Optical fiber shape measurement device and method based on phase demodulation

Through the calibration of the sweep range of the tuned laser in fiber shape sensing technology and the differential phase denoising processing of the artificial neural network model, the problem of sweep range deviation and phase noise interference in fiber shape sensing is solved, and the sensing accuracy is improved And reduces system complexity.

CN118565373BActive Publication Date: 2024-11-19HUAZHONG UNIV OF SCI & TECH
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411017876.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-19
Estimated Expiration
2044-07-29

AI Technical Summary

Technical Problem

In the existing fiber-shaped sensing technology, the phase demodulation optical frequency domain reflectometer is positional deviation due to the swept frequency range deviation during resampling, and is susceptible to phase noise and error interference, affecting the sensing accuracy.

Method used

By calibrating the sweeping range of the tuned laser, the differential phase is denoised using an artificial neural network model to compensate for the misalignment caused by strain, and the phase jump points are removed through Fourier transform and SG filtering smoothing treatment to achieve shape reconstruction .

Benefits of technology

It improves the accuracy of fiber shape sensing, reduces system complexity and cost, enhances the denoising effect of differential phases, and avoids complex calculation processes, direct mapping from phase to shape.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118565373B_ABST
    Figure CN118565373B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of optical fiber shape sensing, and specifically relates to an optical fiber shape measurement device and method based on phase demodulation, which performs shape sensing on a multi-core optical fiber. The present invention first calibrates the tunable laser sweep range of the measurement signal and the reference signal, obtains the differential phase by subtracting the phase of the measurement signal from the reference signal, and further filters it to remove the phase jump. Finally, the denoised differential phase and the artificial neural network model are used to reconstruct the shape. The present invention solves the problem of position deviation caused by the deviation of the sweep range during the resampling process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of optical fiber shape sensing, and in particular relates to an optical fiber shape measuring device and method based on phase demodulation and neural network, and performs shape sensing on multi-core optical fibers. Background Art

[0002] Fiber shape sensing measures the strain of the cross section of a multi-core fiber and constructs the shape and position of the fiber through the relationship between the strains of multiple cores. Distributed fiber shape sensing technology based on optical frequency domain reflectometry has the advantages of no need for visual assistance, anti-electromagnetic interference, and compact results. Therefore, this technology has a wide range of application value in civil engineering, aerospace, biomedicine and other fields. How to improve the accuracy of shape sensing is a key issue for the technology to move towards practical application. To this end, the present invention proposes a shape sensing method based on phase demodulation optical frequency domain reflectometry and artificial neural network to improve the accuracy of shape sensing.

[0003] First of all, shape sensing based on phase demodulation optical frequency domain reflectometer requires two measurements of the optical fiber, that is, the optical fiber is used as a reference signal in a non-deformed state, and as a measurement signal in a state where the optical fiber is deformed. When the reference signal and the measurement signal are resampled by software, there will be a slight difference in the sweep range of the tuned laser during the two measurements, resulting in a slight difference in the theoretical spatial resolution of the two measurements, that is, the step size of adjacent data points. This deviation will accumulate with the increase of the optical fiber length, causing the error of the demodulated differential phase to gradually increase, seriously affecting the sensing accuracy. To address this problem, most of the instruments that add calibration frequency to the system, such as the HCN gas chamber, but this will undoubtedly increase the complexity and cost of the system. Therefore, it is necessary to develop a simple and efficient method to solve this problem.

[0004] Secondly, the phase demodulation optical frequency domain reflectometer is more susceptible to interference from phase noise and errors. For example, when using optical fiber to measure static strain, the strain will cause the length of the optical fiber to change. This change will cause the corresponding data points to be misaligned, resulting in errors in the differential phase. At present, relevant scholars have proposed some algorithms to solve this problem. This problem also exists in optical fiber shape sensing based on phase demodulation optical frequency domain reflectometer, but no relevant papers have been seen to calibrate it. In addition, after the differential phase is unwound, there will be phase jump points of varying degrees on the trace. Since the phase needs to be derived to obtain the strain later, these jump points will seriously affect the strain results. Therefore, it is necessary to develop a fast and accurate algorithm to denoise the differential phase.

[0005] Third, machine learning has extraordinary advantages in classification, pattern recognition, and prediction, and has been widely used in various fields. Machine learning-based methods can obtain rich information from raw data, and establish a mapping model between phase and curvature by directly learning the relationship between the measured phase information and curvature. In actual measurement, the shape of the optical fiber is directly demodulated based on the measured phase information and the mapping model, avoiding the complex calculations from phase to shape fitting. Summary of the invention

[0006] In view of the defects of the prior art, the present invention proposes a device and method for measuring the shape of an optical fiber based on phase demodulation and neural network. First, the scanning range of the tuning laser of the measurement signal and the reference signal is calibrated, and the differential phase is obtained by subtracting the phase of the measurement signal from the reference signal, and then the differential phase is further filtered to remove the phase jump. Finally, the shape is reconstructed using the denoised differential phase and the artificial neural network model. The present invention solves the problem of position deviation caused by the deviation of the scanning range during the resampling process.

[0007] The technical solution of the present invention is:

[0008] Tuned laser sweep range calibration:

[0009] Two experiments were conducted. The first experiment was conducted with the fiber in its natural state as the reference signal. The second experiment was conducted with the fiber shaped as the measurement signal. The auxiliary interferometer data of the reference signal and the measurement signal were taken to obtain the sweep frequency range of the tuned laser for the two measurements using the following formula: ; ; v r 、v m They refer to the frequency sweep range of the reference signal and the frequency sweep range of the measurement signal respectively. I r , I m are the original data of the reference signal and the measurement signal auxiliary interferometer, is the auxiliary interferometer delay, H {} is the Hilbert transform. Compare v r 、v m The same part and construct a new sweep range v , v Included v r 、v m Same sweep section. v The specific expression is as follows: ; vr_min and v m_min for v r and v m The minimum value of v r_max and v m_max for v r and v m The maximum value of A and B, max[A,B] means taking the maximum value of A and B, and min[A,B] means taking the minimum value of A and B.

[0010] And v The main interferometer data of the reference signal and the measurement signal are respectively I R , I M Resample to get resampled data , ,in I R , I M Only include the same sweep frequency range v data, rather than the original data.

[0011] Misalignment calibration due to strain:

[0012] Data after resampling the reference signal and the measured signal , Fourier transform is performed to obtain the data in the distance domain and extract the phase. The difference between the phases of the two sets of signals is obtained to obtain the differential phase Δ φ , using Δ φ The methods for compensating for misalignment caused by strain are as follows:

[0013] Step 1: Divided into N segments, and for each segment, Δ φ Unwind and get the differential phase ΔΦ after each unwinding N ; Take ΔΦ N The last data ΔΦ in N_end Calculate the cumulative length change Δ of this segment L : , where n eff is the effective refractive index of the optical fiber core, λ0 is the initial wavelength, P e is the elastic-optical coefficient;

[0014] Step 2, calculate ΔL The phase shift Δ ϕ : , where v is the frequency sweep range, v 0 is the initial light frequency, c is the speed of light;

[0015] Step 3: Calculate the distance domain data of the compensated measurement signal I z : ;

[0016] Step 4: By cyclically compensating the cumulative length changes of all segments, the final compensated distance domain data is obtained, and the differential phase Δ is recalculated. φ .

[0017] Differential Phase Denoising:

[0018] The differential phase Δ φ The unwrapping is performed to obtain the unwrapped differential phase ΔΦ, and the differential phase is divided into multiple discontinuous parts due to the large phase jump to compensate for the existing scattered phase jump points. The specific method is as follows:

[0019] Step 1, take the derivative of ΔΦ to get d ΔΦ, find d All the values ​​in ΔΦ that are greater than the threshold S point z i , find the point z i forward N After N The mean of the ΔΦ data M 1 and M 2;

[0020] Step 2: If M 2− M 1 is greater than or equal to the set threshold S , it is believed that at this point z i After that, all phases have a large phase jump, so all ΔΦ data after that are subtracted from M 2− M 1;

[0021] Step 3: If the conditions in step 2 are not met, then it is considered that z i There is a single point with a large phase jump, and the ΔΦ value at this point, ΔΦ zi Replace with ;

[0022] Step 4: traverse all large phase jump positions in a loop and compensate for them, and perform SG filtering and smoothing processing on ΔΦ after removing the large phase jump.

[0023] Artificial neural network shape reconstruction principle:

[0024] Derivative the denoised differential phase ΔΦ to obtain d ΔΦ, trained using the training data set d ΔΦ vs. curvature κ The specific method is as follows:

[0025] Step 1: Bend the multi-core optical fiber into many different curvatures κ , and the derivative of the differential phase ΔΦ corresponding to different curvatures is calculated by the previous algorithm d ΔΦ, a large number of different curvatures κ and the corresponding differential phase derivative d ΔΦ is used to train the mapping model;

[0026] Step 2: bend the multi-core optical fiber into different curvatures and calculate the corresponding d ΔΦ, acts as test data, which is used as the input of the model to obtain the curvature predicted by the model and compared with the real curvature of the multi-core optical fiber to measure the curvature prediction accuracy of the model;

[0027] Step 3: bend the multi-core optical fiber into a two-dimensional curve. Different locations of the multi-core optical fiber have different curvatures. The derivative of the differential phase is calculated from the collected data. d ΔΦ is input into the trained model to obtain the curvature size at different positions of the multi-core optical fiber, and the Frenet-Serret equation is used to restore the shape of the multi-core optical fiber. The restored multi-core optical fiber shape and the actual multi-core optical fiber shape are used to measure the shape reconstruction accuracy.

[0028] Beneficial effects and innovative features of the present invention:

[0029] 1. The present invention calibrates the tuning laser sweep range of the reference signal and the measurement signal through a simple and efficient algorithm to solve the problem of position deviation caused by the deviation of the sweep range during the resampling process. Compared with other software calibration methods, the calibration method is simpler in principle and code, and can achieve good results. Compared with other hardware calibration methods, no additional device is added to the system to monitor the wavelength of the tuning laser, which reduces the complexity and cost of the system.

[0030] 2. The algorithm for calibrating the misalignment caused by strain in static strain measurement based on phase demodulation optical frequency domain reflectometer is used in optical fiber shape sensing. The misalignment caused by strain is calibrated to improve the accuracy of phase demodulation. A new algorithm is studied to denoise the phase noise of differential phase. The algorithm detects the position of large phase jumps in differential phase and compensates for it. While ensuring the denoising effect, it also retains the original phase information to the greatest extent.

[0031] 3. For the first time, the artificial neural network shape reconstruction method was used in optical fiber shape sensing based on phase demodulation optical time domain reflectometry. The curvature information of the optical fiber was directly obtained according to the measured multi-core optical fiber phase information and the trained mapping model, avoiding the complex calculation from phase to strain and then to curvature. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A diagram of a shape sensing system based on phase demodulation optical frequency domain reflectometry.

[0033] Figure 2 It is the flow chart of phase demodulation algorithm.

[0034] Figure 3 This is the differential phase diagram before and after sweep range calibration.

[0035] Figure 4 is the differential phase map after correction for the misalignment caused by strain.

[0036] Figure 5 These are the images of the unwrapped differential phase before and after denoising.

[0037] Figure 6 It is a designed artificial neural network model.

[0038] Explanation of the reference numerals: tunable laser 1, first 1:99 fiber coupler 2, first 1:1 fiber coupler 3, delay fiber 4, second 1:1 fiber coupler 5, first balanced detector 6, second 1:99 fiber coupler 7, polarization controller 8, fiber circulator 9, 1:1 polarization-maintaining fiber coupler 10, first polarization beam splitter 11, second polarization beam splitter 12, second balanced detector 13, third balanced detector 14, optical fiber 15, data acquisition card 16. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention and to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0040] First, the optical frequency domain reflectometer system used in the present invention is introduced. Figure 1 The optical fiber shape measurement device based on phase demodulation of the present invention comprises: a tunable laser 1, an auxiliary interferometer, a main interferometer and a data acquisition card 16. The tunable laser 1 of the present invention is divided into two optical paths through a first 1:99 optical fiber coupler 2, and the two optical paths are respectively connected to the auxiliary interferometer and the main interferometer, and the outputs of the auxiliary interferometer and the main interferometer are connected to the data acquisition card 16.

[0041] like Figure 1 As shown, the auxiliary interferometer of the present invention includes a first 1:1 fiber coupler 3, a delay fiber 4, a second 1:1 fiber coupler 5, and a first balanced detector 6. The light of the fiber coupler 2 with an energy share of 1% is connected to the first 1:1 fiber coupler 3. The first 1:1 fiber coupler 3 divides the light into two optical paths, one of which is connected to the second 1:1 fiber coupler 5 through a 60-meter delay fiber 4, and the other is directly connected to the second 1:1 fiber coupler 5. The two signals beat on the first balanced detector 6 through the second 1:1 fiber coupler 5. Generally, the length of the delay fiber 4 needs to be more than twice the length of the optical fiber 15.

[0042] like Figure 1 As shown, the main interferometer includes a second 1:99 fiber coupler 7, a polarization controller 8, a fiber circulator 9, a 1:1 polarization-maintaining fiber coupler 10, a first polarization beam splitter 11, a second polarization beam splitter 12, a second balanced detector 13, a third balanced detector 14, and an optical fiber 15. The light with an energy proportion of 99% of the first 1:99 fiber coupler 2 is connected to the second 1:99 fiber coupler 7, and the light with an energy proportion of 1% of the second 1:99 fiber coupler 7 is connected to the polarization controller 8. The output of the polarization controller 8 is connected to the 1:1 polarization-maintaining fiber coupler 10. The polarization controller 8 is used to adjust The intensities of P light and S light are made equal. The light with 99% of the energy of the second 1:99 fiber coupler 7 passes through the fiber circulator 9 and enters the optical fiber 15. The backward Rayleigh scattered light of the optical fiber 15 passes through the fiber circulator 9 and enters the 1:1 polarization-maintaining fiber coupler 10. The two output lights of the 1:1 polarization-maintaining fiber coupler 10 are respectively connected to the first polarization beam splitter 11 and the second polarization beam splitter 12. The polarization beam splitter divides the incident light into P light and S light. The two beams of P light beat on the second balanced detector 13, and the two beams of S light beat on the third balanced detector 14. The beat signals are collected by the data acquisition card 16.

[0043] The optical fiber 15 of the present invention is composed of an intermediate fiber core and six outer fiber cores. The distances from the six outer fiber cores to the center are equal, and the angle between adjacent fiber cores is 60°. Only three outer fiber cores will be used in the experiment, and the angle between adjacent fiber cores of the three fiber cores used is 120°. The measurement signal and the reference signal in the experiment include the collected auxiliary interferometer data and the main interferometer data, and the main interferometer data and the auxiliary interferometer data include the collected signals of the three fiber cores used in the experiment.

[0044] Two experiments were conducted. In the first experiment, the optical fiber was in a naturally extended state, and the signal collected in the first experiment was the reference signal. In the second experiment, the optical fiber was bent into shapes with different curvatures, and the signal collected in the second experiment was the measurement signal. The processing of the reference signal and the measurement signal is referred to Figure 2 ,The processing process is divided into three parts, namely: ,calibration of the frequency sweep range of the tunable laser, calibration of the ,misalignment caused by strain, and differential phase ,denoising.

[0045] The present invention also discloses a method for measuring optical fiber shape based on phase demodulation and neural network, comprising the following steps:

[0046] Step 1: Calibrate the frequency sweep range of the tuned laser

[0047] Step 11: Obtain auxiliary interferometer data of reference signal and measurement signal respectively I r , I m , perform Hilbert transform and find the frequency sweep range of the tunable laser for the reference signal and the measurement signal v r and v m ,Right now , , , is the length of the delay optical fiber 4. L d = 60 meters, is the effective refractive index of the fiber core, , c is the speed of light, is the delay of the optical fiber.

[0048] Step 12: Find the frequency sweep range v r and v m The minimum value of and find the maximum value between these two values ​​is recorded as v min ; Then find the sweep frequency range v r and v mThe maximum value of and find the minimum value between these two values ​​is recorded as v max ,use v min and v max Construct a new sweep range v =( v min , v max );

[0049] Step 13, obtain the reference signal and the measurement signal main interferometer data, and respectively extract the sweep frequency range from the original data as v The corresponding data acts as I R , I M ,use v The main interferometer data for the reference signal and the measured signal are I R , I M Perform cubic spline interpolation to obtain resampled data , ;

[0050] Step 14: Resample the data , Perform Fourier transform, extract the phase of the reference signal and the measured signal and make a difference to get the differential phase Δ φ .

[0051] Down Figure 3 The differential phase Δ is obtained before and after frequency sweep range calibration. φ From the results, we can see that the uncalibrated differential phase is very chaotic, and it is almost impossible to extract useful information from it. The calibrated differential phase shows regularity, but the misalignment caused by strain causes the differential phase to become very chaotic as the distance increases, so it needs to be compensated and subsequently denoised.

[0052] Step 2: Calibration of misalignment caused by strain

[0053] Step 21: The Δ obtained in step 1 φ Divided into N segments, and for each segment, Δ φ Unwind and get the differential phase ΔΦ after each unwinding N , take ΔΦ N The last data ΔΦ in N_end Calculate the cumulative length change Δ of this segment L ,Right now: ,in neff is the effective refractive index of the optical fiber core, λ0 is the initial wavelength, P e is the elastic-optical coefficient; λ 0 = 1510 μm, P e =0.22;

[0054] Step 22, calculate The phase change caused , , get the resampled data of the measured signal To compensate for it, , i represents the imaginary unit, and Fourier transform is performed on it to obtain the calibrated measurement signal distance domain data I z ,Right now ;

[0055] Step 23, Extract I z The phase of the reference signal is then subtracted to obtain the differential phase after the first calibration. ;

[0056] Step 24, by looping steps 21 to 23, the cumulative length changes of all segments are compensated to obtain the final compensated distance domain data. I z , and recalculate the calibration differential phase Δ φ .

[0057] Down Figure 4 is the Δ after strain calibration φ , it can be seen from the figure that compared with the differential phase before calibration, the Δ φ There is no more confusion at the far end, and from the local magnified image we can see that strain information can be extracted from it.

[0058] Step 3: Differential Phase Denoising

[0059] The differential phase Δ after calibration in step 24 φ Unwinding is performed to obtain ΔΦ. The differential phase is divided into multiple discontinuous parts due to the large phase jump, and the scattered phase jump points are compensated. The specific method is as follows:

[0060] Step 31, derive ΔΦ to obtain d ΔΦ, find d All points in ΔΦ that are greater than the threshold π / 2 z i , find the point z i The mean of the first 10 and last 10 ΔΦ data M 1 andM 2;

[0061] Step 32, if M 2− M 1 is greater than or equal to the set threshold π / 2, it is considered that z i After that, all phases have a large phase jump, so all ΔΦ data after that are subtracted from M 2− M 1;

[0062] Step 33: If the condition of step 32 is not met, it is considered that z i There is a single point with a large phase jump, and the ΔΦ value at this point, ΔΦ zi Replace with ;

[0063] Step 34, by looping through all large phase jump positions and compensating for them, SG filtering and smoothing are performed on the ΔΦ after the large phase jump is removed. The data collected by the three fiber cores are processed respectively, and in step 34, the data ΔΦ1, ΔΦ2, and ΔΦ3 after the data collected by the three fiber cores are processed can be obtained respectively.

[0064] Down Figure 5 Figure 3 is a schematic diagram of ΔΦ before and after denoising. From the local enlarged image, it can be seen that there are many large phase jumps in ΔΦ before denoising, which causes it to be divided into many discontinuous parts and some scattered phase jump points. After denoising, the discontinuous parts in ΔΦ are compensated, and many large phase jump points are removed, making the entire ΔΦ smoother.

[0065] Step 4: Artificial Neural Network Shape Reconstruction

[0066] The model of the designed artificial neural network can be found in Figure 6 The model is divided into three layers, namely the input layer, the hidden layer, and the output layer. The data collected by the three cores of the multi-core optical fiber used for sensing are processed as described above to obtain ΔΦ1, ΔΦ2, and ΔΦ3 (in this embodiment, only three of the six outer cores are used, so when collecting data, only the data of three cores will be collected, and the other three cores will not be used). After taking the derivative, the input layer data is obtained d ΔΦ1, d ΔΦ2, d ΔΦ3, the output layer is the corresponding multi-core optical fiber bending curvature κ, During the experiment, the optical fiber was bent into different curvatures using a curvature calibration plate. The curvature range was 3 -1 m to 20 -1m, a total of 200 different sets of data were collected within this curvature range, of which 150 sets of data were used to train the model as training sets, and 50 sets of data were used as test sets. Note that the test set data cannot be used for model training. d ΔΦ data is input into the trained model, and the curvature of the model output is compared. κ The correctness of the model is measured by comparing it with the actual curvature.

[0067] Finally, the multi-core optical fiber is bent into a curve with different curvatures at different positions. d The ΔΦ data is input into the model to obtain the curvature of the multi-core optical fiber at different positions. The shape of the curve is restored using the Frenet-Serret equation. The specific restoration method is as follows:

[0068] Step 41, assuming that the natural parameter equation of the multi-core optical fiber is: r ( s )= x ( s ) i+y ( s ) j , s is the arc length of the curve, i , j for x , y axis unit vector;

[0069] Step 42: unit tangent vectors at each point of the multi-core optical fiber , the normal vector ;

[0070] Step 43, restore the curve using the following formula: , , , is the curvature function. Finally, the restored curve shape is compared with the original curve to evaluate the reconstruction accuracy.

[0071] A person skilled in the art will easily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptive change of the present invention, which follows the general principles of the present invention and includes common knowledge or customary technical means in the technical field not disclosed by the present invention. The specification and examples are to be regarded as exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0072] It will be appreciated that the present invention is not limited to the precise construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof, which is governed solely by the appended claims.

Claims

1. A method for measuring optical fiber shape based on phase demodulation, comprising the following steps: Step 1: calibrating the frequency sweep range of a tuned laser; Step 2: calibrating the misalignment caused by strain; Step 3: differential phase denoising; Step 4: artificial neural network shape reconstruction; It is characterized by: The specific working process of step one is: Step 11: Obtain auxiliary interferometer data of reference signal and measurement signal I r , I m , perform Hilbert transform and find the frequency sweep range of the tunable laser for the reference signal and the measurement signal v r and v m ,Right now , , , is the length of the delay fiber, is the effective refractive index of the fiber core, c is the speed of light, H {} is the Hilbert transform, It is the delay of the optical fiber; Step 12: Find the frequency sweep range v r and v m The minimum value of and find the maximum value between these two values ​​is recorded as v min ; Find the sweep frequency range v r and v m The maximum value of and find the minimum value between these two values ​​is recorded as v max ,use v min and v max Construct a new sweep range v =( v min , v max ); Step 13, take the reference signal and the measurement signal main interferometer data, and cut the sweep frequency range from the original data to v The corresponding data acts as I R , I M ,use v The main interferometer data for the reference signal and the measured signal are I R , I M Perform cubic spline interpolation to obtain resampled data , ; Step 14: Resample the data , Perform Fourier transform, extract the phase of the reference signal and the measured signal and make a difference to get the differential phase Δ φ .

2. The optical fiber shape measurement method based on phase demodulation according to claim 1, characterized in that: The specific process of step 2 is: Step 21: The Δ obtained in step 1 φ Divided into N segments, and for each segment, Δ φ Unwind and get the differential phase ΔΦ after each unwinding N , take ΔΦ N The last data ΔΦ in N_end Calculate the cumulative length change Δ of this segment L ,Right now: ; in is the effective refractive index of the optical fiber core, λ0 is the initial wavelength, P e is the elastic-optical coefficient; Step 22, calculate The phase change caused , , get the resampled data of the measured signal To compensate for it, , i represents the imaginary unit, and Fourier transform is performed on it to obtain the calibrated measurement signal distance domain data I z ,Right now ; Step 23, Extract I z The phase of the reference signal is then subtracted to obtain the differential phase after the first calibration. ; Step 24, by looping steps 21 to 23, the cumulative length changes of all segments are compensated to obtain the final compensated distance domain data. I z , and recalculate the calibration differential phase .

3. The optical fiber shape measurement method based on phase demodulation according to claim 2, characterized in that: The specific process of step three is: Step 31, the differential phase Δ after calibration in step 24 φ Unwind to get ΔΦ, and take the derivative of ΔΦ to get d ΔΦ, find d All points in ΔΦ that are greater than the threshold π / 2 z i , find the point z i The mean of the first 10 and last 10 ΔΦ data M 1 and M 2; Step 32, if M 2− M 1 is greater than or equal to the set threshold π / 2, then all ΔΦ data after this point will be subtracted M 2− M 1; Step 33: If the condition of step 32 is not met, the ΔΦ value at the point is ΔΦ zi Replace with ; Step 34, by looping through all large phase jump positions and compensating them, SG filtering and smoothing processing is performed on ΔΦ after the large phase jump is removed, and data ΔΦ1, ΔΦ2, and ΔΦ3 after the data collection of the three fiber cores are processed are obtained respectively.

4. The optical fiber shape measurement method based on phase demodulation according to claim 3, characterized in that: The specific process of step 4 is as follows: The model of the artificial neural network consists of an input layer, a hidden layer, and an output layer. The input layer data is obtained by taking the derivative of ΔΦ1, ΔΦ2, and ΔΦ3. d ΔΦ1, d ΔΦ2, d ΔΦ3, the output layer is the corresponding multi-core optical fiber bending curvature κ; The Frenet-Serret equation is used to restore the shape of the curve. The specific restoration method is as follows: Step 41, assuming that the natural parameter equation of the multi-core optical fiber is: r ( s )= x ( s ) i+y ( s ) j , s is the arc length of the curve, i , j for x , y axis unit vector; Step 42: unit tangent vector of each point of the multi-core optical fiber , the normal vector ; Step 43, restore the curve using the following formula: , , , is the curvature function.

Citation Information

Patent Citations

  • Swept-source-based correction method of phase errors in optical coherence tomography

    CN105342558A

  • Single-frame interference graph demodulation method based on convolutional neural network

    CN113674370A

  • Distributed bidirectional polarization measurement device based on double-beat-frequency single auxiliary interferometer

    CN113804303A

  • Optical fiber distributed sound wave sensing system and method

    CN115265750A

  • High-precision dynamic interference measurement method and device based on deep learning

    CN117629104A