CN116106012A - Rolling bearing domain adaptive fault diagnosis method based on attention mechanism - Google Patents

Rolling bearing domain adaptive fault diagnosis method based on attention mechanism Download PDF

Info

Publication number
CN116106012A
CN116106012A CN202111330197.1A CN202111330197A CN116106012A CN 116106012 A CN116106012 A CN 116106012A CN 202111330197 A CN202111330197 A CN 202111330197A CN 116106012 A CN116106012 A CN 116106012A
Authority
CN
China
Prior art keywords
domain
attention
fault
rolling bearing
data
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
CN202111330197.1A
Other languages
Chinese (zh)
Inventor
杜劲松
王煜
高洁
王伟
杨旭
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.)
Shenyang Institute of Automation of CAS
Original Assignee
Shenyang Institute of Automation of CAS
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 Shenyang Institute of Automation of CAS filed Critical Shenyang Institute of Automation of CAS
Priority to CN202111330197.1A priority Critical patent/CN116106012A/en
Publication of CN116106012A publication Critical patent/CN116106012A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00Testing of machine parts
    • G01M13/04Bearings
    • G01M13/045Acoustic or vibration analysis

Landscapes

  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • General Physics & Mathematics (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention relates to a fault diagnosis method for rolling bearing domain adaptation based on an attention mechanism, which is characterized in that a feature extractor which is formed by one-dimensional separable convolution and embedded with a channel attention mechanism and a length attention mechanism extracts deep fault features from collected rolling bearing vibration monitoring signals; the local attention domain adaptation module and the global attention domain adaptation module are constructed to screen signals and signal fragments with good mobility, so that the generalization capability of the model is improved, and the model can better cope with the problem of fault diagnosis under variable working conditions; compared with an intelligent fault diagnosis algorithm for transfer learning, the algorithm considers the different signals and the different migratability of the signal fragments, and improves the interpretability of the model; experiments are carried out by applying various bearing vibration data, so that the algorithm is verified to have good performance stability, and excellent diagnosis results can be still maintained under various working condition change conditions.

