CN106017522A - Rapid high-precision signal demodulation method of fiber F-P sensor - Google Patents

Abstract

The invention provides a rapid high-precision signal demodulation method of a fiber F-P sensor. With a wavelet threshold denoising method, denoising processing is carried out on obtained spectral data; on the basis of a Fourier transform demodulation algorithm, a cavity length of a fiber F-P sensor is calculated and is used as rough measurement cavity length value, and a hill-climbing searching starting point of the cavity length is determined; according to a variable-step-length hill-climbing searching algorithm, discrete cavity length coefficients of the cavity length and a previous cavity length under each step length are compared until a previous coefficient is larger than a next coefficient, a peak value under the step length is searched and is used as a next hill-climbing searching starting point, the step length set for each searching is less than the step length of the last time until preset target precision is reached, and the set step length searched at the last time is a demodulation resolution ratio and a cavity length corresponding to a discrete cavity length coefficient maximum value obtained at the last step is the need result. According to the invention, with the novel variable-step-length hill-climbing searching algorithm, the calculation load of the discrete cavity length coefficient demodulation algorithm is reduced, so that rapid high-precision signal demodulation is realized.

Description

A kind of quick high accuracy signal demodulating method of fibre-optical F-P sensor

Technical field

The invention belongs to technical field of optical fiber sensing, be specifically related to the quick high of a kind of fiber F-P (Fabry-Perot) sensor Precision signal demodulation method.

Background technology

The cavity length demodulating algorithm of fibre-optical F-P sensor has the biggest impact to the precision and stability of measuring results, mainly It is divided into intensity demodulation and phase demodulating two class.Intensity demodulation method typically uses monochromatic laser light source, utilizes linear zone internal interference light intensity Change be demodulated, this demodulation method response speed is the fastest, it is adaptable to Dynamic Signal is (such as acoustical signal, vibration signal Deng) measurement, but its measurement dynamic range is limited, and is easily fluctuated by the intensity of light source and is affected；Phase demodulating method typically uses Wideband light source or swept light source, solve the change of phase place by gained spectrum and obtain the change of F-P cavity chamber length, this demodulation Method, demodulation accuracy is high, and response speed there is not intensity demodulation method fast, it is adaptable to the measurement of quasi-static signal (pressure, temperature etc.).

Currently used widest phase demodulating method is exactly unimodal method and multimodal method, and the most unimodal method demodulation accuracy is high still can only Carry out relative measurement and measurement dynamic range is the least, only in λ/4.Multimodal method can realize absolute chamber length and measure, but due to Error during spectrum peak-seeking, certainty of measurement is relatively low.Fourier transformation (FFT) demodulating algorithm demodulation speed is fast, but demodulation essence It is poor to spend, and resolution is low, it is impossible to realize high-precision measurement.The discrete cavities that the S.M.Musa of the Virginia Institute of Technology proposes is long (DGT) demodulation method improves demodulation accuracy by reducing the long step-size in search in chamber, and under high precision, its amount of calculation is inevitable the biggest. Other high accuracy demodulating algorithm such as cross-correlation demodulating algorithm, Curve Fitting Method Solution adjusts algorithm, and lowest mean square root demodulating algorithm etc. is all There is the problem that amount of calculation is huge, be unfavorable for the engineer applied of reality.

Summary of the invention

The technical problem to be solved in the present invention is: provide the quick high accuracy signal demodulating method of a kind of fibre-optical F-P sensor.

The present invention solves that the technical scheme that above-mentioned technical problem is taked is: the quick high accuracy letter of a kind of fibre-optical F-P sensor Number demodulation method, the light that wideband light source sends passes through bonder, returns, then pass through after the fibre-optical F-P sensor that chamber length is to be measured Bonder exports, and eventually passes opto-electronic conversion and conditioning processes and obtains spectroscopic data, it is characterised in that: it comprises the following steps:

S1, by wavelet threshold denoising method, obtained spectroscopic data is carried out denoising；

S2, according to the spectrum after denoising, by Fourier transformation demodulating algorithm, the chamber calculating fibre-optical F-P sensor is long, As the long bigness scale in chamber value, and determine the search by hill climbing starting point of chamber length；

S3, the result that S2 is obtained, utilize variable step ramping constraint more each step-length cavity of resorption length and previous chamber length from Dissipate the long coefficient in chamber, until a coefficient searches the peak value under this step-length, as climbing the mountain next time greatly after previous coefficient ratio Search starting point, the step-length that search sets every time is respectively less than last step-length, until it reaches default aimed at precision, for the last time Step-length set by search is the demodulation resolution of algorithm, and the chamber that the discrete cavities long coefficient maximum that obtains of final step is corresponding Long, it is required fibre-optical F-P sensor chamber long.

As stated above, described S1 carries out interpolation increase sampling particularly as follows: first spectroscopic data is transformed to wave-number domain from wavelength domain Point, then carries out wavelet decomposition, arranges layer signal every after decomposing after threshold value is filtered and reconstructs.

As stated above, in described S2, take little 2～5 microns of bigness scale value longer than chamber a little as the search by hill climbing starting point that chamber is long.

As stated above, described S3 particularly as follows:

Take the long g in chamber of search by hill climbing starting point, bring into below equation calculate discrete cavities long coefficient a (0):

$a\left(0\right)=|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)\cos (j4\mathrm{\π}\·k(n)\·g){|}^2+|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)\sin (j4\mathrm{\π}\·k(n)\·g){|}^2$

In formula: k (n) is wave number, x (n) is the output spectrum of sensor, and n is the pixel sequence number of ccd array, and N is sampling number,

First with a larger step size, make the long g in chamber increase successively, and to calculate discrete cavities corresponding under the long g of different cavity long simultaneously Coefficient a (s) (s=0,1,2 ...), and the coefficient of subsequent calculations gained is compared with current coefficient successively, until a (s-1) > a (s), then Being search by hill climbing starting point with the long g in chamber that a (s) is corresponding, opposite direction makes the long g in chamber be sequentially reduced with less step-length, and calculates different cavity Long coefficient a (t) of discrete cavities corresponding under long g (t=0,1,2 ...), compare, equally as a (t-1) > a (t) time, more corresponding with a (t) The long g in chamber be search by hill climbing starting point, make the long g in chamber increase successively with less step-length, compare, through the most so reducing Step-length scans for back and forth, and until reaching default aimed at precision, the step-length set by last search is the solution of algorithm Adjust resolution, and chamber corresponding to discrete cavities long coefficient maximum that final step obtains is long, is required F-P sensor cavity long.

The invention have the benefit that

1, utilize novel variable step ramping constraint to reduce the amount of calculation of discrete cavities length demodulating algorithm, thus improve computing speed Degree, searches plain step-length by setting simultaneously and reaches to put forward high-precision purpose, finally realize the quick high accuracy of fibre-optical F-P sensor Signal demodulates.

2, after wavelet decomposition, every layer signal is arranged after threshold value is filtered and reconstructs again, improve the essence of final demodulation result further Degree.

Accompanying drawing explanation

Fig. 1 is the demodulating system structural representation of one embodiment of the invention.

Fig. 2 is the spectroscopic data figure directly collected through micro spectrometer.

Fig. 3 is interference spectrum signal result after sym4 small echo 5 layers decomposes.

Fig. 4 is the spectrogram after discrete wavelet threshold denoising.

Fig. 5 is variable step search by hill climbing result of calculation.

In figure: 1-wideband light source, 2-fibre-optical F-P sensor, 3-photoelectric conversion module, 4-data acquisition module, 5-processor, 6-bonder.

Detailed description of the invention

Below in conjunction with instantiation and accompanying drawing, the present invention will be further described.

The present invention provides the quick high accuracy signal demodulating method of a kind of fibre-optical F-P sensor, as it is shown in figure 1, wideband light source 1 The light sent passes through bonder 6, returns after the fibre-optical F-P sensor 2 that chamber length is to be measured, then is exported by bonder 6, After carry out opto-electronic conversion and conditioning through photoelectric conversion module 3 and process and obtain spectroscopic data, data acquisition module 4 gather, enter Entering the calculation process that processor carries out being correlated with, it comprises the following steps:

S1, by wavelet threshold denoising method, obtained spectroscopic data is carried out denoising；

S2, according to the spectrum after denoising, by Fourier transformation demodulating algorithm, the chamber calculating fibre-optical F-P sensor is long, As the long bigness scale in chamber value, and determine the search by hill climbing starting point of chamber length；

S3, the result that S2 is obtained, utilize variable step ramping constraint more each step-length cavity of resorption length and previous chamber length from Dissipate the long coefficient in chamber, until a coefficient searches the peak value under this step-length, as climbing the mountain next time greatly after previous coefficient ratio Search starting point, the step-length that search sets every time is respectively less than last step-length, until it reaches default aimed at precision, for the last time Step-length set by search is the demodulation resolution of algorithm, and the chamber that the discrete cavities long coefficient maximum that obtains of final step is corresponding Long, it is required fibre-optical F-P sensor chamber long.

Preferably, described S1 carries out interpolation increase sampled point particularly as follows: first spectroscopic data is transformed to wave-number domain from wavelength domain, Then carry out wavelet decomposition, layer signal every after decomposing is arranged after threshold value is filtered and reconstructs again.Spectral signal is improved by denoising Signal to noise ratio, to reach to improve the purpose of final demodulation result precision.

S2 finds the serial number of fundametal compoment peak value corresponding data corresponding to chamber length after spectroscopic data is carried out fast Fourier transform P, then it is long, as the long bigness scale in chamber value to obtain the chamber of fibre-optical F-P sensor by formula (2) after being corrected p by formula (1).

$p_p=p+\frac{ln\frac{A(p-1)}{A\left(p\right)}-ln\frac{A(p+1)}{A\left(p\right)}}{2(ln\frac{A(p-1)}{A\left(p\right)}+ln\frac{A(p+1)}{A\left(p\right)})}---\left(1\right),$

$d=\frac{p_p}{2N\·\mathrm{\δ}k}---\left(2\right),$

In formula, Pp be calibrated after data sequence number, A (p) is the Fourier Transform Coefficients corresponding to serial number p, A (p+1) being the Fourier Transform Coefficients corresponding to serial number p+1, A (p-1) is the Fourier transformation corresponding to serial number p-1 Coefficient, N is sampling number, and δ k is wave number interval, and d is required chamber long value.

Preferably, in described S2, take little 2～5 microns of bigness scale value longer than chamber a little as the search by hill climbing starting point that chamber is long.

It is further preferred that described S3 particularly as follows:

Take the long g in chamber of search by hill climbing starting point, bring into below equation (3) calculate discrete cavities long coefficient a (0):

$a\left(0\right)=|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)cos(j4\mathrm{\π}\·k(n)\·g){|}^2+|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)sin(j4\mathrm{\π}\·k(n)\·g){|}^2---\left(3\right),$

In formula: k (n) is wave number, x (n) is the output spectrum of sensor, and n is the pixel sequence number of ccd array, and N is sampling number. Wherein, the parts during ccd array is photoelectric conversion module.

First with a larger step size, make the long g in chamber increase successively, and to calculate discrete cavities corresponding under the long g of different cavity long simultaneously Coefficient a (s) (s=0,1,2 ...), and the coefficient of subsequent calculations gained is compared with current coefficient successively, until a (s-1) > a (s), then Being search by hill climbing starting point with the long g in chamber that a (s) is corresponding, opposite direction makes the long g in chamber be sequentially reduced with less step-length, and calculates different cavity Long coefficient a (t) of discrete cavities corresponding under long g (t=0,1,2 ...), compare, equally as a (t-1) > a (t) time, more corresponding with a (t) The long g in chamber be search by hill climbing starting point, make the long g in chamber increase successively with less step-length, compare, through the most so reducing Step-length scans for back and forth, and until reaching default aimed at precision, the step-length set by last search is the solution of algorithm Adjust resolution, and chamber corresponding to discrete cavities long coefficient maximum that final step obtains is long, is required F-P sensor cavity long.

In the present embodiment, take a chamber length and be about the optical fiber EFPI sensor of 180 μm as test object, according to Fig. 1 be System carries out test sampling, wherein, the centre wavelength of wideband light source 1 is 1550nm, bandwidth 100nm, data acquisition module 4 Spectrum acquisition scope be 1520nm-1570nm, the host computer journey that processor 5 is write according to demodulating algorithm proposed by the invention Sequence locates the spectroscopic data that reason data acquisition module 4 gathers in real time.Fig. 2 is the light directly collected through data acquisition module 4 Modal data, it can be seen that there is a lot of burr in spectral peak.Fig. 3 is that interference spectrum signal is after sym4 small echo 5 layers decomposes Result, based on mallat algorithm, the data length of every layer is the 2 of primary signal^-j, j is Decomposition order, can also see simultaneously It is concentrated mainly on D to high-frequency noise₅-D₁Detail coefficients in, therefore need D₅-D₁Detail coefficients set threshold value carry out denoising, Here use heuristic threshold value as Rule of judgment.By the D after threshold denoising₅-D₁Detail coefficients is together with approximation coefficient A₅Logical Crossing the interference spectrum signal of discrete wavelet inverse transformation reconstruct F-P sensor, its result is as shown in fig. 4, it can be seen that in spectral peak Burr is removed, and spectrum becomes smooth.

The chamber length of this optical fiber EFPI sensor calculated by fft algorithm as bigness scale value, result of calculation is 181.534245μm.Arranging at 3 μm less than this value is search by hill climbing starting point, with 1 μm as step-length, starts search by hill climbing, directly To the later number finding first mountain top.The long coefficient of calculated discrete cavities is about shown in distribution such as Fig. 5 (a) of chamber length. Last number 180.534245 μm that first time is obtained by search by hill climbing process for the second time is as starting point, and opposite direction carries out climbing the mountain searching Rope, using 100nm as step-length, the long coefficient of calculated discrete cavities is about shown in distribution such as Fig. 5 (b) of chamber length.For the third time Search by hill climbing is starting point by last number 179.834245 μm secondary, using second time search opposite direction as searcher To, with 10nm as step-length, shown in its result such as Fig. 5 (c).Last number of 4th search by hill climbing third time 179.974245 μm are starting point, using third time search opposite direction as the direction of search, with 1nm as step-length, its result such as Fig. 5 Shown in (d).The mountain peak finally searched out, at 179.960245 μm, is required chamber long.Through 4 search, algorithm Resolution has reached nanometer scale, as needed higher resolution, and only need to be with less step-length repeat the above steps.Above-mentioned Calculating process has done 43 discrete cavities Long operation altogether, if centered by 181.534245 μm, 3 μm as search radius, 1nm is step-size in search, i.e. uses general search algorithm, then needs to do 6000 discrete cavities Long operation, if demodulation chamber length is differentiated Rate orientates 1pm as, then need to carry out the search by hill climbing process under 3 different step-lengths on the basis of the 4th search by hill climbing result again, Variable step search by hill climbing method need to carry out about 80 computings altogether, and general search method need to carry out 6000000 computings, it is seen that uses Variable step hill-climbing algorithm greatly reduces the amount of calculation needed for search peak, lower than the amount of calculation of general search algorithm 4 quantity Level, reduces the computation complexity of program, improves measurement efficiency so that demodulating algorithm can realize the solution of fast hi-resolution Adjust.

Above example is merely to illustrate design philosophy and the feature of the present invention, its object is to make those skilled in the art's energy Solution present disclosure much of that is also implemented according to this, and protection scope of the present invention is not limited to above-described embodiment.So, all according to this Equivalent variations that bright disclosed principle, mentality of designing are made or modification, all within protection scope of the present invention.

Claims (4)

1. a quick high accuracy signal demodulating method for fibre-optical F-P sensor, the light that wideband light source sends passes through bonder, warp Return after the fibre-optical F-P sensor that chamber length is to be measured, then exported by bonder, eventually pass opto-electronic conversion and conditioning processes and obtains Spectroscopic data, it is characterised in that: it comprises the following steps:

S1, by wavelet threshold denoising method, obtained spectroscopic data is carried out denoising；

S2, according to the spectrum after denoising, by Fourier transformation demodulating algorithm, the chamber calculating fibre-optical F-P sensor is long, As the long bigness scale in chamber value, and determine the search by hill climbing starting point of chamber length；

S3, the result that S2 is obtained, utilize variable step ramping constraint more each step-length cavity of resorption length and previous chamber length from Dissipate the long coefficient in chamber, until a coefficient searches the peak value under this step-length, as climbing the mountain next time greatly after previous coefficient ratio Search starting point, the step-length that search sets every time is respectively less than last step-length, until it reaches default aimed at precision, for the last time Step-length set by search is the demodulation resolution of algorithm, and the chamber that the discrete cavities long coefficient maximum that obtains of final step is corresponding Long, it is required fibre-optical F-P sensor chamber long.

The quick high accuracy signal demodulating method of fibre-optical F-P sensor the most according to claim 1, it is characterised in that: institute The S1 stated carries out interpolation increase sampled point particularly as follows: first spectroscopic data is transformed to wave-number domain from wavelength domain, then carries out little wavelength-division Solve, layer signal every after decomposing is arranged after threshold value is filtered and reconstructs again.

The quick high accuracy signal demodulating method of fibre-optical F-P sensor the most according to claim 1, it is characterised in that: institute In the S2 stated, take little 2～5 microns of bigness scale value longer than chamber a little as the search by hill climbing starting point that chamber is long.

The quick high accuracy signal demodulating method of fibre-optical F-P sensor the most according to claim 1, it is characterised in that: institute The S3 that states particularly as follows:

Take the long g in chamber of search by hill climbing starting point, bring into below equation calculate discrete cavities long coefficient a (0):

$a\left(0\right)=|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)cos(j4\mathrm{\π}\·k(n)\·g){|}^2+|\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)sin(j4\mathrm{\π}\·k(n)\·g){|}^2$

In formula: k (n) is wave number, x (n) is the output spectrum of sensor, and n is the pixel sequence number of ccd array, and N is sampling number；

First with a larger step size, make the long g in chamber increase successively, and to calculate discrete cavities corresponding under the long g of different cavity long simultaneously Coefficient a (s) (s=0,1,2 ...), and the coefficient of subsequent calculations gained is compared with current coefficient successively, until a (s-1) > a (s), then Being search by hill climbing starting point with the long g in chamber that a (s) is corresponding, opposite direction makes the long g in chamber be sequentially reduced with less step-length, and calculates different cavity Long coefficient a (t) of discrete cavities corresponding under long g (t=0,1,2 ...), compare, equally as a (t-1) > a (t) time, more corresponding with a (t) The long g in chamber be search by hill climbing starting point, make the long g in chamber increase successively with less step-length, compare, through the most so reducing Step-length scans for back and forth, and until reaching default aimed at precision, the step-length set by last search is the solution of algorithm Adjust resolution, and chamber corresponding to discrete cavities long coefficient maximum that final step obtains is long, is required F-P sensor cavity long.