Seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction

Abstract

The invention discloses a seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction, and belongs to the technical field of precise instrument manufacturing and measurement. According to the method, the influence of the noise of the optical fiber sensing system on the shape reconstruction precision is inhibited by utilizing the characteristics of redundant sensing information and sensing optical fiber shape curve continuity of a seven-core optical fiber and adopting the idea of curvature vector prediction and clustering, so that the problem that salt and pepper noise cannot be inhibited in a traditional shape reconstruction method can be solved, and the shape reconstruction precision is improved. According to the method, on the premise that an existing multi-core optical fiber shape sensing system is not changed, the shape sensing precision of the seven-core optical fiber sensor can be remarkably improved only through data processing.

Description

Technical Field

[0001] The invention belongs to the technical field of precision instrument manufacturing and measurement, and relates to a seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction. Background Art

[0002] In recent years, curve and surface online testing technology has received increasing attention in the fields of medicine, aerospace, soft robotics, and intelligent structures. In order to achieve accurate morphological control and high-precision spatial positioning, it is necessary to accurately measure the three-dimensional shape and curve coordinates of the structure. Therefore, there is an urgent need for a sensing technology that can efficiently measure the shape of three-dimensional curves. Shape sensors based on multi-core optical fibers have the characteristics of small size and high precision, but they are easily affected by the measurement accuracy and noise of the optical fiber sensing system, and their shape measurement accuracy needs to be improved.

[0003] At present, the main methods for 3D curve reconstruction include the following:

[0004] 1. Jason P. Moore of NASA Research Center in the United States first used the Frenet frame to reconstruct three-dimensional curves in space. He took the rate of change of the bending direction as the torsion and combined it with the curvature as the input parameter of the Frenet frame. By solving differential equations, he obtained the three-dimensional tangent vector, normal vector and binormal vector of each small section of optical fiber, and then integrated the tangent vector to obtain the three-dimensional spatial coordinates of the optical fiber.

[0005] 2. Zeng Jie and others from Nanjing University of Aeronautics and Astronautics proposed to use the minimum rotation frame to reconstruct three-dimensional curves in space, and use curvature and curvature direction to describe the tangent vector, the first normal vector and the second normal vector, without the need to calculate the torsion. Compared with the Frenet frame, the curve reconstruction accuracy of the minimum rotation frame is higher, and there is no reconstruction error caused by zero curvature.

[0006] In summary, shape sensors based on the principle of multi-core optical fibers have the advantages of strong applicability, high accuracy, and long sensing distance, and have been widely used. However, there are still some problems:

[0007] 1. The accuracy and noise of the fiber optic sensing system directly affect the shape sensing accuracy. Conventional methods can only suppress the influence of Gaussian noise, but have limited effect on salt and pepper noise.

[0008] 2. The Frenet framework cannot overcome the problem of instantaneous reversal of the principal normal vector caused by the zero curvature, which will result in reconstruction errors and completely wrong shape reconstruction after the reconstruction errors. Summary of the invention

[0009] The purpose of the present invention is to propose a high-precision, noise-resistant shape reconstruction method for a multi-core optical fiber-based shape sensor, which suppresses the influence of the noise of the optical fiber sensing system on the shape reconstruction accuracy. The present invention proposes a seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction, which utilizes the characteristics of redundant sensing information of the seven-core optical fiber and the continuity of the sensing optical fiber shape curve, adopts the idea of ​​curvature vector prediction and clustering, suppresses the influence of the noise of the optical fiber sensing system on the shape reconstruction accuracy, and can solve the problem that the traditional shape reconstruction method cannot suppress salt and pepper noise. This method can significantly improve the shape sensing accuracy of the seven-core optical fiber sensor only through data processing without changing the existing multi-core optical fiber shape sensing system.

[0010] The working principle of the present invention is to utilize the characteristics of redundant sensing information and continuity of the shape curve of the seven-core optical fiber, use the curvature vector of the sensing unit before the current optical fiber sensing unit to predict the curvature vector of the current optical fiber sensing unit, and perform clustering data analysis based on the prediction error data to reduce the influence of salt and pepper noise, and then sort according to the size of the prediction error, calculate the average value of the curvature vector after clustering, reduce the influence of Gaussian noise, and use the curvature vector after clustering and averaging and the minimum rotation framework to reconstruct the shape of the seven-core optical fiber to achieve high-precision shape sensing.

[0011] The technical solution of the present invention is: a seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction, the seven-core optical fiber shape sensor contains seven cores inside, namely core one, core two, core three, core four, core five, core six and core seven, wherein core one, core two, core three, core four, core five and core six are evenly distributed in a 60° ring shape on the cross section of the seven-core optical fiber shape sensor and the core distances are consistent. If the angle direction of core one is 0°, the angle directions of core two, core three, core four and core five are 60°, 120°, 120° and 120° respectively. The seven-core optical fiber shape sensor is divided into three parts according to the length area, namely, the non-shape sensing part, the shape sensing starting end and the shape sensing part. When implementing the sensing, the seven-core optical fiber shape sensor near the shape sensing starting end needs to be a straight line, the Z-axis of the initial coordinate frame of shape reconstruction coincides with the straight line and the positive direction is from the shape sensing starting end to the shape sensing part, the X-axis of the initial coordinate frame of shape reconstruction is from the core seven to the core one, and the Y-axis of the initial coordinate frame of shape reconstruction is The directions are respectively perpendicular to the X-axis and the Z-axis and satisfy the right-hand coordinate system, the initial minimum rotation frame coincides with the initial coordinate frame of shape reconstruction, and the T vector of the initial minimum rotation frame is a unit vector in the Z-axis direction, the N1 vector of the initial minimum rotation frame is a unit vector in the X-axis direction, and the N2 vector of the initial minimum rotation frame is a unit vector in the Y-axis direction. The non-shape sensing part of the seven-core optical fiber shape sensor is connected to the seven-channel optical frequency domain distributed optical fiber sensing device, and the seven-channel optical frequency domain distributed optical fiber sensing device collects the distributed strain data of the seven cores inside the shape sensing part of the seven-core optical fiber shape sensor. , where distributed strain data one corresponds to core one, distributed strain data two corresponds to core two, distributed strain data three corresponds to core three, distributed strain data four corresponds to core four, distributed strain data five corresponds to core five, distributed strain data six corresponds to core six, distributed strain data seven corresponds to core seven, and the spatial resolution of the sensing point of the seven-channel optical frequency domain distributed optical fiber sensing device is the length interval of each sensing point on the seven-core optical fiber shape sensor. Then, data processing is performed according to the following steps to reconstruct the sensing shape of the shape sensing part of the seven-core optical fiber shape sensor, thereby realizing the shape sensing of the seven-core optical fiber shape sensor:

[0012] Step 1: Group the distributed strain data 1, distributed strain data 2, distributed strain data 3, distributed strain data 4, distributed strain data 5, distributed strain data 6, and distributed strain data 7, for a total of nine distributed strain combinations, which are:

[0013] Distributed strain combination 1: strain data 1, strain data 3, strain data 5,

[0014] Distributed strain combination 2: strain data 2, strain data 4, strain data 6,

[0015] Distributed strain combination three: strain data one, strain data two, strain data three,

[0016] Distributed strain combination 4: strain data 2, strain data 3, strain data 4,

[0017] Distributed strain combination five: strain data three, strain data four, strain data five,

[0018] Distributed strain combination six: strain data four, strain data five, strain data six,

[0019] Distributed strain combination seven: strain data one, strain data five, strain data six,

[0020] Distributed strain combination eight: strain data one, strain data two, strain data six,

[0021] Distributed strain combination nine: strain data one, strain data two, strain data three, strain data four, strain data five, strain data six,

[0022] Calculate the apparent curvature vector K of each sensing point of the nine sets of distributed strain combinations respectively ij , where the apparent curvature vector K ij =[K1 ij , K2 ij ], i = 1, 2, 3 ... N, N is the number of sensing points of distributed strain data, j = 1, 2, 3 ... 9, corresponding to nine distributed strain combinations, distributed strain combinations 1 to 8 are calculated according to the following formula:

[0023]

[0024] Among them, e iα 、e iβ 、e iΓ is the strain data of the three fiber cores at the i-th sensing point in the corresponding combination, α, β, γ are the distribution angles of the corresponding three fiber cores on the cross section of the seven-core optical fiber shape sensor, r is the fiber core distance, e i7 is the strain data at the i-th sensing point,

[0025] The distributed strain combination nine is calculated according to the following formula:

[0026]

[0027] Among them, e i1 、e i2 、e i3 …e i7 is strain data 1 to strain data 7, r is the fiber core distance;

[0028] Step 2: The apparent curvature vector K of the nine distributed strain combinations ij The apparent curvature vector prediction value KF1 of the first sensing point is initialized, where the apparent curvature vector filter value KF1 of the first sensing point is [0,0], and the apparent curvature vector filter value KF2 of the second sensing point is [0,0];

[0029] Step 3: Calculate the predicted value KF of the apparent curvature vector of the next sensing point in turn i , KF i =2×KF (i-1) -KF (i-2) , where i = 3, 4…N, N is the number of sensing points for distributed strain data;

[0030] Step 4: At the i-th sensing point, where i = 3, 4, ... N, the nine sets of apparent curvature vectors K obtained in step 1 are ij The apparent curvature vector predicted value KF in step 3 i Calculate the magnitude of the apparent curvature vector error ΔK ij , calculated as ΔK ij =|KP i -K ij |, where i = 3, 4…N, N is the number of sensing points for distributed strain data, and j = 1, 2, 3…9;

[0031] Step 5: Modulus ΔK of the apparent curvature vector error in step 4 ij Sorting from large to small, the new apparent curvature vector sequence is K is , where s = 1, 2…9, and then clustering is performed, in the apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the eighth to ninth digits is , s = 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, N is the number of sensing points for distributed strain data;

[0032] Step 6: Repeat steps 3 to 5 until the apparent curvature vector filter value KF of all sensing points is obtained. i =[KF1 i ,KF2 i ], where i = 1, 2…N, N is the number of sensing points for distributed strain data;

[0033] Step 7: The minimum rotation frame of each position of the seven-core optical fiber shape sensor (101) is iteratively reconstructed using the differential equation form of the minimum rotation frame. The differential equation form of the minimum rotation frame is: In each step of the iterative calculation, the apparent curvature vector filter value KF of the i-th sensor point is converted into i =[KF1 i , KF2 i ] as the coefficient matrix, using the minimum rotation frame [T i , N1 i , N2 i ] as the initial value of the differential equation, and then solve the minimum rotation frame [T i+1 , N1 i+1 , N2 i+1 ], in the first iteration, the initial minimum rotating frame is used as the initial value of the differential equation, where i = 1, 2…N, N is the number of sensing points of the distributed strain data;

[0034] Step 8: Use the T of the i-th sensor point i vector and reconstructed coordinates R(x i ,y i ,z i ) to solve the reconstructed coordinates of the i+1th sensing point: R(x i+1 ,y i+1 ,z i+1 )=T i Δs+R(x i ,y i ,z i ), where Δs is the spatial resolution of the sensing point (115), i = 1, 2...N, N is the number of sensing points for the distributed strain data, and in the first iteration, the T vector (206) of the initial minimum rotation frame is used as T1, R (x i ,y i ,z i ) is the reconstructed coordinate of the i-th sensor point. In the first iteration, R(x1,y1,z1)=[0,0,0];

[0035] Step 9: Repeat steps 7 to 8 until the reconstructed coordinates R(x i ,y i ,z i ), where i = 1, 2…N, and N is the number of sensing points for distributed strain data.

[0036] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the seventh to ninth digits is, s = 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0037] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the sixth to ninth digits is , s = 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0038] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fifth to ninth digits is , s = 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0039] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fourth to ninth digits is , s = 4, 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0040] The advantages of the present invention are:

[0041] 1. The method of the present invention has the characteristics of high precision, low cost, small size, simple structure and short measurement time.

[0042] 2. The method of the present invention has a significant effect on suppressing salt and pepper noise in optical fiber sensing systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a schematic diagram of a seven-core optical fiber distributed sensing shape sensing device based on clustering and prediction;

[0044] In the figure: 101. seven-core optical fiber shape sensor, 102. core one, 103. core two, 104. core three, 105. core four, 106. core five, 107. core six, 108. core seven, 109. core distance, 110. non-shape sensing part, 111. shape sensing starting end, 112. shape sensing part, 113. straight line, 114. seven-channel optical frequency domain distributed optical fiber sensing device.

[0045] Figure 2 It is a schematic diagram of the initial coordinate frame and the initial minimum rotation frame of shape reconstruction;

[0046] In the figure: 201. Initial coordinate frame for shape reconstruction, 202. Z axis, 203. X axis, 204. Y axis, 205. Initial minimum rotation frame, 206. T vector, 207. N1 vector, 208. N2 vector.

[0047] Figure 3 It is a schematic diagram of the distributed strain data of the seven fiber cores inside the seven-core optical fiber shape sensor;

[0048] In the figure: 301. distributed strain data one, 302. distributed strain data two, 303. distributed strain data three, 304. distributed strain data four, 305. distributed strain data five, 306. distributed strain data six, 307. distributed strain data seven. DETAILED DESCRIPTION

[0049] The specific implementation of the present invention is further described in detail below with reference to the accompanying drawings:

[0050] A seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction, the seven-core optical fiber shape sensor 101 contains seven cores, namely core 102, core 103, core 104, core 105, core 5 106, core 6 107 and core 7 108, wherein core 102, core 103, core 104, core 105, core 5 106 and core 6 107 are evenly distributed in a 60° ring on the cross section of the seven-core optical fiber shape sensor 101 and the core distance 109 is consistent. If the angle direction of core 102 is 0°, the angle directions of core 2 103, core 3 104, core 4 105 and core 5 106 are 60°, 120°, 18 ... 0°, 240° and 300°, the core seven 108 is at the cross-sectional center of the seven-core optical fiber shape sensor 101, and the seven-core optical fiber shape sensor 101 is divided into three components according to the length region, namely, a non-shape sensing portion 110, a shape sensing starting end 111 and a shape sensing portion 112. When implementing sensing, the seven-core optical fiber shape sensor 101 near the shape sensing starting end 111 needs to be a straight line 113, and the Z axis 202 of the initial coordinate frame 201 of shape reconstruction coincides with the straight line 113 and the positive direction is from the shape sensing starting end 111 to the shape sensing portion 112, and the X axis 203 of the initial coordinate frame 201 of shape reconstruction is from the core seven 108 to the direction of the core one 102, and the initial coordinate frame 201 of shape reconstruction The directions of the Y-axis 204 of the seven-core optical fiber shape sensor 101 are respectively perpendicular to the X-axis 203 and the Z-axis 202 and satisfy the right-hand coordinate system. The initial minimum rotation frame 205 coincides with the initial coordinate frame 201 of the shape reconstruction and the T vector 206 of the initial minimum rotation frame 205 is a unit vector in the direction of the Z-axis 202. The N1 vector 207 of the initial minimum rotation frame 205 is a unit vector in the direction of the X-axis 203. The N2 vector 208 of the initial minimum rotation frame 205 is a unit vector in the direction of the Y-axis 204. The non-shape sensing part 110 of the seven-core optical fiber shape sensor 101 is connected to the seven-channel optical frequency domain distributed optical fiber sensing device 114. The seven-channel optical frequency domain distributed optical fiber sensing device 114 collects the internal shape sensing part 112 of the seven-core optical fiber shape sensor 101. The distributed strain data of the seven cores of the seven optical fiber shape sensor 101, wherein the distributed strain data one 301 corresponds to the core one 102, the distributed strain data two 302 corresponds to the core two 103, the distributed strain data three 303 corresponds to the core three 104, the distributed strain data four 304 corresponds to the core four 105, the distributed strain data five 305 corresponds to the core five 106, the distributed strain data six 306 corresponds to the core six 107, the distributed strain data seven 307 corresponds to the core seven 108, the sensing point spatial resolution 115 of the seven-channel optical frequency domain distributed optical fiber sensing device 114 is the length interval of each sensing point on the seven-core optical fiber shape sensor 101, and then the data processing is performed according to the following steps to reconstruct the sensing shape of the shape sensing part 112 of the seven-core optical fiber shape sensor 101,Realize shape sensing of seven-core optical fiber shape sensor 101:

[0051] Step 1: group the distributed strain data 1 301, the distributed strain data 2 302, the distributed strain data 303, the distributed strain data 4 304, the distributed strain data 5 305, the distributed strain data 6 306, and the distributed strain data 7 307, and there are a total of nine distributed strain combinations, which are:

[0052] Distributed strain combination 1: strain data 1 301, strain data 3 303, strain data 5 305,

[0053] Distributed strain combination 2: strain data 2 302, strain data 4 304, strain data 6 306,

[0054] Distributed strain combination three: strain data one 301, strain data two 302, strain data three 303,

[0055] Distributed strain combination four: strain data two 302, strain data three 303, strain data four 304,

[0056] Distributed strain combination five: strain data three 303, strain data four 304, strain data five 305,

[0057] Distributed strain combination six: strain data four 304, strain data five 305, strain data six 306,

[0058] Distributed strain combination seven: strain data one 301, strain data five 305, strain data six 306,

[0059] Distributed strain combination eight: strain data one 301, strain data two 302, strain data six 306,

[0060] Distributed strain combination nine: strain data one 301, strain data two 302, strain data three 303, strain data four 304, strain data five 305, strain data six 306,

[0061] Calculate the apparent curvature vector K of each sensing point of the nine sets of distributed strain combinations respectively ij , where the apparent curvature vector K ij =[K1 ij , K2 ij ], i = 1, 2, 3 ... N, N is the number of sensing points of distributed strain data, j = 1, 2, 3 ... 9, corresponding to nine distributed strain combinations, distributed strain combinations 1 to 8 are calculated according to the following formula:

[0062]

[0063] Among them, e iα 、eiβ 、e iΓ is the strain data of the three fiber cores at the i-th sensing point in the corresponding combination, α, β, γ are the distribution angles of the corresponding three fiber cores on the cross section of the seven-core optical fiber shape sensor 101, r is the fiber core distance 109, e i7 is the strain data 7307 at the i-th sensing point,

[0064] The distributed strain combination nine is calculated according to the following formula:

[0065]

[0066] Among them, e i1 、e i2 、e i3 …e i7 is strain data one 301 to strain data seven 307, r is the fiber core distance 109;

[0067] Step 2: The apparent curvature vector K of the nine distributed strain combinations ij The apparent curvature vector prediction value KF i Initialization is performed, where the apparent curvature vector filter value KF1 of the first sensor point is [0,0], and the apparent curvature vector filter value KF2 of the second sensor point is [0,0];

[0068] Step 3: Calculate the predicted value KF of the apparent curvature vector of the next sensing point in turn i , KF i =2×KF (i-1) -KF (i-2) , where i = 3, 4…N, N is the number of sensing points for distributed strain data;

[0069] Step 4: At the i-th sensing point, where i = 3, 4, ... N, the nine sets of apparent curvature vectors K obtained in step 1 are ij The apparent curvature vector predicted value KF in step 3 i Calculate the magnitude of the apparent curvature vector error ΔK ij , calculated as ΔK ij =|KP i -K ij |, where i = 3, 4…N, N is the number of sensing points for distributed strain data, and j = 1, 2, 3…9;

[0070] Step 5: Modulus ΔK of the apparent curvature vector error in step 4 ij Sorting from large to small, the new apparent curvature vector sequence is K is , where s = 1, 2…9, and then clustering is performed, in the apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the eighth to ninth digitsis , s = 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, N is the number of sensing points for distributed strain data;

[0071] Step 6: Repeat steps 3 to 5 until the apparent curvature vector filter value KF of all sensing points is obtained. i =[KF1 i ,KF2 i ], where i = 1, 2…N, N is the number of sensing points for distributed strain data;

[0072] Step 7: The minimum rotation frame of each position of the seven-core optical fiber shape sensor (101) is iteratively reconstructed using the differential equation form of the minimum rotation frame. The differential equation form of the minimum rotation frame is: In each step of the iterative calculation, the apparent curvature vector filter value KF of the i-th sensor point is converted into i =[KF1 i , KF2 i ] as the coefficient matrix, using the minimum rotation frame [T i , N1 i , N2 i ] as the initial value of the differential equation, and then solve the minimum rotation frame [T i+1 , N1 i+1 , N2 i+1 ], in the first iteration, the initial minimum rotation frame (205) is used as the initial value of the differential equation, where i = 1, 2 ... N, N is the number of sensing points of the distributed strain data;

[0073] Step 8: Use the T of the i-th sensor point i vector and reconstructed coordinates R(x i ,y i ,z i ) to solve the reconstructed coordinates of the i+1th sensing point: R(x i+1 ,y i+1 ,z i+1 )=T i Δs+R(x i ,y i ,z i), where Δs is the spatial resolution of the sensing point (115), i=1, 2…N, N is the number of sensing points for the distributed strain data, and in the first iteration, the T vector (206) of the initial minimum rotation frame (205) is used as T1, R(x i ,y i ,z i ) is the reconstructed coordinate of the i-th sensor point. In the first iteration, R(x1,y1,z1)=[0,0,0];

[0074] Step 9: Repeat steps 7 to 8 until the reconstructed coordinates R(x i ,y i ,z i ), where i = 1, 2…N, and N is the number of sensing points for distributed strain data.

[0075] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the seventh to ninth digits is , s = 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0076] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the sixth to ninth digits is , s = 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0077] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fifth to ninth digits is , s = 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0078] The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fourth to ninth digits is , s = 4, 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

[0079] The working process of the present invention is as follows:

[0080] The seven-channel optical frequency domain distributed optical fiber sensing device 114 collects distributed strain data of the seven optical cores inside the shape sensing part 112 of the seven-core optical fiber shape sensor 101, wherein the distributed strain data one 301 corresponds to the core one 102, the distributed strain data two 302 corresponds to the core two 103, the distributed strain data three 303 corresponds to the core three 104, the distributed strain data four 304 corresponds to the core four 105, the distributed strain data five 305 corresponds to the core five 106, the distributed strain data six 306 corresponds to the core six 107, and the distributed strain data seven 307 corresponds to the core seven 108. The sensing point spatial resolution 115 of the seven-channel optical frequency domain distributed optical fiber sensing device 114 is the length interval of each sensing point on the seven-core optical fiber shape sensor 101. Then, data processing is performed according to the following steps to reconstruct the sensing shape of the shape sensing part 112 of the seven-core optical fiber shape sensor 101, thereby realizing shape sensing of the seven-core optical fiber shape sensor 101:

[0081] Step 1: Group the distributed strain data 1 301, the distributed strain data 2 302, the distributed strain data 3 303, the distributed strain data 4 304, the distributed strain data 5 305, the distributed strain data 6 306, and the distributed strain data 7 307, a total of nine distributed strain combinations, and calculate the apparent curvature vector K of each sensing point of the nine distributed strain combinations respectively. ij , where the apparent curvature vector K ij =[K1 ij , K2 ij], i = 1, 2, 3 ... N, N is the number of sensing points of distributed strain data, j = 1, 2, 3 ... 9, corresponding to nine distributed strain combinations, distributed strain combinations 1 to 8 are calculated according to the following formula:

[0082]

[0083] The distributed strain combination nine is calculated according to the following formula:

[0084]

[0085] Step 2: The apparent curvature vector K of the nine distributed strain combinations ij The apparent curvature vector prediction value KF i Initialization is performed, where the apparent curvature vector filter value KF1 of the first sensor point is [0,0], and the apparent curvature vector filter value KF2 of the second sensor point is [0,0];

[0086] Step 3: Calculate the predicted value KF of the apparent curvature vector of the next sensing point in turn i , KF i =2×KF (i-1) -KF (i-2) ;

[0087] Step 4: The nine sets of apparent curvature vectors K obtained in step 1 ij The apparent curvature vector predicted value KF in step 3 i Calculate the magnitude of the apparent curvature vector error ΔK ij , calculated as ΔK ij =|KP i -K ij |;

[0088] Step 5: Modulus ΔK of the apparent curvature vector error in step 4 ij Sorting from large to small, the new apparent curvature vector sequence is K is , and then clustering, in the apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the eighth to ninth digits is , s = 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ];

[0089] Step 6: Repeat steps 3 to 5 until the apparent curvature vector filter value KF of all sensing points is obtained. i =[KF1i ,KF2 i ];

[0090] Step 7: Iteratively reconstruct the minimum rotation frame of each position of the seven-core optical fiber shape sensor 101 using the differential equation form of the minimum rotation frame;

[0091] Step 8: Use the T of the i-th sensor point i vector and reconstructed coordinates R(x i ,y i ,z i ) to solve the reconstructed coordinates of the i+1th sensing point: R(x i+1 ,y i+1 ,z i+1 )=T i Δs+R(x i ,y i ,z i ), where Δs is the spatial resolution of the sensing point (115), i=1, 2…N, N is the number of sensing points for the distributed strain data, and in the first iteration, the T vector (206) of the initial minimum rotation frame (205) is used as T1, R(x i ,y i ,z i ) is the reconstructed coordinate of the i-th sensor point. In the first iteration, R(x1,y1,z1)=[0,0,0];

[0092] Step 9: Repeat steps 7 to 8 until the reconstructed coordinates R(x i ,y i ,z i ), where i = 1, 2…N, and N is the number of sensing points for distributed strain data.

[0093] The technical innovation and beneficial technical effect of the seven-core optical fiber distributed sensing shape sensing reconstruction method based on clustering and prediction are: it can make full use of the redundant sensing data of multi-core optical fibers, and suppress the influence of noise and error of the optical fiber sensing system on the curvature vector, while conventional algorithms, such as conditional adjustment method, are only effective for suppressing Gaussian noise, while the method of the present invention can not only suppress Gaussian noise but also suppress salt and pepper noise, thereby improving the accuracy of three-dimensional curve reconstruction. At the same time, based on the continuous curvature characteristics of the curve, the method of the present invention predicts the curvature of the current sensing unit using the curvature of the adjacent sensing unit, and has the advantages of high accuracy and smooth reconstruction of curvature. In addition, without changing the existing multi-core optical fiber shape sensing system, the method of the present invention can significantly improve the shape sensing accuracy of the seven-core optical fiber sensor only through data processing, and has the advantages of wide application range and low application cost.

Claims

1. A seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction, wherein the seven-core optical fiber shape sensor (101) contains seven fiber cores, namely, fiber core one (102), fiber core two (103), fiber core three (104), fiber core four (105), fiber core five (106), fiber core six (107) and fiber core seven (108), wherein the fiber core one (102), fiber core two (103), fiber core three (104), fiber core four (105), fiber core five (106) and fiber core six (107) are evenly distributed in a 60° circular shape on the cross section of the seven-core optical fiber shape sensor (101) and the fiber core distance (109) is consistent. If the angle direction of the fiber core one (102) is 0°, the fiber core two (103), fiber core four (105), fiber core five (106) and fiber core six (107) are uniformly distributed in a 60° circular shape on the cross section of the seven-core optical fiber shape sensor (101) and the fiber core distance (109) is consistent. The angle directions of core three (104), core four (105), and core five (106) are 60°, 120°, 180°, 240°, and 300°, respectively. Core seven (108) is located at the cross-sectional center of the seven-core optical fiber shape sensor (101). The seven-core optical fiber shape sensor (101) is divided into three components according to the length region, namely, a non-shape sensing part (110), a shape sensing starting end (111), and a shape sensing part (112). When sensing is implemented, the seven-core optical fiber shape sensor (101) near the shape sensing starting end (111) needs to be a straight line (113), and the Z axis (202) of the initial coordinate frame (201) of shape reconstruction coincides with the straight line (113), and the positive direction is the shape sensing starting end. (111) points to the shape sensing part (112), the X-axis (203) of the initial coordinate frame (201) for shape reconstruction is the direction of the fiber core seven (108) pointing to the fiber core one (102), the Y-axis (204) of the initial coordinate frame (201) for shape reconstruction is perpendicular to the X-axis (203) and the Z-axis (202) respectively and satisfies the right-hand coordinate system, the initial minimum rotation frame (205) coincides with the initial coordinate frame (201) for shape reconstruction and the T vector (206) of the initial minimum rotation frame (205) is the unit vector in the direction of the Z-axis (202), the N1 vector (207) of the initial minimum rotation frame (205) is the unit vector in the direction of the X-axis (203), and the N2 vector (204) of the initial minimum rotation frame (205) is the unit vector in the direction of the X-axis (203). 08) is a unit vector in the direction of the Y axis (204), the non-shape sensing part (110) of the seven-core optical fiber shape sensor (101) is connected to the seven-channel optical frequency domain distributed optical fiber sensing device (114), the seven-channel optical frequency domain distributed optical fiber sensing device (114) collects distributed strain data of seven fiber cores inside the shape sensing part (112) of the seven-core optical fiber shape sensor (101), wherein the distributed strain data one (301) corresponds to the fiber core one (102), the distributed strain data two (302) corresponds to the fiber core two (103), the distributed strain data three (303) corresponds to the fiber core three (104), the distributed strain data four (304) corresponds to the fiber core four (105), and the distributed strain data five (305) corresponds to the fiber core five (106),The distributed strain data six (306) corresponds to the fiber core six (107), the distributed strain data seven (307) corresponds to the fiber core seven (108), the sensing point spatial resolution (115) of the seven-channel optical frequency domain distributed optical fiber sensing device (114) is the length interval of each sensing point on the seven-core optical fiber shape sensor (101), and then the data is processed according to the following steps to reconstruct the sensing shape of the shape sensing part (112) of the seven-core optical fiber shape sensor (101), thereby realizing the shape sensing of the seven-core optical fiber shape sensor (101):, Step 1: group the distributed strain data one (301), distributed strain data two (302), distributed strain data three (303), distributed strain data four (304), distributed strain data five (305), distributed strain data six (306), and distributed strain data seven (307), and there are a total of nine distributed strain combinations, namely: distributed strain combination one: strain data one (301), strain data three (303), strain data five (305), distributed strain combination two: strain data two (302), strain data four (304), strain data six (306), distributed strain combination three: strain data one (301), strain data two (302), strain data three (303), distributed strain combination four: strain data two (302), strain data six Strain data three (303), strain data four (304), distributed strain combination five: strain data three (303), strain data four (304), strain data five (305), distributed strain combination six: strain data four (304), strain data five (305), strain data six (306), distributed strain combination seven: strain data one (301), strain data five (305), strain data six (306), distributed strain combination eight: strain data one (301), strain data two (302), strain data six (306), distributed strain combination nine: strain data one (301), strain data two (302), strain data three (303), strain data four (304), strain data five (305), strain data six (306), Calculate the apparent curvature vector K of each sensing point of the nine sets of distributed strain combinations respectively ij , where the apparent curvature vector K ij =[K1 ij , K2 ij ], i = 1, 2, 3 ... N, N is the number of sensing points of distributed strain data, j = 1, 2, 3 ... 9, corresponding to nine distributed strain combinations, distributed strain combinations 1 to 8 are calculated according to the following formula: in, e iα 、e iβ 、e iΓ are the strain data of the three fiber cores at the i-th sensing point position in the corresponding combination, α, β, γ are the distribution angles of the corresponding three fiber cores on the cross section of the seven-core optical fiber shape sensor (101), r is the fiber core distance (109), e i7 is the strain data (307) at the i-th sensing point, The distributed strain combination nine is calculated according to the following formula: Among them, e i1 、e i2 、e i3 …e i7 is strain data one (301) to strain data seven (307), r is the fiber core distance (109); Step 2: The apparent curvature vector K of the nine distributed strain combinations ij The apparent curvature vector prediction value KF i Initialization is performed, where the apparent curvature vector filter value KF1 of the first sensor point is [0,0], and the apparent curvature vector filter value KF2 of the second sensor point is [0,0]; Step 3: Calculate the predicted value KF of the apparent curvature vector of the next sensing point in turn i , KF i =2×KF (i-1) -KF (i-2) , where i = 3, 4…N, N is the number of sensing points for distributed strain data; Step 4: At the i-th sensing point, where i = 3, 4, ... N, the nine sets of apparent curvature vectors K obtained in step 1 are ij The apparent curvature vector predicted value KF in step 3 i Calculate the magnitude of the apparent curvature vector error ΔK ij , calculated as ΔK ij =|KF i -K ij |, where i = 3, 4…N, N is the number of sensing points for distributed strain data, and j = 1, 2, 3…9; Step 5: Modulus ΔK of the apparent curvature vector error in step 4 ij Sorting from large to small, the new apparent curvature vector sequence is K is , where s = 1, 2…9, and then clustering is performed, in the apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the eighth to ninth digits is , s = 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector K i As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, N is the number of sensing points for distributed strain data; Step 6: Repeat steps 3 to 5 until the apparent curvature vector filter value KF of all sensing points is obtained. i =[KF1 i , KF2 i ], where i = 1, 2…N, N is the number of sensing points for distributed strain data; Step 7: The minimum rotation frame of each position of the seven-core optical fiber shape sensor (101) is iteratively reconstructed using the differential equation form of the minimum rotation frame. The differential equation form of the minimum rotation frame is: In each step of the iterative calculation, the apparent curvature vector filter value KF of the i-th sensor point is converted into i =[KF1 i , KF2 i ] as the coefficient matrix, using the minimum rotation frame [T i , N1 i , N2 i ] as the initial value of the differential equation, and then solve the minimum rotation frame [T i+1 , N1 i+1 , N2 i+1 ], in the first iteration, the initial minimum rotation frame (205) is used as the initial value of the differential equation, where i = 1, 2 ... N, N is the number of sensing points of the distributed strain data; Step 8: Use the T of the i-th sensor point i vector and reconstructed coordinates R(x i ,y i ,z i ) Solve the reconstructed coordinates of the i+1th sensing point: R(x i+1 ,y i+1 ,z i+1 )=T i Δs+R(x i ,y i ,z i ) Wherein, Δs is the spatial resolution of the sensing point (115), i=1, 2…N, N is the number of sensing points of the distributed strain data, and in the first iteration, the T vector (206) of the initial minimum rotation frame (205) is used as T1, R(x i ,y i ,z i ) is the reconstructed coordinate of the i-th sensor point. In the first iteration R(x1,y1,z1)=[0,0,0]; Step 9: Repeat steps 7 to 8 until the reconstructed coordinates R(x i ,y i ,z i ), where i = 1, 2…N, and N is the number of sensing points for distributed strain data.

2. The seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction according to claim 1 is characterized in that: The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the seventh to ninth digits is , s = 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector K i As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

3. The seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction according to claim 1 is characterized in that: The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the sixth to ninth digits is , s = 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

4. The seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction according to claim 1 is characterized in that: The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fifth to ninth digits is , s = 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

5. The seven-core optical fiber distributed sensing shape sensor reconstruction method based on clustering and prediction according to claim 1 is characterized in that: The data processing process in step 5 can also be: in the new apparent curvature vector sequence K is Select the apparent curvature vector K corresponding to the fourth to ninth digits is , s = 4, 5, 6, 7, 8, 9, and calculate the average value of the apparent curvature vector after clustering The average value of the clustered apparent curvature vector As the apparent curvature vector filter value KF at the i-th sensor point position i =[KF1 i , KF2 i ], where i = 3, 4…N, and N is the number of sensing points for distributed strain data.

Images