Description

Rolling bearing domain adaptive fault diagnosis method based on attention mechanism
Technical Field
The invention belongs to the field of fault diagnosis of a variable working condition bearing, and particularly relates to a fault diagnosis method for a rolling bearing domain adaptation based on an attention mechanism.
Background
The rolling bearing is used as a key part of a rotary machine, the running environment is severe, the rolling bearing is extremely easy to damage in the service period, so that unexpected industrial accidents which are difficult to predict and control are caused, the running of local equipment is influenced by light weight, the production efficiency is reduced, serious production safety accidents are caused by heavy weight, and irrecoverable consequences are caused.
With the rapid rise of technologies such as industrial internet of things and the like, mechanical fault diagnosis gradually steps into the 'big data' era, and intelligent fault diagnosis algorithms based on methods such as deep learning and the like gradually become research hotspots in recent years. However, existing fault diagnosis algorithms based on deep learning are generally based on the assumption that the training set test set is in independent and same distribution, which is not consistent with the condition that a large amount of noise environment and complex variable working conditions exist in actual work. Therefore, the migration learning algorithm capable of coping with this becomes a research hotspot of the intelligent fault diagnosis algorithm.
However, the existing intelligent fault diagnosis algorithm for transfer learning lacks of interpretability, and the fact that different signals and signal fragments may have different migratability is not considered; therefore, it is important to provide an intelligent fault diagnosis technology with stronger generalization capability and a certain interpretability.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides the rolling bearing domain adaptive fault diagnosis method based on the attention mechanism, which improves the generalization capability of the model, can cope with various fault diagnosis problems and improves the interpretability of the model.
The technical scheme adopted by the invention for achieving the purpose is as follows:
an attention mechanism-based rolling bearing domain adaptation fault diagnosis method comprises the following steps:
vibration monitoring data of various bearing health conditions under different working conditions are respectively collected aiming at the same bearing system, and labels are marked for the data according to fault types to obtain bearing data sets aiming at different working conditions;
dividing data in the bearing data set into a source domain data set and a target domain data set, and respectively dividing the data into a training set and a verification set;
constructing a rolling bearing domain adaptive fault diagnosis model, and using a training set of a source domain data set and a training set training model of a target domain data set;
optimizing a rolling bearing domain adaptive fault diagnosis model;
and inputting the test set in the target domain data into the optimized rolling bearing domain adaptive fault diagnosis model to obtain the fault type of the target domain data.
The source domain data are vibration monitoring data with the number larger than a threshold value and complete labels, and the target domain data are vibration monitoring data under the working condition to be detected.
The construction of the rolling bearing domain adaptive fault diagnosis model comprises the following steps:
the feature extractor is used for extracting fault features in the bearing data set and dividing the fault features into a plurality of segments;
the local attention domain adaptation module is used for calculating the local attention value of each segment fault characteristic, carrying out weighting treatment on the local attention value, and carrying out dimension combination on the local attention value weighted by each segment;
the global attention field adaptation module is used for calculating a global attention value according to the local attention values after the dimensions are combined, weighting the global attention value into a classification loss and completing the fault type classification of the input data.
The feature extractor includes sequentially connected spatial convolution, length attention module, channel convolution, channel attention module, and pooling layer.
The local attention domain adaptation module comprises: the device comprises a maximum mean value difference module, a plurality of domain classifiers and a residual error connection module, wherein the fault characteristics of a plurality of fragments output by a characteristic extractor are processed through the maximum mean value difference module, the fault characteristics of each fragment are input into one domain classifier, the probability that the corresponding fragment belongs to a source domain is calculated, the local attention value of the fault characteristics of each fragment is calculated by using an entropy function, the fault characteristics are weighted, the fault characteristics after all the fragments are weighted are combined in dimensions, and the fault characteristics are output through the residual error connection module.
