CN115409052A - Fault diagnosis method and system for wind generating set bearing under data imbalance - Google Patents
Fault diagnosis method and system for wind generating set bearing under data imbalance Download PDFInfo
- Publication number
- CN115409052A CN115409052A CN202210883053.7A CN202210883053A CN115409052A CN 115409052 A CN115409052 A CN 115409052A CN 202210883053 A CN202210883053 A CN 202210883053A CN 115409052 A CN115409052 A CN 115409052A
- Authority
- CN
- China
- Prior art keywords
- data
- fault
- meta
- fault diagnosis
- learning
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 97
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000003062 neural network model Methods 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 14
- 238000006243 chemical reaction Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000004931 aggregating effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000012423 maintenance Methods 0.000 abstract description 13
- 230000000694 effects Effects 0.000 abstract description 10
- 238000005096 rolling process Methods 0.000 description 11
- 238000013528 artificial neural network Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 8
- 238000013136 deep learning model Methods 0.000 description 8
- 238000010248 power generation Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000007547 defect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000035945 sensitivity Effects 0.000 description 6
- 101100455978 Arabidopsis thaliana MAM1 gene Proteins 0.000 description 5
- 238000011161 development Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000003137 locomotive effect Effects 0.000 description 5
- 238000013508 migration Methods 0.000 description 5
- 230000005012 migration Effects 0.000 description 5
- 238000013526 transfer learning Methods 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000012952 Resampling Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 229910052799 carbon Inorganic materials 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 238000004141 dimensional analysis Methods 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000036039 immunity Effects 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008092 positive effect Effects 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012795 verification 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Computational Linguistics (AREA)
- Acoustics & Sound (AREA)
- Molecular Biology (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention belongs to the technical field of wind power bearing fault diagnosis, and discloses a fault diagnosis method and system for a wind generating set bearing under data imbalance, wherein continuous vibration signals of a plurality of time periods are collected through a signal acquisition device; performing STFT (time-frequency transformation) on original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the characteristic information of the original signals as an original data sample; constructing a meta-learning data set for meta-training and meta-testing of the deep neural network model according to the meta-learning strategy on the two-dimensional image data with unbalanced quantity; and constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance through training and fine adjustment work of a plurality of element learning subtask sets on the network. The bearing fault diagnosis method and the bearing fault diagnosis device have more obvious bearing fault diagnosis effect under complex working conditions, provide theoretical basis and technical reference for cross-equipment fault diagnosis work in the field of wind power operation and maintenance in future, and reduce technical gap between operation and maintenance of different equipment.
Description
Technical Field
The invention belongs to the technical field of wind power bearing fault diagnosis, and particularly relates to a fault diagnosis method and system for a wind generating set bearing under data imbalance.
Background
At present, with the implementation of the two-carbon policy, the new energy industry in China is undergoing a great revolution. Wind power generation is one of the most mature, most scale development conditions and power generation modes with commercial development prospects in the field of renewable energy sources, and will become the main battlefield of future new energy revolution, and wind power will play an important role. As the most important complex mechanical part in wind power generation, the rolling bearing of the generator is in a high-load operation state for a long time, and the performance of the rolling bearing directly influences the operation state of the whole device, thereby influencing the safety of industrial production. Therefore, the operation and maintenance work of the rolling bearing becomes important. At present, mechanical vibration signals collected by industry contain abundant bearing working condition information, and have important significance for understanding the health state of the inside of machinery. Therefore, from the professional knowledge in the mechanical field, the scholars extract important characteristic information capable of reflecting the bearing state from the non-stationary vibration signal through a signal processing technology. For example: li et al propose a method that combines improved adaptive parameterless Empirical Wavelet Transform (EWT) and adaptive coefficient coding shrinkage de-noising (ASCSD) algorithms to enhance periodic pulse characteristics, thereby improving the accuracy of rolling bearing fault identification. Sun et al propose a bearing diagnostic method based on Empirical Mode Decomposition (EMD) and improved chebyshev distance. The above-mentioned bearing diagnosis method based on signal processing technique achieves very excellent results. However, in these methods, there is a case where the data is too single, so that the application range of the model is very strictly limited. For example, EMD, which is very sensitive to environmental noise, the IMF signal component decomposed by EMD does not necessarily fit the vibration signal itself when the acquired signal is affected by noise beyond a certain range. The obtained fault diagnosis model cannot effectively diagnose the original vibration data. Therefore, for industrial application, how to enable the model to quickly understand the characterization information of the bearing in different fault states on the basis of not spending a large amount of learning time cost becomes a core value of industrial fault identification application. Machine learning draws the attention of researchers by way of data to fit complex functions representing the nature of the distribution of things. The K nearest neighbor method, the Bayes algorithm, the decision tree, the Support Vector Machine (SVM) and the like are well applied to bearing fault diagnosis. However, the machine learning models are shallow in structure and are often poor in application effect when applied to complex working condition environments.
