CN104268516A - Rolling bearing early failure recognition method - Google Patents
Rolling bearing early failure recognition method Download PDFInfo
- Publication number
- CN104268516A CN104268516A CN201410481861.6A CN201410481861A CN104268516A CN 104268516 A CN104268516 A CN 104268516A CN 201410481861 A CN201410481861 A CN 201410481861A CN 104268516 A CN104268516 A CN 104268516A
- Authority
- CN
- China
- Prior art keywords
- sample
- rolling bearing
- support vector
- feature
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/06—Recognition of objects for industrial automation
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to the field of mechanical device failure recognition, in particular to a rolling bearing early failure recognition method. The method includes the following steps of extracting nine time-domain parameters from rolling bearing data signals to form an original characteristic space, conducting characteristic compression on the original characteristic space through the LE algorithm, establishing an adjacent map through a neighbor method when characteristic extraction is conducted through the LE algorithm, endowing each edge with a weight through a thermonuclear equation to establish an adjacent weight matrix, and conducting failure classification and recognition on compressed characteristic samples through a one-class support vector machine. By means of the method, the nonlinearity characteristics of the failure samples can be effectively extracted, effective fusion of characteristic space information is achieved, and sensitive characteristics are extracted, and intelligent recognition of bearing early failures is achieved.
Description
Technical field
The present invention relates to mechanical hook-up Fault Identification field, be specifically related to the recognition methods of rolling bearing initial failure.
Technical background
In rotating machinery, the running status of rolling bearing often directly affects the performance of entire machine, therefore, realize to the state-detection of rolling bearing and fault diagnosis significant.At present, the method effectively extracting feature from fault-signal is constantly applied to Rolling Bearing Fault Character and is extracted, but under Rolling Bearing Fault Character is usually submerged in powerful noise background, the feature that most of feature extraction decomposites comprises a large amount of redundant informations, thus reduce the susceptibility of feature, have impact on the Intelligent Recognition of subsequent classification.
Manifold learning, as a kind of advanced method extracting data nonlinear characteristic, compares to traditional linear dimension reduction method, and it more effectively can find the essential structure of nonlinear data when the fault-signal of process collection gained higher-dimension, complex nonlinear.Manifold learning according to its inherent linear or nonlinear relationship, can extract sensitive features by optimization fusion strategy, is beneficial to and carries out Dimensionality Reduction and data analysis, be employed in mechanical fault diagnosis field at present.
The sample of more than two classes or two classes is usually needed relative to general support vector machine, one-class support vector machine only needs two class sample object, in mechanical fault diagnosis field, the sample of malfunction often less and show multi-mode, be difficult in practical application obtain and cost is higher, as long as and the sample that one-class support vector machine records a kind of fault just can set up corresponding sorter, thus identify the state of machine.
Summary of the invention
In order to solve the problem, the present invention, according to the feature of bearing initial failure, proposes the bearing initial failure recognition methods combined with one-class support vector machine based on manifold learning.First corresponding time domain parameter is extracted to original fault-signal and form original feature space; Then utilize the laplacian eigenmaps algorithm (Laplacianeigenmap, LE) in manifold learning to carry out the compression of information fusion realization character to the sample of original feature space, extract responsive feature; One-class support vector machine is finally adopted to carry out Classification and Identification to the feature samples of Feature Compression gained.The present invention effectively can extract the nonlinear characteristic of fault sample, reaches the effective integration of feature-space information thus extracts sensitive features, achieving the Intelligent Recognition of bearing initial failure.
Rolling bearing initial failure of the present invention recognition methods, comprises the following steps:
The first step, first from rolling bearing data-signal, extract 9 time domain parameters form original feature space, described 9 features respectively: mean value, effective value, peak value, waveform factor, Impact Index, crest factor, flexure value, kurtosis value, nargin index;
Second step, adopts LE algorithm to realize Feature Compression to it, carries out in the process of feature extraction, first adopt neighbour's mode to build adjacent map at employing LE algorithm, then uses thermonuclear equation to give weights to every bar limit and constructs neighboring rights value matrix;
3rd step, adopts one-class support vector machine to realize failure modes identification to the feature samples after compression, for the sample of often kind of state, adopts one-class support vector machine to realize Fault Identification: Stochastic choice one half-sample is as training sample; Remain a half-sample as positive class testing sample, simultaneously as a part for the negative class testing sample set of other states, repeat 10 to take turns, for the penalty coefficient related in one-class support vector machine and kernel functional parameter, employing grid type is searched for, cross validation, the ratio limiting training sample shared by support vector when setting up disaggregated model corresponding to sample must not more than 10% and on this basis to these two parameter optimizations.
The present invention effectively can extract the nonlinear characteristic of fault sample, reaches the effective integration of feature-space information thus extracts sensitive features, achieving the Intelligent Recognition of bearing initial failure.
Embodiment
Rolling bearing initial failure of the present invention recognition methods, comprises the following steps:
The first step, first from rolling bearing data-signal, extract 9 time domain parameters form original feature space, described 9 features respectively: mean value, effective value, peak value, waveform factor, Impact Index, crest factor, flexure value, kurtosis value, nargin index;
Second step, adopts LE algorithm to realize Feature Compression to it, carries out in the process of feature extraction, first adopt neighbour's mode to build adjacent map at employing LE algorithm, then uses thermonuclear equation to give weights to every bar limit and constructs neighboring rights value matrix;
3rd step, adopts one-class support vector machine to realize failure modes identification to the feature samples after compression, for the sample of often kind of state, adopts one-class support vector machine to realize Fault Identification: Stochastic choice one half-sample is as training sample; Remain a half-sample as positive class testing sample, simultaneously as a part for the negative class testing sample set of other states, repeat 10 to take turns, for the penalty coefficient related in one-class support vector machine and kernel functional parameter, employing grid type is searched for, cross validation, the ratio limiting training sample shared by support vector when setting up disaggregated model corresponding to sample must not more than 10% and on this basis to these two parameter optimizations.
Select the rolling bearing experimental data experiment Analysis in electro-engineering laboratory.The data selected come from SKF6205 rolling bearing, sample frequency 12kHz, select lesion size to be data under inner ring under four kinds of rotating speeds of 0.007mm, outer ring, ball fault and normal condition.Often kind of rotating speed under often kind of state, totally four kinds of rotating speeds, select 25 samples, sample dimension all gets 4096 dimensions, and namely a kind of sample of state is 100, has got altogether 400 samples.
Respectively 9 time domain parameters are extracted to the sample signal of often kind of state, form original feature space.Adopt LE method to carry out Feature Compression to original feature space, select Neighbourhood parameter to be 6 by experiment with computing, the parameter of thermonuclear equation gets 10, and embedding people's dimension gets 3.LE method is adopted to carry out the feasibility that Feature Compression realizes intelligent diagnosing method in order to verify further, the sample adopting LE method to carry out Feature Compression is compared with the sample adopting PCA method to carry out Feature Compression simultaneously, adopt LE method to carry out Feature Compression respectively and adopt PCA method to carry out Feature Compression, compared to PCA method, LE method is adopted to carry out inner ring sample after Feature Compression, the sample of outer ring sample and other two states is very open, mix individually although normal sample and ball sample have, but will easier Classification and Identification than the situation of PCA method.Visible LE method more can extract the sensitive features for distinguishing sample than PCA method.
Respectively to feature samples, the feature samples of PCA method compression, the feature samples of LE method compression of uncompressed, adopt one-class support vector machine to realize Intelligent Recognition simultaneously.The support vector ratio of one-class support vector machine learning outcome gained model is little, show that model has very strong generalization ability, demonstrate the validity of institute's established model, after adopting LE method to carry out Feature Compression to feature space sample, when Embedded dimensions is 3, adopt LE method to carry out the sample mean discrimination of Feature Compression gained four kinds of states and overall average discrimination compared to not adopting Feature Compression and adopting PCA method to carry out both of these case after Feature Compression, LE ways and means acquired results is optimum.This reflects it is when Embedded dimensions is 3 equally on the whole, adopts LE method to carry out Feature Compression than the sensitive features adopting PCA method and carry out Feature Compression and more can effectively extract in feature space.As can be seen here, can from feature space, effectively extract sensitive features thus realize Fault Identification after adopting LE method to carry out Feature Compression, thus reflect that context of methods is feasible, effective.
Claims (1)
1. the recognition methods of rolling bearing initial failure, is characterized in that, comprises the following steps:
The first step, first from rolling bearing data-signal, extract 9 time domain parameters form original feature space, described 9 features respectively: mean value, effective value, peak value, waveform factor, Impact Index, crest factor, flexure value, kurtosis value, nargin index;
Second step, adopts LE algorithm to realize Feature Compression to it, carries out in the process of feature extraction, first adopt neighbour's mode to build adjacent map at employing LE algorithm, then uses thermonuclear equation to give weights to every bar limit and constructs neighboring rights value matrix;
3rd step, adopts one-class support vector machine to realize failure modes identification to the feature samples after compression, for the sample of often kind of state, adopts one-class support vector machine to realize Fault Identification: Stochastic choice one half-sample is as training sample; Remain a half-sample as positive class testing sample, simultaneously as a part for the negative class testing sample set of other states, repeat 10 to take turns, for the penalty coefficient related in one-class support vector machine and kernel functional parameter, employing grid type is searched for, cross validation, the ratio limiting training sample shared by support vector when setting up disaggregated model corresponding to sample must not more than 10% and on this basis to these two parameter optimizations.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410481861.6A CN104268516A (en) | 2014-09-19 | 2014-09-19 | Rolling bearing early failure recognition method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410481861.6A CN104268516A (en) | 2014-09-19 | 2014-09-19 | Rolling bearing early failure recognition method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104268516A true CN104268516A (en) | 2015-01-07 |
Family
ID=52160036
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410481861.6A Pending CN104268516A (en) | 2014-09-19 | 2014-09-19 | Rolling bearing early failure recognition method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104268516A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104596767A (en) * | 2015-01-13 | 2015-05-06 | 北京工业大学 | Method for diagnosing and predicating rolling bearing based on grey support vector machine |
CN107451515A (en) * | 2016-06-01 | 2017-12-08 | 易程(苏州)电子科技股份有限公司 | A kind of rotating machinery fault recognition method and system |
CN107480731A (en) * | 2017-09-06 | 2017-12-15 | 西安西热电站信息技术有限公司 | A kind of EARLY RECOGNITION method of thermal power plant's automobile assembly welding Iine fault signature |
CN108871761A (en) * | 2018-06-07 | 2018-11-23 | 广东石油化工学院 | A kind of initial failure of gear feature extracting method |
CN110411724A (en) * | 2019-07-30 | 2019-11-05 | 广东工业大学 | A kind of rotary machinery fault diagnosis method, device, system and readable storage medium storing program for executing |
-
2014
- 2014-09-19 CN CN201410481861.6A patent/CN104268516A/en active Pending
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104596767A (en) * | 2015-01-13 | 2015-05-06 | 北京工业大学 | Method for diagnosing and predicating rolling bearing based on grey support vector machine |
CN104596767B (en) * | 2015-01-13 | 2017-04-26 | 北京工业大学 | Method for diagnosing and predicating rolling bearing based on grey support vector machine |
CN107451515A (en) * | 2016-06-01 | 2017-12-08 | 易程(苏州)电子科技股份有限公司 | A kind of rotating machinery fault recognition method and system |
CN107480731A (en) * | 2017-09-06 | 2017-12-15 | 西安西热电站信息技术有限公司 | A kind of EARLY RECOGNITION method of thermal power plant's automobile assembly welding Iine fault signature |
CN108871761A (en) * | 2018-06-07 | 2018-11-23 | 广东石油化工学院 | A kind of initial failure of gear feature extracting method |
CN110411724A (en) * | 2019-07-30 | 2019-11-05 | 广东工业大学 | A kind of rotary machinery fault diagnosis method, device, system and readable storage medium storing program for executing |
CN110411724B (en) * | 2019-07-30 | 2021-07-06 | 广东工业大学 | Rotary machine fault diagnosis method, device and system and readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104268516A (en) | Rolling bearing early failure recognition method | |
Ren et al. | Bearing remaining useful life prediction based on deep autoencoder and deep neural networks | |
CN105628383A (en) | Bearing fault diagnosis method and system based on improved LSSVM transfer learning | |
CN103398769B (en) | Transformer on-line fault detecting method based on sampling integrated SVM (support vector machine) under wavelet GGD (general Gaussian distribution) feather and unbalanced K-mean value | |
CN103234767A (en) | Nonlinear fault detection method based on semi-supervised manifold learning | |
Geng et al. | Bogie fault diagnosis under variable operating conditions based on fast kurtogram and deep residual learning towards imbalanced data | |
CN103941131B (en) | Based on the unbalanced SVM transformer faults detection method of reduced set yojan | |
CN103743980A (en) | Power quality disturbance recognition and classification method based on PSO (Particle Swarm Optimization) for SVM (Support Vector Machine) | |
CN109612729A (en) | A kind of malfunction recognition methods of port Machine Trolley Wheel Bearings | |
Huang et al. | A novel method for detecting image forgery based on convolutional neural network | |
CN103530657B (en) | A kind of based on weighting L2 extraction degree of depth study face identification method | |
CN104749334A (en) | Mode-recognition-based design method for biological abnormal water quality evaluation system | |
CN109187022A (en) | A kind of Fault Diagnosis of Roller Bearings based on accidental resonance and autocoder | |
CN103258217A (en) | Pedestrian detection method based on incremental learning | |
CN108038493A (en) | Based on the industrial big data procedure fault monitoring method for returning intelligent distinguishing insertion | |
CN105574489A (en) | Layered stack based violent group behavior detection method | |
CN103279744A (en) | Multi-scale tri-mode texture feature-based method and system for detecting counterfeit fingerprints | |
Cai et al. | A novel improved local binary pattern and its application to the fault diagnosis of diesel engine | |
CN105626502A (en) | Plunger pump health assessment method based on wavelet packet and Laplacian Eigenmap | |
CN104732236A (en) | Intelligent crowd abnormal behavior detection method based on hierarchical processing | |
CN104102726A (en) | Modified K-means clustering algorithm based on hierarchical clustering | |
CN202693304U (en) | Gear fault diagnosis platform | |
Zhao et al. | Hybrid generative/discriminative scene classification strategy based on latent Dirichlet allocation for high spatial resolution remote sensing imagery | |
CN104318079B (en) | Fault predicting characteristic selecting method based on fault evolution analysis | |
CN104680190A (en) | Target detection method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150107 |
|
WD01 | Invention patent application deemed withdrawn after publication |