The global attention domain adaptation module comprises: and a domain classifier, which inputs the weighted fault characteristics into the domain classifier, calculates the probability that the fault characteristics belong to the source domain, and calculates the global attention value of the fault characteristics by using the entropy function so as to weight the fault characteristics.
In the construction of the rolling bearing domain adaptive fault diagnosis model, four types of loss functions are set, specifically:
1) Bearing failure classification loss for source domain dataset:
Figure BDA0003348502370000031
wherein n is S For the number of source domain samples, D S For the source domain sample space,
Figure BDA00033485023700000312
g is a cross entropy loss function y Is a classifier, h i Weighting features for local attention, y i Is a sample label;
2) MMD penalty for reducing classification differences between source and target domains:
Figure BDA0003348502370000032
wherein n is t For the number of target domain samples,
Figure BDA0003348502370000033
and->
Figure BDA0003348502370000034
Signal samples from source and target domains; phi (·) is a nonlinear mapping for mapping source and target domain samples to the same Hilbert space, k (·, ·) is a kernel function;
3) Local attention domain adaptation loss for extracting local and global domain invariance features from signal segments and signal samples, respectively
Figure BDA0003348502370000035
Global attention domain adaptation loss->
Figure BDA0003348502370000036
Figure BDA0003348502370000037
Figure BDA0003348502370000038
Wherein D is S Representing a source domain sample space; d (D) T Representing the target domain sample space, d i Representing signal samples x i Is a domain label of (2);
Figure BDA0003348502370000039
is the cross entropy penalty for domain classification;
4) Global weighted entropy penalty for auxiliary classification:
Figure BDA00033485023700000310
wherein c represents the number of types of faults; p is p i,j Representing signal samples x i Is divided into probability values for tag j.
The total loss function is:
Figure BDA00033485023700000311
wherein θ f ,θ y ,θ d And
Figure BDA0003348502370000041
models of feature extractor, label classifier, global attention domain adaptation module and local attention domain adaptation module, respectivelyThe parameters, η, γ and λ are the correction weights, respectively.
The rolling bearing domain adaptive fault diagnosis model is optimized using back propagation and Adam optimization algorithms, even though the total loss function is minimized.
The model parameters are updated using the following formula:
Figure 3
Figure BDA0003348502370000043
Figure BDA0003348502370000044
Figure 2
wherein epsilon is the learning rate.
The invention has the following beneficial effects and advantages:
1. the invention can autonomously select the signal and the signal segment with better transferability, enhance the domain self-adaptive capacity of the model and improve the accuracy of fault diagnosis.
2. Compared with the current various popular intelligent fault diagnosis algorithms for transfer learning, the intelligent fault diagnosis method has the advantages of optimal performance and robustness, and can be used for solving the fault diagnosis problems of various variable working conditions.
3. The invention has a certain interpretability in the choice of the migratable features.
Drawings
Fig. 1 is a diagram showing a structural model of a fault diagnosis method for rolling bearing domain adaptation based on an attention mechanism according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
An attention mechanism-based rolling bearing domain adaptation fault diagnosis method comprises the following steps:
step 1: vibration monitoring signals of various bearing health conditions under different working conditions are respectively collected aiming at the same bearing system, and data are labeled according to fault types, so that bearing data sets aiming at different working conditions are obtained;
step 2: selecting two data sets under different working conditions obtained in the step 1 as a source domain and a target domain respectively, and dividing corresponding training sets and testing sets;
step 3: training set data of a source domain and a target domain are used as model input to train the models, a grid search method is used for optimizing super parameters, a training model with highest diagnosis precision is selected as a final diagnosis model, and the model is fixed;
step 4: and (3) performing performance test on the final model obtained in the step (3) on the test set data of the target domain.
The one-dimensional separable convolution embedded with the channel attention mechanism and the length attention mechanism is: the present invention applies one-dimensional separable convolutions to replace traditional convolutions and embeds two attention mechanisms therein. The model mainly comprises the following parts: filtering the signal samples in each channel and having a spatial convolution of a particular convolution kernel size; the length attention mechanism module is used for acquiring the importance degree of each signal segment in fault diagnosis by aggregating the characteristic information of each channel; a group of convolution kernels are used for learning the relation among channels and simultaneously improving the channel convolution of the characteristic dimension; and a channel attention module for judging the importance of the features and promoting the important features of the network attention by capturing the interdependence relationship among the channels.
