CN114004252A - Bearing fault diagnosis method, device and equipment - Google Patents
Bearing fault diagnosis method, device and equipment Download PDFInfo
- Publication number
- CN114004252A CN114004252A CN202111205972.0A CN202111205972A CN114004252A CN 114004252 A CN114004252 A CN 114004252A CN 202111205972 A CN202111205972 A CN 202111205972A CN 114004252 A CN114004252 A CN 114004252A
- Authority
- CN
- China
- Prior art keywords
- source domain
- data set
- fault diagnosis
- target domain
- domain sample
- 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 144
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 32
- 230000007246 mechanism Effects 0.000 claims abstract description 18
- 238000000605 extraction Methods 0.000 claims abstract description 15
- 238000003860 storage Methods 0.000 claims abstract description 9
- 238000009826 distribution Methods 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 30
- 230000001133 acceleration Effects 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 239000004576 sand Substances 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000002372 labelling Methods 0.000 claims description 4
- 238000013508 migration Methods 0.000 description 21
- 230000005012 migration Effects 0.000 description 21
- 238000010586 diagram Methods 0.000 description 11
- 238000005096 rolling process Methods 0.000 description 6
- 238000013526 transfer learning Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012800 visualization Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000306 component Substances 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Signal Processing (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a bearing fault diagnosis method, a bearing fault diagnosis device, bearing fault diagnosis equipment and a computer readable storage medium, wherein the bearing fault diagnosis method comprises the following steps: acquiring a bearing vibration signal to construct a multi-source domain and target domain data set, constructing a bearing fault diagnosis model, processing the data set of the multi-source domain and the target domain, inputting the data set into a feature learner to perform feature extraction, solving a moment distance according to the extracted sample features of the multi-source domain and the target domain, inputting each source domain sample feature into a corresponding classifier, outputting a prediction label, calculating cross entropy loss of the classifier with a real label, constructing a target function of the model by using the moment distance and the cross entropy loss of the classifier, and training the model by using an intra-class alignment training strategy; the target domain data set is input into a trained model, and a comprehensive prediction result is output through a weighted classification mechanism.
Description
Technical Field
The present invention relates to the technical field of mechanical fault diagnosis and machine learning, and in particular, to a method, an apparatus, a device, and a computer-readable storage medium for bearing fault diagnosis.
Background
With the ubiquitous existence of rotary machines in modern industrial production, rolling bearings are used as key parts in industrial production equipment and widely applied to various important fields such as machinery, electric power, chemical industry, aviation and the like. As a core component of a rotary machine, a rolling bearing operates at a high speed, under a load and under vibration for a long time, and thus, a failure easily occurs, causing a great economic loss, and even a great deal of casualties. The development of fault diagnosis technology has been to solve this problem and avoid such losses. The rolling bearing monitoring system can monitor the running condition of the rolling bearing, even predict the future fault trend of the rolling bearing, avoid accidents and avoid personnel and economic losses. Therefore, the fault diagnosis technology has important social significance and economic value.
In recent years, fault diagnosis technology has been a popular research direction in academia. With the development of artificial intelligence related theory, the fault diagnosis technology based on machine learning is also widely researched. This technique comprises three main steps: 1. and (2) acquiring signals by using a sensor, manually extracting features, and (3) training and classifying a model. These methods are very effective and have achieved various successful applications. However, the limitations of the above-mentioned fault diagnosis method based on machine learning are obvious, such as time and labor consumption for manually extracting features.
Many scholars have proposed solutions using deep learning methods. These methods can overcome the disadvantages of the conventional methods because they have a strong ability to learn features. The deep neural network is composed of a plurality of hidden layers, and high-dimensional features can be learned from raw data. By training a model, the deep neural network can automatically select high-dimensional features capable of improving the accuracy of a prediction result by using parameter adjustment, so that accurate judgment is made. However, the precondition for achieving good effect of the deep learning method is that: the data used at diagnosis and the data used at training follow the same distribution. This precondition cannot be satisfied in real life, and the working conditions of the rotary machine often change, so that the effect of the deep learning method in fault diagnosis under variable working conditions is unsatisfactory.
The theory related to the transfer learning method has been introduced into the field of fault diagnosis. The goal of migration learning is to extract knowledge from one or more source domains and then apply it to a target domain. The fault diagnosis method based on the transfer learning can effectively apply the knowledge learned from the source task under certain working conditions to the target task under other working conditions, and enhances the capability of the model for processing vibration data under different working conditions. Therefore, migration-based fault diagnosis methods have attracted extensive attention and research in academia.
Currently, most fault diagnosis studies based on migration learning focus on single-source domain migration. However, this approach cannot handle the actual situation, i.e., the situation where multiple source domains must be migrated. Given the diversity of bearing operating conditions, the marked samples are typically from different operating conditions, thereby forming multiple source domains.
In summary, it can be seen that how to establish a multi-source domain migration learning fault diagnosis model and apply the model to actual processing conditions is a problem to be solved at present.
Disclosure of Invention
The invention aims to provide a bearing fault diagnosis method, a bearing fault diagnosis device, bearing fault diagnosis equipment and a computer readable storage medium, which aim to solve the problem that a fault diagnosis model of single-source domain transfer learning in the prior art cannot diagnose actual bearing faults.
