Communication cable structure health safety monitoring system based on distributed fiber sensing
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANXI ELECTRIC POWER CO POWER COMM CENT
- Filing Date
- 2026-06-23
- Publication Date
- 2026-07-21
Smart Images
Figure CN122437602A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of optical cable safety monitoring technology, and in particular to a communication optical cable structural health and safety monitoring system based on distributed optical fiber sensing. Background Technology
[0002] Optical fiber cables are the core transmission carriers of modern communication networks. During long-term operation in the field, they are continuously subjected to environmental stress, external disturbances, and structural aging, making them prone to internal structural deformation and hidden damage. Currently, structural health monitoring of optical fiber cables generally adopts local fixed-point monitoring schemes, which cannot achieve long-distance distributed sensing using the cable itself. The optical fiber signal acquisition method is also limited, and it cannot continuously acquire backscattered light change data along the entire cable, resulting in a significant deficiency in the ability to achieve comprehensive status sensing.
[0003] Existing monitoring data processing methods rely on rigid computational logic, and the coherent demodulation process lacks multi-scale adaptive adjustment capabilities, making it difficult to effectively extract continuous vibration and strain distribution data along the optical fiber. A single, fixed algorithm cannot adapt to signal variation patterns in complex environments, has limited ability to identify various structural anomalies in optical cables, struggles to accurately pinpoint the location of anomalies, cannot classify and determine abnormal events, and is unable to identify early-stage, hidden structural hazards.
[0004] The existing monitoring output only includes basic anomaly data. The monitoring system is not integrated with standardized geographic information resources, anomaly locations cannot be matched with corresponding geographic coordinates, and monitoring results lack spatial information support. This singular data output format cannot meet the usage standards for large-scale, intensive operation and maintenance management of optical cable lines. To address the issues of lack of comprehensive sensing, insufficient signal demodulation adaptability, and limited monitoring information dimensions, it is necessary to build an integrated distributed optical fiber sensing monitoring architecture, optimize optical signal data processing algorithms, and establish a fusion mechanism between monitoring data and geographic information to meet the control requirements for all-weather structural health and stability monitoring of communication optical cables. Summary of the Invention
[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a health and safety monitoring system for communication optical cable structures based on distributed optical fiber sensing.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: a health and safety monitoring system for communication optical cable structures based on distributed optical fiber sensing, comprising:
[0007] The signal acquisition module embeds a dedicated sensing fiber inside the communication optical cable and injects pulsed modulated laser into the sensing fiber.
[0008] The signal processing module acquires the backscattered light signal generated in the sensing fiber, performs photoelectric conversion and digital sampling on the backscattered light signal, and obtains the raw scattering data.
[0009] The event recognition module uses an improved phase-sensitive optical time-domain reflectometry algorithm to process the original scattering data. The improved phase-sensitive optical time-domain reflectometry algorithm is based on multi-scale adaptive coherent demodulation technology to identify the vibration and strain distribution along the fiber length direction. Based on the vibration and strain distribution along the fiber length direction, the location and type of potential abnormal events on the communication optical cable structure are determined.
[0010] The information fusion module associates and maps the location and type of the potential abnormal events with the geographic information database of the communication optical cable, generating monitoring alarm information with geographic coordinates.
[0011] As a further aspect of the present invention, the improved phase-sensitive optical time-domain reflectometry algorithm is used to process the original scattering data. This improved algorithm, based on multi-scale adaptive coherent demodulation technology, identifies the vibration and strain distribution along the fiber length direction, and includes:
[0012] The raw scattering data is segmented according to a preset length to form multiple raw data segments;
[0013] For each of the original data segments, a set of coherent demodulators with different local oscillator frequencies are applied to perform parallel demodulation processing to generate demodulation result signals at multiple demodulation scales.
[0014] Instant phase extraction is performed on the demodulation result signals at multiple different demodulation scales corresponding to each of the original data segments to obtain the instant phase sequence at each demodulation scale;
[0015] An adaptive weighted fusion strategy is designed, which dynamically calculates the weight coefficient of the instantaneous phase sequence under each demodulation scale in the fusion based on the signal-to-noise ratio level of each original data segment.
[0016] Based on the weighting coefficients, the instantaneous phase sequences under all demodulation scales corresponding to the same original data segment are weighted and fused to generate the final fused phase trajectory of the original data segment.
[0017] Along the length of the sensing fiber, the fused phase trajectories of all original data segments are sequentially spliced together to form a phase change curve for the entire fiber length.
[0018] By performing time-domain difference calculations on the phase change curve of the entire fiber length, the phase change is converted into vibration intensity and strain intensity information distributed along the fiber length, thus identifying the vibration and strain distribution along the fiber length direction.
[0019] As a further aspect of the present invention, the location and type of potential abnormal events on the communication optical cable structure are determined based on the vibration and strain distribution along the fiber length direction, including:
[0020] Set vibration intensity threshold and strain intensity threshold, and mark the points in the vibration and strain distribution along the fiber length direction that exceed the corresponding threshold as abnormal response points;
[0021] Spatial clustering analysis is performed on the consecutive abnormal response points to form multiple candidate regions for abnormal events;
[0022] Extract the time-frequency characteristics of vibration signals and the variation patterns of strain signals within each candidate region of the abnormal event;
[0023] The extracted time-frequency features of the vibration signal and the change pattern of the strain signal are input into a pre-trained event classification model.
