A method and system for optical fiber fault detection and separation based on two-step neural network

Abstract

The present invention discloses a method and system for optical fiber fault detection and separation based on a two-step neural network. The method comprises acquiring optical fiber source data; performing B-spline approximation on the optical fiber source data to obtain a probability density function of light intensity data; performing static modeling on the probability density function and calculating weights at each moment; using a neural network to approximate the weights at each moment and establishing a nonlinear model between the input light intensity data and the weights; constructing a nonlinear filter for fault detection and a nonlinear observer for fault diagnosis based on the nonlinear model; and inputting the optical fiber source data into the nonlinear filter and the nonlinear observer to obtain the fault size. The present application detects fault intrusion through the nonlinear filter of the fault detection method, and then diagnoses different fault information through the nonlinear observer of the fault diagnosis method to obtain the fault size, thereby achieving the purpose of fault separation.

Description

Technical Field

[0001] The present invention relates to the technical field of optical fiber fault detection, and in particular to an optical fiber fault detection and separation method and system based on a two-step neural network. Background Art

[0002] With the development of the oil and gas pipeline industry, pipelines are increasingly susceptible to various faults during operation, creating an urgent need for more advanced pipeline safety early warning technologies. To better monitor the operational status of oil and gas pipelines, optical fiber is typically buried in the same trench as the pipeline, utilizing fiber optic vibration source detection technology to indirectly assess the pipeline's operational status. This method enables rapid, accurate, and widespread detection and early warning of oil and gas pipelines, and has become a key application for oil and gas pipeline safety early warning.

[0003] The fiber optic vibration source detection system analyzes the vibration signal generated by the optical fiber pipeline and determines whether it is abnormal. Whether the vibration signal can be correctly identified from the optical fiber signal determines the accuracy of the fiber optic vibration source detection system. The front-end hardware system of the fiber optic vibration source detection system uses the optical time domain reflectometry method (OTDR). Among them, OTDR technology has many advantages and is therefore widely used. However, due to the abnormal sensitivity of the sensors used in OTDR technology, it has low suppression ability for information such as transient noise and harmless interference in the environment. The back-end software of the fiber optic detection system is a single-stage open-loop vibration source detection architecture, which directly processes the collected optical fiber signal. Random non-stationary interference will cause the fiber optic detection system to have false alarm / real alarm aliasing problems, which affects the application effect of this technology in the field of fiber optic vibration source detection.

[0004] For example, Patent No. CN201110459814.8 discloses a fiber fault detection method and device. This method is easy to implement, convenient to operate, unrestricted by test conditions, and does not interrupt ongoing tasks in the passive optical network. It can be applied to various types of passive optical networks, improving fiber fault detection efficiency and reducing detection costs. However, this invention can only detect the presence of a fault in the fiber, but cannot estimate the fault's magnitude. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for optical fiber fault detection and separation based on a two-step neural network to solve the problem in the prior art that the fault size cannot be estimated and different faults cannot be distinguished.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention discloses a method for optical fiber fault detection and separation based on a two-step neural network, comprising:

[0008] Obtain fiber optic vibration source data;

[0009] Performing B-spline approximation on the optical fiber source data to obtain a probability density function of the light intensity data, performing static modeling on the probability density function, and calculating the weight at each moment;

[0010] A neural network is used to approximate the weights at each moment and a nonlinear model is established between the input light intensity data and the weights.

[0011] constructing a nonlinear filter for fault detection and a nonlinear observer for fault diagnosis according to the nonlinear model;

[0012] The fiber source data is input into a nonlinear filter and a nonlinear observer to obtain the fault size.

[0013] Furthermore, the expression of the nonlinear model is:

[0014]

[0015] Where x(t) is the state vector, x(t)∈R m , R m are dimensions, A, G, H and D represent the known parameter matrices of the dynamic part of the weight model, g(x(t)) represents the nonlinear vector function of the nonlinear dynamics of the weight model; F1 and F2 represent different types of faults in the fiber optic vibration source data, u(t) is the system input, and V(t) is the corresponding weight of the B-spline expansion.

[0016] Furthermore, the expression of the nonlinear filter is:

[0017]

[0018] in, is the estimated value of the state vector x(t), L is the unknown gain of the detection observer, L = R m×p , R m×p is an m×p dimensional matrix, where ε(t) represents the measured probability density function γ(z,u(t),F1,F2) and the estimated probability density function The integral of the difference.

[0019] Furthermore, the expression of the nonlinear observer is:

[0020]

[0021] in, and is the estimate for faults F1 and F2, is the estimated value of the state vector x(t), L is the unknown gain of the detection observer, L = R m×p, R m×p is an m×p dimensional matrix, where ε(t) represents the measured probability density function γ(z,u(t),F1,F2) and the estimated probability density function The integral of the difference.

[0022] Furthermore, the obtaining of optical fiber vibration source data includes:

[0023] The light source is injected into the optical fiber head end, and the optical fiber signal is collected at the optical fiber tail end;

[0024] The optical fiber signal is converted into an electrical signal through a photoelectric conversion module and the electrical signal is amplified through an amplifier circuit;

[0025] The dual-channel data acquisition module records the amplified electrical signal to complete the acquisition of the fiber optic vibration source data.

[0026] Furthermore, after obtaining the optical fiber vibration source data, the method further includes:

[0027] Converting the acquired optical fiber vibration source data into readable data, wherein the readable data includes I and Q data;

[0028] The I and Q data are fused to obtain data that has undergone differential operation, square sum operation, and modulus operation.

[0029] In a second aspect, the present invention discloses a fiber fault detection and separation system based on a two-step neural network, comprising a processor and a storage medium;

[0030] The storage medium is used to store instructions;

[0031] The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 6.

[0032] In a third aspect, the present invention discloses a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.

[0033] According to the above technical solution, the embodiments of the present invention have at least the following effects: the present application constructs a nonlinear filter and a nonlinear observer based on a nonlinear model. After the fiber optic vibration source data is input into the nonlinear filter and the nonlinear observer, both can detect the input. When multiple faults are simultaneously present in the fiber optic vibration source data, the fault intrusion is detected by the nonlinear filter of the fault detection, and then the nonlinear observer of the fault diagnosis diagnoses different fault information, obtains the fault size, and achieves the purpose of fault separation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of the optical fiber fault detection and separation method of the present invention;

[0035] Figure 2 is a schematic diagram of an optical fiber fault detection and separation system of the present invention;

[0036] Figure 3 This is the fusion flow chart of fiber optic vibration source data;

[0037] Figure 4 Output PDF 3D graph for actual measurement with faults;

[0038] Figure 5 is the response diagram of the residual vector ε(t);

[0039] Figure 6 is the fault F1 and its diagnostic observer response data diagram;

[0040] Figure 7 is the fault F2 and its diagnostic observer response data diagram;

[0041] Figure 8 Figure 2 is the response data diagram of faults F1, F2 and their fault observer. DETAILED DESCRIPTION

[0042] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0043] The present invention collects output data from a fiber optic vibration source detection system to establish its output probability density functions (PDFs). A nonlinear dynamic model is established using the relationship between input light intensity and weights associated with the output PDFs. Based on the nonlinear dynamic model, a nonlinear filter and a nonlinear observer are constructed to diagnose multiple faults, thereby achieving the purpose of fault isolation. To avoid false alarms, the fiber optic vibration source detection system uses a constant false alarm detection method to determine the threshold for fiber optic vibration source detection.

[0044] Example 1

[0045] The present invention discloses a method for detecting and separating optical fiber faults based on a two-step neural network. Figures 1 to 8 As shown, the method includes the following steps: obtaining optical fiber vibration source data; performing B-spline approximation on the optical fiber vibration source data to obtain the probability density function of the light intensity data, performing static modeling on the probability density function, and calculating the weight at each moment; using a neural network to approximate the weight at each moment, and establishing a nonlinear model between the input light intensity data and the weight; constructing a nonlinear filter for fault detection and a nonlinear observer for fault diagnosis based on the nonlinear model; and inputting the optical fiber vibration source data into the nonlinear filter and the nonlinear observer to obtain the fault size.

[0046] This application constructs a nonlinear filter and a nonlinear observer based on a nonlinear model. After the fiber optic vibration source data is input into the nonlinear filter and the nonlinear observer, both can detect the input. When multiple faults are simultaneously present in the fiber optic vibration source data, the fault intrusion is detected by the nonlinear filter of the fault detection, and then the nonlinear observer of the fault diagnosis diagnoses different fault information, obtains the fault size, and achieves the purpose of fault separation.

[0047] The following describes this application through specific steps.

[0048] Step 1: Use the fiber optic vibration source hardware system to collect fiber optic data, process the fiber optic data through MATLAB, and convert the collected data into readable .mat data.

[0049] Since the collected data includes the optical signal data returned in the pipeline, the optical fiber signal is composed of I and Q data. Since false alarms generally do not appear in both I and Q channels at the same time, the false alarm rate is initially reduced by joint screening of the two channels. The collected matArray data is not processed. It is necessary to fuse the I and Q data and export the data after differential operation, square sum operation and modulus operation in Matlab. The vibration data fusion process is as follows: Figure 3 shown.

[0050] In step 1, the optical fiber data is collected as follows Figure 2 As shown, the optical fiber fault acquisition principle of the present invention is that when the optical fiber is interfered with or invaded at any position, the pipeline in the optical fiber will generate a certain amount of micro-vibration, affecting the value of the optical fiber signal. Utilizing this characteristic, a light source is injected into the optical fiber's head end, and the optical fiber signal is collected at the tail end. The optical signal is converted into an electrical signal and amplified by a photoelectric conversion module and an amplifier circuit. Finally, the data is recorded by a dual-channel data acquisition module to complete the acquisition of the optical fiber vibration source data. Among them, fault F1 is the collected pickaxe data, and F2 is the collected mechanical excavation data. Pickaxe is selected as optical fiber fault F1, and mechanical excavation is selected as optical fiber fault F2.

[0051] Step 2: Use the square root B-spline expansion method to approximate the probability density function PDFs of the output light intensity data, perform static modeling on the PDFs of the output light intensity data, and calculate the weight at each moment.

[0052] We use the square root B-spline model with approximation error as shown below to approximate the output PDFs and construct the static model as Σ1:

[0053]

[0054] Among them, v i (u(t),F1,F2)(i=1,2,…,n) are the corresponding weights of B-spline expansion, bi (z)(i=1,2,…,n) is a pre-specified basis function on the interval [a,b], F1 and F2 represent different types of faults in the fiber optic vibration source system, u(t) is the system input, and ω0(z,u(t),F1,F2) represents the error obtained after approximating the probability density function.

[0055]

[0056] Among them, Λ1∈R (n-1)×(n-1) ,Λ2∈R (n-1)×1 ,Λ3∈R 1×1 Is a known matrix or constant. Simplify V(u(t), F1, F2) to V(t), where V(t) is the corresponding weight of the B-spline expansion, corresponding to the output of the dynamic model of the weight model to be identified. Let Then the inequality (1-ω1(z,u(t),F1,F2))Λ3-V T (t)Λ0V(t)≥0 holds. Where:

[0057]

[0058] According to the inequality Σ3:(1-ω1(z,u(t),F1,F2))Λ3-V T (t)Λ0V(t)≥0,

[0059] V T (t)Λ0V(t)≤(1-ω1(z,u(t),F1,F2))Λ3, where Λ0>0,1-ω1(z,u(t),F1,F2)>0,

[0060] choose

[0061] Then B(z)=[b1(z)b2(z)], V(u(t),F1,F2)=[v1(u(t),F1,F2)v2(u(t),F1,F2)], from which we can get:

[0062]

[0063] The rewritten static model is Σ4:

[0064] Among them, h0(V(t),ω1) is the correlation function of V(t) and ω1(z,u(t),F1,F2), among which,

[0065]

[0066] From this we can further obtain Σ5:

[0067] in, According to the boundedness of V(t) and the known Λ i (i=1,2,3), |ω0(z,u(t),F1,F2)|≤δ0, we can assume that |ω(z,u(t),F1,F2)|≤δ holds for all {z,u(t),F1,F2}, where δ is a known positive constant, and we choose δ=0.1.

[0068] Assume that h(V(t)) satisfies the Lipschitz condition, that is, for any V1(t) and V2(t), there exists a known matrix U1 such that Σ6:||h(V1(t))-h(V2(t))||≤||U1(V1(t)-V2(t))||.

[0069] Among them, select U1 = [1 1].

[0070] Step 3: For the weight coefficients at each moment, a neural network is used to approximate the nonlinear model between the input light intensity and the weight value; a nonlinear dynamic model such as Σ7 is considered between V(t) and u(t):

[0071]

[0072] Where x(t)∈R m is the state vector, A, G, H and D represent the known parameter matrices of the dynamic part of the weight system. g(x(t)) is a nonlinear vector function representing the nonlinear dynamics of the weight model. Assume that g(x(t)) satisfies g(0) = 0 and the following norm Σ8 holds for any x1(t) and x2(t), and U2 is a known matrix

[0073] Σ8: ||g(x1(t))-g(x2(t))||≤||U2(x1(t)-x2(t))||.

[0074] choose

[0075]

[0076] Step 4a: Construct a nonlinear filter for fault detection based on the system model established in step 3, analyze the residual, and thus detect the fault; first, construct a nonlinear filter such as Σ9 based on the system model in step 2:

[0077]

[0078] in, is the estimated value of the state vector x(t), L = R m×p is the unknown gain of the detection observer, R m×pis an m×p dimensional matrix, σ(z)=R p×1 is a pre-specified weight vector defined on [a, b]. The residual ε(t) represents the probability density function of the measurement γ(z,u(t),F1,F2) and the estimated probability density function The integral of the difference between . Define the system estimation error From equations (11) and (13), we can get the first-order derivative of e(t) with respect to time t:

[0079]

[0080] in:

[0081]

[0082] The residual ε(t) can be further expressed as Σ 11 :

[0083] choose △(t)≤0.15,η=1.5, and the fault detection threshold is α=0.5023.

[0084] Step 4b: Construct a nonlinear observer for fault diagnosis based on the system model established in the third step. By selecting an appropriate observer gain, the estimation error is stabilized within a range. After the fault is detected according to the results of the fourth step, fault diagnosis is required to estimate the size of different faults and achieve the purpose of fault separation. To this end, the following Σ 12 The fault diagnostic filter shown:

[0085]

[0086] in, and is the estimate for faults F1 and F2, Λ i (i=5,6,7,8) is and estimation error Related learning operators. The fault estimation error is and Fault diagnosis observer gain L=P -1 The error of the R system estimation is Σ 13 As shown in the formula:

[0087]

[0088] Further according to the following linear matrix inequality Σ 14 and Schur complement to solve the parameters P>0, R and Λ i(i=5,6,7,8):

[0089]

[0090] in,

[0091] Π2=[λ1RΓ2 λ2PG θ1R]

[0092] Π3=[θ2Λ6 θ3Λ6Γ2]

[0093] Π4=[θ2Λ8 θ3Λ8Γ2]

[0094]

[0095] Then the estimated error system Σ is 15 is stable, and the estimation error satisfies:

[0096]

[0097] Select λ1=λ2=1, θ1=θ2=2, θ3=-2, κ=0.1, △(t)≤0.15 Solve the linear matrix inequality to obtain:

[0098]

[0099]

[0100]

[0101] Step 5: Input the fiber optic vibration source data into a nonlinear filter and nonlinear observer to determine the fault size. Specifically, the collected data is processed using MATLAB to generate a .mat file. This processed data is then treated as a fault and simulated using MATLAB / Simlink. The designed nonlinear filter is used to detect the fault, and the nonlinear observer is used for fault diagnosis. The magnitude of each fault is estimated, achieving fault isolation.

[0102] Example 2

[0103] An embodiment of the present application also provides a two-step neural network optical fiber fault detection and separation system, which includes a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method described in Example 1.

[0104] The present application also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method described in fact and strength 1 when the program is executed by a processor.

[0105] The memory is used to store all model data, as well as various data such as the two-step neural network optical fiber fault detection and separation method provided in the embodiment of the present application and the corresponding computing program instructions of the system. The memory can be a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable read-only memory (EPROM), etc.

[0106] The processor is used to read and run the computer program instructions corresponding to the two-step neural network optical fiber fault detection and separation method stored in the memory, and execute the emergency decision control method provided in the embodiment of the present application.

[0107] A processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, including a central processing unit (CPU) or a network processor (NP); or a digital signal processor (DSP), discrete gate or transistor logic devices, or discrete hardware components.

[0108] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0109] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0112] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. A method for optical fiber fault detection and separation based on a two-step neural network, characterized in that: include: Obtain fiber optic vibration source data; Performing B-spline approximation on the optical fiber source data to obtain a probability density function of the light intensity data, performing static modeling on the probability density function, and calculating the weight at each moment; A neural network is used to approximate the weights at each moment and a nonlinear model is established between the input light intensity data and the weights. constructing a nonlinear filter for fault detection and a nonlinear observer for fault diagnosis according to the nonlinear model; Inputting the optical fiber source data into a nonlinear filter and a nonlinear observer to obtain the fault size; The expression of the nonlinear model is: ； in, is the state vector, , For the dimension, 、 、 and represents the known parameter matrix of the dynamic part of the weight model, An already nonlinear vector function representing the nonlinear dynamics of the weight model; Indicates different types of faults in the fiber optic vibration source data, is the system input, is the corresponding weight of the B-spline expansion; The expression of the nonlinear filter is: ； in, is the state vector The estimated value of is the unknown gain of the detection observer, , is an m×p dimensional matrix, The probability density function representing the measurement and the estimated probability density function The integral of the difference; The expression of the nonlinear observer is: ； in, and For faults and Estimates, is the state vector The estimated value of is the unknown gain of the detection observer, , is an m×p dimensional matrix, The probability density function representing the measurement and the estimated probability density function The integral of the difference.

2. The optical fiber fault detection and separation method based on a two-step neural network according to claim 1 is characterized in that: The obtaining of optical fiber vibration source data comprises: The light source is injected into the optical fiber head end, and the optical fiber signal is collected at the optical fiber tail end; The optical fiber signal is converted into an electrical signal through a photoelectric conversion module and the electrical signal is amplified through an amplifier circuit; The dual-channel data acquisition module records the amplified electrical signal to complete the acquisition of the fiber optic vibration source data.

3. The optical fiber fault detection and separation method based on a two-step neural network according to claim 1, characterized in that: After obtaining the fiber optic vibration source data, the following steps are also included: Converting the acquired optical fiber vibration source data into readable data, wherein the readable data includes I and Q data; The I and Q data are fused to obtain data that has undergone differential operation, square sum operation, and modulus operation.

4. A fiber fault detection and separation system based on a two-step neural network, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images