In order to solve the above technical problem, the present invention provides a method for diagnosing a bearing fault, including: acquiring vibration signals of a bearing in operation under each working condition by using an acceleration sensor, constructing a multi-source domain data set and a target domain data set, and constructing a bearing fault diagnosis model; processing the multi-source domain data set and the target domain data set by using fast Fourier transform, performing two-dimensional processing, and outputting processed multi-source domain sample pictures and target domain sample pictures; inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space; calculating a moment distance by using the multi-source domain sample characteristics and the target domain sample characteristics, inputting each source domain sample characteristic into a corresponding source domain classifier, processing by each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label and a real label of each source domain sample characteristic; constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier, training and searching for the optimal parameter of the objective function through the multi-source domain data set, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic by using an in-class alignment training strategy until the bearing fault diagnosis model is trained; inputting the target domain data set into the bearing fault diagnosis model which is trained, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
In an embodiment of the present invention, the acquiring, by an acceleration sensor, a vibration signal of a bearing during operation under each working condition, and constructing a multi-source domain data set and a target domain data set, wherein constructing a bearing fault diagnosis model includes:
collecting vibration signals of bearings under the k working condition by using an acceleration sensor to serve as a k source domain data set
Forming K source domain data sets into the multi-source domain data set DsAnd labeling a bearing fault type label of the multi-source domain data set, wherein,
the number of the samples is the number of the samples,
for the ith tagged data set from the kth source domain,
a failure label for the kth source domain;
selecting a vibration signal acquired by an acceleration sensor when a working condition bearing runs as the target domain data set
Wherein
An unlabeled data set representing the ith target domain;
and establishing the bearing fault diagnosis model by using the multi-source domain data set and the target domain data set.
In an embodiment of the present invention, said calculating a moment distance using the multi-source domain sample features and the target domain sample features comprises:
using the multi-source domain sample features and the target domain sample features by
Calculating the distance of moment；
Wherein, MD (D)s,Dt) Is DsAnd DtThe distance of the moment between the two elements,
respectively, a set of i.i.d. sample characteristics, p being the total order,
numerically satisfying all combinations of 2 sample features selected from K of the source domain sample features
Is the p-order origin moment of X, M is the total number of samples in X, XiIs the ith sample in X.
In an embodiment of the present invention, said calculating the classifier cross-entropy loss by using the prediction label and the true label of each source domain sample feature comprises:
the prediction label and the real label of each source domain sample feature are processed by L (r, p) ═ sigmairilog(pi) Calculating the classifier cross entropy loss
Wherein r isiR is an index function when i is a real labeliEqual to 1, when i is not a true tag riIs equal to 0, piAnd outputting the output probability of the prediction label in the ith category for the bearing fault diagnosis model.
In an embodiment of the present invention, said constructing a model objective function using said moment distances and said classifier cross-entropy losses comprises:
constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier
Wherein G is the parameter of the feature learner, C is the parameter of all classifiers,
as a classifier CiAnd (3) softmax cross entropy loss on the ith source domain sample characteristic, wherein lambda is a hyperparameter for adjusting the weight between the classifier loss and the time matching loss.
In an embodiment of the present invention, the reducing the edge distribution and conditional distribution difference of the source domain sample feature and the target domain sample feature by using the intra-class alignment training strategy includes:
s61: for said each source domain classifier CiConstructing a companion classifier C'iForm a classifier pair (C)i，Ci) Then K classifier pairs form the set C { (C)1,C′1),(C2,C′2),…,(CK,C′k)}；
S62: optimizing the parameters of the feature learner G and the set C' so that the bearing fault diagnosis model can correctly classify the multi-source domain sample pictures;
s63: fixing the parameters of the feature learner G, optimizing the parameters of the set C', and expanding the difference of each classifier pair on the target domain sample features;
s64: fixing the parameters of the set C', optimizing the parameters of the feature learner G, and reducing the difference of each classifier pair on the target domain sample features;
s65: and training the target function based on the multi-source domain data set, and circularly executing the steps S62 to S64 until the target function is converged and the training is completed, and reducing the edge distribution and condition distribution difference of the multi-source domain sample characteristics and the target domain sample characteristics.
In an embodiment of the present invention, the inputting the target domain data set into the trained bearing fault diagnosis model, performing comprehensive fault prediction on the target domain sample through a joint weighted classification mechanism, and outputting the bearing fault diagnosis result of the target domain data set includes:
processing the target domain data set by using the fast Fourier change to obtain a target domain sample picture;
inputting the target domain sample picture into the feature learner, and extracting the target domain sample feature ZtThe target domain sample is characterized by ZtInputting the result into each source domain classifier, and forming an output vector by the result output by each source domain classifier
Based on weight vectors
And the output vector
Constructing the joint weighted classification mechanism, wherein the weight vector
Satisfy the constraint condition
Outputting the comprehensive fault diagnosis result of the target domain data set through the combined weighted classification mechanism
Wherein, ytIntegrated bearing fault diagnosis of the target domain dataset for a joint weighted classification mechanismPrediction result, ωiIs the weight of the ith classifier,
sample features Z in input target domain for ith classifiertAnd (6) outputting the data.
The invention also provides a bearing fault diagnosis device, which comprises:
the acquisition module is used for acquiring vibration signals of the bearing in operation under each working condition by using the acceleration sensor, constructing a multi-source domain data set and a target domain data set and constructing a bearing fault diagnosis model;
the processing module is used for processing the multi-source domain data set and the target domain data set by utilizing fast Fourier transform, performing two-dimensional processing and outputting processed multi-source domain sample pictures and target domain sample pictures;
the feature extraction module is used for inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space;
the calculation module is used for calculating moment distance by using the multi-source domain sample characteristics and the target domain sample characteristics, inputting each source domain sample characteristic into a corresponding source domain classifier, processing the characteristics through each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label of each source domain sample characteristic and a real label;
the training module is used for constructing an objective function of the bearing fault diagnosis model by using the moment distance and the classifier cross entropy loss, searching the optimal parameter of the objective function through the multi-source domain data set training, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic by using an in-class alignment training strategy until the bearing fault diagnosis model finishes training;
and the test module is used for inputting the target domain data set into the bearing fault diagnosis model which finishes training, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
The invention also provides a bearing fault diagnosis device, which comprises:
a memory for storing a computer program; a processor for implementing the steps of a method for bearing fault diagnosis described above when executing the computer program.
