CN112432790A - Rolling bearing fault diagnosis method and device and storage medium - Google Patents
Rolling bearing fault diagnosis method and device and storage medium Download PDFInfo
- Publication number
- CN112432790A CN112432790A CN202010706629.3A CN202010706629A CN112432790A CN 112432790 A CN112432790 A CN 112432790A CN 202010706629 A CN202010706629 A CN 202010706629A CN 112432790 A CN112432790 A CN 112432790A
- Authority
- CN
- China
- Prior art keywords
- domain features
- rolling bearing
- features
- mixed
- 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.)
- Granted
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
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A rolling bearing failure diagnosis method and apparatus and a storage medium are disclosed. The rolling bearing fault diagnosis method comprises the following steps: acquiring a vibration acceleration signal of a rolling bearing; extracting m mixed domain features from the vibration acceleration signal, wherein m is an integer greater than 1, and the m mixed domain features comprise a plurality of time domain features, a plurality of frequency domain features and a plurality of time-frequency domain features; ranking the m mixed domain features by using a semi-supervised laplacian score algorithm, and selecting top-ranked n features of the m mixed domain features, wherein n is an integer greater than 1 and n < m; performing feature dimension reduction on the n features by using a dimension reduction algorithm; and determining a fault state of the rolling bearing by using the classifier based on the reduced-dimension features.
Description
Technical Field
The present disclosure relates to a rolling bearing fault diagnosis method and apparatus, and a storage medium.
Background
In automotive production plants, elevators are common transport devices. For example, an elevator may be used to displace vehicles between adjacent production lines. Rolling bearings are one of the key components in elevators. If the rolling bearing fails, the elevator may be shut down, which may cause shutdown of multiple production lines. Currently, security workers will manually inspect rolling bearings in elevators on a regular basis in the hope of finding a potential failure at an early stage.
Disclosure of Invention
The disclosure provides a fault diagnosis method for a rolling bearing, which comprises the following steps: acquiring a vibration acceleration signal of a rolling bearing; extracting m mixed domain features from the vibration acceleration signal, wherein m is an integer greater than 1, and the m mixed domain features comprise a plurality of time domain features, a plurality of frequency domain features and a plurality of time-frequency domain features; ranking the m mixed domain features by using a semi-supervised laplacian score algorithm, and selecting top-ranked n features of the m mixed domain features, wherein n is an integer greater than 1 and n < m; performing feature dimension reduction on the n features by using a dimension reduction algorithm; and determining a fault state of the rolling bearing by using the classifier based on the reduced-dimension features.
Other features and advantages of the present disclosure will become apparent from the following description with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the disclosure and together with the description, serve to explain, without limitation, the principles of the disclosure. In the drawings, like numbering is used to indicate like items.
Fig. 1 is a block diagram of an example rolling bearing fault diagnostic device, according to some embodiments of the present disclosure.
Fig. 2 is a flow chart illustrating an exemplary rolling bearing fault diagnostic method according to some embodiments of the present disclosure.
Fig. 3 illustrates time domain waveform diagrams of vibration acceleration signals of several typical rolling bearings.
Fig. 4 is a flow diagram illustrating an exemplary feature selection process according to some embodiments of the present disclosure.
Fig. 5A illustrates a fault diagnosis result according to a comparative example, and fig. 5B illustrates a fault diagnosis result according to one embodiment of the present disclosure.
Fig. 6 illustrates a general hardware environment in which the present disclosure may be applied, in accordance with some embodiments of the present disclosure.
Detailed Description
In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the described exemplary embodiments. It will be apparent, however, to one skilled in the art, that the described embodiments may be practiced without some or all of these specific details. In the described exemplary embodiments, well-known structures or processing steps have not been described in detail in order to avoid unnecessarily obscuring the concepts of the present disclosure.
The blocks within each block diagram shown below may be implemented by hardware, software, firmware, or any combination thereof to implement the principles of the present disclosure. It will be appreciated by those skilled in the art that the blocks described in each block diagram can be combined or divided into sub-blocks to implement the principles of the disclosure.
The steps of the methods presented in this disclosure are intended to be illustrative. In some embodiments, the method may be accomplished with one or more additional steps not described and/or without one or more of the steps discussed. Further, the order in which the steps of the method are illustrated and described is not intended to be limiting.
Fig. 1 is a block diagram of an example rolling bearing fault diagnostic device 100, according to some embodiments of the present disclosure. As shown in fig. 1, the apparatus 100 may include: a vibration signal acquisition unit 110 configured to acquire a vibration acceleration signal of the rolling bearing; a feature extraction section 120 configured to extract a plurality of mixed domain features from the vibration acceleration signal; a feature selection part 130 configured to select a plurality of more important features from the plurality of mixed domain features by using a Semi-Supervised Laplace Score (SSLS) algorithm; a feature dimension reduction component 140 configured to perform feature dimension reduction (also referred to as dimension reduction) on a plurality of more important features by using a dimension reduction algorithm; and a failure state determination section 150 configured to determine a failure state of the rolling bearing by using the classifier based on the dimensionality reduced features.
The operation of the various components shown in fig. 1 will be described in further detail below.
Fig. 2 is a flow chart illustrating an exemplary rolling bearing fault diagnostic method 200, according to some embodiments of the present disclosure.
The method 200 starts at step S210, where the vibration signal acquisition part 110 acquires a vibration acceleration signal of the rolling bearing at step S210. Here, the vibration acceleration signal refers to a radial vibration acceleration signal of a drive shaft of the rolling bearing. The vibration signal acquisition section 110 may acquire a vibration acceleration signal from an acceleration sensor provided at a drive end of the rolling bearing. The vibration signal obtaining part 110 may sample the vibration acceleration signal at a predetermined sampling frequency (for example, 12kHz) and obtain data samples having a predetermined length (for example, 2048 sampling points). The vibration signal acquisition section 110 may acquire a plurality of (e.g., q being an integer greater than 2, q may be 100, for example) data samples in various known manners. The vibration signal acquisition section 110 may output a plurality of data samples to the feature extraction section 120 and the feature selection section 130.
Fig. 3 illustrates time domain waveform diagrams of several typical vibration acceleration signals. A rolling bearing generally includes rolling bodies, an inner ring, an outer ring, and a cage. Accordingly, the types of failure of the rolling bearing generally include the following three types: rolling element failure, inner race failure, and outer race failure. Fig. 3 (a) shows a vibration acceleration signal segment in a normal state of the rolling bearing; fig. 3 (b) shows a vibration acceleration signal segment in a rolling element failure state; fig. 3 (c) shows a vibration acceleration signal segment in an inner ring failure state; fig. 3 (d) shows a vibration acceleration signal section in the outer ring failure state.
The method 200 proceeds to step S220, where at step S220, the feature extraction unit 120 extracts m mixed-domain features from the vibration acceleration signal, where m is an integer greater than 1, the m mixed-domain features including a plurality of time-domain features, a plurality of frequency-domain features, and a plurality of time-frequency-domain features. More specifically, the feature extraction section 120 extracts m mixed-domain features from each data sample received from the vibration signal acquisition section 110. When the feature extraction unit 120 receives q data samples from the vibration signal acquisition unit 110, the feature extraction unit 120 generates a q × m feature matrix, that is, a q-row and m-column feature matrix.
In some embodiments, the feature extraction component 120 extracts 26 mixed-domain features, i.e., m-26, from the vibration acceleration signal. These 26 mixed domain features include: 16 time domain features p1~p165 frequency domain features p17~p21And 5 time-frequency domain features p22~p26. The extracted features p are listed in Table 1 below1~p21。
TABLE 1
As shown in Table 1, 16 time-domain features p1~p16The method comprises the following steps: 10 dimensional statistics p1~p10I.e., mean, standard deviation, square root amplitude, mean amplitude, slope, kurtosis, root mean square value, maximum value, minimum value, difference between maximum and minimum values; and 6 dimensionless statistics p11~p16I.e., a waveform index, a peak index, a pulse index, a margin index, a skewness index, and a kurtosis index. 5 frequency domain features p17~p21The method comprises the following steps: mean square frequency, root mean square frequency, frequency standard deviation, kurtosis frequency, and center frequency, where p17Reflecting the magnitude of the vibration energy in the frequency domain, p18~p20Reflecting the change in the position of the main band, p21Indicating the degree of spectral dispersion or concentration. And, 5 time-frequency domain features p22~p26Is the sum of the squares of the respective energy characteristics of the first five eigenmode function (i.e., IMF1-IMF5) components of Empirical Mode Decomposition (EMD). Here, the energy feature sum of squares reflects the energy magnitude (i.e., the magnitude of the signal fluctuation amplitude) of each IMF component. More specifically, the energy feature sum of squares refers to the sum of squares of the magnitudes of the respective IMF components.
It should be understood that the extracted features are not limited to the 26 features listed above. Alternatively, more or fewer features may be extracted according to actual needs.
When the feature extraction unit 120 receives q data samples from the vibration signal acquisition unit 110, the feature extraction unit 120 outputs a q × m feature matrix generated by feature extraction to the feature selection unit 130. For example, in the case where 100 data samples are received and 26 mixed domain features are extracted for each data sample, the feature extraction section 120 outputs a 100 × 26 feature matrix to the feature selection section 130.
Next, the method 200 proceeds to step S230, at step S230, the feature selection component 130 sorts the m mixed domain features as described above by using the SSLS algorithm, and selects top n features of the m mixed domain features, where n is an integer greater than 1 and n < m. In some embodiments, the first 8 features ordered by importance may be selected from, for example, 26 mixed-domain features for subsequent processing, i.e., n-8. Here, the value of n may be determined according to actual requirements. In the case where the feature selection section 130 receives the q × m feature matrix from the feature extraction section 120, the feature selection section 130 outputs the q × n feature matrix after feature selection to the feature dimensionality reduction section 140.
The feature selection process 400 performed by the feature selection component 130 is described in detail below with reference to fig. 4. The feature selection process 400 includes the following four sub-steps S231-S234.
In the first substep S231, (l + u) data samples X ═ X are used1,x2,…,xl,xl+1,…,xl+uH, to construct a neighbor map G in which data samples xi(i 1, 2.... l + u) corresponds to the ith node and is based on every second data sample xiAnd xj(j ≠ 1, 2., l + u, and j ≠ i) to establish an edge between the ith and jth nodes. Data sample xiRepresenting a vibration acceleration signal sampled over a certain time period. l and u are each an integer greater than 1, and Xl={x1,x2,…,xlIs 1 labeled data sample, labeled { z }1,z2,…,zlThe labels indicate different fault states, Xu={xl+1,xl+2,…,xl+uAre u unlabeled data samples.
Here, l + u is q. That is, the feature selection section 130 constructs the neighbor map G using q data samples received from the vibration signal acquisition section 110. For example, in the case where there are 100 data samples, a neighbor graph G having 100 nodes may be constructed, and in the case where the similarity of each two data samples is greater than or equal to a predetermined value, an edge connecting the nodes is established between the respective two nodes, and in the case where the similarity of each two data samples is less than the predetermined value, an edge is not established between the respective two nodes.
In some embodiments, the number of tagged data samples/may be less than the number of untagged data samples u. In other embodiments, the number of tagged data samples/may account for 20% or more of the total number of samples (l + u).
In a second substep S232, a weighting matrix S is calculated from the neighbour map G. Specifically, the weighting matrix S is calculated according to the following equations (1) and (2):
if xi∈Xl,xiIs Nk(xi) Then, then
If xi∈Xu,xiIs Nk(xi) Then, then
Whereint is a constant, | | · | |, represents the euclidean distance, and the neighborhood Nk(xi) Is predetermined by training.
As can be understood from the above equations (1) and (2), it is assumed that samples arexjIs a sample xiA point in the neighborhood of (2), if xjHas a label and xiWith the same label, x is increasediAnd xjIn between, i.e. increasing the weight SijThe point is enabled to play a greater role in minimizing the objective function; if xjWith a label but xiWith different labels, x is removediAnd xjThe connecting edge therebetween, i.e. the weight SijSet 0, make this point not play a role in the minimization of the objective function; if xjIf there is no label, the weight SijKeeping the same; and, in any other case, weight SijAnd setting 0.
In a third sub-step S233, laplacian scores of the m mixed-domain features are calculated based on the calculated weighting matrix S, wherein for each data sample xiM mixed domain features are determined.
In some embodiments, for the r (r ═ 1, 2.., m) th feature, the following equations (3) - (5) are defined:
fr=(fr1,fr2,…frq)T...(3)
D=diag(SI)...(4)
L=D-S...(5)
wherein f isriThe ith feature of the ith data sample is represented, I is a q-dimensional unit vector, and the matrix L is a laplacian matrix of the neighbor graph G. To make the calculation more accurate, the individual features are de-averaged according to equation (6) below:
further, the SSLS of the r-th feature is calculated according to the following equation (7) and equation (8):
wherein Var (f)r) Is the variance of the r-th feature. SijThe larger (f)ri-frj)2The smaller the difference in the data sample over the feature, i.e., the stronger the local information retention of the feature. With respect to equation (7), a smaller numerator indicates that a neighboring sample has a smaller difference in the feature, i.e., the feature has a stronger local information retention capability; a larger denominator indicates that the samples differ more in the feature, i.e., the feature is more discriminative. Therefore, the score Lr of a feature is inversely proportional to the importance of the feature.
In a fourth substep S234, the m mixed domain features are ordered in order of small to large laplacian scores. The more forward the ranking, the higher the importance of the feature is explained.
Although not shown, the feature selection process 400 may also include a sub-step of selecting the top n ranked features.
The SSLS characteristic selection processing extracts the intrinsic information structure of the data by directly learning the characteristic set, fully utilizes the label information of partial data, simplifies the characteristic space, and greatly reserves the integral geometric structure information contained in the fault signal characteristic set, thereby being beneficial to judging and diagnosing the fault of the rolling bearing.
Next, the method 200 proceeds to step S240, at which step S240, the feature dimensionality reduction component 140 performs feature dimensionality reduction on the n features by using a dimensionality reduction algorithm.
In some embodiments, a Principal Component Analysis (PCA) method is employed to perform feature dimension reduction on the n features. In the case where the feature dimensionality reduction component 140 receives the q × n feature matrix from the feature selection component 130, the step of PCA is as follows. The q x n feature matrix comprises q n-dimensional feature vectors xi(i=1,2,...,q)。
(1) Calculate the mean value according to equation (9):
(2) the covariance matrix of the feature vector is calculated according to equation (10):
(3) calculating an eigenvalue λ of the covariance matrix C according to equation (11)iAnd a feature vector vi(i=1,2,…,n):
Cvi=λivi...(11)。
(4) Will be lambdaiArranged from large to small, the first o maximum eigenvalues and corresponding eigenvectors form Δ ═ λ1,λ2,…,λo) And V ═ V (V)1,v2,…,vo)。
(5) An o-dimensional feature vector is calculated according to equation (12) (o is an integer greater than 1 and o < n):
P=VTX...(12),
here, X denotes the q × n feature matrix as described above. The pivot P has the main information of the original characteristic vector and can approximately represent the original characteristic vector, thereby achieving the purpose of reducing the dimension.
After PCA dimensionality reduction, a q x o feature matrix is obtained, and the feature matrix comprises q o-dimensional feature vectors xi(i ═ 1, 2.., q). In some embodiments, o is taken to be 2. That is, the original feature vector of n-dimension can be characterized using 2 principal elements. In other words, the 2 principal elements represent the fused n-dimensional original feature vector. In other embodiments, o may also take 3. The value of o depends mainly on the cumulative contribution rate of each principal element.
Although the use of PCA to perform feature dimensionality reduction is described above, the present disclosure is not so limited. In some embodiments, feature dimension reduction may be performed using Kernel Principal Component Analysis (KPCA) methods. In other embodiments, feature dimensionality reduction may be performed using at least one of the following known dimensionality reduction algorithms: linear Discriminant Analysis (LDA) method, equidistant feature mapping (ISOMAP) method, Local Linear Embedding (LLE) method, and local tangent space permutation (LTSA) method. Those skilled in the art will understand how to perform feature dimension reduction using one or more of these dimension reduction algorithms, and will not be described in detail herein.
Next, the method 200 proceeds to step S250, and at step S250, the failure state determination section 150 determines the failure state of the rolling bearing by using the classifier based on the feature after the dimensionality reduction. The fault condition of the rolling bearing may include: normal state, rolling element fault state, inner ring fault state and outer ring fault state.
The fault state determination section 150 may input the q × o feature matrix received from the feature dimension reduction section 140 into the classifier, and output a fault state corresponding to the classification result output from the classifier.
In some embodiments, the classifier may be a K Nearest Neighbor (KNN) classifier. The predetermined parameter (e.g., the value of the parameter k) in the KNN classifier may be predetermined through training. In other embodiments, other known classifiers may be used to implement the classification of the fault, such as a BP neural network (BPNN), a Support Vector Machine (SVM), or the like. Those skilled in the art can understand how to use a known classifier to classify the fault, and the description thereof is omitted here.
Although not shown, the method 200 may include a training step. In the training step, a plurality of vibration acceleration signal segments marked with a plurality of fault states are used for carrying out neighborhood N in the SSLS algorithmk(xi) And a parameter k in the KNN classifier (or a parameter in another type of classifier).
The diagnostic effect of the rolling bearing failure diagnostic method of the present disclosure is described below with reference to fig. 5A and 5B. Fig. 5A illustrates a fault diagnosis result according to a comparative example, and fig. 5B illustrates a fault diagnosis result according to one embodiment of the present disclosure. The experimental data used are the vibration acceleration data of the rolling bearing of the electrical engineering laboratory of the university of Kaiser-Cisco, USA. In one embodiment of the present disclosure related to fig. 5B, the feature extraction process, the feature selection process using the SSLS algorithm, the feature dimension reduction process using the PCA algorithm, and the fault state determination process using the KNN classifier as described previously are performed. As can be seen in FIG. 5B, in this embodiment of the present disclosure, the selected features are reduced to 2 principal elements. The comparative example differs from this one embodiment of the present disclosure in that: instead of using the SSLS algorithm, in the feature selection process, a plurality of features are randomly selected.
As noted in fig. 5A and 5B, o represents normal data, Δ represents rolling element failure data,represents inner ring fault data, □ represents outer ring fault data, andrepresenting the test data.
As can be seen from fig. 5A, the test data cannot be accurately classified. In contrast, as shown in fig. 5B, the test data can be accurately classified into the corresponding categories.
In accordance with the present disclosure, optimization of features is achieved by selecting features using an SSLS algorithm and dimensionality reduction of the selected features. And the accuracy of fault diagnosis is improved.
Compared with the supervised Laplace algorithm, the SSLS algorithm utilizes the characteristics of selecting a small number of marked samples and a large number of unmarked samples, reduces the workload of marking the samples, and simultaneously ensures the accuracy of fault identification.
In contrast to manual inspection of the operating condition of a rolling bearing, the present disclosure proposes a data-driven fault diagnosis method. Under the condition that the vibration acceleration signal of the rolling bearing indicates the fault and the fault type, the fault can be solved by timely and effectively taking action. This saves labor costs to a great extent.
Hardware implementation
Fig. 6 illustrates a general hardware environment 600 in which the present disclosure may be applied, according to an exemplary embodiment of the present disclosure.
Referring to fig. 6, a computing device 600 will now be described as an example of a hardware device applicable to aspects of the present disclosure. Computing device 600 may be any machine configured to perform processing and/or computing, and may be, but is not limited to, a workstation, a server, a desktop computer, a laptop computer, a tablet computer, a personal digital assistant, a smart phone, a portable camera, or any combination thereof. The apparatus 100 described above may be implemented in whole or at least in part by a computing device 600 or similar device or system.
The bus 602 may include, but is not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, an enhanced ISA (eisa) bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.
Software elements may be located in working memory 614 including, but not limited to, an operating system 616, one or more application programs 618, drivers, and/or other data and code. Instructions for performing the above-described methods and steps may be included in one or more application programs 618, and the above-described components of apparatus 100 may be implemented by processor 604 reading and executing the instructions of one or more application programs 618. More specifically, the vibration signal acquisition section 110 may be implemented, for example, by the processor 604 when executing the application 618 with instructions to perform step S210. The feature extraction component 120 may be implemented, for example, by the processor 604 when executing the application 618 with instructions to perform step S220. The feature selection component 130 may be implemented, for example, by the processor 604 when executing the application 618 with instructions to perform step S230. Also, similarly, the feature dimensionality reduction component 140 and the failure state determination component 150 can be implemented, for example, by the processor 604 when executing the application 618 with instructions to perform steps S240 and S250, respectively. Executable or source code for the instructions of the software elements may be stored in a non-transitory computer-readable storage medium, such as the storage device(s) 610 described above, and may be read into working memory 614, where possible compiled and/or installed. Executable code or source code for the instructions of the software elements may also be downloaded from a remote location.
From the above embodiments, it is apparent to those skilled in the art that the present disclosure can be implemented by software and necessary hardware, or can be implemented by hardware, firmware, and the like. Based on this understanding, embodiments of the present disclosure may be implemented partially in software. The computer software may be stored in a computer readable storage medium, such as a floppy disk, hard disk, optical disk, or flash memory. The computer software includes a series of instructions that cause a computer (e.g., a personal computer, a service station, or a network terminal) to perform a method or a portion thereof according to various embodiments of the disclosure.
Having thus described the disclosure, it will be apparent that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the disclosure, and all such modifications as would be obvious to one skilled in the art are intended to be included within the scope of the following claims.
Claims (10)
1. A rolling bearing failure diagnosis method characterized by comprising:
obtaining a vibration acceleration signal of the rolling bearing,
extracting m mixed-domain features from the vibration acceleration signal, where m is an integer greater than 1, the m mixed-domain features including a plurality of time-domain features, a plurality of frequency-domain features, and a plurality of time-frequency-domain features,
ranking the m mixed domain features by using a semi-supervised Laplacian score algorithm, and selecting top-ranked n features of the m mixed domain features, where n is an integer greater than 1 and n < m,
performing feature dimensionality reduction on the n features by using a dimensionality reduction algorithm, an
Based on the features after the dimension reduction, the fault state of the rolling bearing is determined by using a classifier.
2. The method of claim 1, wherein the fault condition of the rolling bearing comprises: normal state, rolling element fault state, inner ring fault state and outer ring fault state.
3. The method of claim 1, wherein the plurality of time-domain features comprises at least one of: a mean, a standard deviation, a square root amplitude, an average amplitude, a slope, a kurtosis, a root mean square value, a maximum, a minimum, a difference between a maximum and a minimum, a waveform indicator, a peak indicator, a pulse indicator, a margin indicator, a skewness indicator, and a kurtosis indicator, the plurality of frequency domain features comprising at least one of: a mean square frequency, a root mean square frequency, a frequency standard deviation, a kurtosis frequency, and a center frequency, and the plurality of time-frequency domain features comprise at least one of: the respective energy feature square sums of the first five eigenmode function components of the empirical mode decomposition.
4. The method of claim 1, wherein ordering the m mixed-domain features by using a semi-supervised laplacian scoring algorithm comprises:
using (1+ u) data samples X ═ X1,x2,…,xl,xl+1,…,xl+uH, to construct a neighbor map G in which data samples xi(i 1, 2.... l + u) corresponds to the ith node and is based on every second data sample xiAnd xj(j ≠ 1, 2.. l., l + u, and j ≠ i) to establish an edge between the ith and jth nodes, with each data sample xiRepresenting a vibration acceleration signal sampled over a certain period of time, l and u each being an integer greater than 1, and wherein Xl={x1,x2,…,xlIs the l labeled data samples, labeled z1,z2,…,zlThe labels indicate different fault states, Xu={xl+1,xl+2,…,xl+uIs u unlabeled data samples,
a weighting matrix S is calculated from the neighbour graph G, wherein,
if xi∈Xl,xiIs Nk(xi) Then, then
If xi∈Xu,xiIs Nk(xi) Then, then
Whereint is a constant, | | · | |, represents the euclidean distance, and the neighborhood Nk(xi) Is predetermined by training,
calculating Laplace scores of the m mixed domain features based on the calculated weighting matrix S, wherein x is the number of data samples per data sampleiDetermining the m mixed domain features, an
And sequencing the m mixed domain features according to the order of the Laplace scores from small to large.
5. The method of claim 1, wherein the dimension reduction algorithm comprises at least one of: principal component analysis, linear discriminant analysis, equidistant feature mapping, kernel principal component analysis, local linear embedding, and local tangent space arrangement.
6. The method of claim 1, wherein the classifier comprises a K-nearest neighbor classifier, and wherein the predetermined parameter in the K-nearest neighbor classifier is predetermined by training.
7. The method of claim 1, wherein the rolling bearing comprises a rolling bearing installed in an elevator in an automotive production plant.
8. A rolling bearing failure diagnosis device characterized by comprising: means for performing the method of any one of claims 1-7.
9. A rolling bearing failure diagnosis device characterized by comprising:
at least one processor; and
at least one storage device storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any of claims 1-7.
10. A non-transitory computer-readable storage medium having stored thereon instructions which, when executed by a processor, cause performance of the method according to any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010706629.3A CN112432790B (en) | 2020-07-21 | 2020-07-21 | Rolling bearing fault diagnosis method and device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010706629.3A CN112432790B (en) | 2020-07-21 | 2020-07-21 | Rolling bearing fault diagnosis method and device and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112432790A true CN112432790A (en) | 2021-03-02 |
CN112432790B CN112432790B (en) | 2023-03-10 |
Family
ID=74689588
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010706629.3A Active CN112432790B (en) | 2020-07-21 | 2020-07-21 | Rolling bearing fault diagnosis method and device and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112432790B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113158769A (en) * | 2021-03-03 | 2021-07-23 | 安徽大学 | CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method |
CN115862681A (en) * | 2023-03-01 | 2023-03-28 | 深圳市新凯来技术有限公司 | Rotor rub-impact fault acoustic signal diagnosis method and device and electronic equipment |
CN116560895A (en) * | 2023-07-10 | 2023-08-08 | 太原理工大学 | Fault diagnosis method for mechanical equipment |
CN116625686A (en) * | 2023-05-04 | 2023-08-22 | 中国航发沈阳发动机研究所 | On-line diagnosis method for bearing faults of aero-engine |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101334786A (en) * | 2008-08-01 | 2008-12-31 | 浙江大学 | Formulae neighborhood based data dimensionality reduction method |
CN103234767A (en) * | 2013-04-21 | 2013-08-07 | 蒋全胜 | Nonlinear fault detection method based on semi-supervised manifold learning |
CN104849050A (en) * | 2015-06-02 | 2015-08-19 | 安徽工业大学 | Rolling bearing fault diagnosis method based on composite multi-scale permutation entropies |
CN105760839A (en) * | 2016-02-22 | 2016-07-13 | 重庆大学 | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine |
CN106153340A (en) * | 2016-06-29 | 2016-11-23 | 北京工业大学 | A kind of Fault Diagnosis of Roller Bearings |
CN106769049A (en) * | 2017-01-18 | 2017-05-31 | 北京工业大学 | A kind of Fault Diagnosis of Roller Bearings based on Laplce's score value and SVMs |
US20200209109A1 (en) * | 2018-12-28 | 2020-07-02 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for fault diagnosis |
-
2020
- 2020-07-21 CN CN202010706629.3A patent/CN112432790B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101334786A (en) * | 2008-08-01 | 2008-12-31 | 浙江大学 | Formulae neighborhood based data dimensionality reduction method |
CN103234767A (en) * | 2013-04-21 | 2013-08-07 | 蒋全胜 | Nonlinear fault detection method based on semi-supervised manifold learning |
CN104849050A (en) * | 2015-06-02 | 2015-08-19 | 安徽工业大学 | Rolling bearing fault diagnosis method based on composite multi-scale permutation entropies |
CN105760839A (en) * | 2016-02-22 | 2016-07-13 | 重庆大学 | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine |
CN106153340A (en) * | 2016-06-29 | 2016-11-23 | 北京工业大学 | A kind of Fault Diagnosis of Roller Bearings |
CN106769049A (en) * | 2017-01-18 | 2017-05-31 | 北京工业大学 | A kind of Fault Diagnosis of Roller Bearings based on Laplce's score value and SVMs |
US20200209109A1 (en) * | 2018-12-28 | 2020-07-02 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for fault diagnosis |
Non-Patent Citations (12)
Title |
---|
刘波等: "基于特征融合和KPCA_GA-SVM的滚动轴承故障诊断", 《现代计算机(专业版)》 * |
姜景升;王华庆;柯燕亮;向伟;: "基于LTSA与K-最近邻分类器的故障诊断" * |
杜伟;房立清;齐子元;: "基于独立特征选择与流形学习的故障诊断" * |
王冠超;赵荣珍;张维强;: "基于一种流形学习新算法的转子故障数据集降维研究" * |
王振亚等: "基于ASL-Isomap流形学习的滚动轴承故障诊断方法", 《噪声与振动控制》 * |
王雪冬;赵荣珍;邓林峰;张亚龙;: "局部费歇尔判别分值在滚动轴承故障诊断中的应用" * |
程军圣等: "基于部分集成局部特征尺度分解与拉普拉斯分值的滚动轴承故障诊断模型", 《振动工程学报》 * |
陈法法;李冕;陈保家;陈从平;: "基于混合域特征集与加权KNN的滚动轴承故障诊断" * |
陈法法等: "基于混合域特征集与加权KNN的滚动轴承故障诊断", 《机械传动》 * |
黄宏臣等: "基于拉普拉斯特征映射的滚动轴承故障识别", 《振动与冲击》 * |
黄海安;王友仁;孙灿飞;陈伟;王俊;: "行星齿轮系故障特征多目标进化选择方法" * |
黄骥等: "局部切空间排列和改进模糊C-均值聚类的滚动轴承故障诊断模型", 《机械设计与制造》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113158769A (en) * | 2021-03-03 | 2021-07-23 | 安徽大学 | CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method |
CN115862681A (en) * | 2023-03-01 | 2023-03-28 | 深圳市新凯来技术有限公司 | Rotor rub-impact fault acoustic signal diagnosis method and device and electronic equipment |
CN116625686A (en) * | 2023-05-04 | 2023-08-22 | 中国航发沈阳发动机研究所 | On-line diagnosis method for bearing faults of aero-engine |
CN116560895A (en) * | 2023-07-10 | 2023-08-08 | 太原理工大学 | Fault diagnosis method for mechanical equipment |
CN116560895B (en) * | 2023-07-10 | 2023-09-12 | 太原理工大学 | Fault diagnosis method for mechanical equipment |
Also Published As
Publication number | Publication date |
---|---|
CN112432790B (en) | 2023-03-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112432790B (en) | Rolling bearing fault diagnosis method and device and storage medium | |
US11282295B2 (en) | Image feature acquisition | |
CN108334574B (en) | Cross-modal retrieval method based on collaborative matrix decomposition | |
EP3047475B1 (en) | System and method for evaluating a cognitive load on a user corresponding to a stimulus | |
US9471886B2 (en) | Class discriminative feature transformation | |
US20210056127A1 (en) | Method for multi-modal retrieval and clustering using deep cca and active pairwise queries | |
CN113360701A (en) | Sketch processing method and system based on knowledge distillation | |
Zhang et al. | An intelligent fault diagnosis method based on domain adaptation for rolling bearings under variable load conditions | |
Zou et al. | Handwritten chinese character recognition by convolutional neural network and similarity ranking | |
Chen et al. | Max-margin discriminant projection via data augmentation | |
CN114463587A (en) | Abnormal data detection method, device, equipment and storage medium | |
Sanchez et al. | Corrosion grade classification: a machine learning approach | |
US11593673B2 (en) | Systems and methods for identifying influential training data points | |
Singh et al. | Dimensionality reduction for classification and clustering | |
KR101064617B1 (en) | Method and apparatus for classifying multivariate stream data | |
US20220277194A1 (en) | Storage medium and inference method | |
US20170293863A1 (en) | Data analysis system, and control method, program, and recording medium therefor | |
CN113033683B (en) | Industrial system working condition monitoring method and system based on static and dynamic joint analysis | |
Zhao et al. | Distance metric learning based on the class center and nearest neighbor relationship | |
CN116086804A (en) | Rolling bearing fault diagnosis method and device based on entropy | |
Zhang et al. | Rotated target recognition of sonar images via convolutional neural networks with rotated inputs | |
CN112115705B (en) | Screening method and device of electronic resume | |
Tang et al. | Introductory Chapter: Development of Data Clustering | |
Svensson | A comparison between Feed-forward and Convolutional Neural Networks for classification of invoice documents | |
Huang et al. | Hybrid Distance Metric Learning for Real-Time Pedestrian Detection and Re-identification |
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 |