CN107576380A - A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques - Google Patents
A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques Download PDFInfo
- Publication number
- CN107576380A CN107576380A CN201710854451.5A CN201710854451A CN107576380A CN 107576380 A CN107576380 A CN 107576380A CN 201710854451 A CN201710854451 A CN 201710854451A CN 107576380 A CN107576380 A CN 107576380A
- Authority
- CN
- China
- Prior art keywords
- signal
- algorithm
- duration
- primary signal
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention relates to a kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques, the field of signal processing belonged in distributed optical fiber sensing technology.It is divided into following steps:Step 1: determine the average signal-to-noise ratio and signal effective range of primary signal;Step 2: the pretreatment of primary signal;Step 3: the feature extraction of primary signal;Step 4: the Modulation recognition of random forest.
Description
Technical field
The present invention relates to a kind of three-dimensional vibrating Modulation recognition method towards Φ-OTDR technique, belong to fully distributed fiber
Field of signal processing in sensing technology.
Background technology
Along with the fast development of optical fiber communication technology, optical fiber sensing technology obtained in recent years further investigation and it is extensive
Health monitoring problem applied to multiple fields.The one kind of distributed optical fiber sensing technology as optical fiber sensing technology, because
Its good distributed nature is more and more widely paid attention to.Distributed optical fiber sensing technology can be divided into from principle
Three major types, it is Brillouin scattering, Rayleigh scattering, Raman scattering respectively.Brillouin scattering is mainly to monitor deformation and temperature, is drawn
Graceful scattering is mainly monitoring temperature, and Rayleigh scattering is mainly monitoring vibration.Method based on Rayleigh scattering detection vibration has
A lot, wherein coherent light time domain reflection technology (Ф-OTDR) is a kind of than more typical and relative maturity of detection vibration signal
Method, therefore present invention is generally directed to the signal of Φ-OTDR technique collection to be studied.
The three-dimensional vibrating signal for the different oscillatory types that Φ-OTDR technique collects is often widely different, therefore to difference
The tagsort of signal becomes the underlying issue to entirely studying.In actual applications, optical fiber sensing vibration be pass through by
Detection object conduction, and for convenience of analysis time domain and frequency domain representation.Feature extraction for fiber-optic signal is divided into
Feature extraction mode based on time domain, frequency domain and time-frequency domain, wherein in time-frequency domain, have wavelet coefficient, wavelet-packet energy,
The methods of Hilbert spectrums and marginal spectrum, extracts signal characteristic;In a frequency domain, there is Fast Fourier Transform (FFT);Having time in the time domain
Series model method (AR models, arma modeling etc.).Sum up two judging quotas, one is in " when m- amplitude " plane
The estimation of signal to noise ratio, another is the useful signal length range in " length-amplitude " plane.Therefore existing feature is combined to believe
Number research theory, still, the vibration data that Φ-OTDR are monitored include length, three information of time and amplitude, above-mentioned side
Method can not accurately extract the three-dimensional vibrating signal that Φ-OTDR are gathered.The advantages of present invention is directed to the above method and actual feelings
Condition, devise it is a kind of can be according to the method for signal characteristic Accurate classification three-dimensional vibrating signal.
The content of the invention
In order to overcome the shortcomings of above-mentioned technology, the invention provides a kind of three-dimensional vibrating signal towards Φ-OTDR technique
Denoising method.
In terms of the denoising of coherent noise, because coherent noise is regular in time domain, therefore the method for the present invention
It is desirable to be finally reached with the function of periodic transformation and carry out coherent noise one effective counteracting by establishing one.
In terms of the denoising of random noise, adding window of the invention is smooth-Wiener-Hopf equation filtering algorithm, it is ensured that signal to noise ratio and signal are effective
Got a promotion while scope.
The present invention provides a kind of three-dimensional vibrating signal antinoise method towards Φ-OTDR technique, main including following several
Step:
Step 1: determine the average signal-to-noise ratio and signal effective range of primary signal;
Step 2: the pretreatment of primary signal;
Step 3: the feature extraction of primary signal;
Step 4: the Modulation recognition of random forest.
