CN103994816A - Identification method based on optical fiber multiple events - Google Patents

Abstract

The invention discloses an identification method based on optical fiber multiple events, which comprises the following steps: detecting a vibration signal of the optical fiber; establishing a vibration source identification model of the vibration signal, and acquiring a fundamental frequency variation coefficient of the vibration source identification model; and identifying the vibration source of the vibration signal according to the variation coefficient of the fundamental frequency so as to accurately identify the vibration source of the vibration signal. Through the mode, the vibration source of the vibration signal can be accurately identified.

Description

A kind of recognition methods based on the multiple event of optical fiber

Technical field

The present invention relates to a kind of recognition methods based on the multiple event of optical fiber, belong to fiber-optic vibration and measure the crossing domain of processing subject with random signal.

Background technology

People are increasing for the demand of the energy, and pipeline is carried becomes the major way of carrying the energy.Protection becomes with oil and gas pipes, optical cable near zone or the optical cable itself of optical cable companion row the key issue that current predispersed fiber alarm system need to solve.

Existing predispersed fiber alarm system generally adopts single-stage model to carry out vibration source and detects identification, be direct-detection method of identification, by analyzing the feature of one piece of data, the type that judges vibration source according to feature is (as machinery excavates, vehicle process), and then make early warning judgement according to recognition result.Existing predispersed fiber alarm system is identified by neural network, normalization kurtosis, intrinsic mode function or Chaotic Analysis Method, is mainly used in the situation of high s/n ratio.But the vibration signal that vibration source produces is non-stationary signal complicated and changeable, and pipeline distance and real time environment complexity, causes existing predispersed fiber alarm system cannot identify exactly vibration source.

Summary of the invention

The invention provides a kind of recognition methods based on the multiple event of optical fiber, to solve the problem that cannot identify exactly vibration source.

A kind of recognition methods based on the multiple event of optical fiber provided by the invention, it comprises: the vibration signal of detection fiber; Set up the Recognition of Vibration Sources model of vibration signal, and obtain the fundamental frequency coefficient of variation of Recognition of Vibration Sources model; According to the vibration source of fundamental frequency coefficient of variation identification vibration signal, accurately to identify the vibration source of vibration signal.

Wherein, the Recognition of Vibration Sources model of vibration signal meets following relation:

$V\left(m\right)=\frac{E\left\{\right[X_n-E\left(X_n\right)\left]\right[X_{n-m}-E\left(X_{n-m}\right)\left]\right\}}{\sqrt{D\left(X_n\right)}*\sqrt{D\left(X_{n-m}\right)}};$

Wherein, X _nfor the data sequence of vibration signal, X _n-mfor sequence X _nthe data sequence of the vibration signal that time delay m is ordered, E{[X _n-E (X _n)] [X _n-m-E (X _n-m)] be sequence X _nwith sequence X _n-mcovariance, D (X _n) be sequence X _nvariance, D (X _n-m) be sequence X _n-mvariance.

Wherein, the fundamental frequency coefficient of variation of Recognition of Vibration Sources model is:

$=\frac{\mathrm{\σ}}{\left|\mathrm{\μ}\right|};$

$\mathrm{\σ}=\sqrt{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(x_i-\mathrm{\μ});}$

$\mathrm{\μ}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{x}_i;$

Wherein, N is observed reading quantity, x _iit is the amplitude of the vibration signal of i sampled point; The fundamental frequency coefficient of variation is a dimensionless number, for describing significantly different overall discreteness of average, and the unit of elimination or the impact of average difference on the comparison of two or more data degree of variation, to add up the parameter of vibration source.

Wherein, the method also comprises: extract the vibration data of vibration signal, and obtain the energy of vibration signal from vibration data; Set up the Distribution Entropy model of energy according to information entropy.

Wherein, the Distribution Entropy model H of energy _e(X) be:

$H_E\left(X\right)=E\left(I\left(X\right)\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)I\left(x_i\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)\log p\left(x_i\right);$

Wherein, the energy domain that X is vibration signal, the quantity of information function that I (X) is X, p is the probability density function of energy x.

