CN106017954A - Turnout point machine fault early warning system and method based on audio analysis - Google Patents

Turnout point machine fault early warning system and method based on audio analysis Download PDF

Info

Publication number
CN106017954A
CN106017954A CN201610316282.5A CN201610316282A CN106017954A CN 106017954 A CN106017954 A CN 106017954A CN 201610316282 A CN201610316282 A CN 201610316282A CN 106017954 A CN106017954 A CN 106017954A
Authority
CN
China
Prior art keywords
audio
signal
module
point machine
early warning
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
Application number
CN201610316282.5A
Other languages
Chinese (zh)
Inventor
周杨
沙立
杜俊
李子涵
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.)
Communication signal branch of Beijing Metro Operation Co., Ltd.
METRO OPERATION TECHNOLOGY R & D CENTER, BEIJING SUBWAY OPERATION CO., LTD.
Original Assignee
Nanjing Yaxin Technology Group Co 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 Nanjing Yaxin Technology Group Co Ltd filed Critical Nanjing Yaxin Technology Group Co Ltd
Priority to CN201610316282.5A priority Critical patent/CN106017954A/en
Publication of CN106017954A publication Critical patent/CN106017954A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M99/00Subject matter not provided for in other groups of this subclass
    • G01M99/005Testing of complete machines, e.g. washing-machines or mobile phones

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention discloses a turnout point machine fault early warning system based on audio analysis, comprising an audio collection module, a storage module, a signal processing module, a signal judging module, and an early warning module. Through the early warning system, the fault risk of a turnout point machine can be predicted in a timely and reliable manner. In the working process, the audio collection module collects audio signals of a point machine, the system does not contact the electric control system of the point machine, the reliability and accuracy of the result are ensured, and a basis can be provided for prevention and resolution of equipment failure. Moreover, in a fault early warning method of the invention, multiple collected audio signals are filtered by a band-pass filter, framed, FFT-transformed, normalized and cepstrum-increased, and the feature code is calculated out. The interference of background noise is reduced, and the accuracy of audio signal recognition is improved.

