CN108776017A - A kind of rolling bearing method for predicting residual useful life improving CHSMM - Google Patents
A kind of rolling bearing method for predicting residual useful life improving CHSMM Download PDFInfo
- Publication number
- CN108776017A CN108776017A CN201810325011.5A CN201810325011A CN108776017A CN 108776017 A CN108776017 A CN 108776017A CN 201810325011 A CN201810325011 A CN 201810325011A CN 108776017 A CN108776017 A CN 108776017A
- Authority
- CN
- China
- Prior art keywords
- state
- chsmm
- bearing
- useful life
- predicting residual
- 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
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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of rolling bearing method for predicting residual useful life improving CHSMM, it is characterized in that：The first time domain of extraction bear vibration data and the feature vector of time-frequency domain, and dimensionality reduction is carried out to feature vector using PCA algorithms；Then each degenerate state data are obtained using k-means algorithms, establishes degenerate state identification model, and predicting residual useful life model is established using bearing Life cycle data；The low problem of predicting residual useful life precision caused by reality is not met for state duration probability density function, Gaussian-mixture probability density function is introduced into CHSMM；State duration probability distribution can preferably be approached based on the predicting residual useful life model for improving CHSMM foundation compared to the predicting residual useful life model established based on original CHSMM, so as to more accurately predict the remaining life of bearing.
Description
Technical field
The invention belongs to bearing residual lifes to predict field, and in particular to a kind of rolling bearing improving CHSMM remaining longevity
Order prediction technique.
Background technology
As industry and the continuous improvement of scientific and technological level, mechanical equipment are constantly changed in complicated, efficient, light-duty etc.
Into, while also facing harsher working environment.Once the critical component of equipment breaks down, it is possible to entire life can be influenced
The problems such as production process causes huge economic loss, results even in casualties.Therefore, maintenance of equipment is just by traditional thing
To the condition maintenarnce transformation based on state, and as the premise for establishing rational maintenance strategy, equipment is surplus for repair and planned maintenance afterwards
Remaining life prediction also begins to be concerned.
Rolling bearing directly affects whole and sets as one of the key components and parts in rotating machinery, the quality of performance state
Standby operational reliability.In general, rolling bearing can be all undergone in use from normally to the mistake degenerated up to failure
Journey, and a series of different performance degradation states are usually undergone during this.If the mistake that can be degenerated in rolling bearing performance
The remaining life of bearing is monitored in journey, then can targetedly organize to produce and formulate rational maintenance plan, is prevented
The only generation of unit exception failure.
Currently, can be divided into for rolling bearing predicting residual useful life：Based on model, data-driven two major classes method.Base
It is mainly the physical arrangement according to bearing in the method for model, bearing is established using statistical principle or from the angle of mechanics
Life model, these methods need a large amount of expertise and more complex failure mechanism knowledge, limit its application range.Data
The method of driving is mainly the status data run according to bearing, using machine learning algorithm, is carried out to the remaining life of bearing
Prediction.The main method used has deep neural network (deep neuralnetworks, DNN), support vector machines
(supportvectormachine, SVM), Kalman filtering (the Kalman particle filtering
Algorithm, KF) and CHSMM (Continuous hidden semi-Markov model, continuous hidden semi-Markov mould
Type), wherein CHSMM is a dual random process, between the degenerative process and status data that can be very good description bearing
Relationship, many scholars apply it to bearing residual life prediction field.But during application CHSMM, assume its state
Residence time probability density function meets Gaussian Profile, and in practice, the true distribution function of state duration be it is unknown,
This hypothesis can reduce precision of prediction.
Invention content
The present invention proposes a kind of axis of rolling improving CHSMM to more accurately predict the remaining life of bearing
Hold method for predicting residual useful life.
To achieve the goals above, the present invention is achieved through the following technical solutions：
Step (1)：Bearing Life cycle vibration data is obtained, denoising is carried out and normalization pre-processes；Extract vibration number
According to time domain, time and frequency domain characteristics vector；
Step (2)：Using PCA (Principle component analysis, principal component analysis) algorithms to multiple domain spy
Sign vector carries out Feature Dimension Reduction；
Step (3)：The bearing Life cycle data that step (2) obtains are divided into five degenerate states, i.e., normal condition,
Degenerate state 1, degenerate state 2, degenerate state 3, degenerate state 4, are used in combination k-means algorithms to gather Life cycle data
Alanysis obtains each degenerate state data；
Step (4)：Gaussian-mixture probability density function is introduced into CHSMM, improved CHSMM is obtained, utilizes step
(3) each degenerate state data obtained train five degenerate state identification models, the state classifier as bearing；
Step (5)：The Life cycle data obtained using step (2) train a predicting residual useful life model, obtain
To the state transition probability of Life cycle, for testing data, using step (1), (2) institute extracting method extract its feature to
Amount, and be entered into the state classifier of step (4), the current degenerate state of bearing is obtained, remaining life is then utilized
Calculation formula calculates the current remaining life of bearing.
According to above technical solution, advantageous effect below may be implemented：
(1) present invention extraction bear vibration data time domain and the time-frequency characteristics based on wavelet packet, and carried out based on PCA algorithms
Feature Dimension Reduction reduces redundancy in high dimensional feature vector, accelerates the training speed of model；
(2) Gaussian-mixture probability density function is introduced into CHSMM by the present invention, compared to existing CHSMM, after improvement
Model preferably the probability distribution of state duration can be approached, so as to more accurately to the residue of bearing
Service life is predicted.
Description of the drawings
A kind of Fig. 1 flow diagrams for the rolling bearing method for predicting residual useful life improving CHSMM of the present invention.
Specific implementation mode
To make the object, technical solutions and advantages of the present invention etc. be more clearly understood, below in conjunction with example, and with reference to attached
Figure, the present invention is described in more detail.
As shown in Figure 1, a kind of rolling bearing method for predicting residual useful life improving CHSMM, the method includes following steps
Suddenly：
Step (1)：Extract time domain, the time and frequency domain characteristics vector of vibration data；
Using the Life cycle vibration data of one bearing 1,2,3 of operating mode, its temporal signatures RMS (root mean square), AM are extracted
(absolute mean), SMR (root amplitude), Kurtosis (kurtosis), Skewness (degree of skewness), Peak (peak value) are small using db8
Wave carries out three layers of WAVELET PACKET DECOMPOSITION to data, the normalized value of eight node energies is obtained, as time-frequency characteristics；
Step (2)：Feature Dimension Reduction is carried out to multi-domain characteristics vector using PCA algorithms, steps are as follows：
1) eigenvectors matrix X is subjected to zero averaging, and seeks its covariance matrix；
2) eigenvalue λ of covariance matrix is soughtiAnd corresponding feature vector ri(i=1,2, n) and, n is characterized
Vector dimension；
3) contribution rate k is sought：
4) feature vector is arranged in matrix from big to small by character pair value, k rows composition matrix J before taking；
5) Y=JX is the feature vector after dimensionality reduction is tieed up to k；
Step (3)：Degenerate state data acquisition；
The bearing Life cycle data that step (2) obtains are divided into five degenerate states, i.e. normal condition, degenerate state
1, degenerate state 2, degenerate state 3, degenerate state 4 are used in combination k-means algorithms to carry out clustering to Life cycle data,
Obtain each degenerate state data；
Step (4)：Gaussian-mixture probability density letter is introduced into CHSMM, improved CHSMM is obtained, utilizes step
(3) each degenerate state data obtained train five degenerate state identification models, and as the state classifier of bearing, step is such as
Under：
1) parameter of CHSMM is τ=(N, M, π, A, B, C), and concrete meaning is as follows：
①N：In model, the state number of Markov chain, N number of state is denoted as S1, S2..., SN,qtIndicate arbitrary t moment
Markov chain state in which；
②M：The corresponding possible observation number of each state, o in modeltIndicate the observation of arbitrary t moment；
③π：Initial time, the probability distribution over states of model, π=(π1,π2,···,πN), wherein πi=P (q1=Si)1
≤i≤N；
④A：State transition probability matrix, A=(aij)N×N, wherein aij=P (qt+1=Sj|qt=Si), 1≤i, j≤N, and
⑤B：Observation probability density function, B={ bj(ok), 1≤j≤N, 1≤k≤M }, bj(ok)=P (ok|qt=
Sj)；
⑥C：The probability density function of state duration D, C={ cj(d), 1≤j≤N, 1≤d≤E }, cj(d)=P (d |
qt=Sj), E is maximum residence time；
Assessment, decoding and the problem concerning study to be solved for CHSMM, it is proposed that corresponding algorithm:" forward-backward " algorithm
Solves evaluation problem, i.e., given observation sequence O and parameter τ calculates the probability of a certain observation sequence；Viterbi algorithm solves
Decoding problem, i.e., given observation sequence O and parameter τ, finds observation sequence optimal in some sense；Baum Welch algorithms
It solves problem concerning study, that is, gives a sequence of observations, it may be determined that a τ so that and P (O | τ) it is maximum；
Define forward variable αt(i)=P (o1,o2,···,ot,qt=i, qt+1≠i),Backward variable βt(i)=P (ot+1, ot+2, oT, | qt=i, qt+1≠
I),Then have：
For given parameters τ and observation sequence O=(o1,o2,···,oT), T is observation sequence length, can obtain P (O | τ)
Log expressions are：
Whereinδt(i, d)=P (qt=
i,d|O),1≤i,j≤N,1≤d≤T-1,1≤t≤
T；
2) parametric solution：
Above formula is sought into local derviation to variables A, π：
It for observation probability density B, is fitted using gauss hybrid models, model parameter expression formula is:N(ok|Ujg,Σjg) it is Gaussian probability-density function, G represents its number.To B
In each variable seek local derviation：
Wherein,⊙ representation vector dot products.
Its symbol is assumed during application CHSMM carries out predicting residual useful life to bearing for state duration D
Gaussian Profile is closed, and in practice, the true distribution function of state duration is unknown, and this hypothesis can reduce prediction essence
Degree；
To overcome the above disadvantages, the probability density using Gaussian-mixture probability density function as state duration D
Function, model expression are：F represents the number of Gaussian probability-density function；
The model parameter of state duration D derives as follows：
Wherein,
Using above-mentioned improvement CHSMM training algorithms, the state classifier of training bearing, for the feature vector newly inputted,
Be entered into 5 move back state model among, using anterior-posterior to algorithm, calculate each model output probability, wherein output probability most
Big model is the degenerate state that bearing is presently in；
Step (5)：Predicting residual useful life；
For the sample to be tested (can be obtained by step (4)) of its known degenerate state, bearing is obtained by following recurrence equation
Remaining life is (assuming that bearing is in degenerate state i, RULiIndicate that remaining life when bearing is in health status i, D (i) indicate
The residence time of i macrostate)：
When bearing is in health status N-1：
RULN-1=aN-1,N-1[D(N-1)+D(N)]+aN-1,ND(N)
When bearing is in health status N-2：
RULN-2=aN-2,N-2[D(N-2)+RULN-1]+aN-2,N-1RULN-1
When bearing is in health status i：
RULi=ai,i[D(i)+RULi+1]+ai,i+1RULi+1
Wherein,
Claims (2)
1. a kind of rolling bearing method for predicting residual useful life improving CHSMM, specifically includes following steps：
Step (1)：Bearing Life cycle vibration data is obtained, denoising is carried out and normalization pre-processes；Extract vibration data
Time domain, time and frequency domain characteristics vector；
Step (2)：Feature Dimension Reduction is carried out to multi-domain characteristics vector using PCA algorithms；
Step (3)：The bearing Life cycle data that step (2) obtains are divided into five degenerate states, i.e. normal condition, degeneration
State 1, degenerate state 2, degenerate state 3, degenerate state 4 are used in combination k-means algorithms to carry out cluster point to Life cycle data
Analysis, obtains each degenerate state data；
Step (4)：Gaussian-mixture probability density letter is introduced into CHSMM, improved CHSMM is obtained, is obtained using step (3)
To each degenerate state data train five degenerate state identification models, the state classifier as bearing；
Step (5)：The Life cycle data obtained using step (2) train a predicting residual useful life model, obtain complete
The state transition probability of life cycle extracts its feature vector for testing data using step (1), (2) institute extracting method, and
It is entered into the state classifier of step (4), obtains the current degenerate state of bearing, then utilize residual Life Calculation public
Formula calculates the current remaining life of bearing.
2. a kind of rolling bearing method for predicting residual useful life improving CHSMM according to claim 1, it is characterized in that：It is described
Improvement in step (4) to CHSMM, includes the following steps：
1) parameter of CHSMM is τ=(N, M, π, A, B, C), and concrete meaning is as follows：
①N：In model, the state number of Markov chain, N number of state is denoted as S1, S2..., SN,qtIndicate arbitrary t moment Markov
Chain state in which；
②M：The corresponding possible observation number of each state, o in modeltIndicate the observation of arbitrary t moment；
③π：Initial time, the probability distribution over states of model, π=(π1,π2,…,πN), wherein πi=P (q1=Si)1≤i≤N；
④A：State transition probability matrix, A=(aij)N×N, wherein aij=P (qt+1=Sj|qt=Si), 1≤i, j≤N, and
⑤B：Observation probability density function, B={ bj(ok), 1≤j≤N, 1≤k≤M }, bj(ok)=P (ok|qt=Sj)；
⑥C：The probability density function of state duration D, C={ cj(d), 1≤j≤N, 1≤d≤E }, cj(d)=P (d | qt=
Sj), E is maximum residence time；
Assessment, decoding and the problem concerning study to be solved for CHSMM, it is proposed that corresponding algorithm:" forward-backward " algorithm solution
Evaluation problem, i.e., given observation sequence O and parameter τ, calculates the probability of a certain observation sequence；Viterbi algorithm solves to understand
Code problem, i.e., given observation sequence O and parameter τ, finds observation sequence optimal in some sense；Baum Welch algorithms solve
Problem concerning study gives a sequence of observations, it may be determined that τ so that and P (O | τ) it is maximum；
Define forward variable αt(i)=P (o1,o2,…,ot,qt=i, qt+1≠i),Backward variable βt(i)=P (ot+1,ot+2,…,oT,|qt=i, qt+1≠ i),Then have：
For given parameters τ and observation sequence O=(o1,o2,…,oT), T is observation sequence length, can obtain the expression of P (O τ) logarithm
Formula is：
Whereinδt(i, d)=P (qt=i, d |
O), 1≤i,j≤N,1≤d≤T-1,1≤t≤T；
2) parametric solution：
It, which meets height, is assumed during application CHSMM carries out predicting residual useful life to bearing for state duration D
This distribution, and in practice, the true distribution function of state duration is unknown, and this hypothesis can reduce precision of prediction；
To overcome the above disadvantages, the probability density function using Gaussian-mixture probability density function as state duration D,
Model expression is：F represents the number of Gaussian probability-density function；
The model parameter of state duration D derives as follows：
Wherein,
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810325011.5A CN108776017B (en) | 2018-04-12 | 2018-04-12 | Method for predicting residual life of rolling bearing by improving CHSMM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810325011.5A CN108776017B (en) | 2018-04-12 | 2018-04-12 | Method for predicting residual life of rolling bearing by improving CHSMM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108776017A true CN108776017A (en) | 2018-11-09 |
| CN108776017B CN108776017B (en) | 2020-04-24 |
Family
ID=64033684
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810325011.5A Active CN108776017B (en) | 2018-04-12 | 2018-04-12 | Method for predicting residual life of rolling bearing by improving CHSMM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108776017B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472241A (en) * | 2018-11-14 | 2019-03-15 | 上海交通大学 | Combustion engine bearing remaining life prediction technique based on support vector regression |
| CN111597682A (en) * | 2020-04-14 | 2020-08-28 | 新疆大学 | Method for predicting remaining life of bearing of gearbox of wind turbine |
| CN111896254A (en) * | 2020-08-10 | 2020-11-06 | 山东大学 | Fault prediction system and method for variable-speed variable-load large rolling bearing |
| CN113298240A (en) * | 2021-07-27 | 2021-08-24 | 北京科技大学 | Method and device for predicting life cycle of servo drive system |
| EP4040134A4 (en) * | 2019-09-30 | 2023-10-18 | Osaka University | Remaining life prediction system, remaining life prediction device, and remaining life prediction program |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599920A (en) * | 2016-12-14 | 2017-04-26 | 中国航空工业集团公司上海航空测控技术研究所 | Aircraft bearing fault diagnosis method based on coupled hidden semi-Markov model |
-
2018
- 2018-04-12 CN CN201810325011.5A patent/CN108776017B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599920A (en) * | 2016-12-14 | 2017-04-26 | 中国航空工业集团公司上海航空测控技术研究所 | Aircraft bearing fault diagnosis method based on coupled hidden semi-Markov model |
Non-Patent Citations (3)
| Title |
|---|
| YUAN ZHOU 等: "A Gaussian mixture model representation of endmember variability in hyperspectral unmixing", 《IEEE论文》 * |
| 李巍华 等: "连续隐半马尔科夫模型在轴承性能退化评估中的应用", 《振动工程学报》 * |
| 杜鹏: "基于谱聚类和CHSMM的非线性***的剩余寿命预测", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472241A (en) * | 2018-11-14 | 2019-03-15 | 上海交通大学 | Combustion engine bearing remaining life prediction technique based on support vector regression |
| EP4040134A4 (en) * | 2019-09-30 | 2023-10-18 | Osaka University | Remaining life prediction system, remaining life prediction device, and remaining life prediction program |
| CN111597682A (en) * | 2020-04-14 | 2020-08-28 | 新疆大学 | Method for predicting remaining life of bearing of gearbox of wind turbine |
| CN111597682B (en) * | 2020-04-14 | 2023-03-31 | 新疆大学 | Method for predicting remaining life of bearing of gearbox of wind turbine |
| CN111896254A (en) * | 2020-08-10 | 2020-11-06 | 山东大学 | Fault prediction system and method for variable-speed variable-load large rolling bearing |
| CN113298240A (en) * | 2021-07-27 | 2021-08-24 | 北京科技大学 | Method and device for predicting life cycle of servo drive system |
| CN113298240B (en) * | 2021-07-27 | 2021-11-05 | 北京科技大学 | Method and device for predicting life cycle of servo drive system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108776017B (en) | 2020-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108776017A (en) | A kind of rolling bearing method for predicting residual useful life improving CHSMM | |
| CN108737406B (en) | Method and system for detecting abnormal flow data | |
| CN106885697B (en) | The performance degradation assessment method of rolling bearing based on FCM-HMM | |
| Zhou et al. | Bearing fault recognition method based on neighbourhood component analysis and coupled hidden Markov model | |
| US20210334658A1 (en) | Method for performing clustering on power system operation modes based on sparse autoencoder | |
| CN111737911A (en) | Bearing recession trend prediction method for deep confidence network and extreme learning machine | |
| CN111459144A (en) | Airplane flight control system fault prediction method based on deep cycle neural network | |
| CN110417005B (en) | Transient stability serious fault screening method combining deep learning and simulation calculation | |
| Yang et al. | Enhanced hierarchical symbolic dynamic entropy and maximum mean and covariance discrepancy-based transfer joint matching with Welsh loss for intelligent cross-domain bearing health monitoring | |
| CN106647650A (en) | Distributed industrial process monitoring method based variable weighting PCA (Principal Component Analysis) model | |
| Bie et al. | An integrated approach based on improved CEEMDAN and LSTM deep learning neural network for fault diagnosis of reciprocating pump | |
| Zheng et al. | Real-time transient stability assessment based on deep recurrent neural network | |
| CN112199670A (en) | Log monitoring method for improving IFOREST (entry face detection sequence) to conduct abnormity detection based on deep learning | |
| CN110598955B (en) | Maximum instantaneous wind speed probability prediction method for high-speed train | |
| CN113225346A (en) | Network operation and maintenance situation assessment method based on machine learning | |
| Lee et al. | Hidden markov models for forex trends prediction | |
| Rahadian et al. | Image encoding selection based on Pearson correlation coefficient for time series anomaly detection | |
| CN115908842A (en) | Transformer partial discharge data enhancement and identification method | |
| Luo et al. | Multi-mode non-Gaussian variational autoencoder network with missing sources for anomaly detection of complex electromechanical equipment | |
| CN111475986B (en) | LSTM-AON-based gear residual life prediction method | |
| Kai et al. | Notice of Retraction: A Novel Forecasting Model of Fuzzy Time Series Based on K-means Clustering | |
| CN117194903A (en) | Network traffic data complement method and system based on generation of countermeasure network | |
| CN104200222B (en) | Object identifying method in a kind of picture based on factor graph model | |
| Wang et al. | Similarity-based echo state network for remaining useful life prediction | |
| Xiao et al. | An Integrated Approach Fusing CEEMD Energy Entropy and Sparrow Search Algorithm‐Based PNN for Fault Diagnosis of Rolling Bearings |
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 |