The invention also provides a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of a method of bearing fault diagnosis as described above.
According to the bearing fault diagnosis method provided by the invention, firstly, vibration signals of the bearing are collected through the acceleration sensor, a multi-source domain data set and a target domain data set are constructed, and a bearing fault diagnosis model is constructed, so that more knowledge can be better provided for a target domain by adopting the multi-source domain data set, and the bearing fault diagnosis model can be better used for fault diagnosis under various working conditions; preprocessing by fast Fourier transform, inputting the preprocessed data into a feature learner for feature extraction, outputting multi-source domain sample features and target domain sample features, calculating moment distance, reducing the difference between a multi-source domain and a target domain, facilitating better feature alignment, inputting the source domain sample features into a corresponding classifier to obtain a prediction label, calculating out cross entropy loss of the classifier by using the prediction label and a real label of each source domain, constructing a bearing fault diagnosis model target function according to the distance and the cross entropy loss of the classifier, training the model by using an in-class alignment strategy, reducing the edge distribution and condition distribution difference of the source domain and the target domain features, leading the fault diagnosis result to be high in accuracy and stronger in robustness, completing the training until the function convergence, and inputting the target domain data into the trained bearing fault diagnosis model, and outputting a bearing fault diagnosis result of the target domain data set through a joint classification mechanism. The method constructs a multi-source domain bearing fault diagnosis model, adopts the multi-source domain to provide more knowledge for a target domain, can be suitable for bearing fault detection under various working conditions, and adopts an intra-class alignment training strategy to reduce the difference of edge distribution and condition distribution among the domains.
Drawings
In order to more clearly illustrate the embodiments or technical solutions of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a flow chart of a first embodiment of a method of bearing fault diagnosis provided by the present invention;
FIG. 2 is a flow chart of a second embodiment of a method of bearing fault diagnosis provided by the present invention;
FIG. 3 is a test chart of a rolling bearing data generating test bed according to the present invention;
FIG. 4a is a sample graph of a bearing vibration signal at a load condition of 0 kN;
FIG. 4b is a sample exemplary graph of a bearing vibration signal at a load of 1 kN;
FIG. 4c is an exemplary graph of a bearing vibration signal sample at a load condition of 2 kN;
FIG. 4d is an exemplary graph of a bearing vibration signal sample at a load condition of 3 kN;
FIG. 5 is a schematic diagram of the structure of the joint weighted classifier of the present invention;
FIG. 6 is a graph of fault diagnosis accuracy for each migration domain task for the method and variation of the present invention;
FIG. 7a is a fault diagnosis confusion matrix diagram for multi-source domain migration task 1,2, 3-0;
FIG. 7b is a fault diagnosis confusion matrix diagram for multi-source domain migration task 0,2, 3-1;
FIG. 7c is a fault diagnosis confusion matrix diagram for multi-source domain migration task 0,1, 3-2;
FIG. 7d is a fault diagnosis confusion matrix diagram for multi-source domain migration tasks 0,1, 2-3;
FIG. 8a is a schematic diagram of the visualization of fault diagnosis characteristics of variant 1 migration tasks 2-0;
FIG. 8b is a schematic diagram of the visualization of the fault diagnosis characteristics of variant 2 migration tasks 1,2, 3-0;
FIG. 8c is a schematic diagram of the visualization of fault diagnosis features of variant 3 migration tasks 1,2, 3-0;
FIG. 8d is a schematic diagram illustrating the visualization of fault diagnosis features of migration tasks 1,2,3-0 according to the present invention;
fig. 9 is a block diagram of a bearing fault diagnosis apparatus according to an embodiment of the present invention.
Detailed Description
The core of the invention is to provide a bearing fault diagnosis method, a device, equipment and a computer readable storage medium, construct a multi-source domain transfer learning fault diagnosis model, and apply the model to bearing fault diagnosis for processing various variable working conditions in actual life.
In order that those skilled in the art will better understand the disclosure, the invention will be described in further detail with reference to the accompanying drawings and specific embodiments. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a bearing fault diagnosis method according to a first embodiment of the present invention; the specific operation steps are as follows:
step S101: acquiring vibration signals of a bearing in operation under each working condition by using an acceleration sensor, constructing a multi-source domain data set and a target domain data set, and constructing a bearing fault diagnosis model;
step S102: processing the multi-source domain data set and the target domain data set by using fast Fourier transform, performing two-dimensional processing, and outputting processed multi-source domain sample pictures and target domain sample pictures;
step S103: inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space;
step S104: calculating a moment distance by using the multi-source domain sample characteristics and the target domain sample characteristics, inputting each source domain sample characteristic into a corresponding source domain classifier, processing by each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label and a real label of each source domain sample characteristic;
step S105: constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier, training and searching for the optimal parameter of the objective function through the multi-source domain data set, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic by using an in-class alignment training strategy until the bearing fault diagnosis model is trained;
step S106: inputting the target domain data set into the bearing fault diagnosis model which is trained, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
According to the method provided by the embodiment, the acceleration sensor is used for collecting bearing vibration signals under different working conditions to construct the multi-source domain data set and the target domain data set, the bearing fault diagnosis model is constructed, and the multi-source domain data set can better provide more knowledge for the target domain, so that the bearing fault diagnosis model can be better used for fault diagnosis under various working conditions; the moment distance in the model reduces the difference between a multi-source domain and a target domain, is beneficial to better aligning characteristics, and trains the model by adopting an in-class alignment training strategy, so that the difference between the edge distribution and the condition distribution of the multi-source domain sample characteristics and the target domain sample characteristics is reduced.