[0024] The event classification model outputs the event type discrimination result for each abnormal event candidate region. The event types include excavation construction, vehicle passage, human theft, static load compression, and slight bending of the optical cable itself.
[0025] The event type determination result is combined with the geographical center point of the corresponding abnormal event candidate area to determine the location and type of the potential abnormal event.
[0026] As a further aspect of the present invention, spatial clustering analysis is performed on the consecutive abnormal response points to form multiple candidate regions for abnormal events, including:
[0027] Using fiber optic distance as coordinates, all the aforementioned abnormal response points are projected onto a one-dimensional space;
[0028] Calculate the distance between adjacent abnormal response points and group abnormal response points whose distance is less than the preset spatial clustering radius into the same initial cluster;
[0029] For each initial cluster, calculate the average vibration intensity and average strain intensity of all anomalous response points within it;
[0030] If the average vibration intensity and average strain intensity of the initial cluster are lower than the preset cluster strength threshold, the initial cluster is disbanded, and the abnormal response points contained therein are marked as discrete noise points.
[0031] For the retained initial cluster, check its spatial span. If the spatial span is greater than the preset maximum event length, then divide the initial cluster into two or more sub-clusters based on the local minimum points in the vibration intensity or strain intensity distribution curve along the fiber distance direction within the initial cluster.
[0032] All the final retained initial clusters and subclusters are defined as the anomalous event candidate regions, and the fiber distance coordinates of the start and end of each anomalous event candidate region are recorded.
[0033] As a further aspect of the present invention, the event classification model is pre-trained through the following steps:
[0034] Collect historical monitoring data, which includes multiple abnormal event samples of confirmed event types. Each abnormal event sample includes vibration signal time-frequency characteristics, strain signal change pattern, and real event type label.
[0035] The time-frequency characteristics of the vibration signal and the variation pattern of the strain signal are standardized to eliminate dimensional differences;
[0036] The standardized feature and pattern data are divided into training subset and validation subset according to the proportion.
[0037] The training subset is used to train a preset deep neural network model, which includes convolutional layers and fully connected layers.
[0038] During training, the validation subset is used to evaluate the classification accuracy of the deep neural network model, and the hyperparameters of the deep neural network model are adjusted until the accuracy reaches the preset standard.
[0039] The trained deep neural network model is then solidified into the pre-trained event classification model.
[0040] As a further aspect of the present invention, the system further includes:
[0041] The data processing module is used to perform noise reduction and trend term removal on the phase change curve of the entire fiber length, specifically including:
[0042] A moving average filter is applied to the phase change curve of the entire fiber length to smooth high-frequency random noise, resulting in a first-order smooth phase curve.
[0043] Polynomial fitting is performed on the first-order smooth phase curve to extract the phase trend term that characterizes the slow-changing environmental temperature factor.
[0044] Subtracting the phase trend term from the phase change curve of the entire fiber length yields the phase fluctuation curve after removing the phase trend term.
[0045] The phase fluctuation curve after removing the phase trend term is subjected to wavelet threshold denoising again to suppress residual noise;
[0046] The signal after wavelet threshold denoising is used as the high-quality phase curve for the final differential calculation.
[0047] As a further aspect of the present invention, the location and type of the potential abnormal events are associated and mapped with the geographic information database of communication optical cables, including:
[0048] The geographic information database of the communication optical cable stores the longitude, latitude, and altitude information corresponding to each sampling point on the sensing optical fiber.
[0049] Based on the location of the potential anomaly, i.e., the fiber optic distance coordinates, the corresponding longitude, latitude, and altitude information are queried from the geographic information database of the communication optical cable.
[0050] The longitude, latitude, and altitude information obtained from the query are bound to the type of the potential abnormal event to form a complete monitoring and alarm information record;
[0051] Add a timestamp and a unique event identifier to each of the aforementioned monitoring alarm records.
[0052] As a further aspect of the present invention, the system further includes:
[0053] The visualization module is used to display monitoring and alarm information in real time, and specifically includes:
[0054] Establish a data interface with the geographic information system to receive newly generated monitoring and alarm information records in real time;
[0055] On the electronic map, using the longitude and latitude information recorded in the monitoring and alarm information as coordinates, corresponding graphic markers are drawn, and different event types are distinguished by graphic markers of different colors and shapes;
[0056] Dynamically label graphic markers using timestamp information;
[0057] When a user clicks on a graphic marker on the electronic map, an information box pops up, displaying detailed information about the event, including the event type, time of occurrence, geographical location, and altitude.
[0058] As a further aspect of the present invention, the system further includes:
[0059] The risk prediction module analyzes event patterns and predicts risks based on historical alarm information, specifically including:
[0060] Periodically extract all monitoring and alarm information records within a specified time period from the historical database;
[0061] The extracted monitoring and alarm information records are statistically analyzed by grid-based partitioning according to geographical location, and the frequency and time patterns of various events in different partitions are analyzed.
