CN116026593A - Cross-working-condition rolling bearing fault targeted migration diagnosis method and system - Google Patents

Cross-working-condition rolling bearing fault targeted migration diagnosis method and system Download PDF

Info

Publication number
CN116026593A
CN116026593A CN202211631915.3A CN202211631915A CN116026593A CN 116026593 A CN116026593 A CN 116026593A CN 202211631915 A CN202211631915 A CN 202211631915A CN 116026593 A CN116026593 A CN 116026593A
Authority
CN
China
Prior art keywords
rolling bearing
cross
loss
fault
feature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211631915.3A
Other languages
Chinese (zh)
Inventor
张法业
刘福政
姜明顺
张雷
隋青美
贾磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University
Original Assignee
Shandong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shandong University filed Critical Shandong University
Priority to CN202211631915.3A priority Critical patent/CN116026593A/en
Publication of CN116026593A publication Critical patent/CN116026593A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis method and a cross-working-condition rolling bearing fault targeted migration diagnosis system, which solve the problems that the traditional rolling bearing fault diagnosis algorithm is difficult to extract network deep characteristic information in a source domain and a target domain and cannot realize effective cross-domain fault diagnosis. The invention utilizes the characteristic encoder to accurately extract the high-dimensional mapping characteristic of the signal from the input rolling bearing vibration signal; the features are further input into a graph construction layer, deep features of data are mined, and a multi-channel kernel graph rolling network is utilized to model an instance graph; with the difference and challenge based training to minimize the distance between the source and target domain distributions, the classifier uses the extracted domain invariant features to accomplish cross domain fault identification. Compared with other methods, the method can better extract deep features for cross-domain transmission under the cross-working condition of the rolling bearing, and greatly improves the diagnosis accuracy.

Description

Cross-working-condition rolling bearing fault targeted migration diagnosis method and system
Technical Field
The invention relates to the technical field of bearing fault diagnosis, in particular to a cross-working-condition rolling bearing fault targeted migration diagnosis method and system.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The rolling bearing is used as a key part of most rotary mechanical equipment, and is frequently failed due to the characteristics of severe working environment, working condition change, long overload running time and the like, so that the mechanical failure of the whole machine is brought and certain economic loss is caused. Therefore, the method has important significance in monitoring the running state of the rolling bearing under the cross-working condition.
In recent years, with the advent of mass data, a data driving method based on deep learning has been attracting attention. It differs from the traditional shallow network architecture but is realized by stacking multiple layers of nonlinear processing units, and provides an end-to-end solution, so many students have also conducted corresponding research. Jia et al propose a SAE-based deep neural network DNN for identifying faults in motors and gearboxes. Ding et al propose the use of convolutional neural networks to mine multi-scale features of energy fluctuations from wavelet packet energy images for fault diagnosis of spindle bearings. Chen et al propose an automatic feature learning neural network that takes as input the original vibration signal, and automatically extracts different frequency signal features from the original data using two CNNs with different kernel sizes, and then uses LSTM to identify the fault type based on the learned features. The deep learning algorithm has good effect in the field of fault diagnosis of the rolling bearing under constant working conditions, but is difficult to extract inter-domain difference characteristics under complex cross-working conditions, so that the generalization performance of the model is reduced.
The migration learning is used as a method for reducing the cross-domain feature distribution difference, provides a new thought for establishing knowledge migration from source domain marked data to target domain unmarked data, and effectively solves the problem of inter-domain distribution difference under the cross-working condition. In the field of intelligent fault diagnosis, a number of migration learning methods have been proposed to solve the cross-domain diagnosis problem, which can be basically classified into instance-based, model-based and feature-based. Xiao et al use TrAdaBoost to adjust the weight factor of each training sample to enhance the diagnostic performance of the fault classifier. Wang et al propose a conditional MMD based on estimating pseudo tags to shorten the distribution distance of bearing fault diagnostics. By minimizing MMD loss, edge distribution and conditional distribution are aligned simultaneously in multiple layers. Han et al propose a deep antagonistic convolutional neural network DACNN that utilizes an antagonistic-based loss function to reduce inter-domain differences for improving the generalization performance of gearbox and motor fault diagnosis. Li et al take graph data with a topological structure as input, mine the data relationship more effectively, model a graph rolling network GCN with more powerful characteristic representation for mechanical fault diagnosis, and obtain excellent performance. It can be seen that the information of category labels, domain labels and data structures plays an important role in reducing the source domain and target domain differences and that they should be perfect and enhance each other.
However, the inventor discovers that the existing method only considers two kinds of information of a source domain and a target domain, does not integrate a data structure into a deep neural network model, has large data distribution difference under different working conditions, and causes low fault recognition accuracy and insufficient generalization performance.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis method and a cross-working-condition rolling bearing fault targeted migration diagnosis system, which can better extract deep features for cross-domain transmission under the cross-working-condition of the rolling bearing, and greatly improve diagnosis accuracy.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the first aspect of the invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis method.
A cross-working-condition rolling bearing fault targeted migration diagnosis method comprises the following steps:
acquiring a vibration signal of a rolling bearing;
extracting deep high-dimensional features in the vibration signals through a feature encoder, a graph construction layer and a multi-channel kernel graph convolution network;
obtaining a rolling bearing fault diagnosis result according to the extracted deep high-dimensional characteristics and the classifier;
the method comprises the steps of obtaining classification loss according to deep high-dimensional characteristics obtained by combining a classifier with a source domain training set, obtaining structural difference loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and a target domain training set, and obtaining countermeasures loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and the target domain training set and a countermeasures network;
and optimizing parameters of the feature extractor, the classifier and the discriminator through a back propagation algorithm according to the overall objective function by taking the sum of the classification loss, the product of the first parameter and the structural difference loss and the product of the second parameter and the counterattack loss as an overall loss function.
As an optional implementation manner of the first aspect of the present invention, the graph construction layer is configured to obtain an adjacency matrix, and includes:
obtaining a high-dimensional feature map, i.e., x=g (X), from the sample data according to the feature encoder network;
the extracted high-dimensional characteristic map is input into a linear layer and expressed as after Softmax
Figure BDA0004006128250000031
By performing matrix multiplication calculation between the features of the linear layer and its transpose, an adjacency matrix a is obtained,
Figure BDA0004006128250000032
through KNN algorithm, constructing edge relation:
Figure BDA0004006128250000033
as a further definition of the first aspect of the invention, a multi-channel kernel graph rolling network comprises:
Figure BDA0004006128250000034
Figure BDA0004006128250000041
wherein X represents the input, A represents the adjacency matrix,
Figure BDA0004006128250000042
represents a trainable weight matrix, G represents a multi-channel kernel graph convolution operation, ++>
Figure BDA0004006128250000043
Represents the kth i High-dimensional characterization of individual channels at layer L, [. Cndot.]Representing feature stitching, H represents the output features after passing through the multi-channel kernel graph convolution network.
As an optional implementation manner of the first aspect of the present invention, cross entropy loss L C Comprising:
Figure BDA0004006128250000044
wherein ,
Figure BDA0004006128250000045
representing the prediction result of the classifier, E representing the mathematical expectation,/->
Figure BDA0004006128250000046
For the source domain sample, ++>
Figure BDA0004006128250000047
For its tag.
As an optional implementation manner of the first aspect of the present invention, the structural difference loss L s Comprising:
Figure BDA0004006128250000048
wherein ,
Figure BDA0004006128250000049
and
Figure BDA00040061282500000410
Respectively representing the feature mapping of the ith source domain sample and the jth target domain sample after the ith source domain sample passes through a feature extractor, phi represents the nonlinear feature mapping, omega k Representing distance metrics of embedded extracted features into regenerated kernel Hilbert space RKHS, employing convex combinations k of m kernels u To effectively estimate the mapping:
Figure BDA00040061282500000411
wherein ,αu Is a weighting parameter of different kernels and +.>
Figure BDA00040061282500000412
E represents a mathematical expectation value, +.>
Figure BDA00040061282500000413
For the source domain sample, ++>
Figure BDA00040061282500000414
Is a target domain sample.
As an optional implementation manner of the first aspect of the invention, the countering loss L AD Comprising:
Figure BDA00040061282500000415
wherein D (·) is the feature output after passing through the discriminator,
Figure BDA00040061282500000416
and
Figure BDA00040061282500000417
Respectively representing the feature mapping of the ith source domain sample and the jth target domain sample after the feature extractor, E represents the mathematical expectation value,>
Figure BDA00040061282500000418
for the source domain sample, ++>
Figure BDA00040061282500000419
Is a target domain sample.
As an optional implementation manner of the first aspect of the present invention, optimizing parameters of the feature extractor, the classifier and the arbiter by a back propagation algorithm includes:
Figure BDA0004006128250000051
wherein ,
Figure BDA0004006128250000052
represents partial differential operator, eta represents learning rate, theta F Parameters representative of feature extractor, θ C Representing the parameters of the classifier, θ D Representing parameters of the arbiter, L C For cross entropy loss, L s For structural differential loss, L AD To combat losses.
The second aspect of the invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis system.
A cross-regime rolling bearing fault targeted migration diagnostic system comprising:
a data acquisition module configured to: acquiring a vibration signal of a rolling bearing;
a feature extraction module configured to: extracting deep high-dimensional features in the vibration signals through a feature encoder, a graph construction layer and a multi-channel kernel graph convolution network;
a fault diagnosis module configured to: obtaining a rolling bearing fault diagnosis result according to the extracted deep high-dimensional characteristics and the classifier;
the method comprises the steps of obtaining classification loss according to deep high-dimensional characteristics obtained by combining a classifier with a source domain training set, obtaining structural difference loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and a target domain training set, and obtaining countermeasures loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and the target domain training set and a countermeasures network;
and optimizing parameters of the feature extractor, the classifier and the discriminator through a back propagation algorithm according to the overall objective function by taking the sum of the classification loss, the product of the first parameter and the structural difference loss and the product of the second parameter and the counterattack loss as an overall loss function.
A third aspect of the present invention is a computer-readable storage medium having stored thereon a program which, when executed by a processor, implements the steps in the cross-regime rolling bearing fault-targeted migration diagnostic method according to the first aspect of the present invention.
A fourth aspect of the present invention provides an electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, the processor implementing the steps in the cross-condition rolling bearing fault targeted migration diagnostic method according to the first aspect of the present invention when the program is executed.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention adopts the feature encoder to adaptively extract signal features from the input signal, and utilizes the graph construction layer to acquire the data structure in the captured features of the feature encoder, so as to construct an example graph, and applies the multi-channel kernel graph convolution network to model the feature encoder, so as to further excavate the high-dimensional features of the signal, and solve the problem of difficult extraction of deep features of a source domain and a target domain under the cross-working condition.
2. According to the invention, aiming at the problem of large data difference between a rolling bearing vibration signal source domain and a target domain under a cross-working condition, a loss function based on the data structure difference and a loss function based on antagonism are adopted to jointly reduce the inter-domain difference, and meanwhile, a classifier uses extracted domain invariant features to complete cross-domain fault identification.
3. The invention solves the problems of difficulty in obtaining the marking data in the industrial scene, large data distribution difference under different working conditions, low fault recognition accuracy and insufficient generalization performance, and can extract deep high-dimensional characteristics to reduce inter-domain difference and obtain better diagnosis performance.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flow chart of a method for diagnosing fault-targeted migration of a rolling bearing under a cross-working condition provided in embodiment 1 of the present invention;
fig. 2 is a KNN diagram construction flow based on sample data features provided in embodiment 1 of the present invention;
fig. 3 is a schematic diagram of an HFZZ rotary machine fault simulation platform provided in embodiment 1 of the present invention;
fig. 4 is an experimental result under different cross-domain diagnostic tasks provided in example 1 of the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Example 1:
as shown in fig. 1, embodiment 1 of the present invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis method, which includes the following steps:
s1: signal acquisition and data set partitioning
Collecting data under different working conditions through a mechanical fault test platform, wherein the data comprises a labeled source domain data set
Figure BDA0004006128250000071
And the target domain data set without tag +.>
Figure BDA0004006128250000072
And the sample data is further divided into training data and test data; />
S2: network model construction and high-dimensional feature extraction
Extracting deep high-dimensional characteristics of sample data through a feature encoder, a graph construction layer and a multichannel kernel graph convolution network construction model;
s3: overall objective function and model optimization
Reducing the classification error by classifying the loss function, jointly reducing the inter-domain difference by using a transfer learning method based on the loss function of the data structure difference and the inter-domain difference based on antagonism, and optimizing the parameters in the overall objective function by using a back propagation algorithm;
s4: model test and diagnostic result output
In the test stage, the target domain test data can adopt the optimized network in the S2 to extract deep features, and the classifier can be directly used for fault classification.
S2, constructing a network model and extracting high-dimensional characteristics, wherein the constructing process comprises the following steps:
(1) Feature encoder
The system mainly comprises a generator G and a classifier C, wherein the generator G is used for encoding input data to obtain a high-dimensional characteristic representation, and the classifier C is used for finally classifying source and target tasks;
in the embodiment, a one-dimensional convolutional neural network CNN is constructed for feature extraction and fault classification, a hidden layer of the CNN is used for nonlinear feature mapping corresponding to G to obtain depth feature representation, and a Softmax of an output layer of the CNN is used as a classifier C to obtain probability output of each fault type;
the high-dimensional property representation in CNN can be expressed as:
G(x)=G L (G L-1 (…G 2 (G 1 (x,w 1 )))) (1)
where x is the input, G is the feature map of CNN, w i Is the learning weight of the ith layer in the CNN, L is the number of layers of the CNN.
(2) Graph construction layer
After the samples pass through the feature encoder, a graph construction layer can be generated by using a KNN algorithm, the adjacency relation between all outputs is defined, the local characteristics among the samples are reflected, and the construction flow is shown in fig. 2.
The configuration of the edges may be represented by the following formula:
A ij =KNN(k,L iji ),A ij ∈A (2)
wherein ,Ωi ={L i1 ,L i2 ,…,L in The node h i And a distance set with all other nodes, wherein k is a super parameter.
The graph construction layer is used for obtaining an adjacent matrix A and obtaining an example graph from the minimum input matrix, and mainly comprises the following steps:
(A) Obtaining a high-dimensional feature map, i.e., x=g (X), from the sample data using the feature encoder network of (1);
(B) The extracted high-dimensional feature map is input into the linear layer and after Softmax can be expressed as:
Figure BDA0004006128250000091
(C) By matrix multiplication between the features of the linear layer and its transpose, the adjacency matrix A is obtained, i.e
Figure BDA0004006128250000092
(D) By KNN algorithm, edge relations are constructed, i.e
Figure BDA0004006128250000093
Thus, the construction of the adjacency matrix is summarized as follows: />
Figure BDA0004006128250000094
Wherein A is the adjacency matrix of the construction,
Figure BDA0004006128250000095
is the output of the high-dimensional features after passing through the linear layer and Softmax, and norm (·) represents the normalization function, ++>
Figure BDA0004006128250000096
Is a sparse adjacency matrix, KNN (·) returns an index of the first k maxima of adjacency matrix a in the row direction.
(3) Multi-channel kernel graph convolution network
A graph convolutional network is a network that can be represented by the geometry and structure of data, can provide more information representation, and performs network learning based on the connection relationships between nodes. In general, g= (a, X) is simplified, where a is an adjacency matrix, and may reflect a connection relationship between nodes, and X is a feature of a node. L=i N -D -1/2 AD -1/2 Is a Laplacian matrix, where D can be obtained from an adjacency matrix, i.e
Figure BDA0004006128250000097
I N Is an identity matrix, and a filter g is used for graph convolution θ =diag (θ) to smooth the input signal, which can be expressed as:
g θ * G x=Ug θ U T x (4)
where θ is a parameter that can be learned G Is a graph convolution operation, U is a eigenvector of the Laplacian matrix, U T x represents the fourier transform of the signal on the plot.
The graph convolution operation defined in equation (4) is not localized and has a high computational load. The convolution kernel is constrained to a polynomial expansion using equation (5):
Figure BDA0004006128250000101
where K is the order of the polynomial,
Figure BDA0004006128250000102
lambda represents the eigenvalue of the Laplacian matrix.
A multi-core graph convolution network is employed to obtain a representation of features of a broader receptive field, whose convolution operations can be defined as:
Figure BDA0004006128250000103
wherein ,
Figure BDA0004006128250000104
is a parameter that can be learned, < >>
Figure BDA0004006128250000105
Represents the kth i The high-dimensional features of the convolution kernels are represented.
Assume that the number of multi-channel convolution kernels in the network is k i I.e. the number of channels, the number of network layers is l= (1, 2, …, L), i.e. the network depth, and the multi-core graph convolution network with different receptive fields is used to respectively perform graph convolution on different channels to construct a multi-channel core graph convolution network:
Figure BDA0004006128250000106
Figure BDA0004006128250000107
wherein X represents the input, A represents the adjacency matrix,
Figure BDA0004006128250000108
represents a trainable weight matrix, G represents a multi-channel kernel graph convolution operation, ++>
Figure BDA0004006128250000109
Represents the kth i High-dimensional characterization of individual channels at layer L, [. Cndot.]Representing feature stitching, H represents the output features after passing through the multi-channel kernel graph convolution network.
S3, a construction process of overall objective function and model optimization comprises the following steps:
in order to fully utilize the characteristics of data and the depth network structure to reduce the characteristic difference of the data of the source domain and the target domain after mapping, three parts of loss functions, namely classification loss, structure difference loss and countermeasures loss, are adopted to form an overall objective function.
(1) Classification loss
The classification loss between the real label and the predictive label is estimated by a cross entropy loss, which can be defined as:
Figure BDA0004006128250000111
wherein ,
Figure BDA0004006128250000112
representing the prediction result of the tag classifier, L C Representing a cross entropy loss function. E represents a mathematical expectation.
(2) Structural differential loss
The following measurement mode is adopted to reduce the structural difference loss between the source domain and the target domain, and the method is defined as:
Figure BDA0004006128250000113
wherein ,
Figure BDA0004006128250000114
and
Figure BDA0004006128250000115
Representing the feature map of the ith source domain sample and the jth target domain sample after passing through a feature extractor (referred to herein as a feature encoder and a multi-channel kernel graph convolution network), respectively. Phi represents nonlinear feature mapping, omega k Representing a distance measure of the embedded extracted features into the regenerated kernel Hilbert space RKHS, in this embodiment, a convex combination k of m kernels is used u To effectively estimate the mapping:
Figure BDA0004006128250000116
wherein ,αu Is a weighted parameter of different cores, and
Figure BDA0004006128250000117
in the present invention, the ratio of the two to the total of [0,1 ]]The average allocation is made.
(3) Countering losses
The problems of domain ramping and domain shifting are handled in an countermeasure training way, whether the extracted features come from the source domain or the target domain is judged by the domain discriminant, and the feature extractor is trained to deceptively discriminant. When both reach the minmax gaming equilibrium, domain invariant properties can be obtained, employing the loss function in equation (12) as the counterloss:
Figure BDA0004006128250000121
wherein D (·) is the feature output after passing through the discriminator, and the value is between 0 and 1, so that the source domain or the target domain can be distinguished.
(4) Overall objective function
The overall objective function may be defined as:
Figure BDA0004006128250000122
wherein τ and
Figure BDA0004006128250000123
is an adjustable parameter.
(5) Parameter optimization
For the overall objective function in equation (13), the parameters of the individual parts are optimized by the back propagation algorithm as follows:
Figure BDA0004006128250000124
Figure BDA0004006128250000125
Figure BDA0004006128250000126
wherein ,
Figure BDA0004006128250000127
represents partial differential operator, eta represents learning rate, theta F Parameters representative of feature extractor, θ C Representing the parameters of the classifier, θ D Representing parameters of the arbiter.
The detailed network structure of the rolling bearing fault targeted migration diagnosis method under the cross-working condition provided by the embodiment is shown in table 1, wherein C represents the fault type:
table 1: detailed network structure table of rolling bearing fault targeted migration diagnosis method under cross-working condition
Figure BDA0004006128250000131
The present embodiment provides the following specific cases:
and building an HFZZ rotating machinery fault simulation platform, wherein the platform is composed of a motor, a control system, a radial loading device, an acceleration sensor and the like as shown in figure 3. The original data are collected by an acceleration sensor, and the collection frequency is 12.8kHz. The bearing operating states under different working conditions, namely working condition A (1750 rpm), working condition B (2000 rpm) and working condition C (2250 rmp), are simulated by adopting 3 different rotating speeds. By machining, bearing faults were co-produced into 9 health conditions including Normal (NM), inner ring faults (IR, IRU), outer ring faults (OR), rolling body faults (BA), and various mixed faults, including single faults and compound faults, the detailed information of which is shown in table 2. In the experiment, 100 sets of data were collected from each health condition at each operating speed, each data containing 1024 data points. For a total of 900 samples (100 x 9 health conditions) were obtained for each condition, the dataset contained a total of 2700 samples (900 x 3 different speeds), as specified in table 3, and according to 6:4 to divide the training set and the test set.
Table 2: rolling bearing health status detailed description table
Figure BDA0004006128250000141
Table 3: rolling bearing migration task detailed description table
Figure BDA0004006128250000142
In order to examine the superiority of the proposed cross-condition rolling bearing fault targeted migration diagnostic method, several most advanced deep neural network algorithms, such as CNN (Baseline), MKMMD, JAN, DANN, CDAN, are used to compare cross-domain diagnostic tasks under cross-conditions. Ten experiments were performed in the fault diagnosis experiment, and the average value was taken as an experimental result as shown in table 4 and fig. 4. From the results, the method according to the embodiment achieves superior performance and improves the overall average accuracy. As source domain and target domain differences become larger and cross-domain tasks become more difficult, the proposed algorithm may still achieve performance improvement.
Table 4: experimental results table for rolling bearing data set
Figure BDA0004006128250000143
Figure BDA0004006128250000151
Example 2:
the embodiment 2 of the invention provides a cross-working-condition rolling bearing fault targeted migration diagnosis system, which comprises:
a data acquisition module configured to: acquiring a vibration signal of a rolling bearing;
a feature extraction module configured to: extracting deep high-dimensional features in the vibration signals through a feature encoder, a graph construction layer and a multi-channel kernel graph convolution network;
a fault diagnosis module configured to: obtaining a rolling bearing fault diagnosis result according to the extracted deep high-dimensional characteristics and the classifier;
the method comprises the steps of obtaining classification loss according to deep high-dimensional characteristics obtained by combining a classifier with a source domain training set, obtaining structural difference loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and a target domain training set, and obtaining countermeasures loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and the target domain training set and a countermeasures network;
and optimizing parameters of the feature extractor, the classifier and the discriminator through a back propagation algorithm according to the overall objective function by taking the sum of the classification loss, the product of the first parameter and the structural difference loss and the product of the second parameter and the counterattack loss as an overall loss function.
The working method of the system is the same as the cross-working-condition rolling bearing fault targeted migration diagnosis method provided in embodiment 1, and is not described here again.
Example 3:
embodiment 3 of the present invention provides a computer-readable storage medium having a program stored thereon, which when executed by a processor, implements the steps in the cross-operating mode rolling bearing fault targeted migration diagnostic method according to embodiment 1 of the present invention.
Example 4:
the embodiment 4 of the invention provides an electronic device, which comprises a memory, a processor and a program stored on the memory and capable of running on the processor, wherein the processor realizes the steps in the cross-working-condition rolling bearing fault targeted migration diagnosis method according to the embodiment 1 of the invention when executing the program.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A cross-working-condition rolling bearing fault targeted migration diagnosis method is characterized by comprising the following steps:
acquiring a vibration signal of a rolling bearing;
extracting deep high-dimensional features in the vibration signals through a feature encoder, a graph construction layer and a multi-channel kernel graph convolution network;
obtaining a rolling bearing fault diagnosis result according to the extracted deep high-dimensional characteristics and the classifier;
the method comprises the steps of obtaining classification loss according to deep high-dimensional characteristics obtained by combining a classifier with a source domain training set, obtaining structural difference loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and a target domain training set, and obtaining countermeasures loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and the target domain training set and a countermeasures network;
and optimizing parameters of the feature extractor, the classifier and the discriminator through a back propagation algorithm according to the overall objective function by taking the sum of the classification loss, the product of the first parameter and the structural difference loss and the product of the second parameter and the counterattack loss as an overall loss function.
2. The method for diagnosing the fault-targeted migration of the rolling bearing under the cross-working condition of claim 1, wherein the method comprises the following steps of,
the graph construction layer is used for acquiring an adjacency matrix and comprises the following steps:
obtaining a high-dimensional feature map, i.e., x=g (X), from the sample data according to the feature encoder network;
the extracted high-dimensional characteristic map is input into a linear layer and expressed as after Softmax
Figure FDA0004006128240000011
By performing matrix multiplication calculation between the features of the linear layer and its transpose, an adjacency matrix a is obtained,
Figure FDA0004006128240000012
by KNN algorithm, edge relations are constructed, i.e
Figure FDA0004006128240000013
3. A cross-operating mode rolling bearing fault targeted migration diagnostic method as claimed in claim 2, wherein,
a multi-channel kernel graph rolling network comprising:
Figure FDA0004006128240000021
Figure FDA0004006128240000022
wherein X represents the input, A represents the adjacency matrix,
Figure FDA0004006128240000023
represents a trainable weight matrix, G represents a multi-channel kernel graph convolution operation, ++>
Figure FDA0004006128240000024
Represents the kth i High-dimensional characterization of individual channels at layer L, [. Cndot.]Representing feature stitching, H represents the output features after passing through the multi-channel kernel graph convolution network.
4. The method for diagnosing the fault-targeted migration of the rolling bearing under the cross-working condition of claim 1, wherein the method comprises the following steps of,
cross entropy loss L C Comprising:
Figure FDA0004006128240000025
wherein ,
Figure FDA0004006128240000026
representing the prediction result of the classifier, E representing the mathematical expectation,/->
Figure FDA0004006128240000027
For the source domain sample, ++>
Figure FDA0004006128240000028
For its tag.
5. The method for diagnosing the fault-targeted migration of the rolling bearing under the cross-working condition of claim 1, wherein the method comprises the following steps of,
structural differential loss L s Comprising:
Figure FDA0004006128240000029
wherein ,
Figure FDA00040061282400000210
and
Figure FDA00040061282400000211
Respectively representing the feature mapping of the ith source domain sample and the jth target domain sample after the ith source domain sample passes through a feature extractor, phi represents the nonlinear feature mapping, omega k Representing distance metrics of embedded extracted features into regenerated kernel Hilbert space RKHS, employing convex combinations k of m kernels u To effectively estimate the mapping:
Figure FDA00040061282400000212
wherein ,αu Is a weighting parameter of different kernels and +.>
Figure FDA00040061282400000213
E represents a mathematical expectation value, +.>
Figure FDA0004006128240000031
For the source domain sample, ++>
Figure FDA0004006128240000032
Is a target domain sample.
6. The method for diagnosing the fault-targeted migration of the rolling bearing under the cross-working condition of claim 1, wherein the method comprises the following steps of,
countering loss L AD Comprising:
Figure FDA0004006128240000033
wherein D (·) is the feature output after passing through the discriminator,
Figure FDA0004006128240000034
and
Figure FDA0004006128240000035
Respectively representing the feature mapping of the ith source domain sample and the jth target domain sample after the feature extractor, E represents the mathematical expectation value,>
Figure FDA0004006128240000036
for the source domain sample, ++>
Figure FDA0004006128240000037
Is a target domain sample.
7. The method for diagnosing the fault-targeted migration of the rolling bearing under the cross-working condition of claim 1, wherein the method comprises the following steps of,
optimizing parameters of the feature extractor, classifier and arbiter by a back propagation algorithm, comprising:
Figure FDA0004006128240000038
wherein ,
Figure FDA0004006128240000039
represents partial differential operator, eta represents learning rate, theta F Parameters representative of feature extractor, θ C Representing the parameters of the classifier, θ D Representing parameters of the arbiter, L C For cross entropy loss, L s For structural differential loss, L AD To combat losses.
8. A cross-condition rolling bearing fault targeted migration diagnostic system, comprising:
a data acquisition module configured to: acquiring a vibration signal of a rolling bearing;
a feature extraction module configured to: extracting deep high-dimensional features in the vibration signals through a feature encoder, a graph construction layer and a multi-channel kernel graph convolution network;
a fault diagnosis module configured to: obtaining a rolling bearing fault diagnosis result according to the extracted deep high-dimensional characteristics and the classifier;
the method comprises the steps of obtaining classification loss according to deep high-dimensional characteristics obtained by combining a classifier with a source domain training set, obtaining structural difference loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and a target domain training set, and obtaining countermeasures loss according to the deep high-dimensional characteristics obtained by combining the source domain training set and the target domain training set and a countermeasures network;
and optimizing parameters of the feature extractor, the classifier and the discriminator through a back propagation algorithm according to the overall objective function by taking the sum of the classification loss, the product of the first parameter and the structural difference loss and the product of the second parameter and the counterattack loss as an overall loss function.
9. A computer readable storage medium having stored thereon a program, which when executed by a processor, implements the steps of the cross-regime rolling bearing fault targeted migration diagnostic method according to any one of claims 1 to 7.
10. An electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor performs the steps in the cross-regime rolling bearing fault targeted migration diagnostic method of any one of claims 1-7 when the program is executed.
CN202211631915.3A 2022-12-19 2022-12-19 Cross-working-condition rolling bearing fault targeted migration diagnosis method and system Pending CN116026593A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211631915.3A CN116026593A (en) 2022-12-19 2022-12-19 Cross-working-condition rolling bearing fault targeted migration diagnosis method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211631915.3A CN116026593A (en) 2022-12-19 2022-12-19 Cross-working-condition rolling bearing fault targeted migration diagnosis method and system

Publications (1)

Publication Number Publication Date
CN116026593A true CN116026593A (en) 2023-04-28

Family

ID=86073278

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211631915.3A Pending CN116026593A (en) 2022-12-19 2022-12-19 Cross-working-condition rolling bearing fault targeted migration diagnosis method and system

Country Status (1)

Country Link
CN (1) CN116026593A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116952554A (en) * 2023-07-05 2023-10-27 北京科技大学 Multi-sensor mechanical equipment fault diagnosis method and device based on graph rolling network
CN116977708A (en) * 2023-06-14 2023-10-31 北京建筑大学 Bearing intelligent diagnosis method and system based on self-adaptive aggregation visual view
CN117194983A (en) * 2023-09-08 2023-12-08 北京理工大学 Bearing fault diagnosis method based on progressive condition domain countermeasure network
CN117232577A (en) * 2023-09-18 2023-12-15 杭州奥克光电设备有限公司 Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116977708A (en) * 2023-06-14 2023-10-31 北京建筑大学 Bearing intelligent diagnosis method and system based on self-adaptive aggregation visual view
CN116977708B (en) * 2023-06-14 2024-04-12 北京建筑大学 Bearing intelligent diagnosis method and system based on self-adaptive aggregation visual view
CN116952554A (en) * 2023-07-05 2023-10-27 北京科技大学 Multi-sensor mechanical equipment fault diagnosis method and device based on graph rolling network
CN117194983A (en) * 2023-09-08 2023-12-08 北京理工大学 Bearing fault diagnosis method based on progressive condition domain countermeasure network
CN117194983B (en) * 2023-09-08 2024-04-19 北京理工大学 Bearing fault diagnosis method based on progressive condition domain countermeasure network
CN117232577A (en) * 2023-09-18 2023-12-15 杭州奥克光电设备有限公司 Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box
CN117232577B (en) * 2023-09-18 2024-04-05 杭州奥克光电设备有限公司 Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box

Similar Documents

Publication Publication Date Title
CN111898634B (en) Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption
CN109635677B (en) Compound fault diagnosis method and device based on multi-label classification convolutional neural network
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
CN116026593A (en) Cross-working-condition rolling bearing fault targeted migration diagnosis method and system
CN112101085B (en) Intelligent fault diagnosis method based on importance weighted domain antagonism self-adaptation
CN111709448A (en) Mechanical fault diagnosis method based on migration relation network
CN112926641A (en) Three-stage feature fusion rotating machine fault diagnosis method based on multi-modal data
Fang et al. ANS-net: anti-noise Siamese network for bearing fault diagnosis with a few data
CN114358123B (en) Generalized open set fault diagnosis method based on deep countermeasure migration network
CN114429152A (en) Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption
CN116894187A (en) Gear box fault diagnosis method based on deep migration learning
Graening et al. Shape mining: A holistic data mining approach for engineering design
Jiao et al. A mixed adversarial adaptation network for intelligent fault diagnosis
CN113239610A (en) Domain self-adaptive rolling bearing fault diagnosis method based on Wasserstein distance
Huang et al. Intelligent fault identification for industrial internet of things via prototype-guided partial domain adaptation with momentum weight
CN114648044B (en) Vibration signal diagnosis and analysis method based on EEMD and depth domain countermeasure network
Zhang et al. Convolutional neural network based two-layer transfer learning for bearing fault diagnosis
Sun et al. Intelligent fault diagnosis scheme for rolling bearing based on domain adaptation in one dimensional feature matching
CN118035448A (en) Method, device and medium for classifying paper fields in citation network based on pseudo tag depolarization
CN116150668B (en) Rotating equipment fault diagnosis method based on double-stage alignment partial migration network
Zou et al. Research on mechanical fault diagnosis based on MADS evidence fusion theory
CN117235490A (en) Fault self-adaptive diagnosis method integrating deep volume and self-attention network
Xiang et al. Spacecraft Intelligent Fault Diagnosis under Variable Working Conditions via Wasserstein Distance‐Based Deep Adversarial Transfer Learning
Yu et al. A Novel 1D‐CNN‐Based Diagnosis Method for a Rolling Bearing with Dual‐Sensor Vibration Data Fusion
CN115600134A (en) Bearing transfer learning fault diagnosis method based on domain dynamic impedance self-adaption

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