With the successful application of the deep learning model CNN in the field of image classification, a series of fault diagnosis applications based on the deep neural network emerge endlessly. However, these models often need to be trained correctly under the condition of data balance, because when the data is unbalanced, the models tend to pay heavy attention and learn to most samples, and the accuracy obtained by the models is high overall. From an industrial application perspective, the few classes of samples often represent anomaly and fault data, which is the part of the engineer that needs significant attention. Therefore, the fault diagnosis of the network model under the data imbalance can be regarded as a few-sample learning problem. Common few-sample learning methods include data enhancement, transfer learning and original learning. The data enhancement techniques proceed by fitting the original data to generate new samples, which become the equilibrium data set. The transfer learning technology is characterized in that a model is trained under a large number of balanced data sets, fine tuning work is carried out when the trained model is applied to a data set with data imbalance, and the difference between the two data sets is calculated and reduced to be stable, so that a fault diagnosis model applicable to a data imbalance scene can be obtained. However, the two technologies have certain disadvantages:
data enhancement techniques: if SMOTE exists data enhancement, the influence of noise becomes large, so that the classification boundary between classes is blurred, and the diagnosis precision is reduced. The GAN has the problems of instability, modal collapse, weak gradient, etc. of the generated samples due to the instability of random noise signals and the definition mode of generating the anti-network loss function.
Transfer learning: the premise of the transfer learning is that a large and data-balanced data set with source domain data close to a target domain exists for the prior knowledge learning of the model. However, in the field of fault diagnosis of rotating machines, no data set can be pretrained well to perform model migration. Therefore, the main application of the migration learning is mainly the migration between different working conditions of the same equipment, and the application between cross-machines is still not mature.
The prior art includes:
(1) The signal processing technology comprises the following steps: extracting the characteristics of different analysis domains (time domain, frequency domain and time frequency) to form an original characteristic data set;
(2) Sample generation techniques: SMOTE (synthetic minority class of oversampling techniques), GAN (generating an antagonistic network);
(3) The model carries out characteristic screening from the aspects of evaluation criteria such as distance, information entropy, relevance and the like according to characteristic values, and finally selects characteristics with higher sensitivity to carry out bearing fault diagnosis.
The current method for diagnosing the fault of the wind generating set bearing under the condition of data unbalance mainly has two problems:
(1) The PHM system related to wind power operation and maintenance in most of the existing wind farms has defects in the aspect of intelligent signal processing function. The motor bearing is used as a rotating mechanical part with a complex structure, can generate various different types (even unknown types) of nonlinear vibration signals containing impact attenuation in severe and complex operating environments, but a mechanism characteristic evaluation model widely used at present has certain limitation, and a characteristic data set extracted from a common time domain, a common frequency domain and a common time-frequency domain cannot comprehensively depict the condition of state information of mechanical equipment represented by the vibration signals. And certain defects exist in the stability and generalization capability of the model under big data.
(2) During the service period of the wind turbine generator, most of the data collected by the PHM system are sample data in a normal state. The fault sample data used for model training is rare, so that the trained model has weak self-adaptive capability, and the sensitivity to the data in a fault state is greatly reduced, thereby influencing the precision of the model on fault diagnosis. Currently, for such training sample imbalance problem, data enhancement and model cross-domain migration techniques are commonly used. Where data enhancement is more commonly applied in various domains. The new sample data is generated by resampling of samples or GAN (generation of countermeasure network) technique to achieve the effect of training sample balance. However, since these techniques only generate samples of known classes and have weak noise immunity, the finally trained model has poor robustness. In the fault diagnosis of the bearing, the instability of a diagnosis model is easily caused, so that the learning and self fine adjustment of the model on subsequent data are influenced, and the integral classification performance and the accuracy of the fault diagnosis are further influenced.
In recent years, in order to establish a sound green low-carbon circular development economic system, china vigorously develops clean energy. Wind power generation is one of the most mature and most extensive development conditions and power generation modes with commercial development prospects in the field of renewable energy sources, and will become the main battlefield of future new energy revolution, and the wind power industry will play an important role. As the most important complex mechanical part in wind power generation equipment, a rolling bearing of a fan unit is in a high-load operation state for a long time, and the performance of the rolling bearing directly influences the operation state of the whole device, so that the safety of industrial production is influenced. In fact, in all machines, the failure of the rotating machine accounts for 80%, while about 70% of them are caused by the rolling bearing. Therefore, the equipment operation and maintenance work for the rolling bearing becomes important. The traditional fault diagnosis model can achieve high-precision fault discrimination capability only under the training of a large amount of balance sample data. However, in practical applications, due to the complexity of the actual working conditions and the lack of fault samples, it is very difficult to train a fault diagnosis model that works effectively under the complex working conditions. Meanwhile, the definition of the bearing fault category in some extreme states is very fuzzy, which easily causes trouble to the operation and maintenance work of mechanical equipment, thereby causing the increase of the operation and maintenance cost and even the casualties of personnel. Therefore, the fault diagnosis method for the wind generating set bearing under data imbalance is constructed, stability and reliability of the bearing fault diagnosis and identification result are improved, and the fault diagnosis method has important significance on normal operation and safe maintenance of wind generating equipment.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) In most of existing PHM systems related to wind power operation and maintenance in wind farms, the used mechanism characteristic evaluation model has defects in the aspect of intelligent signal processing function, and the condition of state information of mechanical equipment represented by a vibration signal cannot be comprehensively described, so that the stability and generalization capability of the model are poor.
(2) The existing training samples have the problem of unbalance, namely fault sample data for model training are rare, so that the trained model has weak adaptive capacity, the sensitivity to data in a fault state is greatly reduced, and the precision of the model on fault diagnosis is influenced.