The particular convolution kernel size of the spatial convolution is equivalent to the convolution kernel size of the conventional convolution being replaced.
The migration learning module consisting of the local attention domain adaptation module and the global attention domain adaptation module is as follows:
the basic structure of the local attention domain adaptation module and the global attention domain adaptation module applied by the invention refers to the domain division in the DANN modelA class device; the features output from the feature extractor are divided into K segments and input to a local attention domain adaptation module comprising K domain classifiers, one for each signal segment
Figure BDA0003348502370000051
Calculating the probability that the corresponding fragment belongs to the source domain, judging the migration capability of the fragment, and calculating the local attention value by using an entropy function to realize weighting of the characteristics; meanwhile, in order to ensure that the characteristic difference between the source domain and the target domain is overlarge to cause negative migration, the invention uses the maximum mean value difference (MMD) to reduce the distribution difference of the source domain and the target domain before the local attention domain adaptation module is input; the local attention weighting features are input to a global attention domain adaptation module, which is composed of a domain classifier G, to measure the mobility of those signals more strongly D The output of which represents the probability that the corresponding input signal belongs to the source domain, is similarly used to calculate the attention value by means of an entropy function and weighted into a classification penalty in order to assist classification.
The block number K of the input feature in the local attention domain adaptation module is determined according to the output size of the last layer of the feature extractor, and the block number K is the product of the height and the width of the feature map, and the height is 1 and the block number K is equal to the width of the input feature map because the invention aims at the one-dimensional vibration signal.
The global parameters in the training process of the model are set as follows: the learning rate was 0.001, the batch size was 32, and the training times was 500.
Fig. 1 is a diagram showing a model structure of a fault diagnosis method for rolling bearing domain adaptation based on an attention mechanism.
The invention provides a rolling bearing domain adaptive fault diagnosis method based on an attention mechanism, which comprises two parts: a feature extractor and domain adaptation module; the feature extractor is composed of one-dimensional separable convolution stacks embedded with a channel attention mechanism and a length attention mechanism and is used for extracting fault related features; the domain adaptation module is composed of a local attention domain adaptation module and a global attention domain adaptation module, and signals and signal fragments with good mobility are screened, so that the generalization capability of the model is improved, and the model can better cope with the problem of fault diagnosis of the working conditions.
The method comprises the following specific steps:
step 1: and (5) acquiring and preprocessing data.
Aiming at the same bearing system, vibration monitoring signals of various bearing health conditions under different working conditions are respectively collected, and labeling is carried out on data according to fault types, so that bearing data sets aiming at different working conditions are obtained, two of the obtained data sets under different working conditions are respectively selected as a source domain and a target domain, and corresponding training sets and test sets are divided;
step 2: and extracting fault characteristics by the optimized characteristic extractor.
The one-dimensional separable convolution is applied to replace the traditional convolution, so that the model parameter number can be reduced under the condition of not influencing the diagnosis precision. And meanwhile, the embedded length attention mechanism and the channel attention mechanism are distributed after two steps of one-dimensional separable convolution. The reason is that the spatial convolution in one-dimensional separable convolution applies mutually independent convolution kernels to filter different channels respectively, so that each channel has unique characteristics. The length attention mechanism mainly gathers the characteristic information of different channels, and can play a richer role after being placed in a space convolution layer. At the same time, the spatial convolution does not change the feature dimension. In the present invention, the channel convolution in one-dimensional separable convolution is used not only to learn the relationship between channels by aggregating the outputs of the spatial convolution, but also to increase the dimension of the feature space. The present invention places the channel attention mechanism after the channel convolution to ensure that this attention module is applied to a higher dimensional feature space. Input signal sample x i Through a feature extractor G f The fault-related characteristics can be obtained well
