Optical cable fault prediction method, system, device and medium based on multi-source data

Abstract

The application discloses a cable fault prediction method, system, device and medium based on multi-source data, relates to the technical field of optical cables, and the method comprises the following steps: acquiring distributed test data and benchmark test data; based on the distributed test data, a first state response vector and an individual state response matrix are constructed; based on the benchmark test data, a second state response vector and a group state response matrix are constructed; based on the individual state response matrix, a dominant response vector is extracted; based on the group state response matrix, a common health vector is extracted; based on the first state response vector and the dominant response vector, a state deviation index is determined, and an individual deviation degree is determined according to the state deviation index; according to the second state response vector and the common health vector, a group deviation degree is determined; and a fault prediction result is obtained according to the individual deviation degree and the group deviation degree. The application has the effect of improving the accuracy of optical cable fault prediction.

Description

Technical Field

[0001] This application relates to the technical field of optical cables, and in particular to optical cable fault prediction methods, systems, devices and media based on multi-source data. Background Technology

[0002] As the core physical carrier of modern communication networks, the reliability of optical cables directly affects the stability of the entire communication system. Optical cable faults, especially the latent degradation and sudden interruption of optical fibers, can lead to communication service interruptions and cause huge economic losses. Therefore, accurate monitoring of the operating status of optical cables and early prediction of faults are of great practical significance.

[0003] Currently, optical cable monitoring and diagnosis mainly rely on testing instruments such as optical time-domain reflectometers (OTDRs) to obtain state parameters such as fiber attenuation, loss point, and reflection events by analyzing backscattered signals. Most existing mainstream methods are based on data obtained under single, fixed test conditions, using static empirical thresholds for fault diagnosis and alarms. These methods are increasingly revealing their inherent limitations when facing complex optical cable operating environments and diverse fault modes.

[0004] First, the testing perspective is limited, making it difficult to comprehensively capture fault signs. The actual fault modes of optical cables are complex and diverse, potentially caused by a variety of factors such as stress micro-bending, aging, joint deterioration, or external environmental corrosion. Different potential faults exhibit varying forms and degrees of severity under different testing conditions (e.g., different test wavelengths, signal power, or modulation modes). Existing technologies rely on data from a single testing configuration, failing to acquire sufficient dimensional information and easily overlooking early, subtle fault characteristics that only manifest under specific testing conditions, leading to delayed warnings or missed diagnoses.

[0005] Secondly, static threshold determination has poor adaptability and a high risk of false alarms and missed alarms. Existing methods typically use a uniform, preset static threshold to determine whether all optical fibers are in normal condition. This method ignores the inherent differences in characteristics between different optical fibers due to variations in manufacturing batches, laying paths, stress levels, and historical loads. For optical fibers whose performance parameters are already at the boundary, false alarms are easily generated due to normal fluctuations; while for optical fibers with slow performance and gradual degradation, their parameter values ​​may remain within the general threshold range for a long period, leading to missed alarms and failing to achieve true early prediction.

[0006] Therefore, there is an urgent need in this field for a fiber optic cable fault prediction technology that can overcome the above-mentioned defects. It can comprehensively perceive the fiber optic status from multiple testing dimensions and establish a more intelligent and adaptive health assessment mechanism to achieve more accurate and forward-looking prediction of potential faults. Summary of the Invention

[0007] To improve the accuracy of optical cable fault prediction, this application provides a method, system, device, and medium for optical cable fault prediction based on multi-source data.

[0008] Firstly, this application provides a method for optical cable fault prediction based on multi-source data, employing the following technical solution: Optical cable fault prediction methods based on multi-source data include: The distributed test data of each optical fiber in the target optical cable under various test configurations and the benchmark test data under the benchmark test parameters are obtained. Each test configuration is obtained by changing a single test parameter. For each optical fiber and each test configuration, based on the distributed test data, multiple first state response vectors for a single optical fiber and a single test configuration are constructed and aggregated into an individual state response matrix for the single optical fiber and the single test configuration; and, based on the benchmark test data, a second state response vector corresponding to the single optical fiber is constructed and aggregated into a group state response matrix. Based on the individual state response matrices, the dominant response vectors of each optical fiber under different test configurations are extracted; and based on the group state response matrix, the common health vectors of each optical fiber under the benchmark test parameters are extracted. For each optical fiber, based on the first state response vector of the optical fiber under each test configuration and the corresponding dominant response vector, multiple state deviation indicators are determined, and the individual deviation degree is determined according to the state deviation indicators; and the group deviation degree is determined according to the second state response vector of the optical fiber and the common health vector. By jointly analyzing the individual deviation and the group deviation, the fault prediction results for each optical fiber are obtained.

[0009] By adopting the above technical solution, distributed test data of each optical fiber in the target optical cable under various test configurations and benchmark test data under benchmark test parameters are first obtained. Each test configuration is obtained by changing a single test parameter. Then, for each optical fiber and each test configuration, based on the distributed test data, multiple first state response vectors for a single optical fiber and a single test configuration are constructed and aggregated into an individual state response matrix for a single optical fiber and a single test configuration. Based on the benchmark test data, a second state response vector corresponding to a single optical fiber is constructed and aggregated into a group state response matrix. Then, based on each individual state response matrix, the dominant response vector of each optical fiber under different test configurations is extracted. Based on the group state response matrix, the common health vector of each optical fiber under the benchmark test parameters is extracted. Then, for each optical fiber, based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration, multiple state deviation indicators are determined, and the individual deviation degree is determined according to the state deviation indicators. The group deviation degree is determined according to the second state response vector and the common health vector of the optical fiber. Finally, the individual deviation degree and the group deviation degree are jointly analyzed to obtain the fault prediction results of each optical fiber. By using the above methods, a dynamic and adaptive health assessment system is established, which effectively overcomes the limitations of traditional static threshold criteria. While reducing the false alarm rate, it enhances the detection capability of progressive performance degradation, realizes multi-dimensional panoramic perception of the health status of optical cables, improves the sensitivity and reliability of early fault identification, helps reduce operation and maintenance costs, and ensures the continuous and stable operation of communication networks.

[0010] Optionally, the step of constructing multiple first-state response vectors for a single optical fiber and a single test configuration based on the distributed test data includes: Based on the distributed test data, multiple sets of test data for a single optical fiber under a single test configuration are extracted. The single test configuration includes multiple sets of test parameters, and each set of test data is data obtained by performing multiple tests under a single set of test parameters. For each set of test data, state features are extracted from each single test data under that set of test data to obtain multiple sets of first state feature values, wherein the state feature values ​​include at least one of attenuation coefficient, total link loss, signal-to-noise ratio, and reflection coefficient; and the average value of each set of first state feature values ​​is calculated to obtain the first average value of each state feature. Based on the first average value of each item, normalization is performed to generate the first state response vector of the single optical fiber under the single set of test parameters.

[0011] By adopting the above technical solution, in order to construct the first state response vector, multiple sets of test data for a single optical fiber under a single test configuration are extracted based on distributed test data. Each single test configuration includes multiple sets of test parameters, and each set of test data is obtained by performing multiple tests under those parameters. Then, for each set of test data, state features are extracted from each individual test data to obtain multiple sets of first state feature values. These state feature values ​​include at least one of attenuation coefficient, total link loss, signal-to-noise ratio, and reflection coefficient. Furthermore, the first state feature values ​​are averaged to obtain the first average value corresponding to each state feature. Then, normalization is performed based on these first average values ​​to generate the first state response vector for a single optical fiber under the single set of test parameters.

[0012] Optionally, the step of constructing the second state response vector corresponding to the single optical fiber based on the benchmark test data includes: Based on the benchmark test data, extract a single set of benchmark test data obtained by performing multiple tests on a single optical fiber under the benchmark test parameters. State features are extracted from each single benchmark test data under the single set of benchmark test data to obtain multiple sets of second state feature values; and the average value of each set of second state feature values ​​is calculated to obtain the second average value of each state feature. Normalization is performed based on the second average value of each item to generate the second state response vector of the single optical fiber under the benchmark test parameters.

[0013] By adopting the above technical solution, in order to construct the second state response vector, based on the benchmark test data, a single set of benchmark test data is extracted from multiple tests conducted on a single optical fiber under the benchmark test parameters. Then, state features are extracted from each single benchmark test data under the single set of benchmark test data to obtain multiple sets of second state feature values. Furthermore, the average values ​​of each set of second state feature values ​​are calculated to obtain the second average value corresponding to each state feature. Then, normalization is performed based on the second average value to generate the second state response vector corresponding to a single optical fiber under the benchmark test parameters.

[0014] Optionally, the step of extracting the dominant response vector of each optical fiber under different test configurations based on the individual state response matrices includes: For each individual state response matrix, the individual state response matrix is ​​decomposed into a nonnegative matrix to obtain a corresponding first basis matrix and a first coefficient matrix. Each column vector of the first basis matrix represents a potential response pattern, and each row vector of the first coefficient matrix represents the association weight between the response pattern of a single test and the potential response pattern. Determine the first L2 norm corresponding to each column vector in the first coefficient matrix, and use the first L2 norm as the first contribution of the corresponding column vector in the first basis matrix; Sort the column vectors of the first basis matrix in descending order of the first contribution to form a first set of column vectors; Based on a preset first cumulative contribution threshold, the first k column vectors are selected from the first set of column vectors as the dominant response vectors of the corresponding single optical fiber under the corresponding single test configuration.

[0015] By adopting the above technical solution, in order to extract the dominant response vector, for each individual state response matrix, non-negative matrix decomposition is performed on the individual state response matrix to obtain the corresponding first basis matrix and first coefficient matrix. Each column vector of the first basis matrix represents a potential response mode, and each row vector of the first coefficient matrix represents the association weight between the response mode of a single test and the potential response mode. Then, the first L2 norm corresponding to each column vector in the first coefficient matrix is ​​determined, and the first L2 norm is used as the first contribution of the corresponding column vector in the first basis matrix. Then, the column vectors of the first basis matrix are sorted in descending order of the first contribution to form a first column vector set. Then, based on a preset first cumulative contribution threshold, the first k column vectors are selected from the first column vector set as the dominant response vector of the corresponding single optical fiber under the corresponding single test configuration.

[0016] Optionally, the step of extracting the common health vector of each optical fiber under the benchmark test parameters based on the group state response matrix includes: The group state response matrix is ​​decomposed into a nonnegative matrix to obtain the corresponding second basis matrix and second coefficient matrix. Each column vector of the second basis matrix represents a potential health mode, and each row vector of the second coefficient matrix represents the correlation weight between the health state of a single optical fiber and the potential health mode. Determine the second L2 norm corresponding to each column vector in the second coefficient matrix, and use the second L2 norm as the second contribution of the corresponding column vector in the second basis matrix; The column vectors of the second basis matrix are sorted in descending order of the second contribution to form a second set of column vectors; Based on a preset second cumulative contribution threshold, the first m column vectors are selected from the second set of column vectors as the common health vectors of each optical fiber under the benchmark test parameters.

[0017] By adopting the above technical solution, in order to extract the common health vector, the population state response matrix is ​​decomposed into a non-negative matrix to obtain the corresponding second basis matrix and second coefficient matrix. Each column vector of the second basis matrix represents a potential health mode, and each row vector of the second coefficient matrix represents the correlation weight between the health status of a single optical fiber and the potential health mode. Then, the second L2 norm corresponding to each column vector in the second coefficient matrix is ​​determined, and the second L2 norm is used as the second contribution of the corresponding column vector in the second basis matrix. Then, the column vectors of the second basis matrix are sorted in descending order of the second contribution to form a second column vector set. Then, based on the preset second cumulative contribution threshold, the first m column vectors are selected from the second column vector set as the common health vector of each optical fiber under the benchmark test parameters.

[0018] Optionally, the step of determining multiple state deviation indices based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration includes: For each first state response vector of the optical fiber under any test configuration, a first feature subspace is spanned based on the dominant response vector corresponding to that test configuration. The first state response vector is projected onto the first feature subspace to obtain the first projection vector; Calculate the difference between the first state response vector and the first projection vector to obtain the first residual vector; The L2 norm of the first residual vector is determined as the state deviation index corresponding to the first state response vector.

[0019] By adopting the above technical solution, in order to determine multiple state deviation indices, for each first state response vector of the optical fiber under any test configuration, a first feature subspace is spanned based on the dominant response vector corresponding to the test configuration. Then, the first state response vector is projected onto the first feature subspace to obtain a first projection vector. Then, the difference between the first state response vector and the first projection vector is calculated to obtain a first residual vector. Then, the L2 norm of the first residual vector is determined as the state deviation index corresponding to the first state response vector.

[0020] Optionally, the step of determining the population deviation based on the second state response vector of the optical fiber and the common health vector includes: Based on the aforementioned common health vector, a second feature subspace is formed; Calculate the covariance matrix of the group state response matrix; The second state response vector of the optical fiber is projected onto the second feature subspace to obtain the second projection vector; Based on the covariance matrix, the Mahalanobis distance of the second projection vector in the second feature subspace is calculated, and the Mahalanobis distance is used as the population deviation of the optical fiber.

[0021] By adopting the above technical solution, in order to determine the population deviation, a second feature subspace is spanned based on the common health vector, and then the covariance matrix of the population state response matrix is ​​calculated. Then, the second state response vector of the optical fiber is projected onto the second feature subspace to obtain the second projection vector. Then, based on the covariance matrix, the Mahalanobis distance of the second projection vector in the second feature subspace is calculated, and the Mahalanobis distance is used as the population deviation of the optical fiber.

[0022] Secondly, this application also provides an optical cable fault prediction system based on multi-source data, which adopts the following technical solution: A fiber optic cable fault prediction system based on multi-source data includes: The data acquisition module is used to acquire distributed test data of each optical fiber in the target optical cable under various test configurations and benchmark test data under benchmark test parameters. Each test configuration is obtained by changing a single test parameter. The feature construction module is used to construct multiple first state response vectors for a single optical fiber and a single test configuration based on the distributed test data for each optical fiber and each test configuration, and aggregate them into an individual state response matrix for the single optical fiber and the single test configuration; and to construct a second state response vector corresponding to the single optical fiber based on the benchmark test data, and aggregate it into a group state response matrix. The feature extraction module is used to extract the dominant response vector of each optical fiber under different test configurations based on the individual state response matrices; and to extract the common health vector of each optical fiber under the benchmark test parameters based on the group state response matrix. The deviation determination module is used to determine multiple state deviation indicators for each optical fiber based on the first state response vector of the optical fiber under each test configuration and the corresponding dominant response vector, and to determine the individual deviation degree according to the state deviation indicators; and to determine the group deviation degree according to the second state response vector of the optical fiber and the common health vector. The result generation module is used to perform joint analysis of the individual deviation and the group deviation to obtain the fault prediction results of each optical fiber.

[0023] Thirdly, this application also provides a computer device, which adopts the following technical solution: A computer device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the method described in the first aspect.

[0024] Fourthly, this application also provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the method described in the first aspect.

[0025] In summary, this application includes at least the following beneficial technical effects: First, distributed test data and benchmark test data under benchmark test parameters are obtained for each optical fiber in the target optical cable under various test configurations. Each test configuration is obtained by changing a single test parameter. Then, for each optical fiber and each test configuration, multiple first state response vectors for a single optical fiber and a single test configuration are constructed based on the distributed test data and aggregated into an individual state response matrix for a single optical fiber and a single test configuration. Based on the benchmark test data, a second state response vector corresponding to a single optical fiber is constructed and aggregated into a group state response matrix. Then, based on each individual state response matrix, the dominant response vector of each optical fiber under different test configurations is extracted. Based on the group state response matrix, the common health vector of each optical fiber under benchmark test parameters is extracted. Then, for each optical fiber, multiple state deviation indicators are determined based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration. The individual deviation degree is determined based on the state deviation indicators. The group deviation degree is determined based on the second state response vector and the common health vector of the optical fiber. Finally, the individual deviation degree and the group deviation degree are jointly analyzed to obtain the fault prediction results of each optical fiber. By using the above methods, a dynamic and adaptive health assessment system is established, which effectively overcomes the limitations of traditional static threshold criteria. While reducing the false alarm rate, it enhances the detection capability of progressive performance degradation, realizes multi-dimensional panoramic perception of the health status of optical cables, improves the sensitivity and reliability of early fault identification, helps reduce operation and maintenance costs, and ensures the continuous and stable operation of communication networks. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall process of an embodiment of this application.

[0027] Figure 2 This is a schematic diagram of the system structure of this application.

[0028] Figure 3 This is a structural block diagram of the computer device described in this application. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0030] This application discloses a method for predicting optical cable faults based on multi-source data.

[0031] Reference Figure 1 A method for optical cable fault prediction based on multi-source data includes: Step S11: Obtain distributed test data of each optical fiber in the target optical cable under various test configurations and benchmark test data under benchmark test parameters. Each test configuration is obtained by changing a single test parameter.

[0032] It should be noted that, through step S11, optical fiber status data is collected from multiple dimensions. Multiple test configurations refer to a series of test environments formed by systematically and independently changing single test parameters such as test wavelength, signal power, and modulation frequency. This design enables the observation of the optical fiber's response behavior under different physical conditions (e.g., different wavelengths have different sensitivities to microbending loss), thereby obtaining richer status information than a single test condition. At the same time, the benchmark test data collected under the reference test parameters (i.e., a set of standard, conventional test conditions) is used to establish a unified reference benchmark for the optical fiber under normal conditions.

[0033] Step S12: For each optical fiber and each test configuration, based on distributed test data, construct multiple first state response vectors for a single optical fiber and a single test configuration, and aggregate them into an individual state response matrix for a single optical fiber and a single test configuration; and based on benchmark test data, construct a second state response vector corresponding to a single optical fiber, and aggregate it into a group state response matrix.

[0034] It should be noted that step S12 involves structuring and characterizing the raw data. For non-benchmark test data, a first state response vector is constructed for each set of test results (such as power tests with multiple variations at a certain wavelength) of a single optical fiber under a specific test configuration. This vector is a normalized comprehensive representation of multiple state characteristics (such as attenuation, signal-to-noise ratio, etc.). These vectors are aggregated into an individual state response matrix, which aims to characterize the set of all observed state modes of the optical fiber under this specific test environment. For benchmark test data, a second state response vector is constructed for each optical fiber to characterize its health state under standard conditions. The benchmark state vectors of all optical fibers are aggregated into a group state response matrix, which constitutes a sample space reflecting the distribution of the health state of the entire optical cable under standard conditions.

[0035] Step S13: Based on the individual state response matrix, extract the dominant response vector of each optical fiber under different test configurations; and based on the group state response matrix, extract the common health vector of each optical fiber under the benchmark test parameters.

[0036] It should be noted that the core health pattern is extracted through step S14. By performing nonnegative matrix decomposition on the state response matrix of each individual fiber, the dominant response vector hidden behind multiple test data can be extracted, which represents the most important and typical behavior pattern of the fiber under the corresponding test configuration. This is equivalent to establishing a personalized health fingerprint for the fiber under different test scenarios. Similarly, by decomposing the group state response matrix, a common health vector that represents the common health characteristics of most fibers in the entire optical cable under the baseline conditions can be extracted, which constitutes a group-level health reference standard.

[0037] Step S14: For each optical fiber, based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration, determine multiple state deviation indicators, and determine the individual deviation degree according to the state deviation indicators; and determine the group deviation degree according to the second state response vector and the common health vector of the optical fiber.

[0038] It should be noted that step S14 involves a quantitative assessment of the deviation. For a single optical fiber, the first state response vector obtained from each actual measurement under various test configurations is compared with the locally stored dominant response vector (i.e., its health fingerprint) under that configuration. The degree of deviation is calculated, resulting in multiple state deviation indices. These indices are then combined (e.g., taking the maximum value, average value, etc.) to ultimately form an "individual deviation" scalar value that reflects the overall deviation of the current state of the optical fiber from its own historical health pattern. The "group deviation" measures the degree of abnormality of the optical fiber relative to the entire healthy group of optical cables at the current moment by calculating the difference between the second state response vector of the optical fiber under the baseline conditions and the common health vector extracted from the group (e.g., Mahalanobis distance).

[0039] Step S15: Perform joint analysis of individual deviation and group deviation to obtain the fault prediction results for each optical fiber.

[0040] It should be noted that step S15 is the final decision-making stage. A single deviation indicator may lead to misjudgment. For example, the individual deviation of a fiber may temporarily increase due to environmental noise, but remain normal within the group; conversely, it may appear slightly unusual within the group, but its behavior pattern is very stable. By jointly analyzing individual and group deviations (e.g., by setting two-dimensional decision boundaries, using weighted scoring, or logistic regression fusion strategies), cross-validation and information complementarity can be achieved. This fusion strategy can more comprehensively and reliably assess the health risks of optical fibers, ultimately outputting a more accurate prediction of whether each fiber is in a pre-fault state, thus achieving early warning.

[0041] In the above implementation, distributed test data and benchmark test data under benchmark test parameters are first obtained for each optical fiber in the target optical cable under various test configurations. Each test configuration is obtained by changing a single test parameter. Then, for each optical fiber and each test configuration, multiple first state response vectors for a single optical fiber and a single test configuration are constructed based on the distributed test data and aggregated into an individual state response matrix for a single optical fiber and a single test configuration. Based on the benchmark test data, a second state response vector corresponding to a single optical fiber is constructed and aggregated into a group state response matrix. Then, based on each individual state response matrix, the dominant response vector of each optical fiber under different test configurations is extracted. Based on the group state response matrix, the common health vector of each optical fiber under the benchmark test parameters is extracted. Then, for each optical fiber, multiple state deviation indicators are determined based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration. The individual deviation degree is determined based on the state deviation indicators. The group deviation degree is determined based on the second state response vector and the common health vector of the optical fiber. Finally, the individual deviation degree and the group deviation degree are jointly analyzed to obtain the fault prediction results of each optical fiber. By using the above methods, a dynamic and adaptive health assessment system is established, which effectively overcomes the limitations of traditional static threshold criteria. While reducing the false alarm rate, it enhances the detection capability of progressive performance degradation, realizes multi-dimensional panoramic perception of the health status of optical cables, improves the sensitivity and reliability of early fault identification, helps reduce operation and maintenance costs, and ensures the continuous and stable operation of communication networks.

[0042] As a further implementation of the method, the step of constructing multiple first-state response vectors for a single optical fiber and a single test configuration based on distributed test data includes: Step S21: Based on the distributed test data, extract multiple sets of test data for a single optical fiber under a single test configuration. The single test configuration includes multiple sets of test parameters, and each set of test data is the data obtained by performing multiple tests under a single set of test parameters.

[0043] Step S22: For each set of test data, extract state features from the single test data under the set of test data to obtain multiple sets of first state feature values, wherein the state feature values ​​include at least one of attenuation coefficient, total link loss, signal-to-noise ratio, and reflection coefficient; and calculate the average of each set of first state feature values ​​to obtain the first average value of each state feature.

[0044] Step S23: Normalize based on the first average value of each item to generate the first state response vector of a single optical fiber under a single set of test parameters.

[0045] It should be noted that from steps S21 to S23, through multiple layers of data processing, the original distributed test data is transformed into a standardized state response vector. Specifically, firstly, step S21 performs structured grouping of the test data, establishing a systematic data organization architecture from test configuration to specific test parameters. Step S22, through dual processing of feature extraction and average calculation, effectively suppresses measurement random errors while retaining key state information, obtaining a feature set that can represent the stable state of the optical fiber under a single set of test parameters. The normalization process in step S23 eliminates the differences in the dimensions and orders of magnitude of different state feature values, making the final generated first state response vector have unified mathematical properties and comparability. Thus, a complete data processing flow is constructed, providing high-quality, standardized input data for subsequent steps and improving the stability and reliability of the entire prediction system.

[0046] In the above implementation, to construct the first state response vector, multiple sets of test data for a single optical fiber under a single test configuration are extracted based on distributed test data. Each single test configuration includes multiple sets of test parameters, and each set of test data is obtained by performing multiple tests under those parameters. Then, for each set of test data, state features are extracted from each individual test data to obtain multiple sets of first state feature values. These state feature values ​​include at least one of attenuation coefficient, total link loss, signal-to-noise ratio, and reflection coefficient. Furthermore, the first state feature values ​​are averaged to obtain the first average value corresponding to each state feature. Then, normalization is performed based on these first average values ​​to generate the first state response vector for a single optical fiber under the single set of test parameters.

[0047] As a further implementation of the method, the step of constructing a second state response vector corresponding to a single optical fiber based on benchmark test data includes: Step S31: Based on the benchmark test data, extract a single set of benchmark test data obtained by performing multiple tests on a single optical fiber under the benchmark test parameters.

[0048] Step S32: Extract state features from the single benchmark test data under the single set of benchmark test data to obtain multiple sets of second state feature values; and calculate the average of the second state feature values ​​of each set to obtain the second average value of each state feature.

[0049] Step S33: Normalize based on the second average value of each item to generate the second state response vector of a single optical fiber under the benchmark test parameters.

[0050] It should be noted that steps S31 to S33 complete the system transformation process from benchmark test data to standardized state vectors. Specifically, step S31 ensures the uniformity and comparability of data sources. Step S32, through dual processing of feature extraction and averaging, retains key features that reflect the essential state of the optical fiber while effectively eliminating random fluctuations from single measurements, obtaining characteristic average values ​​that represent the stability performance of the optical fiber under the benchmark state. The normalization process in step S33 further eliminates the influence of differences in the dimensions of various feature parameters, making the final generated second-state response vector have unified mathematical properties and scale standards. This constructs a feature expression system for the optical fiber benchmark state, providing a standardized and comparable state characterization basis for subsequent group health benchmark extraction and deviation analysis, and improving the reliability and accuracy of the entire fault prediction method.

[0051] In the above implementation, to construct the second state response vector, based on benchmark test data, a single set of benchmark test data is extracted from multiple tests conducted on a single optical fiber under benchmark test parameters. Then, state features are extracted from each single benchmark test data set to obtain multiple sets of second state feature values. Furthermore, the average values ​​of each set of second state feature values ​​are calculated to obtain the second average value corresponding to each state feature. Then, normalization is performed based on the second average value to generate the second state response vector corresponding to a single optical fiber under the benchmark test parameters.

[0052] As a further implementation of the method, the step of extracting the dominant response vector of each optical fiber under different test configurations based on the individual state response matrix includes: Step S41: For each individual state response matrix, perform non-negative matrix decomposition on the individual state response matrix to obtain the corresponding first basis matrix and first coefficient matrix. Each column vector of the first basis matrix represents a potential response pattern, and each row vector of the first coefficient matrix represents the association weight between the response pattern of a single test and the potential response pattern.

[0053] Step S42: Determine the first L2 norm corresponding to each column vector in the first coefficient matrix, and use the first L2 norm as the first contribution of the corresponding column vector in the first basis matrix.

[0054] Step S43: Sort the column vectors of the first basis matrix in descending order of the first contribution to form the first set of column vectors.

[0055] Step S44: Based on the preset first cumulative contribution threshold, select the first k column vectors from the first column vector set as the dominant response vectors of the corresponding single optical fiber under the corresponding single test configuration.

[0056] It should be noted that from steps S41 to S44, through a systematic matrix decomposition and pattern selection process, the essential features are extracted from complex individual state data. Specifically, step S41 uses non-negative matrix decomposition to decompose the individual state response matrix into a first basis matrix and a first coefficient matrix with clear physical meaning. The column vectors of the basis matrix represent the potential response modes that the optical fiber may have under various test conditions, while the coefficient matrix quantifies the degree of performance of these potential modes in specific tests. Step S42 evaluates the contribution of each potential mode by calculating the L2 norm of the column vectors of the first coefficient matrix, providing a quantitative basis for ranking the importance of the modes. Steps S43 and S44, through a ranking and threshold selection mechanism, identify the most representative dominant response modes from among the many potential modes. This series of processes not only achieves data dimensionality reduction and feature enhancement, but more importantly, it can uncover the dominant response features that characterize the essential properties of the optical fiber.

[0057] In the above implementation, in order to extract the dominant response vector, for each individual state response matrix, nonnegative matrix decomposition is performed to obtain the corresponding first basis matrix and first coefficient matrix. Each column vector of the first basis matrix represents a potential response mode, and each row vector of the first coefficient matrix represents the association weight between the response mode of a single test and the potential response mode. Then, the first L2 norm corresponding to each column vector in the first coefficient matrix is ​​determined, and the first L2 norm is used as the first contribution of the corresponding column vector in the first basis matrix. Then, the column vectors of the first basis matrix are sorted in descending order of the first contribution to form a first column vector set. Then, based on a preset first cumulative contribution threshold, the first k column vectors are selected from the first column vector set as the dominant response vector of the corresponding single optical fiber under the corresponding single test configuration.

[0058] As a further implementation of the method, the step of extracting the common health vector of each optical fiber under benchmark test parameters based on the group state response matrix includes: Step S51: Perform nonnegative matrix decomposition on the group state response matrix to obtain the corresponding second basis matrix and second coefficient matrix. Each column vector of the second basis matrix represents a potential health mode, and each row vector of the second coefficient matrix represents the correlation weight between the health state of a single optical fiber and the potential health mode.

[0059] Step S52: Determine the second L2 norm corresponding to each column vector in the second coefficient matrix, and use the second L2 norm as the second contribution of the corresponding column vector in the second basis matrix.

[0060] Step S53: Sort the column vectors of the second basis matrix in descending order of the second contribution to form the second column vector set.

[0061] Step S54: Based on the preset second cumulative contribution threshold, select the first m column vectors from the second column vector set as the common health vectors of each optical fiber under the benchmark test parameters.

[0062] It should be noted that from steps S51 to S54, a feature system capable of characterizing the health benchmark of the entire optical cable was established through the system's matrix decomposition and pattern screening process. Specifically, step S51 performs non-negative matrix decomposition on the group state response matrix, decoupling the multidimensional group data into a second basis matrix and a second coefficient matrix with clear physical meaning. The column vectors of the basis matrix represent the potential health patterns shared by the optical cable group, while the coefficient matrix quantifies the degree of correlation between each fiber and these common patterns. Step S52 provides a quantitative indicator for evaluating the group contribution of each potential health pattern by calculating the L2 norm of the column vectors of the second coefficient matrix. Steps S53 and S54 identify the most representative common health patterns from numerous potential patterns through a ranking and threshold screening mechanism. This processing flow not only effectively mines the essential health characteristics of the optical cable system but also ensures the representativeness and reliability of the extracted common health vectors through contribution ranking.

[0063] In the above implementation, in order to extract the common health vector, the population state response matrix is ​​decomposed into a nonnegative matrix to obtain the corresponding second basis matrix and second coefficient matrix. Each column vector of the second basis matrix represents a potential health mode, and each row vector of the second coefficient matrix represents the correlation weight between the health state of a single optical fiber and the potential health mode. Then, the second L2 norm corresponding to each column vector in the second coefficient matrix is ​​determined, and the second L2 norm is used as the second contribution of the corresponding column vector in the second basis matrix. Then, the column vectors of the second basis matrix are sorted in descending order of the second contribution to form a second column vector set. Then, based on a preset second cumulative contribution threshold, the first m column vectors are selected from the second column vector set as the common health vectors of each optical fiber under the benchmark test parameters.

[0064] As a further implementation of the method, the step of determining multiple state deviation indices based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration includes: Step S61: For each first state response vector of the optical fiber under any test configuration, a first feature subspace is spanned based on the dominant response vector corresponding to that test configuration.

[0065] Step S62: Project the first state response vector onto the first feature subspace to obtain the first projection vector.

[0066] Step S63: Calculate the difference between the first state response vector and the first projection vector to obtain the first residual vector.

[0067] Step S64: The L2 norm of the first residual vector is determined as the state deviation index corresponding to the first state response vector.

[0068] It should be noted that from steps S61 to S64, an anomaly detection mechanism based on subspace projection is constructed to quantify the deviation between the actual state of the optical fiber and the health benchmark. Specifically, step S61 establishes a reference space characterizing the health state of the optical fiber under a specific test configuration by using the first feature subspace spanned by the dominant response vector; step S62 projects the real-time acquired first state response vector onto this health subspace to obtain its best approximate representation under the health mode; step S63 effectively separates the anomalous components in the signal that cannot be explained by the health mode by calculating the residual between the original vector and the projected vector; step S64 uses the L2 norm of the residual vector as a state deviation index. This index is highly sensitive to anomalous modes and can accurately reflect the deviation of the optical fiber state from the health benchmark, thereby realizing the conversion from a multi-dimensional state vector to a single-dimensional deviation index, providing a quantitative basis for subsequent comprehensive deviation analysis, and ensuring the sensitivity and reliability of fault detection.

[0069] In the above implementation, in order to determine multiple state deviation indices, for each first state response vector of the optical fiber under any test configuration, a first feature subspace is spanned based on the dominant response vector corresponding to the test configuration. Then, the first state response vector is projected onto the first feature subspace to obtain a first projection vector. Then, the difference between the first state response vector and the first projection vector is calculated to obtain a first residual vector. Then, the L2 norm of the first residual vector is determined as the state deviation index corresponding to the first state response vector.

[0070] As a further implementation of the method, the step of determining the population deviation based on the second state response vector and the common health vector of the optical fiber includes: Step S71: Based on the common health vector, expand the second feature subspace.

[0071] Step S72: Calculate the covariance matrix of the group state response matrix.

[0072] Step S73: Project the second state response vector of the optical fiber onto the second feature subspace to obtain the second projection vector.

[0073] Step S74: Based on the covariance matrix, calculate the Mahalanobis distance of the second projection vector in the second feature subspace, and use the Mahalanobis distance as the population deviation of the optical fiber.

[0074] It should be noted that from steps S71 to S74, a group health status assessment mechanism based on multivariate statistics is constructed. Specifically, step S71 establishes a reference space characterizing the health benchmark of the optical cable group through the second feature subspace spanned by the common health vector; step S72 calculates the covariance matrix of the group state response matrix, which contains the correlation and variability information between various health features; step S73 projects the second state response vector of a single optical fiber onto the group health subspace to obtain its representation vector under the group health mode; step S74 obtains a group deviation index that considers both the correlation between features and the shape of data distribution by calculating the Mahalanobis distance of the second projection vector in the feature subspace considering the covariance structure. This index can accurately reflect the degree of statistical anomaly of the tested optical fiber relative to the entire optical cable group and has higher detection sensitivity for abnormal states in the group.

[0075] In the above implementation, in order to determine the population deviation, a second feature subspace is spanned based on the common health vector, and then the covariance matrix of the population state response matrix is ​​calculated. Then, the second state response vector of the optical fiber is projected onto the second feature subspace to obtain the second projection vector. Then, based on the covariance matrix, the Mahalanobis distance of the second projection vector in the second feature subspace is calculated, and the Mahalanobis distance is used as the population deviation of the optical fiber.

[0076] This application also discloses an optical cable fault prediction system based on multi-source data.

[0077] refer to Figure 2 A fiber optic cable fault prediction system based on multi-source data includes: The data acquisition module is used to acquire distributed test data of each optical fiber in the target optical cable under various test configurations and benchmark test data under benchmark test parameters. Each test configuration is obtained by changing a single test parameter. The feature construction module is used to construct multiple first state response vectors for a single fiber and a single test configuration based on distributed test data for each fiber and each test configuration, and aggregate them into an individual state response matrix for a single fiber and a single test configuration; and to construct a second state response vector corresponding to a single fiber based on benchmark test data, and aggregate it into a group state response matrix. The feature extraction module is used to extract the dominant response vector of each optical fiber under different test configurations based on the individual state response matrix; and to extract the common health vector of each optical fiber under the benchmark test parameters based on the group state response matrix. The deviation determination module is used to determine multiple state deviation indices for each optical fiber based on the first state response vector and the corresponding dominant response vector of the optical fiber under each test configuration, and to determine the individual deviation degree based on the state deviation indices; and to determine the group deviation degree based on the second state response vector and the common health vector of the optical fiber. The results generation module is used to perform joint analysis of individual deviation and group deviation to obtain the fault prediction results for each optical fiber.

[0078] The optical cable fault prediction system based on multi-source data of the present invention can implement any of the methods of optical cable fault prediction based on multi-source data, and the specific working process of the optical cable fault prediction system based on multi-source data of the present invention can refer to the corresponding process in the above-mentioned optical cable fault prediction method based on multi-source data.

[0079] This application also discloses a computer device.

[0080] refer to Figure 3 A computer device includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement any of the above-described methods for optical cable fault prediction based on multi-source data.

[0081] This application also discloses a computer-readable storage medium.

[0082] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed any of the above-described methods for optical cable fault prediction based on multi-source data.

[0083] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0084] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for predicting optical cable faults based on multi-source data, characterized in that, include: The distributed test data of each optical fiber in the target optical cable under various test configurations and the benchmark test data under the benchmark test parameters are obtained. Each test configuration is obtained by changing a single test parameter. For each optical fiber and each test configuration, based on the distributed test data, multiple first state response vectors for a single optical fiber and a single test configuration are constructed and aggregated into an individual state response matrix for the single optical fiber and the single test configuration; and, based on the benchmark test data, a second state response vector corresponding to the single optical fiber is constructed and aggregated into a group state response matrix. Based on the individual state response matrices, the dominant response vectors of each optical fiber under different test configurations are extracted; and based on the group state response matrix, the common health vectors of each optical fiber under the benchmark test parameters are extracted. For each optical fiber, based on the first state response vector of the optical fiber under each test configuration and the corresponding dominant response vector, multiple state deviation indicators are determined, and the individual deviation degree is determined according to the state deviation indicators; and the group deviation degree is determined according to the second state response vector of the optical fiber and the common health vector. By jointly analyzing the individual deviation and the group deviation, the fault prediction results for each optical fiber are obtained.

2. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of constructing multiple first-state response vectors for a single optical fiber and a single test configuration based on the distributed test data includes: Based on the distributed test data, multiple sets of test data for a single optical fiber under a single test configuration are extracted. The single test configuration includes multiple sets of test parameters, and each set of test data is data obtained by performing multiple tests under a single set of test parameters. For each set of test data, state features are extracted from each single test data under that set of test data to obtain multiple sets of first state feature values, wherein the state feature values ​​include at least one of attenuation coefficient, total link loss, signal-to-noise ratio, and reflection coefficient; and the average value of each set of first state feature values ​​is calculated to obtain the first average value of each state feature. Based on the first average value of each item, normalization is performed to generate the first state response vector of the single optical fiber under the single set of test parameters.

3. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of constructing the second state response vector corresponding to the single optical fiber based on the benchmark test data includes: Based on the benchmark test data, extract a single set of benchmark test data obtained by performing multiple tests on a single optical fiber under the benchmark test parameters. State features are extracted from each single benchmark test data under the single set of benchmark test data to obtain multiple sets of second state feature values; and the average value of each set of second state feature values ​​is calculated to obtain the second average value of each state feature. Normalization is performed based on the second average value of each item to generate the second state response vector of the single optical fiber under the benchmark test parameters.

4. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of extracting the dominant response vector of each optical fiber under different test configurations based on the individual state response matrices includes: For each individual state response matrix, the individual state response matrix is ​​decomposed into a nonnegative matrix to obtain a corresponding first basis matrix and a first coefficient matrix. Each column vector of the first basis matrix represents a potential response pattern, and each row vector of the first coefficient matrix represents the association weight between the response pattern of a single test and the potential response pattern. Determine the first L2 norm corresponding to each column vector in the first coefficient matrix, and use the first L2 norm as the first contribution of the corresponding column vector in the first basis matrix; Sort the column vectors of the first basis matrix in descending order of the first contribution to form a first set of column vectors; Based on a preset first cumulative contribution threshold, the first k column vectors are selected from the first set of column vectors as the dominant response vectors of the corresponding single optical fiber under the corresponding single test configuration.

5. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of extracting the common health vector of each optical fiber under the benchmark test parameters based on the population state response matrix includes: The group state response matrix is ​​decomposed into a nonnegative matrix to obtain the corresponding second basis matrix and second coefficient matrix. Each column vector of the second basis matrix represents a potential health mode, and each row vector of the second coefficient matrix represents the correlation weight between the health state of a single optical fiber and the potential health mode. Determine the second L2 norm corresponding to each column vector in the second coefficient matrix, and use the second L2 norm as the second contribution of the corresponding column vector in the second basis matrix; The column vectors of the second basis matrix are sorted in descending order of the second contribution to form a second set of column vectors; Based on a preset second cumulative contribution threshold, the first m column vectors are selected from the second set of column vectors as the common health vectors of each optical fiber under the benchmark test parameters.

6. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of determining multiple state deviation indices based on the first state response vector and the corresponding dominant response vector of the optical fiber under various test configurations includes: For each first state response vector of the optical fiber under any test configuration, a first feature subspace is spanned based on the dominant response vector corresponding to that test configuration. The first state response vector is projected onto the first feature subspace to obtain the first projection vector; Calculate the difference between the first state response vector and the first projection vector to obtain the first residual vector; The L2 norm of the first residual vector is determined as the state deviation index corresponding to the first state response vector.

7. The optical cable fault prediction method based on multi-source data according to claim 1, characterized in that, The step of determining the population deviation based on the second state response vector of the optical fiber and the common health vector includes: Based on the aforementioned common health vector, a second feature subspace is formed; Calculate the covariance matrix of the group state response matrix; The second state response vector of the optical fiber is projected onto the second feature subspace to obtain the second projection vector; Based on the covariance matrix, the Mahalanobis distance of the second projection vector in the second feature subspace is calculated, and the Mahalanobis distance is used as the population deviation of the optical fiber.

8. A fiber optic cable fault prediction system based on multi-source data, characterized in that, include: The data acquisition module is used to acquire distributed test data of each optical fiber in the target optical cable under various test configurations and benchmark test data under benchmark test parameters. Each test configuration is obtained by changing a single test parameter. The feature construction module is used to construct multiple first state response vectors for a single optical fiber and a single test configuration based on the distributed test data for each optical fiber and each test configuration, and aggregate them into an individual state response matrix for the single optical fiber and the single test configuration; and to construct a second state response vector corresponding to the single optical fiber based on the benchmark test data, and aggregate it into a group state response matrix. The feature extraction module is used to extract the dominant response vector of each optical fiber under different test configurations based on the individual state response matrices; and to extract the common health vector of each optical fiber under the benchmark test parameters based on the group state response matrix. The deviation determination module is used to determine multiple state deviation indicators for each optical fiber based on the first state response vector of the optical fiber under each test configuration and the corresponding dominant response vector, and to determine the individual deviation degree according to the state deviation indicators; and to determine the group deviation degree according to the second state response vector of the optical fiber and the common health vector. The result generation module is used to perform joint analysis of the individual deviation and the group deviation to obtain the fault prediction results of each optical fiber.

9. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and execute the method of any one of claims 1 to 7.

Images