(3) The data enhancement and model cross-domain migration technology only generates samples of known classes and is weak in noise immunity, so that the finally trained model is poor in robustness.
(4) In the fault diagnosis of the bearing, the prior art is easy to cause the instability of a diagnosis model, so that the learning and self fine adjustment of the model on subsequent data are influenced, and the integral classification performance and the fault diagnosis accuracy are further influenced.
(5) The traditional fault diagnosis model can achieve high-precision fault discrimination capability only under the training of a large amount of balance sample data, but in practical application, due to the complexity of actual working conditions and the lack of fault samples, the training of a fault diagnosis model which works effectively under the complex working conditions is very difficult.
(6) The definition of the bearing fault category in some extreme states is very fuzzy, and is easy to cause trouble to the operation and maintenance work of mechanical equipment, thereby causing the increase of the operation and maintenance cost and even the casualties of people.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method, a system, a medium, equipment and a terminal for diagnosing the fault of a wind generating set bearing under data imbalance, and particularly relates to a method and a system for diagnosing the fault of the wind generating set bearing under data imbalance by using a novel meta-learning framework.
The invention is realized in such a way, the fault diagnosis method of the wind generating set bearing under the condition of data imbalance comprises the following steps:
when a generator set bearing in the wind driven generator carries out wind power generation work outdoors, equipment information acquisition and state monitoring are carried out on the generator set bearing in real time; the method combines a signal processing technology, a deep convolution neural network, transfer learning and a meta learning technology to realize fault characterization, model construction and optimization of a rear bearing of a generator of the wind turbine generator, and ensures normal operation of the wind turbine generator through uninterrupted analysis of received data by a system.
Further, the fault diagnosis method for the wind generating set bearing under the data unbalance comprises the following steps:
collecting continuous vibration signals of a plurality of time periods through an accelerometer sensor signal acquisition device;
performing STFT time-frequency conversion on the original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the characteristic information of the original signals as an original data sample;
step three, constructing a meta-learning data set for the unbalanced two-dimensional image data according to a meta-learning strategy, and using the meta-learning data set for meta-training and meta-testing of the deep neural network model;
and step four, training and fine-tuning the network through a plurality of element learning subtask sets, and constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance.
Further, in the second step, the short-time Fourier transform (STFT) is adopted to convert the original signal into a two-dimensional time-frequency image, and the STFT passes through a window functionWill signalCut into a plurality of time periods, and for each time axisThe FFT is performed on the signal segment to finally form a two-dimensional image with time-frequency domain information.
Wherein, the mathematical formula of the STFT is as follows:
wherein,respectively representing the angular frequency and time of the called number;expressing the Euler formula-j(ii) a N represents the number of signal sampling points after windowing, and a hanning window function with the scale size of N =1024 is applied, as shown in the following formula:
further, in the third step, a meta-learning method based on gradient optimization is adopted to learn previous experiences from multiple related tasks and rely on different learning subtasksAccumulate different element knowledge。
After constructing a meta-learning data set, inputting a subtask set into a deep neural network model with a multi-level feature fusion function; and aggregating the bottom-layer features and the middle-layer features through residual learning identity mapping, so that the model determines the fault categories corresponding to the feature vectors under different view planes.
Another object of the present invention is to provide a system for diagnosing a fault of a bearing of a wind turbine generator set under data imbalance, which applies the method for diagnosing a fault of a bearing of a wind turbine generator set under data imbalance, the system for diagnosing a fault of a bearing of a wind turbine generator set under data imbalance includes:
the signal acquisition module is used for collecting continuous vibration signals of a plurality of time periods through the signal acquisition device;
the time-frequency conversion module is used for carrying out STFT (standard time transform) time-frequency conversion on the original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the time-frequency characteristic information of the original signals as an original data sample;
the data set construction module is used for constructing a meta-learning data set for meta-training and meta-testing of the deep neural network model according to the meta-learning strategy on the unbalanced two-dimensional image data;
and the fault diagnosis module is used for training and fine-tuning the network through a plurality of meta-learning subtask sets, and constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance.
Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, causes the processor to execute the method for diagnosing the fault of the bearing of the wind generating set under the data imbalance.
Another object of the present invention is to provide a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the processor is enabled to execute the method for diagnosing the fault of the wind generating set bearing under the data imbalance.
It is another object of the present invention to provide a computer program product stored on a computer readable medium, comprising a computer readable program for providing a user input interface for applying said system for diagnosing a fault of a wind turbine generator set bearing under data imbalance when executed on an electronic device.
Another object of the present invention is to provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to apply the system for diagnosing a fault of a bearing of a wind turbine generator system under imbalance of data.
The invention also aims to provide an information data processing terminal, which is used for realizing the fault diagnosis system of the wind generating set bearing under the data unbalance.
