CN114444218A - Small sample rolling bearing fault diagnosis method under multiple working conditions - Google Patents

Small sample rolling bearing fault diagnosis method under multiple working conditions Download PDF

Info

Publication number
CN114444218A
CN114444218A CN202111212782.1A CN202111212782A CN114444218A CN 114444218 A CN114444218 A CN 114444218A CN 202111212782 A CN202111212782 A CN 202111212782A CN 114444218 A CN114444218 A CN 114444218A
Authority
CN
China
Prior art keywords
maml
sample
rolling bearing
model
fault
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111212782.1A
Other languages
Chinese (zh)
Inventor
向玲
苏浩
胡爱军
杨鑫
陈凯乐
陈锦鹏
姚青陶
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN202111212782.1A priority Critical patent/CN114444218A/en
Publication of CN114444218A publication Critical patent/CN114444218A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Geometry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Biology (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

A fault diagnosis method for a small-sample rolling bearing under multiple working conditions comprises the following steps: a. constructing a task set; b. inputting data in the task set into the ISDAE for reconstruction to obtain a reconstructed signal matrix which retains effective characteristics in an original signal and reduces noise; c. classifying the reconstructed signals by using the MAML, training model parameters of the MAML and obtaining an optimal network model; d. and inputting the original vibration signal of the rolling bearing to be monitored into the trained MAML model, and judging whether the rolling bearing has faults or not and judging the type of the faults. The method for diagnosing the bearing fault by combining the model independence with the improved sparse noise reduction self-coding not only can extract the separability characteristic in the original vibration signal and improve the anti-noise capability of the signal, but also can improve the generalization capability of the model, thereby accurately diagnosing the bearing fault of small sample data under multiple working conditions and ensuring the safe operation of mechanical equipment.

Description

Small sample rolling bearing fault diagnosis method under multiple working conditions
Technical Field
The invention relates to a method for diagnosing faults of a small-sample rolling bearing under a multi-working condition by combining model-independent meta-learning (MAML) with improved sparse noise reduction self-encoding (ISDAE), and belongs to the technical field of diagnosis.
Background
Rolling bearings are one of the key components of rotary machines and are widely used in modern large-scale mechanical equipment. Statistically, in a rotating machine using a rolling bearing, about 30% of mechanical failures are caused by the bearing. In recent years, intelligent fault diagnosis methods for rolling bearings based on deep learning are diversified, and powerful tools are provided for monitoring safe operation of mechanical equipment. However, with the continuous improvement of the processing precision and the material performance of the bearing, the service life of the bearing is continuously prolonged, and the bearing is replaced in time when the bearing is detected to be damaged, so that sufficient bearing fault data cannot be obtained. Because a large amount of data is needed to be used as a research object for training the deep learning model, the training of the model is trapped by the difficulty due to the fact that sufficient bearing fault data cannot be obtained, the generalization capability of the deep learning model is more limited, and the recognition accuracy is difficult to improve. In addition, the rolling bearing is time-varying in operating environment, fault conditions are complicated, and the extraction of dominant separable depth features is very difficult due to different sampling frequencies, different fault degrees, different fault positions, different fault types, different fault sizes and the like, so that the difficulty in intelligent diagnosis of bearing faults is increased. Therefore, how to effectively utilize the small sample data to diagnose the fault of the rolling bearing under multiple working conditions is of great importance.
Disclosure of Invention
The invention aims to provide a method for diagnosing the fault of a small-sample rolling bearing under multiple working conditions aiming at the defects of the prior art so as to accurately diagnose the fault of the rolling bearing and ensure the safe operation of mechanical equipment.
The problems of the invention are solved by the following technical scheme:
a fault diagnosis method for a small-sample rolling bearing under multiple working conditions comprises the following steps:
a. classifying original vibration signals of a rolling bearing with faults, which are acquired by a data acquisition system, according to different working conditions, different sampling frequencies, different fault degrees and different fault types to construct a task set T, and obtaining a data sample matrix composed of the original vibration signals
Figure BDA0003309437120000021
Wherein x ismDenotes the mth sample, M denotes the number of samples,
Figure BDA0003309437120000022
i-th data representing an m-th sample, n representing a length of the sample;
b. inputting the data in the task set into the ISDAE for reconstruction to obtain a reconstructed signal matrix which retains the effective characteristics in the original signal and reduces the noise
Figure BDA0003309437120000023
c. Classifying the reconstructed signals by using the MAML, training model parameters of the MAML and obtaining an optimal network model;
d. and inputting the original vibration signal of the monitored rolling bearing acquired by the data acquisition system into the trained MAML model, and judging whether the rolling bearing has faults or not and judging the type of the faults.
The method for diagnosing the fault of the small-sample rolling bearing under the multiple working conditions comprises the following specific processes of reconstructing data in a task set:
random noise is added to a data sample matrix X formed by original vibration signals to form a signal matrix
Figure RE-GDA0003580874930000024
Adding a probability distribution metric MMD (X, Y) and a sparse penalty term to a loss function
Figure RE-GDA0003580874930000025
Beta represents a penalty coefficient, p represents a sparse parameter,
Figure RE-GDA0003580874930000026
represents the activation value of the jth hidden unit, which is equal to the sparse parameter, s represents the number of hidden units, and KL (.) is used to measure ρ and
Figure RE-GDA0003580874930000027
relative entropy between, the loss function of ISDAE is obtained:
Figure BDA0003309437120000028
in the formula:
Figure BDA0003309437120000029
by means of iteration, minimizing the loss function JISDAEThereby obtaining a reconstructed signal matrix
Figure BDA00033094371200000210
The method for diagnosing the fault of the small-sample rolling bearing under the multiple working conditions utilizes the MAML to classify the reconstructed signals, and the specific process of training the model parameters of the MAML is as follows:
a. randomly sampling N types of samples from the reconstructed task Set T, randomly sampling K +1 samples from each type of sample, then randomly extracting K samples from the K +1 samples of each type of sample to form a training Set (Support Set), and forming a test Set (Query Set) by using the rest samples of each type of sample;
b. building an MAML model;
c. and (3) meta learning process: randomly selecting a sample from each class in a training set to form a group of training data, inputting the training data into the MAML model for training, and updating parameters of the MAML model;
d. and c, repeating the step c, updating the parameters of the MAML for multiple times to obtain an optimal network model, then randomly extracting a sample from the test set, and judging the type of the rolling bearing fault by using the trained MAML to finish the test of the MAML model.
According to the fault diagnosis method for the small-sample rolling bearing under the multiple working conditions, the MAML model comprises four groups of modules, and each group of modules consists of a 3 x 3 convolution kernel, a batch regularization layer, a ReLU activation function layer, a 2 x 2 maximum pooling layer and 64 output units.
Advantageous effects
The method for diagnosing the bearing fault by combining the model independence and the improved sparse noise reduction self-coding is adopted, so that the separability characteristic in the original vibration signal can be extracted, the anti-noise capability of the signal is improved, the generalization capability of the model can be improved, the bearing fault of small sample data under multiple working conditions can be accurately diagnosed, and the safe operation of mechanical equipment is ensured.
The invention has the following advantages:
a. the bearing fault signal is reconstructed by using the ISDAE, so that the sparse representation capability of the model is improved, the separability characteristic in the original vibration signal can be extracted, and the anti-noise capability of the signal is improved;
b. the method utilizes the MAML to classify and recognize the reconstructed signals, can well extract deep meta-features in small sample data, quickly train a deep learning model, improve the generalization capability of the model, and is favorable for diagnosing bearing faults;
c. by adopting the method and the device for analyzing the fault signals of the small sample rolling bearing under the multiple working conditions, higher classification precision of the bearing fault can be achieved, and deep meta-characteristic information hidden in the signals can be effectively extracted, so that the health condition of the bearing of the small sample data under the multiple working conditions can be comprehensively and accurately discriminated.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings.
FIG. 1 is an overall flow diagram of the present invention;
FIG. 2 is a time domain diagram of a multi-fault experimental signal of a rolling bearing under different working conditions, wherein FIG. 2(a) is a CWRU time domain diagram; FIG. 2(b) is an HD time domain diagram;
FIG. 3 is a graph of the results of an analysis of the United states university of Kaiser storage (CWRU) bearing fault signal in accordance with the present invention;
FIG. 4 is a diagram showing the result of analyzing the bearing fault signal of the present invention for the tin-free Thick Instrument Co.
The symbols herein are represented as: t represents a task set, X represents a matrix of raw data samples,
Figure RE-GDA0003580874930000031
representing the matrix of data samples after the addition of noise, M representing the number of samples, xmDenotes the mth sample, n denotes the length of the sample, Y denotes the reconstructed signal matrix, MMD (X, Y) denotes a probability distribution metric, β denotes a penalty factor, ρ denotes a sparse parameter,
Figure RE-GDA0003580874930000032
represents the activation value of the jth hidden unit, s represents the number of hidden units, and KL (is) is used for measuring rho and rho
Figure RE-GDA0003580874930000033
Relative entropy between, JISDAERepresenting loss function, fθRepresenting parameterized functions, alpha and beta being hyperparameters, TiRepresenting a new task.
Detailed Description
The invention provides a fault diagnosis method for a small-sample rolling bearing under multiple working conditions by combining MAML and ISDAE, which can effectively extract deep-layer meta-characteristics of a rolling bearing fault under multiple working conditions, thereby realizing accurate identification of the fault state of the rolling bearing.
As shown in fig. 1, the present invention comprises the steps of:
1) firstly, in order to meet the requirements of a meta-learning task, original vibration signals are classified according to different working conditions, different sampling frequencies, different fault degrees and different fault types to construct a task set.
2) Inputting the task set data constructed in the step 1) into an ISDAE (iterative reconstruction of independent discriminant analysis) for reconstruction, and giving a data sample matrix consisting of original vibration signals
Figure BDA0003309437120000041
Wherein xmDenotes the mth sample, M denotes the number of samples,
Figure BDA0003309437120000042
(i ═ 1,2,. n) denotes the ith data of the mth sample, and n denotes the length of the sample; reconstructing the original vibration signal by adding different loss terms to obtain a reconstructed signal matrix
Figure BDA0003309437120000043
The number of samples and the signal length are the same as those of the original vibration signal, so that the reconstructed signal can keep effective characteristics in the original signal and reduce the noise of the signal. The specific reconstruction process comprises the following steps:
the goal of ISDAE is to minimize the loss function J in an iterative mannerISDAESo as to obtain the parameter optimal solution h of the model h (.)W,bAnd realizing the reconstruction of the sample X. In order to improve the anti-noise interference capability of the reconstructed signal, random noise is added to original signal samples X to form a signal
Figure RE-GDA0003580874930000045
In order to improve the sparse representation capability of the model, a sparse penalty term is added into the loss function
Figure RE-GDA0003580874930000046
Where beta represents a penalty factor, p represents a sparsity parameter,
Figure RE-GDA0003580874930000047
represents the activation value of the jth hidden unit, which is equal to the sparse parameter, s represents the number of hidden units, and KL (.) is used to measure ρ and
Figure RE-GDA0003580874930000048
relative entropy therebetween; is composed ofMinimizing the edge distribution between the original vibration signal and the reconstructed signal, and adding a probability distribution metric MMD (X, Y) to the loss function, so that the reconstructed signal maximally maintains the probability distribution of the original vibration signal. I.e., the loss function of the ISDAE is
Figure BDA0003309437120000048
In the formula:
Figure BDA0003309437120000049
3) classifying the reconstructed signals after ISDAE processing in the step 2) by using the MAML, training model parameters, and specifically realizing the following processes:
firstly, randomly sampling N types of samples from a task set T, and randomly sampling K +1 samples from each type of sample; then, of the K +1 samples, K samples are randomly selected from each class of samples to form a training Set (Support Set), and the remaining one sample from each class forms a testing Set (Query Set).
Building an MAML (maximum likelihood model): the model has four groups of modules, each group of modules consists of a 3 × 3 convolution kernel, a batch regularization layer, a ReLU activation function layer, a 2 × 2 maximum pooling layer and 64 output units.
And (3) meta learning process: randomly selecting a sample from each type in the Support Set to form a group of training data, and inputting a model for training. MAML is essentially treated as a parameterized function fθThe goal is to obtain the optimum parameters such that fθCan be adapted to a new task Ti。
The first step of parameter updating is:
Figure BDA0003309437120000051
in the formula: alpha is a hyperparameter, thetai' is the parameter after the first update of the parameter theta,
Figure BDA0003309437120000052
the derivative of the cross entropy loss function is represented. Then continue on the new task TiAnd a new sample is sampled for testing.
The second step of parameter updating is:
Figure BDA0003309437120000053
in the formula: beta is a hyper-parameter,
Figure BDA0003309437120000054
multiple slave T in representing element training processiSample acquisition loss
Figure BDA0003309437120000055
The sum of (a) and (b).
4) And repeating the step 3), updating the parameters of the MAML through multiple iterations to obtain an optimal network model, randomly extracting a sample from the Query Set, judging the type of the rolling bearing fault by using the trained MAML, and completing the test of the MAML.
5) And inputting the original vibration signal of the monitored rolling bearing acquired by the data acquisition system into the trained MAML model, and judging whether the rolling bearing has faults or not and judging the type of the faults.
The fault classification result is determined according to the meta-feature learning capability of the MAML model.
The effectiveness of the invention is verified by the analysis of the fault signals of the two sets of rolling bearings.
The invention was verified using two sets of rolling bearing fault data, one from the open source data of bearing laboratories of the university of Kaiser Sichu, USA, and the other from the fault data collected at the bearing laboratory bench of Wuxi-Thick instruments, Inc. The detailed data are shown in tables 1 and 2, wherein the sampling frequency of the bearing data of Kaiser Sichu university laboratory is fs112kHz and fs248kHz, bench Bearings data sampling frequency f for Wuxi Thick instrumentss12.8kHz, the number of samples used per condition is 20, each sampleHas a data length of 1024:
TABLE 1 CWRU bearing data
Figure BDA0003309437120000056
Figure BDA0003309437120000061
TABLE 2 HD bearing data
Figure BDA0003309437120000062
The time domain waveform of the multi-fault experimental signal of the rolling bearing is shown in fig. 2, and it can be seen from fig. 2 that the effective fault characteristics are difficult to distinguish by a simple time domain signal, so that the fault existing in the bearing cannot be judged.
The invention is adopted to analyze two groups of bearing signals, and the specific implementation process is as follows:
firstly, the ISDAE is utilized to reconstruct the original vibration signal, the sparse representation capability and the anti-noise interference capability of the model are improved by constructing a loss function containing a sparse term, a mean square error term and an MMD term, redundant information and impurity information in the original vibration signal are removed, the probability distribution of the reconstructed signal consistent with the original vibration signal is kept, and the characteristic distinguishability of the reconstructed signal is enhanced.
Then, the obtained reconstructed signals are classified by using the MAML, and model parameters are trained.
And finally, updating parameters of the MAML through multiple iterations to obtain an optimal network model, randomly extracting a sample from the Query Set, judging the type of the rolling bearing fault by using the trained MAML, and respectively obtaining the test results as shown in FIGS. 3 and 4. In fig. 3, the test precision reaches more than 90%, and the training loss is approximately in a descending trend; in fig. 4, the testing precision reaches more than 80%, the testing precision is kept in a stable state, and the training loss is approximately in a downward trend. As can be seen from fig. 3 and fig. 4, the training loss (dotted line) approximately presents a downward trend, which indicates that the model training process is relatively stable, and the trained model can be used for judging the fault type; the test precision (real line) is changed from rapid rising to slow rising in the early stage, and finally shows a relatively mild growth trend, which indicates that the model is updated and iterated all the time in the training process, the parameters are gradually optimized, and finally, higher test precision is obtained.
Therefore, the method can effectively dig out deep layer element characteristic information, accurately identify the fault condition of the small sample rolling bearing under multiple working conditions and realize accurate diagnosis of the fault state of the bearing.

Claims (4)

1. A fault diagnosis method for a small sample rolling bearing under multiple working conditions is characterized by comprising the following steps:
a. classifying original vibration signals of a rolling bearing with faults, which are acquired by a data acquisition system, according to different working conditions, different sampling frequencies, different fault degrees and different fault types to construct a task set T, and obtaining a data sample matrix consisting of the original vibration signals
Figure FDA0003309437110000011
Wherein xmDenotes the mth sample, M denotes the number of samples,
Figure FDA0003309437110000012
i-th data representing an m-th sample, n representing a length of the sample;
b. inputting the data in the task set into the ISDAE for reconstruction to obtain a reconstructed signal matrix which retains the effective characteristics in the original signal and reduces the noise
Figure FDA0003309437110000013
c. Classifying the reconstructed signals by using the MAML, training model parameters of the MAML and obtaining an optimal network model;
d. and inputting the original vibration signal of the monitored rolling bearing acquired by the data acquisition system into the trained MAML model, and judging whether the rolling bearing has faults or not and judging the type of the faults.
2. The method for diagnosing the fault of the small-sample rolling bearing under the multiple working conditions according to claim 1, wherein the specific process of reconstructing the data in the task set comprises the following steps:
random noise is added to a data sample matrix X formed by original vibration signals to form a signal matrix
Figure RE-FDA0003580874920000014
Adding probability distribution metric MMD (X, Y) and sparse penalty term to loss function
Figure RE-FDA0003580874920000015
Beta represents a penalty factor, p represents a sparseness parameter,
Figure RE-FDA0003580874920000016
represents the activation value of the jth hidden unit, which is equal to the sparse parameter, s represents the number of hidden units, and KL (.) is used to measure ρ and
Figure RE-FDA0003580874920000017
relative entropy between, the loss function of ISDAE is obtained:
Figure RE-FDA0003580874920000018
in the formula:
Figure RE-FDA0003580874920000019
by means of iteration, minimizing the loss function JISDAEThereby obtaining a reconstructed signal matrix
Figure RE-FDA00035808749200000110
3. The method for diagnosing the fault of the small-sample rolling bearing under the multiple working conditions according to the claim 1 or 2, wherein the MAML is used for classifying the reconstructed signals, and the specific process of training the model parameters of the MAML is as follows:
a. randomly sampling N types of samples from the reconstructed task set T, randomly sampling K +1 samples from each type of sample, then randomly extracting K samples from the K +1 samples of each type of sample to form a training set, and forming a test set by using the rest samples of each type of sample;
b. building an MAML model;
c. and (3) meta learning process: randomly selecting a sample from each class in a training set to form a group of training data, inputting the training data into the MAML model for training, and updating parameters of the MAML model;
d. and c, repeating the step c, updating the parameters of the MAML through multiple iterations to obtain an optimal network model, then randomly extracting a sample from the test set, judging the type of the fault of the rolling bearing by using the trained MAML, and completing the test of the MAML model.
4. The method for diagnosing the fault of the rolling bearing with the small sample under the multiple working conditions according to claim 3, wherein the MAML model comprises four groups of modules, and each group of modules consists of a 3 x 3 convolution kernel, a batch regularization layer, a ReLU activation function layer, a 2 x 2 maximum pooling layer and 64 output units.
CN202111212782.1A 2021-10-19 2021-10-19 Small sample rolling bearing fault diagnosis method under multiple working conditions Pending CN114444218A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111212782.1A CN114444218A (en) 2021-10-19 2021-10-19 Small sample rolling bearing fault diagnosis method under multiple working conditions

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111212782.1A CN114444218A (en) 2021-10-19 2021-10-19 Small sample rolling bearing fault diagnosis method under multiple working conditions

Publications (1)

Publication Number Publication Date
CN114444218A true CN114444218A (en) 2022-05-06

Family

ID=81362443

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111212782.1A Pending CN114444218A (en) 2021-10-19 2021-10-19 Small sample rolling bearing fault diagnosis method under multiple working conditions

Country Status (1)

Country Link
CN (1) CN114444218A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114993677A (en) * 2022-05-11 2022-09-02 山东大学 Rolling bearing fault diagnosis method and system based on unbalanced small sample data

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114993677A (en) * 2022-05-11 2022-09-02 山东大学 Rolling bearing fault diagnosis method and system based on unbalanced small sample data

Similar Documents

Publication Publication Date Title
CN112036301B (en) Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion
CN111037365B (en) Cutter state monitoring data set enhancing method based on generative countermeasure network
CN113780056A (en) Rolling bearing fault diagnosis method based on vibration signal
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN111238843B (en) Fan health evaluation method based on rapid spectrum kurtosis analysis
CN113567130A (en) Bearing fault diagnosis method based on multiple working conditions of equipment
CN107451760B (en) Rolling bearing fault diagnosis method based on time window slip limited Boltzmann machine
CN110044623A (en) The rolling bearing fault intelligent identification Method of empirical mode decomposition residual signal feature
CN115424635B (en) Cement plant equipment fault diagnosis method based on sound characteristics
CN111665050B (en) Rolling bearing fault diagnosis method based on clustering K-SVD algorithm
CN112364706A (en) Small sample bearing fault diagnosis method based on class imbalance
CN114444218A (en) Small sample rolling bearing fault diagnosis method under multiple working conditions
CN115659143A (en) Fault real-time diagnosis method based on experimental design
CN113865872B (en) Bearing fault diagnosis method based on wavelet packet reconstruction imaging and CNN
CN103926097B (en) Method for collecting and extracting fault feature information of low-speed and heavy-load device
CN114487129A (en) Flexible material damage identification method based on acoustic emission technology
CN112945556B (en) Bearing fault diagnosis method based on wavelet packet decomposition and optimal rejection classification strategy
CN112215307B (en) Method for automatically detecting signal abnormality of earthquake instrument by machine learning
Jianhui et al. New fault diagnosis method for rolling bearing based on PCA
CN114692695B (en) Fault feature selection and classification method research based on multi-index fusion
CN111275003A (en) Microseismic signal identification method based on class-optimal Gaussian kernel multi-classification support vector machine
CN116340812A (en) Transformer partial discharge fault mode identification method and system
CN116361723A (en) Bearing fault diagnosis and classification method based on multi-scale characteristics and attention
CN112603337A (en) Electroencephalogram signal identification method

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