Step 3: extracting domain invariance characteristics through the optimized domain adaptation module.
Signal sample x i Through a feature extractor G f The latter features are dividedFor K fragments, each fragment is characterized by
Figure 4
Input it into the corresponding local classifier +.>
Figure BDA0003348502370000072
After that, the probability of the fragment belonging to the source domain can be obtained>
Figure 1
The domain classifier and the feature extractor form a countermeasure relation, and the domain classifier cannot distinguish whether the features generated by the feature extractor belong to a source domain or a target domain through training, so that the extraction of the domain invariance features is realized. When probability value->
Figure BDA0003348502370000074
Approaching 0 or 1, this means that the signal segment can be classified by a domain classifier, which has poor mobility. The probability value cannot be used directly to weight the feature, so the entropy function is used +.>
Figure BDA0003348502370000075
Calculating a local attention value of each segment, wherein X is a variable value in a discrete random variable X, and p is a probability distribution function of the discrete random variable X, and the final calculation formula of the local attention value of each segment is as follows:
Figure BDA0003348502370000076
meanwhile, the local classifier is added with a residual error connection module to prevent negative migration caused by an incorrect attention value, so that the characteristics of each segment after being weighted by local attention are as follows:
Figure 5
then the corresponding segments weighted by local attention are dimension combined according to the position of each segment before the segmentation, and the obtainedFeature h i Is consistent with the dimensions of the features before segmentation. Weighting the local attention by a feature h i Input global domain classifier G D After that, the probability that the signal belongs to the source domain can be obtained
Figure BDA0003348502370000081
The attention value is also calculated using the entropy function, resulting in a global attention value of:
Figure BDA0003348502370000082
step 4: and (5) loss function setting.
The loss function of the present invention includes the following four classes in total:
1) Bearing failure classification loss for source domain dataset:
Figure BDA0003348502370000083
wherein n is S For the number of source domain samples, D S For the source domain sample space,
Figure BDA0003348502370000084
g is a cross entropy loss function y Is a classifier, h i Weighting features for local attention, y i Is a sample label;
2) MMD penalty for reducing classification differences between source and target domains:
Figure BDA0003348502370000085
wherein n is t For the number of target domain samples,
Figure BDA0003348502370000086
and->
Figure BDA0003348502370000087
From a source domainSignal samples in the target domain; phi (·) is a nonlinear mapping for mapping source and target domain samples to the same Hilbert space, k (·, ·) is a kernel function for computing the inner product of the two mappings, in the above formula +.>
Figure BDA0003348502370000088
For the inner product of the mapped source and target domain samples, i.e
Figure BDA0003348502370000089
3) Local attention domain adaptation loss for extracting local and global domain invariance features from signal segments and signal samples, respectively
Figure BDA00033485023700000810
Global attention domain adaptation loss->
Figure BDA00033485023700000811
The signal segments are samples which are uniformly divided according to a certain number, the dividing number is consistent with the dividing block number when the feature extractor divides the fault feature,
Figure BDA00033485023700000812
Figure BDA00033485023700000813
wherein D is S Representing a source domain sample space; d (D) T Representing a target domain sample space; d, d i Representing a vibration signal sample x i Is a domain label of (2);
Figure BDA0003348502370000091
is the cross entropy penalty for domain classification;
4) Global weighted entropy penalty for auxiliary classification:
Figure BDA0003348502370000092
wherein c represents the number of types of faults; p is p i,j Representing signal samples x i Is divided into probability values for tag j.
Thus, the total loss function of the present invention is:
Figure BDA0003348502370000093
wherein θ f ,θ y ,θ d And
Figure BDA0003348502370000094
model parameters of the feature extractor, the tag classifier, the global domain classifier and the local domain classifier are respectively, and eta, gamma and lambda are respectively correction weights.
Step 5: and optimizing strategy setting.
The network uses standard back propagation and Adam optimization algorithms to minimize the total loss function when the training process is completed. The parameters of each network structure are updated according to the following formula:
Figure BDA0003348502370000095
Figure BDA0003348502370000096
Figure BDA0003348502370000097
Figure BDA0003348502370000098
wherein epsilon is the learning rate.
Step 6: and performing fault diagnosis on the target domain test set.
And searching an optimal super-parameter combination by using a grid search method, fixing a training model under the optimal super-parameter combination, and performing fault diagnosis test model performance in a target domain test set.