[0062] High-risk grid zones are identified where the frequency of events consistently exceeds a preset risk frequency threshold.
[0063] Association rule mining was performed on historical event sequences within high-risk grid partitions to discover frequently co-occurring combinations of event types and their time interval patterns;
[0064] Based on the results of association rule mining, a prediction model is established to probabilistically predict the types of events that will occur in a specific high-risk grid partition within a specific time period in the future, and a risk prediction report is generated.
[0065] As a further aspect of the present invention, the step of performing association rule mining on historical event sequences within high-risk grid partitions includes:
[0066] The event sequence within each high-risk grid partition is converted into a list of event types sorted by time.
[0067] Set the minimum support threshold and the minimum confidence threshold;
[0068] Using a priori algorithm, scan the event type list of all high-risk grid partitions to find all frequent event type itemsets that meet the minimum support threshold;
[0069] Based on the identified frequent event type itemsets, generate all association rules;
[0070] Calculate the confidence score of each association rule and filter out strong association rules with a confidence score higher than the minimum confidence score threshold;
[0071] The antecedent event type, consequent event type, and average time interval between the occurrence of the antecedent event and the occurrence of the consequent event are recorded as the results of the event pattern analysis.
[0072] Compared with the prior art, the advantages and positive effects of the present invention are as follows:
[0073] Leveraging an improved phase-sensitive optical temporal reflectance algorithm coupled with multi-scale adaptive coherent demodulation technology, this method performs layered analytical processing on the raw scattering data along the entire optical cable, overcoming the constraints of fixed demodulation parameters. It continuously captures vibration values and strain distribution along the fiber's length, refining the signal characteristic differences corresponding to varying degrees of structural changes and expanding the data acquisition coverage of the optical cable's structural state. Furthermore, it continuously analyzes the mechanical change parameters across the entire optical cable route, refining the accuracy of identifying subtle structural variations, expanding the criteria for judging abnormal events, dividing the signal characteristic intervals corresponding to different anomalies, and improving the identification conditions for latent structural problems in the optical cable.
[0074] The system synchronously extracts location and type information of abnormal events and performs data association mapping by connecting to a dedicated geographic information database for optical cables. It integrates and matches monitoring business data generated by equipment parsing with line spatial data, supplementing the spatial location attributes of monitoring results and standardizing the output format of monitoring information. This data fusion processing enriches the composition structure of monitoring content, providing standardized coordinate-based alarm content for line operation and maintenance.
[0075] By leveraging the built-in sensing fibers within the optical cable, integrated signal acquisition is achieved across the entire line, forming an uninterrupted distributed sensing mode and maintaining the operational conditions for long-distance, continuous monitoring. The entire process of optical signal acquisition, data conversion, algorithm analysis, and information fusion is seamlessly integrated, standardizing the overall operational flow of optical cable health monitoring. This adapts to a unified management model for large-scale, long-distance communication lines, continuously maintaining the operational stability of online monitoring of the optical cable's structural status. Attached Figure Description
[0076] Figure 1 This is a state diagram of the communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing as described in this invention.
[0077] Figure 2 A flowchart for determining the location and type of potential abnormal events;
[0078] Figure 3 A flowchart for pre-training an event classification model. Detailed Implementation
[0079] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0080] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0081] See Figure 1 This invention provides a health and safety monitoring system for communication optical cable structures based on distributed optical fiber sensing. The specific system includes:
[0082] The system comprises a signal acquisition module, a signal processing module, an event recognition module, and an information fusion module. The signal acquisition module injects pulse-modulated laser light into a dedicated sensing fiber embedded inside the communication optical cable. The signal processing module acquires the backscattered light signal generated in the sensing fiber and performs photoelectric conversion and digital sampling on the signal to obtain the raw scattering data. The event recognition module uses an improved phase-sensitive optical time-domain reflectometry algorithm to process the raw scattering data. This algorithm, based on multi-scale adaptive coherent demodulation technology, can identify the vibration and strain distribution along the fiber length and determine the location and type of potential abnormal events on the communication optical cable structure based on this distribution. The information fusion module associates and maps the identified potential abnormal event locations and types with the geographic information database of the communication optical cable to generate monitoring and alarm information with precise geographic coordinates.
[0083] In one embodiment of the present invention, during signal processing, the raw scattering data is segmented into multiple raw data segments according to a preset length. For each raw data segment, a set of coherent demodulators with different local oscillator frequencies are applied for parallel demodulation processing to generate demodulation result signals at multiple demodulation scales. Instantaneous phase extraction is performed on the demodulation result signals at multiple different demodulation scales corresponding to each raw data segment to obtain the instantaneous phase sequence at each demodulation scale. An adaptive weight fusion strategy is designed to dynamically calculate each demodulation scale based on the signal-to-noise ratio level of each raw data segment. The instantaneous phase sequence is assigned a weight coefficient in the fusion process. Based on the weight coefficient, the instantaneous phase sequences under all demodulation scales corresponding to the same original data segment are weighted and fused to generate the final fused phase trajectory of the original data segment. The fused phase trajectories of all original data segments are sequentially spliced along the length of the sensing fiber to form a phase change curve along the entire fiber length. By performing time-domain difference calculation on the phase change curve along the entire fiber length, the phase change is converted into vibration intensity and strain intensity information distributed along the fiber length, thereby identifying the vibration and strain distribution along the fiber length.
