CN114755017B - Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network - Google Patents
Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network Download PDFInfo
- Publication number
- CN114755017B CN114755017B CN202210382997.6A CN202210382997A CN114755017B CN 114755017 B CN114755017 B CN 114755017B CN 202210382997 A CN202210382997 A CN 202210382997A CN 114755017 B CN114755017 B CN 114755017B
- Authority
- CN
- China
- Prior art keywords
- domain
- data set
- bearing
- domain data
- fault
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Acoustics & Sound (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a variable-speed bearing fault diagnosis method of a cross-domain data-driven unsupervised field shared network, which designs a maximum mean difference evaluation index based on cauchy nucleus induction to measure the difference between the distribution of a source domain and a target domain, and adopts an unbiased estimation technology to improve the calculation efficiency; a balance factor with adjustable segments is designed to balance the importance degree of the Cauchy nucleus induced maximum mean difference loss and the classification cross entropy loss and is used for constructing an overall loss function. The method can accurately extract the migratable features between the source domain and the target domain, improves the robustness of the domain self-adaptation process, effectively improves the calculation efficiency, saves a large amount of calculation cost, realizes higher diagnosis precision under the task of cross-domain fault migration at variable speed when the stable speed is reached, and is superior to the prior art.
Description
Technical Field
The invention relates to the technical field of cross-domain fault diagnosis, in particular to a variable-speed bearing fault diagnosis method of a cross-domain data driving unsupervised domain shared network.
Background
In recent years, with the rapid development of modern industries, the requirements for reliability and safety of intelligent and integrated machine systems have increased dramatically. After long-term operation under severe conditions, various faults occur in a key rotating transmission component of the bearing, and damage or even major accidents can be caused to machine equipment. The advanced intelligent bearing fault diagnosis and identification technology based on data driving is researched and developed, various fault modes of the bearing are accurately classified and diagnosed, and the operation reliability and the economic maintainability of machine equipment can be effectively improved. However, in an industrial scenario, it is very difficult to obtain a large number of samples of the same-equipment fault bearing with labeled information, and moreover, the variability of the working conditions and the complexity of the machine system cause a certain distribution difference between the collected fault samples, and the diagnosis of the bearing fault is more challenging.
Some researchers use an unsupervised migration learning method based on features to perform fault identification, and perform cross-domain migration on existing source domain data knowledge to solve the problem that the amount of labeled bearing samples in a target domain is insufficient, for example, a domain adaptive migration method based on maximum mean difference induced by gaussian kernel or polynomial kernel, a domain antagonistic migration learning method, and the like, have been widely applied to bearing fault diagnosis and identification research, but still have some problems to be improved: (1) At present, most of bearing fault source domain data come from a fault simulation test bed in a laboratory, the data are more suitable for researching the general rule of fault phenomena, the aim of establishing the simulation test bed with certain precision is not all the time, a large amount of lasting resource investment is needed, and the fault data requirements under various working conditions are difficult to flexibly meet. By means of a numerical simulation technology, a fault simulation model reflecting the real operation condition of a mechanical system can be established, and a large number of bearing fault samples with abundant fault information and sufficient label data are obtained, so that the problem of insufficient training samples is solved, and the resource dependence on a simulation test bed is reduced. (2) The self-adaptive migration method based on the Gaussian kernel induced maximum mean difference field has the advantages that the diagnostic performance is sensitive to the width parameter of the Gaussian kernel, the calculation complexity is high, and the model is difficult to effectively realize optimal convergence. The self-adaptive migration method based on the maximum mean difference field of polynomial nuclear induction needs to adjust three parameters simultaneously, and the high-order calculation time is relatively long; (3) The existing method research mainly focuses on cross-domain migration diagnosis under a stable rotating speed, but in order to meet the requirements of production tasks such as starting and braking in an actual industrial scene, cross-domain fault migration diagnosis under a time-varying rotating speed generally needs to be considered. The relationship between the signal at time varying speed and the fault category is much more complex than for a steady rotational speed.
Therefore, new techniques are introduced to accurately and rapidly perform unsupervised cross-domain migration fault diagnosis of bearing components from a stable speed to an instantaneous speed.
