Optical cable fault positioning method based on OTDR signal characteristic adaptive denoising and event identification

Abstract

The invention provides an optical cable fault positioning method based on OTDR signal characteristic adaptive denoising and event identification, and the method comprises the steps: classifying modal components after OTDR signal adaptive noise decomposition based on a preset entropy threshold value, and obtaining a noise dominant component, a mixed component and a signal dominant component; singular value difference spectrum abrupt change point detection is carried out on the noise dominant component, and residual useful signals are extracted; verifying the entropy value of the mixed component, and reconstructing the component meeting the entropy threshold value, the signal dominant component and the residual useful signal into a de-noised signal; synchronously optimizing hyper-parameters and feature selection subsets of the support vector machine by adopting a swarm intelligence algorithm, wherein search parameters of the algorithm are dynamically updated according to an exponential decay mechanism; and outputting fault point space position information based on the optimized support vector machine model.

Description

Technical Field

[0001] The invention belongs to the technical field of optical fiber communication monitoring, and in particular relates to an optical cable fault locating method based on OTDR signal characteristic adaptive denoising and event recognition. Background Art

[0002] Optical Time Domain Reflectometry (OTDR), as a fiber optic fault detection instrument, can measure fiber length, transmission attenuation, and fault location. It can effectively measure splice loss, fusion loss, bend loss, and the location of certain fault points at any location within a sensing fiber. This instrument is crucial for timely detecting fiber optic network faults and ensuring network security and stability, and has been widely used in various fields of optical fiber and cable. The OTDR signal curve reflects the loss distribution of backscattered light along the fiber. However, due to environmental and system influences, event points on the curve are easily obscured by noise. When the curve is severely contaminated by noise, events on the curve are difficult to identify. Furthermore, locating and identifying fiber optic faults requires manual operation, which not only wastes manpower and material resources, but also relies on the maintenance personnel's experience to interpret the OTDR curve, resulting in relatively low accuracy. This impacts the smooth progress of repair work and fails to meet the requirements for effective fiber optic fault repair. Therefore, in order to solve the problem that the use of OTDR in optical fiber maintenance fault detection relies on the experience of workers and to improve the accuracy of OTDR in fault location and fault type identification, it is very necessary to study the OTDR signal processing algorithm. Summary of the Invention

[0003] To address the shortcomings of existing technologies, such as severe OTDR signal noise interference, difficulty in extracting fault event features, and low efficiency in optimizing classification model parameters, the present invention provides a method for locating optical cable faults based on adaptive denoising and event identification of OTDR signal characteristics. This method innovatively integrates adaptive noise processing with an intelligent optimization mechanism, solving core technical challenges through a three-level processing framework:

[0004] Noise signal directional separation

[0005] Using an improved adaptive noise decomposition algorithm (such as ICEEMDAN), modal components are intelligently classified into noise-dominated, mixed, and signal-dominated components using a preset entropy threshold. Singular value difference spectrum mutation point detection is performed on the noise-dominated components. Through Hankel matrix reconstruction and first negative cutoff point truncation techniques, useful signals submerged in noise are accurately extracted.

[0006] Mixed-Signal Quality Assurance Reconfiguration

[0007] A secondary entropy validation is performed on the mixed components, retaining only those components that meet the threshold requirements. These components are then fused with the dominant signal components and the extracted useful signal to reconstruct a high-quality signal. A sliding window with specific window parameters is used to fully capture the event characteristics, extracting multidimensional features including the spectral peak stability index.

[0008] Dynamic Tuning of Classification Models

[0009] The search parameters of the intelligent optimization algorithm are dynamically adjusted through an exponential decay mechanism, and the key hyperparameters of the support vector machine (including the penalty factor and kernel parameters for wide-range search) and the optimal feature subset are simultaneously optimized, significantly improving the accuracy of fault classification.

[0010] This solution achieves sub-hundred-meter positioning accuracy (experimental error <15 meters) on OTDR equipment (dynamic range ≥36dB, measurement distance 1-180km), significantly improves the signal-to-noise ratio to over 42dB, and achieves a fault identification accuracy rate exceeding 97%. The core advantages are:

[0011] A three-layer separation mechanism of noise components guided by entropy threshold;

[0012] Directional signal extraction based on the first negative point of the singular value difference spectrum;

[0013] A dynamic optimization framework with exponential decay of search parameters;

[0014] Break through the technical bottlenecks of incomplete noise suppression, fuzzy event boundaries, and poor model generalization in traditional optical cable fault location.

[0015] The technical solution specifically adopted by the present invention to solve the technical problem is:

[0016] A method for locating optical cable faults based on adaptive denoising and event recognition of OTDR signal characteristics:

[0017] Based on the preset entropy threshold, the modal components of the OTDR signal after adaptive noise decomposition are classified to obtain the noise-dominant component, the mixed component and the signal-dominant component;

[0018] Performing singular value difference spectrum mutation point detection on the noise dominant component to extract the residual useful signal;

[0019] Verify the entropy value of the mixed component, and reconstruct the component that meets the entropy value threshold, the signal dominant component and the residual useful signal into a denoised signal;

[0020] A swarm intelligence algorithm is used to simultaneously optimize the hyperparameters and feature selection subsets of the support vector machine, and the search parameters of the algorithm are dynamically updated according to an exponential decay mechanism;

[0021] The spatial location information of the fault point is output based on the optimized support vector machine model.

[0022] Furthermore, the algorithm used in the adaptive noise decomposition is ICEEMDAN, k is the number of adjustments, and the noise adjustment coefficient β is k satisfy:

[0023] When k = 1, β1 is proportional to the inverse of the initial signal-to-noise ratio ε0 of the input signal;

[0024] When k≥2, β k Proportional to the standard deviation of the k-1th order residual.

[0025] Furthermore, the entropy value is sample entropy, the threshold is 0.1, the embedding dimension m=2, and the similarity tolerance r=0.2×signal standard deviation.

[0026] Furthermore, the singular value difference spectrum mutation point detection is performed on the noise-dominant component to extract the residual useful signal. Specifically, the noise component is constructed as a Hankel matrix, the singular value of the noise component is obtained by the SVD algorithm, and the residual useful information in the noise component is extracted using the singular value difference spectrum method; the number of rows l of the Hankel matrix satisfies: when the signal length n is an even number, l=n / 2; when n is an odd number, l=(n+1) / 2.

[0027] Furthermore, the singular value difference spectrum mutation point detection includes: calculating a sequence of adjacent difference values ​​of the singular values ​​arranged in descending order; and taking the point where a negative value first appears in the sequence as a dividing point.

[0028] Furthermore, the swarm intelligence algorithm is the vulture optimization algorithm BES, which dynamically adjusts the search parameter IP to meet the following requirements:

[0029]

[0030] Among them, α is a preset constant, t is the current number of iterations, T max is the maximum number of iterations.

[0031] Furthermore, the support vector machine is a twin support vector machine TWSVM, and the optimized hyperparameters include penalty factors C1, C2 and kernel parameter γ, where C1, C2∈[0.1, 100], γ∈[0.001, 10].

[0032] Furthermore, a sliding window is used to extract local features of the denoised signal, wherein the window width covers at least one complete event feature, and adjacent windows partially overlap.

[0033] Furthermore, the local features include time domain statistical features, frequency domain features and nonlinear features.

[0034] And, an optical cable fault location system, comprising:

[0035] A signal acquisition module configured to acquire an OTDR original signal;

[0036] A processing module configured to execute the above method;

[0037] The output module is configured to display the spatial location information of the fault point.

[0038] And, a computer device includes a memory, a processor and a computer program stored in the memory, and the processor implements the above method when executing the computer program.

[0039] A non-transitory computer-readable storage medium stores a computer program, which implements the method described above when executed by a processor.

[0040] Compared with the prior art, the present invention and its preferred embodiments have at least the following beneficial effects:

[0041] 1. Breakthrough the technical bottleneck of noise signal aliasing

[0042] Through the three-layer component classification mechanism (noise-dominated component, mixed component and signal-dominated component) driven by entropy threshold, combined with the targeted purification processing of the noise component's mutation points, the signal-to-noise ratio is significantly improved, effectively solving the problem of OTDR fault event characteristics being submerged by noise.

[0043] 2. Optimize algorithm convergence efficiency and generalization ability

[0044] An exponential decay mechanism is used to dynamically adjust the search parameters of the swarm intelligence algorithm, balance global exploration and local development capabilities, and simultaneously optimize the hyperparameters and key feature subsets of the support vector machine to avoid the risk of model overfitting and improve the accuracy and real-time performance of fault classification in complex scenarios.

[0045] 3. Ensure the integrity and positioning accuracy of fault events

[0046] An event feature extraction mechanism based on a sliding window (window width covers the complete event, and feature types are multi-dimensionally integrated), supplemented by secondary verification of the entropy value of the mixed component, ensures that event boundary information is captured without omission and achieves sub-hundred-meter spatial positioning accuracy.

[0047] 4. Expand the engineering applicability of the method

[0048] Supports industrial-grade OTDR equipment with a dynamic range of ≥36dB and a ranging range of 1-180km, providing a standardized fault location process to replace reliance on manual experience.

[0049] 5. Reduce the complexity of system implementation

[0050] Build an end-to-end processing chain from signal acquisition, noise purification, feature optimization to positioning decision-making, and reduce the cost of manual parameter adjustment through adaptive selection of algorithm parameters (such as entropy threshold 0.1 and dynamic coefficient α). BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0052] Figure 1 This is a flow chart of the ICEEMDAN-SVD joint adaptive denoising algorithm according to an embodiment of the present invention;

[0053] Figure 2 This is a signal decomposition flow chart of the ICEEMDAN algorithm according to an embodiment of the present invention;

[0054] Figure 3 This is a comparison diagram of the ideal OTDR signal and the noisy signal according to an embodiment of the present invention;

[0055] Figure 4 Schematic diagram of IMFs decomposition of noisy OTDR signals by ICEEMDAN according to an embodiment of the present invention;

[0056] Figure 5 This is a comparison chart of the signal before and after denoising using the ICEEMDAN-SVD algorithm according to an embodiment of the present invention;

[0057] Figure 6 This is a diagram of the reconstruction error between the signal denoised by the ICEEMDAN-SVD algorithm according to an embodiment of the present invention and the ideal OTDR signal;

[0058] Figure 7 This is a flowchart of IBES-TWSVM event classification according to an embodiment of the present invention;

[0059] Figure 8 This is a diagram showing the classification results of the OTDR dataset after denoising by IBES-TWSVM according to an embodiment of the present invention;

[0060] Figure 9 This is a diagram showing the classification results of the OTDR dataset without denoising using IBES-TWSVM according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the features and advantages of the present invention more clearly understood, the following embodiments are given for detailed description:

[0062] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which this application belongs.

[0063] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0064] To overcome the shortcomings of the existing technology, an embodiment of the present invention provides an optical cable fault location algorithm based on ICEEMDAN-SVD and improved BES-TWSVM, which realizes denoising of the OTDR curve signal, judges the denoising effect by the signal-to-noise ratio and root mean square error of the denoised signal, and reduces the difficulty of identifying event points in the signal. The vulture optimization algorithm is used to adjust the penalty factors C1, C2 and kernel function parameter γ in the twin support vector machine (TWSVM) algorithm. The C1 and C2 parameters in TWSVM are sensitive to the classification effect of TWSVM. Improper settings will lead to poor TWSVM classification effect. Therefore, C1 and C2 are optimized and solved by a swarm intelligence optimization algorithm to improve the classification performance of TWSVM.

[0065] In the design scheme of this embodiment, an OTDR instrument is used to obtain an event curve signal in an optical fiber when an optical cable fails; then the obtained OTDR curve is decomposed by ICEEMDAN, decomposing it into multiple IMFs and calculating the sample entropy of each IMF. If the sample entropy is higher than a threshold, it means that the IMF is mainly composed of noise components, and if it is lower than the threshold, it is mainly composed of useful signal components; the IMF above the threshold is denoised by SVD, and the sample entropy of the denoised signal is calculated again. If the sample entropy after denoising is still higher than the threshold, it means that the IMF is entirely noise, and the IMF component is discarded. If the sample entropy after denoising is lower than the threshold, it means that most of the noise in the IMF component has been removed. The processed IMF and the useful signal component are used to reconstruct the signal to obtain the denoised OTDR signal; the signal-to-noise ratio and root mean square error of the denoised signal are calculated to determine the denoising effect; the improved condor optimization algorithm is used as the optimization algorithm, the penalty factors C1, C2 and kernel parameter γ of TWSVM and the binary feature are selected as the optimization target, and the optimal feature subset and TWSVM parameters C1, C2 and γ are obtained by solving; event discrimination is performed on the OTDR denoised signal to obtain the location information of the optical cable breakpoint. The present invention can efficiently denoise the OTDR signal, help improve the signal-to-noise ratio of the signal, reduce the root mean square error and restore the original signal, and efficiently and accurately locate the optical cable fault point.

[0066] The implementation process includes the following steps:

[0067] S1: Obtain the original OTDR signal x(t) when the optical cable fails through the OTDR device;

[0068] S2: Determine the embedding dimension m and similarity tolerance r of the sample entropy, as well as the sample entropy threshold SE.

[0069] S3: Decompose the acquired original OTDR signal x(t) into a set of intrinsic mode functions (IMFs) and residual terms using the ICEEMDAN algorithm

[0070] S4: Calculate the sample entropy of each IMF separately. If the sample entropy is greater than a preset value of 0.1, it is marked as a noise component; if it is less than the preset value, it is marked as a useful component.

[0071] S5: The noise component is constructed as a Hankel matrix, and the singular values ​​of the noise component are obtained through the SVD algorithm. The singular value difference spectrum method is used to extract the remaining useful information in the noise component.

[0072] S6: Calculate the sample entropy of the denoised component using the SVD algorithm. If the sample entropy is still greater than 0.1, discard it. If it is less than 0.1, reconstruct it with the useful component in step 2 to obtain the denoised OTDR signal.

[0073] S7: The algorithm denoising effect is obtained by calculating the signal-to-noise ratio (SNR) and root mean square error (RMSE) of the reconstructed signal.

[0074] S8: The denoised OTDR signal is passed through a sliding window with a width of 50 data points and a sliding step of 20 data points to extract the local features of the signal.

[0075] S9: Using the improved vulture optimization algorithm (BES) as the optimization algorithm, the penalty factors C1, C2 and kernel parameter γ of TWSVM and the binary features are selected as the optimization targets to obtain the optimal feature subset and TWSVM parameter C1 best 、C2 best with γ best .

[0076] S10: The feature subset of the optimal solution, C1 best 、C2 best with γ best As the parameters of TWSVM, the denoised OTDR signal data is used for event discrimination to obtain the classification accuracy of different events.

[0077] S11: Output the distance information of the optical cable fault point to complete the entire optical cable fault location process.

[0078] As a preferred solution of this embodiment, the dynamic range of the original OTDR equipment is 36dB, the minimum sampling interval is 0.25m, the number of sampling points does not exceed 32,000, and the measurement range is 1km-180km.

[0079] As a preferred solution of this embodiment, in step S2, m defines the length of the vector formed in the time series. By selecting the parameter m of the sample entropy, the time series can be Converted into a series of m-dimensional vectors. The choice of M affects the sensitivity of sample entropy. Smaller values ​​of m may lead to highly complex evaluations, while larger values ​​of m can capture longer-term data dependencies. r is a threshold used to determine whether two m-dimensional vectors are similar. If the maximum absolute difference between the corresponding elements of the two vectors does not exceed r, the two vectors are considered similar. Sample entropy is defined as: .

[0080] As a preferred solution of this embodiment, in step S3 is the average value, Original signal, is the signal after adding white noise, is white noise with zero mean and unit variance, is the local mean of the signal, is the kth modal component of the EMD decomposition, is the coefficient of adding noise during the kth adjustment. The specific decomposition process of ICEEMDAN is as follows:

[0081] Use EMD to construct . Find their local mean to get the first residual .

[0082] Subtracting the first residual gives the first component obtained from the original signal The weight is .

[0083] As a preferred solution of this embodiment, the first residual The local mean of the average is used to obtain the second residual , and use the first residual minus the second residual to get the second modal component for .

[0084] for Repeat the above steps to calculate the kth residual, The kth modal component is .

[0085] Repeat the previous step until all components are obtained. Decomposed into a group Component and residual r, that is: .

[0086] Select Constant To adjust the signal-to-noise ratio of the added noise and residual. hour, ;when hour, in is the inverse of the required signal-to-noise ratio of the input signal and the first noise added, is the standard deviation calculation.

[0087] As a preferred solution of this embodiment, in step S4, after the original OTDR signal is decomposed into multiple IMFs by the ICEEMDAN algorithm, the sample entropy of each IMF is calculated respectively. If the IMF sample entropy is greater than the threshold value 0.1, it is marked as a noise component; if it is less than the threshold value, it is marked as a useful component.

[0088] As a preferred solution of this embodiment, in step S5, the IMF marked as the noise component is decomposed using SVD to extract useful components therein, so as to ensure that the useful signal is extracted to the maximum extent.

[0089] The noise component Construct a Hankel matrix R that can be processed by SVD. In the Hankel matrix, It can be expressed as:

[0090]

[0091] in, , For signal The length of the Hankel matrix. Since the number of rows and number of columns Different values ​​of will affect the signal denoising effect. In order to separate the useful components and noise components in the signal as much as possible, The value of should be as large as possible, so when When it is an even number, ;when When is an odd number, .

[0092] Perform singular value decomposition on the two-dimensional matrix R to obtain the left singular vector , singular values , and the right singular vectors The constructed trajectory matrix can be expressed as: .

[0093] The singular value difference spectrum method is selected to determine the threshold. The singular values ​​after the mutation point are set to 0, and the singular values ​​before the mutation point are retained and inverse SVD transformation is performed to obtain the denoised signal.

[0094] As a preferred solution of this embodiment, the sample entropy of the noise component after denoising by the SVD algorithm in step S6 is calculated again. If the sample entropy is still greater than the set threshold, it is discarded. If it is less than the threshold, the signal is reconstructed with the IMFs previously marked as useful signals to obtain the denoised OTDR signal curve and a clearer event waveform diagram.

[0095] As a preferred solution of this embodiment, in step S7, the signal-to-noise ratio and the root mean square error are used as evaluation indicators of the denoising performance. The higher the signal-to-noise ratio, the more obvious the noise suppression effect; the smaller the root mean square error, the better the denoising performance. The two evaluation indicators are defined as:

[0096]

[0097]

[0098] in is a noisy signal, is the denoised signal, is the length of the signal.

[0099] As a preferred solution of this embodiment, step S8 extracts local signal features by setting a sliding window with a width of 50 data points to ensure that each window can cover the complete event feature information. A window that is too small cannot contain the entire event data, and a window that is too large easily concentrates multiple events in one window, causing feature aliasing. At the same time, in order to avoid the problem of boundary information loss that may occur during window switching, the step size of each window sliding is set to 20 data points. There is an overlapping area between adjacent windows to ensure that at least one window can completely cover the event features. By dividing the complete signal into multiple overlapping small signals, it is possible to capture event details and improve the efficiency of signal feature extraction. Multiple signal feature values ​​are extracted in each window, including statistical features, dynamic features, spectral features, etc.

[0100] As a preferred solution of this embodiment, when step S9 uses the BES algorithm to simultaneously optimize TWSVM parameters and feature selection, the first part is to optimize the kernel parameters C1, C2 and kernel parameter γ of TWSVM, and the second part is the features in the original data set.

[0101] Prepare a training set and a test set of OTDR denoised signal data, extract feature data from the OTDR signal and perform normalization processing in step S8.

[0102] Set the vulture population size, optimization range of γ, C1, and C2, binary encoding of feature sets, maximum number of iterations, objective function dimension, and initial value boundary conditions.

[0103] Calculate the fitness values ​​of individuals in a vulture population.

[0104] In the BES algorithm, the forward speed of the hunting space phase is dynamically adjusted to avoid falling into the local optimal solution. In this embodiment, the originally fixed iteration parameter is improved to: .

[0105] The improved BES algorithm (IBES) is used to search TWSVM parameters γ, C1, C2 and binary features.

[0106] Parameters γ, C1, C2 and data features are calculated and input into the TWSVM classifier to select feature subsets.

[0107] Calculate the fitness value of cross validation , where acc is the correct classification accuracy; R and N are the number of feature subsets selected by feature selection and the total number of features in the original feature set, respectively, and the optimal individual position and fitness value of the vulture population are updated.

[0108] Determine whether the maximum number of iterations has been reached. If not, return to the step of calculating the fitness value of the vulture population individuals. If reached, output the selected feature subset, parameters γ, C1, C2, and classification accuracy.

[0109] As a preferred solution of this embodiment, step S10 is to input the parameter values ​​when obtaining the best classification effect into the algorithm, and train and test the OTDR data set to finally obtain the algorithm classification effect and the optical path distance information of the event.

[0110] As a preferred solution of this embodiment, step S11 is to deploy the algorithm on the device to obtain the distance information of the optical cable fault point.

[0111] The following further demonstrates and introduces the above solutions of the embodiment of the present invention through more specific test examples:

[0112] like Figures 1-9 The optical cable fault location solution based on ICEEMDAN-SVD and improved BES-TWSVM is shown in Figure 1 and Figure 7 , its design and implementation process includes the following steps:

[0113] Step 1: Use the ICEEMDAN algorithm to decompose the noisy OTDR signal into a set of IMFs components. Figure 2 Flowchart of signal decomposition for ICEEMDAN, Figure 3This is a comparison chart of an ideal OTDR signal and a noisy OTDR signal after adding Gaussian white noise, which is common in actual measurements. Figure 4 Figure 1 shows the waveforms of the various IMFs obtained after the noisy OTDR signal is decomposed using the ICEEMDAN algorithm. The sample entropy of each IMF is calculated and statistically reported in Table 1.

[0114] Table 1 Sample entropy of each IMF component

[0115] IMF components Sample entropy IMF components Sample entropy IMF1 0.1354 IMF9 0.0145 IMF2 0.2344 IMF10 0.0071 IMF3 0.2605 IMF11 0.0051 IMF4 0.3075 IMF12 0.0054 IMF5 0.2641 IMF13 0.0045 IMF6 0.1918 IMF14 0.0023 IMF7 0.1564 IMF15 0.0014 IMF8 0.0916 IMF16 0.0001

[0116] Step 2: The noise inherent in the noisy OTDR signal is decomposed into a subset of IMFs. To enable the algorithm to automatically separate these components into useful and noise components, the sample entropy metric is introduced. Due to the randomness and irregularity of noise, the sample entropy of noise components is generally higher than the sample entropy of useful components. The reconstruction dimension and reconstruction threshold parameters of the sample entropy are set to 2 and 0.2, respectively. The sample entropy threshold is set to 0.1. When the sample entropy of an IMF is greater than 0.1, the component is considered to be primarily noise-dominated; when it is less than 0.1, it is considered to be a useful component.

[0117] Step 3: Using ICEEMDAN decomposition can better separate the signal and noise, but there is still useful signal in the noise component after decomposition. Therefore, the noise component is decomposed by SVD to extract the residual useful signal from the noise component. The useful components obtained after the two decompositions are reconstructed, and finally the comparison diagram of the OTDR signal after noise filtering and the noisy OTDR signal and the reconstruction error between the denoised OTDR signal and the ideal OTDR signal are obtained. Figure 5 and Figure 6 As shown in Figure 2, after the noisy OTDR signal is denoised by ICEEMDAN-SVD, the signal-to-noise ratio is 42.65 and the root mean square error is 0.10.

[0118] Step 4: Construct an OTDR dataset. During this process, we eliminated most similar samples and tried to maintain a balanced distribution of event points. The resulting dataset contained 1042 samples. Common event points in OTDR signal analysis fall into five categories: fiber start (event 1); connector reflection (event 2); splice or bend (event 3); fiber break (event 4); and fiber end (event 5). The dataset was divided into training and test sets in a 7:3 ratio. The distribution of event types in the training and test sets is shown in Table 2.

[0119] Table 2 Sample data distribution

[0120] type Event 1 Event 2 Event 3 Event 4 Event 5 training set 730 207 1633 143 730 Test set 312 95 653 61 312

[0121] Step 5: Convert the OTDR data set to Figure 1 The denoising process is performed according to the process shown in the figure to obtain a denoised OTDR data set. Local signal features are extracted by setting a sliding window with a width of 50 data points, and the step size of each window sliding is set to 20 data points. The complete signal is divided into multiple overlapping small signals, and various signal feature values ​​are extracted within each window, including statistical features, dynamic features, and spectral features. The specific extracted feature types and calculation formulas are shown in Table 3. The window also needs to include parameters related to position information, such as the horizontal coordinates of the window start and end points, and the peak position. After the classification algorithm determines the presence of an event within the window and identifies its type, it outputs the window boundary information and combines it with information from adjacent windows to perform error correction and event fusion processing, thereby improving positioning accuracy.

[0122] Table 3 OTDR signal characteristic values

[0123] Absolute difference#timg# Peak Factor Average value Margin factor #timg# Absolute average#timg# Coefficient of variation Peak Spectrum Center#timg# Slope#timg# Spectrum Width Standard deviation Mean square frequency#timg# Mean square error#timg# Frequency variance#timg# Square root amplitude#timg# RMS frequency#timg# Skewness Frequency standard deviation#timg# Kurtosis Spectral entropy#timg# Peak-to-peak value Peak stability index#timg#

[0124] Step 6: Calculate and normalize the feature dataset processed in Step 5. When a window contains an event, label the corresponding event information in the label column. Label events 1-5 from Step 4 with numbers in the label column. Use the IBES algorithm to optimize the parameters of the TWSVM algorithm. Initial values ​​are set for the parameters and used as individuals in the vulture population to find the optimal value during the algorithm iteration process. The recognition accuracy of each round of the TWSVM algorithm determines whether to update the values ​​of each parameter. At the end of the final iteration, the highest event recognition accuracy achieved by the TWSVM algorithm on the training set is obtained.

[0125] Step 7: The final optimal values ​​of the TWSVM penalty factor and kernel function parameters are passed into the algorithm, and classification experiments are performed on the OTDR dataset after denoising and the OTDR dataset without denoising. The classification accuracy rates are as follows: Figure 8 and Figure 9 shown.

[0126] Step 8: The algorithm output includes key parameters such as the event type, the optical path distance of the event, the signal loss value at that point, and the distance between the previous event points. The positioning results of the spare optical fiber and the comparison optical fiber in the grid's operating optical cable are compared with the optical cable line data within the grid to obtain the positioning error and positioning time, as shown in Table 4.

[0127] Table 4 Final test data

[0128]

[0129] Specific implementation functions of the positioning algorithm: Based on the signal characteristics, the ICEEMDAN-SVD algorithm adaptively performs denoising and extracts event feature information from each part of the signal. The IBES-TWSVM algorithm identifies and classifies each event in the signal and outputs information such as the distance between event points to achieve accurate identification and positioning.

[0130] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0131] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, performs the above-described method. The storage medium may be any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0132] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0133] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other manner. Any person skilled in the art may utilize the above-disclosed technical content to modify or modify the present invention into equivalent embodiments. However, any simple modifications, equivalent variations, and modifications to the above embodiments that do not depart from the technical content of the present invention and are based on the technical essence of the present invention remain within the scope of protection of the present invention.

[0134] The present invention is not limited to the above-mentioned best embodiment. Anyone can derive various other forms of an optical cable fault location method based on adaptive denoising and event identification of OTDR signal characteristics under the guidance of the present invention. All equal changes and modifications made within the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A method for locating optical cable faults based on adaptive denoising and event recognition of OTDR signal characteristics, characterized by: Based on the preset entropy threshold, the modal components of the OTDR signal after adaptive noise decomposition are classified to obtain the noise-dominant component, the mixed component and the signal-dominant component; Performing singular value difference spectrum mutation point detection on the noise dominant component to extract the residual useful signal; Verify the entropy value of the mixed component, and reconstruct the component that meets the entropy value threshold, the signal dominant component and the residual useful signal into a denoised signal; A swarm intelligence algorithm is used to simultaneously optimize the hyperparameters and feature selection subsets of the support vector machine, and the search parameters of the algorithm are dynamically updated according to an exponential decay mechanism; The spatial location information of the fault point is output based on the optimized support vector machine model.

2. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The algorithm used in the adaptive noise decomposition is ICEEMDAN, k is the number of adjustments, and the noise adjustment coefficient β is k satisfy: When k = 1, β1 is proportional to the inverse of the initial signal-to-noise ratio ε0 of the input signal; When k≥2, β k Proportional to the standard deviation of the k-1th order residual.

3. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The entropy value is the sample entropy, the threshold is 0.1, the embedding dimension m=2, and the similarity tolerance r=0.2×signal standard deviation.

4. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The method of performing singular value difference spectrum mutation point detection on the noise-dominant component and extracting the residual useful signal is specifically as follows: constructing the noise component into a Hankel matrix, obtaining the singular value of the noise component through the SVD algorithm, and extracting the residual useful information in the noise component using the singular value difference spectrum method; the number of rows l of the Hankel matrix satisfies: when the signal length n is an even number, l=n / 2; when n is an odd number, l=(n+1) / 2.

5. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The singular value difference spectrum mutation point detection includes: calculating a sequence of adjacent difference values ​​of the singular values ​​arranged in descending order; and taking the point where a negative value first appears in the sequence as a dividing point.

6. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The swarm intelligence algorithm is the vulture optimization algorithm BES, which dynamically adjusts the search parameter IP to meet the following requirements: Among them, α is a preset constant, t is the current number of iterations, T max is the maximum number of iterations.

7. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The support vector machine is a twin support vector machine TWSVM, and the optimized hyperparameters include penalty factors C1, C2 and kernel parameter γ, where C1, C2∈[0.1, 100], γ∈[0.001, 10].

8. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The local features of the denoised signal are extracted using a sliding window, where the window width covers at least one complete event feature and adjacent windows partially overlap.

9. The optical cable fault location method based on OTDR signal characteristic adaptive denoising and event recognition according to claim 1, characterized in that: The local features include time domain statistical features, frequency domain features and nonlinear features.

10. An optical cable fault location system, characterized in that: include: A signal acquisition module configured to acquire an OTDR original signal; A processing module configured to execute the method according to any one of claims 1 to 10; The output module is configured to display the spatial location information of the fault point.

Images