The advantage of the invention is that:
(1) the present invention is in order to adapt to the feature extracting method of Φ-OTDR technique.
(2) the present invention is in order to adapt to the sorting technique of Φ-OTDR technique.
Brief description of the drawings
Fig. 1 is algorithm flow chart in the present invention;
Fig. 2 is TESP single order features in the present invention;
Fig. 3 is TESP second order features in the present invention;
Fig. 4 is TESP union features in the present invention;
Embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
A kind of three-dimensional vibrating Modulation recognition method towards Φ-OTDR technique provided by the invention, system block diagram such as Fig. 1 institutes
Show, it is as follows to specifically include step:
Step 1: determine the average signal-to-noise ratio and signal effective range of primary signal.
Signal to noise ratio, English name are called SNR or S/N (SIGNAL-NOISE RATIO), also known as signal to noise ratio.
Signal effective range, refer under the premise of ensureing that signal to noise ratio is effective in " when m- amplitude " plane, in " length-width
Maximum effective distance in degree " plane.
Step 2: the pretreatment of primary signal;
Distributed fiberoptic sensor has the characteristics of high sensitivity, can detect small signal intensity, but be also vulnerable to
The influence of noise, especially when propagation distance is farther out and when signal intensity is weaker, it is easy to flooded by noise, so carrying out
Denoising is carried out to the signal collected before other operations.Currently used fiber-optic signal denoising method is Wavelet Denoising Method,
But find it for the excavation signal denoising DeGrain that collects after testing, it is impossible to preferably recovering signal, therefore take
A kind of smooth filtering method carries out denoising to excavating signal.
Step 3: the feature extraction of primary signal;
Sequence pattern algorithm (abbreviation TESP) with temporal characteristics.The main thought of the algorithm is by the consecutive numbers in time domain
According to limited discrete model is converted to, a feature unit is considered to each model, constituent parts are assigned with different numerical value
Or symbol.Continuous signal can be thus converted to the symbol stream being made up of simple elements.
Step 4: the Modulation recognition of random forest.
At present, SVMs (SVM) method being used for fiber-optic signal identification, this method generalization ability is higher, but
Kernel functional parameter is very big on result influence, and kernel function is complex, therefore proposes to enter fiber-optic signal using random forests algorithm
Row identification.
Embodiment one
The present invention is described in detail with reference to the accompanying drawings and examples.
A kind of three-dimensional vibrating signal antinoise method towards Φ-OTDR technique provided by the invention, specifically includes step such as
Under:
Step 1: determine the average signal-to-noise ratio and signal effective range of primary signal.
Signal to noise ratio, English name are called SNR or S/N (SIGNAL-NOISE RATIO), also known as signal to noise ratio.Refer to one
The ratio of signal and noise in electronic equipment or electronic system.The measurement unit of signal to noise ratio is dB, its computational methods such as formula
(1) shown in.
The effective power of wherein Ps and Pn difference representation signals and noise.Because what instrument and equipment finally monitored is all electricity
Signal, therefore signal to noise ratio can also be converted into the ratio of voltage magnitude, as shown in formula (2).
Signal effective range, refer under the premise of ensureing that signal to noise ratio is effective in " when m- amplitude " plane, in " length-width
Maximum effective distance in degree " plane.
Step 2: the pretreatment of primary signal;
Distributed fiberoptic sensor has the characteristics of high sensitivity, can detect small signal intensity, but be also vulnerable to
The influence of noise, especially when propagation distance is farther out and when signal intensity is weaker, it is easy to flooded by noise, so carrying out
Denoising is carried out to the signal collected before other operations.Currently used fiber-optic signal denoising method is Wavelet Denoising Method,
But find it for the excavation signal denoising DeGrain that collects after testing, it is impossible to preferably recovering signal, therefore take
A kind of smooth filtering method carries out denoising to excavating signal.