Disclosure of Invention
The invention aims to provide a variable-speed bearing fault diagnosis method of a cross-domain data driving unsupervised field shared network, which is used for effectively and quickly carrying out cross-domain fault migration diagnosis on a bearing when the bearing reaches a stable speed and changes speed; the method solves the problems that the existing feature migration diagnosis method based on maximum mean difference is high in computational complexity and difficult to effectively realize optimal convergence, and the existing migration method is low in diagnosis precision when a fault migration task is performed at a variable speed from a stable speed to a time, accurately extracts migratable features between a source domain and a target domain, improves robustness of a domain self-adaption process, effectively improves computational efficiency, saves a large amount of computational cost, realizes higher diagnosis precision when the fault migration task is performed at the variable speed from the stable speed to the time, realizes reliable bearing fault migration diagnosis, and can solve technical problems related in the background technology.
The technical scheme of the invention is as follows:
a fault diagnosis method for a variable-speed bearing of a cross-domain data drive unsupervised field shared network comprises the following steps:
s1, constructing a rotor-rolling bearing coupled system dynamic model, acquiring simulation data samples of a bearing under different fault modes to serve as a source domain data set, and meanwhile, taking the simulation data samples as training samples of a source domain;
s2, collecting a non-label bearing experiment data sample at a time-varying rotating speed as a target domain data set, and dividing a target domain training sample and a test sample;
s3, measuring the difference between the source domain data set and the target domain data set by adopting a maximum mean difference evaluation index based on cauchy nucleus induction, and improving the calculation efficiency by adopting an unbiased estimation technology;
s4, balancing the importance degree of the maximum mean difference loss and the classification cross entropy loss induced by the Kexi nucleus by adopting a balance factor with adjustable segments, and constructing a total loss function;
s5, constructing an unsupervised field shared network based on the maximum mean difference evaluation index induced by Kernel and the sectionally adjustable balance factor, inputting training samples in a source field data set and a target field data set into the unsupervised field shared network simultaneously to extract migratable features, training to obtain a bearing migration fault diagnosis model, and detecting the fault classification effect of the bearing migration fault diagnosis model by using test samples in the target field data set;
and S6, identifying the bearing fault type under the time-varying speed working condition by adopting the trained bearing migration fault diagnosis model.
As a preferable improvement of the present invention, step S3 specifically includes the following steps:
s31, taking a Cauchy kernel as a nonlinear mapping function kernel function of the maximum mean difference, and mapping a source domain data set and a target domain data set to a Hilbert high-dimensional inner product space of a regeneration kernel to measure the difference of two-domain distribution;
wherein, the first and the second end of the pipe are connected with each other,
is a source domain data set comprising n 1 Source domain samples
Is a target domain data set comprising n 2 Individual target domain samples
Is based on the maximum mean difference evaluation index of cauchy nucleus induction, H is the Hilbert high-dimensional inner product space of the regenerated nucleus, sigma C Is the width of the nucleus;
s32, under the precondition that the sample numbers of the source domain data set and the target domain data set are equal, reducing the calculation complexity of the Cauchy' S nucleus induced maximum mean difference evaluation index by adopting an unbiased estimation technology;
as a preferable improvement of the present invention, step S4 specifically includes the following steps:
s41, adopting a balance factor with adjustable segmentation, wherein the value of the balance factor is dynamically changed along with the increase of the iteration times;
wherein λ is C The method comprises the steps of (1) obtaining a balance factor which is adjustable in a segmented mode, wherein epoch is the current iteration number; max-epoch is a preset maximum iteration number; mid-epoch is a preset hyper-parameter;
step S42, introducing a balance factor with adjustable segmentation into a model overall loss function to balance the importance degree of the maximum mean difference loss and the classification cross entropy loss induced by the Kensive nucleus:
wherein, L represents the overall loss function,
an output value representing the i-th sample passing through the Softmax classification layer,
the label representing the ith sample.
As a preferred improvement of the present invention, in step S5, the classification layer activation function of the unsupervised domain shared network is Softmax, and the pooling layer activation function is ReLu.
The variable-speed bearing fault diagnosis method of the cross-domain data driving shared network in the unsupervised field has the following beneficial effects:
1. the difference between the source domain data set and the target domain data set is measured by using a maximum mean difference evaluation index based on Cauchy nucleus induction, so that the robustness of the domain self-adaption process is improved;
2. the computing process of the maximum mean difference evaluation index based on cauchy nucleus induction is improved by adopting an unbiased estimation technology, the computing efficiency is improved, and a large amount of computing cost is saved;
3. balancing the importance degree of the maximum mean difference loss and the classification cross entropy loss induced by the Kexi nucleus by adopting a sectional adjustable balance factor, constructing a total loss function, and accurately extracting the migratable characteristic between a source domain and a target domain by using an iterative training model;
4. the method solves the problems that the existing characteristic migration diagnosis method based on the maximum mean difference is high in calculation complexity and difficult to effectively realize optimal convergence, and the existing migration method is low in diagnosis precision when a fault migration task is carried out at a variable speed from a stable speed to an intermediate speed, can realize higher diagnosis precision when the fault migration task is carried out at the variable speed from the stable speed to the intermediate speed, and realizes reliable bearing fault migration diagnosis.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without inventive efforts, wherein:
FIG. 1 is a flow chart of a variable speed bearing fault diagnosis method of a cross-domain data driven unsupervised domain shared network of the present invention;
FIG. 2 is a diagram of a model of the rotor-rolling bearing coupling system dynamics of the present invention;
FIG. 3 is a graph of the time domain signals of the original vibration acceleration for various fault conditions of the bearing of the target domain of the present invention;
FIG. 4 is a diagram of vibration acceleration time domain signals for various fault conditions of the target domain of the present invention, with the various fault conditions being de-averaged and normalized;
FIG. 5 is a multi-classification confusion matrix diagram of the bearing fault of the fault diagnosis method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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.
In addition, the technical solutions in the embodiments of the present invention may be combined with each other, but it must be based on the realization of those skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination of technical solutions should not be considered to exist, and is not within the protection scope of the present invention.
Referring to fig. 1, the present invention provides a fault diagnosis method for a variable speed bearing of a cross-domain data-driven unsupervised domain shared network, which comprises the following steps:
s1, constructing a rotor-rolling bearing coupled system dynamic model, acquiring simulation data samples of a bearing under different fault modes to serve as a source domain data set, and meanwhile, taking the simulation data samples as training samples of a source domain;
s2, collecting a non-label bearing experiment data sample at a time-varying rotating speed as a target domain data set, and dividing a target domain training sample and a test sample;
s3, measuring the difference between the source domain data set and the target domain data set by adopting a maximum mean difference evaluation index based on cauchy nucleus induction, and improving the calculation efficiency by adopting an unbiased estimation technology, wherein the method specifically comprises the following steps:
s31, taking the Cauchy kernel as a nonlinear mapping function kernel function of the maximum mean difference, and mapping a source domain data set and a target domain data set to a Hilbert high-dimensional inner product space of a regeneration kernel to measure the difference of two-domain distribution;
wherein, the first and the second end of the pipe are connected with each other,
is a source domain data set containing n 1 Individual source domain samples
Is a target domain data set comprising n 2 Sample of individual target domain
Is the maximum mean difference evaluation index based on KexieNuclear induction, H is the Hilbert high-dimensional inner product space of the regenerated nucleus, and sigma C Is the cauchy nucleus width;
s32, under the precondition that the sample numbers of the source domain data set and the target domain data set are equal, reducing the calculation complexity of the Cauchy' S nucleus induced maximum mean difference evaluation index by adopting an unbiased estimation technology;
s4, balancing the importance degree of the Cauchy nucleus induced maximum mean difference loss and the classified cross entropy loss by using a piecewise adjustable balance factor, and constructing a total loss function, wherein the method specifically comprises the following steps:
s41, adopting a balance factor with adjustable segmentation, wherein the value of the balance factor is dynamically changed along with the increase of the iteration times;
wherein λ is C The method comprises the steps of (1) obtaining a balance factor which is adjustable in a segmented mode, wherein epoch is the current iteration number; max-epoch is a preset maximum iteration number; mid-epoch is a preset hyper-parameter;
step S42, introducing a piecewise adjustable balance factor into a model overall loss function to balance the importance degree of the Cauchy nucleus induced maximum mean difference loss and the classification cross entropy loss:
wherein, L represents the overall loss function,
an output value representing the i-th sample passing through the Softmax classification layer,
the label representing the ith sample.
S5, constructing an unsupervised field shared network based on the maximum mean difference evaluation index induced by Kernel and the sectionally adjustable balance factor, inputting training samples in a source field data set and a target field data set into the unsupervised field shared network simultaneously to extract migratable features, training to obtain a bearing migration fault diagnosis model, and detecting the fault classification effect of the bearing migration fault diagnosis model by using test samples in the target field data set;
specifically, the classification layer activation function of the unsupervised domain shared network is Softmax, and the pooling layer activation function is ReLu.
And S6, identifying the bearing fault type under the time-varying speed working condition by adopting the trained bearing migration fault diagnosis model.
The following describes in detail a variable speed bearing fault diagnosis method of a cross-domain data-driven unsupervised domain shared network according to the present invention in specific embodiment 1.
Example 1
In example 1, a rotor-bearing system simulation model is used to simulate an analog signal to obtain a source domain data set, all of which are used as source domain training samples, the model of a test bearing is JIS6306, an outer ring of the bearing is fixed on a bearing seat, and an inner ring of the bearing is fixed on a rotating shaft (see fig. 2). The stiffness and damping of the main components of the simulation model are assumed to be constant. The rotational frequency of the rotor was fixed at 1909rpm. The bearing fault state comprises four different bearing fault states of a normal state, an inner ring fault, a rolling fault and an outer ring fault.
And installing a vibration sensor on a QPZZ-II rotary mechanical vibration analysis and fault diagnosis test platform system to acquire data to construct a target domain data set, and dividing a training sample and a test sample of a target domain. The failure was set as a square groove with a wire cut in the bearing surface. The motor speed was ramped up from 640rpm to 1500rpm and then ramped down to 640rpm to collect acceleration data at a time varying speed. The target domain data set contains the same four health states as the source domain. The sampling frequency was 25.6kHz.
Constructing a bearing fault diagnosis model of an unsupervised field shared network based on Kexi nucleus induction, wherein the hyper-parameters are set as follows: width σ of cauchy nucleus C The method comprises the following steps that =1, the batch size is 28, the weight attenuation coefficient lambda =0.002, the learning rate eta =0.001, the iteration time t =100, the maximum iteration time is max-epoch =1000, and a preset hyper-parameter mid-epoch =50, wherein table 1 shows the shared network structure and parameters in the unsupervised field used by the invention. Table 2 shows the comparison results of the method of the present invention with intelligent identification methods such as migration component analysis, deep convolutional neural network, domain adaptive neural network, and joint distribution adaptation after ten independent operations.
The input of the deep convolution neural network, the domain adaptive neural network and the joint distribution adaptation is the same as that of the method of the invention, namely the original time domain vibration signal, and the input of the migration component analyzer is a frequency domain amplitude spectrum signal converted from the time domain vibration signal. Specifically, when the recognition result of one time in 10 times of operation of the method is analyzed, the overall classification accuracy of intelligent fault recognition is 80.03%, and a multi-classification confusion matrix is shown in fig. 5. As can be seen from Table 2 and FIG. 5, the average identification accuracy is highest by using the method provided by the invention, and 4 fault states of the bearing can be effectively distinguished based on the original time domain vibration signal.
