CN111678699A - Early fault monitoring and diagnosing method and system for rolling bearing - Google Patents

Early fault monitoring and diagnosing method and system for rolling bearing Download PDF

Info

Publication number
CN111678699A
CN111678699A CN202010557768.4A CN202010557768A CN111678699A CN 111678699 A CN111678699 A CN 111678699A CN 202010557768 A CN202010557768 A CN 202010557768A CN 111678699 A CN111678699 A CN 111678699A
Authority
CN
China
Prior art keywords
rolling bearing
wavelet packet
sequence
graph
singular value
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
Application number
CN202010557768.4A
Other languages
Chinese (zh)
Other versions
CN111678699B (en
Inventor
卢国梁
文新
闫鹏
杨少华
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 CN202010557768.4A priority Critical patent/CN111678699B/en
Publication of CN111678699A publication Critical patent/CN111678699A/en
Application granted granted Critical
Publication of CN111678699B publication Critical patent/CN111678699B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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

Abstract

The invention discloses a rolling bearing-oriented early fault monitoring and diagnosis method, which solves the problem of missed detection or delayed detection of early faults in the prior art, has the beneficial effects of realizing online monitoring of the running state of a rolling bearing and timely and accurately monitoring the early faults of the rolling bearing, and has the following specific scheme: a method for monitoring and diagnosing early faults of a rolling bearing comprises the steps of collecting vibration signals when the bearing runs, dividing the collected vibration signals into a plurality of signal segments by using a non-overlapped sliding window, and decomposing each signal segment by using a wavelet packet decomposition technology with self-adaptive time-frequency resolution to obtain wavelet packet coefficients of each segment in different frequency bands; and carrying out graph modeling on each wavelet packet coefficient sequence by utilizing a graph model capable of capturing data correlation to form a wavelet packet coefficient graph set, and carrying out fault monitoring and fault diagnosis according to the graph modeling.

Description

Early fault monitoring and diagnosing method and system for rolling bearing
Technical Field
The invention belongs to the technical field of state monitoring and fault diagnosis in a mechanical system, and particularly relates to a rolling bearing-oriented early fault monitoring and 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 early fault monitoring and diagnosis of the rolling bearing is a hot spot in the research of the current industrial field, particularly, along with the development of the intelligent technology, the automation degree of mechanical equipment is higher and higher, and the rolling bearing is a key component of the mechanical equipment, so that higher requirements are also put forward on the running state monitoring of the rolling bearing. The inventor finds that if the signals are processed by the current common frequency domain monitoring technology, effective fault information can be diluted by the whole frequency band, so that the early faults are missed to be detected, and the obvious faults of the rolling bearing can be detected, so that economic property loss is caused in serious cases; meanwhile, in the fault diagnosis stage, a dynamic model needs to be accurately established for the structure of the bearing, and the structural parameters of the dynamic model need to be very accurate in the establishing process, otherwise, the establishing precision of the model is influenced, and the models established by different bearings are greatly different, which is not beneficial to the establishment of an intelligent and automatic system; if the short-time Fourier transform of the time-frequency method is adopted, the weak information of the early fault is also influenced by the time-frequency resolution, the effective fault information can be weakened for the fixed time-frequency resolution, the detection omission or the delayed detection of the early fault is caused, and the economic loss is caused.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a rolling bearing-oriented early fault monitoring and diagnosis method, which can enhance weak characteristic information in an early fault signal and reduce the probability of missed detection; and the signals obtained by monitoring are analyzed and processed, the running state of the rolling bearing is monitored on line, the rolling bearing is timely and accurately monitored when an early fault occurs, early warning is given, and then the fault type of the early fault is judged.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a method for monitoring and diagnosing early faults of a rolling bearing is characterized by collecting vibration signals when the bearing runs, dividing the collected vibration signals into a plurality of signal segments by using a non-overlapped sliding window, and decomposing each signal segment by using a wavelet packet decomposition technology with self-adaptive time-frequency resolution to obtain wavelet packet coefficients of each segment in different frequency bands; and carrying out graph modeling on each wavelet packet coefficient sequence by utilizing a graph model capable of capturing data correlation to form a wavelet packet coefficient graph set, and carrying out fault monitoring and fault diagnosis according to the graph modeling.
The method for monitoring and diagnosing the early fault of the rolling bearing comprises the following steps: the wavelet packet coefficient graph on each frequency band carries out similarity measurement according to time sequence to obtain a similarity score, and the multi-dimensional similarity scores are fused into a one-dimensional abnormal score so as to monitor multiple faults through the one-dimensional abnormal score; and evaluating the health status of the abnormal degree score sequence to confirm whether early faults occur.
According to the rolling bearing early fault monitoring and diagnosis oriented method, the multidimensional similarity scores are fused into the multidimensional abnormal score through an adaptive multidimensional sequence fusion algorithm.
According to the rolling bearing early fault monitoring and diagnosing method, the health state of the abnormal degree score sequence is evaluated by using hypothesis testing.
The early fault monitoring and diagnosing method for the rolling bearing comprises the following steps: and after the abnormal state is monitored, further extracting fault resolution characteristics of the wavelet packet coefficient atlas of the signal segment under the monitored abnormal state, constructing a singular value characteristic sequence, judging the obtained characteristics to be diagnosed and the characteristics of the training set, and diagnosing the type of the fault.
According to the monitoring and diagnosis method for the early failure of the rolling bearing, the failure resolution feature extraction comprises singular value extraction, singular value extraction is carried out on each graph model of a signal segment by adopting a singular value decomposition technology, and a singular value feature sequence is constructed according to ordinal numbers.
A face scrolling as described aboveThe method for monitoring and diagnosing the early failure of the bearing utilizes a graph model capable of capturing data correlation to carry out graph modeling on each wavelet packet coefficient sequence and comprises the following steps of carrying out graph modeling on any two coefficients in the wavelet packet coefficient sequence
Figure BDA0002545075060000021
And
Figure BDA0002545075060000022
as the vertex of the graph model, a line connecting the two vertices is taken as an edge l (p, q) of the graph model; calculating two coefficients
Figure BDA0002545075060000023
And
Figure BDA0002545075060000024
the Euclidean distance therebetween is taken as the weight d (p, q) of the edge l (p, q); graph model constructed by wavelet packet coefficient sequence
Figure BDA0002545075060000025
Represented by adjacency matrix α, where α ═ { d (p, q) };
and modeling each wavelet packet coefficient in the wavelet packet coefficient sequence set by adopting the graph modeling mode to obtain the corresponding wavelet coefficient graph set.
The method for monitoring and diagnosing early faults of the rolling bearing comprises the following steps:
in order to eliminate the difference of the scales between the multidimensional data, the data is normalized, and the similarity is determined according to the following formula
Figure BDA0002545075060000026
Carrying out standardization processing to obtain ith standardized data of jth dimension
Figure BDA0002545075060000027
Figure BDA0002545075060000028
Wherein, min { s }3,jAnd max s3,jAre multidimensional similarity scores, respectively
Figure BDA0002545075060000031
Medium and maximum values, k being the kth signal segment monitored;
according to the formula of the variance, the data is obtained,
Figure BDA0002545075060000032
calculating weights w for adaptive multi-dimensional fusion3，j,
Figure BDA0002545075060000033
Wherein the content of the first and second substances,
Figure BDA0002545075060000034
the mean value of the ith normalized data in the jth dimension;
according to the obtained weight of the multi-dimensional fusion, the similarity scores of the multiple dimensions are fused into an abnormal score q of the dimensionk：
Figure BDA0002545075060000035
Converting the multidimensional similarity score into a one-dimensional abnormal degree score (q)1,q2,....,qk}。
According to the rolling bearing-oriented early fault monitoring and diagnosis method, singular value extraction is performed on each graph model of the signal segment by adopting a singular value decomposition technology, and the method comprises the following steps:
for any one of the graph models
Figure BDA0002545075060000036
Performing singular value decomposition
Figure BDA0002545075060000037
Wherein U and V are respectively a left singular matrix and a right singular matrix, U1Is the 1 st row vector, v, of the left singular matrix U1Is the 1 st column vector of the right singular matrix V, ∑ is the diagonal matrix, and the diagonal element [ sigma ]1,σ2,...,σn]The singular value sequence after decomposition is obtained, and n is a natural number; selecting the first 10% singular value section
Figure BDA0002545075060000038
Constructing a singular value characteristic sequence;
performing the above operation on each graph model in the wavelet packet coefficient graph set to obtain a plurality of singular value segments, and performing series arrangement on the singular value segments according to the sequence in the wavelet packet coefficient graph set to obtain a singular value characteristic sequence
Figure BDA0002545075060000039
According to the rolling bearing-oriented early fault monitoring and diagnosis method, the fault type is judged on the singular value feature sequence through a KNN classifier (nearest neighbor classifier).
In a second aspect, the invention further provides an early fault monitoring and diagnosis system for the rolling bearing, which comprises a rolling bearing state signal acquisition module, a rolling bearing state signal modeling module, a rolling bearing fault monitoring module and a rolling bearing fault diagnosis module.
The rolling bearing state signal acquisition module is used for acquiring state signals of the rolling bearing to be monitored;
the rolling bearing state signal modeling module is connected with the rolling bearing state signal acquisition module and decomposes the acquired rolling bearing state signal by using a wavelet packet decomposition method to acquire a wavelet packet coefficient; the rolling bearing state signal modeling module utilizes a graph model to model the wavelet packet coefficient to form a wavelet packet coefficient graph set so as to represent the state information of the rolling bearing;
the rolling bearing fault monitoring module is connected with the rolling bearing state signal modeling module and is used for measuring the similarity of the wavelet packet coefficient diagram on each frequency band to obtain a similarity score; the multi-dimensional similarity scores are fused into the abnormal degree scores of the dimensions; the evaluation of the health state of the rolling bearing is realized by adopting an abnormal decision detection algorithm based on hypothesis testing on the abnormal degree score sequence;
and (3) fusing the multidimensional similarity scores into the multidimensional abnormal degree scores by utilizing an adaptive multidimensional sequence fusion algorithm.
The rolling bearing fault diagnosis module is connected with the rolling bearing fault monitoring module, and constructs a singular value characteristic sequence as a fault resolution characteristic for a wavelet packet coefficient atlas of a signal segment in an abnormal state; this feature is fed into a classifier to identify the type of rolling bearing failure.
The beneficial effects of the invention are as follows:
1) according to the invention, a graph model is adopted to model the wavelet packet coefficients, the correlation among the wavelet packet coefficients is considered, the integrity of the wavelet packet coefficients is described, the expressive force of the early fault weak signal characteristics is increased, and compared with other enhancement modes of wavelet packet technology signal processing, such as kurtosis, energy, entropy and multi-scale entropy arrangement, the data correlation can be well described, and the wavelet packet technology is enhanced; the method has the advantages that the mining and the expression of data information among wavelet packet coefficients and the whole data information are more complete, weak characteristic information in early fault signals can be enhanced, the probability of missed detection is reduced, and the subsequent monitoring and diagnosis are more accurate and comprehensive.
2) The invention adopts the self-adaptive multi-dimensional sequence fusion algorithm to fuse the information on a single frequency band into a monitoring index, and can more intelligently meet the problem of comprehensively monitoring the diversified faults in the actual industry; compared with the existing frequency domain analysis method, the method has the advantages that the fault is diagnosed in a specific frequency range, the method does not need prior structural characteristic frequency, and the monitoring of a plurality of frequency bands can be realized.
3) The method adopts a singular value decomposition technology to extract the characteristic information of the graph model, can completely retain the characteristic information of the graph model, can simplify the complexity of calculation, quickens the early fault diagnosis speed, and cannot cause delay; compared with other diagnostic methods, the method has higher diagnostic precision under a smaller training sample.
4) The method provided by the invention can be used for timely and effectively monitoring the running state of the vehicle on line, sending out early warning indication when early fault symptoms occur, and judging the fault type of the early fault, so that relevant operators can conveniently make a reasonable maintenance scheme, and the economic loss caused by shutdown is reduced.
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 incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is an overall flowchart of a rolling bearing-oriented early failure monitoring and diagnosis method provided in embodiment 1 of the present invention;
FIG. 2 is a diagram of an example of applying an adaptive multi-dimensional algorithm to a simulation signal in embodiment 1 of the present invention;
FIG. 3 is a data analysis chart of an example of early failure monitoring obtained in a specific experiment in embodiment 1 of the present invention;
FIG. 4 is a three-dimensional visual representation of the early failure diagnosis resolution feature set obtained in the specific experiment in example 1 of the present invention;
fig. 5 is a graph showing the results of early failure diagnosis obtained in a specific experiment in example 1 of the present invention.
Fig. 6 is a schematic flow chart of an early failure monitoring and diagnosing system for a rolling bearing according to embodiment 2 of the present invention;
Detailed Description
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/or "the" are intended to include the plural forms as well, unless the invention expressly state otherwise, 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;
for convenience of description, the words "up", "down", "left" and "right" in the present invention, if any, merely indicate correspondence with up, down, left and right directions of the drawings themselves, and do not limit the structure, but merely facilitate the description of the invention and simplify the description, rather than indicate or imply that the referenced device or element must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the invention.
As introduced by the background technology, the problem of early fault missing detection exists in the prior art, and in order to solve the technical problem, the invention provides a method for monitoring and diagnosing early faults of a rolling bearing.
Example 1
In an exemplary embodiment of the present invention, a method for monitoring and diagnosing early failure of a rolling bearing, which uses the rolling bearing as a monitoring object, includes the following steps:
step 1, collecting vibration signals X of a monitored bearing by using a vibration acceleration sensor, and dividing the collected vibration signals into a plurality of signal fragments { X by adopting a non-overlapped sliding window1,X2,X3,...,Xk}; performing 3-layer wavelet packet decomposition on each signal segment to obtain the k signal segment XkFor example, the sequence can be decomposed into wavelet packet coefficient sequence sets
Figure BDA0002545075060000061
Wherein
Figure BDA0002545075060000062
The third layer, the jth (j ═ 0,1,..,7) wavelet packet coefficient sequence.
Step 2, utilizing a graph model to obtain the wavelet packet coefficient sequence
Figure BDA0002545075060000063
Carrying out graph modeling; any two coefficients in the wavelet packet coefficient sequence
Figure BDA0002545075060000064
And
Figure BDA0002545075060000065
as the vertex of the graph model, a line connecting the two vertices is taken as an edge l (p, q) of the graph model; then two coefficients are calculated
Figure BDA0002545075060000066
And
Figure BDA0002545075060000067
the Euclidean distance therebetween is taken as the weight d (p, q) of the edge l (p, q); graph model constructed by wavelet packet coefficient sequence
Figure BDA0002545075060000068
May be represented by adjacency matrix α, where α ═ { d (p, q) }.
Wavelet packet coefficient sequence set by adopting the graph modeling mode
Figure BDA0002545075060000069
Each wavelet packet coefficient in the wavelet packet coefficient map is modeled to obtain a corresponding wavelet coefficient map set composed of wavelet packet coefficient maps
Figure BDA00025450750600000610
Early fault monitoring:
step 3, for a wavelet packet coefficient graph model of a certain specific frequency band, the graph model has similar characteristic information in time sequence, namely the graph model of the jth sub-frequency band in time sequence is
Figure BDA00025450750600000611
Whether an abnormality occurs in the frequency band can be judged by measuring a similarity score between the two graph models. Similarity score at a particular frequency band for k segments
Figure BDA00025450750600000612
Can be obtained by the following calculation,
Figure BDA00025450750600000613
n denotes the dimension of the graph model, which is defined by the above-mentioned sequence of similarity of the specific frequency bands
Figure BDA00025450750600000614
It is possible to evaluate whether or not an abnormality has occurred in the frequency band.
And 4, in order to meet the requirement that various faults occur in practical industrial application, because different types of faults are reflected on different frequency bands, when some faults are monitored by adopting frequency domains, the working condition and the corresponding structural model of the rolling bearing need to be given, and the fault monitoring is time-consuming and needs to be completed by professionals. Therefore, the embodiment provides a self-adaptive multi-dimensional fusion algorithm which fuses multi-dimensional similarity scores into the dimensional abnormal score without researching the working condition and structural characteristics of mechanical equipment, and further realizes the intellectualization of early fault monitoring.
The specific calculation of the adaptive multi-dimensional fusion algorithm is as follows:
(1) in order to eliminate the difference of the scales between the multi-dimensional data, the data is normalized, and the similarity score series is processed
Figure BDA0002545075060000071
Carrying out standardization processing to obtain ith standardized data of jth dimension
Figure BDA0002545075060000072
Figure BDA0002545075060000073
Wherein, min { s }3,jAnd max s3,jAre multidimensional similarity scores, respectively
Figure BDA0002545075060000074
Medium and maximum values, k being the kth signal segment monitored.
(2) The variance is an important index capable of reflecting the fluctuation condition of the data, and here, the smaller variance is considered to play a more important role in the subsequent decision, and the calculation formula is,
Figure BDA0002545075060000075
wherein the content of the first and second substances,
Figure BDA0002545075060000076
the mean value of the ith normalized data in the jth dimension;
(3) the weight w of the adaptive multi-dimensional fusion can be calculated according to the variance3，j,
Figure BDA0002545075060000077
(4) According to the multidimensional weight obtained by calculation, the multidimensional similarity score q can be fused into a multidimensional abnormal degree score qk
Figure BDA0002545075060000078
And finally converting the multidimensional similarity score into a one-dimensional abnormality score q1,q2,....,qk}。
And 5, adopting an abnormal decision based on hypothesis test to detect whether an abnormal state is generated or not for the abnormal degree score, wherein the detection process is as follows:
H0:qk≤|μk-1±3σk-1|
H1:qk>|μk-1±3σk-1|
wherein H0Indicates that no abnormal state has occurred, and H1Indicating the occurrence of an abnormal state; mu.sk-1And σk-1Mean and variance of the first k-1 outlier scores, respectively.
Fig. 2 shows a multidimensional similarity score and a one-dimensional abnormality score of a simulation signal, and a fault point from a state 1 normal state to a state 2 fault state is detected on the abnormality score.
Fig. 3 shows an example of early fault monitoring of a measured signal from a database of life-cycle bearing experiments from the university of siemens traffic, which shows that the method can detect a fault early and timely and give a corresponding indication.
Diagnosis of early failure:
step 6, after the occurrence of the fault is monitored, a wavelet packet coefficient atlas of a signal segment of which the abnormal point is detected is subjected to
Figure BDA0002545075060000081
And further extracting fault discrimination characteristics. Because the graph model is a two-dimensional feature, directly judging multiple graphs increases the complexity of recognition, which is not favorable for timely judging the type of a fault after an early fault point is monitored. Because the graph models are adjacency matrixes and the singular value decomposition can keep original characteristic information, the singular value decomposition technology is used for decomposing each graph model in the graph set, extracting a singular value sequence of each graph model and constructing the singular value characteristic sequence. It is calculated specifically as follows:
(1) for any one of the graph models
Figure BDA0002545075060000082
Performing singular value decomposition
Figure BDA0002545075060000083
Where U and V are the left and right singular matrices, respectively, ∑ is the diagonal matrix, and the diagonal element [ sigma ]1,σ2,...,σn]Is a decomposed singular value sequence. Since the singular value has information concentrated in the front part, the singular value segment of the top 10% (m ═ 0.1n) is selected
Figure BDA0002545075060000084
And constructing a singular value characteristic sequence.
(2) By performing the above operations on each graph model in the graph set, 8 singular value segments can be obtained, and the singular value segments are arranged in series according to the sequence in the graph set to obtain a singular value feature sequence
Figure BDA0002545075060000085
And the obtained singular value characteristic sequence is sent to a classifier for judging the fault type.
Fig. 4 shows a visualization diagram of a singular value feature sequence extracted in an early fault, which can clearly see that the same class is gathered together, but the different classes are separated greatly, which indicates that the feature sequence has strong separability, and can accurately and effectively diagnose the early fault. The early failure experimental data for this section is from the university of Kaiser storage digital bearing database.
And 7, judging the fault type of the singular value characteristic sequence by using the simplest and most convenient KNN classifier, wherein the types of the classifiers are multiple.
Fig. 5 shows the results of monitoring early failures in a 10% training sample, and it can be clearly seen that the proposed method can identify the type of failure with an accuracy of 100%.
Example 2
Referring to fig. 6, the early fault monitoring and diagnosis system for the rolling bearing comprises a rolling bearing state signal acquisition module, a rolling bearing state signal modeling module, a rolling bearing fault monitoring module and a rolling bearing fault diagnosis module.
The rolling bearing state signal acquisition module is used for acquiring state signals of the rolling bearing to be monitored;
the rolling bearing state signal modeling module is connected with the rolling bearing state signal acquisition module, and firstly decomposes the acquired rolling bearing state signal by using a wavelet packet decomposition method to acquire a wavelet packet coefficient; and carrying out graph modeling on the obtained wavelet packet coefficient on each frequency band by using a graph model to obtain a graph set containing a plurality of graph models so as to represent the state information of the rolling bearing.
The rolling bearing fault monitoring module is connected with the rolling bearing state signal modeling module, and carries out similarity measurement on the wavelet packet coefficient diagram on each frequency band according to time sequence to obtain a similarity score; then, a self-adaptive multi-dimensional sequence fusion algorithm is used for fusing the multi-dimensional similarity scores into a dimensional abnormal score, namely an index for monitoring the health state of the rolling bearing; and then, evaluating the abnormal degree score sequence by adopting an abnormal decision detection algorithm based on hypothesis testing so as to realize the evaluation of the health state of the rolling bearing and complete the monitoring of the faults of the rolling bearing.
The rolling bearing fault diagnosis module is connected with the rolling bearing fault monitoring module, the rolling bearing fault diagnosis module further extracts fault characteristics of the monitored map set, firstly, each map model in the map set is decomposed by using a singular value decomposition algorithm to obtain a singular value of each map model, and the first 10% of the map models are used for reconstructing fault resolution characteristics, namely singular value sequence characteristics; this feature is fed into the KNN classifier to identify the type of rolling bearing failure.
The rolling bearing state signal acquisition module comprises a vibration acceleration sensor, a displacement sensor and the like.
The rolling bearing state signal modeling module, the rolling bearing fault monitoring module and the rolling bearing fault diagnosis module are integrated in a controller, the controller is connected with the rolling bearing state signal acquisition module, and the controller can be an industrial personal computer or other types of controllers.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement 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 rolling bearing-oriented early fault monitoring and diagnosis method is characterized in that vibration signals generated when a bearing runs are collected, the collected vibration signals are divided into a plurality of signal segments through non-overlapping sliding windows, and each signal segment is decomposed through a wavelet packet decomposition technology with self-adaptive time-frequency resolution, so that wavelet packet coefficients of each segment in different frequency bands are obtained; and carrying out graph modeling on each wavelet packet coefficient sequence by utilizing a graph model capable of capturing data correlation to form a wavelet packet coefficient graph set, and carrying out fault monitoring and fault diagnosis according to the graph modeling.
2. The early failure monitoring and diagnosis method for rolling bearings according to claim 1, characterized in that the failure monitoring comprises the following: the wavelet packet coefficient graph on each frequency band carries out similarity measurement according to time sequence to obtain a similarity score, and the multi-dimensional similarity scores are fused into a one-dimensional abnormal score so as to monitor multiple faults through the one-dimensional abnormal score; and evaluating the health status of the abnormal degree score sequence to confirm whether early faults occur.
3. The rolling bearing-oriented early fault monitoring and diagnosis method according to claim 2, wherein the multidimensional similarity scores are fused into one-dimensional abnormality scores by an adaptive multi-dimensional sequence fusion algorithm.
4. The rolling bearing-oriented early fault monitoring and diagnosis method according to claim 2, wherein the abnormality degree score sequence is evaluated for health using hypothesis testing.
5. The early failure monitoring and diagnosis method for a rolling bearing according to claim 1 or 2, characterized in that the failure diagnosis comprises the following: and after the abnormal state is monitored, further extracting fault resolution characteristics of the wavelet packet coefficient atlas of the signal segment under the monitored abnormal state, constructing a singular value characteristic sequence, judging the obtained characteristics to be diagnosed and the characteristics of the training set, and diagnosing the type of the fault.
6. The early failure monitoring and diagnosis method for rolling bearings according to claim 5, wherein the failure resolution feature extraction comprises singular value extraction, the singular value extraction is performed on each graph model of the signal segment by using a singular value decomposition technique, and a singular value feature sequence is constructed according to ordinal numbers.
7. The rolling bearing-oriented early fault monitoring and diagnosing method as claimed in claim 1, wherein the graph modeling of each wavelet packet coefficient sequence by using a graph model capable of capturing data correlation comprises the step of performing graph modeling on any two coefficients in the wavelet packet coefficient sequence
Figure FDA0002545075050000011
And
Figure FDA0002545075050000012
as the vertex of the graph model, a line connecting the two vertices is taken as an edge l (p, q) of the graph model; calculating two coefficients
Figure FDA0002545075050000013
And
Figure FDA0002545075050000014
the Euclidean distance therebetween is taken as the weight d (p, q) of the edge l (p, q); graph model constructed by wavelet packet coefficient sequence
Figure FDA0002545075050000015
Represented by adjacency matrix α, where α ═ { d (p, q) }.
8. The rolling bearing-oriented early fault monitoring and diagnosis method according to claim 3, wherein the algorithm of the adaptive multi-dimensional sequence fusion comprises the following contents:
in order to eliminate the difference of the scales between the multidimensional data, the data is normalized, and the similarity is determined according to the following formula
Figure FDA0002545075050000021
And (3) carrying out standardization treatment:
Figure FDA0002545075050000022
where min { s }3,jAnd max s3,jAre multidimensional similarity scores, respectively
Figure FDA0002545075050000023
Medium and maximum values;
according to the formula of the variance, the data is obtained,
Figure FDA0002545075050000024
calculating weights w for adaptive multi-dimensional fusion3，j,
Figure FDA0002545075050000025
According to the obtained weight of the multi-dimensional fusion, the similarity scores of the multiple dimensions are fused into an abnormal score q of the dimensionk：
Figure FDA0002545075050000026
9. The rolling bearing-oriented early fault monitoring and diagnosis method according to claim 6, wherein singular value extraction is performed on each graph model of the signal segment by using a singular value decomposition technique, and the method comprises the following steps:
for any one of the graph models
Figure FDA0002545075050000027
Performing singular value decomposition
Figure FDA0002545075050000028
Where U and V are the left and right singular matrices, respectively, ∑ is the diagonal matrix, and the diagonal element [ sigma ]1,σ2,...,σn]Decomposing the singular value sequence; selecting the first 10% singular value section
Figure FDA0002545075050000029
Constructing a singular value characteristic sequence;
performing the above operation on each graph model in the wavelet packet coefficient graph set to obtain a plurality of singular value segments, and performing series arrangement on the singular value segments according to the sequence in the wavelet packet coefficient graph set to obtain a singular value characteristic sequence
Figure FDA00025450750500000210
10. An early fault monitoring and diagnosing system for a rolling bearing, comprising:
the rolling bearing state signal acquisition module is used for acquiring state signals of the rolling bearing to be monitored;
the rolling bearing state signal modeling module is connected with the rolling bearing state signal acquisition module and decomposes the acquired rolling bearing state signal by using a wavelet packet decomposition method to acquire a wavelet packet coefficient; the rolling bearing state signal modeling module utilizes a graph model to model the wavelet packet coefficient to form a wavelet packet coefficient graph set so as to represent the state information of the rolling bearing;
the rolling bearing fault monitoring module is connected with the rolling bearing state signal modeling module and is used for measuring the similarity of the wavelet packet coefficient diagram on each frequency band to obtain a similarity score; the multi-dimensional similarity scores are fused into the abnormal degree scores of the dimensions; the evaluation of the health state of the rolling bearing is realized by adopting an abnormal decision detection algorithm based on hypothesis testing on the abnormal degree score sequence;
the rolling bearing fault diagnosis module is connected with the rolling bearing fault monitoring module, and constructs a singular value characteristic sequence as a fault resolution characteristic for a wavelet packet coefficient atlas of a signal segment in an abnormal state; this feature is fed into a classifier to identify the type of rolling bearing failure.
CN202010557768.4A 2020-06-18 2020-06-18 Early fault monitoring and diagnosing method and system for rolling bearing Active CN111678699B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010557768.4A CN111678699B (en) 2020-06-18 2020-06-18 Early fault monitoring and diagnosing method and system for rolling bearing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010557768.4A CN111678699B (en) 2020-06-18 2020-06-18 Early fault monitoring and diagnosing method and system for rolling bearing

Publications (2)

Publication Number Publication Date
CN111678699A true CN111678699A (en) 2020-09-18
CN111678699B CN111678699B (en) 2021-06-04

Family

ID=72455518

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010557768.4A Active CN111678699B (en) 2020-06-18 2020-06-18 Early fault monitoring and diagnosing method and system for rolling bearing

Country Status (1)

Country Link
CN (1) CN111678699B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112182773A (en) * 2020-10-16 2021-01-05 北京航天自动控制研究所 Online identification method for aircraft steering engine fault based on linear frequency modulation Z transformation
CN113310693A (en) * 2021-06-07 2021-08-27 华润电力技术研究院有限公司 Mechanical fault detection method, device, equipment and storage medium
CN113670611A (en) * 2021-07-23 2021-11-19 美核电气（济南）股份有限公司 Bearing early degradation evaluation method, system, medium and electronic equipment
CN117630679A (en) * 2023-11-30 2024-03-01 湖北工业大学 Battery fault diagnosis method and system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108426713A (en) * 2018-02-26 2018-08-21 成都昊铭科技有限公司 Rolling bearing Weak fault diagnostic method based on wavelet transformation and deep learning
CN108960328A (en) * 2018-07-04 2018-12-07 温州大学 A method of the diagnosis automobile hub bearing failure based on Hankel matrix
CN109858104A (en) * 2019-01-10 2019-06-07 山东大学 A kind of rolling bearing health evaluating and method for diagnosing faults and monitoring system
CN110057588A (en) * 2019-05-09 2019-07-26 山东大学 The bearing incipient fault detection and diagnostic method and system merged based on singular value with characteristics of graph theory
WO2020095303A1 (en) * 2018-11-09 2020-05-14 Augury Systems Ltd. Automated analysis of non-stationary machine performance
CN111259330A (en) * 2020-01-13 2020-06-09 北京化工大学 Rotary machine early warning method based on vibration signal

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108426713A (en) * 2018-02-26 2018-08-21 成都昊铭科技有限公司 Rolling bearing Weak fault diagnostic method based on wavelet transformation and deep learning
CN108960328A (en) * 2018-07-04 2018-12-07 温州大学 A method of the diagnosis automobile hub bearing failure based on Hankel matrix
WO2020095303A1 (en) * 2018-11-09 2020-05-14 Augury Systems Ltd. Automated analysis of non-stationary machine performance
CN109858104A (en) * 2019-01-10 2019-06-07 山东大学 A kind of rolling bearing health evaluating and method for diagnosing faults and monitoring system
CN110057588A (en) * 2019-05-09 2019-07-26 山东大学 The bearing incipient fault detection and diagnostic method and system merged based on singular value with characteristics of graph theory
CN111259330A (en) * 2020-01-13 2020-06-09 北京化工大学 Rotary machine early warning method based on vibration signal

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
AKOGLUL等: "Graphbasedanomalydetectionanddescription;asurvey", 《DATAMININGANDKNOWLEDGEDISCOVERY》 *
曾世琛: "基于内相关信息数据挖掘的滚动轴承健康状态监测技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112182773A (en) * 2020-10-16 2021-01-05 北京航天自动控制研究所 Online identification method for aircraft steering engine fault based on linear frequency modulation Z transformation
CN113310693A (en) * 2021-06-07 2021-08-27 华润电力技术研究院有限公司 Mechanical fault detection method, device, equipment and storage medium
CN113670611A (en) * 2021-07-23 2021-11-19 美核电气（济南）股份有限公司 Bearing early degradation evaluation method, system, medium and electronic equipment
CN117630679A (en) * 2023-11-30 2024-03-01 湖北工业大学 Battery fault diagnosis method and system

Also Published As

Publication number Publication date
CN111678699B (en) 2021-06-04

Similar Documents

Publication Publication Date Title
CN111678699B (en) Early fault monitoring and diagnosing method and system for rolling bearing
CN113469060A (en) Multi-sensor fusion convolution neural network aeroengine bearing fault diagnosis method
CN104712542B (en) A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults
Kwan et al. A novel approach to fault diagnostics and prognostics
KR101316486B1 (en) Error detection method and system
CN108444708A (en) The method for building up of rolling bearing intelligent diagnostics model based on convolutional neural networks
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN111505424A (en) Large experimental device power equipment fault diagnosis method based on deep convolutional neural network
CN111914883A (en) Spindle bearing state evaluation method and device based on deep fusion network
CN115688018B (en) Method for monitoring state and diagnosing faults of bearing under multiple working conditions
JP7340265B2 (en) Abnormality detection device, abnormality detection method, and program
JP2012018066A (en) Device for inspecting abnormality
Son et al. Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge
CN113391207A (en) Motor fault detection method, medium and system
CN110553789A (en) state detection method and device of piezoresistive pressure sensor and brake system
CN110108474B (en) Online monitoring and evaluating method and system for operation stability of rotary machine
Neupane et al. Deep learning-based bearing fault detection using 2-D illustration of time sequence
Lu et al. Early fault warning and identification in condition monitoring of bearing via wavelet packet decomposition coupled with graph
KR20210006832A (en) Method and apparatus for machine fault diagnosis
CN110057588B (en) Bearing early fault detection and diagnosis method and system based on fusion of singular value and graph theory characteristics
Jianhui et al. New fault diagnosis method for rolling bearing based on PCA
CN114279728B (en) Fault diagnosis method and system for vibrating screen body
CN112069621B (en) Method for predicting residual service life of rolling bearing based on linear reliability index
CN114266013A (en) Deep learning virtual perception network-based transmission system vibration decoupling method
Tang et al. Fault Diagnosis of Rolling Bearing Based on BP Nural Network Technology

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