Description

Point machine fault early warning system based on audio analysis and method
Technical field
The present invention relates to a kind of fault early warning system and method, be specifically related to fault early warning system and the side of point machine Method;Belong to technical field of rail traffic.
Background technology
Along with the fast development of transit's routes, goat is more and more wider in the application of the field of track traffic such as railway, subway General.As the actuator of switch control system, the regular event of goat is most important for rail transportation operation, once turns Rut machine breaks down, the most then cause late, proteges of the powerful who stay with their benefactions like parasites maintenance, the normal trip of puzzlement passenger, heavy then cause serious traffic safety Accident and heavy economic losses, even entail dangers to common people life.In conventional art, rush to repair after mostly goat is to use fault Scheme, but be as traffic Fast Construction, Rush Repair Scheme cannot meet the most for ageing high request;Thus, existing Have in technology the fault pre-alarming of goat, realize preventative maintenance and done a lot of research work.
Application publication number is that the Chinese patent of CN 101893667 A discloses a kind of exchange point machine fault detect System and method, specifically includes current sampling device, voltage sampling apparatus, failure analysis module, wherein, current sampling device and The input of voltage sampling apparatus respectively with the connection to be measured of point machine equipment machine room, its outfan and accident analysis Module connects;Failure analysis module is divided into state monitoring module and track switch powered period state monitoring module during track switch action, For detecting the running status of point machine, procedure parameter and fault message, and export above-mentioned fortune by status display interface Row state, procedure parameter and fault message.In this scenario, using the method indirectly measured, on-line real-time measuremen goes out track switch and turns Stressing conditions during rut machine actual motion and other procedure parameters (such as voltage, electric current), it was predicted that and find point machine early The phenomenon that state is bad, especially finds and forecasts the situation that in relay circuit, contact is good, safeguarding in time, from And reduce fault, the purpose of raising reliability, ensure the properly functioning of point machine, the good side of fault pre-alarming of can yet be regarded as Method.But, in this scenario, fault detection system needs to access in the circuit structure of goat, may steady to goat Determine to run and produce impact, cause goat that uncertain fault occurs, thus such detection method does not the most obtain Popularization and application to scale.
Summary of the invention
For solving the deficiencies in the prior art, it is an object of the invention to provide a kind of based on audio analysis contactless Point machine fault early warning system and method, it is possible to accurately and in time goat is carried out fault pre-alarming, it is ensured that changing points Machine properly functioning.
In order to realize above-mentioned target, the present invention adopts the following technical scheme that:
Point machine fault early warning system based on audio analysis, including:
Audio collection module, carries out multisample collection, and the audio frequency that will collect to audio signal during goat action Signal amplifies;
Memory module, accepts the audio signal after processing and amplifying, carries out analog digital conversion real-time storage;
Signal processing module, carries out band-pass filter, signal framing, FFT, returns the audio signal collected One change and cepstrum promote and calculate condition code, are supported vector machine training according to condition code and generate model file storage of classifying To terminal handler;
Signal judge module, reads in Real-time Collection processed audio signal, is predicted according to model file Classification, then classification results is judged;
Warning module, the running status of output point machine and fault pre-alarming.
Specifically, aforementioned audio acquisition module includes: pick up, mounting bracket and signal amplifier, and pick up is installed on It is connected in mounting bracket and with signal amplifier.
Preferably, aforementioned memory module is arranged in DVR.
More preferably, aforementioned signal judge module and warning module are arranged in terminal handler, described terminal handler For computer.
Additionally, the invention also discloses the method utilizing aforesaid fault early warning system to carry out fault pre-alarming, specifically include Following steps:
S1, utilize audio collection module that audio signal during point machine action is carried out multiple sample collection, utilize Signal amplification circuit carries out signal amplification, then stores to memory module after analog digital conversion;
S2, signal processing module carry out band-pass filter, signal framing, FFT to the multiple audio signals collected Conversion, normalization and cepstrum promote and calculate condition code, are supported vector machine training according to condition code and generate model file also Carry out classification storage;
S3, the audio signal of audio collection module Real-time Collection point machine, use the method meter identical with step S2 Calculate the condition code of real-time audio signal, and compare with model file, be predicted classification by signal judge module, come Judge whether point machine is in normal operating conditions;
S4, etc. next section audio data to be collected repeat step S3;If point machine is in malfunction, then same Time outputting alarm to warning module, so, staff can predict the potential faults of goat quickly and accurately, keeps away Exempt to rush to repair afterwards.
Preferably, in abovementioned steps S2, the classification of model file includes following three kinds: when point machine normally works Background sound when audio signal when audio signal, point machine operation irregularity and point machine do not work.