Based on the above embodiment, the above embodiment is explained in more detail in this embodiment, a transferable data set including seven healthy bearings under four non-operating loads is constructed from the collected vibration signals, and a bearing fault model is trained, please refer to fig. 2, where fig. 2 is a flowchart of a second specific embodiment of the method for diagnosing a bearing fault provided by the present invention; the specific operation steps are as follows:
step S201: acquiring vibration signals of a bearing in operation under seven working conditions by using an acceleration sensor, constructing a multi-source domain data set and a target domain data set, and constructing a bearing fault diagnosis model;
the signals collected by the test stand shown in fig. 3 are used to construct a transferable data set containing seven healthy bearings under four dead loads, and different condition distributions and edge distributions exist in each data set. The test bearings were set to four basic health conditions (normal, inner ring failure, roller failure and outer ring failure) and two failure sizes (0.2 mm and 0.3 mm). Thus, seven health states were obtained, as listed in table 1. There are 100 training samples and 100 test samples per state of health. Each sample consists of 1024 sampling points.
TABLE 1 seven types of faulty bearing information for each domain
Four fields are established based on four operating conditions, namely four load conditions (0, 1,2 and 3kN), a sample of which is shown in fig. 4. In addition, four multi-source domain migration tasks are set. For clarity of description, migration tasks and their abbreviations are provided in table 2.
Table 2 all migration tasks and abbreviations
Collecting vibration signals of bearings under the k working condition by using an acceleration sensor to serve as a k source domain data set
Forming K source domain data sets into the multi-source domain data set DsAnd labeling a bearing fault type label of the source domain data set, wherein,
the number of the samples is the number of the samples,
for the ith tagged data set from the kth source domain,
a failure label for the kth source domain;
selecting a vibration signal acquired by an acceleration sensor when a working condition bearing runs as the target domain data set
Wherein
An unlabeled data set representing the ith target domain;
and establishing the bearing fault diagnosis model by using the multi-source domain data set and the target domain data set.
Step S202: processing the multi-source domain data set and the target domain data set by using fast Fourier transform, performing two-dimensional processing, and outputting processed multi-source domain sample pictures and target domain sample pictures;
unifying the multi-source domain data set and the target domain data set into three-channel data. The number of channels is 1, the sample is copied into three parts, and the three parts are layered on the channel dimension to form three-channel data;
a Fast Fourier Transform (FFT) is performed on the multi-source domain dataset and the target domain dataset. The FFT transforms the samples from a time domain signal to a frequency domain signal while preserving the 3 x 1024 shape, but helps the feature learner to extract features more easily;
the multi-source and target domain datasets are reshaped from 3 x 1024 to 3 x 32. 1024 points on each channel were divided into 32 shares, each with 32 points, and reconstructed in a new dimension to obtain the ideal shape of 3 × 32 × 32.
Step S203: inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space;
constructing a feature learner according to the parameters in the table 3, wherein the feature learner is essentially a convolutional neural network mainly composed of three convolutional layers and two fully-connected layers;
table 3 structural parameters of feature learner
And taking the sample subjected to data preprocessing as an input, and obtaining the feature representation of the sample by a feature extractor.
Step S204: calculating a moment distance MD (D) using the multi-source domain sample features and the target domain sample featuress,Dt)；
Using the multi-source domain sample features and the target domain sample features by
Wherein, MD (D)s,Dt) Is DsAnd DtThe distance of the moment between the two elements,
the edge is a set consisting of i.i.d. samples, p is the total order (the larger the order, the more accurate the calculation, but the more calculation effort is consumed),
for the moment of origin of order p of the ith source domain sample X,
for the moment of origin of order p of the target domain sample X,
numerically satisfy all combinations of 2 sample features selected from the K source domain sample features
Is the p-order origin moment of X, M is the total number of samples in X, XiIs the ith sample in X.
Step S205: inputting each source domain sample characteristic into a corresponding source domain classifier, processing through each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label and a real label of each source domain sample characteristic;
the prediction label and the real label of each source domain sample feature are processed by L (r, p) ═ sigmairilog(pi) Calculating the classifier cross entropy loss
Wherein r isiIs an index function, i is trueTime of labelling riEqual to 1, when i is not a true tag riIs equal to 0, piAnd outputting the output probability of the prediction label in the ith category for the bearing fault diagnosis model.
Step S206: constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier;
constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier
Wherein G is the parameter of the feature learner, C is the parameter of all classifiers,
as a classifier CiSoftmax cross entropy loss on the ith source domain, λ is a hyperparameter that adjusts the weight between classifier loss and time matching loss.
Step S207: constructing a classifier pair set by utilizing the multi-source domain data set intra-class alignment training strategy, and optimizing the parameters of the feature learner and the classifier pair set;
for said each source domain classifier CiConstructing a companion classifier C'iForm a classifier pair (C)i，C′i) Then K classifier pairs form the set C { (C)1,C′1),(C2,C′2),…,(CK,C′k)}；
Optimizing the parameters of the feature learner G and the classifier pair set C' so that the bearing fault diagnosis model can correctly classify multi-source domain samples, and the objective function at this stage is the core objective function of the method;
the correct classification is the accuracy of the predicted label and the real label output by the source domain classifier.
Step S208: fixing the parameters of the feature learner, and optimizing the parameters of the classifier pair set;
fixing the parameters of the feature learner G, optimizing the set of classifier pairs C′Expands the difference of each classifier pair on the target domain samples, which is expressed mathematically as
Step S209: fixing the parameters of the classifier pair set, and optimizing the parameters of the feature learner;
fixing the parameters of the classifier pair set C', optimizing the parameters of the feature learner G, and reducing the difference of each classifier pair on the target domain sample, wherein the mathematical expression of the target is
And (5) repeating the training steps S207 to S209 through the multi-source domain data set until the target function is converged and the training is completed, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic.
Step S2010: inputting the target domain data set into the bearing fault diagnosis model which is trained, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
Processing the target domain data set by using the fast Fourier change to obtain a target domain sample picture;
inputting the target domain sample picture into the feature learner, and extracting the target domain sample feature ZtThe target domain sample is characterized by ZtInputting the result into all the source domain classifiers, and forming an output vector by the result output by all the source domain classifiers
Based on weight vectors
And the output vector
The framework of the joint weighted classification mechanism, the joint weighted classifier, is shown in FIG. 5, wherein the weight vector
Satisfy the constraint condition
Outputting the comprehensive fault diagnosis result of the target domain data set through the weighted classification mechanism
Wherein, ytSynthetic bearing fault diagnosis prediction results, ω, for the target domain data set for a joint weighted classification mechanismiIs the weight of the ith classifier,
sample features Z in input target domain for ith classifiertAnd (6) outputting the data.
In the testing process, for a target domain sample needing fault diagnosis, comprehensive fault prediction is carried out on the target domain sample through a combined weighted classification mechanism. For a target domain sample to be detected, extracting a feature representation Z of the sample by a feature learnert. Then, ZtIs fed into all classifiers, all results of which form an output vector
Meanwhile, the introduction of the constraint condition
Weight vector of
The goal of the weight vector is to evaluate the degree of association of each source domain with the target domain. The method adopts the performance of the ith classifier only on corresponding source domain data to determine the weight omegaiIt can be calculated with the following formula:
wherein, acciAnd the classification accuracy of the ith classifier on the ith source domain data is represented.
The composite fault diagnosis result is defined as the dot product of the output and the weight vector, and can be calculated by the following formula:
wherein, ytSynthetic bearing fault diagnosis prediction results, ω, for the target domain data set for a joint weighted classification mechanismiIs the weight of the ith classifier,
inputting target domain characteristics Z for ith classifiertAnd (6) outputting the data.
Fig. 6, fig. 7, and fig. 8 show the fault diagnosis accuracy of the present method and each variation method under each migration task, respectively, and the present method visualizes the fault diagnosis confusion matrix of the four multi-source domain migration tasks and the fault diagnosis characteristics of the method set variation method, and from the above results, the average fault diagnosis accuracy of the present method reaches 99.96%, and the characteristic extraction effect is better. The invention has high fault diagnosis precision and good robustness, and can well process fault diagnosis under variable working conditions.
In summary, the invention discloses a bearing fault diagnosis method based on a multi-source domain adaptive network in a moment matching class, which adopts a feature learner to extract domain variable features and adopts a moment distance to evaluate domain differences. The superiority and robustness of the method are proved through experimental verification and comparison with several methods. Some of the results herein are summarized as follows: research finds that the proposed moment distance can effectively evaluate the difference between the multi-source domain and the target domain, which helps better feature alignment; the developed intra-class alignment training strategy can correctly align the condition distribution of each domain, which is crucial in the complex migration task under study; compared with single-source transfer learning, the multi-source transfer learning method can provide more knowledge for the target domain, so that fault diagnosis under variable working conditions can be more effectively carried out. Therefore, multi-source migration learning is a prospective direction to implement fault diagnosis in complex cases.
The method provided by the embodiment comprises the steps of collecting a multi-source domain data set and a target domain data set of seven data sets under different working conditions to train and test a bearing fault diagnosis model, wherein the multi-source domain data set can provide more knowledge for target domain testing, optimizing parameters of a feature learner and a classifier pair set by adopting an in-class alignment training strategy, and facilitating correct classification of sample features by the model, fixing the parameters of the feature learner, optimizing the parameters of the classifier pair set, fixing the parameters of the classifier pair set and optimizing the parameters of the feature learner; and (4) repeatedly training the model by using the multi-source domain data set, and reducing the edge distribution and the condition distribution of the target domain. Compared with other variant methods, the invention has the advantages of high average fault diagnosis accuracy rate of 99.96%, good feature extraction effect, good robustness and capability of processing fault diagnosis under variable working conditions.
Referring to fig. 9, fig. 9 is a block diagram of a bearing fault diagnosis apparatus according to an embodiment of the present invention; the specific device may include:
the acquisition module 100 is used for acquiring vibration signals of the bearing in operation under each working condition by using an acceleration sensor, constructing a multi-source domain data set and a target domain data set, and constructing a bearing fault diagnosis model;
a processing module 200, configured to process the multi-source domain data set and the target domain data set by using fast fourier transform, perform two-dimensional processing, and output processed multi-source domain sample pictures and target domain sample pictures;
a feature extraction module 300, configured to input the multi-source domain sample picture and the target domain sample picture into a feature learner, perform feature extraction by the feature learner, output a multi-source domain sample feature and a target domain sample feature, and map the multi-source domain sample feature and the target domain sample feature in the same feature space;
a calculating module 400, configured to calculate a moment distance by using the multi-source domain sample features and the target domain sample features, input each source domain sample feature into a corresponding source domain classifier, perform processing by each source domain classifier to obtain a prediction label of each source domain sample, and calculate a classifier cross entropy loss by using the prediction label and a real label of each source domain sample;
the training module 500 is configured to construct an objective function of the bearing fault diagnosis model by using the moment distance and the classifier cross entropy loss, search an optimal parameter of the objective function by training the multi-source domain data set, and reduce edge distribution and condition distribution differences of the source domain sample features and the target domain sample features by using an intra-class alignment training strategy until the bearing fault diagnosis model completes training;
the testing module 600 is configured to input the target domain data set into the bearing fault diagnosis model that has completed training, perform comprehensive fault prediction on the target domain data set through a joint weighted classification mechanism, and output a bearing fault diagnosis result of the target domain data set.
The bearing fault diagnosis device of this embodiment is used for implementing the foregoing bearing fault diagnosis method, and therefore a specific implementation manner of the bearing fault diagnosis device may refer to the foregoing embodiment parts of the bearing fault diagnosis method, for example, the acquisition module 100, the processing module 200, the feature extraction module 300, the calculation module 400, the training module 500, and the test module 600, which are respectively used for implementing steps S101, S102, S103, S104, S105, and S106 in the bearing fault diagnosis method, and therefore, the specific implementation manner thereof may refer to descriptions of corresponding respective part embodiments, and is not repeated herein.