Wherein, vibration source generation vibration signal at the Distribution Entropy model of one minute vibration data energy is:

$H\left(X\right)=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_2\left(p\left(x_i\right)\right).$

Wherein, set p (x _i) log _bp (x _i)=0 according to Shannon's theorems, occurs that maximum entropy is in the time that the sub-range of energy probability equates:

$H_{\max }=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_{b}p\left(x_i\right)={\log }_b\left(\mathrm{mn}\right).$

Wherein,, can obtain the entropy normalization of per minute according to maximum entropy:

$\stackrel{\mathrm{\Λ}}{H}\left(X\right)=\frac{H\left(X\right)}{H_{\max }}.$

Pass through such scheme, the invention has the beneficial effects as follows: the present invention is by setting up the Recognition of Vibration Sources model of vibration signal, and obtain the fundamental frequency coefficient of variation of Recognition of Vibration Sources model, according to the vibration source of fundamental frequency coefficient of variation identification vibration signal, can accurately identify the vibration source of vibration signal.

Brief description of the drawings

In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing of required use during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.Wherein:

Fig. 1 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of first embodiment of the invention;

Fig. 2 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of second embodiment of the invention;

Fig. 3 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of third embodiment of the invention;

Fig. 4 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of fourth embodiment of the invention;

Fig. 5 is that vibration source is the testing result of vibration source in eight minutes of raining;

Fig. 6 is that vibration source is the testing result of knocking in eight minutes;

Fig. 7 is that vibration source is the testing result of train in eight minutes;

Fig. 8 is the testing result of vibration source Wei Po road signal in eight minutes.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtaining under performing creative labour prerequisite, belong to the scope of protection of the invention.

Shown in Figure 1, Fig. 1 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of first embodiment of the invention.As shown in Figure 1, the recognition methods based on the multiple event of optical fiber that the present embodiment discloses comprises:

S101: the vibration signal of detection fiber;

S102: set up the Recognition of Vibration Sources model of vibration signal, and obtain the fundamental frequency coefficient of variation of Recognition of Vibration Sources model;

S103: according to the vibration source of fundamental frequency coefficient of variation identification vibration signal, accurately to identify the vibration source of vibration signal;

S104: extract the vibration data of vibration signal, and obtain the energy of vibration signal from vibration data;

S105: the Distribution Entropy model of setting up energy according to information entropy.

In S102, the Recognition of Vibration Sources model of vibration signal meets following relation:

$V\left(m\right)=\frac{E\left\{\right[X_n-E\left(X_n\right)\left]\right[X_{n-m}-E\left(X_{n-m}\right)\left]\right\}}{\sqrt{D\left(X_n\right)}*\sqrt{D\left(X_{n-m}\right)}}---\left(1\right)$

Wherein, X _nfor the data sequence of vibration signal, X _n-mfor sequence X _nthe data sequence of the vibration signal that time delay m is ordered, E{[X _n-E (X _n)] [X _n-m-E (X _n-m)] be sequence X _nwith sequence X _n-mcovariance, D (X _n) be sequence X _nvariance, D (X _n-m) be sequence X _n-mvariance.

The fundamental frequency coefficient of variation of Recognition of Vibration Sources model is:

$=\frac{\mathrm{\σ}}{\left|\mathrm{\μ}\right|}---\left(2\right)$

$\mathrm{\σ}=\sqrt{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(x_i-\mathrm{\μ})}---\left(3\right)$

$\mathrm{\μ}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{x}_i---\left(4\right)$

Wherein, N is observed reading quantity, x _iit is the amplitude of the vibration signal of i sampled point; The fundamental frequency coefficient of variation is a dimensionless number, for describing significantly different overall discreteness of average, and the unit of elimination or the impact of average difference on the comparison of two or more data degree of variation, to add up the parameter of vibration source.

In S105, the Distribution Entropy model H of energy _e(X) be:

$H_E\left(X\right)=E\left(I\left(X\right)\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)I\left(x_i\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)\log p\left(x_i\right)---\left(5\right)$

Wherein, the energy domain that X is vibration signal, the quantity of information function that I (X) is X, p is the probability density function of energy x.

Vibration source produces vibration signal:

$H\left(X\right)=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_2\left(p\left(x_i\right)\right)---\left(6\right)$

Set p (x _i) log _bp (x _i)=0 according to Shannon's theorems, occurs that maximum entropy is in the time that the sub-range of energy probability equates:

$H_{\max }=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_{b}p\left(x_i\right)={\log }_b\left(\mathrm{mn}\right)---\left(7\right)$

,, can obtain to facilitate calculating the entropy normalization of per minute according to maximum entropy:

$\stackrel{\mathrm{\Λ}}{H}\left(X\right)=\frac{H\left(X\right)}{H_{\max }}---\left(8\right)$

The present embodiment passes through according to the vibration source of fundamental frequency coefficient of variation identification vibration signal, and sets up the Distribution Entropy model of energy according to information entropy, can accurately identify the vibration source of vibration signal.

Shown in Figure 2, Fig. 2 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of second embodiment of the invention.The recognition methods that the present embodiment discloses comprises:

S201: the vibration signal of detection fiber, obtains the vibration data of vibration signal;

S202: vibration data is sampled and auto-correlation processing in short-term;

S203: coefficient of autocorrelation threshold value is set;

S204: calculate the number N of the extreme point that is greater than coefficient of autocorrelation threshold value, and record the time interval T of adjacent two extreme points;

S205: whether the number N that judges extreme point is greater than default first threshold N0; If so, enter S206; If not, enter S210;

S206: judge that whether the minimum time Tmin between the extreme point of coefficient of autocorrelation threshold value is greater than default Second Threshold T0, if so, enters S207; If not, return to S203, and increase coefficient of autocorrelation threshold value;

S207: computing time interval T fundamental frequency coefficient of variation cv;

S208: judge that whether fundamental frequency coefficient of variation cv is less than the 3rd default threshold value cv0, if so, enters S209; If not, enter S210;

S209: the vibration source of vibration signal is mechanical vibration source;

S210: the vibration source of vibration signal is on-mechanical vibration source, in the time that the duration is greater than 4s, the vibration source of vibration signal is train vibration source.

In S202, vibration data is carried out to sampling processing, extract fundamental frequency information from vibration data, get the vibration data that length is 20 frames, front 10 frame data territory remainder datas carry out relevant treatment, and the data length of at every turn participating in computing is 10 frames, and meet formula (1).Wherein, X _nfor length is the data sequence of the vibration signal of 10 frames; X _n-mfor sequence X _ntime delay m point, length is the data sequence of the vibration signal of 10 frames.

Vibration data is carried out to auto-correlation processing in short-term, vibration data is carried out to 10 times of down-sampled processing, down-sampled formula is:

x‘(n)＝x(m) (9)

Wherein, m=10n, n=1,2,3 ...Vibration data after down-sampled is carried out to auto-correlation processing in short-term, can obtain the coefficient of autocorrelation in short-term of vibration data.

In S203, the initial value of coefficient of autocorrelation threshold value is 0.3.

In S205, first threshold N0 is preferably setting value.

In S206, Second Threshold T0 is preferably Fixed Time Interval.

In S207, according to the fundamental frequency coefficient of variation cv of formula (2), (3), (4) interval T computing time.

In S208, the 3rd threshold value cv0 is preferably and sets coefficient of variation threshold value.

The vibration source that the present embodiment can accurately be identified vibration signal is mechanical vibration source or on-mechanical vibration source.

Shown in Figure 3, Fig. 3 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of third embodiment of the invention.The recognition methods that the present embodiment discloses comprises:

S301: the vibration signal of detection fiber, obtains the vibration data of vibration signal;

S302: energy and the energy distribution of obtaining vibration signal from vibration data;

S303: divide energy and cut apart sub-range;

S304: the Probability p (x that calculates each sub-range energy distribution _i);

S305: calculate the normalization entropy of i minute

S306: the fundamental frequency coefficient of variation cv of T0 minute before calculating;

S307: judge that whether fundamental frequency coefficient of variation cv is less than the 3rd default threshold value cv0, if so, enters S308; If not, enter S310;

S308: before judgement, whether the normalization entropy of T0 minute is greater than the 4th default threshold value H0, if so, enters S309; Enter if not S310;

S309: the vibration source of vibration signal is the vibration source that rains;

S310: the vibration source of vibration signal is non-rainy vibration source.

In S302, obtain the energy of vibration signal from vibration data, calculate the energy of vibration signal, wherein the energy of every frame vibration signal is:

$E_s=10\log \left(\underset{i=1}{\overset{1024}{\mathrm{\Σ}}}{x}_i^2\right)---\left(10\right)$

Wherein, E _sfor the energy of every frame vibration signal, unit is dB; x _iit is the amplitude of the vibration signal of i sampled point.

In S305, calculate the normalization entropy of i minute according to formula (5), (6), (7), (8)

In S306, the fundamental frequency coefficient of variation cv of T0 minute before calculating according to formula (2), (3), (4), wherein the fundamental frequency coefficient of variation is less, illustrates that measured value degree of variation is less, also more stable.

