CN114861737A - Remote disturbance feature extraction method and system of distributed optical fiber sensing system - Google Patents

Remote disturbance feature extraction method and system of distributed optical fiber sensing system Download PDF

Info

Publication number
CN114861737A
CN114861737A CN202210637622.XA CN202210637622A CN114861737A CN 114861737 A CN114861737 A CN 114861737A CN 202210637622 A CN202210637622 A CN 202210637622A CN 114861737 A CN114861737 A CN 114861737A
Authority
CN
China
Prior art keywords
dictionary
disturbance
matrix
polymorphic
feature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210637622.XA
Other languages
Chinese (zh)
Inventor
刘玉申
许人东
陶宇
胥国祥
石明强
滕诣迪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changshu Institute of Technology
Jiangsu Hengtong Marine Cable Systems Co Ltd
Original Assignee
Changshu Institute of Technology
Jiangsu Hengtong Marine Cable Systems Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Changshu Institute of Technology, Jiangsu Hengtong Marine Cable Systems Co Ltd filed Critical Changshu Institute of Technology
Priority to CN202210637622.XA priority Critical patent/CN114861737A/en
Publication of CN114861737A publication Critical patent/CN114861737A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention belongs to the technical field of signal processing of a distributed optical fiber sensing system, and discloses a remote disturbance feature extraction method of the distributed optical fiber sensing system, which comprises the following steps: dividing distance dimensional intervals of a signal Rayleigh scattering spectrum in the optical fiber channel, establishing corresponding dictionary learning models for the intervals, and learning and observing an original dictionary to form a polymorphic equivalent dictionary; based on the polymorphic equivalent dictionary, establishing a joint sparse representation model under the cascade of the polymorphic equivalent dictionaries to form joint sparse representation of the multi-echo scattering spectrum; and constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary, and extracting the characteristics of the remote disturbance signal. The problem that far-end interference signals in a distributed optical fiber sensing system are submerged due to inherent attenuation of Rayleigh scattering spectrums can be solved. The accuracy of extracting the characteristics of the disturbance signals is greatly improved.

Description

Remote disturbance feature extraction method and system of distributed optical fiber sensing system
Technical Field
The invention belongs to the technical field of signal processing of a distributed optical fiber sensing system, and relates to a method and a system for extracting remote disturbance characteristics of the distributed optical fiber sensing system.
Background
The optical fiber sensor is widely applied to the fields of security, monitoring and surveying by virtue of the advantages of no electromagnetic interference, high flexibility and easiness in networking. The distributed optical fiber sensing system takes points on optical fibers as independent sensing units, utilizes the phase-sensitive (OTDR) technology, and has the advantages of remarkably improving the measurement range, the measurement sensitivity and the response capability compared with the traditional point type optical fiber sensing system. Similarly, the high sensitivity and fast response speed of the distributed optical fiber sensing system can cause the sensitivity of the system to noise and environmental disturbances. Meanwhile, in a long-distance transmission detection scene, due to the fading characteristic of the rayleigh scattering spectrum, a far-end interference signal is submerged by a near-end echo signal, so that accurate extraction of the interference signal is influenced.
Application number 201810359506X discloses a distributed optical fiber vibration sensing positioning method and device based on disturbance signal feature extraction, wherein an optical cable to be detected is divided into a plurality of sections, pulse peak signals generated by vibration are received in each section respectively, and then the number of pulse peaks is recorded, wherein the number of pulse peaks on the jth section of the ith frame of the optical cable is recorded as N (i, j); respectively counting the average value and the variance of the number of pulse peaks in each interval of the optical cable; comparing the average value and the variance of the number of the pulse peaks with preset parameters in each optical cable interval, determining whether the optical cable in the current interval has a disturbance signal according to the comparison result, and determining each interval of the optical cable to be detected having the disturbance signal by analogy; and acquiring an optical cable interval with disturbance signals, and respectively confirming the positions of disturbance points in the interval. This method is based on statistical findings and requires a certain amount of data. The actual operation is complicated, and when the data volume is small, an accurate result cannot be obtained.
Application number 2017103708538 discloses a method for rapidly positioning a phase-sensitive optical time domain reflection distributed optical fiber sensing system, which comprises the steps of constructing a Rayleigh scattering light digital signal matrix corresponding to a plurality of optical pulses, selecting a test window and a test column at intervals of a certain length on the signal matrix, obtaining the phase of each test column, roughly positioning a disturbance source interval according to the phase of the test column of the adjacent test window, and extracting an interval signal containing a disturbance source for accurately positioning disturbance. The method does not solve the technical problem that the far-end interference signal is submerged by the near-end echo signal due to the fading characteristic of the Rayleigh scattering spectrum, and the accurate disturbance characteristic cannot be extracted.
Disclosure of Invention
The invention aims to provide a method and a system for extracting remote disturbance characteristics of a distributed optical fiber sensing system. The accuracy of extracting the disturbance signal features is improved.
The technical solution for realizing the purpose of the invention is as follows:
a remote disturbance feature extraction method of a distributed optical fiber sensing system comprises the following steps:
s01: dividing distance dimensional intervals of a signal Rayleigh scattering spectrum in the optical fiber channel, establishing corresponding dictionary learning models for the intervals, and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
s02: based on the polymorphic equivalent dictionary, establishing a joint sparse representation model under the cascade of the polymorphic equivalent dictionaries to form joint sparse representation of the multi-echo scattering spectrum;
s03: and constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary, and extracting the characteristics of the remote disturbance signal.
In a preferred embodiment, the method for forming a polymorphic equivalent dictionary in step S01 includes:
s11: constructing an intrinsic feature dictionary for all detection distances:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ]；
wherein psi l Represents the unit response at the l-th detection range;
obtaining the ground state spectrum
Figure BDA0003681145050000031
Expressed as:
Figure BDA0003681145050000032
wherein eta is a ground state characteristic coefficient;
s12: ζ for nth disturbance event n The perturbation feature space of
Figure BDA0003681145050000033
The perturbation characteristic complementary space is
Figure BDA0003681145050000034
Based on
Figure BDA0003681145050000035
Obtaining corresponding characteristic coefficient vector
Figure BDA0003681145050000036
Comprises the following steps:
Figure BDA0003681145050000037
wherein eta is l And
Figure BDA0003681145050000038
respectively represent eta and
Figure BDA0003681145050000039
the l element of (1);
s13: ζ for disturbance event n And constructing a disturbance modal observation matrix:
Figure BDA00036811450500000310
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a weight coefficient vector;
s14: based on the disturbance modal observation matrix, learning observation is carried out on the intrinsic feature dictionary to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ；
and carrying out cascade combination on the modal dictionary and the intrinsic feature dictionary corresponding to all the disturbance events N to obtain a polymorphic equivalent dictionary:
Figure BDA0003681145050000041
in a preferred embodiment, the method for forming the joint sparse representation of the multiple echo scattering spectrum in step S02 includes:
a joint echo matrix composed of M echo light pulses is represented as:
Figure BDA0003681145050000042
wherein, P i Represents a sampled Rayleigh curve corresponding to the ith echo light pulse [. degree] T Representing a transpose of a vector and a matrix;
based on the polymorphic equivalent dictionary, obtaining a joint sparse representation model of the joint echo matrix as follows:
Figure BDA0003681145050000043
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
In a preferred technical solution, the method for extracting the far-end disturbance signal feature in step S03 includes:
s31: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ
wherein,
Figure BDA0003681145050000044
I (N+1)×(N+1) a unit matrix of dimensions (N +1) × (N +1) is represented,
Figure BDA0003681145050000051
representing a kronecker product, and theta is a reconstruction characteristic vector;
s32: initializing a reconstructed feature vector θ (0) 0, residual epsilon (0) R; index set
Figure BDA0003681145050000052
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
s33: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix gamma l Each gamma is l Multiplying the residual epsilon by 1,2 … L, and finding the index lambda corresponding to the maximum value of the product t I.e. by
Figure BDA0003681145050000053
S34: update index set Λ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual correlation degree in the found polymorphic equivalent dictionary, and reconstructing the original subset into gamma t ＝[γ t-1 ,Γ l ]；
S35: solving to obtain theta (t) ＝argmin||r-γ t θ (t) || 2 Updating the residual epsilon (t) ＝r-γ t θ (t) ，t＝t+1；
S36: determining the residual error epsilon (t) If less than xi, if yes, thenStopping iteration and outputting a reconstructed feature vector theta; otherwise, go to step S33 to continue execution;
s37: and reconstructing the reconstructed feature vector theta to obtain a joint feature matrix theta.
In a preferred technical solution, after the step S03, the method further includes setting zero to the first L column of the reconstructed joint feature matrix Θ, then determining the remaining non-zero columns, obtaining an index vector υ of the non-zero columns, and pointing the index vector υ to the disturbance event type.
The invention also discloses a remote disturbance feature extraction system of the distributed optical fiber sensing system, which comprises the following steps:
the polymorphic equivalent dictionary building module is used for dividing distance dimensional intervals of a Rayleigh scattering spectrum of a signal in the optical fiber channel, building a corresponding dictionary learning model aiming at each interval and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
the combined sparse representation module of the multi-echo scattering spectrum is used for establishing a combined sparse representation model under cascade connection of polymorphic equivalent dictionaries based on the polymorphic equivalent dictionaries to form combined sparse representation of the multi-echo scattering spectrum;
and the disturbance signal feature extraction module is used for constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary and extracting the far-end disturbance signal features.
In a preferred technical solution, the method for forming the polymorphic equivalent dictionary comprises:
s11: constructing an intrinsic feature dictionary for all detection distances:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ]；
wherein psi l Represents the unit response at the l-th detection range;
obtaining the ground state spectrum
Figure BDA0003681145050000061
Expressed as:
Figure BDA0003681145050000062
wherein eta is a ground state characteristic coefficient;
s12: ζ for nth disturbance event n The perturbation feature space of
Figure BDA0003681145050000063
The perturbation characteristic complementary space is
Figure BDA0003681145050000064
Based on
Figure BDA0003681145050000065
Obtaining corresponding characteristic coefficient vector
Figure BDA0003681145050000066
Comprises the following steps:
Figure BDA0003681145050000067
wherein eta is l And
Figure BDA0003681145050000068
respectively represent eta and
Figure BDA0003681145050000069
the l element of (1);
s13: ζ for disturbance event n And constructing a disturbance modal observation matrix:
Figure BDA00036811450500000610
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a weight coefficient vector;
s14: based on the disturbance modal observation matrix, learning observation is carried out on the intrinsic feature dictionary to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ；
and carrying out cascade combination on the modal dictionary and the intrinsic feature dictionary corresponding to all the disturbance events N to obtain a polymorphic equivalent dictionary:
Figure BDA0003681145050000071
in a preferred embodiment, the method for jointly sparsely representing multiple echo scattering spectra includes:
a joint echo matrix composed of M echo light pulses is represented as:
Figure BDA0003681145050000072
wherein, P i Represents a sampled Rayleigh curve corresponding to the ith echo light pulse [. degree] T Representing a transpose of a vector and a matrix;
based on the polymorphic equivalent dictionary, obtaining a joint sparse representation model of the joint echo matrix as follows:
Figure BDA0003681145050000073
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
In a preferred technical solution, the method for extracting the far-end disturbance signal feature includes:
s31: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ
wherein,
Figure BDA0003681145050000074
I (N+1)×(N+1) a unit matrix of dimensions (N +1) × (N +1) is represented,
Figure BDA0003681145050000075
expresses the kronecker productTheta is a reconstructed feature vector;
s32: initializing a reconstructed feature vector θ (0) 0, residual ε (0) R; index set
Figure BDA0003681145050000076
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
s33: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix gamma l Each gamma is l Multiplying the residual epsilon by 1,2 … L, and finding the index lambda corresponding to the maximum value of the product t I.e. by
Figure BDA0003681145050000081
S34: update index set Λ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual correlation degree in the found polymorphic equivalent dictionary, and reconstructing the original subset into gamma t ＝[γ t-1 ,Γ l ]；
S35: solved to obtain
Figure BDA0003681145050000082
Updating the residual ε (t) ＝r-γ t θ (t) ，t＝t+1；
S36: determining the residual error epsilon (t) If the value is smaller than xi, stopping iteration and outputting a reconstructed feature vector theta; otherwise, go to step S33 to continue execution;
s37: and reconstructing the reconstructed feature vector theta to obtain a joint feature matrix theta.
In the preferred technical scheme, the method further comprises a disturbance event type identification module, wherein the disturbance event type identification module is used for setting the first L columns of the reconstructed combined feature matrix theta to zero, then determining the remaining non-zero columns, obtaining a non-zero column index vector upsilon, and pointing the index vector upsilon to the disturbance event type.
Compared with the prior art, the invention has the following remarkable advantages:
the invention can solve the problem that the far-end interference signal in the distributed optical fiber sensing system is submerged due to the inherent attenuation of the Rayleigh scattering spectrum. The method comprises the steps of carrying out learning observation on an original dictionary to form a polymorphic equivalent dictionary aiming at different Rayleigh scattering spectrum attention intervals, then establishing a combined sparse representation model under cascade connection of the polymorphic equivalent dictionary based on the polymorphic equivalent dictionary to form combined sparse representation of the multi-echo scattering spectrum, and finally completing identification and extraction of a far-end disturbance signal through an optimization reconstruction algorithm. The accuracy of extracting the characteristics of the disturbance signals is greatly improved, and the types of disturbance events can be identified.
Drawings
FIG. 1 is a flow chart of a remote disturbance feature extraction method of a distributed optical fiber sensing system according to a preferred embodiment;
fig. 2 is a schematic block diagram of a remote disturbance feature extraction system of the distributed optical fiber sensing system according to the preferred embodiment.
Detailed Description
The principle of the invention is as follows: aiming at the defects of the existing universal fixture and the special fixture, the fixture which is simple and easy to clamp, does not damage precise parts, can completely detect all form and position errors and has universality when the sizes of the parts of the same type are measured and the sizes of the detected objects are deviated is designed, so that the disassembly and the unnecessary repeated positioning of the fixture can be reduced, the labor intensity of workers is reduced, and the measuring efficiency is improved.
Example 1:
as shown in fig. 1, a method for extracting far-end disturbance features of a distributed optical fiber sensing system includes the following steps:
s01: dividing distance dimensional intervals of a signal Rayleigh scattering spectrum in the optical fiber channel, establishing corresponding dictionary learning models for the intervals, and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
s02: based on the polymorphic equivalent dictionary, establishing a joint sparse representation model under the cascade of the polymorphic equivalent dictionaries to form joint sparse representation of the multi-echo scattering spectrum;
s03: and constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary, and extracting the characteristics of the remote disturbance signal.
It should be noted that: for an optical cable which is already wired, a rayleigh scattering spectrum of the optical cable has a fixed component combination under the condition of no disturbance, the fixed component combination is called a ground state spectrum of the rayleigh scattering spectrum of the optical cable, which is called the ground state spectrum for short, if the interval of distributed sensors in the optical cable is considered to be delta, and the total length of detection is L delta, in order to facilitate the conciseness of a formula in subsequent contents, in the invention, the delta is assumed to be 1, and the generality of the contents of the invention under the condition that the delta is taken as other values is not influenced by the assumption.
In a preferred embodiment, the method for forming the polymorphic equivalent dictionary in step S01 includes:
s11: constructing an intrinsic feature dictionary for all detection distances:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ]；
wherein psi l Represents the unit response at the l-th detection range;
obtaining the ground state spectrum
Figure BDA0003681145050000101
Expressed as:
Figure BDA0003681145050000102
wherein eta is a ground state characteristic coefficient;
s12: ζ for nth disturbance event n The perturbation feature space of
Figure BDA0003681145050000103
The perturbation characteristic complementary space is
Figure BDA0003681145050000104
Based on
Figure BDA0003681145050000105
Obtaining corresponding characteristic coefficient vector
Figure BDA0003681145050000106
Comprises the following steps:
Figure BDA0003681145050000107
wherein eta is l And
Figure BDA0003681145050000108
respectively represent eta and
Figure BDA0003681145050000109
the l element of (1);
s13: ζ for disturbance event n And constructing a disturbance modal observation matrix:
Figure BDA00036811450500001010
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a weight coefficient vector;
s14: based on the disturbance modal observation matrix, learning observation is carried out on the intrinsic feature dictionary to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ；
and carrying out cascade combination on the modal dictionary and the intrinsic feature dictionary corresponding to all the disturbance events N to obtain a polymorphic equivalent dictionary:
Figure BDA0003681145050000111
in a preferred embodiment, the method for forming the joint sparse representation of the multiple echo scattering spectrum in step S02 includes:
a joint echo matrix composed of M echo light pulses is represented as:
Figure BDA0003681145050000112
wherein, P i Indicating the ith echo light pulseCorresponding sampled Rayleigh curve [ ·] T Representing a transpose of a vector and a matrix;
based on the polymorphic equivalent dictionary, obtaining a joint sparse representation model of the joint echo matrix as follows:
Figure BDA0003681145050000113
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
In a preferred embodiment, the method for extracting the far-end disturbance signal feature in step S03 includes:
s31: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ
wherein,
Figure BDA0003681145050000114
I (N+1)×(N+1) a unit matrix of dimensions (N +1) × (N +1) is represented,
Figure BDA0003681145050000115
representing a kronecker product, and theta is a reconstruction characteristic vector;
s32: initializing a reconstructed feature vector θ (0) 0, residual epsilon (0) R; index set
Figure BDA0003681145050000116
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
s33: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix gamma l Each gamma is l Multiplying the residual epsilon by 1,2 … L, and finding the index lambda corresponding to the maximum value of the product t I.e. by
Figure BDA0003681145050000121
S34: updating index setsΛ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual correlation degree in the found polymorphic equivalent dictionary, and reconstructing the original subset into gamma t ＝[γ t-1 ,Γ l ]；
S35: solving to obtain theta (t) ＝argmin||r-γ t θ (t) || 2 Updating the residual epsilon (t) ＝r-γ t θ (t) ，t＝t+1；
S36: determining the residual error epsilon (t) If the value is smaller than xi, stopping iteration and outputting a reconstructed feature vector theta; otherwise, go to step S33 to continue execution;
s37: and reconstructing the reconstructed feature vector theta to obtain a joint feature matrix theta.
In another embodiment, after the step S03, the method further includes setting zero to the first L columns of the reconstructed joint feature matrix Θ, and then determining the remaining non-zero columns, obtaining a non-zero column index vector v, where the index vector v points to the disturbance event type. So that the type of disturbance event can be identified.
In another embodiment, as shown in fig. 2, the present invention further discloses a remote disturbance feature extraction system of a distributed optical fiber sensing system, including:
the polymorphic equivalent dictionary building module is used for dividing distance dimensional intervals of a Rayleigh scattering spectrum of a signal in the optical fiber channel, building a corresponding dictionary learning model aiming at each interval and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
the combined sparse representation module of the multi-echo scattering spectrum is used for establishing a combined sparse representation model under cascade connection of polymorphic equivalent dictionaries based on the polymorphic equivalent dictionaries to form combined sparse representation of the multi-echo scattering spectrum;
and the disturbance signal feature extraction module is used for constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary and extracting the far-end disturbance signal features.
Specifically, the work flow of the remote disturbance feature extraction system of the distributed optical fiber sensing system is as follows:
firstly, establishing an echo Rayleigh spectrum of a distributed optical fiber sensing systemThe invention relates to a line feature dictionary, wherein for an optical cable which is already wired, a Rayleigh scattering spectrum of the optical cable under the condition of not receiving disturbance has a fixed component combination, the fixed component combination is called as a ground state spectrum of the Rayleigh scattering spectrum of the optical cable, if the interval of distributed sensors in the optical cable is considered to be delta, and the total detection length is L delta, in order to facilitate the conciseness of a formula in subsequent contents, the delta is assumed to be 1 in the invention, and the generality of the content of the invention under other value-taking conditions of the delta is not influenced by the assumption. Let psi l Representing the unit response at the ith probe range, an intrinsic feature dictionary may be constructed for all probe ranges:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ] (1)
the ground state spectrum
Figure BDA0003681145050000131
Can be expressed as:
Figure BDA0003681145050000132
where η is the ground state eigen coefficient. Due to the ground state spectrum
Figure BDA0003681145050000133
Is a continuous spectrum, therefore the above formula is not an ideal sparse problem, so that the mode decomposition and other means can be adopted to obtain the ground state spectrum of the optical cable
Figure BDA0003681145050000134
And (5) decomposing to obtain a ground state characteristic coefficient vector eta.
The invention designs a multi-state dictionary learning observation method based on a ground state characteristic coefficient eta, which comprises the following steps:
let the number of the types of the disturbance events concerned by the distributed optical fiber sensing system be N in total, then the nth disturbance event is recorded as ζ n . Based on the previous disturbance event sample training, the disturbance feature distribution condition of various disturbance events, namely the disturbance feature space, can be described. ζ for disturbance event n Let us order
Figure BDA0003681145050000135
Representing the perturbation feature space, further defining the perturbation feature complementary space
Figure BDA0003681145050000136
Based on
Figure BDA0003681145050000137
Obtaining corresponding characteristic coefficient vector
Figure BDA0003681145050000138
The method specifically comprises the following steps:
Figure BDA0003681145050000141
wherein eta is l And
Figure BDA0003681145050000142
respectively represent eta and
Figure BDA0003681145050000143
the ith element of (1).
Further, ζ is aimed at disturbance event n And constructing a disturbance modal observation matrix:
Figure BDA0003681145050000144
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a vector of weight coefficients. Based on the disturbance modal observation matrix, the intrinsic feature dictionary can be learned and observed to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ (5)
and carrying out cascade combination on the modal dictionaries corresponding to all concerned disturbance events and the intrinsic feature dictionary to obtain a polymorphic equivalent dictionary:
Figure BDA0003681145050000145
secondly, the invention carries out joint optimization representation on the multi-echo light pulse of the distributed optical fiber sensing system on the basis. If P i And representing a sampling rayleigh curve corresponding to the ith echo light pulse, a joint echo matrix composed of M echo light pulses can be represented as:
Figure BDA0003681145050000146
wherein [ ·] T Representing the transpose of the vector and matrix.
Based on the polymorphic cascade dictionary shown in the formula (6), a joint sparse representation form of a joint echo matrix can be obtained:
Figure BDA0003681145050000147
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
Based on the joint sparse representation model, the process of obtaining the joint feature matrix from the joint echo matrix through reconstruction can be summarized as the following steps:
step 1: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ (9)
wherein,
Figure BDA0003681145050000151
I (N+1)×(N+1) a unit array having dimensions of (N +1) × (N +1),
Figure BDA0003681145050000152
representing the kronecker product.
Step 2: initializing a reconstructed feature vector θ (0) 0, residual epsilon (0) R; index set
Figure BDA0003681145050000153
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
and step 3: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix gamma l Each gamma is l Multiplying the residual epsilon by 1,2 … L, and finding the index lambda corresponding to the maximum value of the product t I.e. by
Figure BDA0003681145050000154
And 4, step 4: update index set Λ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual correlation degree in the found polymorphic cascade dictionary, and reconstructing the original subset into gamma t ＝[γ t-1 ,Γ l ]；
And 5: solving to obtain theta (t) ＝argmin||r-γ t θ (t) || 2 Updating the residual epsilon (t) ＝r-γ t θ (t) ，t＝t+1；
Step 6: determining the residual error epsilon (t) If the value is smaller than xi, stopping iteration and outputting a reconstructed feature vector theta; otherwise, jumping to step 3 and continuing the steps.
And 7: and recovering the reconstructed feature vector theta to obtain a joint feature matrix theta.
In another embodiment, after the joint feature matrix Θ is obtained, the numerical values in Θ are analyzed, the first L columns of Θ are set to be zero, then the remaining non-zero columns are determined, a non-zero column index vector v is obtained, and the index vector v points to the disturbance event type. Therefore, the invention realizes the synchronous identification of the multi-type disturbance events including the remote disturbance event.
The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be regarded as equivalent replacements within the protection scope of the present invention.

Claims (10)

1. A remote disturbance feature extraction method of a distributed optical fiber sensing system is characterized by comprising the following steps:
s01: dividing distance dimensional intervals of a signal Rayleigh scattering spectrum in the optical fiber channel, establishing corresponding dictionary learning models for the intervals, and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
s02: based on the polymorphic equivalent dictionary, establishing a joint sparse representation model under the cascade of the polymorphic equivalent dictionaries to form joint sparse representation of the multi-echo scattering spectrum;
s03: and constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary, and extracting the characteristics of the remote disturbance signal.
2. The method for extracting the far-end disturbance features of the distributed optical fiber sensing system according to claim 1, wherein the method for forming the multi-state equivalent dictionary in the step S01 includes:
s11: constructing an intrinsic feature dictionary for all detection distances:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ]；
wherein psi l Represents the unit response at the l-th detection range;
obtaining the ground state spectrum
Figure FDA0003681145040000011
Expressed as:
Figure FDA0003681145040000012
wherein eta is a ground state characteristic coefficient;
s12: ζ for nth disturbance event n With a perturbation feature space of
Figure FDA0003681145040000013
The perturbation characteristic complementary space is
Figure FDA0003681145040000014
Based on
Figure FDA0003681145040000015
Obtaining corresponding characteristic coefficient vector
Figure FDA0003681145040000016
Comprises the following steps:
Figure FDA0003681145040000017
wherein eta l And
Figure FDA0003681145040000018
respectively represent eta and
Figure FDA0003681145040000019
the l element of (1);
s13: ζ for disturbance event n And constructing a disturbance modal observation matrix:
Figure FDA0003681145040000021
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a weight coefficient vector;
s14: based on the disturbance modal observation matrix, learning observation is carried out on the intrinsic feature dictionary to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ；
and carrying out cascade combination on the modal dictionary and the intrinsic feature dictionary corresponding to all the disturbance events N to obtain a polymorphic equivalent dictionary:
Figure FDA0003681145040000022
3. the method for extracting the remote disturbance features of the distributed optical fiber sensing system according to claim 1 or 2, wherein the method for forming the joint sparse representation of the multiple echo scattering spectrum in the step S02 includes:
a joint echo matrix composed of M echo light pulses is represented as:
Figure FDA0003681145040000023
wherein, P i Represents a sampled Rayleigh curve corresponding to the ith echo light pulse [. degree] T Represents a transpose of a vector and a matrix;
based on the polymorphic equivalent dictionary, obtaining a joint sparse representation model of a joint echo matrix as follows:
Figure FDA0003681145040000024
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
4. The method for extracting the far-end disturbance feature of the distributed optical fiber sensing system according to claim 3, wherein the method for extracting the far-end disturbance signal feature in the step S03 includes:
s31: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ
wherein,
Figure FDA0003681145040000031
a unit matrix of dimensions (N +1) × (N +1) is represented,
Figure FDA0003681145040000032
representing a kronecker product, and theta is a reconstruction characteristic vector;
s32: initializing a reconstructed feature vector θ (0) 0, residual epsilon (0) R; index set
Figure FDA0003681145040000033
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
s33: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix T l Each gamma is l Multiplying the residual epsilon by 1,2 … L, and finding the index lambda corresponding to the maximum value of the product t I.e. by
Figure FDA0003681145040000034
S34: update index set Λ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual error correlation degree in the found polymorphic equivalent dictionary, and reconstructing the original subset into gamma t ＝[Υ t-1 ,Γ l ]；
S35: solving to obtain theta (t) ＝argmin||r-Υ t θ (t) || 2 Updating the residual epsilon (t) ＝r-Υ t θ (t) ，t＝t+1；
S36: determining the residual error epsilon (t) If the value is smaller than xi, stopping iteration and outputting a reconstructed feature vector theta; otherwise, go to step S33 to continue execution;
s37: and reconstructing the reconstructed feature vector theta to obtain a joint feature matrix theta.
5. The method for extracting the remote disturbance features of the distributed optical fiber sensing system according to claim 4, wherein after the step S03, the method further includes zeroing the first L columns of the reconstructed joint feature matrix Θ, then determining the remaining non-zero columns, obtaining an index vector v of the non-zero columns, and pointing the index vector v to the disturbance event type.
6. The utility model provides a distributed optical fiber sensing system's far-end disturbance feature extraction system which characterized in that includes:
the polymorphic equivalent dictionary building module is used for dividing distance dimensional intervals of a Rayleigh scattering spectrum of a signal in the optical fiber channel, building a corresponding dictionary learning model aiming at each interval and learning and observing an original dictionary to form a polymorphic equivalent dictionary;
the combined sparse representation module of the multi-echo scattering spectrum is used for establishing a combined sparse representation model under cascade connection of polymorphic equivalent dictionaries based on the polymorphic equivalent dictionaries to form combined sparse representation of the multi-echo scattering spectrum;
and the disturbance signal feature extraction module is used for constructing a combined optimization reconstruction algorithm based on the polymorphic cascade dictionary and extracting the far-end disturbance signal features.
7. The remote disturbance feature extraction system of the distributed optical fiber sensing system according to claim 6, wherein the method for forming the polymorphic equivalent dictionary comprises:
s11: constructing an intrinsic feature dictionary for all detection distances:
Ψ＝[ψ 1 ,ψ 2 ,…,ψ L ]；
wherein psi l Represents the unit response at the l-th detection range;
obtaining the ground state spectrum
Figure FDA0003681145040000041
Expressed as:
Figure FDA0003681145040000042
wherein eta is a ground state characteristic coefficient;
s12: ζ for nth disturbance event n Which perturb the feature spaceIs omega ζn The perturbation characteristic complementary space is
Figure FDA0003681145040000043
Based on
Figure FDA0003681145040000044
Obtaining corresponding characteristic coefficient vector
Figure FDA0003681145040000045
Comprises the following steps:
Figure FDA0003681145040000046
wherein eta is l And
Figure FDA0003681145040000047
respectively represent eta and
Figure FDA0003681145040000048
the l element of (1);
s13: ζ for disturbance event n And constructing a disturbance modal observation matrix:
Figure FDA0003681145040000051
wherein diag (z) denotes a matrix formed by diagonal elements of the vector z, β (n) Is a weight coefficient vector;
s14: based on the disturbance modal observation matrix, learning observation is carried out on the intrinsic feature dictionary to form a disturbance event zeta n The corresponding modality dictionary:
Ψ (n) ＝Φ (n) Ψ；
and carrying out cascade combination on the modal dictionary and the intrinsic feature dictionary corresponding to all the disturbance events N to obtain a polymorphic equivalent dictionary:
Figure FDA0003681145040000052
8. the remote disturbance feature extraction system of the distributed optical fiber sensing system according to claim 6 or 7, wherein the method of joint sparse representation of the multiple echo scattering spectra comprises:
a joint echo matrix composed of M echo light pulses is represented as:
Figure FDA0003681145040000053
wherein, P i Represents a sampled Rayleigh curve corresponding to the ith echo light pulse [. degree] T Representing a transpose of a vector and a matrix;
based on the polymorphic equivalent dictionary, obtaining a joint sparse representation model of the joint echo matrix as follows:
Figure FDA0003681145040000054
and theta is a joint feature matrix and is also a column sparse matrix, and a nonzero list of the matrix characterizes disturbance events corresponding to the corresponding polymorphic dictionary.
9. The remote disturbance feature extraction system of a distributed optical fiber sensing system according to claim 8, wherein the method for extracting the remote disturbance signal feature comprises:
s31: vectorizing the joint echo matrix to obtain r ═ vec (r), and further obtaining an expansion joint sparse representation:
r＝Γθ
wherein,
Figure FDA0003681145040000061
a unit matrix of dimensions (N +1) × (N +1) is represented,
Figure FDA0003681145040000062
representing a kronecker product, and theta is a reconstruction characteristic vector;
s32: initializing a reconstructed feature vector θ (0) 0, residual epsilon (0) R; index set
Figure FDA0003681145040000063
Setting a termination iteration threshold xi; t is 1, Θ 0 Is an empty matrix;
s33: using each mode dictionary psi in gamma (n) Decimating with the l-th column to form a word matrix gamma l Each gamma is l L ═ 1,2.. L is multiplied by the residual epsilon, and the index λ corresponding to the product having the maximum value is found t I.e. by
Figure FDA0003681145040000064
S34: update index set Λ t ＝Λ t-1 ∪{λ t Recording the column combination with the highest residual error correlation degree in the found polymorphic equivalent dictionary, and reconstructing the original subset into gamma t ＝[Υ t-1 ,Γ l ]；
S35: solving to obtain theta (t) ＝argmin||r-Υ t θ (t) || 2 Updating the residual epsilon (t) ＝r-Υ t θ (t) ，t＝t+1；
S36: determining the residual error epsilon (t) If the value is smaller than xi, stopping iteration and outputting a reconstructed feature vector theta; otherwise, go to step S33 to continue execution;
s37: and reconstructing the reconstructed feature vector theta to obtain a joint feature matrix theta.
10. The remote disturbance feature extraction system of the distributed optical fiber sensing system according to claim 9, further comprising a disturbance event type identification module, wherein the first L columns of the reconstructed joint feature matrix Θ are set to be zero, then remaining non-zero columns are determined, a non-zero column index vector v is obtained, and the index vector v points to the disturbance event type.
CN202210637622.XA 2022-06-07 2022-06-07 Remote disturbance feature extraction method and system of distributed optical fiber sensing system Pending CN114861737A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210637622.XA CN114861737A (en) 2022-06-07 2022-06-07 Remote disturbance feature extraction method and system of distributed optical fiber sensing system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210637622.XA CN114861737A (en) 2022-06-07 2022-06-07 Remote disturbance feature extraction method and system of distributed optical fiber sensing system

Publications (1)

Publication Number Publication Date
CN114861737A true CN114861737A (en) 2022-08-05

Family

ID=82624442

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210637622.XA Pending CN114861737A (en) 2022-06-07 2022-06-07 Remote disturbance feature extraction method and system of distributed optical fiber sensing system

Country Status (1)

Country Link
CN (1) CN114861737A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115790815A (en) * 2023-01-17 2023-03-14 常熟理工学院 Method and system for rapidly identifying disturbance of distributed optical fiber sensing system
CN118089915A (en) * 2024-04-29 2024-05-28 浙江大学 Disturbance event estimation method and system for optical fiber sensing system background interference suppression

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115790815A (en) * 2023-01-17 2023-03-14 常熟理工学院 Method and system for rapidly identifying disturbance of distributed optical fiber sensing system
CN118089915A (en) * 2024-04-29 2024-05-28 浙江大学 Disturbance event estimation method and system for optical fiber sensing system background interference suppression

Similar Documents

Publication Publication Date Title
CN114861737A (en) Remote disturbance feature extraction method and system of distributed optical fiber sensing system
US11562224B2 (en) 1D-CNN-based distributed optical fiber sensing signal feature learning and classification method
CN110995339B (en) Method for extracting and identifying time-space information of distributed optical fiber sensing signal
CN109344195B (en) HMM model-based pipeline security event recognition and knowledge mining method
CN111442827B (en) Optical fiber passive online monitoring system for transformer winding vibration
CN111222743B (en) Method for judging vertical offset distance and threat level of optical fiber sensing event
CN113687433B (en) Bi-LSTM-based magnetotelluric signal denoising method and system
CN110542831A (en) Fault traveling wave detection method based on variational modal decomposition and S transformation
CN107678064B (en) Real-time extraction method for sound wave time difference
CN113358239A (en) FBG-based wavelength feature identification method
CN111832176A (en) Sea surface wind field inversion method and system of full-polarization microwave radiometer under rainfall condition
CN115265750A (en) Optical fiber distributed acoustic wave sensing system and method
CN105698915B (en) A kind of optical fiber Recognition of Vibration Sources method and device, system
CN111427046B (en) Terahertz pulse echo positioning method for improving detection precision
CN110333506B (en) Method for extracting inhaul cable position parameters of cable force measurement radar
CN112798142A (en) Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation
CN104278990B (en) Log data quality restoration methods and device
CN116046145A (en) System and method for detecting consistency of traffic road optical cable areas
CN113687192B (en) Method for collecting and positioning discharge signal of power transmission line
CN114034327A (en) Brillouin scattering signal measurement method based on sparse sampling and artificial neural network
CN112484837B (en) Optical fiber space positioning system and implementation method thereof
CN115265614A (en) Distance domain dislocation correction method based on cross-correlation spectrum and curve fitting
CN114577360B (en) Automatic analysis calibration and inversion method for Raman temperature measurement radar signals
CN109061725B (en) Automatic identification method of elastic wave data acquisition jumping point and equipment used by same
Velis Statistical segmentation of geophysical log data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination