually need be gone on under fan shutdown state, and receives the influence of multiple factors such as ambient light, sensor angle and wind speed, hardly accomplishes real-time identification and monitoring. The patent with application numbers CN201820791284.4 and CN201620757688.2 adopts an optical fiber strain monitoring system adhered to the blade to monitor the structural crack and damage of the blade, the damage can monitor the corrosion of the front edge of the structural layer, but the damage type and degree of the front edge corrosion cannot be accurately distinguished and positioned, and the corrosion condition cannot be identified and early warned in the initial stage of the front edge corrosion. Patent with application number CN201210205204.X discloses a wind power blade damage monitoring and positioning system based on a wireless acoustic emission sensor network, and the system adopts a wireless acoustic emission technology and does not explicitly refer to the identification of blade leading edge corrosion. The patent with the application number of CN202010241827.7 discloses a wind power blade external field unmanned aerial vehicle double-light-source detection system and method, and although the appearance and the structural damage of a blade can be visually identified, the system only belongs to a semi-intelligent blade damage detection method, but not a set of complete online monitoring system, and the system does not have real-time performance.
In the prior art, a set of targeted online monitoring system is urgently needed, the front edge corrosion state of each machine position can be identified in time, and the electric quantity loss caused by corrosion is avoided.
Disclosure of Invention
The application provides a monitoring method and a monitoring system for corrosion of a front edge of a blade, which aim to adopt aerodynamic noise signals and characteristics of the blade to carry out online monitoring on the corrosion of the front edge of the blade so as to identify corrosion damage of the front edge, judge the damage type of the front edge and provide early warning for early identification and maintenance of the corrosion of the front edge.
The method for monitoring the corrosion of the leading edge of the blade comprises the following steps:
(1) collecting a wind sweeping signal of the blade through a sound sensor;
(2) acquiring sound data meeting the conditions in the running process of the fan at fixed intervals by a condition filtering module;
(3) performing quality detection on the voice data through a data quality processing module, and screening out the voice data with poor data quality;
(4) and identifying whether the corrosion of the leading edge of the blade exists or not and evaluating the severity of the corrosion, and giving early warning information when the corrosion exceeds a set threshold value.
Preferably, the method further comprises the step (5): and displaying the corrosion damage rate according to the historical factor data, and giving a position with the damage rate exceeding a threshold value.
In the step (3), the quality detection of the sound data includes two steps of training the sound sample and judging the test data, wherein the training of the sound sample includes the following steps:
(1.1) accumulating normal noise-free sound samples as training data, and taking the rest noise-free sound samples as a verification set;
(1.2) transforming each sample in the training data to obtain a spectrogram matrix S;
(1.3) transforming the spectrogram matrix S to obtain an energy spectrum matrix S';
(1.4) setting a frequency dimension length such that a time dimension length is greater than a frequency dimension length;
(1.5) reducing the length of the time dimension to the length of the frequency dimension to form a square matrix Sx；
(1.6) processing all training data to form a training set
(1.7) recording the global mean and standard deviation of the training set data;
(1.8) constructing a self-encoder model for model training, selecting an optimization mode by taking the minimized normalized root mean square error as a target, and storing the model after training;
(1.9) carrying out the characteristic extraction and normalization processing on the data of the verification set, and predicting by using a trained self-encoder model;
(1.10) carrying out reconstruction error calculation on the prediction results of all verification set data to obtain an error threshold value alpha;
the judgment of the test data comprises the following steps:
(2.1) performing the operations of the above steps (1.1) - (1.6) on any test data x;
(2.2) based on the training set
The global mean and the standard deviation of the test data are normalized;
(2.3) inputting the trained self-encoder model for prediction to obtain a reconstructed sample
(2.4) calculating x and
the reconstructed error epsilon between the two is compared with an error threshold alpha, and the quality of the sound signal is judged.
And (4) selecting a damage judgment module according to the type of the input blade leading edge protection scheme, wherein the blade leading edge protection scheme comprises the step of respectively arranging a protective film or protective paint on the leading edge of the blade for damage monitoring.
Preferably, the method for providing the protective film on the leading edge of the blade for damage monitoring comprises the following steps:
(1) collecting a plurality of packets of sound signals of wind swept by a fan blade in real time within a period of time, and carrying out short-time Fourier transform on the collected wind swept signals to obtain a spectrogram;
(2) converting and dividing the sound signal, calculating the energy mutual difference factor value of each packet of data in the time period, and judging whether the energy mutual difference factor value exceeds a threshold value; meanwhile, the whistle forms in the images are identified;
(3) identifying the position of the whistle in the spectrogram and positioning the maximum frequency and the minimum frequency corresponding to the whistle outline;
(4) calculating the distance from the whistle sounding position to the center of the hub;
(5) and judging the damage type and the damage position of the fan blade.
Wherein, the step of judging whether the energy mutual difference factor value exceeds the threshold value comprises the following steps:
(1) dividing the blade wind sweeping signal spectrogram according to a local maximum search algorithm to obtain a time domain division point of each blade wind sweeping signal on the spectrogram;
(2) calculating the sum of the wind sweeping energy of each blade in different time intervals based on the spectrogram;
(3) calculating the average wind sweeping energy sum of each blade by taking the number of the fan blades as a wind sweeping period;
(4) carrying out normalization calculation on the blade wind sweeping energy to obtain an energy mutual difference factor value;
(5) and judging whether the energy mutual difference factor exceeds a set threshold value or not for each packet of sound signal data.
The whistle recognition and damage positioning method comprises the following steps:
(1) storing the spectrogram obtained by transformation into an image format;
(2) recognizing the whistle sound form in the image based on an image target detection algorithm, recognizing the position of the whistle sound in a spectrogram, and positioning the maximum frequency and the minimum frequency corresponding to the whistle sound outline;
(3) and calculating the distance from the whistle sounding position to the center of the hub.
The method for monitoring the damage by arranging the protective paint on the front edge of the blade comprises a training phase and an online operation phase.
Preferably, the training phase comprises the steps of:
(1) collecting a plurality of groups of audio signals within a period of time, and acquiring an average rotating speed value of each group of audio corresponding to the time;
(2) converting the audio signal to obtain a spectrogram and converting the spectrogram into an energy level;
(3) screening working conditions, and selecting data in a certain rotating speed range;
(4) each group of audio signals is processed to obtain a sample, the rotating speed is used as an independent variable, the maximum energy level of the target frequency band is used as a dependent variable, linear fitting is carried out on the rotating speed and the energy level of a plurality of sample points, and residual distribution of the energy level is calculated;
(5) and (4) respectively carrying out the operation of the step (4) on each frequency band to obtain a linear model.
Preferably, the on-line operational phase comprises the steps of:
(1) acquiring a group of audio signals and corresponding rotating speed values thereof in real time, and judging whether the rotating speed values are in a rotating speed range screened by training data;
(2) predicting the predicted energy level of the target frequency band according to the rotating speed by adopting the model obtained by training, and calculating the residual error between the energy level of the group of audio signals and the predicted energy level;
(3) selecting a plurality of groups of collected effective data, and calculating the average value of residual errors;
(4) estimating the damage level of the protective paint by adopting a prediction factor according to the residual distribution of the training data and the average value of the online running residual;
(5) when a plurality of frequency bands are selected for fitting, the final prediction factor takes the median of the prediction factors of each frequency band at the same time;
(6) and judging the stage of the fault according to the current factor and the historical factor by using the rule.
The present application also relates to a system for monitoring the erosion of the leading edge of a blade, comprising at least one sound sensor for acquiring sound signals, at least one machine end collector, and a software system comprising an algorithm for carrying out the monitoring method according to the above.
The software system at least comprises an algorithm module, a data management module, a configuration and user management module and a monitoring and early warning display module; the algorithm module identifies the corrosion condition of the front edge of the blade according to the collected sound signal and the type of the protection scheme of the front edge of the blade, judges the damage degree, and sends out early warning information when the damage exceeds a set threshold value; the data management module manages the collected sound data and the data calculated by the algorithm module; the configuration and user management module is used for providing configuration input of a system parameter by a user and management and maintenance of user information; the monitoring and early warning display module is used for providing information display of monitoring and early warning results for a user.
The software system is arranged on a machine-end collector, a station-end server or a cloud-end server; the monitoring system also comprises an acquisition cabinet, wherein the acquisition cabinet comprises at least one of a lightning protection module, a power filter, an air switch, a signal acquisition device and a wireless transceiver module; and a data processing device for performing condition filtering on the initial data is arranged in the acquisition cabinet.
The monitoring method and the monitoring system for the corrosion of the front edge of the blade can timely identify the corrosion of the front edge of the blade based on the machine position level, provide alarm information before the corrosion damage of the front edge of the blade to the structural layer, support targeted timely maintenance, avoid the serious corrosion from being damaged to the structural layer, reduce the damage of the pneumatic performance, shorten the maintenance period, improve the maintenance quality and avoid the increase of fatigue load. Meanwhile, a corrosion rate curve of the front edge of the blade can be provided, and targeted upgrading and technical transformation of a front edge protection system can be performed aiming at a machine position with a high corrosion rate, so that economic benefits are optimized.
Drawings
FIG. 1 is a schematic diagram of the present invention for monitoring blade leading edge erosion.
FIG. 2 is a flow chart illustrating steps of a method for monitoring erosion of a leading edge of a blade according to the present application.
Figure 3 shows a schematic diagram of a flight level exhibiting a damage rate based on historical factor data and giving the damage rate exceeding a threshold.
FIG. 4 is a graph of the leading edge protective paint damage factor as a function of time as monitored by the blower in the examples.
FIG. 5 is a graph of statistics of the rate of change of the erosion rate of the leading edge of the wind turbine in the examples.
Detailed Description
To make the objects, technical solutions and advantages of the present application more apparent, embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that the embodiments and features of the embodiments in the present application may be arbitrarily combined with each other without conflict.
This application aims at in time discerning the corrosion of blade leading edge based on the machine position rank, provides alarm information before blade leading edge corrosion damage to structural layer, supports the timely maintenance of pertinence, avoids damaging to the serious corruption of structural layer, reduces the aerodynamic performance damage, shortens the maintenance cycle, improves the maintenance quality, avoids the increase of fatigue load. Meanwhile, on the basis, the monitoring system can provide a corrosion rate curve of the front edge of the blade, and aiming at a machine position with a high corrosion rate, wind field management personnel can carry out targeted upgrading and technical transformation on a front edge protection system, so that the economic benefit is optimized.
As shown in fig. 1, a system for monitoring the corrosion of the leading edge of a blade according to the present application comprises at least one acoustic sensor 1 mounted on the tower of a wind turbine or at a distance from the wind turbine for acquiring the wind sweeping signal of the blade. The sound sensor 1 can also be provided with a mounting bracket which can be provided with wind-proof, rain-proof and anti-freezing functions. The system also comprises a software system comprising an algorithm, data storage, monitoring and early warning result display. The software system can be arranged on the machine-side collector 4, the station-side server 2 or the cloud server 3 according to requirements. In order to ensure the network security of the wind field, the monitoring system can be provided with a forward or reverse type network security isolation device according to the requirement.
Preferably, the monitoring system may include an acquisition cabinet, and the acquisition cabinet may include at least one of a lightning protection module, a power filter, an air switch, a signal acquisition device, and a wireless transceiver module. The air switch, the lightning protection module and the power filter can ensure that the monitoring system has stronger environmental adaptability and is not influenced by lightning and electromagnetic interference. The wireless transceiving module mainly solves the problem that the fan ring network cannot communicate with an external network in time, and is used for upgrading software and algorithms and transmitting collected data back to the cloud in real time when the fan cannot communicate with the external network. The signal collector is used for collecting the operation data and/or the environment data of the fan.
Preferably, the data processing device is arranged in the acquisition cabinet, and the initial data is subjected to condition filtering, namely the condition filtering module is included, and the condition filtering module is mainly used for judging whether the acquired data is stored or not and further analyzing the data according to working conditions such as rotating speed, yaw and variable pitch. Preferably, the monitoring data of each fan is uniformly transmitted back to the station-side server through the wind field ring network, and then the early warning result and the original data are uploaded to the cloud server through the network safety isolation device. The station side service is provided with a software system for modules such as data analysis, algorithm operation, factor calculation, monitoring, early warning display and the like; and the station server supports the connection of a safe cloud algorithm library and supports the mobile terminal app and even an interface of a work order system. The cloud server provides data storage, result display and other functions. In order to adapt to the requirements of real-world situations, the monitoring system also supports the configuration and storage of all software and data in a machine-side collector.
The software system at least comprises an algorithm module, a data management module, a configuration and user management module and a monitoring and early warning display module. The algorithm module identifies the corrosion condition of the front edge of the blade according to the collected sound signal and the selected type of the protection scheme of the front edge of the blade, judges the damage degree, and sends out early warning information when the damage exceeds a set threshold value; the data management module manages the collected sound data and the data calculated by the algorithm module; the configuration and user management module is used for providing configuration input of a system parameter by a user and management and maintenance of user information; the monitoring and early warning display module is used for providing information display of monitoring and early warning results for a user.
As shown in fig. 2, the operation process of the algorithm module in the software system of the present application mainly includes:
(1) the collector obtains sound data in the running process of the fan meeting the conditions and fan state parameters at corresponding moments, such as rotating speed, paddle angle, yaw angle and the like, at fixed intervals through the condition filtering module;
(2) performing quality detection on the voice data through a data quality processing module, and screening out the voice data with poor data quality;
(3) selecting a damage judgment module according to the type of the input blade leading edge protection scheme, identifying whether the corrosion of the leading edge of the blade exists or not by adopting a protective film damage module or a protective paint damage module, evaluating the severity of the corrosion, respectively outputting protective film damage factors or protective paint damage factor values, and giving alarm information according to a threshold value;
(4) and displaying the damage rate according to the historical factor data, and giving the position of the damage rate exceeding a threshold value. As shown in fig. 3, the abscissa is the station number and the ordinate is the damage rate. And judging the severity of the damage through the factor value of the corresponding module. Aiming at the machine position with high corrosion rate, wind field management personnel can carry out targeted upgrading and technical transformation on the front edge protection system.
In the step (2), the quality detection of the sound data includes two steps of training the sound sample and judging the test data, wherein the training of the sound sample includes the following steps:
(1) accumulating a certain amount of normal noise-free sound samples as training data, and taking the rest noise-free sound samples as a verification set;
(2) carrying out short-time Fourier transform on each sample in the training data to obtain a spectrogram matrix S in a complex form;
(3) carrying out logarithmic transformation on the spectrogram matrix S to obtain an energy spectrum matrix S';
the size of the two matrixes S and S 'is determined by the frequency resolution parameter of short-time Fourier transform, and the time dimension length of the two matrixes S and S' is determined by the parameter of short-time Fourier transform and the time length of the original sound signal; the larger the resolution is, the more accurate the result is, but the higher the requirement on the computing capability is, that is, the larger the dimension length is, the slower the computing process is, so that the selection of the frequency dimension length is generally determined by balancing the resolution and the computing efficiency of the model;
(4) the frequency dimension length is fixedly set to a certain length, for example 256, so that the time dimension length is greater than the frequency dimension length;
(5) using Principal Component Analysis (PCA) to reduce the time dimension length to the frequency dimension length to form a square matrix Sx；
(6) All training data are processed to form a training set
Each sample size being 256 × 256, for example;
(7) recording the global mean value and the standard deviation of the training set data so as to carry out normalization processing in the following;
(8) constructing a self-encoder model, wherein the depth of the model is 5+5, the number of convolution kernels in an encoder is reduced from small to large, and the number of convolution kernels in a decoder is symmetrically reduced; the depth of the model refers to the number of convolutional layers, i.e. the encoder and decoder are each composed of 5 convolutional layers, pooling layers and related structures;
(9) performing model training, selecting an optimization mode by taking the minimized normalized root mean square error as a target, and storing the model after the training is finished, wherein an Adam optimization method with regularization is used in the method;
(10) performing the characteristic extraction and normalization processing on the data of the verification set, and predicting by using a trained self-encoder model;
(11) performing reconstruction error calculation on the prediction results of all the verification set data to obtain an error threshold value alpha;
the judgment of the test data comprises the following steps:
(1) performing the operations of the steps (1.1) - (1.6) on any test data x to obtain a feature matrix of 256 × 256, for example;
(2) according to a training set
The global mean and the standard deviation of the test data are normalized;
(3) inputting the trained self-encoder model for prediction to obtain a reconstructed sample
(4) Calculating x and
the reconstruction error epsilon between the two is compared with an error threshold alpha;
(5) if the error epsilon is larger than the threshold alpha, the data is represented as data interfered by noise, the interference severity is determined by the deviation degree of epsilon compared with alpha, and the data quality is poorer when the deviation is larger.
The method and the device carry out detection, evaluation and control on the quality of the sound signal from the aspect of a signal analysis processing algorithm of the mixed noise; the signals interfered by the mixed noise are analyzed through end-to-end integrated modeling, an accurate sound signal quality detection system can be formed in an unsupervised mode without collecting noise data, and reliable input is provided for subsequent diagnosis and analysis.
In step (3), the blade leading edge protection scheme type generally includes providing a protective film or paint on the leading edge of the blade for damage monitoring.
When the scheme of the protective film is adopted, the damage monitoring method comprises the following steps:
(1) collecting a plurality of packets of sound signals of wind swept by a fan blade in real time within a period of time, and carrying out short-time Fourier transform on the collected wind swept signals to obtain a spectrogram;
(2) converting and dividing the sound signal, calculating the energy mutual difference factor value of each packet of data in the time period, and judging whether the energy mutual difference factor value exceeds a threshold value; meanwhile, the whistle forms in the images are identified;
(3) identifying the position of the whistle in the spectrogram and positioning the maximum frequency and the minimum frequency corresponding to the whistle outline;
(4) calculating the distance from the whistle sounding position to the center of the hub;
(5) and judging the damage type and the damage position of the fan blade.
More specifically, the step of calculating the value of the energy mutual difference factor includes:
(1) carrying out short-time Fourier transform on each collected packet of sound signals to obtain a spectrogram;
(2) dividing the blade wind sweeping signal spectrogram according to a local maximum search algorithm to obtain a time domain division point of each blade wind sweeping signal on the spectrogram;
(3) calculating the sum of the wind sweeping energy of each blade in different time intervals based on the spectrogram;
(4) calculating the average wind sweeping energy sum of each blade by taking the number of the fan blades as a wind sweeping period; because most of the fans have 3 blades, the average wind sweeping energy sum can be calculated by taking 3 times of wind sweeping as a wind sweeping period and the wind sweeping energy of the same blade in different periods;
(5) the blade wind sweeping energy is subjected to normalization calculation to obtain an energy mutual difference factor value factor, wherein a normalization formula of the factor value is as follows:
wherein, ene is the energy and the set of each blade, i.e. ene (ene)1,enei,……enen) And n is the number of blades.
(6) And judging whether the energy mutual difference factor exceeds a set threshold value or not for each packet of sound signal data. If the threshold value is not exceeded, the leaves are not damaged; if the threshold is exceeded, damage to the leaf exists.
The whistle recognition and damage positioning method comprises the following steps:
(1) carrying out short-time Fourier transform on the collected wind sweeping signals within a period of time to obtain a spectrogram;
(2) storing the spectrogram into an image format;
(3) based on an image target detection algorithm, the whistle morphology in the image is identified, the position of the whistle in a spectrogram can be identified, and the maximum frequency and the minimum frequency corresponding to the whistle outline can be positioned. When there is a whistle in the picture, the algorithm will output the corresponding coordinates and probability, and when there is no whistle in the picture, the algorithm will not output. The image-based target detection algorithm may employ models such as fast-rcnn, yolo, and the like.
(4) If the whistle exists in the image, the Doppler frequency shift effect is applied to calculate the distance between the whistle sounding position and the center of the hub, and the calculation formula is as follows:
wherein tau is the time required for one rotation of the blade, VsIs the local ambient sound velocity, FHAnd FLAre respectively the corresponding whistle formThe doppler should be shifted by the maximum and minimum frequencies. Wherein, the ambient sound velocity can adopt Vs＝(331.45+0.61*t)m·s-1It is calculated that 331.45 is the speed of sound in air at 0 degrees and t is the ambient temperature (c).
And calculating the actual damage position according to the formula by combining the frequency parameter and the rotating speed parameter obtained by the image recognition and the fan state parameter such as the wind speed and the like synchronously acquired by the fan. When the damaged position is far away from the drain hole, the blade is damaged by the protective film, and when the damaged position is close to the drain hole, the blade may be damaged by the protective film or blocked by the drain hole. When the damage location is close to the blade root, the whistle may be caused by mechanical jamming when the blade pitches. The rotating speed parameters can be calculated by taking a time domain obtained by segmentation in the energy mutual difference calculation module as a period or directly using fan rotating speed parameters acquired by a sensor.
According to the damage monitoring method of the fan blade protective film, non-contact monitoring is carried out on damage of the fan blade protective film through the audio signals, and the purpose of nondestructive testing is achieved while the health state of the fan blade can be monitored in real time. Meanwhile, the damage type can be identified in detail and the damage position can be positioned through the Doppler shift effect of the sound signal and an image target detection method based on a spectrogram, so that the damage monitoring of the protective film is more pertinent, diversified and effective.
When the protective paint scheme is adopted, the damage monitoring method comprises a training phase and an online operation phase, wherein the training phase comprises the following steps:
(1) collecting a plurality of groups of audio signals within a period of time, and acquiring an average rotating speed value of each group of audio corresponding to the time; the average rotating speed value can be obtained from an external system, and when the rotating speed cannot be obtained from the outside, the average rotating speed value can also be calculated by identifying the number of times of wind sweeping in the audio signal per unit time;
(2) performing short-time Fourier transform (STFT) conversion on the audio signal to obtain a spectrogram, and converting the spectrogram into an energy level; the energy level here refers to a parameter representing the energy of the wind sweeping sound, and for example, a sound pressure level and the like can be used as evaluation indexes of the energy level;
(3) and (3) screening data: screening working conditions, selecting data in a certain rotating speed range, and filtering low-rotating-speed noise points and data points in full-time; the selected speed is generally a higher speed interval of the fan; the fan rotating speed range can be determined according to the fan parameters and whether the collection conditions are limited; full-load is the highest rotating speed value, namely after reaching the rotating speed under a certain wind speed, the wind speed is increased again, the rotating speed cannot be increased, and the power can also become approximate rated power;
(4) each group of audio signals is processed to obtain one sample, and multiple groups of audio signals can obtain multiple samples in a period of time. Taking the rotating speed as an independent variable and the maximum energy level of the target frequency band as a dependent variable, performing linear fitting on the rotating speed and the energy level of a plurality of sample points by adopting linear regression, and calculating residual distribution of the energy level, such as an n-sigma value;
(5) and (4) when a plurality of frequency bands are selected, respectively carrying out the operation in the step (4) on each frequency band to obtain a linear model. Because the damage of the protective paint has the same performance in a large range of frequency bands, one or more target frequency bands (5000 Hz-16000 Hz) can be selected, and when a plurality of frequency bands are selected, the range of each frequency band is smaller and the linear relation is stronger.
The on-line operation phase comprises the following steps:
(1) and acquiring a group of audio signals and corresponding rotating speed values thereof in real time, and judging whether the rotating speed values are in a rotating speed range screened by the training data. When the rotating speed is in the screening range, carrying out STFT conversion on the audio signal and calculating the energy level of the audio signal; when the rotating speed is not in the screening range, subsequent operation is not carried out;
(2) predicting the predicted energy level of the target frequency band according to the rotating speed by adopting the model obtained by training, and calculating the residual error between the energy level of the group of audio signals and the predicted energy level;
(3) selecting a plurality of groups of collected effective data, and calculating the average value of residual errors;
(4) and evaluating the damage level of the protective paint according to the residual distribution of the training data and the average value of the online running residual, wherein the formula can be as follows:
wherein x is an online operation residual error average value, y is a prediction factor which is the damage level of the protective paint, base is a current fan damage level reference value, when y is 0, the protective paint is not damaged, and when y is 1, the protective paint is seriously damaged; the nsigma is the sum of the average value of the residual errors of the training data and n times of standard deviation, and n can be determined according to the requirement of monitoring precision;
(5) when a plurality of frequency bands are selected for fitting, finally, the prediction factors take the median of the factors of each frequency band at the same time;
(6) and judging the stage of the fault according to the current factor and the historical factor by using a rule, and outputting the fault stage and whether mutation occurs at the current stage.
The damage monitoring method of the fan blade protective paint has the following beneficial technical effects: the online real-time monitoring and the non-contact monitoring can be realized, the timeliness and the high efficiency of the monitoring are ensured, and the possible damage caused by the monitoring is avoided; based on the mechanism that the rotating speed of the blade and the energy level of wind sweeping sound are in a linear relation when the blade sweeps wind, the damage of the protective paint can be identified only by using an audio signal by utilizing the phenomenon that a relation curve translates along the direction with high energy level when the protective paint is damaged, so that the monitoring and the identification are more efficient and accurate; aiming at the mechanism that the damage failure of the protective paint has different rates at different stages, the damage degree of the protective paint can be judged, the precision of damage monitoring is improved, and a more accurate maintenance suggestion is provided for subsequent maintenance; the method for taking the median value based on the fitting multiple narrow frequency bands can effectively remove the interference caused by other types of faults and various environmental noises without carrying out denoising processing on data at an early stage, thereby saving a data processing program and improving the data processing efficiency; the method and the device map the wind sweeping energy level to the fault level finally according to the wind sweeping energy level variation trend of the blades, can early warn the deteriorating blades in time, and improve early warning efficiency and early warning accuracy.
Examples
Taking the blade of a 1.5MW wind turbine in a certain high rain erosion wind field as an example, the sensor of the monitoring system is installed at a proper position at the bottom of the tower, and the position of the sensor in the height direction is determined as a reference: the nearest distance from the blades is not more than 30m, and the circumferential position of the sensor is determined by referring to the main wind direction position of the fan. The sensor is fixed on the support, and the support is through magnetism inhale with the sticky surface of pasting of structure at a tower section of thick bamboo surface, and the sensor passes through the signal line and links to each other with gathering the cabinet. The acquisition cabinet is arranged on the tower bottom platform through bolt connection, and the acquisition cabinet enters the wind field ring network through the net port and the tower bottom wind field ring network router. The server is arranged at the booster station, is connected with the wind field ring network, is connected with the wind field conventional monitoring system SCADA, is communicated with the Scada through the Modbus soft gateway, acquires state information such as a propeller angle, a yaw angle, active power, rotating speed and the like, and transmits the state information to the collector of the corresponding machine position through the wind field ring network. The server is connected with the external network through the forward and reverse isolation devices and is communicated with the cloud server. The cloud server acquires monitoring data, monitoring system state and results of the station server through the forward isolation device; and issuing an algorithm and a software upgrading package through a reverse isolation device, and supporting remote configuration.
The monitoring system software functions include: intelligent monitoring, intelligent warning, audio display, rain erosion failure mode display and factor display; and the automatic report is that the report of the relevant information such as the rain erosion state, the early warning result, the speed and the like of each machine position blade in any time range is automatically provided, a management page of authority, configuration and an interface is provided, and the self-diagnosis result of the monitoring system is also displayed.
By configuring a condition acquisition strategy of an acquisition unit end, the acquisition unit automatically acquires data with a packet time length of 60s at fixed time intervals, such as 1 hour every interval, by a microphone, combines fan state data transmitted from a station end server at the same moment to judge conditions, stores the data according to the judged conditions, and transmits the data back to the station end server; when the collected sound data does not meet the judgment condition, the system continues to collect the next packet of data until the collection condition is met; the next collection node is 1 hour after the last collection node is finished. As an example, the determination condition may be, for example, a rotation speed of X or more and Y or less; the paddle angle value is unchanged within 60 s; the variation angle of the yaw angle is less than or equal to 30 degrees within 60 s; the active power of the fan is more than 500W.
According to the flow shown in fig. 2, the collector obtains sound data and fan state parameters at corresponding moments, such as rotation speed, pitch angle, yaw angle, active power and the like, in the fan operation process meeting the conditions at fixed intervals through the condition filtering module; the method comprises the steps of cleaning sound data through a data quality processing module, cleaning low-quality data of interference blade pneumatic noise signals of fan mechanical noise, electromagnetic interference from a power supply or other parts, noise insect singing from the environment, bird calls and the like, screening out the noise of cleaning a wind sweeping map, selecting a protective paint damage module according to the type of an input blade leading edge protection scheme, identifying corrosion of the leading edge of the blade by adopting the protective paint damage module, calculating a protective paint damage factor, carrying out early-stage data processing, and calculating the leading edge corrosion factor.
Fig. 4 is a time variation curve of the damage factor of the leading edge protective paint in the period of 3 years monitored by the fan, wherein A, B, C are the factor values corresponding to the initial stage of rain erosion damage (basically no damage or slight damage to the coating), the middle stage (complete damage to the coating and the putty layer, and the glass fiber reinforced plastic is exposed) and the later stage (damage to the structural layer), and the factor values are normalized. When the factor value is greater than or equal to a threshold A0 and less than a threshold B0, the system gives out yellow warning information and advises a customer to pay attention to the blade rain erosion condition; when the factor value is greater than or equal to the threshold B0 and less than the threshold C0, the system gives orange warning information and recommends a customer to carry out maintenance of front edge corrosion on the airplane position in the current season; when the factor value is greater than or equal to the threshold value C0, the system gives a red warning message, and the customer is required to stop the machine as soon as possible to repair the front edge corrosion of the machine.
Fig. 5 is a statistical result of the corrosion rate change of the leading edge of the wind turbine in 3 years, where D, E, F three points are corrosion rates corresponding to the initial stage, the middle stage and the later stage of the rain erosion damage, and the corrosion rate of each point is (the damage factor 1.5 months before the time point-the damage factor 1.5 months after the time point)/3 months; the corrosion rate threshold is D0, E0, F0; when any one of the time points of the corrosion rate of the front edge of the fan exceeds the threshold value, the system gives out warning information to remind a client that the corrosion rate of the front edge of the fan is too high, and the client is advised to consider selecting a front edge protection system with longer service life for technical improvement.
The system mainly adopts a non-contact acquisition system convenient to install and maintain, and through the aerodynamic noise of real-time monitoring blades, through the targeted front edge corrosion damage judgment module, the targeted front edge corrosion and the degree thereof are judged, and compared with other methods for monitoring blade damage based on sound, the system is more targeted, gives clear early warning to the optimal time point of damage maintenance, and is very favorable for targeted intelligent maintenance.
Although the embodiments disclosed in the present application are described above, the descriptions are only for the convenience of understanding the present application, and are not intended to limit the present application. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure as defined by the appended claims.
Claims (14)
1. A method for monitoring corrosion of a leading edge of a blade is characterized by comprising the following steps: the method comprises the following steps:
(1) collecting a wind sweeping signal of the blade through a sound sensor;
(2) acquiring sound data meeting the conditions in the running process of the fan at fixed intervals by a condition filtering module;
(3) performing quality detection on the voice data through a data quality processing module, and screening out the voice data with poor data quality;
(4) identifying whether the corrosion of the front edge of the blade exists or not and evaluating the severity of the corrosion, and giving out early warning information when the corrosion exceeds a set threshold value;
in the step (3), the quality detection of the sound data includes two steps of training the sound sample and judging the test data, wherein the training of the sound sample includes the following steps:
(1.1) accumulating normal noise-free sound samples as training data, and taking the rest noise-free sound samples as a verification set;
(1.2) transforming each sample in the training data to obtain a spectrogram matrixS；
(1.3) spectrogram matrixSTransforming to obtain an energy spectrum matrix;
(1.4) setting a frequency dimension length such that a time dimension length is greater than a frequency dimension length;
(1.5) reducing the length of the time dimension to the length of the frequency dimension to form a square matrix
；
(1.6) processing all training data to form a training set
；
(1.7) recording the global mean and standard deviation of the training set data;
(1.8) constructing a self-encoder model for model training, selecting an optimization mode by taking the minimized normalized root mean square error as a target, and storing the model after training;
(1.9) carrying out feature extraction and normalization processing on the data of the verification set, and predicting by using a trained self-encoder model;
(1.10) carrying out reconstruction error calculation on the prediction results of all verification set data to obtain an error threshold value;
the judgment of the test data comprises the following steps:
(2.1) performing the operations of the steps (1.1) to (1.6) on any test data;
(2.2) based on the training set
Global mean and standard deviation ofOperating;
(2.3) inputting the trained self-encoder model for prediction to obtain a reconstructed sample;
and (2.4) calculating a reconstruction error between the sum and the threshold value, and comparing the reconstruction error with the threshold value to judge the quality of the sound signal.
2. The monitoring method according to claim 1, wherein: and (4) selecting a damage judgment module according to the type of the input blade leading edge protection scheme, wherein the blade leading edge protection scheme comprises the step of respectively arranging a protective film or protective paint on the leading edge of the blade for damage monitoring.
3. The monitoring method according to claim 2, wherein: the method for monitoring the damage by arranging the protective film at the front edge of the blade comprises the following steps:
(1) collecting multi-packet sound signals of wind swept by fan blades in real time for a period of time, and shortening the collected wind swept signals
Performing time Fourier transform to obtain a spectrogram;
(2) converting and dividing the sound signal, calculating the energy mutual difference factor value of each packet of data in the time period, and judging whether the energy mutual difference factor value exceeds a threshold value; meanwhile, the whistle forms in the images are identified;
(3) identifying the position of the whistle in the spectrogram and positioning the maximum frequency and the minimum frequency corresponding to the whistle outline;
(4) calculating the distance from the whistle sounding position to the center of the hub;
(5) and judging the damage type and the damage position of the fan blade.
4. The method of claim 3, wherein the step of determining whether the energy mutual difference factor value exceeds a threshold value comprises:
(1) dividing the blade wind sweeping signal spectrogram according to a local maximum search algorithm to obtain a time domain division point of each blade wind sweeping signal on the spectrogram;
(2) calculating the sum of the wind sweeping energy of each blade in different time intervals based on the spectrogram;
(3) calculating the average wind sweeping energy sum of each blade by taking the number of the fan blades as a wind sweeping period;
(4) carrying out normalization calculation on the blade wind sweeping energy to obtain an energy mutual difference factor value;
(5) and judging whether the energy mutual difference factor exceeds a set threshold value or not for each packet of sound signal data.
5. The monitoring method according to claim 3 or 4, wherein the whistle recognition and the lesion localization comprise the steps of:
(1) storing the spectrogram obtained by transformation into an image format;
(2) recognizing the whistle sound form in the image based on an image target detection algorithm, recognizing the position of the whistle sound in a spectrogram, and positioning the maximum frequency and the minimum frequency corresponding to the whistle sound outline;
(3) and calculating the distance from the whistle sounding position to the center of the hub.
6. The monitoring method according to claim 2, wherein: the method for arranging the protective paint on the front edge of the blade for damage monitoring comprises a training phase and an online operation phase.
7. The monitoring method according to claim 6, wherein: the training phase comprises the following steps:
(1) collecting a plurality of groups of audio signals within a period of time, and acquiring an average rotating speed value of each group of audio corresponding to the time;
(2) converting the audio signal to obtain a spectrogram and converting the spectrogram into an energy level;
(3) screening working conditions, and selecting data in a certain rotating speed range;
(4) each group of audio signals is processed to obtain a sample, the rotating speed is used as an independent variable, the maximum energy level of the target frequency band is used as a dependent variable, linear fitting is carried out on the rotating speed and the energy level of a plurality of sample points, and residual distribution of the energy level is calculated;
(5) and (4) respectively carrying out the operation of the step (4) on each frequency band to obtain a linear model.
8. The monitoring method according to claim 7, wherein: the on-line operation phase comprises the following steps:
(1) acquiring a group of audio signals and corresponding rotating speed values thereof in real time, and judging whether the rotating speed values are in a rotating speed range screened by training data;
(2) predicting the predicted energy level of the target frequency band according to the rotating speed by adopting the model obtained by training, and calculating the residual error between the energy level of the group of audio signals and the predicted energy level;
(3) selecting a plurality of groups of collected effective data, and calculating the average value of residual errors;
(4) estimating the damage level of the protective paint by adopting a prediction factor according to the residual distribution of the training data and the average value of the online running residual;
(5) when a plurality of frequency bands are selected for fitting, the final prediction factor takes the median of the prediction factors of each frequency band at the same time;
(6) and judging the stage of the fault according to the current factor and the historical factor by using the rule.
9. The monitoring method according to any one of claims 1 to 4, 6 to 8, wherein: after the corrosion degree is judged to exceed the set threshold value and early warning information is given, the monitoring method further comprises the following steps:
and displaying the corrosion damage rate according to the historical factor data, and giving a position with the damage rate exceeding a threshold value.
10. A system for monitoring erosion of a leading edge of a blade, comprising: comprising at least one sound sensor for collecting sound signals, at least one terminal collector, and a software system comprising an algorithm for performing the monitoring method according to any one of claims 1-9.
11. The monitoring system of claim 10, wherein: the software system comprises an algorithm module, a data management module, a configuration and user management module and a monitoring and early warning display module;
the algorithm module identifies the corrosion condition of the front edge of the blade according to the collected sound signal and the type of the protection scheme of the front edge of the blade, judges the damage degree, and sends out early warning information when the damage exceeds a set threshold value; the data management module manages the collected sound data and the data calculated by the algorithm module; the configuration and user management module is used for providing configuration input of a system parameter by a user and management and maintenance of user information; the monitoring and early warning display module is used for providing information display of monitoring and early warning results for a user.
12. The monitoring system according to claim 10 or 11, wherein: the software system is arranged on the machine-side collector, the station-side server or the cloud server.
13. The monitoring system according to claim 10 or 11, wherein: the monitoring system further comprises a collection cabinet, and the collection cabinet comprises at least one of a lightning protection module, a power filter, an air switch, a signal collector and a wireless transceiver module.
14. The monitoring system of claim 13, wherein: and a data processing device for performing condition filtering on the initial data is arranged in the acquisition cabinet.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011478576.0A CN112727704B (en) | 2020-12-15 | 2020-12-15 | Method and system for monitoring corrosion of leading edge of blade |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011478576.0A CN112727704B (en) | 2020-12-15 | 2020-12-15 | Method and system for monitoring corrosion of leading edge of blade |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112727704A CN112727704A (en) | 2021-04-30 |
| CN112727704B true CN112727704B (en) | 2021-11-30 |
Family
ID=75602329
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011478576.0A Active CN112727704B (en) | 2020-12-15 | 2020-12-15 | Method and system for monitoring corrosion of leading edge of blade |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112727704B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111306008B (en) * | 2019-12-31 | 2022-03-11 | 远景智能国际私人投资有限公司 | Fan blade detection method, device, equipment and storage medium |
| CN113205514A (en) * | 2021-05-27 | 2021-08-03 | 上海扩博智能技术有限公司 | Method, system, equipment and storage medium for acquiring fan blade corrosion information |
| CN114000989B (en) * | 2021-11-30 | 2023-07-18 | 中国华能集团清洁能源技术研究院有限公司 | Wind generating set blade aerodynamic performance attenuation detection method and test system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10418024B1 (en) * | 2018-04-17 | 2019-09-17 | Salesforce.Com, Inc. | Systems and methods of speech generation for target user given limited data |
| CN110763997A (en) * | 2019-11-04 | 2020-02-07 | 华北电力大学（保定） | Early fault early warning method for synchronous motor stator |
| CN110838302A (en) * | 2019-11-15 | 2020-02-25 | 北京天泽智云科技有限公司 | Audio frequency segmentation method based on signal energy peak identification |
| CN112067701A (en) * | 2020-09-07 | 2020-12-11 | 国电电力新疆新能源开发有限公司 | Fan blade remote auscultation method based on acoustic diagnosis |
Family Cites Families (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4195267B2 (en) * | 2002-03-14 | 2008-12-10 | インターナショナル・ビジネス・マシーンズ・コーポレーション | Speech recognition apparatus, speech recognition method and program thereof |
| US8712567B2 (en) * | 2006-12-22 | 2014-04-29 | Aviat U.S., Inc. | Manufactured product configuration |
| US8082115B2 (en) * | 2009-11-16 | 2011-12-20 | Nrg Systems, Inc. | Data acquisition system for condition-based maintenance |
| CN102750955B (en) * | 2012-07-20 | 2014-06-18 | 中国科学院自动化研究所 | Vocoder based on residual signal spectrum reconfiguration |
| US9395337B2 (en) * | 2013-03-15 | 2016-07-19 | Digital Wind Systems, Inc. | Nondestructive acoustic doppler testing of wind turbine blades from the ground during operation |
| CN103793712B (en) * | 2014-02-19 | 2017-02-08 | 华中科技大学 | Image recognition method and system based on edge geometric features |
| CN103971908B (en) * | 2014-05-06 | 2016-03-09 | 国家电网公司 | A kind of transformer noise suppressing method |
| AU2014101380A4 (en) * | 2014-11-14 | 2015-01-15 | TANG, Richard Zhao Kun Mr | SC-MH (The solar cube for micro-grid hub) |
| CN104677623B (en) * | 2015-03-16 | 2016-08-24 | 西安交通大学 | A kind of blade of wind-driven generator fault acoustics in place diagnostic method and monitoring system |
| US10403269B2 (en) * | 2015-03-27 | 2019-09-03 | Google Llc | Processing audio waveforms |
| JP7393947B2 (en) * | 2017-05-22 | 2023-12-07 | ジェネテシス エルエルシー | Mechanical identification of anomalies in bioelectromagnetic fields |
| EP3447731A1 (en) * | 2017-08-24 | 2019-02-27 | Agfa Nv | A method of generating an enhanced tomographic image of an object |
| KR102137151B1 (en) * | 2018-12-27 | 2020-07-24 | 엘지전자 주식회사 | Apparatus for noise canceling and method for the same |
| CN110705456A (en) * | 2019-09-29 | 2020-01-17 | 华南理工大学 | Micro motor abnormity detection method based on transfer learning |
| CN111306008B (en) * | 2019-12-31 | 2022-03-11 | 远景智能国际私人投资有限公司 | Fan blade detection method, device, equipment and storage medium |
| CN111306010B (en) * | 2020-04-17 | 2022-03-29 | 北京天泽智云科技有限公司 | Method and system for detecting lightning damage of fan blade |
| CN111601433B (en) * | 2020-05-08 | 2022-10-18 | 中国传媒大学 | Method and device for predicting stage lighting effect control strategy |
| CN111986699B (en) * | 2020-08-17 | 2023-07-04 | 西安电子科技大学 | Sound event detection method based on full convolution network |
-
2020
- 2020-12-15 CN CN202011478576.0A patent/CN112727704B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10418024B1 (en) * | 2018-04-17 | 2019-09-17 | Salesforce.Com, Inc. | Systems and methods of speech generation for target user given limited data |
| CN110763997A (en) * | 2019-11-04 | 2020-02-07 | 华北电力大学（保定） | Early fault early warning method for synchronous motor stator |
| CN110838302A (en) * | 2019-11-15 | 2020-02-25 | 北京天泽智云科技有限公司 | Audio frequency segmentation method based on signal energy peak identification |
| CN112067701A (en) * | 2020-09-07 | 2020-12-11 | 国电电力新疆新能源开发有限公司 | Fan blade remote auscultation method based on acoustic diagnosis |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度信念网络的猪咳嗽声识别;黎煊等;《农业机械学报》;20171128(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112727704A (en) | 2021-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112727704B (en) | Method and system for monitoring corrosion of leading edge of blade | |
| CN112233091B (en) | Wind turbine blade image damage detection and positioning method | |
| US7895018B2 (en) | Event monitoring via combination of signals | |
| US9347432B2 (en) | System and method for enhanced operation of wind parks | |
| US10371124B2 (en) | System and method for determining wind farm wake loss | |
| JP7199608B2 (en) | Methods and apparatus for inspecting wind turbine blades, and equipment and storage media therefor | |
| CN109763944B (en) | Non-contact monitoring system and monitoring method for blade faults of offshore wind turbine | |
| CN111539553B (en) | Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree | |
| CN111091298B (en) | Wind power plant flow field coupling characteristic evaluation and intelligent grouping method and system | |
| CN112324629A (en) | Wind power blade early damage monitoring system and method based on vibration and sound | |
| CN111415070A (en) | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data | |
| CN112727710B (en) | Wind field thunderbolt density statistical method and system based on audio signals | |
| CN110362045A (en) | A kind of marine double-fed fan motor unit fault distinguishing method considering maritime meteorology factor | |
| CN112067701B (en) | Fan blade remote auscultation method based on acoustic diagnosis | |
| CN112780503B (en) | Fan blade protective paint damage monitoring method and system based on audio signals | |
| EP3406897B1 (en) | System and method for determining wind farm wake loss | |
| CN112749820A (en) | Wind power prediction method and system | |
| CN219081774U (en) | Wind turbine blade state monitoring system | |
| CN111878326B (en) | Method and system for detecting freezing fault of fan, electronic equipment and medium | |
| CN116153333A (en) | Wind turbine blade fault diagnosis method based on aerodynamic noise | |
| CN115855465A (en) | System and method for monitoring state of wind turbine blade based on reverberation noise | |
| CN116292133B (en) | Method and device for detecting clearance abnormality of wind generating set | |
| CN115788803A (en) | Wind turbine blade icing detection method and system based on aerodynamic performance analysis | |
| CN116163894A (en) | Wind farm fan blade state detection method, system and storage medium | |
| CN117514652A (en) | Method and equipment for monitoring comprehensive state of wind turbine blade |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |