CN113780151B - Bearing fault diagnosis method and system based on bilinear feature fusion - Google Patents

Bearing fault diagnosis method and system based on bilinear feature fusion Download PDF

Info

Publication number
CN113780151B
CN113780151B CN202111043157.9A CN202111043157A CN113780151B CN 113780151 B CN113780151 B CN 113780151B CN 202111043157 A CN202111043157 A CN 202111043157A CN 113780151 B CN113780151 B CN 113780151B
Authority
CN
China
Prior art keywords
features
fusion
time
feature
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111043157.9A
Other languages
Chinese (zh)
Other versions
CN113780151A (en
Inventor
李沂滨
王代超
贾磊
宋艳
郭庆稳
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 CN202111043157.9A priority Critical patent/CN113780151B/en
Publication of CN113780151A publication Critical patent/CN113780151A/en
Application granted granted Critical
Publication of CN113780151B publication Critical patent/CN113780151B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • 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
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Probability & Statistics with Applications (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Signal Processing (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention provides a bearing fault diagnosis method and system with bilinear feature fusion, which are used for acquiring bearing signals detected by various sensors and dividing the bearing signals into time domain data and time-frequency domain data; extracting the characteristics of the two data of each sensor, and carrying out interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism; carrying out deep fusion on the features extracted and fused by each sensor; and classifying the features after the deep fusion to obtain a diagnosis result. The invention can solve the interaction problem among different input characteristics and the effective fusion problem of different signal source characteristics, and improve the precision of bearing fault diagnosis.

Description

Bearing fault diagnosis method and system based on bilinear feature fusion
Technical Field
The invention belongs to the technical field of bearing fault diagnosis, and particularly relates to a bearing fault diagnosis method and system with bilinear feature fusion.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
As an important component of rotating machinery, such as drive motors and wind turbines, bearings are becoming increasingly important in industrial applications. Many bearing failures can result in high maintenance costs and even loss of life due to increasingly complex structures and harsh operating environments. Research shows that bearing faults account for 40% of faults of large mechanical systems and 90% of faults of small mechanical systems. Therefore, accurate diagnosis of bearing faults, particularly in early fault diagnosis, is crucial.
Fault diagnosis methods are mainly divided into two categories, model-based and data-driven. The state space model using the dynamic equation is a common model-based method, but the model-based method requires the establishment of an accurate model and depends on a great deal of professional knowledge. However, since the rotating machine has complicated coupling and a severe working environment, it is difficult to establish an accurate model. At present, the data driving method can be widely applied to bearing fault diagnosis due to the fact that the characteristics can be automatically extracted and the requirement on professional knowledge is low. In addition, with the rapid development of the manufacturing industry, the data acquisition of mechanical equipment is faster and more extensive than before. It provides new opportunities for data-driven fault diagnosis methods, and has also attracted much research attention.
The existing research shows that the vibration signal is most widely applied to mechanical fault diagnosis. Many researchers have employed vibration signals for bearing fault diagnosis and have achieved a great deal of success. Although vibration signals have found widespread use in fault diagnosis, fault diagnosis based on vibration signals still presents many challenges. Therefore, in order to better perform fault diagnosis, researchers have attempted to develop fault diagnosis methods based on other signals, such as acoustic signals, current signals, and the like.
However, in a complicated industrial system, it is difficult to obtain high fault diagnosis performance with a single sensor signal. Research has shown that fault diagnosis using multi-sensor signals has better performance than fault diagnosis using single-sensor data, since multi-sensor signals typically contain complementary fault signatures. Therefore, multi-sensor data fusion is widely applied to fault diagnosis. Data fusion can be classified into data-level fusion, feature-level fusion, and decision-level fusion according to the fusion level. In data level fusion, the original data is processed and recombined, thereby preserving the original information to the maximum extent. It requires a highly consistent data structure, which is difficult to implement. In decision-level fusion, the diagnostic result is determined jointly by the results of multiple classifiers, where conflicting results may affect the accuracy of the final classification. Thus, research has shown that feature level fusion has better performance than the other two methods because it can compress large amounts of information, facilitating real-time data processing.
However, the existing feature level fusion fault diagnosis method has the following two problems:
(1) in the process of feature extraction, most of the existing methods only focus on independently extracting features from different signals. Interaction among features from different inputs is ignored, and the extracted features are poor in fusion effect;
(2) most studies simply relate the features of the last layer. The features of the different signals are not effectively fused, resulting in failure to achieve the desired diagnostic accuracy.
Disclosure of Invention
The invention provides a bearing fault diagnosis method and system with bilinear feature fusion, aiming at solving the problems, and the method and system can solve the interaction problem among different input features and the effective fusion problem of different signal source features and improve the bearing fault diagnosis precision.
According to some embodiments, the invention adopts the following technical scheme:
a bearing fault diagnosis method with bilinear feature fusion comprises the following steps:
acquiring bearing signals detected by each sensor, and dividing the bearing signals into time domain data and time-frequency domain data;
extracting the characteristics of the two data of each sensor, and carrying out interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism;
performing deep fusion on the features extracted and fused by each sensor;
and classifying the features after the deep fusion to obtain a diagnosis result.
As an alternative embodiment, the specific process of extracting the features of the two data includes:
and the time domain features guide the time-frequency domain feature extraction and the time-frequency domain features guide the time-frequency domain feature extraction.
As a further limitation, the specific process of extracting the features of the two data includes: performing feature extraction on the time-frequency domain data by using a two-dimensional convolutional neural network;
performing feature extraction on the time domain data by using a one-dimensional convolutional neural network;
and fusing the extracted features.
As a further limitation, in the stage of time-domain feature extraction guided by time-domain features, the extracted features are subjected to dimensional deformation, then multiplied by another extracted feature, and the product is input to an attention mechanism module and weighted by an attention coefficient.
As an alternative embodiment, when two kinds of features extracted from the respective sensors are deeply fused, the features extracted from the plurality of sensors are sufficiently fused using a bilinear model.
In an alternative embodiment, the features after depth fusion are classified after normalization.
As an alternative embodiment, when the features after the depth fusion are classified, the extracted features are mapped to the classification result according to the bearing state corresponding to the maximum probability by adopting softmax regression.
A bilinear feature fused bearing fault diagnostic system comprising:
the data acquisition module is configured to acquire the bearing signals detected by the sensors and divide the bearing signals into two types of data, namely time domain data and time-frequency domain data;
the characteristic extraction and fusion module is configured to extract the characteristics of the two data of each sensor and perform interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism;
the depth fusion module is configured to extract the fused features from each sensor for depth fusion;
and the classification module is configured to classify the fused features to obtain a diagnosis result.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to carry out the steps of the above-mentioned method.
A terminal device comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is used for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the above-described method.
Compared with the prior art, the invention has the beneficial effects that:
the invention adopts a mutual attention mechanism to realize the interaction between the characteristics of different inputs. Specifically, the feature extraction process of one modality data in different modality data of the same signal source is guided by the extracted feature of the other modality data, and the purpose of this is to make the feature extraction process of the modality data based on the feature of the other modality data. Thereby connecting the two features. The extracted characteristics fully consider the information of the data of the two modes, so that the data has better fusion characteristics, and the performance of fault diagnosis is greatly improved.
The invention adopts a feature fusion module based on a bilinear model to realize deep fusion of different signal source features. The fusion mode is different from simple connection, and is characterized in that the features are fused on the element level, and the deep fusion of the features is realized by adopting an element multiplication summation mode. Through the bilinear model, the characteristics extracted by the two signal sources can be deeply fused, and the final fault diagnosis accuracy rate is greatly improved compared with simple connection.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are included to illustrate an exemplary embodiment of the invention and not to limit the invention.
FIG. 1 is a schematic view of the flow concept of the present invention;
FIG. 2 is a block diagram of a feature extraction and fusion module based on the mutual attention mechanism according to the present invention.
The specific implementation mode is as follows:
the invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. 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 invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
A bilinear fusion network bearing fault diagnosis method based on a mutual attention mechanism is disclosed, the principle of which is shown in FIG. 1, and can be summarized into 3 steps: (1) feature extraction and fusion based on a mutual attention mechanism; (2) fusing features based on a bilinear model; (3) and (4) fault classification. Which will be described in detail below.
In this embodiment, the collected signal is specifically composed of four channel inputs and is divided into two groups. Each group consists of different patterns (time domain and time-frequency domain) of a single sensor signal.
(1) Feature extraction and fusion based on mutual attention mechanism
The method comprises the steps of feature extraction and feature fusion. In the feature extraction part, interaction between time domain data features and time-frequency data features is realized through time domain feature guidance time-frequency domain feature extraction and time-frequency domain feature guidance time-frequency domain feature extraction. As shown in particular in fig. 2.
Assuming that the time-frequency domain data input is
Figure BDA0003250161080000061
The time domain data is input as
Figure BDA0003250161080000062
Firstly, a two-dimensional and one-dimensional Convolutional Neural Network (CNN) is adopted for feature extraction:
Figure BDA0003250161080000071
Figure BDA0003250161080000072
wherein
Figure BDA0003250161080000073
Representing the kth convolution kernel from input X1And K represents the number of convolution kernels. Thereby obtaining the characteristics
Figure BDA0003250161080000074
And
Figure BDA0003250161080000075
wherein Y is1∈Rm×m×K，Y2∈Rm×K. Then a mutual attention mechanism is used to realize the characteristic Y1And Y2The interaction of (2).
In the stage of time domain feature extraction guided by time domain features, firstly, the features Y are extracted2Deformation to Y2'∈R1×m×KThen it is compared with the feature Y1Multiplication is carried out:
H1'＝Y2'×Y1,H1'∈R1×m×K (3)
then H is introduced1' input attention mechanism module:
Z1＝Averagepool(H1') (4)
Z1'＝FC(FC(Z1)) (5)
wherein Z is1'＝[z′1,z′2,...,z′K]∈R1×KIs the attention coefficient. Then using the vector Z1' coming to feature Y1And (3) weighting:
Figure BDA0003250161080000076
after the attention module has been re-weighted,realize X2To X1Guidance of the feature extraction process, hence feature Y1'∈Rm×m×KFully consider Y1And Y2The fault signature of (2). In the same way, the characteristic Y can be obtained2'∈R1×m×K。
In the feature fusion part, the feature Y is fused1' and Y2' making a connection to obtain a new feature F1∈R(m+1)×m×K：
F1＝Conc(Y1',Y2') (7)
Where Conc represents a join operation.
The above process realizes the feature extraction of different modes of a single signal source, and the same operation is adopted to obtain the feature F of another signal source2∈R(m+1)×m×K。
(2) Feature fusion based on bilinear model
Effective feature fusion will greatly improve the accuracy of fault diagnosis. Features extracted from multiple sensors are fully fused by using a bilinear model. From the above steps, we obtain the features extracted from two signal sources
Figure BDA0003250161080000081
And
Figure BDA0003250161080000082
wherein
Figure BDA0003250161080000083
Assume that in K features F1And F2The elements at the (i, j) positions of (a) are respectively
Figure BDA0003250161080000084
And
Figure BDA0003250161080000085
the fusion of features is performed by the following formula:
Figure BDA0003250161080000086
wherein Trans represents the full-time operation and D ∈ RK×K(ii) a Because the element value of the fused features is large, the final classification accuracy is influenced, and therefore the features are normalized by adopting the following formula:
Figure BDA0003250161080000087
wherein d isi,jI, j ═ 1, 2., K constitutes the final feature D'.
Different from simple connection, the feature fusion step based on the bilinear model is to multiply elements of the feature map, so that the deep fusion of the feature map is realized.
(3) Fault classification
The scheme adopts softmax regression to map the extracted features to classification results. First, stretch feature D' into one dimension:
D'＝Flatten(D'),D'∈RK·K (10)
according to the softmax function, each bearing state corresponds to a probability, which can be obtained by adopting the following formula:
Figure BDA0003250161080000091
wherein p isi(x) Is the probability corresponding to the ith bearing state, and outputs the bearing state corresponding to the maximum probability as the final classification result.
Given the classification result and the true label of the data, the classification error can be calculated as follows, based on the cross entropy loss:
Figure BDA0003250161080000092
wherein y isiIs a true label corresponding to the ith bearing state, and when the classification result is the ith class, yiIs 1;otherwise, yiIs 0. The values of the weights and offsets may be updated by the partial derivatives of E (y, p (x)). For example, the value of the weight w can be updated by the following formula:
Figure BDA0003250161080000093
wherein l represents the learning rate when the weight value is updated.
The above-described embodiments can extract and effectively fuse high-quality features. And the mutual attention mechanism is utilized to carry out characteristic interaction between different modes of the single sensor signal, so that an attention coefficient is generated, useful characteristics are highlighted, and redundant characteristics are suppressed. Thus, the features extracted from one modality data fully take into account the information of the other modality data. Then, features extracted from the multiple sensor signals are fused by using a bilinear model, so that deep fusion of the features is realized.
The invention also provides the following product examples:
a bilinear feature fused bearing fault diagnostic system comprising:
the data acquisition module is configured to acquire the bearing signals detected by the sensors and divide the bearing signals into time domain data and time-frequency domain data;
the characteristic extraction and fusion module is configured to extract the characteristics of the two data of each sensor and perform interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism;
the depth fusion module is configured to extract the fused features from each sensor for depth fusion;
and the classification module is configured to classify the fused features to obtain a diagnosis result.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to carry out the steps of the above-mentioned method.
A terminal device comprising a processor and a computer readable storage medium, the processor for implementing instructions; the computer readable storage medium is used for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the above-described method.
As will be appreciated by one skilled in the art, 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 an entirely hardware embodiment, an entirely 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, disk storage, CD-ROM, 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 flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams 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.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.

Claims (6)

1. A bearing fault diagnosis method based on bilinear feature fusion is characterized by comprising the following steps: the method comprises the following steps:
acquiring bearing signals detected by each sensor, and dividing the bearing signals into time domain data and time-frequency domain data;
extracting the characteristics of the two data of each sensor, and carrying out interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism; the specific process is as follows:
interaction of time domain data features and time-frequency data features is realized through time domain feature guidance time-frequency domain feature extraction and time-frequency domain feature guidance time-frequency domain feature extraction;
assuming that the time-frequency domain data input is
Figure FDA0003574241930000011
The time domain data is input as
Figure FDA0003574241930000012
Firstly, extracting features by adopting two-dimensional and one-dimensional convolution neural networks; thereby obtaining the characteristics
Figure FDA0003574241930000013
And
Figure FDA0003574241930000014
wherein Y is1∈Rm×m×K，Y2∈Rm×K(ii) a Then adopt each otherAttention mechanism to achieve feature Y1And Y2The interaction of (1);
in the stage of time domain feature extraction guided by time domain features, firstly, the features Y are extracted2Deformation to Y2'∈R1×m×KThen it is compared with the feature Y1Multiplying, inputting the result of multiplying to the attention mechanism module to obtain Z1'；
Then using the vector Z1' coming to feature Y1Carrying out weighting;
after the attention module reweighs, X is realized2To X1Guidance of the feature extraction process, hence feature Y1'∈Rm ×m×KFully consider Y1And Y2The fault characteristic of (a); in the same way, the characteristic Y can be obtained2'∈R1×m×K；
Carrying out deep fusion on the features extracted and fused by each sensor; when two features extracted from each sensor are deeply fused, a bilinear model is adopted to fully fuse the features extracted from a plurality of sensors; different from simple connection, the characteristic fusion step based on the bilinear model is to multiply elements of the characteristic graph, so that the deep fusion of the characteristic graph is realized;
and classifying the features after the deep fusion to obtain a diagnosis result.
2. The method for diagnosing the bearing fault with the bilinear feature fusion as claimed in claim 1, wherein the method comprises the following steps: and carrying out normalization processing on the features subjected to depth fusion and then classifying.
3. The method for diagnosing the bearing fault with the bilinear feature fusion as claimed in claim 1, wherein the method comprises the following steps: and when the features after the depth fusion are classified, mapping the extracted features to a classification result according to the bearing state corresponding to the maximum probability by adopting softmax regression.
4. A bearing fault diagnosis system with bilinear feature fusion is characterized in that: the method comprises the following steps:
the data acquisition module is configured to acquire the bearing signals detected by the sensors and divide the bearing signals into two types of data, namely time domain data and time-frequency domain data;
the characteristic extraction and fusion module is configured to extract the characteristics of the two data of each sensor and perform interaction and fusion of time domain data characteristics and time frequency data characteristics based on a mutual attention mechanism; the specific process is as follows:
interaction of time domain data features and time-frequency data features is realized through time domain feature guidance time-frequency domain feature extraction and time-frequency domain feature guidance time-frequency domain feature extraction;
assuming that the time-frequency domain data input is
Figure FDA0003574241930000021
The time domain data is input as
Figure FDA0003574241930000022
Firstly, extracting features by adopting two-dimensional and one-dimensional convolution neural networks; thereby obtaining the characteristics
Figure FDA0003574241930000023
And
Figure FDA0003574241930000024
wherein Y is1∈Rm×m×K，Y2∈Rm×K(ii) a Then a mutual attention mechanism is used to realize the characteristic Y1And Y2The interaction of (1);
in the stage of time domain feature extraction guided by time domain features, firstly, the features Y are extracted2Deformation to Y2'∈R1×m×KThen it is compared with the feature Y1Multiplying, inputting the result of multiplying to the attention mechanism module to obtain Z1'；
Then using the vector Z1' coming to feature Y1Carrying out weighting;
after the attention module reweighs, X is realized2To X1Guidance of the feature extraction process, hence feature Y1'∈Rm ×m×KFully consider Y1And Y2The fault characteristic of (a); in the same manner, the characteristic Y can be obtained2'∈R1×m×K；
The depth fusion module is configured to extract the fused features from each sensor for depth fusion; when two features extracted from each sensor are deeply fused, a bilinear model is adopted to fully fuse the features extracted from a plurality of sensors; different from simple connection, the characteristic fusion step based on the bilinear model is to multiply elements of the characteristic graph, so that the deep fusion of the characteristic graph is realized;
and the classification module is configured to classify the fused features to obtain a diagnosis result.
5. A computer-readable storage medium characterized by: in which a plurality of instructions are stored, said instructions being adapted to be loaded by a processor of a terminal device and to carry out the steps of the method according to any one of claims 1 to 3.
6. A terminal device is characterized in that: the system comprises a processor and a computer readable storage medium, wherein the processor is used for realizing instructions; a computer readable storage medium for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the method according to any of claims 1-3.
CN202111043157.9A 2021-09-07 2021-09-07 Bearing fault diagnosis method and system based on bilinear feature fusion Active CN113780151B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111043157.9A CN113780151B (en) 2021-09-07 2021-09-07 Bearing fault diagnosis method and system based on bilinear feature fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111043157.9A CN113780151B (en) 2021-09-07 2021-09-07 Bearing fault diagnosis method and system based on bilinear feature fusion

Publications (2)

Publication Number Publication Date
CN113780151A CN113780151A (en) 2021-12-10
CN113780151B true CN113780151B (en) 2022-05-20

Family

ID=78841562

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111043157.9A Active CN113780151B (en) 2021-09-07 2021-09-07 Bearing fault diagnosis method and system based on bilinear feature fusion

Country Status (1)

Country Link
CN (1) CN113780151B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114818835B (en) * 2022-06-27 2022-09-23 山东大学 Rotary machine fault diagnosis method and system based on difference feature extraction and fusion
CN115143128B (en) * 2022-06-28 2023-05-16 浙江石水泵业科技有限公司 Fault diagnosis method and system for small-sized submersible electric pump

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200097651A1 (en) * 2018-09-26 2020-03-26 General Electric Company Systems and methods to achieve robustness and security in medical devices
CN111275007B (en) * 2020-02-24 2021-10-22 山东大学 Bearing fault diagnosis method and system based on multi-scale information fusion
CN111931806A (en) * 2020-06-23 2020-11-13 广州杰赛科技股份有限公司 Equipment fault diagnosis method and device for multi-sensor data fusion
CN112926641B (en) * 2021-02-22 2022-05-17 山东大学 Three-stage feature fusion rotating machine fault diagnosis method based on multi-mode data

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
应用图像特征和高斯过程的行星齿轮箱齿轮故障诊断研究;王斌;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20200915;第2020年卷(第9期);第C029-17页 *

Also Published As

Publication number Publication date
CN113780151A (en) 2021-12-10

Similar Documents

Publication Publication Date Title
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
Wang et al. An enhanced intelligent diagnosis method based on multi-sensor image fusion via improved deep learning network
CN113780151B (en) Bearing fault diagnosis method and system based on bilinear feature fusion
Yiakopoulos et al. Rolling element bearing fault detection in industrial environments based on a K-means clustering approach
Rojas et al. Practical scheme for fast detection and classification of rolling-element bearing faults using support vector machines
Ainapure et al. Deep learning-based cross-machine health identification method for vacuum pumps with domain adaptation
Shokrolahi et al. A deep network solution for intelligent fault detection in analog circuit
Islam et al. Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal
Tang et al. Feature extraction of multi-sensors for early bearing fault diagnosis using deep learning based on minimum unscented kalman filter
Qi et al. A review on data-driven condition monitoring of industrial equipment
CN113420837A (en) Fault diagnosis method based on multi-source compressed sensing
Zhao et al. An intelligent diagnosis method of rolling bearing based on multi-scale residual shrinkage convolutional neural network
Xu et al. Deep domain adversarial method with central moment discrepancy for intelligent transfer fault diagnosis
Wang et al. Performance analysis and enhancement of deep convolutional neural network: Application to gearbox condition monitoring
Luo et al. A fault diagnosis method for rotating machinery with variable speed based on multi-feature fusion and improved ShuffleNet V2
Yang et al. Improved graph-regularized deep belief network with sparse features learning for fault diagnosis
CN116738367A (en) Rolling bearing fault diagnosis method based on multi-scale map transducer
Liu et al. Fault diagnosis of satellite flywheel bearing based on convolutional neural network
Dong et al. Bearing fault diagnosis under different operating conditions based on cross domain feature projection and domain adaptation
Chen et al. Time–frequency Transformer with shifted windows for journal bearing-rotor systems fault diagnosis under multiple working conditions
Zheng et al. Few-shot intelligent fault diagnosis based on an improved meta-relation network
Zhang et al. The combination model of CNN and GCN for machine fault diagnosis
CN113158769A (en) CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method
CN113724261A (en) Fast image composition method based on convolutional neural network
Yu et al. Fast adaption for multi motor anomaly detection via meta learning and deep unsupervised learning

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