The purpose of smothing filtering is to remove the noise being mingled with data, recovering signal inherent feature, the construction master of wave filter
Consider at 2 points:The selection of filter window size and filtering algorithm.Data and place of the big I of filter window according to actual treatment
The demand setting of reason, filtering algorithm include mean filter, medium filtering, Kalman filtering etc., several filtering are contrasted in this research
The effect of algorithm, it is proposed that a kind of filtering method, can preferably remove influence of noise.
Filtering is divided into three parts progress:First, the DC component of signal is removed;Then, using 10 data points as window
Mouth carries out sliding stack;Finally, it is superimposed again with 10 for window sliding after being taken absolute value to stack result.Such a smooth filtering method
With other filtering method processing data results as shown in the figure:
Upper signal is the signal that gathers when noise jamming is serious, bottom signal be noise jamming it is weaker in the case of collect
Signal, left-side signal is filtered signal, and right-side signal is original signal.The front and rear signal waveform of contrast filtering, can see
Go out, such a smothing filtering mode of proposition, reduced in the case of the signal restored under noise serious conditions and noise jamming are weaker
The signal gone out, although there is certain difference in amplitude, the peak-to-valley value and overall trend of signal are substantially consistent, and remaining is filtered
The signal that the signal that wave method restores under noise serious conditions reduces with noise jamming in the case of weaker has larger difference.Therefore
Such a filtering method proposed has preferable denoising effect.
Step 3: the feature extraction of primary signal;
Sequence pattern algorithm (abbreviation TESP) with temporal characteristics.The main thought of the algorithm is by the consecutive numbers in time domain
According to limited discrete model is converted to, a feature unit is considered to each model, constituent parts are assigned with different numerical value
Or symbol.Continuous signal can be thus converted to the symbol stream being made up of simple elements.The specific implementation of TESP algorithms
Step is as follows:
1) in time domain, continuous signal is cut into continuous some time, cutting principle be each two zero crossing it
Between be a period, being accustomed to each period according to the algorithm of TESP algorithms is referred to as member.
2) two indices be present in each member.One is the duration, is typically represented with D;Another is signal aspect, one
As represented with S.Simultaneously following information is obtained according to the two indexs:
A. multiple sampled points within the duration of each member be present, these sampled points can be converted to frequency by conversion
Information.
B. within the duration, its derivative is solved after carrying out waveform fitting, corresponding extreme point and flex point can be obtained, with
And harmonic information.
3) using the duration and in the form of as two dimensions, establish a two-dimensional matrix, and matrix is encoded.Two
Corresponding symbol is generally Serial No. in dimension matrix.
4) among whole time domain procedures, each symbol can frequently occur, their probability of occurrence can be it is different,
Therefore count their probability of occurrence and be used as feature use.
Step 4: the Modulation recognition of random forest.
At present, SVMs (SVM) method being used for fiber-optic signal identification, this method generalization ability is higher, but
Kernel functional parameter is very big on result influence, and kernel function is complex, therefore proposes to enter fiber-optic signal using random forests algorithm
Row identification.
Random forest is a kind of important sorting algorithm in machine learning, is the grader for including multiple decision trees, output
Classification be by it is each tree output classification mode depending on.Each decision tree determines division category according to the situation of sample characteristics
Property and measurement, sample is split into each subset, and the classification of sample is as far as possible consistent in each subset.It is each in random forest to determine
The training of plan tree is carried out parallel, therefore has fireballing advantage, has good classifying quality on the data set of majority.
The decision tree quantity ultimately generated in random forest and the feature chosen can influence nicety of grading, therefore will be to random forests algorithm
Parameter optimizes, and selects suitable decision tree quantity and feature to obtain optimal grader.
Optimize random forests algorithm using gridding method, two parameters to be optimized are drawn in certain spatial dimension
It is divided into grid, optimized parameter is found by traveling through intersection point all in grid.By the method will travel through the institute of two parameters
There are combination, the low time length of efficiency, therefore a threshold value is set according to result of the test, the stopping time when the accuracy rate of identification is higher than this value
Go through.The feature finally chosen includes frequency-region signal amplitude maximum respective frequencies, the minimum value of signal, signal energy, signal most
The ratio between big value, frequency domain amplitude maximum, peak-to-average force ratio, zero-crossing rate, signal median, signal main peak value and minor peaks, due to decision-making
After quantity is set more than 70, the accuracy rate on training set is all very high, weighs operation efficiency and test result, the number of trade-off decision tree
Measure as 100.
Claims (2)
1. the present invention provides a kind of three-dimensional vibrating Modulation recognition method towards Φ-OTDR technique, main to include following several steps
Suddenly:
Step 1: determine the average signal-to-noise ratio and signal effective range of primary signal;
Step 2: the pretreatment of primary signal;
Step 3: the feature extraction of primary signal;
Step 4: the Modulation recognition of random forest.
2. optimization construction method according to claim 1, it is characterised in that:The feature of primary signal carries in step 3
Take;
Sequence pattern algorithm (abbreviation TESP) with temporal characteristics.The main thought of the algorithm is to turn the continuous data in time domain
Be changed to limited discrete model, a feature unit be considered to each model, constituent parts are assigned different numerical value or
Symbol.Continuous signal can be thus converted to the symbol stream being made up of simple elements.The specific implementation step of TESP algorithms
It is as follows:
1) in time domain, continuous signal is cut into continuous some time, cutting principle is to be between each two zero crossing
One period, it is accustomed to referred to as member of each period according to the algorithm of TESP algorithms.
2) two indices be present in each member.One is the duration, is typically represented with D;Another is signal aspect, general to use
S is represented.Simultaneously following information is obtained according to the two indexs:
A. multiple sampled points within the duration of each member be present, these sampled points can be converted to frequency by conversion to be believed
Breath.
B. within the duration, its derivative is solved after carrying out waveform fitting, corresponding extreme point and flex point, Yi Jixie can be obtained
Ripple information.
3) using the duration and in the form of as two dimensions, establish a two-dimensional matrix, and matrix is encoded.In Two-Dimensional Moment
Corresponding symbol is generally Serial No. in battle array.
4) among whole time domain procedures, each symbol can frequently occur, their probability of occurrence can be it is different, therefore
Count their probability of occurrence and be used as feature use.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710854451.5A CN107576380A (en) | 2017-09-20 | 2017-09-20 | A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710854451.5A CN107576380A (en) | 2017-09-20 | 2017-09-20 | A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107576380A true CN107576380A (en) | 2018-01-12 |
Family
ID=61036354
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710854451.5A Pending CN107576380A (en) | 2017-09-20 | 2017-09-20 | A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107576380A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109470352A (en) * | 2018-10-19 | 2019-03-15 | 威海北洋光电信息技术股份公司 | Distributed optical fiber pipeline safety monitoring algorithm based on adaptive threshold |
CN111160106A (en) * | 2019-12-03 | 2020-05-15 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | Method and system for extracting and classifying optical fiber vibration signal features based on GPU |
CN111256890A (en) * | 2020-02-12 | 2020-06-09 | 金陵科技学院 | Long-period fiber grating axial stress estimation method based on random forest |
JP2022052280A (en) * | 2020-09-23 | 2022-04-04 | アンリツ株式会社 | Optical time domain reflectometer and method for testing optical fiber using optical pulse |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104729667A (en) * | 2015-03-25 | 2015-06-24 | 北京航天控制仪器研究所 | Method for recognizing disturbance type in a distributed optical fiber vibration sensing system |
CN104751580A (en) * | 2015-04-08 | 2015-07-01 | 武汉理工光科股份有限公司 | Distributed optical fiber sensing signal mode identifying method and system |
CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
-
2017
- 2017-09-20 CN CN201710854451.5A patent/CN107576380A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104729667A (en) * | 2015-03-25 | 2015-06-24 | 北京航天控制仪器研究所 | Method for recognizing disturbance type in a distributed optical fiber vibration sensing system |
CN104751580A (en) * | 2015-04-08 | 2015-07-01 | 武汉理工光科股份有限公司 | Distributed optical fiber sensing signal mode identifying method and system |
CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
Non-Patent Citations (3)
Title |
---|
YANZHU HU: "The Research on Information Representation of Φ-OTDR Distributed Vibration Signals", 《JOURNAL OF SENSORS》 * |
白永强: "面向全分布式光纤传感技术的车辆振动信号特征提取方法研究", 《中国科技论文在线》 * |
魏然 等: "随机森林在分布式光纤挖掘信号识别中的应用", 《中国科技论文在线》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109470352A (en) * | 2018-10-19 | 2019-03-15 | 威海北洋光电信息技术股份公司 | Distributed optical fiber pipeline safety monitoring algorithm based on adaptive threshold |
CN111160106A (en) * | 2019-12-03 | 2020-05-15 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | Method and system for extracting and classifying optical fiber vibration signal features based on GPU |
CN111160106B (en) * | 2019-12-03 | 2023-12-12 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | GPU-based optical fiber vibration signal feature extraction and classification method and system |
CN111256890A (en) * | 2020-02-12 | 2020-06-09 | 金陵科技学院 | Long-period fiber grating axial stress estimation method based on random forest |
CN111256890B (en) * | 2020-02-12 | 2021-04-23 | 金陵科技学院 | Long-period fiber grating axial stress estimation method based on random forest |
JP2022052280A (en) * | 2020-09-23 | 2022-04-04 | アンリツ株式会社 | Optical time domain reflectometer and method for testing optical fiber using optical pulse |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Fault diagnosis of diesel engine based on adaptive wavelet packets and EEMD-fractal dimension | |
CN107576380A (en) | A kind of three-dimensional vibrating Modulation recognition method towards Φ OTDR techniques | |
CN109974782B (en) | Equipment fault early warning method and system based on big data sensitive characteristic optimization selection | |
Zheng et al. | Gear fault diagnosis method based on local mean decomposition and generalized morphological fractal dimensions | |
CN104504403B (en) | A kind of rotating machinery fault Forecasting Methodology based on scattering conversion | |
Sheng et al. | Applications in bearing fault diagnosis of an improved Kurtogram algorithm based on flexible frequency slice wavelet transform filter bank | |
CN104729667A (en) | Method for recognizing disturbance type in a distributed optical fiber vibration sensing system | |
Zhong et al. | A study on the stationarity and Gaussianity of the background noise in land-seismic prospecting | |
CN105606198A (en) | Fiber vibration sensing system two-order signal feature extraction and determining method | |
CN104485113A (en) | Multi-fault-source acoustic emission signal separation method | |
CN104297213A (en) | Blood cell analyzing instrument and recognizing method and system for abnormal cells thereof | |
Yan et al. | Bearing fault diagnosis via a parameter-optimized feature mode decomposition | |
Yao et al. | An adaptive seismic signal denoising method based on variational mode decomposition | |
CN105823492A (en) | Method of extracting weak target signal in ocean current interference | |
CN112985574A (en) | High-precision classification identification method for optical fiber distributed acoustic sensing signals based on model fusion | |
Zhang et al. | An adaptive graph morlet wavelet transform for railway wayside acoustic detection | |
CN114077854B (en) | phi-OTDR underwater acoustic signal processing method and device based on self-adaptive VMD | |
Han et al. | A novel rolling bearing fault diagnosis method based on generalized nonlinear spectral sparsity | |
Zhong et al. | Recognition and prediction of ground vibration signal based on machine learning algorithm | |
CN115310499B (en) | Industrial equipment fault diagnosis system and method based on data fusion | |
CN112033656A (en) | Mechanical system fault detection method based on broadband spectrum processing | |
CN111339986A (en) | Frequency law mining method and system for equipment based on time domain/frequency domain analysis | |
Han et al. | Fault diagnosis of rolling bearings using dual-tree complex wavelet packet transform and time-shifted multiscale range entropy | |
CN110160778A (en) | Gearbox fault state identification method based on sequential hypothesis testing | |
WO2023004180A1 (en) | Identification of false transformer humming using machine learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180112 |
|
WD01 | Invention patent application deemed withdrawn after publication |