TABLE 1 shared network architecture and parameters in unsupervised domains
TABLE 2 comparison of predicted results
Referring to fig. 1, the present invention can be mainly divided into three parts. The first part is that simulation time domain vibration signals of a dynamic model of a rotor-rolling bearing coupling system in different health states are collected as a source domain, original time domain vibration signals of a test bearing of machine equipment are collected as a target domain, and training samples and test samples which are divided into the source domain and the target domain respectively after mean value removing and normalization processing are carried out on the training samples and the test samples; the second part is to design a maximum mean difference measurement index based on cauchy nucleus induction, reduce the calculation complexity of the designed measurement index by adopting an unbiased estimation technology, and finally design the maximum mean difference measurement index based on cauchy nucleus induction of unbiased estimation; the third part is to design a balance factor which can be adjusted in a segmented mode, and finally, a total loss function is constructed; and the fourth part is a bearing fault diagnosis model which is trained by target domain test sample inspection and based on a Cauchy nucleus induced unsupervised domain shared network.
Referring to fig. 2, a dynamic model diagram of a rotor-rolling bearing coupled system is shown, wherein two ends of a rotor are supported by 2 identical rolling bearings in the coupled model, and the motions of all parts of a casing vibration rotor system are interacted and mutually coupled.
Referring to fig. 3, the original time domain vibration signals of various fault states of the bearing are tested experimentally, wherein the original time domain vibration signals include four different bearing fault states, namely a normal state, an inner ring fault, a rolling element fault and an outer ring fault. The sampling frequency is 25.6kHz, the sampling time is 16s, the number of sampling points is 970400, the abscissa in the figure represents time in s; the ordinate represents the amplitude in m/s 2 。
Referring to fig. 4, the time-domain vibration signals of various fault states of the tested bearing after mean removal and normalization are represented by the abscissa in the graph, the unit is s, the ordinate is dimensionless, the range is-1 to 1, 265700-368700 points are selected to divide training samples and testing samples, 1024 points are arranged on each sample, the two samples are not overlapped, and the total number is 400 samples, wherein 100 training samples are arranged in each state.
Referring to fig. 5, the multi-classification confusion matrix chart for test identification according to the method of the present invention has the abscissa as the predicted state label, the ordinate as the real state label, and the main diagonal number representing the identification accuracy of the category.
The variable-speed bearing fault diagnosis method of the cross-domain data driving shared network in the unsupervised field has the following beneficial effects:
1. the difference between the distribution of the source domain data set and the target domain data set is measured by using a maximum mean difference evaluation index based on cauchy nucleus induction, so that the robustness of the domain self-adaption process is improved;
2. the computing process of the maximum mean difference evaluation index based on Kexi nucleus induction is improved by adopting an unbiased estimation technology, so that the computing efficiency is improved, and a large amount of computing cost is saved;
3. balancing the importance degree of the maximum mean difference loss and the classification cross entropy loss induced by the Kexi nucleus by adopting a sectional adjustable balance factor, constructing a total loss function, and accurately extracting the migratable characteristic between a source domain and a target domain by using an iterative training model;
4. the method solves the problems that the existing characteristic migration diagnosis method based on the maximum mean difference is high in calculation complexity and difficult to effectively realize optimal convergence, and the existing migration method is low in diagnosis precision when a fault migration task is carried out at a variable speed from a stable speed to a variable speed, can realize higher diagnosis precision when the fault migration task is carried out at the variable speed from the stable speed to the variable speed, and realizes reliable bearing fault migration diagnosis.
While embodiments of the invention have been disclosed above, it is not limited to the applications set forth in the specification and the embodiments, which are fully applicable to various fields of endeavor for which the invention pertains, and further modifications may readily be made by those skilled in the art, it being understood that the invention is not limited to the details shown and described herein without departing from the general concept defined by the appended claims and their equivalents.
