A botdr temperature error estimation method

Abstract

The application discloses a BOTDR temperature error estimation method. In a BOTDR system, the Brillouin frequency shift of the Brillouin spectrum signal is extracted by quadratic polynomial fitting, and the fiber temperature is calculated accordingly. The method is to calculate the standard deviation of the fiber temperature caused by noise according to the formula, and the measurement error of the fiber temperature is represented by the standard deviation. In the formula, σ T is the standard deviation of the fiber temperature; N is the number of sampling points; δ is the sweep interval; SNR A is the ratio of the Brillouin gain peak value to the standard deviation of the noise obeying the normal distribution; C vT is the temperature coefficient of the Brillouin frequency shift change; Δη is the difference between the maximum and minimum values of the quadratic polynomial curve obtained by fitting the normalized Brillouin spectrum in the sweep range, wherein the normalization means that the maximum value of the original signal is scaled to 1. The above parameters are known quantities except Δη, which is obtained according to the sweep range and the Brillouin shape interpolation. The accuracy of the method of the application is better than that of the existing typical method.

Description

TECHNICAL FIELD

[0001] The application relates to a BOTDR temperature error estimation method and belongs to the technical field of measurement. BACKGROUND

[0002] Distributed optical fiber sensors have the characteristics of strong anti-interference ability, low loss, high precision and high sensitivity, and are widely applied in civil engineering, aerospace, power industry and petrochemical industry and other fields. The distributed optical fiber sensor based on Brillouin scattering can effectively measure the strain and temperature change of the optical fiber, and can realize rapid positioning of faults.

[0003] The distributed optical fiber sensor based on Brillouin scattering has been studied by many scholars in recent decades due to its superior performance. The principle is to obtain the Brillouin frequency shift by analyzing the Brillouin spectrum, and then calculate the temperature and strain of the optical fiber according to the Brillouin frequency shift. Due to the existence of noise, the frequency measured by the sensor at the discrete Brillouin spectrum peak is often different from the accurate value of the Brillouin frequency shift. To solve this problem, the least square fitting model method is often used to reconstruct the Brillouin spectrum. Common fitting models include Lorentz model, Gaussian model, Voigt model, pseudo-Voigt model and quadratic polynomial model. Among them, the Lorentz model approximately conforms to the Brillouin spectrum when the incident light pulse width is very large (> 50 ns), the Gaussian model approximately conforms to the Brillouin spectrum when the incident light pulse width is very small (< 10 ns), the Voigt model and the pseudo-Voigt model approximately conform to the Brillouin spectrum under most incident light pulse widths, the former has higher accuracy but the formula is complex, it is obtained by convolution of the Lorentz model and the Gaussian model, and the fitting speed is slow, and the latter is obtained by linear addition of the Lorentz model and the Gaussian model, which improves the fitting speed while ensuring reliability. Compared with the above four models, the quadratic polynomial model has incomparable speed of the above four models under the premise of ensuring reliability, and has great advantages in occasions with real-time requirements. Therefore, the application analyzes the measurement error of the fitting method based on the quadratic polynomial model.

[0004] The Brillouin frequency shift and the temperature and strain of the optical fiber satisfy a certain linear relationship, and in the case that the strain of the optical fiber along the line is unchanged, the temperature of the optical fiber can be calculated according to the size of the Brillouin frequency shift. Due to the existence of noise and other factors in the environment, the actual measured data often has noise, and the temperature of the optical fiber calculated by using the data will have errors. However, the current Brillouin optical time domain reflectometer (BOTDR, Brillouin optical time-domain reflectometry) temperature error estimation method has defects, and its accuracy will be greatly reduced with the increase of the sweep range. There are limitations in application, so there is a certain market demand for BOTDR temperature error estimation methods with higher accuracy and stronger applicability. SUMMARY

[0005] The present application aims at the drawbacks of the prior art, and provides a BOTDR temperature error estimation method, which ensures accuracy and enhances applicability.

[0006] To achieve the above object, the technical scheme adopted by the present application is as follows:

[0007] A BOTDR temperature error estimation method, which is to extract the Brillouin spectrum parameters by using quadratic polynomial fitting in the Brillouin spectrum signal output by a BOTDR distributed optical fiber sensor, and then to calculate the standard deviation of the optical fiber temperature caused by noise according to the following formula, so as to reflect the measurement error of the optical fiber temperature by the standard deviation:

[0008]

[0009] In the formula, σ T is the standard deviation of the optical fiber temperature; N is the number of sampling points; δ is the sweep interval; SNR A is the ratio of the Brillouin gain peak value to the standard deviation of the noise obeying normal distribution; C vT is the temperature coefficient of the Brillouin frequency shift change; Δη is the difference between the maximum and minimum values of the quadratic polynomial curve obtained by fitting the normalized Brillouin spectrum in the sweep range, wherein the normalization means that the maximum value of the original signal is scaled to 1.

[0010] Further, Δη in the formula is obtained by interpolation according to the sweep range and the Brillouin shape, and the specific data is referred to Table 1.

[0011] The present application designs a BOTDR temperature error estimation method, and has the advantages compared with the prior art as follows: the accuracy of the present application is better than that of the prior art under most parameter settings. Especially when the sweep range is large, the present application can still maintain high accuracy, while the accuracy of the prior art is greatly reduced, so the present application has stronger applicability. BRIEF DESCRIPTION OF DRAWINGS

[0012] The present application will be further described in detail below with reference to the accompanying drawings.

[0013] Figure 1 is the calculation result of the optical fiber temperature standard deviation corresponding to different methods, and the simulation spectrum is shown in FIG. 1;

[0014] Figure 2 is the measured Brillouin spectrum obtained after superposition and averaging processing;

[0015] Figure 3 is the calculation result of the optical fiber temperature standard deviation under different calculation methods, and the measured spectrum is shown in FIG. 2;

[0016] Figure 4is the relationship between the error of the standard deviation of the fiber temperature and the Brillouin spectrum shape under different calculation methods, the simulation spectrum, wherein Figure 4 The Brillouin spectrum parameters of (a) are shown in typical value A, Figure 4 The Brillouin spectrum parameters of (b) are shown in typical value B;

[0017] Figure 5 is the relationship between the error of the standard deviation of the fiber temperature and the sweep range under different calculation methods, the simulation spectrum, wherein Figure 5 The Brillouin spectrum parameters of (a) are shown in typical value A, Figure 5 The Brillouin spectrum parameters of (b) are shown in typical value B;

[0018] Figure 6 is the relationship between the error of the standard deviation of the fiber temperature and the signal-to-noise ratio under different calculation methods, the simulation spectrum, wherein Figure 6 The Brillouin spectrum parameters of (a) are shown in typical value A, Figure 6 The Brillouin spectrum parameters of (b) are shown in typical value B;

[0019] Figure 7 is the relationship between the error of the standard deviation of the fiber temperature and the linewidth under different calculation methods, the simulation spectrum, wherein Figure 7 The Brillouin spectrum parameters of (a) are shown in typical value A, Figure 7 The Brillouin spectrum parameters of (b) are shown in typical value B;

[0020] Figure 8 is the relationship between the error of the standard deviation of the fiber temperature and the sweep interval under different calculation methods, the simulation spectrum, wherein Figure 8 The Brillouin spectrum parameters of (a) are shown in typical value A, Figure 8 The Brillouin spectrum parameters of (b) are shown in typical value B; DETAILED DESCRIPTION

[0021] The application will be further described in detail below with reference to the accompanying drawings.

[0022] 1. Introduction of the existing method

[0023] The existing error estimation method is to calculate the standard deviation of the fiber temperature caused by noise after extracting the Brillouin spectrum parameters by using formula (1) or (2). For the convenience of description, two methods are defined as existing method 1 and existing method 2.

[0024]

[0025] In the formula, σ T is the standard deviation of the fiber temperature; SNR A is the ratio of the Brillouin gain peak value to the standard deviation of the noise obeying the normal distribution; C vT is the temperature coefficient of the Brillouin frequency shift change; δ is the sweep interval; ΔvB is the line width; η is the ratio of the minimum value to the peak value of the Brillouin gain in the sweep range.

[0026]

[0027] where v B is the Brillouin shift; Q is the ratio of v B to Δv B ; SNR B is the ratio of the mean square of the Brillouin gain in the line width range to the variance of the noise subject to normal distribution.

[0028] The Brillouin spectrum generated by numerical simulation verifies the shortcomings of the above two methods. The expression of the noisy Brillouin spectrum is as follows:

[0029]

[0030] where g B is the Brillouin gain; v is the frequency; g0 is the peak value of the Brillouin gain; g1 and g2 are the peak values of the Lorentz and Gaussian model Brillouin gains respectively, and the sum of the two is always 1; and randn is a random variable subject to normal distribution with a mean of 0 and a standard deviation of 1.

[0031] For the convenience of calculation, the calculation in the specification assumes that C vT = 1 MHz / ℃. In actual cases, C vT will change with the physical properties of the optical fiber and the wavelength of the incident light, and needs to be measured in application. The specific parameters of the noisy Brillouin spectrum are as follows: the sweep range varies in the range of 1-4 times the line width; the Brillouin shift is 10.7 GHz; the line width is 50 MHz; the sweep interval is 1 MHz; g0 = 1; g1 = 0.5; and SNR A = 100. Based on formula (3), 10,000 sets of Brillouin spectra are generated, and the standard deviations of the fiber temperature calculated by the existing methods 1 and 2 are compared with the simulation results, as shown in FIG. 1. Figure 1

[0032] As can be seen from FIG. 1, the existing methods 1 and 2 have good effects only when the sweep range is appropriate, and their use still has limitations. Figure 1

[0033] 2 Comparison of the method of the present application with the existing methods

[0034] 2.1 Actual spectrum

[0035] ​​The present application builds an optical fiber Brillouin spectrum measurement system by using AV6419 optical time domain reflectometer produced by China Electronics Technology Instrumentation Co., Ltd., and selects a SM 9 / 125 μm optical fiber with a length of about 1 km. The sweep frequency range is 10.52-10.92 GHz, the sweep frequency interval is 1 MHz, the incident light pulse width is 200 ns, the sampling resolution is 10 m, and the superposition average number is 2 10 and 2 18 . The experiment is carried out at room temperature, and the Brillouin spectrum is obtained by 2 18 times of superposition average Figure 2 . In order to improve the accuracy of the optical fiber temperature error estimation, the pseudo-Voigt model is used to fit the Brillouin spectrum with a superposition number of 2 18 , so as to extract the corresponding Brillouin spectrum parameters as a reference. Thus, the Brillouin frequency shift of the Brillouin spectrum is 10.7234 GHz, the Brillouin line width is 41.1 MHz, and g1 is 0.5107. Taking the Brillouin spectrum with a superposition number of 2 18 as a reference, the difference between the Brillouin spectrum with a superposition number of 2 10 is calculated as the size of the noise.

[0036] According to the above Brillouin spectrum parameters and Table 1, the size of Δη under the corresponding sweep frequency range and Brillouin spectrum shape (g1) can be calculated by using the interpolation method. Then, the required parameters are brought into formula (4), and the standard deviation of the optical fiber temperature under the corresponding parameters can be calculated.

[0037]

[0038] In the formula, Δη is the difference between the maximum and minimum values of the quadratic polynomial curve obtained by fitting the normalized Brillouin spectrum in the sweep frequency range.

[0039] The standard deviations of the optical fiber temperature calculated by using the existing methods 1, 2 and the method of the present application are compared with the standard deviation of the optical fiber temperature calculated by using the measured Brillouin spectrum data, and the results are shown in Figure 3 .

[0040] As shown in Figure 3 , the method of the present application has higher accuracy than the existing methods in this experiment, especially in the case of a larger sweep frequency range, the accuracy of the method of the present application is much higher than that of the existing methods. It is found by calculation that for the above measured Brillouin spectrum, the error of the method of the present application is only 1 / 47 of the existing method 1 and 1 / 42 of the existing method 2.

[0041] 2.2 Numerical generation spectrum

[0042] The influence of Brillouin spectrum shape, sweep range, signal-to-noise ratio, line width and sweep interval on the accuracy of the method is studied respectively. The noise-containing Brillouin spectrum is generated based on formula (3), and the relevant parameters are changed in the following ranges: 1) g1 is changed in the range of 0-1; 2) the sweep range is changed in the range of 1-4 times the line width; 3) SNR A is changed in the range of 10-10 5 4) the line width is changed in the range of 30-150 MHz; 5) the sampling interval is changed in the range of 1-5 MHz. When studying a certain factor, the other factors are selected as follows: typical value A: the sweep range is 2 times the line width, SNR A = 10, the line width is 140 MHz, g1 = 0.2, N = 141, δ = 2 MHz, g0 = 1; typical value B: the sweep range is 4 times the line width, SNR A = 100, the line width is 50 MHz, g1 = 0.6, N = 121, δ = 1 MHz, g0 = 1. The errors of the existing methods 1, 2 and the method of the application are explored respectively under different Brillouin spectrum shapes, sweep ranges, signal-to-noise ratios, Brillouin line widths and sweep intervals. Since the number of sweep points is determined by the sweep range and the sweep interval, the size of the error under different sweep point numbers can be indirectly reflected by studying the size of the error under different sweep ranges and sweep intervals, which is not discussed separately here.

[0043] The standard deviations of the fiber temperature calculated by the methods 1, 2 and the method of the application are compared with the standard deviations of the fiber temperature calculated from the numerically generated Brillouin spectrum, and the results are shown in Figures 4 to 8 .

[0044] It can be seen from Figures 4 to 8 that the accuracy of the existing methods 1 and 2 is higher than that of the method of the application only in a few cases. In most cases, the accuracy of the method of the application is much higher than that of the existing methods 1 and 2. Therefore, the method of the application has a wider range of application and higher accuracy than the existing methods. It is found by calculation that for the above numerically generated Brillouin spectrum, the error of the method of the application is only 1 / 26 of that of the existing method 1 and 1 / 12 of that of the existing method 2.

[0045] Table 1 Size of Δη under different sweep ranges and Brillouin spectrum shapes

[0046]

Claims

1. A method for estimating temperature error in a BOTDR system, characterized in that, The method involves extracting Brillouin spectrum parameters from the Brillouin spectrum signal output by the BOTDR distributed optical fiber sensor using a quadratic polynomial fitting, and then calculating the standard deviation of the fiber temperature caused by noise according to the following formula, using the standard deviation to represent the measurement error of the fiber temperature: In the formula, σ T δ is the standard deviation of fiber temperature; N is the number of sampling points; δ is the sweep frequency interval; SNR A C is the ratio of the peak Brillouin gain to the standard deviation of noise that follows a normal distribution; vT η is the temperature coefficient of the Brillouin frequency shift; Δη is the difference between the maximum and minimum values ​​of the quadratic polynomial curve obtained by fitting the normalized Brillouin spectrum within the frequency sweep range, where normalization means scaling the maximum value of the original signal to 1.

2. The BOTDR temperature error estimation method according to claim 1, characterized in that, The Δη in the formula is obtained by interpolation based on the sweep frequency range and the shape of the Brillouin spectrum.

Images