CN105203645B - A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion - Google Patents

A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion Download PDF

Info

Publication number
CN105203645B
CN105203645B CN201510310524.5A CN201510310524A CN105203645B CN 105203645 B CN105203645 B CN 105203645B CN 201510310524 A CN201510310524 A CN 201510310524A CN 105203645 B CN105203645 B CN 105203645B
Authority
CN
China
Prior art keywords
hurt
switch
vibration signal
track switch
trouble
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.)
Expired - Fee Related
Application number
CN201510310524.5A
Other languages
Chinese (zh)
Other versions
CN105203645A (en
Inventor
王小敏
陈虹屹
郭进
潘炜
陈建译
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.)
Southwest Jiaotong University
China Railway Corp
Original Assignee
Southwest Jiaotong University
China Railway Corp
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 Southwest Jiaotong University, China Railway Corp filed Critical Southwest Jiaotong University
Priority to CN201510310524.5A priority Critical patent/CN105203645B/en
Publication of CN105203645A publication Critical patent/CN105203645A/en
Application granted granted Critical
Publication of CN105203645B publication Critical patent/CN105203645B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention discloses a kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion.Switch blade first in high-speed switch, vibration acceleration sensor collection vibration signal is installed in trouble and at trouble tail three, carrying out CEEMD adaptive decompositions respectively to the vibration signals of three measuring points obtains limited individual IMF components;Then main IMF components are filtered out using correlation analysis, and calculates the singular entropy characteristic value of main IMF components;Next the singular entropy characteristic value of multiple measuring point vibration signals is fused into characteristic vector, and this feature vector input LSSVM is trained and tested, realize the judgement of the working condition and hurt type of track switch.The different crackle hurts of the inventive method energy effective district shunting trouble, have good real-time and noise immunity, new technological means are provided for switch breakdown intelligent diagnostics.

Description

A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion
Technical field
The present invention relates to a kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal, belong to railway switch Detection and maintenance technology field.
Background technology
Important component of the track switch as railroad track, it is the indispensable line facility of high-speed railway, is also simultaneously Weak link on circuit, it is closely related with train running speed and security performance.When train is through turnout passing conversion line, by The strong impact power of track is acted in wheel, track switch may be caused contact fatigue, abrasion, crackle even track deformation etc. occur Hurt type.If not detected, being handled to track switch hurt in time, over time, track switch hurt further deteriorates, The major accidents such as train derailing may be triggered, this safe operation to train constitutes serious threat.Therefore, track switch wound is studied Damage identification, obtains turnout work status information in real time, to ensureing that train is efficient, safe operation is significant.
For track switch hurt monitoring problem, each state all conducts in-depth research and develops corresponding Switch monitor system System, the Track and Turnout monitoring monitoring systems that such as French track switch uses, what German track switch used Roadmaster2000 monitoring systems etc., system above can realize to track switch communication apparatus, electric current, voltage, track circuit and The monitoring of goat state, the conversion power of each towing point etc., for turnout work status information real-time monitoring provide it is strong Means.But these systems are there is also weak point, for example lack the detection to hurts such as track switch abrasion, crackle, contact fatigues. At present, the defectoscope that the track switch hurt detection in China is mainly combined with traditional large-scale inspection car and small-sized defectoscope is made as It is main, although this method of detection can detect track switch hurt to a certain extent, there is that detection efficiency is low, detection range has The problems such as limit, road occupying inspection influence train operation efficiency, it is difficult to meet the needs of China Express Railway fast development.
Publication No. CN102175768A " a kind of high ferro rail defects and failures detection method and device based on vibration signal " Patent application, using the power spectrum of first IMF component of calculating after EMD decomposition vibration signals, and rail is used as using the power spectrum The distinguishing characteristic of hurt carries out hurt detection.In theory, EMD modal overlap problem influences the analysis of signal local feature with carrying Take.For technical standpoint, the patent application only carries out hurt detection from a pick-up point collection signal, can not detect hurt The position (along rail line direction) of generation, do not know be capable of detecting when to come from pick-up point hurt how far yet.Due to track Circuit is oversize, to carry out the detection of hurt position, it is necessary to along the line lay enormous amount sensor and set up sensor network (referring to The 26th section of the patent application specification), the consequence so done, first, data volume is difficult to greatly very much handle;Second, vibration signal is through steel The rigidity of rail has an impact after propagating to multiple sensors, i.e., hurt information is special in the power spectrum of the vibration signal of multiple sensors It can be embodied in sign, cause power spectrum characteristic differentiating method to be difficult to hurt.Fig. 8 measured result also demonstrates the power Spectrum signature can not judge turnout work state and hurt type.
The content of the invention
The purpose of the present invention is the deficiency for existing detection technique, there is provided a kind of high-speed switch based on vibration signal splits Line hurt intelligent detecting method.
The technical solution adopted in the present invention specific implementation step is as follows:
(1) according to the changing rule of the single span simply supported beam Mode Shape of track switch, in the switch blade, trouble in high-speed switch and trouble tail Vibration acceleration sensor is installed at three measuring points;
(2) when train is through turnout passing, the track switch vibration signal at three measuring points is gathered respectively.Without loss of generality, switch blade is remembered Vibration signal at measuring point is x (i), i=1 ..., N, and N is sampling number.
(3) adaptive decomposition is carried out to x (i) using CEEMD, obtains the M comprising track switch hurt information intrinsic mode point Measure (IMF), be designated as cj(i), j=1,2 ..., M.
(4) IMF components c is calculatedj(i) with the coefficient correlation of former vibration signal x (i)
Wherein,WithX (i) and c is represented respectivelyj(i) average value.
(5) η is chosenjMain IMF component of the m IMF component as the measuring point vibration signal more than threshold value T, is designated asK=1,2 ..., m.
(6) main IMF components are calculatedSingular entropy EkAnd by this m singular entropy Ek(k=1,2 ..., m) surveyed as switch blade The track switch hurt characteristic value of point, is designated as H1={ E1,E2,…Em}。
(7) similarly, in trouble, the vibration signal of trouble tail measuring point collection perform step (3)-step (6) respectively, obtain Trouble neutralizes the track switch hurt characteristic value H at trouble tail measuring point2、H3.Then to H1、H2、H3Carry out being augmented Fusion Features, form 3m dimensions Track switch hurt characteristic vector H={ H1、H2、H3}。
(8) hurt characteristic vector H is inputted into Least square support vector (LSSVM) model, chooses RBF and make For kernel function, optimizing, Jin Ershi are carried out to LSSVM penalty factors and RBF parameter using grid search and cross validation The judgement of existing turnout work state and hurt type.
In a kind of above-mentioned high-speed switch crackle hurt intelligent detecting method based on vibration signal, it is characterised in that institute State the threshold value T=0.1 in step (5).
In actually implementing, the main IMF components of the step (6)Singular entropy EkCalculation procedure is as follows:
(6.1) it is rightCarry out phase space reconfiguration.WillIt is embedded into (N-n+1) × n dimension phase spaces, is reconstructed Attractor track matrix X
Wherein n is Embedded dimensions, and N counts for signal sampling.
Singular value decomposition is carried out to X, i.e.,
X=U Λ VT (3)
Wherein,WithFor orthogonal matrix, For diagonal matrix, and q meets q=min (N-n+1, n).λj(j=1,2 ..., q) it is referred to as matrix X singular value.
(6.2) calculateSingular entropy Ek.From λjV maximum singular value before being chosen in (j=1,2 ..., q), and v expires FootThen singular entropy EkFor
Wherein,For weight of j-th of singular value in whole singular value.
In actually implementing, the parameter optimization step of the step (8) is as follows:
(8.1) penalty factor γ ∈ [e are initialized-5,e5], kernel functional parameter σ ∈ [e-5,e5], sizing grid takes 10 × 10, 100 groups of parameters pair are obtained.
(8.2) training sample data are divided into 10 groups, following operate is performed to (γ, σ) to each group of parameter in grid:
(8.3) selecting one of which sample data, remaining 9 groups are used as training set, obtain LSSVM prediction as test set Error delta.
(8.4) repeat step (8.3) performs 10 times, selects different subsets to be tested as training set, and by 10 times every time Obtained prediction error is averaged, and obtains the prediction error of the group (γ, σ)
(8.5) parameter sets (γ, σ are changed2), step (8.3) and (8.4) are repeated, obtain various combination ginseng successively Several lower LSSVM prediction errorOne group of minimum parameter of error mean will be predicted as the optimal model parameters group in grid Close.
Compared with prior art, the beneficial effects of the invention are as follows:
1) present invention carries out hurt detection based on vibration signal to track switch, and the signal acquisition of this method is simple, and signal accumulates Abundant switch status information is contained, the real-time detection of track switch operating mode can be realized and need not largely take track equipment.
2) CEEMD methods effectively inhibit modal overlap problem existing for EMD and EEMD noise residual problem, are adapted to In non-linear, non-stationary the track switch vibration signal of processing, and singular entropy has the work(that singular value decomposition excavates matrix modal characteristics The characteristics of energy and comentropy describe signal sequence complexity, the IMF singular entropies extracted can preferably reflect that the hurt of track switch is special Sign.
3) using LSSVM as grader, without decision threshold is manually set, track switch hurt type can be achieved certainly in the present invention It is dynamic to differentiate.Meanwhile optimizing is carried out to LSSVM parameters using grid search and cross validation, the blindness of parameter selection is reduced, Improve the accuracy rate of hurt detection.
Brief description of the drawings
Fig. 1 is flow chart of the present invention;
Fig. 2 is track switch vibrating sensor scheme of installation;
Fig. 3 is the time domain waveform of the different operating modes in track switch tip, wherein:(a) normal track switch, (b) crackle 0.5cm, (c) crackle 1.5cm;
Fig. 4 is normal track switch vibration signal CEEMD decomposition results;
Fig. 5 is the coefficient correlation of each IMF of different operating mode track switch vibration signals and original signal, wherein:(a) track switch tip, (b) In the middle part of track switch, (c) track switch tail end;
Fig. 6 A are the singular entropy distribution map of the different operating modes in track switch tip;
Fig. 6 B are the singular entropy distribution map of crackle 1.5cm diverse locations;
Fig. 7 is grid search and cross validation parameter optimization result;
Fig. 8 is the power spectral density of the first rank IMF under the different operating modes in track switch tip, wherein:(a) normal track switch, (b) crackle 0.5cm, (c) crackle 1.5cm;
Fig. 9 is influence of the different noises to hurt recognition result.
Embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
As shown in figure 1, the embodiment of the present invention is a kind of high-speed switch crackle hurt intelligence based on vibration signal Energy detection method, its step are:
(1) according to the changing rule of the single span simply supported beam Mode Shape of track switch, in the switch blade, trouble in high-speed switch and trouble tail Vibration acceleration sensor is installed at three measuring points;
(2) when train is through turnout passing, the track switch vibration signal at three measuring points is gathered respectively.Without loss of generality, switch blade is remembered Vibration signal at measuring point is x (i), i=1 ..., N, and N is sampling number.
(3) adaptive decomposition is carried out to x (i) using complete set empirical mode decomposition (CEEMD), obtains hindering comprising track switch M intrinsic modal components (IMF) of damage information, are designated as cj(i), j=1,2 ..., M.
(4) IMF components c is calculatedj(i) with the coefficient correlation of former vibration signal
WhereinWithX (i) and c is represented respectivelyj(i) average value.Choose ηjIMF components more than threshold value T=0.1 are made For main IMF.
(5) the singular entropy E of main IMF components is calculatedk, and by this m singular entropy Ek(k=1,2 ..., m) it is used as switch blade measuring point The track switch hurt characteristic value of collection, is designated as H1={ E1,E2,…Em}.The specific calculation procedure of singular entropy is as follows:
(5.1) it is rightCarry out phase space reconfiguration.WillIt is embedded into (N-n+1) × n dimension phase spaces, obtains reconstruct and inhale Introduction track matrix X
Wherein n is Embedded dimensions, and N counts for signal sampling.
(5.2) singular value decomposition is carried out to X, i.e.,
X=U Λ VT (7)
Wherein,WithFor orthogonal matrix, For diagonal matrix, and q meets q=min (N-n+1, n).λj(j=1,2 ..., q) it is referred to as matrix X singular value.
(5.3) calculateSingular entropy Ek.From λjV maximum singular value before being chosen in (j=1,2 ..., q), and v expires FootThen singular entropy EkFor
Wherein,For weight of j-th of singular value in whole singular value.
(6) similarly, in trouble, the vibration signal of trouble tail measuring point collection perform step (3)-step (5) respectively, obtain Trouble neutralizes the track switch hurt characteristic value H at trouble tail measuring point2、H3.Then to H1、H2、H3Carry out being augmented Fusion Features, form 3m dimensions Track switch hurt characteristic vector H={ H1、H2、H3}。
(7) hurt characteristic vector H is inputted into LSSVM models, chooses RBF as kernel function, utilize grid search Optimizing is carried out to LSSVM penalty factors and RBF parameter with cross validation, and then realizes turnout work state and hurt The judgement of type.Parameter optimization step is as follows:
(7.1) penalty factor γ ∈ [e are initialized-5,e5], kernel functional parameter σ ∈ [e-5,e5], sizing grid takes 10 × 10, 100 groups of parameters pair are obtained.
(7.2) training sample data are divided into 10 groups, following operate is performed to (γ, σ) to each group of parameter in grid:
(7.3) selecting one of which sample data, remaining 9 groups are used as training set, obtain LSSVM prediction as test set Error delta.
(7.4) repeat step (c) performs 10 times, selects different subsets to be tested as training set, and by 10 times every time To prediction error average, obtain the prediction error of the group (γ, σ)
(7.5) parameter sets (γ, σ are changed2), step (7.3) and (7.4) are repeated, obtain various combination ginseng successively Several lower LSSVM prediction errorOne group of minimum parameter of error mean will be predicted as the optimal model parameters group in grid Close.
Effect of the present invention is verified below.
(1) vibration signal at track switch different measuring points is gathered by vibrating sensor.Sensor mounting location such as Fig. 2 institutes Show, wherein measuring point 1 is located at track switch tip, and measuring point 2 is located in the middle part of track switch, and measuring point 3 is located at track switch tail end.Fig. 3 is switch blade difference work The time domain waveform of condition.Signal is decomposed using CEEMD, the decomposition result of normal track switch vibration signal is as shown in Figure 4.Can be with Find out, primary signal is decomposed into several IMF by CEEMD, and each IMF contains the different characteristic information of signal.
(2) coefficient correlation of each IMF components and former vibration signal at track switch different measuring points is calculated, as shown in figure 5, first 5 The correlation of IMF components and original signal is larger, characterizes the main hurt feature of original signal, and other high-order IMF and original signal Below 0.1, visual is abandoned correlation for chaff component.Therefore 5 IMF components calculate as singular entropy before choosing Data source.
(3) singular entropy for the main IMF components that calculation procedure (2) obtains.Part singular entropy distribution is as shown in Fig. 6 A, Fig. 6 B. As can be seen that it is normal different in same observation station its singular entropy distribution with two kinds of operating modes of crackle hurt, different crackle hurt degree Singular entropy distribution is similar, but size has differences.Track switch different measuring points are due to conduct vibrations path and the sensitivity to crackle hurt Degree is different, and in diverse location, its singular entropy is distributed there is also difference same operating, i.e., the hurt information that diverse location is included It is different.Therefore, the singular entropy feature of different measuring points can be carried out being augmented fusion and forms hurt characteristic vector.Above analysis shows are strange Different entropy can preferably characterize the hurt feature of track switch diverse location difference operating mode.
(4) track switch difference operating mode vibration signal road totally 172 groups of sample datas are gathered respectively, wherein, 60 groups of nominal situation, split Each 56 groups of line 0.5cm and crackle 1.5cm.According to step (1)~(3)) obtain respective hurt characteristic vector and input LSSVM instructions Practice and test.Optimizing is carried out to LSSVM penalty factors and RBF parameter using grid search and cross validation, obtains γ =1.63, σ=7.31.Grid search and cross validation parameter optimization result are as shown in Figure 7.
Single measuring point and the test result of 3 points of fusions are as shown in table 1, it can be seen that based on the comprehensive of multi-measuring point vibration signal Close analysis, have make full use of multi-sensor information redundancy and it is complementary the advantages of, the characteristic vector of single measuring point passes through at 3 points After fusion treatment, the crackle hurt discrimination of test sample reaches 91.25%.Show that this method can be effectively used for the inspection of track switch hurt Survey.In addition, the hurt recognition time for calculating the inventive method is 6.49s, real-time is preferable.
1 single measuring point of table and the test result of 3 points of fusions
As a comparison, the track switch vibration signal of above-mentioned collection is carried out using Patent No. CN102175768A methods describeds Experimental analysis, as a result as shown in Figure 8.As can be seen that the IMF Power Spectrum Distributions of track switch difference operating mode are complicated, without obvious area Dtex is levied, and the identification of track switch operating mode and hurt type is difficult to according to the patent document methods described.
Noise testing:
Due to actual track switch vibration signal in gatherer process by Environmental Noise Influence, to assess the anti-noise of the inventive method Performance, the white Gaussian noise, gauss heat source model, pulse for adding different signal to noise ratio respectively on the basis of original vibration signal are made an uproar Sound, and carry out emulation experiment.Wherein gauss heat source model is obtained by the white Gaussian noise that variance is 1 by a fourth-order band-pass wave filter Take.Impulsive noise is by formula n'(k)=B (k) G (k) generations, wherein G (k) is that average is 0, and variance is 1 white Gaussian noise;B (k) it is Bernoulli process.
The test result of the inventive method is as shown in Figure 9 under different signal to noise ratio, it can be seen that the hurt of the inventive method is known Rate does not reduce with the increase for adding noise intensity, and when signal to noise ratio is higher than 20dB, this method is small by three kinds of influence of noises, stable Property is good;When signal to noise ratio is less than 20dB, interference of the noise to signal is larger, and discrimination reduces with signal to noise ratio and declined comparatively fast.But i.e. Make in the case where signal to noise ratio is 5dB, discrimination remains to reach 68% or so, shows that the inventive method has good anti-noise Property.Field experiment shows that the signal to noise ratio of vibration signal of the train through being gathered during turnout passing is generally greater than 20dB.Therefore, it is of the invention The automatic detection of track switch crackle hurt can be realized in Railway Site.

Claims (1)

1. a kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal, it is characterised in that comprise the following steps:
(1) according to the changing rule of the single span simply supported beam Mode Shape of track switch, in the switch blade, trouble in high-speed switch and three, tail of trouble Vibration acceleration sensor is installed at measuring point;
(2) when train is through turnout passing, the track switch vibration signal at three measuring points is gathered respectively;Without loss of generality, switch blade measuring point is remembered The vibration signal at place is x (i), i=1 ..., N, and N is sampling number;
(3) adaptive decomposition is carried out to x (i) using complete set empirical mode decomposition CEEMD, obtains including track switch hurt information M intrinsic modal components IMF, be designated as cj(i), j=1,2 ..., M;
(4) IMF components c is calculatedj(i) with the coefficient correlation of former vibration signal x (i)
Wherein,WithX (i) and c is represented respectivelyj(i) average value;
(5) η is chosenjMain IMF component of the m IMF component as the measuring point vibration signal more than threshold value T, is designated as
(6) main IMF components are calculatedSingular entropy EkAnd by this m singular entropy EkTrack switch hurt feature as switch blade measuring point Value, is designated as H1={ E1,E2,…Em};
(7) in trouble, the vibration signal of trouble tail measuring point collection perform step (3)-step (6) respectively, obtain trouble and neutralize trouble tail surveying Track switch hurt characteristic value H at point2、H3;Then to H1、H2、H3Carry out being augmented Fusion Features, form the track switch hurt feature of 3m dimensions Vectorial H={ H1、H2、H3};
(8) hurt characteristic vector H is inputted into Least square support vector LSSVM models, chooses RBF as core letter Number, optimizing is carried out to LSSVM penalty factors and RBF parameter using grid search and cross validation, and then realize track switch The judgement of working condition and hurt type;
Threshold value T=0.1 in the step (5);
The main IMF components of the step (6)Singular entropy EkCalculation procedure is as follows:
(6.1) it is rightCarry out phase space reconfiguration;WillIt is embedded into (N-n+1) × n dimension phase spaces, obtains reconstruct and attract Sub-track matrix X
Wherein n is Embedded dimensions, and N counts for signal sampling;
Singular value decomposition is carried out to X, i.e.,
X=U Λ VT (3)
Wherein,WithFor orthogonal matrix,For Diagonal matrix, and q meets q=min (N-n+1, n);λjReferred to as matrix X singular value, j=1,2 ..., q;
(6.2) calculateSingular entropy Ek;From λjV maximum singular value before middle selection, and v meetsThen singular entropy EkFor
Wherein,For weight of j-th of singular value in whole singular value.
CN201510310524.5A 2015-06-08 2015-06-08 A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion Expired - Fee Related CN105203645B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510310524.5A CN105203645B (en) 2015-06-08 2015-06-08 A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510310524.5A CN105203645B (en) 2015-06-08 2015-06-08 A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion

Publications (2)

Publication Number Publication Date
CN105203645A CN105203645A (en) 2015-12-30
CN105203645B true CN105203645B (en) 2018-03-13

Family

ID=54951436

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510310524.5A Expired - Fee Related CN105203645B (en) 2015-06-08 2015-06-08 A kind of high-speed switch crackle hurt intelligent detecting method based on vibration signal fusion

Country Status (1)

Country Link
CN (1) CN105203645B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105740840B (en) * 2016-02-29 2017-02-22 中南大学 Nonlinear identification method for rock fracture signal and blasting vibration signal
CN105784936B (en) * 2016-03-09 2017-09-26 西安交通大学 The quick determination method and system of a kind of composite dash-board injury
CN106017879B (en) * 2016-05-18 2018-07-03 河北工业大学 Omnipotent breaker mechanical failure diagnostic method based on acoustic signal Fusion Features
CN106203382A (en) * 2016-07-20 2016-12-07 河海大学 A kind of excitation surge current based on kernel function extreme learning machine and fault current recognition methods
CN106556647B (en) * 2016-11-18 2019-02-22 湘潭大学 A kind of impact echo data processing method
CN108937847B (en) * 2017-05-22 2020-10-09 北京大学 Method for evaluating human body movement coordination
CN109085248B (en) * 2018-07-03 2021-04-06 内蒙古科技大学 Method, device and system for positioning impact source of pressure-bearing pipeline wall
CN109827082B (en) * 2019-03-13 2020-10-09 常州大学 Pipeline multi-point leakage accurate positioning method
CN109975013B (en) * 2019-04-30 2020-09-18 南京工程学院 IVMD-SE-based wind turbine generator gearbox fault feature extraction method
CN110274957B (en) * 2019-07-19 2021-01-29 西安交通大学 Interface bonding quality detection method for large-damping laminated structure

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE19912785A1 (en) * 1999-03-12 2000-10-05 Siemens Ag Railway points rail defect detection device
CN102175768A (en) * 2011-02-22 2011-09-07 哈尔滨工业大学 Method and device for detecting defects and failures of high-speed rail based on vibration signals
CN103175897A (en) * 2013-03-13 2013-06-26 西南交通大学 High-speed turnout damage recognition method based on vibration signal endpoint detection
CN103197001A (en) * 2013-03-13 2013-07-10 西南交通大学 High speed turnout injury identification method based on vibration signal wavelet threshold value denoising

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE19912785A1 (en) * 1999-03-12 2000-10-05 Siemens Ag Railway points rail defect detection device
CN102175768A (en) * 2011-02-22 2011-09-07 哈尔滨工业大学 Method and device for detecting defects and failures of high-speed rail based on vibration signals
CN103175897A (en) * 2013-03-13 2013-06-26 西南交通大学 High-speed turnout damage recognition method based on vibration signal endpoint detection
CN103197001A (en) * 2013-03-13 2013-07-10 西南交通大学 High speed turnout injury identification method based on vibration signal wavelet threshold value denoising

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CEEMD-FFT在滚动轴承故障诊断中的应用;陆森林 等;《郑州大学学报(工学版)》;20150131;第36卷(第1期);75 *
奇异值熵和支持向量机的齿轮故障诊断;张超 等;《振动、测试与诊断》;20111031;第31卷(第5期);600-602 *
高速列车转向架故障的经验模态熵特征分析;秦娜 等;《交通运输工程学报》;20140228;第14卷(第1期);57、63-64 *

Also Published As

Publication number Publication date
CN105203645A (en) 2015-12-30

Similar Documents

Publication Publication Date Title
CN104240455B (en) A kind of disturbance event recognition methods in distribution type fiber-optic pipeline safety early warning system
CN106662072B (en) Wind-driven generator method for monitoring state and system
CN106443310B (en) A kind of transformer fault detection method based on SOM neural network
CN104753461B (en) Method for diagnosing and classifying faults of photovoltaic power generation arrays on basis of particle swarm optimization support vector machines
CN103245861B (en) A kind of transformer fault diagnosis method based on Bayesian network
CN102944416B (en) Multi-sensor signal fusion technology-based fault diagnosis method for wind turbine blades
CN103033359B (en) A kind of main transmission in wind generating set method for diagnosing faults of multiple features Multilateral Comprehensive Judge
CN105241660B (en) High-speed rail large bridge performance test methods based on health monitoring data
García Márquez et al. Railway point mechanisms: condition monitoring and fault detection
CN102317770B (en) Method and device for classification of sound-generating processes
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
Gargoom et al. Automatic classification and characterization of power quality events
Li et al. Fault detection method for railway wheel flat using an adaptive multiscale morphological filter
CN102221655B (en) Random-forest-model-based power transformer fault diagnosis method
CN101776710B (en) High voltage DC power transmission line shielding failure current waveform inversion recovery method
CN104809722A (en) Electrical device fault diagnosis method based on infrared thermography
CN106056821B (en) Electric line foreign matter intrusion intelligent video monitors assessment system on-line
CN103335617B (en) A kind of railway track geometric deformation detection method based on vibration signal
CN107219457B (en) Frame-type circuit breaker fault diagnosis and degree assessment method based on operation attachment electric current
CN101975910A (en) Intelligent fault classification and location method for ultra-high voltage direct current transmission line
CN103323718B (en) Capacitive high-voltage equipment insulation aging diagnostic test system and working method thereof
CN105676085A (en) Extra-high voltage GIS partial discharge detection method based on multi-sensor information fusion
CN106526400B (en) The earth fault diagnostic method and device of DC600V power supply system of train
CN101900708B (en) Vibration and audio signal-based high-speed train track defect detecting method
Wu et al. Feature extraction and identification in distributed optical-fiber vibration sensing system for oil pipeline safety monitoring

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20180313

Termination date: 20180608