Claims (4)
1. A fault diagnosis method for a variable-speed bearing of a cross-domain data-driven unsupervised field shared network is characterized by comprising the following steps:
s1, constructing a rotor-rolling bearing coupled system dynamic model, acquiring simulation data samples of a bearing under different fault modes to serve as a source domain data set, and meanwhile, taking the simulation data samples as training samples of a source domain;
s2, collecting a non-label bearing experiment data sample at a time-varying rotation speed as a target domain data set, and dividing a target domain training sample and a test sample;
s3, measuring the difference between the source domain data set and the target domain data set by adopting a maximum mean difference evaluation index based on cauchy nucleus induction, and improving the calculation efficiency by adopting an unbiased estimation technology;
s4, balancing the importance degree of the maximum mean difference loss and the classification cross entropy loss induced by the Cauchy nucleus by adopting a subsection adjustable balance factor, and constructing a total loss function;
s5, constructing an unsupervised field shared network based on the maximum mean difference evaluation index induced by Kernel and the sectionally adjustable balance factor, inputting training samples in a source field data set and a target field data set into the unsupervised field shared network simultaneously to extract migratable features, training to obtain a bearing migration fault diagnosis model, and detecting the fault classification effect of the bearing migration fault diagnosis model by using test samples in the target field data set;
and S6, identifying the bearing fault type under the time-varying speed working condition by adopting the trained bearing migration fault diagnosis model.
2. The method for diagnosing the fault of the variable-speed bearing of the cross-domain data-driven unsupervised domain shared network according to claim 1, is characterized in that: the step S3 specifically comprises the following steps:
s31, taking the Cauchy kernel as a nonlinear mapping function kernel function of the maximum mean difference, and mapping a source domain data set and a target domain data set to a Hilbert high-dimensional inner product space of a regeneration kernel to measure the difference of two-domain distribution;
wherein, the first and the second end of the pipe are connected with each other,
is a source domain data set containing n 1 Individual source domain samples
Is a target domain data set, comprising n 2 Individual target domain samples
Is the maximum mean difference evaluation index based on KexieNuclear induction, and H is the Hilbert high-dimensional index of regenerated nucleusInner product space, σ C Is the cauchy nucleus width;
s32, under the precondition that the sample numbers of the source domain data set and the target domain data set are equal, reducing the calculation complexity of the Cauchy' S nucleus induced maximum mean difference evaluation index by adopting an unbiased estimation technology;
3. the method for diagnosing the fault of the variable-speed bearing of the cross-domain data drive unsupervised domain shared network according to claim 1, characterized in that: the step S4 specifically includes the following steps:
s41, adopting a balance factor with adjustable segmentation, wherein the value of the balance factor is dynamically changed along with the increase of the iteration times;
wherein λ is C The method comprises the steps of (1) obtaining a balance factor which is adjustable in a segmented mode, wherein epoch is the current iteration number; max-epoch is a preset maximum iteration number; mid-epoch is a preset hyper-parameter;
step S42, introducing a piecewise adjustable balance factor into a model overall loss function to balance the importance degree of the Cauchy nucleus induced maximum mean difference loss and the classification cross entropy loss:
wherein, L represents the overall loss function,
represents the output value of the ith sample passing through the Softmax classification layer,
the label representing the ith sample.
4. The method for diagnosing the fault of the variable-speed bearing of the cross-domain data drive unsupervised domain shared network according to claim 1, characterized in that: in step S5, the classification layer activation function of the unsupervised domain shared network is Softmax, and the pooling layer activation function is ReLu.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210382997.6A CN114755017B (en) | 2022-04-13 | 2022-04-13 | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210382997.6A CN114755017B (en) | 2022-04-13 | 2022-04-13 | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114755017A CN114755017A (en) | 2022-07-15 |
| CN114755017B true CN114755017B (en) | 2022-12-20 |
Family
ID=82330042
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210382997.6A Active CN114755017B (en) | 2022-04-13 | 2022-04-13 | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114755017B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115510926B (en) * | 2022-11-23 | 2023-04-18 | 武汉理工大学 | Cross-machine type diesel engine combustion chamber fault diagnosis method and system |
| CN116702304B (en) * | 2023-08-08 | 2023-10-20 | 中建五局第三建设有限公司 | Method and device for grouping foundation pit design schemes based on unsupervised learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103018043A (en) * | 2012-11-16 | 2013-04-03 | 东南大学 | Fault diagnosis method of variable-speed bearing |
| CN105608450A (en) * | 2016-03-01 | 2016-05-25 | 天津中科智能识别产业技术研究院有限公司 | Heterogeneous face identification method based on deep convolutional neural network |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN109902393A (en) * | 2019-03-01 | 2019-06-18 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under a kind of variable working condition based on further feature and transfer learning |
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN111626345A (en) * | 2020-05-15 | 2020-09-04 | 北京航空航天大学 | Multi-stage deep convolution transfer learning fault diagnosis method between different bearing devices |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2440792A1 (en) * | 2002-09-27 | 2004-03-27 | Mechworks Systems Inc. | A method and system for online condition monitoring of multistage rotary machinery |
| US8400472B2 (en) * | 2009-02-25 | 2013-03-19 | Technion Research & Development Foundation Limited | Method and system of geometric deformation |
| CN110285969B (en) * | 2019-07-10 | 2020-05-26 | 西安交通大学 | Rolling bearing fault migration diagnosis method with polynomial nuclear implantation characteristic distribution adaptation |
-
2022
- 2022-04-13 CN CN202210382997.6A patent/CN114755017B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103018043A (en) * | 2012-11-16 | 2013-04-03 | 东南大学 | Fault diagnosis method of variable-speed bearing |
| CN105608450A (en) * | 2016-03-01 | 2016-05-25 | 天津中科智能识别产业技术研究院有限公司 | Heterogeneous face identification method based on deep convolutional neural network |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN109902393A (en) * | 2019-03-01 | 2019-06-18 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under a kind of variable working condition based on further feature and transfer learning |
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN111626345A (en) * | 2020-05-15 | 2020-09-04 | 北京航空航天大学 | Multi-stage deep convolution transfer learning fault diagnosis method between different bearing devices |
Non-Patent Citations (1)
| Title |
|---|
| "面向轴承早期故障检测的多尺度残差注意力深度领域适配模型";毛文涛;《计算机应用》;20200430;第40卷(第10期);第2890-2898页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114755017A (en) | 2022-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN114755017B (en) | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network | |
| CN109186973B (en) | Mechanical fault diagnosis method of unsupervised deep learning network | |
| Yang et al. | Bearing fault automatic classification based on deep learning | |
| Kumar et al. | Tacho-less sparse CNN to detect defects in rotor-bearing systems at varying speed | |
| Castejón et al. | Automated diagnosis of rolling bearings using MRA and neural networks | |
| CN110132554B (en) | Rotary machine fault diagnosis method based on deep Laplace self-coding | |
| CN111160167A (en) | Spindle fault classification and identification method based on S-transform deep convolutional neural network | |
| Kaya et al. | Classification of bearing vibration speeds under 1D-LBP based on eight local directional filters | |
| Wang et al. | Conditionsensenet: A deep interpolatory convnet for bearing intelligent diagnosis under variational working conditions | |
| CN112765890B (en) | Dynamic domain adaptive network-based multi-working-condition rotating machine residual life prediction method | |
| CN111238815B (en) | Bearing fault identification method based on data enhancement under sample imbalance | |
| Pan et al. | A deep learning network via shunt-wound restricted Boltzmann machines using raw data for fault detection | |
| CN113639999A (en) | Fault diagnosis method, system, medium and equipment for rolling bearing under variable rotating speed | |
| CN113569990B (en) | Strong noise interference environment-oriented performance equipment fault diagnosis model construction method | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| CN116793682A (en) | Bearing fault diagnosis method based on iCORAL-MMD and anti-migration learning | |
| Ribeiro Junior et al. | Fault detection and diagnosis in electric motors using convolution neural network and short-time fourier transform | |
| Wang et al. | A quantitative intelligent diagnosis method for early weak faults of aviation high-speed bearings | |
| CN111076934A (en) | Method for diagnosing potential fault of bearing based on S transformation | |
| Qiao et al. | Fault diagnosis method of rolling bearings based on VMD and MDSVM | |
| CN114048787B (en) | Method and system for intelligently diagnosing bearing fault in real time based on Attention CNN model | |
| CN115876473A (en) | Bearing fault diagnosis method based on PWVD and DenseNet | |
| Li et al. | Transformer-based meta learning method for bearing fault identification under multiple small sample conditions | |
| Jianhui et al. | New fault diagnosis method for rolling bearing based on PCA |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |