CN106228107A - A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis - Google Patents

A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis Download PDF

Info

Publication number
CN106228107A
CN106228107A CN201610529802.0A CN201610529802A CN106228107A CN 106228107 A CN106228107 A CN 106228107A CN 201610529802 A CN201610529802 A CN 201610529802A CN 106228107 A CN106228107 A CN 106228107A
Authority
CN
China
Prior art keywords
matrix
data matrix
data
isolated component
vector
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
CN201610529802.0A
Other languages
Chinese (zh)
Other versions
CN106228107B (en
Inventor
唐志峰
徐玉胜
姜晓勇
吕福在
张鹏飞
伍建军
骆苏军
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.)
Hangzhou Zheda Jingyi Electromechanical Technology Corp Ltd
Original Assignee
Hangzhou Zheda Jingyi Electromechanical Technology Corp Ltd
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 Hangzhou Zheda Jingyi Electromechanical Technology Corp Ltd filed Critical Hangzhou Zheda Jingyi Electromechanical Technology Corp Ltd
Priority to CN201610529802.0A priority Critical patent/CN106228107B/en
Publication of CN106228107A publication Critical patent/CN106228107A/en
Application granted granted Critical
Publication of CN106228107B publication Critical patent/CN106228107B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
    • B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
    • B61K9/08—Measuring installations for surveying permanent way
    • B61K9/10—Measuring installations for surveying permanent way for detecting cracks in rails or welds thereof
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4454—Signal recognition, e.g. specific values or portions, signal events, signatures
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/02—Indexing codes associated with the analysed material
    • G01N2291/023—Solids
    • G01N2291/0234—Metals, e.g. steel
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Pathology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Immunology (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Acoustics & Sound (AREA)
  • Mechanical Engineering (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention discloses a kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis.The monitoring signal data installing ultrasonic transducer collection acquisition by being arranged on two ends, rail monitored area is processed: creation analysis data matrix, by testing data matrix and formed initial analysis data matrix by benchmark data matrix, carry out centralization and whitening processing successively, decomposing for the analytical data matrix after albefaction, each isolated component generates and separates vector iterative computation acquisition initially-separate vector；It is arranged to make up separation matrix and calculates weight matrix and isolated component matrix, set reference vector, calculate and obtain its correlation coefficient, compare with alarm threshold value, it is thus achieved that rail break result is reported to the police.The present invention is capable of under strong noise background the accurate differentiation for ultrasonic guided wave signals, and highly sensitive, robustness is good.

Description

A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis
Technical field
The present invention relates to a kind of supersonic guide-wave monitoring algorithm, especially related to a kind of based on independent component analysis super Guided Waves rail break monitoring algorithm.
Background technology
Along with the progressively quickening of railway construction in China, the monitoring of railway track structure integrity also becomes the most important, I The current most of circuit of state is assembled with automatic block system, and the core of automatic block system is " track circuit ", i.e. two Aliving in rail, when train enters block section, car body is by the circuit turn-on in two tracks so that in block section Track circuit " short-circuit ", judge the seizure condition of circuit with this, provide information for train below, can also sentence roughly simultaneously The integrity of disconnected track structure.But rely on track circuit monitoring rail break presence problems with: track circuit originally experience railway roadbed ginseng Number situation impact is relatively big, often occurs between rail short-circuit, red in the little and more southern areas with heavy rainfall of railway roadbed leakage impedance The failure conditions such as light belt wrong report；Currently, due to technical reason, China's major part single-track railway is all Semiautomatic block system, i.e. The installation of rail-free circuit arrangement.Along with the application of the BEI-DOU position system of China's independent research, newly-built circuit and train are all pacified The intelligent traveling location equipment of Zhuan Liao China independent research, i.e. the automatic block system of a new generation, track circuit will progressively be washed in a pan Eliminate.
Above present situation shows, is badly in need of a set of special railway track structure integrity monitoring systems (rail break monitoring system), The structural intergrity of real time on-line monitoring track, accurately grasps railway infrastructure state, scientifically carries out quality control and protect Barrier traffic safety.Supersonic guide-wave monitoring technology is one of more potential monitoring technology of one.But making an uproar by force at railway operation Under sound background, guided wave signals identification becomes abnormal difficult.The invention of Patent No. CN201410246903 proposes a kind of based on super The real-time broken rail detection and location system of guided Waves, but it uses the relative magnitude of signal and absolute amplitude to determine whether Rail break is difficult in the case of background noise is relatively strong, the most not mentioned implementation result in patent.
The invention of Patent No. CN201210342445 proposes a kind of high-speed railway rail based on supersonic guide-wave Yu wireless network Health monitoring systems.Literary composition uses wavelet transformation carry out noise reduction, re-use Hilbert transform extraction envelope and differentiate, with Sample, when signal to noise ratio is less than time to a certain degree, and the method also cannot differentiate whether monitored area ruptures.
Summary of the invention
The object of the invention is to solve the deficiency in background technology field, proposes a kind of based on independent component analysis ultrasonic lead Ripple rail break monitoring algorithm, it is achieved the track integrity based on supersonic guide-wave under strong background noise differentiates.
The present invention is by the following technical solutions:
The present invention will install the monitoring signal number of ultrasonic transducer collection acquisition by being arranged on two ends, rail monitored area Process according to using following steps:
Step one: creation analysis data matrix X, will be converted the testing data matrix D obtained by current Received Signal datac With the benchmark data matrix D being converted acquisition by reference signal databEmploying below equation composition initial analysis data matrix X:
X = D b D c * K
Wherein, K is that yardstick stretches matrix；
Step 2: carry out centralization process: all data in initial analysis data matrix X are deducted the average of matrix:
X ^ = X - X ‾
Wherein,Represent the analytical data matrix after centralization,Represent the average of initial analysis data matrix, at the beginning of X represents Beginning analytical data matrix；
Step 3: carry out whitening processing: relevant between observation signal in the analytical data matrix after removing centralization Property, the analytical data matrix after using below equation to obtain albefaction
X ~ = RD - 1 / 2 R T X ^ ,
Wherein, the analytical data matrix after changing centered by REigenvectors matrix, D is characterized value diagonal matrix；
Step 4: for the analytical data matrix after albefactionDecompose, be divided into the isolated component that each is separate, Presetting the number of the isolated component of decomposition, for each isolated component, stochastic generation one separates vector and is iterated meter Calculate, calculate and obtain initially-separate vector Wd；
Step 5: each initially-separate vector W that above-mentioned steps is obtaineddIt is arranged in order composition as matrix column to separate Matrix W, uses below equation to calculate weight matrix A and isolated component matrix S；
A=W-1
S = W × X ~
Wherein,Represent the analytical data matrix after albefaction；
Step 6: using each behavior signal in isolated component matrix S as an isolated component, right in weight matrix A The each weight vectors being classified as isolated component answered, sets reference vector B, and it is correlated with to calculate acquisition for each isolated component Coefficient；
Step 7: set alarm threshold value TH, if there is more than one correlation coefficient r in weight matrix Ai> TH, i=1,2, 3 ..., M, then it is assumed that rail break, reports to the police, is otherwise not considered as rail break, does not reports to the police.
Testing data matrix D in described step onecWith benchmark data matrix DbIt is all will to gather the reception signal number obtained One group receives the signal row as matrix according to this, and is arranged in order formation acquisition with the time.
Specifically, the data matrix D currently recordedcIt is by current Received Signal data receiving signal as matrix Row, and be arranged in order formation with the time.The row of different matrixes represents the reception signal of different group.
Described benchmark data matrix DbReception signal data be that preliminary orbit gathers when guaranteeing complete and obtains.
In described step one, if testing data matrix DcWith benchmark data matrix DbCorresponding reception signal data frequency Time consistent, the element that yardstick stretches in matrix K is 1.
Yardstick stretches matrix K in testing data matrix DcWith benchmark data matrix DbCorresponding reception signal data Dispersion compensation is realized when frequency is inconsistent.
Analytical data matrix after centralization in described step 3Eigenvectors matrix R and eigenvalue diagonal matrix D uses below equation to calculate:
E{XXT}=RDRT
Wherein, the transposition of subscript T representing matrix；E{.} represents computing of averaging.
Described step 4 particularly as follows:
4.1) number setting the isolated component needing estimation represents iteration count as M, d, when iteration starts, initializes repeatedly In generation, counting d was d=1；
4.2) initially-separate vector W is builtd, WdBeing the column vector of N for data amount check, N is DbAnd DcMiddle signal number it With, initially-separate vector WdMiddle primary data is random real number；
4.3) the Newton iteration mode of below equation is used to calculate in order:
③Wd=Wd/‖Wd‖
In formula, E{.} represents computing of averaging,Representing the analytical data matrix after albefaction, g (x) is a non-linear letter Number, g ' (x) represents the derivative of g (x),Represent initially-separate vector WdTransposition, WIBefore expression, the I time iteration completes Initially-separate vector, I=1,2,3 ..., d-1, as d=1, I=0, WI=0, ‖ Wd‖ represents WdMould；
Nonlinear function g (x) uses below equation to calculate:
g ( x ) = 1 1 + e - x
Wherein, e is the truth of a matter of natural logrithm, and x represents this argument of function；
4.4) if WdStep 4.3 is returned) if not restraining；
4.5) make d=d+1, if d is not more than M, return step 4.2).
Described step 6 particularly as follows:
6.1) using each behavior signal in isolated component matrix S as an isolated component, correspondence in weight matrix A Each weight vectors being classified as isolated component, setting reference vector B:
Wherein, data matrix D on the basis of mbIn signal number, n is testing data matrix DcIn signal number；
6.2) below equation is used to calculate in reference vector B and weight matrix A between the weight vectors of each isolated component again Correlation coefficient:
r i = Σ j = 1 N ( a i , j - a i ‾ ) ( B j - B ‾ ) Σ j = 1 N ( a i , j - a i ‾ ) 2 Σ j = 1 N ( B j - B ‾ ) 2
Wherein, aiRepresent the i-th row in weight matrix A, i=1,2,3 ..., M, M are the columns of weight matrix A, are point The number of the isolated component managed out；aI, jI-th row, the numerical value of jth row in expression weight matrix A, j=1,2,3 ..., N, N are power The line number of weight matrix A,Represent the average of the i-th row in weight matrix A；BjRepresent the jth numerical value in reference vector B,Represent The average of reference vector B；riRepresent the i-th row and the correlation coefficient of reference vector B in weight matrix A.
Line number N of described weight matrix A is testing data matrix DcWith benchmark data matrix DbIn signal number it With, i.e. N=m+n.
The invention has the beneficial effects as follows:
Inventive algorithm can extract isolated component, thus differentiate whether transducer receives under strong noise background Guided wave signals, it is achieved whether monitored area is occurred to the differentiation fast and accurately of fracture, highly sensitive, strong robustness.
Accompanying drawing explanation
Fig. 1 is the flow chart of monitoring algorithm.
The measurement signal of signal and two kinds of situations on the basis of Fig. 2.
Fig. 3 for do not occur fracture time matrix in isolated isolated component and weight vectors thereof.
Fig. 4 is isolated isolated component and weight vectors thereof in signal matrix during generation fracture.
Detailed description of the invention
Below in conjunction with the drawings and specific embodiments, the present invention is further described.
Embodiments of the invention and to be embodied as work process as follows:
Embodiment, in supersonic guide-wave monitoring implementation process, installs transducer, it is ensured that monitored area at two ends, monitored area In the case of Wan Zheng, one section of sinusoidal guided wave signals launched continuously by one end transducer, and the transducer of the other end receives.
Gather and once comprise many group reception signals and receive signal data as benchmark, benchmark receive in signal data with Reception signal is as row, and sort downwards formation benchmark data matrix D by each group of reception signalb.In actual monitoring process In, certain interval of time once gathers, and when gathering, many groups of reception signals are as current Received Signal data every time, currently connect Collection of letters number is using reception signal as row, and sort downwards formation testing data by many groups gathered reception signals every time Matrix Dc。
It is inconsistent that the frequency of current Received Signal data and benchmark receive signal data, then the guided wave modal encouraged differs Cause, now need yardstick matrix K of stretching to realize the dispersion compensation of current Received Signal data, when both frequencies are consistent, K square In Zhen, all elements is all 1.
Under strong noise background, guided wave signals occurs significantly to decay after long range propagation, is finally submerged in background In noise, only cannot judge whether monitored area ruptures from time-domain signal.Such as Fig. 2 is the actual measurement of a certain iron leg route Data, two groups of transducers are arranged on the web of the rail position of rail, at a distance of 2 kilometers, receive signal data, Fig. 2 (b) on the basis of Fig. 2 (a) For there is not measurement data during fracture, measurement data when Fig. 2 (c) is that fracture occurs, in the case of three kinds, tranmitting frequency one Cause.Both (b) and (c) measurement data contrasts in Fig. 2 cannot find out significantly difference to judge whether monitored area breaks Split.
Following example process uses the inventive method to process:
Step one: creation analysis data matrix X, will be converted the testing data matrix D obtained by current Received Signal datac With the benchmark data matrix D being converted acquisition by reference signal databComposition initial analysis data matrix X；
Step 2: carry out centralization process: all data in initial analysis data matrix X are deducted the average of matrix；
Step 3: carry out whitening processing: relevant between observation signal in the analytical data matrix after removing centralization Property, obtain the analytical data matrix after albefaction
Step 4: for the analytical data matrix after albefactionDecompose, be divided into the isolated component that each is separate, Presetting the number of the isolated component of decomposition, for each isolated component, stochastic generation one separates vector and is iterated meter Calculate, calculate and obtain initially-separate vector Wd；
Step 5: each initially-separate vector W that above-mentioned steps is obtaineddIt is arranged in order composition as matrix column to separate Matrix W, calculates weight matrix A and isolated component matrix S；
Step 6: Fig. 3 and Fig. 4 is that monitored area does not occurs fracture and occurs to isolate independently dividing of signal during fracture respectively , it can be seen that when monitored area occurs fracture, there is the weight vectors of an isolated component in amount and weight vectors thereof Middle existence obvious step feature, this isolated component mostlys come from the signal that transducer is launched, when breaking in monitored area When splitting, this isolated component accounts for the main component of signal, and after rupturing, the weight shared by this isolated component is almost nil, therefore, Step feature occurs in weight vectors.All signal number sums during element number is equal to analysis matrix in weight vectors, by The step signal vector B of this one standard of structure is as reference vector, then calculates its phase relation of acquisition for each isolated component Number；
Step 7: when there is step feature in weight vectors, then calculate and there is certain phase relation in the correlation coefficient of gained Number is in close proximity to 1.So an alarm threshold value TH can be set, if weight matrix A exists more than one correlation coefficient ri> TH, i=1,2,3 ..., M, then send alarm, otherwise measure after setting interval time next time.
If measure time monitored area rupture, after this algorithm process, the weight of isolated isolated component to Amount does not has obvious step feature.The isolated component decomposed after three kinds of these algorithm process of measurement data of Fig. 2 is as it is shown on figure 3, a left side Four, side signal is isolated four isolated components, and four, right side signal is corresponding weight vectors, it can be seen that weight to Amount does not has obvious step feature.
If monitored area has occurred and that fracture when measuring, then after this algorithm process, the power of the isolated component that sub-argument goes out Weight vector exists one or above weight vectors has obvious step feature, as shown in Figure 4, first isolated component Weight vectors exists for obvious step feature, from there through the phase of the reference vector calculating weight vectors and have step feature Close coefficient, alarm threshold value can be set and differentiate.
The present invention can differentiate under strong noise background whether transducer receives guided wave signals, thus differentiates monitored area Structural intergrity, highly sensitive, robustness is good.Above-mentioned implementation is used only to explain the present invention, and the present invention's is concrete real Applying method includes but not limited to method mentioned above, in scope of the presently claimed invention to any amendment of the present invention all Belong to protection scope of the present invention.

Claims (8)

1. a supersonic guide-wave rail break monitoring algorithm based on independent component analysis, it is characterised in that will supervise by being arranged on rail The monitoring signal data surveying the installation ultrasonic transducer collection acquisition of two ends, region uses following steps to process:
Step one: creation analysis data matrix X, will be converted the testing data matrix D obtained by current Received Signal datacWith by Reference signal data converts the benchmark data matrix D obtainedbEmploying below equation composition initial analysis data matrix X:
Wherein, K is that yardstick stretches matrix；
Step 2: carry out centralization process: all data in initial analysis data matrix X are deducted the average of matrix:
Wherein,Represent the analytical data matrix after centralization,Representing the average of initial analysis data matrix, X represents initial point Analysis data matrix；
Step 3: carry out whitening processing: dependency between observation signal in the analytical data matrix after removing centralization, adopts The analytical data matrix after albefaction is obtained by below equation
Wherein, the analytical data matrix after changing centered by REigenvectors matrix, D is characterized value diagonal matrix；
Step 4: for the analytical data matrix after albefactionDecompose, be divided into the isolated component that each is separate, in advance Setting the number of the isolated component decomposed, for each isolated component, stochastic generation one separates vector and is iterated calculating, meter Calculate and obtain initially-separate vector Wd；
Step 5: each initially-separate vector W that above-mentioned steps is obtaineddIt is arranged in order composition separation matrix as matrix column W, uses below equation to calculate weight matrix A and isolated component matrix S；
A=W-1
Wherein,Represent the analytical data matrix after albefaction；
Step 6: using each behavior signal in isolated component matrix S as an isolated component, correspondence in weight matrix A Each weight vectors being classified as isolated component, sets reference vector B, and calculates its phase relation of acquisition for each isolated component Number；
Step 7: set alarm threshold value TH, if there is more than one correlation coefficient r in weight matrix Ai> TH, i=1,2,3 ..., M, then it is assumed that rail break, is otherwise not considered as rail break.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 1, its feature exists In: testing data matrix D in described step onecWith benchmark data matrix DbBe all by gather obtain reception signal data with One group receives the signal row as matrix, and is arranged in order formation acquisition with the time.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 2, its feature exists In: described benchmark data matrix DbReception signal data be that preliminary orbit gathers when guaranteeing complete and obtains.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 1, its feature exists In: in described step one, if testing data matrix DcWith benchmark data matrix DbCorresponding reception signal data frequency is consistent Time, the element that yardstick stretches in matrix K is 1.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 1, its feature exists In: the analytical data matrix after the centralization in described step 3Eigenvectors matrix R and eigenvalue diagonal matrix D use Below equation calculates:
E{XXT}=RDRT
Wherein, the transposition of subscript T representing matrix；E{.} represents computing of averaging.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 1, its feature exists In: described step 4 particularly as follows:
4.1) number setting the isolated component needing estimation represents iteration count as M, d, when iteration starts, initializes iteration meter Number d is d=1；
4.2) initially-separate vector W is builtd, WdBeing the column vector of N for data amount check, N is DbAnd DcMiddle signal number sum, initially Separate vector WdMiddle primary data is random real number；
4.3) the Newton iteration mode of below equation is used to calculate in order:
③Wd=Wd/‖Wd‖
In formula, E{.} represents computing of averaging,Representing the analytical data matrix after albefaction, g (x) is a nonlinear function, g ' X () represents the derivative of g (x),Represent initially-separate vector WdTransposition, WIInitial point that before expression, the I time iteration completes From vector, I=1,2,3 ..., d-1, as d=1, I=0, WI=0, ‖ Wd‖ represents WdMould；
Nonlinear function g (x) uses below equation to calculate:
Wherein, e is the truth of a matter of natural logrithm, and x represents the independent variable of nonlinear function g (x)；
4.4) if WdStep 4.3 is returned) if not restraining；
4.5) make d=d+1, if d is not more than M, return step 4.2).
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 1, its feature exists In: described step 6 particularly as follows:
6.1) using each behavior signal in isolated component matrix S as an isolated component, in weight matrix A, correspondence is each It is classified as the weight vectors of isolated component, setting reference vector B:
Wherein, data matrix D on the basis of mbIn signal number, n is testing data matrix DcIn signal number；
6.2) below equation is used to calculate reference vector B and phase between the weight vectors of each isolated component in weight matrix A again Pass coefficient:
Wherein, aiRepresent that i-th in weight matrix A arranges, i=1,2,3 ..., M, M are the columns of weight matrix A；ai,jRepresent weight I-th row, the numerical value of jth row in matrix A, j=1,2,3 ..., N, N are the line number of weight matrix A,Represent in weight matrix A the The average of i row；BjRepresent the jth numerical value in reference vector B,Represent the average of reference vector B；riRepresent in weight matrix A I-th row and the correlation coefficient of reference vector B.
A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis the most according to claim 7, its feature exists In: line number N of described weight matrix A is testing data matrix DcWith benchmark data matrix DbIn signal number sum.
CN201610529802.0A 2016-06-30 2016-06-30 A kind of supersonic guide-wave broken rail monitoring method based on independent component analysis Active CN106228107B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610529802.0A CN106228107B (en) 2016-06-30 2016-06-30 A kind of supersonic guide-wave broken rail monitoring method based on independent component analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610529802.0A CN106228107B (en) 2016-06-30 2016-06-30 A kind of supersonic guide-wave broken rail monitoring method based on independent component analysis

Publications (2)

Publication Number Publication Date
CN106228107A true CN106228107A (en) 2016-12-14
CN106228107B CN106228107B (en) 2019-05-21

Family

ID=57520048

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610529802.0A Active CN106228107B (en) 2016-06-30 2016-06-30 A kind of supersonic guide-wave broken rail monitoring method based on independent component analysis

Country Status (1)

Country Link
CN (1) CN106228107B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107966206A (en) * 2017-11-07 2018-04-27 华北电力大学（保定） A kind of array signal denoising method of fiber ultrasonic sensor array
CN108225329A (en) * 2017-12-28 2018-06-29 广州泽祺信息科技有限公司 A kind of accurate indoor orientation method
CN108709932A (en) * 2018-04-04 2018-10-26 西安理工大学 A kind of track condition detection method based on supersonic guide-wave broken rail detecting system
CN109649432A (en) * 2019-01-23 2019-04-19 浙江大学 Cloud platform rail integrity monitoring systems and method based on guided wave technology
CN111694328A (en) * 2019-03-12 2020-09-22 宁波大学 Distributed process monitoring method based on multiple independent component analysis algorithms
CN113415309A (en) * 2021-06-16 2021-09-21 西安交通大学 Vehicle-mounted steel rail disease monitoring device based on visual analysis and working method thereof

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006208075A (en) * 2005-01-26 2006-08-10 Hitachi Ltd Abnormality diagnosis apparatus, method and program
CN101927773A (en) * 2010-08-17 2010-12-29 华南理工大学 Blind signal separation technology-based railway track safety detection system and method
CN104020221A (en) * 2014-05-30 2014-09-03 杨媛 Real-time broken-rail detecting and positioning system based on ultrasonic guided waves
US20150068296A1 (en) * 2012-04-06 2015-03-12 The Regents Of The University Of California Air-Coupled Ultrasonic Inspection Of Rails

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006208075A (en) * 2005-01-26 2006-08-10 Hitachi Ltd Abnormality diagnosis apparatus, method and program
CN101927773A (en) * 2010-08-17 2010-12-29 华南理工大学 Blind signal separation technology-based railway track safety detection system and method
US20150068296A1 (en) * 2012-04-06 2015-03-12 The Regents Of The University Of California Air-Coupled Ultrasonic Inspection Of Rails
CN104020221A (en) * 2014-05-30 2014-09-03 杨媛 Real-time broken-rail detecting and positioning system based on ultrasonic guided waves

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
DANIEL HESSE ET.AL: "A Single Probe Spatial Averaging Technique for Guided Waves and Its Application to Surface Wave Rail Inspection", 《IEEE TRANSACTIONS ON ULTRASONICS, FERROELECTRICS, AND FREQUENCY CONTROL》 *
李浩 等: "基于超声导波的无缝线路断轨检测***", 《电子设计工程》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107966206A (en) * 2017-11-07 2018-04-27 华北电力大学（保定） A kind of array signal denoising method of fiber ultrasonic sensor array
CN108225329A (en) * 2017-12-28 2018-06-29 广州泽祺信息科技有限公司 A kind of accurate indoor orientation method
CN108225329B (en) * 2017-12-28 2021-10-29 杨艳华 Accurate indoor positioning method
CN108709932A (en) * 2018-04-04 2018-10-26 西安理工大学 A kind of track condition detection method based on supersonic guide-wave broken rail detecting system
CN108709932B (en) * 2018-04-04 2021-04-06 西安理工大学 Track state detection method based on ultrasonic guided wave broken track detection system
CN109649432A (en) * 2019-01-23 2019-04-19 浙江大学 Cloud platform rail integrity monitoring systems and method based on guided wave technology
CN109649432B (en) * 2019-01-23 2020-06-23 浙江大学 System and method for monitoring integrity of steel rail of cloud platform based on guided wave technology
CN111694328A (en) * 2019-03-12 2020-09-22 宁波大学 Distributed process monitoring method based on multiple independent component analysis algorithms
CN111694328B (en) * 2019-03-12 2022-03-18 宁波大学 Distributed process monitoring method based on multiple independent component analysis algorithms
CN113415309A (en) * 2021-06-16 2021-09-21 西安交通大学 Vehicle-mounted steel rail disease monitoring device based on visual analysis and working method thereof

Also Published As

Publication number Publication date
CN106228107B (en) 2019-05-21

Similar Documents

Publication Publication Date Title
CN106228107A (en) A kind of supersonic guide-wave rail break monitoring algorithm based on independent component analysis
CN101900708B (en) Vibration and audio signal-based high-speed train track defect detecting method
Kalhori et al. Non-intrusive schemes for speed and axle identification in bridge-weigh-in-motion systems
CN101614775A (en) Transformer State Assessment system and appraisal procedure thereof based on Multi-source Information Fusion
CN105760934A (en) Bridge abnormity monitoring restoration method based on wavelet and BP neural network
CN105021953A (en) A transformer substation grounding grid corrosion detection system and method based on earth's surface magnetic induction intensity
CN106093701A (en) A kind of cable fault signal detecting method based on empirical mode decomposition filtering
CN105203645A (en) Intelligent detection method of high-speed turnout crack damage based on vibration signal fusion
CN103994817A (en) Vibration source identification method based on long-distance optical fiber frequent occurring events
CN105445624A (en) Cable fault positioning method according to combination of wavelet transformation and curve fitting
CN105138843A (en) Electric system sampling flying spot detection and repair method thereof
Le et al. Distinction between harmonic and structural components in ambient excitation tests using the time–frequency domain decomposition technique
CN105467428A (en) Seismic wave warning method based on short-time energy detection and spectrum feature analysis
CN103438794A (en) Engineering structure impacted position determination method based on piezoelectric sensor network
CN102830421B (en) Method for identifying redundancies and assembly of satellite-borne electronic device
CN103913512B (en) The damage reason location system of suspension cable periodic detection
CN104897995A (en) Transformer station grounding network corrosion detecting system based on earth surface potential, and method thereof
CN106338662A (en) Transformer winding deformation diagnosis method based on mathematical morphology
CN112963740A (en) Method for monitoring and positioning leakage of fire fighting pipeline of convertor station
CN105021954A (en) Bus protection method based on line-mode current S-transformation argument detection
CN103197001A (en) High speed turnout injury identification method based on vibration signal wavelet threshold value denoising
CN204679586U (en) A kind of grounding net of transformer substation corrosion detection system based on earth's surface magnetic induction density
CN105226616A (en) A kind of bus bar protecting method based on row wave height frequency component coefficient correlation
De Andrade et al. Travelling wave based fault location analysis for transmission lines
CN105046859A (en) Optical fiber intrusion detection method based on vibration signal space-time two-dimensional sparse representation K-S detection

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