In combination with the technical solutions and the technical problems to be solved, please analyze the advantages and positive effects of the technical solutions to be protected in the present invention from the following aspects:
first, aiming at the technical problems existing in the prior art and the difficulty in solving the problems, the technical problems to be solved by the technical scheme of the present invention are closely combined with the technical scheme to be protected and the results and data in the research and development process, and some creative technical effects brought after the problems are solved are analyzed in detail and deeply. The specific description is as follows:
the invention collects continuous vibration signals of a plurality of time periods through an accelerometer sensor signal acquisition device; therefore, the original bearing vibration signal corresponding to the current acceleration of the bearing can be detected, and the condition state identification of the health state of the data can be favorably and correctly carried out. The original signal is only a one-dimensional time domain signal, and needs to be converted into a frequency domain signal through Fast Fourier Transform (FFT) to obtain stable frequency domain characteristics for learning. However, the existing model only depends on one-dimensional analysis of a time domain and a frequency domain, and cannot capture internal information of the bearing in various fault states for learning and transferring application. Therefore, two-dimensional time-frequency analysis has become an important method for analyzing mechanical vibration signals to perform fault diagnosis. According to the invention, different time resolutions and frequency resolutions are defined by STFT through different scales of window functions, and different combinations are performed on STFT parameters to obtain the optimal time-frequency image input scale, so that the model can extract fault characteristic information more efficiently.
According to the method, a meta-learning data set is constructed for unbalanced two-dimensional image data according to a meta-learning strategy and is used for meta-training and meta-testing of a deep neural network model; and (3) carrying out scene simulation training on the overall unbalanced data set to be divided into a plurality of meta-task sets, and refining the learning of the model on data details. According to the method, a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance is constructed through training and fine adjustment work of a plurality of element learning subtask sets on the network. Compared with a traditional deep learning model, the difference between the interior and the whole of the meta-task is refined through the internal and external circulation parts in the meta-learning process, and therefore the fault diagnosis precision of the model under the condition of data imbalance is improved.
The invention applies the time-frequency image conversion technology (STFT) to the field of rolling bearing fault diagnosis: the method makes up the defects of the one-dimensional signals in the feature expression content, the fault signal features contained in the two-dimensional image are more comprehensive, and more complex structural distribution can be expressed, so that the model can learn the hidden high-level features which can represent the essence of the fault state. The method is applied to the field of wind power bearing fault diagnosis for the first time.
The invention applies the meta-learning (MAML) based on gradient optimization to the field of rolling bearing fault diagnosis: the meta-learning method is based on a gradient synthesis optimization mode and is carried out on the basis of information entropy optimization aiming at the problem of unbalanced training data. On the premise of solving data imbalance, the recognition sensitivity of the model to different types of fault states is balanced. The method is applied to the field of wind power bearing fault diagnosis for the first time.
The invention applies a deep learning model with multi-level feature fusion to the field of rolling bearing fault diagnosis: firstly, self-adaptive feature extraction is carried out on a signal time-frequency image after two-dimensional conversion through a deep convolution neural network. Combining the extracted features of different levels through a residual error module, thereby obtaining high-dimensional fusion features for multi-classification evaluation; and finally, optimizing the evaluation capability of the model through a plurality of data subsets to finally obtain a diagnosis model capable of accurately identifying different bearing fault states in a data imbalance state. The method is applied to the field of wind power bearing fault diagnosis for the first time.
Due to the particularity of the operating environment of the rotary bearing of the wind generating set, the detected fault mode of the rotary bearing presents stronger complexity and variability, so that the bearing fault diagnosis method based on the traditional mode can cause certain hidden troubles to the safe maintenance work of the bearing. The scheme mainly solves the problems of two aspects:
1. aiming at the current mainstream signal processing method, the state information of the mechanical equipment is comprehensively represented by partial mechanism characteristics extracted from a time domain, a frequency domain and a time-frequency domain. According to the scheme, the time-frequency domain characteristics of the original signal are converted into the two-dimensional image through the time-frequency processing technology, so that more complex signal content can be expressed, and the learning capability of the model on the high-grade characteristics of the signal is enhanced.
2. By means of the meta-learning strategy for the data, the problem of how to efficiently apply a small amount of fault data information is analyzed so as to solve the problems of the data imbalance phenomenon and how to balance the recognition sensitivity of a model to different fault category data under the condition of data imbalance. The model can effectively utilize data to improve the learning ability of the model while reducing the training data amount.
Secondly, considering the technical scheme as a whole or from the perspective of products, the technical effect and advantages of the technical scheme to be protected by the invention are specifically described as follows:
the fault diagnosis model of the wind generating set bearing under the data imbalance is deployed on a wind generating set PHM platform developed by the institute of electric locomotives, china, taoise, china, and the new model is tested and verified through historical data of a wind power part of a middle-sized vehicle.
Third, as inventive supplementary proof of the claims of the present invention, there are several important aspects as follows:
(1) The expected income and commercial value after the technical scheme of the invention is converted are as follows: the invention combines a PHM platform of a wind turbine generator developed by the institute of Electrical locomotives of Zhonghua province in China to carry out operation and maintenance work on the wind turbine generator under complex environment. After the scheme is implemented, an engineer does not need to perform a large amount of data preprocessing work to balance the problem of training data imbalance. Under the condition of ensuring that effective fault data exists, the model can greatly reduce the work of collection, verification, identification and the like of the fault data. Enabling rapid learning and generalization of the model to other applications.
(2) The technical scheme of the invention solves the technical problem that people are eagerly to solve but can not be successfully solved all the time: the technical scheme of the invention solves the problem that the traditional deep learning model can not train an effective fault diagnosis model under the conditions of unbalanced data and complex working conditions. The method utilizes double-layer optimization of parameters to enable a model to refine learning of fault characteristics. Under the condition of data driving, the self-learning process of the model is completed, so that the optimized model can be more effectively applied to other directions.
(3) The technical scheme of the invention overcomes the technical prejudice whether: the method is different from a traditional deep learning model in solving the problem of fan bearing fault diagnosis under data imbalance. Most of the traditional data enhancement modes for balancing data are based on resampling or generating auxiliary samples capable of fitting original fault data, and certain defects exist in authenticity and diversity. The meta-learning strategy starts from local data and enables the model to learn the association and difference between samples by self. Ensuring its authenticity. Meanwhile, in order to ensure the diversity of training data, a data set is divided into a plurality of meta tasks according to K-way N-shot, and the detail learning capacity of the meta learning diagnosis model is enhanced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of a fault diagnosis method for a wind generating set bearing under data imbalance according to an embodiment of the invention;
FIG. 2 is a schematic flow chart of a deep learning fault diagnosis model based on meta learning according to an embodiment of the present invention;
FIG. 3 is a structural block diagram of a fault diagnosis system of a wind turbine generator system bearing under data imbalance according to an embodiment of the invention;
FIG. 4 is a schematic diagram of STFT time-frequency transformation of an original vibration signal according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a MAML algorithm provided by an embodiment of the present invention;
fig. 6 (a) is a schematic diagram of the fault diagnosis accuracy (precision) of the fault diagnosis model trained by the neural network with different iteration numbers in the training set and the test set respectively according to the embodiment of the present invention;
FIG. 6 (b) is a graph of the output value (loss) of the neural network loss function for different iterations provided by the implementation of the present invention;
FIG. 7 is a flowchart of a method for diagnosing a fault of a bearing of a wind turbine generator system under data imbalance according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides a method and a system for diagnosing the fault of a wind generating set bearing under data imbalance, and the invention is described in detail below with reference to the attached drawings.
1. The embodiments are explained. This section is an explanatory embodiment expanding on the claims so as to fully understand how the present invention is embodied by those skilled in the art.
As shown in fig. 7, the method for diagnosing a fault of a bearing of a wind turbine generator set under data imbalance according to the embodiment of the present invention includes the following steps:
s101, collecting continuous vibration signals in a plurality of time periods through an accelerometer sensor signal collecting device;
s102, performing STFT time-frequency conversion on original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing original signal time characteristic information as an original data sample;
s103, constructing a meta-learning data set for the meta-training and the meta-testing of the deep neural network model according to the meta-learning strategy on the unbalanced two-dimensional image data;
and S104, constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance through training and fine adjustment work of the plurality of element learning subtask sets on the network.
The schematic diagram of the fault diagnosis method for the wind generating set bearing under the data imbalance provided by the embodiment of the invention is shown in fig. 1.
The fault diagnosis system for the wind generating set bearing under the data imbalance provided by the embodiment of the invention comprises:
the signal acquisition module is used for collecting continuous vibration signals of a plurality of time periods through the signal acquisition device;
the time-frequency conversion module is used for carrying out STFT (standard time transform) time-frequency conversion on the original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the time-frequency characteristic information of the original signals as an original data sample;
the data set construction module is used for constructing a meta-learning data set for meta-training and meta-testing of the deep neural network model according to the meta-learning strategy on the unbalanced two-dimensional image data;
and the fault diagnosis module is used for training and fine-tuning the network through a plurality of meta-learning subtask sets, and constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance.
The technical solution of the present invention is further described below with reference to specific examples.
The wind turbine generator bearing fault diagnosis method provided by the invention takes a generator set bearing in a wind turbine generator as a research object, when the wind turbine generator works outdoors, the generator set bearing is subjected to equipment information acquisition and state monitoring in real time, and the normal operation of the wind turbine generator is ensured through the uninterrupted analysis of the received data by the system. The fault diagnosis model of the wind generating set bearing under the data imbalance is deployed on a wind generating set PHM platform developed by the institute of electric locomotives, china, taoise, china, and the new model is tested and verified through historical data of a wind power part of a middle-sized vehicle.
According to the invention, continuous vibration signals of a plurality of time periods are collected through a signal acquisition device, and STFT time-frequency conversion is carried out on original bearing fault vibration signals in different fault states, so that a two-dimensional image containing characteristic information of the original signals is obtained and used as an original data sample. And then, constructing a meta-learning data set for the unbalanced two-dimensional image data according to a meta-learning strategy for meta-training and meta-testing of the deep neural network model. And finally, training and fine-tuning the network through a plurality of element learning subtask sets to obtain a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance.
The invention collects continuous vibration signals of a plurality of time periods through an accelerometer sensor signal acquisition device; therefore, the original bearing vibration signal corresponding to the current acceleration of the bearing can be detected, and the condition state identification of the health state of the data can be correctly performed. The original signal is only a one-dimensional time domain signal, and needs to be converted into a frequency domain signal through Fast Fourier Transform (FFT) to obtain stable frequency domain characteristics for learning. However, the existing model only depends on one-dimensional analysis of a time domain and a frequency domain, and cannot capture internal information of the bearing in various fault states for learning and transferring application. Therefore, two-dimensional time-frequency analysis has become an important method for analyzing mechanical vibration signals to perform fault diagnosis. According to the invention, different time resolutions and frequency resolutions are defined by STFT through different scales of window functions, and different combinations are performed on STFT parameters to obtain the optimal time-frequency image input scale, so that the model can extract fault characteristic information more efficiently.
According to the method, a meta-learning data set is constructed for unbalanced two-dimensional image data according to a meta-learning strategy and is used for meta-training and meta-testing of a deep neural network model; and (3) carrying out scene simulation training on the overall unbalanced data set to be divided into a plurality of meta-task sets, and refining the learning of the model on data details. According to the method, a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance is constructed through training and fine adjustment work of a plurality of element learning subtask sets on the network. Compared with a traditional deep learning model, the difference between the interior and the whole of the meta-task is refined through the internal and external circulation parts in the meta-learning process, and therefore the fault diagnosis precision of the model under the condition of data imbalance is improved.
As shown in fig. 2, the overall technical process of the scheme can be summarized, and the feasibility of the method is verified through a "wind turbine PHM platform" developed by institute of electric locomotives in zhou guzhou province, limited. The obtained diagnosis result is as follows: under the condition that the training samples are unbalanced, compared with the traditional neural network model, the training process of the model can be converged quickly, and the accuracy of the final fault diagnosis is higher than that of the traditional fault diagnosis model.
In analyzing machine learning for bearing fault diagnosis applications, it is often necessary to perform a significant amount of preparation work for the feature engineering section. The conventional vibration signal characteristic criterion can only obtain effective results in an ideal experimental environment. However, for an industrial application scenario of wind turbine generator bearing fault diagnosis, a more complex and variable operating environment needs to explore deeper hidden layer features for representing more complex fault categories. Therefore, compared with a mode of converting a time domain signal into a one-dimensional frequency domain signal through Fast Fourier Transform (FFT) in the conventional signal processing so as to obtain stable frequency domain characteristics thereof for model learning, the two-dimensional time-frequency analysis provided by the invention can capture the internal information of the bearing in each fault state, as shown in fig. 4. The invention adopts Short Time Fourier Transform (STFT) to convert the original signal into a two-dimensional Time-frequency image. STFT pass window functionWill signalCut into a plurality of time periods, and for each time axisThe signal segments are FFT processed to finally form a two-dimensional image with time-frequency domain information. The mathematical formula of STFT is shown in formula (1).
Wherein,respectively representing the angular frequency and time of the called number;expressing the Euler formula-j(ii) a n represents the number of signal sampling points after windowing; in order to reduce the frequency spectrum leakage, a Hanning window function with the size of N =1024 is applied. As shown in formula (2).,Referred to as signal frequency (hertz);
a large number of analyses prove that under the condition that the number of training data of the deep neural network diagnosis model is unbalanced, the sensitivity of deep learning model fault diagnosis is biased to the fault category with a large number of training samples. Such a result often causes few sample failure types in some extremely complicated cases to be unable to be identified and detected by the model. And is often a significant accident consequence of such failures. Therefore, aiming at the problem of few samples, the invention adopts a meta-learning method based on gradient optimization. The concept of meta-learning is to be able to learn previous experiences from multiple related tasks and to improve its performance in the target domain task by means of different meta-knowledge θ _ i accumulated across different learning sub-tasks T _ i, as shown in fig. 5.
The method enables the model to have stronger generalization capability and faster adaptability when being applied to task sets p (T) with different distributions, and is considered to be capable of effectively solving the problem of learning with few samples. The invention solves the data imbalance problem based on the MAML algorithm framework. In terms of overall thinking, the MAML algorithm focuses on improving the task learning capability of the model as a whole, but not the capability of solving some specific tasks. The aim is to train a model which can quickly adapt to a new task through a small amount of sample data under a small amount of training iteration times. And is generally used to solve the classification problem of data imbalance in the case of few samples.
In the field of pattern recognition, a CNN model in deep learning has been attracting attention as one of the classical models in fault diagnosis applications. Wherein the feature adaptive extraction part related to the convolution pooling layer is always favored by the CV boundary image classification application. However, for different data sources, different neural network parameters greatly influence the final classification effect. Therefore, people often choose a huge number of data sets containing almost all known classes as a training set to train a good applicable model. However, in the industry, different devices and environments affect the data itself. Compared with the method for acquiring the original vibration information data of almost all bearing devices as the training data set, the method is not as challenging as how to make the model more intelligent to improve the learning ability of the model to deal with unknown data. Therefore, after the meta-learning data set is constructed, the subtask set is input into the deep neural network model with the multi-level feature fusion function. As shown in the in-depth neural network model architecture portion of fig. 4. And aggregating the bottom-layer features and the middle-layer features through residual learning identity mapping, so that the model can more comprehensively know the fault categories corresponding to the feature vectors under different view planes.
2. Evidence of the relevant effects of the examples. The embodiment of the invention has some positive effects in the process of research and development or use, and indeed has great advantages compared with the prior art, and the following contents are described by combining data, charts and the like in the test process.
According to a bearing fault diagnosis model which is successfully deployed in a PHM platform of a wind turbine generator and is developed by the institute of electric locomotives in Zhongzhu province, the invention selects a model with the first three comprehensive diagnosis capabilities for comparison, and the experimental comparison result is shown in the table 1. The classification accuracy of the model trained in the three fault diagnosis experiments is higher than that of other models.
TABLE 1 Classification accuracy of different diagnostic methods
The accuracy rate is over 99 percent, which shows that the method has good fault diagnosis effect under the condition of data set unbalance. Due to the reason that the data set is unbalanced, the fault classification learning effect of a part of decision trees in the model is poor, and the accuracy is reduced. And partial effective fault characteristic information is lost in the noise filtering process of the VMD _ SVM model, so that the data set unbalance condition still exists. However, like the method of the present invention, the WGAN _ CNN performs a local data enhancement on the original data set with respect to the data imbalance problem. However, the WGAN enhanced fault samples are close to the original fault data distribution in data distribution, so the model is not really trained in the case of data equalization. Therefore, the information in the table can be used for obtaining that the deep learning model based on the MAML algorithm has a stronger classification effect for the bearing fault diagnosis method.
The practical application result provided by the embodiment of the invention is verified as shown in fig. 6, and whether the fault diagnosis model has the fault diagnosis capability under the training of the deep neural network is verified; fig. 6 (a) represents the failure diagnosis accuracy (precision) of the failure diagnosis model trained by the neural network with different iteration numbers in the training set and the test set data, respectively; fig. 6 (b) represents the output value (loss) of the neural network loss function for different iteration numbers.
It should be noted that the embodiments of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. It will be appreciated by those skilled in the art that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, for example such code provided on a carrier medium such as a diskette, CD-or DVD-ROM, a programmable memory such as read-only memory (firmware) or a data carrier such as an optical or electronic signal carrier. The apparatus and its modules of the present invention may be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of hardware circuits and software, e.g., firmware.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.
Claims (9)
1. A fault diagnosis method for a wind generating set bearing under data imbalance is characterized by comprising the following steps:
collecting continuous vibration signals of a plurality of time periods through an accelerometer sensor signal acquisition device;
performing STFT (space time transformation) on the original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the characteristic information of the original signals as an original data sample;
step three, constructing a meta-learning data set for meta-training and meta-testing of the deep neural network model according to the meta-learning strategy on the unbalanced two-dimensional image data;
and step four, training and fine-tuning the network through a plurality of element learning subtask sets, and constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance.
2. The method for diagnosing the fault of the bearing of the wind generating set under the data imbalance according to claim 1, wherein in the second step, a Short Time Fourier Transform (STFT) is adopted to convert an original signal into a two-dimensional time-frequency image, and the STFT passes through a window functionWill signalIs cut into a plurality of time segments and is respectively time axisPerforming FFT on the signal section to finally form a two-dimensional image with time-frequency domain information;
wherein the mathematical formula of the STFT is as follows:
wherein,respectively representing the angular frequency and time of the called number;expressing the Euler formula-j(ii) a n represents the number of signal sampling points after windowing;
a hanning window function with a scale size of N =1024 is applied as shown in the following equation:
3. the method for diagnosing the fault of the wind generating set bearing under the data imbalance according to claim 1, wherein in the third step, a meta-learning method based on gradient optimization is adopted to learn the previous experience from a plurality of related tasks, and different meta-knowledge θ _ i is accumulated by means of different learning subtasks T _ i;
after constructing a meta-learning data set, inputting a subtask set into a deep neural network model with a multi-level feature fusion function; and aggregating the bottom-layer characteristics and the middle-layer characteristics through identity mapping of residual learning, so that the model determines the fault categories corresponding to the characteristic vectors under different view planes.
4. A fault diagnosis system of a wind generating set bearing under data imbalance, applying the fault diagnosis method of the wind generating set bearing under data imbalance according to any one of claims 1 to 3, wherein the fault diagnosis system of the wind generating set bearing under data imbalance comprises:
the signal acquisition module is used for collecting continuous vibration signals of a plurality of time periods through the signal acquisition device;
the time-frequency conversion module is used for carrying out STFT (standard time transform) time-frequency conversion on the original bearing fault vibration signals in different fault states to obtain a two-dimensional image containing the time-frequency characteristic information of the original signals as an original data sample;
the data set construction module is used for constructing a meta-learning data set for meta-training and meta-testing of the deep neural network model according to the meta-learning strategy on the unbalanced two-dimensional image data;
and the fault diagnosis module is used for constructing a fault diagnosis model suitable for the wind generating set bearing under the condition of data imbalance through training and fine adjustment work of the plurality of element learning subtask sets on the network.
5. A computer arrangement, characterized in that the computer arrangement comprises a memory and a processor, the memory stores a computer program, and the computer program when executed by the processor causes the processor to execute the method for diagnosing the fault of the bearing of the wind generating set under the data unbalance according to any one of claims 1 to 3.
6. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, causes the processor to execute the method for diagnosing the fault of the bearing of the wind generating set under the data imbalance according to any one of claims 1 to 3.
7. A computer program product stored on a computer readable medium, comprising a computer readable program that, when executed on an electronic device, provides a user input interface to apply the system for diagnosing a fault in a wind turbine generator system bearing with data imbalance of claim 4.
8. A computer readable storage medium storing instructions that, when executed on a computer, cause the computer to apply the system for diagnosing a failure of a wind turbine generator system bearing under imbalance of data as set forth in claim 4.
9. An information data processing terminal, characterized in that the information data processing terminal is used for realizing the fault diagnosis system of the wind generating set bearing under the data unbalance as claimed in claim 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210883053.7A CN115409052A (en) | 2022-07-26 | 2022-07-26 | Fault diagnosis method and system for wind generating set bearing under data imbalance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210883053.7A CN115409052A (en) | 2022-07-26 | 2022-07-26 | Fault diagnosis method and system for wind generating set bearing under data imbalance |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115409052A true CN115409052A (en) | 2022-11-29 |
Family
ID=84157124
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210883053.7A Pending CN115409052A (en) | 2022-07-26 | 2022-07-26 | Fault diagnosis method and system for wind generating set bearing under data imbalance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115409052A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116448425A (en) * | 2023-03-09 | 2023-07-18 | 江苏波司登科技有限公司 | Conveyor belt bearing remote fault diagnosis method and system based on improved random forest |
CN116465630A (en) * | 2023-02-16 | 2023-07-21 | 苏州大学 | Bearing small sample fault diagnosis method and system based on meta-shift learning |
CN117350067A (en) * | 2023-10-19 | 2024-01-05 | 江西嘉逸陶瓷有限公司 | Standard sample generation method and device applied to ceramic nondestructive testing |
CN117574273A (en) * | 2023-11-03 | 2024-02-20 | 佛山职业技术学院 | Fan state and fault diagnosis method based on operation and maintenance big data and application |
CN117668623A (en) * | 2024-02-02 | 2024-03-08 | 中国海洋大学 | Multi-sensor cross-domain fault diagnosis method for leakage of ship pipeline valve |
-
2022
- 2022-07-26 CN CN202210883053.7A patent/CN115409052A/en active Pending
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116465630A (en) * | 2023-02-16 | 2023-07-21 | 苏州大学 | Bearing small sample fault diagnosis method and system based on meta-shift learning |
CN116448425A (en) * | 2023-03-09 | 2023-07-18 | 江苏波司登科技有限公司 | Conveyor belt bearing remote fault diagnosis method and system based on improved random forest |
CN116448425B (en) * | 2023-03-09 | 2023-11-21 | 江苏波司登科技有限公司 | Conveyor belt bearing remote fault diagnosis method and system based on improved random forest |
CN117350067A (en) * | 2023-10-19 | 2024-01-05 | 江西嘉逸陶瓷有限公司 | Standard sample generation method and device applied to ceramic nondestructive testing |
CN117574273A (en) * | 2023-11-03 | 2024-02-20 | 佛山职业技术学院 | Fan state and fault diagnosis method based on operation and maintenance big data and application |
CN117668623A (en) * | 2024-02-02 | 2024-03-08 | 中国海洋大学 | Multi-sensor cross-domain fault diagnosis method for leakage of ship pipeline valve |
CN117668623B (en) * | 2024-02-02 | 2024-05-14 | 中国海洋大学 | Multi-sensor cross-domain fault diagnosis method for leakage of ship pipeline valve |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115409052A (en) | Fault diagnosis method and system for wind generating set bearing under data imbalance | |
Chen et al. | Fault feature extraction and diagnosis of rolling bearings based on wavelet thresholding denoising with CEEMDAN energy entropy and PSO-LSSVM | |
Xu et al. | Multireceptive field denoising residual convolutional networks for fault diagnosis | |
Zhang et al. | Fault diagnosis and prognosis using wavelet packet decomposition, Fourier transform and artificial neural network | |
CN105678343B (en) | Hydropower Unit noise abnormality diagnostic method based on adaptive weighted group of sparse expression | |
Pu et al. | A one-class generative adversarial detection framework for multifunctional fault diagnoses | |
Zhao et al. | Probabilistic principal component analysis assisted new optimal scale morphological top-hat filter for the fault diagnosis of rolling bearing | |
Wang et al. | A factor analysis based transfer learning method for gearbox diagnosis under various operating conditions | |
Liu et al. | Multi-Scale Fusion Attention Convolutional Neural Network for Fault Diagnosis of Aero-Engine Rolling Bearing | |
CN116956215A (en) | Fault diagnosis method and system for transmission system | |
CN115219197A (en) | Bearing fault intelligent diagnosis method, system, medium, equipment and terminal | |
Wong et al. | Fault diagnosis of induction motors under untrained loads with a feature adaptation and improved broad learning framework | |
CN113705405B (en) | Nuclear pipeline fault diagnosis method | |
Shi et al. | A novel rotating machinery fault diagnosis method based on adaptive deep belief network structure and dynamic learning rate under variable working conditions | |
CN114298110A (en) | Rolling bearing fault diagnosis method and system based on interpretable 1DCNN model | |
CN117828531A (en) | Bearing fault diagnosis method based on multi-sensor multi-scale feature fusion | |
An et al. | A Parameter‐Optimized Variational Mode Decomposition Investigation for Fault Feature Extraction of Rolling Element Bearings | |
Wang et al. | Explainable machine learning for motor fault diagnosis | |
CN115356599B (en) | Multi-mode urban power grid fault diagnosis method and system | |
Shen et al. | Fault detection method based on multi-scale convolutional neural network for wind turbine gearbox | |
CN114659785A (en) | Fault detection method and device for transmission chain of wind driven generator | |
Jiang et al. | Rolling bearing weak fault diagnosis utilizing successive variational mode decomposition with sparsity index reconstructing strategy | |
Cao et al. | Bearing remaining useful life prediction based on optimized support vector regression model with denoising technique | |
Hoffmann et al. | Embedding anomaly detection autoencoders for wind turbines | |
Tian et al. | A cross-domain intelligent fault diagnosis method based on feature transfer with improved Inception ResNet for rolling bearings under varying working condition |
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 |