Claims (10)

1. An attention mechanism-based rolling bearing domain adaptation fault diagnosis method is characterized by comprising the following steps of:
vibration monitoring data of various bearing health conditions under different working conditions are respectively collected aiming at the same bearing system, and labels are marked for the data according to fault types to obtain bearing data sets aiming at different working conditions;
dividing data in the bearing data set into a source domain data set and a target domain data set, and respectively dividing the data into a training set and a verification set;
constructing a rolling bearing domain adaptive fault diagnosis model, and using a training set of a source domain data set and a training set training model of a target domain data set;
optimizing a rolling bearing domain adaptive fault diagnosis model;
and inputting the test set in the target domain data into the optimized rolling bearing domain adaptive fault diagnosis model to obtain the fault type of the target domain data.
2. The method for diagnosing a rolling bearing domain adaptation fault based on an attention mechanism according to claim 1, wherein the source domain data is vibration monitoring data with the number being greater than a threshold value and complete labels, and the target domain data is vibration monitoring data under a working condition to be detected.
3. The method for diagnosing a rolling bearing domain adaptation failure based on an attention mechanism according to claim 1, wherein said constructing a rolling bearing domain adaptation failure diagnosis model includes:
the feature extractor is used for extracting fault features in the bearing data set and dividing the fault features into a plurality of segments;
the local attention domain adaptation module is used for calculating the local attention value of each segment fault characteristic, carrying out weighting treatment on the local attention value, and carrying out dimension combination on the local attention value weighted by each segment;
the global attention field adaptation module is used for calculating a global attention value according to the local attention values after the dimensions are combined, weighting the global attention value into a classification loss and completing the fault type classification of the input data.
4. A method of fault diagnosis for rolling bearing domain adaptation based on attention mechanism according to claim 3, wherein the feature extractor comprises sequentially connected spatial convolution, length attention module, channel convolution, channel attention module and pooling layer.
5. A method of diagnosing a rolling bearing domain adaptation fault based on an attention mechanism as claimed in claim 3, wherein the local attention domain adaptation module comprises: the device comprises a maximum mean value difference module, a plurality of domain classifiers and a residual error connection module, wherein the fault characteristics of a plurality of fragments output by a characteristic extractor are processed through the maximum mean value difference module, the fault characteristics of each fragment are input into one domain classifier, the probability that the corresponding fragment belongs to a source domain is calculated, the local attention value of the fault characteristics of each fragment is calculated by using an entropy function, the fault characteristics are weighted, the fault characteristics after all the fragments are weighted are combined in dimensions, and the fault characteristics are output through the residual error connection module.
6. A method of fault diagnosis of a rolling bearing domain adaptation based on an attention mechanism according to claim 3, wherein the global attention domain adaptation module comprises: and a domain classifier, which inputs the weighted fault characteristics into the domain classifier, calculates the probability that the fault characteristics belong to the source domain, and calculates the global attention value of the fault characteristics by using the entropy function so as to weight the fault characteristics.
7. A method for diagnosing a rolling bearing domain adaptation failure based on an attention mechanism according to claim 3, wherein in said constructing a rolling bearing domain adaptation failure diagnosis model, four types of loss functions are set, specifically:
1) Bearing failure classification loss for source domain dataset:
Figure FDA0003348502360000021
wherein n is S For the number of source domain samples, D S For the source domain sample space,
Figure FDA0003348502360000022
g is a cross entropy loss function y Is a classifier, h i Weighting features for local attention, y i Is a sample label;
2) MMD penalty for reducing classification differences between source and target domains:
Figure FDA0003348502360000023
wherein n is t For the number of target domain samples,
Figure FDA0003348502360000024
and->
Figure FDA0003348502360000025
Signal samples from source and target domains; phi (·) is a nonlinear mapping for mapping source and target domain samples to the same Hilbert space, k (·, ·) is a kernel function;
3) Local attention domain adaptation loss for extracting local and global domain invariance features from signal segments and signal samples, respectively
Figure FDA0003348502360000031
Global attention domain adaptation loss->
Figure FDA0003348502360000032
Figure FDA0003348502360000033
Figure FDA0003348502360000034
Wherein D is S Representing a source domain sample space; d (D) T Representing the target domain sample space, d i Representing signal samples x i Is a domain label of (2);
Figure FDA0003348502360000035
is the cross entropy penalty for domain classification;
4) Global weighted entropy penalty for auxiliary classification:
Figure FDA0003348502360000036
wherein c represents the number of types of faults; p is p i,j Representing signal samples x i Is divided into probability values for tag j.
8. The method for diagnosing a rolling bearing domain adaptation failure based on an attention mechanism as recited in claim 7, wherein the total loss function is:
Figure FDA0003348502360000037
wherein θ f ,θ y ,θ d And
Figure FDA0003348502360000038
model parameters of the feature extractor, the tag classifier, the global attention domain adaptation module and the local attention domain adaptation module are respectively, and eta, gamma and lambda are respectively correction weights.
9. The method for diagnosing a rolling bearing domain adaptation fault based on an attention mechanism according to claim 1 or 8, wherein the rolling bearing domain adaptation fault diagnosis model is optimized by using a back propagation and Adam optimization algorithm, even if the total loss function is minimized.
10. The method for diagnosing a rolling bearing domain adaptation fault based on an attention mechanism according to claim 8, wherein the model parameters are updated using the following formula:
Figure FDA0003348502360000039
Figure FDA00033485023600000310
Figure FDA00033485023600000311
Figure FDA0003348502360000041
wherein epsilon is the learning rate.
CN202111330197.1A 2021-11-11 2021-11-11 Rolling bearing domain adaptive fault diagnosis method based on attention mechanism Pending CN116106012A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111330197.1A CN116106012A (en) 2021-11-11 2021-11-11 Rolling bearing domain adaptive fault diagnosis method based on attention mechanism