[0084] Based on the identified vibration and strain distribution along the fiber length, refer to Figure 2Vibration intensity thresholds and strain intensity thresholds are set, and points exceeding the corresponding thresholds are marked as abnormal response points. Spatial clustering analysis is performed on continuous abnormal response points to form multiple candidate regions for abnormal events. The time-frequency characteristics of vibration signals and the change patterns of strain signals in each candidate region are extracted. The extracted time-frequency characteristics of vibration signals and the change patterns of strain signals are input into a pre-trained event classification model. The event classification model outputs the event type discrimination result for each candidate region for abnormal events. Event types include excavation construction, vehicle passage, human theft, static load compression, and micro-bending of the optical cable itself. The event type discrimination result is combined with the geographical center point of the corresponding candidate region to determine the location and type of potential abnormal events.
[0085] In practical implementation, the improved phase-sensitive optical time-domain reflectometry algorithm identifies the vibration and strain distribution along the fiber length based on multi-scale adaptive coherent demodulation technology. In this implementation, after the signal processing module acquires the raw scattering data, it segments the raw scattering data according to a preset length, forming multiple raw data segments. For each raw data segment, a set of coherent demodulators with different local oscillator frequencies is applied for parallel demodulation processing, generating demodulation result signals at multiple different demodulation scales. In practical implementation, instantaneous phase extraction is performed on the demodulation result signals at multiple different demodulation scales corresponding to each raw data segment, obtaining the instantaneous phase sequence at each demodulation scale. An adaptive weighted fusion strategy is designed, dynamically calculating the weight coefficient of the instantaneous phase sequence at each demodulation scale in the fusion based on the signal-to-noise ratio level of each raw data segment. In some embodiments, the weight coefficient is calculated using the following formula:
[0086] ;
[0087] in: Indicates the first The original data segment in the first Weighting coefficients under each demodulation scale Indicates the first The original data segment in the first Signal-to-noise ratio level at each demodulation scale This represents the signal-to-noise ratio level of the i-th raw data segment at the k-th demodulation scale. It is an adjustment index. This refers to the total number of demodulation scales. In practice, based on weighting coefficients, the instantaneous phase sequences under all demodulation scales corresponding to the same original data segment are weighted and fused to generate the final fused phase trajectory of the original data segment. Along the length of the sensing fiber, the fused phase trajectories of all original data segments are sequentially spliced to form a phase change curve along the entire fiber length. By performing time-domain difference calculation on the phase change curve along the entire fiber length, the phase change is converted into vibration intensity and strain intensity information distributed along the fiber length, thus identifying the vibration and strain distribution along the fiber length.
[0088] In specific implementations, based on the vibration and strain distribution along the fiber optic cable length, the location and type of potential abnormal events on the communication optical cable structure are determined. Vibration intensity thresholds and strain intensity thresholds are set, and points exceeding the corresponding thresholds in the vibration and strain distribution along the fiber optic cable length are marked as abnormal response points. Spatial clustering analysis is performed on consecutive abnormal response points to form multiple candidate regions for abnormal events. In some embodiments, the time-frequency features of vibration signals and the variation patterns of strain signals within each candidate region for abnormal events are extracted. These extracted features are then input into a pre-trained event classification model. The model outputs an event type classification result for each candidate region, including events such as excavation work, vehicle passage, theft, static load compression, and micro-bending of the optical cable itself. In specific implementations, the event type classification result is combined with the geographical center point of the corresponding candidate region to determine the location and type of potential abnormal events. Optionally, the pre-trained event classification model is built based on a deep neural network, containing convolutional layers and fully connected layers to process the input time-frequency features of vibration signals and the variation patterns of strain signals. Optionally, spatial clustering analysis employs a distance-based clustering method to merge adjacent anomalous response points into candidate regions. It is understood that the time-frequency characteristics of vibration signals include frequency domain energy distribution or time-frequency plot representation, and strain signal variation patterns include strain-time curves or statistical characteristics. It is also understood that the event classification model is trained using historical monitoring data, which contains multiple anomalous event samples of confirmed event types.
[0089] In one embodiment of the present invention, during the process of forming candidate regions for abnormal events, all abnormal response points are projected into a one-dimensional space using fiber optic distance as coordinates. The distance between adjacent abnormal response points is calculated, and abnormal response points with a distance less than a preset spatial clustering radius are grouped into the same initial cluster. For each initial cluster, the average vibration intensity and average strain intensity of all abnormal response points within it are calculated. If the average vibration intensity and average strain intensity of the initial cluster are lower than a preset cluster intensity threshold, the initial cluster is disbanded, and the abnormal response points contained therein are marked as discrete noise points. For the retained initial clusters, their spatial span is checked. If the spatial span is greater than a preset maximum event length, the initial cluster is divided into two or more sub-clusters based on the local minima in the intensity distribution curve. All the final retained initial clusters and sub-clusters are defined as candidate regions for abnormal events, and the fiber optic distance coordinates of the start and end of each region are recorded.
[0090] The pre-trained event classification model is obtained through the following steps, see below. Figure 3 Historical monitoring data is collected, which includes multiple abnormal event samples of confirmed event types. Each sample includes vibration signal time-frequency characteristics, strain signal change patterns, and a label for the actual event type. The vibration signal time-frequency characteristics and strain signal change patterns are standardized to eliminate dimensional differences. The standardized feature and pattern data are then divided into training and validation subsets according to a set ratio. The training subset is used to train a pre-defined deep neural network model, which includes convolutional and fully connected layers. During training, the validation subset is used to evaluate the model's classification accuracy, and the model's hyperparameters are adjusted until the accuracy reaches the preset standard. The trained deep neural network model is then solidified as a pre-trained event classification model.
[0091] After generating the phase change curve for the entire fiber length, the phase change curve is denoised and the trend term is removed. The specific process includes applying a moving average filter to the phase change curve for the entire fiber length to smooth high-frequency random noise and obtain a first-order smoothed phase curve; performing polynomial fitting on the first-order smoothed phase curve to extract the phase trend term that characterizes the slow-changing environmental temperature factor; subtracting the phase trend term from the phase change curve for the entire fiber length to obtain the phase fluctuation curve after removing the trend term; performing wavelet threshold denoising on the phase fluctuation curve after removing the trend term to suppress residual noise; and using the signal after wavelet threshold denoising as the final high-quality phase curve for differential calculation.
[0092] In practical implementation, after the event recognition module generates the phase change curve for the entire fiber length, the data processing module performs noise reduction and trend term removal on the phase change curve. Specifically, a moving average filter is applied to the phase change curve to smooth high-frequency random noise, resulting in a smoothed phase curve. The window length configuration of the moving average filter affects the smoothing effect; some configurations use sampling points as the unit of window length. The actual physical fiber length needs to be converted based on the system sampling rate (e.g., a sampling rate of [missing information]). At that time, window length correspond (fiber segment).
[0093] See Table 1:
[0094] Table 1: Moving Average Filter Window Length Configuration Table
[0095]
[0096] In specific implementations, a polynomial fitting is performed on the smoothed phase curve to extract the phase trend term characterizing the slowly varying environmental temperature. In some embodiments, the polynomial fitting uses the least squares method, and the fitting order is selected based on the length and complexity of the phase change curve. In specific implementations, the phase trend term is subtracted from the phase change curve along the entire fiber length to obtain the phase fluctuation curve after removing the trend term. In specific implementations, the phase fluctuation curve after removing the trend term is subjected to wavelet thresholding denoising to suppress residual noise. Wavelet thresholding denoising involves wavelet decomposition of the phase fluctuation curve to obtain multi-scale wavelet coefficients, and then thresholding the wavelet coefficients. In some embodiments, a soft thresholding function is used for wavelet coefficient thresholding, with the formula:
[0097] ;
[0098] in: It is the first Layer Wavelet coefficients, These are the coefficients after thresholding. It is a threshold. It is a symbolic function. It is a maximum value function. Optional, threshold. It is obtained from the median estimation of the wavelet coefficients, specifically as follows: ,in It is an adjustment factor. It is the median function. In practice, the signal after wavelet threshold denoising is used as the final high-quality phase curve for difference calculation. Optionally, the basis functions for wavelet decomposition can be Daubechies wavelets, and the number of decomposition levels is determined according to the signal length. It can be understood that the moving average filter is used to initially suppress high-frequency random noise, and polynomial fitting is used to eliminate slowly changing trend terms. It can also be understood that wavelet threshold denoising can effectively preserve the abrupt changes in the phase fluctuation curve while suppressing Gaussian white noise.
[0099] In one embodiment of the present invention, during the information fusion process, the location and type of potential abnormal events are associated and mapped with the geographic information database of the communication optical cable. The geographic information database of the communication optical cable stores the longitude, latitude, and altitude information corresponding to each sampling point on the sensing optical fiber. Based on the location of the potential abnormal event, i.e., the fiber distance coordinates, the corresponding longitude, latitude, and altitude information is queried in the geographic information database of the communication optical cable. The queried longitude, latitude, and altitude information are bound with the type of potential abnormal event to form a complete monitoring alarm information record. A timestamp and a unique event identifier are added to each monitoring alarm information record.
[0100] The system also includes a visualization module for real-time visualization of monitoring and alarm information. The specific process includes establishing a data interface with the geographic information system, receiving newly generated monitoring and alarm information records in real time, drawing corresponding graphic markers on the electronic map using the longitude and latitude information in the monitoring and alarm information records as coordinates, distinguishing different event types using graphic markers of different colors and shapes, dynamically labeling the graphic markers with timestamp information, and popping up an information box when the user clicks on a graphic marker on the electronic map to display the detailed content of the event, including the event type, occurrence time, geographical location, and altitude.
[0101] In practical implementation, the information fusion module maps the location and type of potential anomalies to the geographic information database of the communication optical cable. This database stores the longitude, latitude, and altitude information corresponding to each sampling point on the sensing fiber. In practice, based on the location of the potential anomaly, i.e., the fiber optic distance coordinates, the corresponding longitude, latitude, and altitude information are queried from the geographic information database of the communication optical cable. In some embodiments, for an anomaly candidate region spanning multiple sampling points, the fiber optic distance coordinates of its center point are... It can be done through the formula:
[0102] ;
[0103] in: These are the starting fiber optic distance coordinates of the recorded area. It is the length of the recorded area, based on the fiber optic distance coordinates from the center point. Interpolation queries are performed in the geographic information database of the communication optical cable to obtain precise geographic coordinates. In specific implementation, the obtained longitude, latitude, and altitude information are bound with the type of potential abnormal event to form a complete monitoring and alarm information record. A timestamp and a unique event identifier are added to each monitoring and alarm information record. Optionally, the unique event identifier is generated using a combination rule of "fiber optic identifier-timestamp-serial number".
[0104] In practical implementation, the system includes a visualization module for real-time visualization of monitoring and alarm information. This visualization module establishes a data interface with the geographic information system (GIS) to receive newly generated monitoring and alarm information records in real time. In some embodiments, different event types are distinguished on the electronic map using graphic markers of different colors and shapes, as shown in Table 2.
[0105] Table 2: Mapping Table between Event Types and Graphical Tag Attributes
[0106]
[0107] In practical implementation, on the electronic map, using the longitude and latitude information from the monitoring alarm information records as coordinates, corresponding graphic markers are drawn according to the table above, and these graphic markers are dynamically labeled with timestamp information. In practical implementation, when a user clicks on a graphic marker on the electronic map, an information box pops up, displaying detailed information about the event, including the event type, occurrence time, geographical location, and altitude. Optionally, the dynamic labeling can be displayed by hovering the timestamp text near the graphic marker. It can be understood that the geographical location information includes longitude, latitude, and altitude, and can also be associated with the nearest place name. It can be understood that the visualization module can display multiple monitoring alarm information records simultaneously and supports filtering functions based on event type or time range.
[0108] In one embodiment of the present invention, the system further includes a risk prediction module. This module analyzes event patterns and predicts risks based on historical alarm information. It periodically extracts all monitoring alarm information records within a specified time period from the historical database, performs grid-based statistical analysis on the extracted monitoring alarm information records according to geographical location, analyzes the frequency and time patterns of various events in different zones, identifies high-risk grid zones where the frequency of events consistently exceeds a preset risk frequency threshold, performs association rule mining on historical event sequences within high-risk grid zones, discovers frequently co-occurring event type combinations and their time interval patterns, and establishes a prediction model based on the results of association rule mining to probabilistically predict the types of events that will occur in a specific time period and a specific high-risk grid zone in the future, generating a risk prediction report.
[0109] When performing association rule mining on historical event sequences within high-risk grid partitions, the event sequences within each high-risk grid partition are converted into a time-sorted list of event types. A minimum support threshold and a minimum confidence threshold are set. A priori algorithm is used to scan the event type lists of all high-risk grid partitions to identify all frequent event type itemsets that meet the minimum support threshold. Based on the identified frequent event type itemsets, all association rules are generated. The confidence of each association rule is calculated, and strong association rules with confidence scores higher than the minimum confidence threshold are selected. The antecedent event type, consequent event type, and the average time interval between the occurrence of the antecedent and the occurrence of the consequent for each strong association rule are recorded as the results of the event pattern analysis.
[0110] The risk prediction module periodically extracts all monitoring and alarm information records within a specified time period from the historical database. In practice, the extracted monitoring and alarm information records are statistically divided into grids based on geographical location. The frequency and temporal patterns of various events within different zones are analyzed. The grid partitioning divides the geographical area into several rectangular grids of equal area based on longitude and latitude. In practice, high-risk grid zones with event occurrence frequencies consistently higher than a preset risk frequency threshold are identified. The risk frequency threshold is set based on the historical average event occurrence rate and acceptable risk level. In practice, association rule mining is performed on the historical event sequences within the high-risk grid zones to discover frequently co-occurring event type combinations and their time interval patterns. In some embodiments, association rule mining not only focuses on event co-occurrence but also evaluates the effectiveness of the rules. Lift is a key indicator used to measure the correlation between the antecedent and consequent of a rule, and its calculation can be expressed as:
[0111] ;
[0112] in: This indicates the lift of the association rule "If A occurs, then B occurs". This indicates the confidence level of the association rule's occurrence in historical data. This represents the support level of event type A in the entire historical dataset. This represents the support and lift of event type B throughout the entire historical dataset. A value greater than 1 indicates that A and B are positively correlated.
[0113] In practical implementation, the association rule mining of historical event sequences within high-risk grid partitions includes the following steps: The event sequences within each high-risk grid partition are converted into a time-sorted list of event types, and minimum support and minimum confidence thresholds are set. In practice, a priori algorithm is used to scan the event type lists of all high-risk grid partitions, identifying all frequent event type itemsets that meet the minimum support threshold. Based on these frequent event type itemsets, all association rules are generated, and the confidence of each association rule is calculated. Strong association rules with confidence scores higher than the minimum confidence threshold are selected. Optionally, the minimum support threshold is set to 0.05, and the minimum confidence threshold is set to 0.6. The antecedent event type, consequent event type, and the average time interval between the occurrence of the antecedent and the consequent for each strong association rule are recorded as the result of the event pattern analysis. In practice, based on the results of association rule mining, a prediction model is established to probabilistically predict the types of events that will occur within a specific high-risk grid partition in a future specific time period, generating a risk prediction report. In some embodiments, the predictive model is built based on mined strong association rules and their average time intervals. When an antecedent event type is detected within a high-risk grid partition, the model predicts the probability of a consequent event type occurring within a specific future time window based on the corresponding rules. Optionally, the risk prediction report is presented in the form of a structured document or a visual chart, including the location of the high-risk grid partition, the predicted event type, the predicted occurrence time window, and the estimated probability. It is understood that association rule mining helps to discover potential patterns such as "after excavation work, the optical cable may experience a slight bend within 24 hours." It is also understood that the predictive model can be periodically updated and iterated using the latest historical data to maintain the accuracy of the predictions.
[0114] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A health and safety monitoring system for communication optical cable structures based on distributed optical fiber sensing, characterized in that, The system includes: The signal acquisition module embeds a dedicated sensing fiber inside the communication optical cable and injects pulsed modulated laser into the sensing fiber. The signal processing module acquires the backscattered light signal generated in the sensing fiber, performs photoelectric conversion and digital sampling on the backscattered light signal, and obtains the raw scattering data. The event recognition module uses an improved phase-sensitive optical time-domain reflectometry algorithm to process the original scattering data. The improved phase-sensitive optical time-domain reflectometry algorithm is based on multi-scale adaptive coherent demodulation technology to identify the vibration and strain distribution along the fiber length direction. Based on the vibration and strain distribution along the fiber length direction, the location and type of potential abnormal events on the communication optical cable structure are determined. The information fusion module associates and maps the location and type of the potential abnormal events with the geographic information database of the communication optical cable, generating monitoring alarm information with geographic coordinates.
2. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 1, characterized in that, The improved phase-sensitive optical time-domain reflectometry algorithm is used to process the original scattering data. This improved algorithm, based on multi-scale adaptive coherent demodulation technology, identifies the vibration and strain distribution along the fiber length, including: The raw scattering data is segmented according to a preset length to form multiple raw data segments; For each of the original data segments, a set of coherent demodulators with different local oscillator frequencies are applied to perform parallel demodulation processing to generate demodulation result signals at multiple demodulation scales. Instant phase extraction is performed on the demodulation result signals at multiple different demodulation scales corresponding to each of the original data segments to obtain the instant phase sequence at each demodulation scale; An adaptive weighted fusion strategy is designed, which dynamically calculates the weight coefficient of the instantaneous phase sequence under each demodulation scale in the fusion based on the signal-to-noise ratio level of each original data segment. Based on the weighting coefficients, the instantaneous phase sequences under all demodulation scales corresponding to the same original data segment are weighted and fused to generate the final fused phase trajectory of the original data segment. Along the length of the sensing fiber, the fused phase trajectories of all original data segments are sequentially spliced together to form a phase change curve for the entire fiber length. By performing time-domain difference calculations on the phase change curve of the entire fiber length, the phase change is converted into vibration intensity and strain intensity information distributed along the fiber length, thus identifying the vibration and strain distribution along the fiber length direction.
3. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 1, characterized in that, Based on the vibration and strain distribution along the fiber length, the location and type of potential abnormal events on the communication optical cable structure are determined, including: Set vibration intensity threshold and strain intensity threshold, and mark the points in the vibration and strain distribution along the fiber length direction that exceed the corresponding threshold as abnormal response points; Spatial clustering analysis is performed on the consecutive abnormal response points to form multiple candidate regions for abnormal events; Extract the time-frequency characteristics of vibration signals and the variation patterns of strain signals within each candidate region of the abnormal event; The extracted time-frequency features of the vibration signal and the change pattern of the strain signal are input into a pre-trained event classification model. The event classification model outputs the event type discrimination result for each abnormal event candidate region. The event types include excavation construction, vehicle passage, human theft, static load compression, and slight bending of the optical cable itself. The event type determination result is combined with the geographical center point of the corresponding abnormal event candidate area to determine the location and type of the potential abnormal event.
4. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 3, characterized in that, Spatial clustering analysis is performed on the consecutive anomalous response points to form multiple candidate regions for anomalous events, including: Using fiber optic distance as coordinates, all the aforementioned abnormal response points are projected onto a one-dimensional space; Calculate the distance between adjacent abnormal response points and group abnormal response points whose distance is less than the preset spatial clustering radius into the same initial cluster; For each initial cluster, calculate the average vibration intensity and average strain intensity of all anomalous response points within it; If the average vibration intensity and average strain intensity of the initial cluster are lower than the preset cluster strength threshold, the initial cluster is disbanded, and the abnormal response points contained therein are marked as discrete noise points. For the retained initial cluster, check its spatial span. If the spatial span is greater than the preset maximum event length, then divide the initial cluster into two or more sub-clusters based on the local minimum points in the vibration intensity or strain intensity distribution curve along the fiber distance direction within the initial cluster. All the final retained initial clusters and subclusters are defined as the anomalous event candidate regions, and the fiber distance coordinates of the start and end of each anomalous event candidate region are recorded.
5. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 3, characterized in that, The event classification model is pre-trained through the following steps: Collect historical monitoring data, which includes multiple abnormal event samples of confirmed event types. Each abnormal event sample includes vibration signal time-frequency characteristics, strain signal change pattern, and real event type label. The time-frequency characteristics of the vibration signal and the variation pattern of the strain signal are standardized to eliminate dimensional differences; The standardized feature and pattern data are divided into training subset and validation subset according to the proportion. The training subset is used to train a preset deep neural network model, which includes convolutional layers and fully connected layers. During training, the validation subset is used to evaluate the classification accuracy of the deep neural network model, and the hyperparameters of the deep neural network model are adjusted until the accuracy reaches the preset standard. The trained deep neural network model is then solidified into the pre-trained event classification model.
6. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 2, characterized in that, The system also includes: The data processing module is used to perform noise reduction and trend term removal on the phase change curve of the entire fiber length, specifically including: A moving average filter is applied to the phase change curve of the entire fiber length to smooth high-frequency random noise, resulting in a first-order smooth phase curve. Polynomial fitting is performed on the first-order smooth phase curve to extract the phase trend term that characterizes the slow-changing environmental temperature factor. Subtracting the phase trend term from the phase change curve of the entire fiber length yields the phase fluctuation curve after removing the phase trend term. The phase fluctuation curve after removing the phase trend term is subjected to wavelet threshold denoising again to suppress residual noise; The signal after wavelet threshold denoising is used as the high-quality phase curve for the final differential calculation.
7. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 1, characterized in that, Associating the location and type of the potential abnormal events with the geographic information database of communication optical cables, including: The geographic information database of the communication optical cable stores the longitude, latitude, and altitude information corresponding to each sampling point on the sensing optical fiber. Based on the location of the potential anomaly, i.e., the fiber optic distance coordinates, the corresponding longitude, latitude, and altitude information are queried from the geographic information database of the communication optical cable. The longitude, latitude, and altitude information obtained from the query are bound to the type of the potential abnormal event to form a complete monitoring and alarm information record; Add a timestamp and a unique event identifier to each of the aforementioned monitoring alarm records.
8. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 7, characterized in that, The system also includes: The visualization module is used to display monitoring and alarm information in real time, and specifically includes: Establish a data interface with the geographic information system to receive newly generated monitoring and alarm information records in real time; On the electronic map, using the longitude and latitude information recorded in the monitoring and alarm information as coordinates, corresponding graphic markers are drawn, and different event types are distinguished by graphic markers of different colors and shapes; Dynamically label graphic markers using timestamp information; When a user clicks on a graphic marker on the electronic map, an information box pops up, displaying detailed information about the event, including the event type, time of occurrence, geographical location, and altitude.
9. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 8, characterized in that, The system also includes: The risk prediction module analyzes event patterns and predicts risks based on historical alarm information, specifically including: Periodically extract all monitoring and alarm information records within a specified time period from the historical database; The extracted monitoring and alarm information records are statistically analyzed by grid-based partitioning according to geographical location, and the frequency and time patterns of various events in different partitions are analyzed. High-risk grid zones are identified where the frequency of events consistently exceeds a preset risk frequency threshold. Association rule mining was performed on historical event sequences within high-risk grid partitions to discover frequently co-occurring combinations of event types and their time interval patterns; Based on the results of association rule mining, a prediction model is established to probabilistically predict the types of events that will occur in a specific high-risk grid partition within a specific time period in the future, and a risk prediction report is generated.
10. The communication optical cable structure health and safety monitoring system based on distributed optical fiber sensing according to claim 9, characterized in that, The process of mining association rules for historical event sequences within high-risk grid partitions includes: The event sequence within each high-risk grid partition is converted into a list of event types sorted by time. Set the minimum support threshold and the minimum confidence threshold; Using a priori algorithm, scan the event type list of all high-risk grid partitions to find all frequent event type itemsets that meet the minimum support threshold; Based on the identified frequent event type itemsets, generate all association rules; Calculate the confidence score of each association rule and filter out strong association rules with a confidence score higher than the minimum confidence score threshold; The antecedent event type, consequent event type, and average time interval between the occurrence of the antecedent event and the occurrence of the consequent event are recorded as the results of the event pattern analysis.