The specific embodiment of the present invention further provides a bearing fault diagnosis device, including: a memory for storing a computer program; a processor for implementing the steps of a method for bearing fault diagnosis described above when executing the computer program.
The embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the method for diagnosing bearing faults as described above.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The bearing fault diagnosis method, device, equipment and computer readable storage medium provided by the invention are described in detail above. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Claims (10)
1. A method of bearing fault diagnosis, comprising:
acquiring vibration signals of a bearing in operation under each working condition by using an acceleration sensor, constructing a multi-source domain data set and a target domain data set, and constructing a bearing fault diagnosis model;
processing the multi-source domain data set and the target domain data set by using fast Fourier transform, performing two-dimensional processing, and outputting processed multi-source domain sample pictures and target domain sample pictures;
inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space;
calculating a moment distance by using the multi-source domain sample characteristics and the target domain sample characteristics, inputting each source domain sample characteristic into a corresponding source domain classifier, processing by each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label and a real label of each source domain sample characteristic;
constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier, training and searching for the optimal parameter of the objective function through the multi-source domain data set, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic by using an in-class alignment training strategy until the bearing fault diagnosis model is trained;
inputting the target domain data set into the bearing fault diagnosis model which is trained, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
2. The method of claim 1, wherein the step of collecting vibration signals of the bearing during operation under each working condition by using the acceleration sensor to construct a multi-source domain data set and a target domain data set, and the step of constructing the bearing fault diagnosis model comprises the following steps:
collecting vibration signals of bearings under the k working condition by using an acceleration sensor to serve as a k source domain data set
Forming the K source domain data sets into a multi-source domain data set Ds, and labeling a bearing fault type label of the multi-source domain data set Ds, wherein,
the number of the samples is the number of the samples,
for the ith tagged data set from the kth source domain,
a failure label for the kth source domain;
selecting a vibration signal acquired by an acceleration sensor when a working condition bearing runs as the target domain data set
Wherein
An unlabeled data set representing the ith target domain;
and establishing the bearing fault diagnosis model by using the multi-source domain data set and the target domain data set.
3. The method of claim 2, wherein the calculating a moment distance using the multi-source domain sample features and the target domain sample features comprises:
using the multi-source domain sample features and the target domain sample features by
Calculating a moment distance;
wherein, MD (D)s，Dt) Is DsAnd DtThe distance of the moment between the two elements,
respectively, a set of i.i.d. sample characteristics, p being the total order,
numerically satisfying all combinations of 2 sample features selected from K of the source domain sample features
Is the p-order origin moment of X, M is the total number of samples in X, XiIs the ith sample in X.
4. The method of claim 3, wherein said computing classifier cross-entropy losses using the prediction label and the true label for each source domain sample feature comprises:
the prediction label and the real label of each source domain sample feature are processed by L (r, p) ═ sigmairilog(pi) Calculating the classifier cross entropy loss
Wherein r isiR is an index function when i is a real labeliEqual to 1, when i is not a true tag riIs equal to 0, piAnd outputting the output probability of the prediction label in the ith category for the bearing fault diagnosis model.
5. The method of claim 4, wherein said constructing a model objective function using said moment distances and said classifier cross-entropy losses comprises:
constructing an objective function of the bearing fault diagnosis model by using the moment distance and the cross entropy loss of the classifier
Wherein G is the parameter of the feature learner, C is the parameter of all classifiers,
as a classifier CiAnd (3) softmax cross entropy loss on the ith source domain sample characteristic, wherein lambda is a hyperparameter for adjusting the weight between the classifier loss and the time matching loss.
6. The method of claim 5, wherein the utilizing an intra-class alignment training strategy to narrow down edge distribution and conditional distribution differences of the source domain sample features and the target domain sample features comprises:
s61: for said each source domain classifier CiConstructing a companion classifier C'iForm a classifier pair (C)i，C′i) Then K classifier pairs form the set C { (C)1，C′1)，(C2，C′2)，...，(CK，C′k)}；
S62: optimizing the parameters of the feature learner G and the set C' so that the bearing fault diagnosis model can correctly classify the multi-source domain sample pictures;
s63: fixing the parameters of the feature learner G, optimizing the parameters of the set C', and expanding the difference of each classifier pair on the target domain sample features;
s64: fixing the parameters of the set C', optimizing the parameters of the feature learner G, and reducing the difference of each classifier pair on the target domain sample features;
s65: and training the target function based on the multi-source domain data set, and circularly executing the steps S62 to S64 until the target function is converged and the training is completed, and reducing the edge distribution and condition distribution difference of the multi-source domain sample characteristics and the target domain sample characteristics.
7. The method of claim 1, wherein the inputting the target domain data set into the trained bearing fault diagnosis model, performing comprehensive fault prediction on the target domain samples through a joint weighted classification mechanism, and outputting the bearing fault diagnosis result of the target domain data set comprises:
processing the target domain data set by using the fast Fourier change to obtain a target domain sample picture;
inputting the target domain sample picture into the feature learner, and extracting the target domain sample featureSign ZtThe target domain sample is characterized by ZtInputting the result into each source domain classifier, and forming an output vector by the result output by each source domain classifier
Based on weight vectors
And the output vector
Constructing the joint weighted classification mechanism, wherein the weight vector
Satisfy the constraint condition
Outputting the comprehensive fault diagnosis result of the target domain data set through the combined weighted classification mechanism
Wherein, ytSynthetic bearing fault diagnosis prediction results, ω, for the target domain data set for a joint weighted classification mechanismiIs the weight of the ith classifier,
sample features Z in input target domain for ith classifiertAnd (6) outputting the data.