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111330197.1A CN116106012A (en) 2021-11-11 2021-11-11 Rolling bearing domain adaptive fault diagnosis method based on attention mechanism

Publications (1)

Publication Number Publication Date
CN116106012A true CN116106012A (en) 2023-05-12

Family

ID=86266088

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111330197.1A Pending CN116106012A (en) 2021-11-11 2021-11-11 Rolling bearing domain adaptive fault diagnosis method based on attention mechanism

Country Status (1)

Country Link
CN (1) CN116106012A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117574259A (en) * 2023-10-12 2024-02-20 南京工业大学 Attention twin intelligent migration interpretability diagnosis method suitable for high-end equipment
CN118035766A (en) * 2024-04-12 2024-05-14 太原理工大学 Variable working condition bearing fault diagnosis method based on similarity countermeasure and contrast learning

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117574259A (en) * 2023-10-12 2024-02-20 南京工业大学 Attention twin intelligent migration interpretability diagnosis method suitable for high-end equipment
CN117574259B (en) * 2023-10-12 2024-05-07 南京工业大学 Attention twin intelligent migration interpretability diagnosis method suitable for high-end equipment
CN118035766A (en) * 2024-04-12 2024-05-14 太原理工大学 Variable working condition bearing fault diagnosis method based on similarity countermeasure and contrast learning

Similar Documents

Publication Publication Date Title
CN107133943A (en) A kind of visible detection method of stockbridge damper defects detection
CN116106012A (en) Rolling bearing domain adaptive fault diagnosis method based on attention mechanism
CN111882620B (en) Road drivable area segmentation method based on multi-scale information
CN114048568A (en) Rotating machine fault diagnosis method based on multi-source migration fusion contraction framework
CN113375941A (en) Open set fault diagnosis method for high-speed motor train unit bearing
CN113239720B (en) Subway vehicle running gear fault diagnosis method based on deep migration learning
CN111353373A (en) Correlation alignment domain adaptive fault diagnosis method
CN112784872B (en) Cross-working condition fault diagnosis method based on open set joint transfer learning
CN117315380B (en) Deep learning-based pneumonia CT image classification method and system
CN113435321A (en) Method, system and equipment for evaluating state of main shaft bearing and readable storage medium
CN116977725A (en) Abnormal behavior identification method and device based on improved convolutional neural network
CN118447297A (en) Knowledge distillation-based light crop disease identification method
CN107609216B (en) Mechanical fault diagnosis method based on probability box model correction
CN116912576B (en) Self-adaptive graph convolution brain disease classification method based on brain network higher-order structure
CN117111464A (en) Self-adaptive fault diagnosis method under multiple working conditions
CN117171713A (en) Cross self-adaptive deep migration learning method and system based on bearing service life
CN116842358A (en) Soft measurement modeling method based on multi-scale convolution and self-adaptive feature fusion
CN114941890B (en) Central air conditioner fault diagnosis method and system based on image and depth blur
CN116166997A (en) Intelligent main shaft service state diagnosis method, system, equipment and medium
CN113344005B (en) Image edge detection method based on optimized small-scale features
CN109993218A (en) A kind of identification of maize diseases based on convolutional neural networks and disaggregated model
CN115753102A (en) Bearing fault diagnosis method based on multi-scale residual error sub-domain adaptation
CN118522083B (en) Fault diagnosis method and device for automobile transmission system
CN117197095B (en) Surface defect detection method and system for generating countermeasure segmentation model based on semi-supervision
CN116698410B (en) Rolling bearing multi-sensor data monitoring method based on convolutional neural network

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