In S308, the 4th threshold value H0 is preferably setting data energy average information entropy.

The vibration source that the present embodiment can accurately be identified vibration signal is vibration source or the non-rainy vibration source of raining.

Shown in Figure 4, Fig. 4 is the process flow diagram of the recognition methods based on the multiple event of optical fiber of fourth embodiment of the invention.The recognition methods that the present embodiment discloses comprises:

S401: the vibration signal of detection fiber;

S402: the Distribution Entropy model of setting up energy;

S403: whether the energy that judges vibration signal possesses three rainy features, if so, enters S404; If not, enter S405;

S404: vibration source is the vibration source that rains;

S405: judge whether to there is high-energy, if so, enter S406; If not, enter S407;

S406: vibration source is knocking;

S407: whether the duration that judges continuous signal is greater than 4s, if so, enters S408; If not, enter S411;

S408: judge whether fundamental frequency coefficient of variation cv is greater than 0.1, if so, enters S409; If not, enter S410;

S409: vibration source is train;

S410: vibration source Wei Po road signal;

S411: the vibration data of vibration signal is carried out to auto-correlation processing in short-term, and vibration source is for knocking road surface.

In S402, set up the Distribution Entropy model of energy according to formula (5), (6), (7), (8).

In S403, three that rain are characterized as: uniformly, when temporal evolution, stably, and exist for a long time.

In S408, obtain fundamental frequency coefficient of variation cv according to formula (2), (3), (4).

In S411, the vibration data of vibration signal is carried out to auto-correlation processing is identical with S202 in short-term.

The below test data for obtaining according to disclosed recognition methods.Wherein, vibration source priority 6～1 represents by numbering respectively on earth successively by height, and wherein 6 represent that warning levels are the highest, i.e. the threaten degree maximum of such vibration source to system, 1 represent little to system threaten degree.

As shown in Figure 5, vibration source is the testing result of vibration source in eight minutes of raining:

Warning: have rain signal within the scope of 9.7366Km centered by the 1st minute 17.5312Km;

Warning: have rain signal within the scope of 9.8864Km centered by the 2nd minute 17.6819Km;

Warning: have rain signal within the scope of 10.0951Km centered by the 3rd minute 17.3596Km;

Warning: have rain signal within the scope of 10.1494Km centered by the 4th minute 17.6434Km;

Warning: have rain signal within the scope of 10.3396Km centered by the 5th minute 18.1845Km;

Warning: have rain signal within the scope of 9.2Km centered by the 6th minute 17.4669Km;

Warning: have rain signal within the scope of 9.7906Km centered by the 7th minute 17.0463Km;

Warning: have rain signal within the scope of 10.8672Km centered by the 8th minute 17.7801Km.

As shown in Figure 6, vibration source is the testing result of knocking in eight minutes:

Warning: there was knocking at 42.9891Km place in the 4th minute, continues 119 times;

Warning: there was knocking at 41.4774Km place in the 5th minute, continues 135 times;

Warning: there was knocking at 42.3657Km place in the 6th minute, continues 123 times;

Warning: there was knocking at 42.0206Km place in the 7th minute, continues 146 times;

Warning: there was knocking at 42.3902Km place in the 8th minute, continues 151 times.

As shown in Figure 7, vibration source is the testing result of train in eight minutes:

Warning: there was train signal at 15.4279Km place in the 1st minute, continues 19.3192 seconds;

Warning: there was the signal of dripping at 8.6761Km place in the 2nd minute, continues 34 times;

Warning: there was the signal of dripping at 9.3593Km place in the 3rd minute, continues 36 times;

Warning: there was train signal at 15.4272Km place in the 5th minute, continues 22.7736 seconds;

Warning: there was the signal of dripping at 9.9338Km place in the 8th minute, continues 50 times.

As shown in Figure 8, the testing result of vibration source Wei Po road signal in eight minutes:

Warning: there was mechanical signal at 3.3484Km place in the 1st minute, continues 5.4921 seconds;

Warning: there was mechanical signal at 2.3263Km place in the 2nd minute, continues 22.1698 seconds;

Warning: there was mechanical signal at 2.3454Km place in the 3rd minute, continues 23.8993 seconds;

Warning: there was mechanical signal at 2.3494Km place in the 4th minute, continues 15.0891 seconds;

Warning: there was mechanical signal at 2.2681Km place in the 5th minute, continues 26.3486 seconds;

Warning: there was mechanical signal at 2.4743Km place in the 6th minute, continues 40.3896 seconds;

Warning: there was mechanical signal at 2.2774Km place in the 7th minute, continues 36.4499 seconds;

Warning: there was mechanical signal at 2.2664Km place in the 8th minute, continues 39.3185 seconds.

In sum, the present invention is by identifying the vibration source of vibration signal according to the fundamental frequency coefficient of variation, and the vibration source that can accurately identify vibration signal is mechanical vibration source or on-mechanical vibration source; And by set up the Distribution Entropy model of energy according to information entropy, the vibration source that can accurately identify vibration signal is vibration source or the non-rainy vibration source of raining, and then can accurately identify the vibration source of vibration signal.

These are only embodiments of the present invention; not thereby limit the scope of the claims of the present invention; every equivalent structure or conversion of equivalent flow process that utilizes instructions of the present invention and accompanying drawing content to do; or be directly or indirectly used in other relevant technical fields, be all in like manner included in scope of patent protection of the present invention.

Claims (8)

1. the recognition methods based on the multiple event of optical fiber, is characterized in that, described method comprises:

The vibration signal of detection fiber;

Set up the Recognition of Vibration Sources model of described vibration signal, and obtain the fundamental frequency coefficient of variation of described Recognition of Vibration Sources model;

Identify the vibration source of described vibration signal according to the described fundamental frequency coefficient of variation, accurately to identify the vibration source of described vibration signal.

2. recognition methods according to claim 1, is characterized in that, the Recognition of Vibration Sources model of described vibration signal meets following relation:

$V\left(m\right)=\frac{E\left\{\right[X_n-E\left(X_n\right)\left]\right[X_{n-m}-E\left(X_{n-m}\right)\left]\right\}}{\sqrt{D\left(X_n\right)}*\sqrt{D\left(X_{n-m}\right)}};$

Wherein, X _nfor the data sequence of described vibration signal, X _n-mfor described sequence X _nthe data sequence of the vibration signal that time delay m is ordered, E{[X _n-E (X _n)] [X _n-m-E (X _n-m)] be described sequence X _nwith described sequence X _n-mcovariance, D (X _n) be described sequence X _nvariance, D (X _n-m) be described sequence X _n-mvariance.

3. recognition methods according to claim 2, is characterized in that, the fundamental frequency coefficient of variation of described Recognition of Vibration Sources model is:

$=\frac{\mathrm{\σ}}{\left|\mathrm{\μ}\right|};$

$\mathrm{\σ}=\sqrt{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(x_i-\mathrm{\μ});}$

$\mathrm{\μ}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{x}_i;$

Wherein, N is observed reading quantity, x _iit is the amplitude of the vibration signal of i sampled point; The described fundamental frequency coefficient of variation is a dimensionless number, for describing significantly different overall discreteness of average, and the unit of elimination or the impact of average difference on the comparison of two or more data degree of variation, to add up the parameter of described vibration source.

4. recognition methods according to claim 1, is characterized in that, described method also comprises:

Extract the vibration data of described vibration signal, and obtain the energy of described vibration signal from described vibration data;

Set up the Distribution Entropy model of described energy according to information entropy.

5. recognition methods according to claim 4, is characterized in that, the Distribution Entropy model H of described energy _e(X) be:

$H_E\left(X\right)=E\left(I\left(X\right)\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)I\left(x_i\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}p\left(x_i\right)\log p\left(x_i\right);$

Wherein, the energy domain that X is described vibration signal, the quantity of information function that I (X) is X, p is the probability density function of energy x.

6. recognition methods according to claim 5, is characterized in that, described vibration source produces described vibration signal and at the Distribution Entropy model of one minute vibration data energy is:

$H\left(X\right)=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_2\left(p\left(x_i\right)\right).$

7. recognition methods according to claim 6, is characterized in that, sets according to Shannon's theorems, in the time that equating, the sub-range of described energy probability occurs that maximum entropy is:

$H_{\max }=-\underset{i=1}{\overset{m\×n}{\mathrm{\Σ}}}p\left(x_i\right){\log }_{b}p\left(x_i\right)={\log }_b\left(\mathrm{mn}\right).$

8. recognition methods according to claim 7, is characterized in that,, can obtain the entropy normalization of per minute according to described maximum entropy:

$\stackrel{\mathrm{\Λ}}{H}\left(X\right)=\frac{H\left(X\right)}{H_{\max }}.$