More preferably, in abovementioned steps S2, the multi-category support vector machines using kernel to be rbf carries out mould to sample file Type training, cross validation selects optimization c, g parameter to generate model file, and wherein c represents that penalty coefficient, g represent in kernel function Gamma arrange.
It is highly preferred that in step S3, carry out real-time audio signal by the decoding call back function of SDK bag in DVR Collection.
Further, in abovementioned steps S2, the detailed process of signal framing is: first take individual for N (N is usually 256 or 512) Sampling point assembles an observation unit, referred to as sound frame (Frame), has one section of overlapping area between two adjacent tone frames, and this overlaps Region contains M sample point, and the value of M is the 1/2 or 1/3 of N, is thus avoided that the change of adjacent two sound frames is excessive, Improve sound frame seriality, thus improve the accuracy that condition code is extracted further.
It is further preferred that owing to audio signal change in time domain is difficult to find out its characteristic, thus, in abovementioned steps In S2, Energy distribution audio signal being transformed on frequency domain by FFT is analyzed, owing to FFT requires every frame Data signal comprises a complete signal period, thus use take advantage of Hamming window to add the seriality of forte frame left and right end, so The extraction that can make subsequent characteristics code is more accurate, it is assumed that the signal of sound frame is S (n), n=0,1 ... N-1, N are that sound frame is big Little, it is S ' (n)=S (n) * W (n) after taking advantage of Hamming window, the formula of Hamming window W (n) is as follows: W (n, a)=(1-a)-a*cos (2pi* N/ (N-1)), wherein, a represents the constant parameter of setting, 0 < a < 1.
The invention have benefit that: the fault early warning system of the present invention can in time, reliably predict changing points The potential faults of machine, is acquired the audio signal of goat by pick up in work process, not with the electrical equipment of goat Control system contacts, it is ensured that the reliability of result and accuracy, it is possible to prevention and solution for equipment fault provide foundation;And And, in the fault early warning method of the present invention, by the multiple audio signals collected carry out band-pass filter, FFT becomes Change, signal framing, normalization and cepstrum lifting calculate condition code, reduce the interference of background noise, improve audio signal identification Accuracy.
Accompanying drawing explanation
Fig. 1 is the training flow chart of the fault early warning method of the point machine based on audio analysis of the present invention;
Fig. 2 is the overhaul flow chart of the fault early warning method of the point machine based on audio analysis of the present invention;
Fig. 3 be the present invention fault early warning method in the sampling value scattergram of signal framing.
Detailed description of the invention
Below in conjunction with the drawings and specific embodiments, the present invention made concrete introduction.
The fault early warning system of the present invention, based on the audio analysis to point machine, specifically includes: audio collection module, Memory module, signal processing module, signal judge module and warning module.Wherein, audio collection module includes: pick up, peace Dress support and signal amplifier, pick up is installed in mounting bracket and is connected with signal amplifier, passing through audio collection module Audio signal during goat action carries out multisample (normal, fault and non-duty) gather, and the sound that will collect Frequently signal amplifies;Memory module is arranged in DVR, the audio signal after accepting processing and amplifying, carries out modulus and turns Change and real-time storage;Signal processing module is a core component, it is possible to the audio signal collected is carried out band filter filter Ripple, FFT, signal framing and normalization cepstrum promote and calculate condition code, are supported vector machine training according to condition code Generate model file and classification stores to terminal handler;Signal judge module in terminal handler is by Real-time Collection and through place The audio signal managed is read in, and is predicted classification according to model file, then judges classification results, it is judged that result is passed through Warning module exports.
In order to the present invention is better achieved, below the fault early warning method of this fault early warning system is described in detail, This fault early warning method can be divided into training flow process and the testing process of Fig. 2 of Fig. 1.
Wherein, training flow process sees Fig. 1, particularly as follows: (1), utilize audio collection module to during point machine action Audio signal carries out multiple sample collection, utilizes signal amplification circuit to carry out signal amplification, then store after analog digital conversion to Memory module;(2), signal processing module the multiple audio signals collected are carried out band-pass filter, signal framing, FFT, normalization and cepstrum promote and calculate condition code, are supported vector machine training according to condition code and generate model file And carry out classification storage.Here model file can be divided three classes: audio signal, track switch when point machine normally works turn Background sound when audio signal during rut machine operation irregularity and point machine do not work, so, by all kinds of Signals collecting the process of state store again, call and carry out judging and early warning, especially, by track switch during for subsequent detection flow process After background sound when goat does not works is also carried out gathering, it is possible to avoid erroneous judgement, improve the accuracy of testing result.
It should be noted that when generating model file, the multi-category support vector machines using kernel to be rbf is to sample literary composition Part carries out model training, and cross validation selects optimization c, g parameter to generate model file, and wherein c represents that penalty coefficient, g represent Gamma in kernel function is arranged, and is so obtained in that relatively accurately, disturbs few model file for transferring, improves subsequent detection Accuracy.
In testing process, signal acquisition process and training flow process basic simlarity above, from the point of view of Fig. 2, its concrete mistake Cheng Wei: (1), the pick up of audio detection module is close to the outer wall of point machine, by the solution of the SDK bag of DVR Code call back function carries out the collection of real-time audio signal, carries out signal amplification through signal amplification circuit, then after analog digital conversion Store to memory module;(2), the signal processing module real-time audio signal to collecting carries out band-pass filter, signal Framing, FFT, normalization and cepstrum promote and calculate condition code, and enter with the model file (training flow process obtains) transferred Row comparison, is predicted classification by signal judge module, judges whether point machine is in normal operating conditions;(4)、 Etc. next section audio data to be collected and repeat step (3);If point machine is in malfunction, then outputting alarm while To warning module, so, staff can predict the potential faults of goat quickly and accurately, it is to avoid rush to repair afterwards.
In the present invention, in order to improve accuracy and the reliability of testing result further, signal processing is also entered (other do not make specified otherwise, the common technology being in this area, such as " band-pass filter: enter frequency spectrum to have gone optimization Row smoothing, and the effect of harmonic carcellation, highlight the formant of original voice, also can reduce operand "), specifically include following Several aspects:
A, signal framing: owing to point machine running duration is shorter, the time of only four or five seconds, in order to obtain relatively Many samples, first assemble an observation unit, referred to as sound frame (Frame) by N number of sample point, and the value of usual N is 256 or 512, The time contained is about 20~about 30ms.In order to avoid the change of adjacent two sound frames is excessive, thus we can allow two adjacent because of There is one section of overlapping area between frame, improve sound frame seriality, thus improve the accuracy that condition code is extracted, this overlapping district further Territory contains M sample point, and the value of usual M is approximately the 1/2 or 1/3 of N, as shown in Figure 3.In the present invention, DVR samples Frequency is 8KHz, and sound frame length is 256 sample points, then corresponding time span is 256/8000*1000=32ms.
B, FFT (fast fourier transform): owing to audio signal change in time domain is difficult to find out its characteristic, because of And the energy it being transformed on frequency domain is analyzed, different Energy distribution represents different characteristics.Left in order to add forte frame The seriality of right-hand member, improves the accuracy that condition code is extracted, have employed the calculation taking advantage of Hamming window, it is assumed that sound frame in the present invention The signal changed is S (n), n=0,1 ... N-1, N are sound frame size, are S ' (n)=S (n) * W (n), Hamming window W after taking advantage of Hamming window N the formula of () is as follows: (n, a)=(1-a)-a*cos (2pi*n/ (N-1)), wherein, a represents the constant parameter of setting, 0 < a to W < 1.
C, normalization: data are mapped to (-1,1) scope and process, dimension expression formula will be had to be changed into dimensionless expression formula:Wherein, D ' (i) represents the result after normalized, and D (i) represents and works as Front numerical value, D represents the data set needing normalized, and U represents the normalized data upper limit 1, and L represents under normalized data Limit-1.
D, cepstrum promote (merging residual quantity parameters of cepstrum): energy frequency spectrum energy is multiplied by one group of 20 band filter, Try to achieve the logarithmic energy of each band filter output Wherein, M represents the quantity of band filter, HmK () is the frequency response of band filter,
H m ( k ) = 0 , k < f ( m - 1 ) 2 ( k - f ( m - 1 ) ) ( f ( m + 1 ) - f ( m - 1 ) ) ( f ( m ) - f ( m - 1 ) ) , f ( m - 1 ) &le; k &le; f ( m ) 2 ( f ( m + 1 ) - k ) ( f ( m + 1 ) - f ( m - 1 ) ) ( f ( m ) - f ( m - 1 ) ) , f ( m ) &le; k &le; f ( m + 1 ) 0 , k &GreaterEqual; f ( m + 1 )
In formula
In formula, x (n) is the audio signal of input, this Place N represents counting of Fourier transform;Then, 20 logarithmic energy are obtained MFCC coefficient through discrete cosine transform, obtains L rank Cepstrum parameter, L-value is 12, and discrete cosine transform formula is as follows: Show that the feature of each sound frame has 13 dimensions, comprise 1 logarithmic energy and 12 parameters of cepstrums, then ask parameters of cepstrum relative to The slope of time, formula is as follows:Wherein, CtRepresent t position Cepstrum coefficient;Finally, add that residual quantity computing i.e. produces the characteristic vector of 26 dimensions, obtain condition code.
To sum up, in the fault early warning system and method for the present invention, gather audio signal and to adopting by cordless Collect to audio signal carry out band-pass filter, FFT, signal framing, normalization and cepstrum promote and calculate feature Code, final realization is accurately and reliably predicted the potential faults of goat, is reduced fault, raising reliability, avoids robbing afterwards The purpose repaiied.
The ultimate principle of the present invention, principal character and advantage have more than been shown and described.The technical staff of the industry should Understanding, above-described embodiment limits the present invention the most in any form, and the mode of all employing equivalents or equivalent transformation is obtained Technical scheme, all falls within protection scope of the present invention.

Claims (10)

1. point machine fault early warning system based on audio analysis, it is characterised in that including:
Audio collection module, carries out multisample collection, and the audio signal that will collect to audio signal during goat action Amplify;
Memory module, accepts the audio signal after processing and amplifying, carries out analog digital conversion real-time storage;
Signal processing module, carries out band-pass filter, signal framing, FFT, normalization to the audio signal collected And cepstrum promotes and calculates condition code, it is supported vector machine training generation model file according to condition code and classification stored to end End processor;
Signal judge module, reads in Real-time Collection processed audio signal, is predicted classification according to model file, Again classification results is judged;
Warning module, the running status of output point machine and fault pre-alarming.
Point machine fault early warning system based on audio analysis the most according to claim 1, it is characterised in that described Audio collection module includes: pick up, mounting bracket and signal amplifier.
Point machine fault early warning system based on audio analysis the most according to claim 1, it is characterised in that described Memory module is arranged in DVR.
Point machine fault early warning system based on audio analysis the most according to claim 1, it is characterised in that described Signal judge module and warning module are arranged in terminal handler, and described terminal handler is computer.
5. utilize the method that the fault early warning system described in claim 1 carries out fault pre-alarming, it is characterised in that include walking as follows Rapid:
S1, utilize audio collection module that audio signal during point machine action is carried out multiple sample collection, utilize signal Amplifying circuit carries out signal amplification, then stores to memory module after analog digital conversion;
S2, signal processing module the multiple audio signals collected are carried out band-pass filter, signal framing, FFT, Normalization and cepstrum promote and calculate condition code, are supported vector machine training according to condition code and generate model file and carry out point Class stores;
S3, the audio signal of audio collection module Real-time Collection point machine, use the method identical with step S2 to calculate The condition code of real-time audio signal, and compare with model file, it is predicted classification by signal judge module, judges Whether point machine is in normal operating conditions;
S4, etc. next section audio data to be collected repeat step S3;If point machine is in malfunction, the most defeated Go out alarm to warning module.
The method of fault pre-alarming the most according to claim 5, it is characterised in that in described step S2, dividing of model file Class includes following three kinds: audio signal when audio signal when point machine normally works, point machine operation irregularity And the background sound that point machine is not when working.
The method of fault pre-alarming the most according to claim 5, it is characterised in that in described step S2, employing kernel is rbf Multi-category support vector machines sample file is carried out model training, cross validation selects optimization c, g parameter to generate model literary composition Part, wherein c represents that the gamma that penalty coefficient, g represent in kernel function is arranged.
The method of fault pre-alarming the most according to claim 5, it is characterised in that in step S3, by DVR The decoding call back function of SDK bag carries out the collection of real-time audio signal.
The method of fault pre-alarming the most according to claim 5, it is characterised in that in described step S2, the tool of signal framing Body process is: N number of sample point first assembles an observation unit, referred to as sound frame, has one section of overlapping between two adjacent tone frames Region, this overlapping area contains M sample point, and the value of M is the 1/2 or 1/3 of N.
The method of fault pre-alarming the most according to claim 5, it is characterised in that in described step S2, pass through FFT Energy distribution audio signal being transformed on frequency domain is analyzed, use take advantage of Hamming window to add the seriality of forte frame left and right end, The signal assuming sound frame is S (n), n=0,1 ... N-1, N are sound frame size, take advantage of after Hamming window as S ' (n)=S (n) * W N (), the formula of Hamming window W (n) is as follows: (n, a)=(1-a)-a*cos (2pi*n/ (N-1)), wherein, a represents the normal of setting to W Amount parameter, 0 < a < 1.
CN201610316282.5A 2016-05-13 2016-05-13 Turnout point machine fault early warning system and method based on audio analysis Pending CN106017954A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610316282.5A CN106017954A (en) 2016-05-13 2016-05-13 Turnout point machine fault early warning system and method based on audio analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610316282.5A CN106017954A (en) 2016-05-13 2016-05-13 Turnout point machine fault early warning system and method based on audio analysis

Publications (1)

Publication Number Publication Date
CN106017954A true CN106017954A (en) 2016-10-12

Family

ID=57099770

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610316282.5A Pending CN106017954A (en) 2016-05-13 2016-05-13 Turnout point machine fault early warning system and method based on audio analysis

Country Status (1)

Country Link
CN (1) CN106017954A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109507583A (en) * 2018-11-13 2019-03-22 交控科技股份有限公司 A kind of electronic switching machine monitoring method and system
EP3564091A1 (en) * 2018-04-30 2019-11-06 Siemens Aktiengesellschaft Method and assembly for detecting errors in a points system
CN110866655A (en) * 2019-11-25 2020-03-06 武汉地铁运营有限公司 Intelligent turnout jamming fault early warning method based on power numerical analysis
CN111016964A (en) * 2019-12-13 2020-04-17 西南交通大学 Switch state multi-mode intelligent identification and confirmation platform system
CN112214634A (en) * 2020-09-24 2021-01-12 交控科技股份有限公司 Processing method and system for switch conversion sound
CN113574398A (en) * 2019-02-27 2021-10-29 西门子交通有限公司 Method for monitoring a switch of a railway track system
CN113672859A (en) * 2021-08-17 2021-11-19 郑州铁路职业技术学院 Switch point machine fault acoustic diagnosis system
US11579012B1 (en) 2021-07-13 2023-02-14 Wistron Corporation Abnormal sound detection method and apparatus

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101893667A (en) * 2010-05-26 2010-11-24 广州市地下铁道总公司 Fault detection system of AC turnout switch machine and method thereof
CN102156262A (en) * 2011-03-24 2011-08-17 西南交通大学 On-line monitoring system and method of railway switch machine motor
CN103810374A (en) * 2013-12-09 2014-05-21 中国矿业大学 Machine fault prediction method based on MFCC feature extraction
CN104932485A (en) * 2015-05-29 2015-09-23 中国铁路通信信号上海工程局集团有限公司 Railway track switch point machine fault prediction system and method
CN105260595A (en) * 2015-04-02 2016-01-20 北京交通大学 Feature extraction method for switch action current curve and switch fault diagnosis method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101893667A (en) * 2010-05-26 2010-11-24 广州市地下铁道总公司 Fault detection system of AC turnout switch machine and method thereof
CN102156262A (en) * 2011-03-24 2011-08-17 西南交通大学 On-line monitoring system and method of railway switch machine motor
CN103810374A (en) * 2013-12-09 2014-05-21 中国矿业大学 Machine fault prediction method based on MFCC feature extraction
CN105260595A (en) * 2015-04-02 2016-01-20 北京交通大学 Feature extraction method for switch action current curve and switch fault diagnosis method
CN104932485A (en) * 2015-05-29 2015-09-23 中国铁路通信信号上海工程局集团有限公司 Railway track switch point machine fault prediction system and method

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
吴朝晖 等: "《说话人识别模型与方法》", 31 March 2009 *
徐建立 等: "基于声音信号的结构损伤识别方法", 《金属热处理》 *
机械工程学会设备与维修工程分会: "《设备与维修工程 现代技术及应用》", 31 January 2009 *
杨榆 等: "《音频信息隐藏与数字水印》", 31 October 2015, 北京邮电大学出版社 *
魏维 等: "《媒体计算与内容分析》", 31 January 2012, 清华大学出版社 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3564091A1 (en) * 2018-04-30 2019-11-06 Siemens Aktiengesellschaft Method and assembly for detecting errors in a points system
CN109507583A (en) * 2018-11-13 2019-03-22 交控科技股份有限公司 A kind of electronic switching machine monitoring method and system
CN113574398A (en) * 2019-02-27 2021-10-29 西门子交通有限公司 Method for monitoring a switch of a railway track system
CN110866655A (en) * 2019-11-25 2020-03-06 武汉地铁运营有限公司 Intelligent turnout jamming fault early warning method based on power numerical analysis
CN110866655B (en) * 2019-11-25 2024-04-05 武汉地铁运营有限公司 Intelligent switch blocking fault early warning method based on power numerical analysis
CN111016964A (en) * 2019-12-13 2020-04-17 西南交通大学 Switch state multi-mode intelligent identification and confirmation platform system
CN112214634A (en) * 2020-09-24 2021-01-12 交控科技股份有限公司 Processing method and system for switch conversion sound
CN112214634B (en) * 2020-09-24 2024-04-23 交控科技股份有限公司 Method and system for processing switch conversion sound
US11579012B1 (en) 2021-07-13 2023-02-14 Wistron Corporation Abnormal sound detection method and apparatus
CN113672859A (en) * 2021-08-17 2021-11-19 郑州铁路职业技术学院 Switch point machine fault acoustic diagnosis system
CN113672859B (en) * 2021-08-17 2024-03-29 郑州铁路职业技术学院 Fault acoustic diagnosis system for switch machine

Similar Documents

Publication Publication Date Title
CN106017954A (en) Turnout point machine fault early warning system and method based on audio analysis
Márquez et al. New methods for the condition monitoring of level crossings
CN105882687A (en) Method for analyzing categories of faults of point machines
Cheung et al. The application of statistical pattern recognition methods for damage detection to field data
CN108802535B (en) Screening method, main interference source identification method and device, server and storage medium
Chen et al. Structural nonlinear damage identification algorithm based on time series ARMA/GARCH model
CN104636746B (en) A kind of GIS device Eigenvalue Extraction Method based on regional area wavelet transformation
CN107844067A (en) A kind of gate of hydropower station on-line condition monitoring control method and monitoring system
CN110553789A (en) state detection method and device of piezoresistive pressure sensor and brake system
CN108180967A (en) A kind of vehicle dynamic weighing system and its cheat detection method based on audio auxiliary
CN110450819A (en) A kind of onboard ATC equipment operational system based on non-contact data acquisition
CN113988326A (en) Subway equipment maintenance optimization method and system
CN107105092A (en) A kind of human body tumble recognition methods based on dynamic time warping
KR20220127209A (en) Equipment failure prediction system and method
CN111487491A (en) Health state detection and evaluation system and method for oil immersed smoothing reactor
CN110823552B (en) Electronic parking caliper noise measurement device and measurement method
CN115618205A (en) Portable voiceprint fault detection system and method
Ding et al. Structural damage assessment using improved Dempster-Shafer data fusion algorithm
CN207992717U (en) A kind of gate of hydropower station on-line condition monitoring system
CN109782168A (en) Induction electromotor rotor broken bar fault diagnostic method based on wavelet packet support vector machines
CN109598309A (en) A kind of detection system and its monitoring method of mental package press machine
CN117607598A (en) Transformer fault detection method and system based on voiceprint characteristics
CN110426192A (en) Acoustic fingerprint detection system and method for circuit breaker
US20210116504A1 (en) Method for monitoring circuit breaker and apparatus and internet of things using the same
CN115238733A (en) Method for evaluating operation state of switching-on and switching-off coil of high-voltage circuit breaker and related equipment

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Cao Hongsheng

Inventor after: Chang Liukai

Inventor before: Zhou Yang

Inventor before: Sha Li

Inventor before: Du Jun

Inventor before: Li Zihan

TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20180125

Address after: Nanjing Qinhuai District 210006 Jiangsu Province Yingtian Avenue No. 388 (Road No. 1) the 1865 D1 Creative Park Building No. 1002

Applicant after: Nanjing Yaxin Technology Group Co., Ltd.

Applicant after: Communication signal branch of Beijing Metro Operation Co., Ltd.

Applicant after: METRO OPERATION TECHNOLOGY R & D CENTER, BEIJING SUBWAY OPERATION CO., LTD.

Address before: Nanjing Qinhuai District 210006 Jiangsu Province Yingtian Avenue No. 388 (Road No. 1) the 1865 D1 Creative Park Building No. 1002

Applicant before: Nanjing Yaxin Technology Group Co., Ltd.

RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20161012