8. An apparatus for bearing fault diagnosis, comprising:
the acquisition module is used for acquiring vibration signals of the bearing in operation under each working condition by using the acceleration sensor, constructing a multi-source domain data set and a target domain data set and constructing a bearing fault diagnosis model;
the processing module is used for processing the multi-source domain data set and the target domain data set by utilizing fast Fourier transform, performing two-dimensional processing and outputting processed multi-source domain sample pictures and target domain sample pictures;
the feature extraction module is used for inputting the multi-source domain sample picture and the target domain sample picture into a feature learner, performing feature extraction through the feature learner, outputting multi-source domain sample features and target domain sample features, and mapping the multi-source domain sample features and the target domain sample features to the same feature space;
the calculation module is used for calculating moment distance by using the multi-source domain sample characteristics and the target domain sample characteristics, inputting each source domain sample characteristic into a corresponding source domain classifier, processing the characteristics through each source domain classifier to obtain a prediction label of each source domain sample characteristic, and calculating cross entropy loss of the classifier by using the prediction label of each source domain sample characteristic and a real label;
the training module is used for constructing an objective function of the bearing fault diagnosis model by using the moment distance and the classifier cross entropy loss, searching the optimal parameter of the objective function through the multi-source domain data set training, and reducing the edge distribution and condition distribution difference of the source domain sample characteristic and the target domain sample characteristic by using an in-class alignment training strategy until the bearing fault diagnosis model finishes training;
and the test module is used for inputting the target domain data set into the bearing fault diagnosis model which finishes training, performing comprehensive fault prediction on the target domain sample through a combined weighted classification mechanism, and outputting a bearing fault diagnosis result of the target domain data set.
9. An apparatus for bearing fault diagnosis, comprising:
a memory for storing a computer program;
a processor for implementing the steps of a method of bearing fault diagnosis as claimed in any one of claims 1 to 7 when executing said computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of a method of bearing fault diagnosis according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111205972.0A CN114004252A (en) | 2021-10-13 | 2021-10-13 | Bearing fault diagnosis method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111205972.0A CN114004252A (en) | 2021-10-13 | 2021-10-13 | Bearing fault diagnosis method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114004252A true CN114004252A (en) | 2022-02-01 |
Family
ID=79923114
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111205972.0A Pending CN114004252A (en) | 2021-10-13 | 2021-10-13 | Bearing fault diagnosis method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114004252A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115165362A (en) * | 2022-06-24 | 2022-10-11 | 鞍钢集团矿业有限公司 | Intelligent bearing fault detection system and method based on inspection and vibration signals |
| CN115356105A (en) * | 2022-08-31 | 2022-11-18 | 北京理工大学 | Bearing fault diagnosis method and device, electronic equipment and readable storage medium |
| CN115659224A (en) * | 2022-11-07 | 2023-01-31 | 苏州大学 | Probability-guided domain confrontation bearing fault diagnosis method and system |
| CN116188449A (en) * | 2023-03-13 | 2023-05-30 | 哈尔滨市科佳通用机电股份有限公司 | Rail wagon relief valve pull rod split pin loss fault identification method and equipment |
| CN116304820A (en) * | 2023-03-14 | 2023-06-23 | 广东石油化工学院 | Bearing fault type prediction method and system based on multi-source domain transfer learning |
| CN116541713A (en) * | 2023-06-27 | 2023-08-04 | 清华大学 | Bearing fault diagnosis model training method based on local time-frequency characteristic transfer learning |
| CN116756483A (en) * | 2023-05-08 | 2023-09-15 | 苏州大学 | Mechanical fault diagnosis method, device and equipment under condition that target working condition data are unavailable |
| CN116894190A (en) * | 2023-09-11 | 2023-10-17 | 江西南昌济生制药有限责任公司 | Bearing fault diagnosis method, device, electronic equipment and storage medium |
| CN117313000A (en) * | 2023-09-19 | 2023-12-29 | 北京交通大学 | Motor brain learning fault diagnosis method based on sample characterization topology |
| CN117349749A (en) * | 2023-10-09 | 2024-01-05 | 石家庄铁道大学 | Multi-source domain bearing fault diagnosis method based on mixed convolution |
| CN118190414A (en) * | 2024-01-31 | 2024-06-14 | 佛山科学技术学院 | Intelligent bearing fault diagnosis method and equipment based on multilayer alignment strategy domain adaptation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190325299A1 (en) * | 2018-04-18 | 2019-10-24 | Element Ai Inc. | Unsupervised domain adaptation with similarity learning for images |
| CN111506862A (en) * | 2020-05-01 | 2020-08-07 | 西北工业大学 | Rolling bearing fault diagnosis method based on multi-source weighted integration transfer learning |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Method for diagnosing similar self-adaptive bearing fault under variable working conditions |
| CN112629863A (en) * | 2020-12-31 | 2021-04-09 | 苏州大学 | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions |
| CN112860183A (en) * | 2021-01-07 | 2021-05-28 | 西安交通大学 | Multisource distillation-migration mechanical fault intelligent diagnosis method based on high-order moment matching |
-
2021
- 2021-10-13 CN CN202111205972.0A patent/CN114004252A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190325299A1 (en) * | 2018-04-18 | 2019-10-24 | Element Ai Inc. | Unsupervised domain adaptation with similarity learning for images |
| CN111506862A (en) * | 2020-05-01 | 2020-08-07 | 西北工业大学 | Rolling bearing fault diagnosis method based on multi-source weighted integration transfer learning |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Method for diagnosing similar self-adaptive bearing fault under variable working conditions |
| CN112629863A (en) * | 2020-12-31 | 2021-04-09 | 苏州大学 | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions |
| CN112860183A (en) * | 2021-01-07 | 2021-05-28 | 西安交通大学 | Multisource distillation-migration mechanical fault intelligent diagnosis method based on high-order moment matching |
Non-Patent Citations (2)
| Title |
|---|
| KUNIAKI SAITO: "Maximum Classifier Discrepancy for Unsupervised Domain Adaptation", 《PROCEEDINGS OF THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| XINGCHAO PENG: "Moment Matching for Multi-Source Domain Adaptation", 《HTTPS://ARXIV.ORG/ABS/1812.01754V3》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115165362A (en) * | 2022-06-24 | 2022-10-11 | 鞍钢集团矿业有限公司 | Intelligent bearing fault detection system and method based on inspection and vibration signals |
| CN115356105A (en) * | 2022-08-31 | 2022-11-18 | 北京理工大学 | Bearing fault diagnosis method and device, electronic equipment and readable storage medium |
| CN115659224A (en) * | 2022-11-07 | 2023-01-31 | 苏州大学 | Probability-guided domain confrontation bearing fault diagnosis method and system |
| CN115659224B (en) * | 2022-11-07 | 2023-08-25 | 苏州大学 | Probability-guided domain countermeasure bearing fault diagnosis method and system |
| CN116188449A (en) * | 2023-03-13 | 2023-05-30 | 哈尔滨市科佳通用机电股份有限公司 | Rail wagon relief valve pull rod split pin loss fault identification method and equipment |
| CN116188449B (en) * | 2023-03-13 | 2023-08-08 | 哈尔滨市科佳通用机电股份有限公司 | Rail wagon relief valve pull rod split pin loss fault identification method and equipment |
| CN116304820A (en) * | 2023-03-14 | 2023-06-23 | 广东石油化工学院 | Bearing fault type prediction method and system based on multi-source domain transfer learning |
| CN116304820B (en) * | 2023-03-14 | 2023-12-01 | 广东石油化工学院 | Bearing fault type prediction method and system based on multi-source domain transfer learning |
| CN116756483A (en) * | 2023-05-08 | 2023-09-15 | 苏州大学 | Mechanical fault diagnosis method, device and equipment under condition that target working condition data are unavailable |
| CN116541713A (en) * | 2023-06-27 | 2023-08-04 | 清华大学 | Bearing fault diagnosis model training method based on local time-frequency characteristic transfer learning |
| CN116541713B (en) * | 2023-06-27 | 2023-12-12 | 清华大学 | Bearing fault diagnosis model training method based on local time-frequency characteristic transfer learning |
| CN116894190A (en) * | 2023-09-11 | 2023-10-17 | 江西南昌济生制药有限责任公司 | Bearing fault diagnosis method, device, electronic equipment and storage medium |
| CN116894190B (en) * | 2023-09-11 | 2023-11-28 | 江西南昌济生制药有限责任公司 | Bearing fault diagnosis method, device, electronic equipment and storage medium |
| CN117313000A (en) * | 2023-09-19 | 2023-12-29 | 北京交通大学 | Motor brain learning fault diagnosis method based on sample characterization topology |
| CN117313000B (en) * | 2023-09-19 | 2024-03-15 | 北京交通大学 | Motor brain learning fault diagnosis method based on sample characterization topology |
| CN117349749A (en) * | 2023-10-09 | 2024-01-05 | 石家庄铁道大学 | Multi-source domain bearing fault diagnosis method based on mixed convolution |
| CN117349749B (en) * | 2023-10-09 | 2024-03-15 | 石家庄铁道大学 | Multi-source domain bearing fault diagnosis method based on mixed convolution |
| CN118190414A (en) * | 2024-01-31 | 2024-06-14 | 佛山科学技术学院 | Intelligent bearing fault diagnosis method and equipment based on multilayer alignment strategy domain adaptation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114004252A (en) | Bearing fault diagnosis method, device and equipment | |
| Zhao et al. | Deep multi-scale convolutional transfer learning network: A novel method for intelligent fault diagnosis of rolling bearings under variable working conditions and domains | |
| Grezmak et al. | Interpretable convolutional neural network through layer-wise relevance propagation for machine fault diagnosis | |
| CN111898634B (en) | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption | |
| CN111721536B (en) | Rolling bearing fault diagnosis method for improving model migration strategy | |
| Zhang et al. | A fault diagnosis method based on improved convolutional neural network for bearings under variable working conditions | |
| Chen et al. | A novel fault diagnosis model for gearbox based on wavelet support vector machine with immune genetic algorithm | |
| CN108510153B (en) | Multi-working-condition rotary machine fault diagnosis method | |
| Sugumaran et al. | Effect of number of features on classification of roller bearing faults using SVM and PSVM | |
| CN113065581B (en) | Vibration fault migration diagnosis method for reactance domain self-adaptive network based on parameter sharing | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| CN106628097A (en) | Ship equipment fault diagnosis method based on improved radial basis function neutral network | |
| CN117009916A (en) | Actuator fault diagnosis method based on multi-sensor information fusion and transfer learning | |
| CN114091504A (en) | Rotary machine small sample fault diagnosis method based on generation countermeasure network | |
| Huang et al. | Intelligent fault identification for industrial internet of things via prototype-guided partial domain adaptation with momentum weight | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN113869451B (en) | Rolling bearing fault diagnosis method under variable working condition based on improved JGSA algorithm | |
| Zhang et al. | Inferable deep distilled attention network for diagnosing multiple motor bearing faults | |
| Zhang et al. | Few-shot bearing fault diagnosis based on meta-learning with discriminant space optimization | |
| CN117113170A (en) | Lightweight rotary machine fault diagnosis method based on multi-scale information fusion | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| Xu et al. | An optimal method based on HOG-SVM for fault detection | |
| Chen et al. | Gearbox fault diagnosis method based on deep learning multi-task framework | |
| CN117390411B (en) | Variable working condition bearing fault diagnosis method based on meta-shift learning | |
| Li et al. | Transformer-based meta learning method for bearing fault identification under multiple small sample conditions |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220201 |