CN102607449A - Signal processing method for simultaneously improving BOTDR (Brillion optical time domain reflectometer) spatial resolution ratio and frequency shift measuring precision - Google Patents

Abstract

A signal processing method for simultaneously improving a BOTDR (Brillion optical time domain reflectometer) spatial resolution ratio and frequency shift measuring precision is used for processing time domain signals of a BOTDR by means of Cohen time-frequency analysis. The signal processing method includes the steps: acquiring broadband time domain signals of the BOTDR and leading a sampling rate fs to be four times of the highest frequency fm of the signals, namely fs>=4fm; processing the acquired time domain signals of the BOTDR by means of Cohen time-frequency analysis to obtain corresponding time-frequency distribution; and acquiring frequency spectrum information of time points, namely acquiring time-frequency distribution of the signals, finding positions with strain or temperature change in the frequency spectrum information and performing Cohen time-frequency analysis for the signals at the positions after STFT (short time Fourier transformation) processing in order to decrease calculated amount. By the aid of the signal processing method, the contradiction between the BOTDR spatial resolution ratio and the frequency shift measuring precision caused by the uncertainty principle in signal processing can be overcome.

Description

Improve the signal processing method of BOTDR spatial resolution and frequency displacement measuring accuracy simultaneously

Technical field

The present invention relates to the signal Processing of Brillouin light time-domain reflectomer (BOTDR).In the BOTDR signal Processing, adopt Cohen class Time-Frequency Analysis Method, can alleviate the spatial resolution that causes because of uncertainty principle in the classical signals disposal route and the contradiction of frequency displacement measuring accuracy, make these two performance index obtain effect preferably simultaneously.

Background technology

Brillouin light time-domain reflectomer (BOTDR) is a kind of fully distributed fiber sensor-based system based on spontaneous brillouin scattering in the optical fiber, can be used for monitoring optical fiber strain along the line and temperature information.Compare with the Brillouin optical time domain analysis appearance (BOTDA) based on stimulated Brillouin scattering (SBS), BOTDR has and can single-endedly measure, and can record advantages such as breakpoint.But, accurately record the BOTDR signal and it is handled comparatively difficulty because spontaneous brillouin scattering is extremely faint.In the signal Processing of BOTDR, need from the broadband Brillouin's signal that records, extract and obtain each instantaneous spectrum constantly.The time varying spectrum of the signal that is obtained is used to estimate the Brillouin shift of position on the optical fiber, thereby obtains the distribution of temperature or strain according to the linear relationship of Brillouin shift and dut temperature and strain.The time varying spectrum of signal is called time-frequency distributions, and corresponding signal processing method is called time frequency analysis.The time-frequency distributions of a signal can be estimated with the time-frequency aggregation, about the time-frequency aggregation, multiple different definition is arranged, wherein a kind of being defined as: ${\mathrm{\Δ}}_{\mathrm{Tf}}=\frac{1}{E}\∫\∫(\frac{t^2}{T^2}+T^2{f}^2)P(t,f)\mathrm{Dtdf}.$ In the formula wherein T be a non-zero duration arbitrarily, (t f) is time-frequency distributions to P.

Spatial resolution and frequency displacement measuring accuracy are two important performance index of BOTDR.The various signals disposal route can obtain different time-frequency distributions, and these two performance index are exerted an influence.

At present; In the BOTDR system that adopts relevant process of heterodyning; Two kinds of Time-Frequency Analysis Method are suggested: (1) Short Time Fourier Transform (STFT): directly collect broadband signal; Use the computing of Fast Fourier Transform (FFT) (FFT) realization STFT, obtain the spectrum information of each time point, i.e. the time-frequency distributions of signal.(2) frequency sweep method: behind detector, add wave filter,, be combined into the time-frequency distributions of signal through the time domain track of each frequency component in the frequency sweep acquisition Brillouin signal bandwidth.We provable STFT is consistent with the time-frequency distributions that frequency sweep obtains, and studies so can regard frequency sweep and STFT with a kind of method as.

If the response of detector is enough fast, the spatial resolution δ z of BOTDR can be expressed as δ z=vW/2, and wherein v is the transmission speed of pulse in optical fiber, and W is the time width of pulse.For the STFT method, because window function has covered one section finite length of signal, and the frequency spectrum of this segment signal gained only is used to the information of representing that the window function center position is instantaneous, so the spatial resolution of system can worsen to some extent.Obviously, accurate as far as possible in order to make spatial resolution, hope that the signal length that window function covers trends towards O.But because the restriction of uncertainty principle also is unallowed even obtain the width of window function very little, because this will cause the maximum error that frequency displacement is measured according to uncertainty principle.

The uncertainty principle of STFT can be described as: for the signal after the windowing, and its time width and frequency domain width long-pending more than or equal to a constant, B _tT _t>=1/4 π.If the width of window function is very little, the time width T of signal that window function covers then _tAlso less, can get by uncertainty principle, frequency domain mean square deviation width is bigger, and it is also bigger that corresponding with it Brillouin composes full width at half maximum (FWHM).And the Brillouin shift δ v of FWHM and minimum detectable _BRelation is arranged

Wherein SN R representes signal to noise ratio (S/N ratio), so FWHM is big more, the frequency displacement measuring accuracy is poor more.Therefore, the uncertainty principle in the signal Processing has caused the contradiction between BOTDR spatial resolution and the frequency displacement measuring accuracy.In the BOTDR system that adopts STFT and frequency sweep method, these two performance index are difficult to reach simultaneously effect preferably.

Summary of the invention

The objective of the invention is to, propose a kind of signal processing method that improves BOTDR spatial resolution and frequency displacement measuring accuracy simultaneously, overcome BOTDR spatial resolution that the uncertainty principle in the signal Processing causes and the contradiction between the frequency displacement measuring accuracy.This method can obtain the better time-frequency distributions of time-frequency aggregation, thereby weakens the contradiction between spatial resolution and the frequency displacement measuring accuracy, and then makes these two performance index reach effect preferably simultaneously.

Technical solution of the present invention is: improve the signal processing method of BOTDR spatial resolution and frequency displacement measuring accuracy simultaneously, adopt Cohen class Time-Frequency Analysis Method that the BOTDR time-domain signal is carried out signal Processing, may further comprise the steps:

1) collects the broadband time-domain signal of BOTDR, sample rate f _sAt least be signal highest frequency f _m4 times, i.e. f _s>=4f _m

2) use Cohen class Time-Frequency Analysis Method to handle to the BOTDR time-domain signal that collects, obtain corresponding time-frequency distributions.In order to reduce calculated amount, can only handle interested a part of signal.For example, can at first adopt STFT to handle after, find the position that strain or temperature variation are arranged, again this part signal is carried out Cohen class time frequency analysis.

The general formula of Cohen class time frequency analysis can be expressed as:

$C(t,\mathrm{\ω})=\frac{1}{4{\mathrm{\π}}^2}{\∫\∫\∫s}^{*}(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{\φ}(\mathrm{\θ},\mathrm{\τ})e^{-\mathrm{j\θt}-\mathrm{j\τ\ω}+\mathrm{j\θu}}\mathrm{dud\τd\θ}$

Wherein, t, ω are respectively time and frequency variable, and s (t) is a time-domain signal, u, and τ, θ is aleatory variable, and (θ τ) is kernel function to φ, and different kernel functions can generate different time-frequency distributions.The generation of cross term need be chosen suitable kernel function in the Cohen time-frequency distributions.

Several typical C ohen class time-frequency distributions:

Choi-Williams distributes:

When $\mathrm{\φ}(\mathrm{\θ},\mathrm{\τ})=e^{-{\mathrm{\θ}}^2{\mathrm{\τ}}^2/\mathrm{\σ}}$ The time,

$C_{\mathrm{CW}}(t,\mathrm{\ω})=\frac{1}{4{\mathrm{\π}}^{3/2}}\∫\∫\frac{1}{\sqrt{{\mathrm{\τ}}^2/\mathrm{\σ}}}\exp [-\frac{{(u-t)}^2}{4{\mathrm{\τ}}^2/\mathrm{\σ}}-\mathrm{j\τ\ω}]s*(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{dud\τ}$

Zhao-Altas-Marks distributes:

When $\mathrm{\φ}(\mathrm{\θ},\mathrm{\τ})=g\left(t\right)\left|\mathrm{\τ}\right|\frac{\mathrm{Sin}\mathrm{A\θ\; \τ}}{\mathrm{A\θ\; \τ}}$ The time,

$C_{\mathrm{ZAM}}(t,\mathrm{\ω})=\frac{1}{4\mathrm{\πa}}\∫g\left(\mathrm{\τ}\right)e^{-\mathrm{j\τ\ω}}{\∫}_{t-a\left|\mathrm{\τ}\right|}^{t+a\left|\mathrm{\τ}\right|}s*(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{dud\τ}$

Born-Jorda n distributes:

When $\mathrm{\φ}(\mathrm{\θ},\mathrm{\τ})=\frac{\mathrm{Sin}\mathrm{A\θ\; \τ}}{\mathrm{A\θ\; \τ}}$ The time

$C_{\mathrm{ZAM}}(t,\mathrm{\ω})=\frac{1}{4\mathrm{\πa}}\∫\frac{1}{\left|\mathrm{\τ}\right|}{e}^{-\mathrm{j\τ\ω}}{\∫}_{t-a\left|\mathrm{\τ}\right|}^{t+a\left|\mathrm{\τ}\right|}s*(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{dud\τ}$

Above σ in three formulas, a is an arbitrary parameter, g (τ) is an arbitrary function, t, ω are respectively time and frequency variable, s (t) is a time-domain signal, u, τ, θ is aleatory variable.

The time-frequency aggregation parameter that STFT and Choi-Williams distribute satisfies Δ respectively _Tf-STFT>=1/ π and Δ _Tf-CW>=1/2 π, wherein the time-frequency aggregation does ${\mathrm{\Δ}}_{\mathrm{Tf}}=\∫\∫(\frac{t^2}{T^2}+T^2{f}^2)P(t,f)\mathrm{Dtdf},$ T is a non-zero duration arbitrarily, and (t f) is time-frequency distributions to P, promptly is limited to the twice that Choi-Williams distributes under the time-frequency aggregation of STFT.Therefore the Cohen class can weaken the contradiction between spatial resolution and the frequency displacement measuring accuracy.

3) time-frequency distributions to obtaining adopts follow-up data processing means such as match, obtains the Brillouin shift of position on the optical fiber, thereby realizes optical fiber temperature or Strain Distribution along the line.

Beneficial effect of the present invention is following: the present invention can obtain the better time-frequency distributions of time-frequency aggregation, thereby weakens the contradiction between spatial resolution and the frequency displacement measuring accuracy, and then makes these two performance index reach effect preferably simultaneously.BOTDR spatial resolution and frequency displacement measuring accuracy have been improved simultaneously.Therefore compare with traditional STFT or frequency sweep method, can under same spatial resolution, obtain better frequency displacement measuring accuracy, perhaps under same frequency displacement measuring accuracy, obtain better spatial resolution.

Description of drawings

Fig. 1 is for obtaining the experiment device schematic diagram of BOTDR time-domain signal.

The time-frequency distributions result that Fig. 2 obtains for the Choi-Williams method.

Fig. 3 is the Brillouin shift curve.

Fig. 4 is the instant bandwidth curve of time-frequency distributions.

Specific embodiments

The present invention at first uses Cohen class Time-Frequency Analysis Method to handle to the BOTDR time-domain signal that collects, and obtains corresponding time-frequency distributions.Adopt follow-up data processing means such as match again, obtain the Brillouin shift of position on the optical fiber, thereby realize optical fiber temperature or Strain Distribution along the line.The present invention can obtain the better time-frequency distributions of time-frequency aggregation, thereby weakens the contradiction between spatial resolution and the frequency displacement measuring accuracy, and then makes these two performance index reach effect preferably simultaneously.

Below in conjunction with accompanying drawing and specific embodiment the present invention is further specified, but do not limit protection scope of the present invention with this.

Fig. 1 is for obtaining the experiment device schematic diagram of BOTDR time-domain signal.Laser is divided into two-way through 95: 5 coupling mechanisms.95% detection light one tunnel uses EOM to be modulated to the 20ns pulse, amplifies through EDFA, injects a single-mode fiber behind scrambler and the circulator.Optical fiber is about 360m, and the long part of the about 1Om of tail end is heated to 70 ℃, and room temperature is 30 ℃.5% reference light is leaded up to an optical frequency shift mechanism and is produced the downshift of 10.52GHz.Two-way light is balanced detector after through coupling mechanism and surveys, and carries out signals collecting and data processing again.

Use Cohen class Time-Frequency Analysis Method to handle to the BOTDR time-domain signal that collects, only choose the Choi-Williams of wherein a kind of kernel function here and be distributed as example for

.Its mathematical form is:

$C_{\mathrm{CW}}(t,\mathrm{\ω})=\frac{1}{4{\mathrm{\π}}^{3/2}}\∫\∫\frac{1}{\sqrt{{\mathrm{\τ}}^2/\mathrm{\σ}}}\exp [-\frac{{(u-t)}^2}{4{\mathrm{\τ}}^2/\mathrm{\σ}}-\mathrm{j\τ\ω}]s*(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{dud\τ}$

Wherein get σ=1, t, ω are respectively time and frequency variable, and s (t) is a time-domain signal, u, and τ, θ is aleatory variable.

There is one section of temperature variation to handle to the BOTDR signal, the time-frequency distributions result that Fig. 2 obtains for the Choi-Williams method, average time is 1000 times.Horizontal ordinate representation space position, ordinate is represented the frequency of coherent signal.

Time-frequency distributions to obtaining is carried out the Lorentz match, and Fig. 3 is the Brillouin shift curve, and Fig. 4 is the instant bandwidth curve of time-frequency distributions.Compare for ease, also made three kinds of results of STFT among the figure, be respectively the long 3ns of Gaussian window, 11ns, 23ns.Bandwidth curve 348m place outstanding is that sudden change by Brillouin shift causes among Fig. 4.For STFT, there are contradiction in spatial resolution and frequency displacement measuring accuracy.When window was elongated, rising edge was elongated, and instant bandwidth narrows down, also just explanation, and the spatial resolution variation, the frequency displacement measuring accuracy improves.When the long 4ns of window, spatial resolution is 1.6m, but because the about 185MHz of instant bandwidth, the fluctuation of frequency displacement curve is serious, i.e. and frequency displacement measuring accuracy is relatively poor.When the long 23ns of window, spatial resolution is 3.0m, but because instant bandwidth 68MHz only, the frequency displacement curve is comparatively level and smooth, i.e. and frequency displacement measuring accuracy is higher.For the result that the Choi-Williams method obtains, spatial resolution is coming to the same thing of 1.6m and the shorter window of STFT; The result of about 65MHz of instant bandwidth and the longer window of STFT is approaching.

Therefore, the spatial resolution of using the Choi-Williams method to obtain weakening to cause because of uncertainty principle in the classical signals disposal route and the contradiction of frequency displacement measuring accuracy make these two performance index obtain effect preferably simultaneously.

Claims (2)

1. improve the signal processing method of BOTDR spatial resolution and frequency displacement measuring accuracy simultaneously, it is characterized in that adopting Cohen class Time-Frequency Analysis Method that the BOTDR time-domain signal is carried out signal Processing, may further comprise the steps:

1) collects the broadband time-domain signal of BOTDR, sample rate f _sAt least be signal highest frequency f _m4 times, i.e. f _s>=4f _m

2) use Cohen class Time-Frequency Analysis Method to handle to the BOTDR time-domain signal that collects, obtain corresponding time-frequency distributions; In order to reduce calculated amount; After time-domain signal adopts Short Time Fourier Transform STFT to handle; The spectrum information that obtains each time point is the time-frequency distributions of signal, finds the position that strain or temperature variation are arranged in the spectrum information, again this part signal is carried out Cohen class time frequency analysis;

The general formula of Cohen class time frequency analysis is expressed as:

$C(t,\mathrm{\ω})=\frac{1}{4{\mathrm{\π}}^2}{\∫\∫\∫s}^{*}(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{\φ}(\mathrm{\θ},\mathrm{\τ})e^{-\mathrm{j\θt}-\mathrm{j\τ\ω}+\mathrm{j\θu}}\mathrm{dud\τd\θ}$

Wherein, t, ω are respectively time and frequency variable, and s (t) is a time-domain signal, u, and τ, θ is aleatory variable, and (θ τ) is kernel function to φ, and different kernel functions can generate different time-frequency distributions.

Adopt Choi-Williams to distribute: when kernel function

$C_{\mathrm{CW}}(t,\mathrm{\ω})=\frac{1}{4{\mathrm{\π}}^{3/2}}\∫\∫\frac{1}{\sqrt{{\mathrm{\τ}}^2/\mathrm{\σ}}}\exp [-\frac{{(u-t)}^2}{4{\mathrm{\τ}}^2/\mathrm{\σ}}-\mathrm{j\τ\ω}]s*(u-\frac{1}{2}\mathrm{\τ})s(u+\frac{1}{2}\mathrm{\τ})\mathrm{dud\τ}$

Wherein σ is an arbitrary parameter, and t, ω are respectively time and frequency variable, and s (t) is a time-domain signal, u, and τ, θ is aleatory variable.

The time-frequency aggregation parameter that STFT and Choi-Williams distribute satisfies Δ respectively _Tf-STFT>=1/ π and Δ _Tf-CW>=1/2 π, wherein the time-frequency aggregation does ${\mathrm{\Δ}}_{\mathrm{Tf}}=\∫\∫(\frac{t^2}{T^2}+T^2{f}^2)P(t,f)\mathrm{Dtdf},$ T is a non-zero duration arbitrarily, and (t f) is time-frequency distributions to P, promptly is limited to the twice that Choi-Williams distributes under the time-frequency aggregation of STFT; Therefore Cohen class time frequency analysis can weaken the contradiction between spatial resolution and the frequency displacement measuring accuracy.

2. the signal processing method that improves BOTDR spatial resolution and frequency displacement measuring accuracy simultaneously according to claim 1; It is characterized in that time-frequency distributions to obtaining; Adopt the match follow-up data to handle means; Obtain the Brillouin shift of position on the optical fiber, thereby realize optical fiber temperature or Strain Distribution along the line.

Images