An artificial intelligence-based optical fiber temperature measurement data management system and method

Abstract

This invention discloses an artificial intelligence-based fiber optic temperature measurement data management system and method, belonging to the field of data management technology. Based on the physical characteristics of optical fibers, the invention divides the optical fiber into several segments; constructs an optical fiber topology model based on the divided segments; calculates the reference light propagation speed for each segment based on its physical characteristics; collects the temperature and optical signal sequences of the entire optical fiber link; matches the optical signal sequences to the corresponding topology segments one by one according to the optical fiber segment based on the optical fiber topology model; constructs an AI time-series feature classification model to filter abnormal topology segments based on temperature and optical signal sequences; locates fiber breakpoints in abnormal topology segments; analyzes the impact mechanism of fiber breakpoints on optical signal propagation and establishes causal links; based on the optical fiber topology model and causal links, infers the scope of the abnormal impact caused by the breakpoints and determines the processing priority of the fiber breakpoints. This invention achieves intelligent management and control of the entire process of optical fiber temperature measurement data, improving the efficiency of fault handling.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to an artificial intelligence-based fiber optic temperature measurement data management system and method. Background Technology

[0002] Fiber optic temperature measurement technology, with its advantages of resistance to electromagnetic interference, wide temperature measurement range, and high spatial resolution, is widely used in industrial scenarios such as power cables, chemical storage tanks, and building security. The core requirement is to promptly detect faults such as fiber optic breaks through accurate temperature measurement and optical signal analysis, avoiding service interruptions and security risks. However, as industrial scenarios demand higher accuracy in temperature measurement and faster fault response, traditional fiber optic temperature measurement data management methods are increasingly unable to meet the needs of processing massive amounts of time-series data, dynamically locating faults, and making efficient operational and maintenance decisions.

[0003] In existing technologies, matching optical signal sequences with fiber segments lacks precise algorithms under topological constraints, making it susceptible to interference signals across segments and resulting in significant matching errors. Anomaly screening often relies on fixed thresholds, making it difficult to distinguish between genuine anomalies caused by breakpoints and pseudo-anomalies caused by electromagnetic interference or link loss. Breakpoint location simply combines optical signal propagation parameters without incorporating the segment's reference optical propagation speed for accurate derivation, leading to location errors exceeding operational and maintenance requirements. When extrapolating the impact range of anomalies, existing technologies do not incorporate topological structure and causal links for transmission analysis, only assessing the direct impact of a single breakpoint and failing to cover the chain reactions of series and parallel links. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based fiber optic temperature measurement data management system and method to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Firstly, this application provides an artificial intelligence-based method for managing fiber optic temperature measurement data, comprising the following steps:

[0007] Based on the physical properties of optical fiber, the optical fiber is divided into several segments; an optical fiber topology model is constructed based on the divided optical fiber segments; and the reference light propagation speed of each segment is calculated based on the physical properties of each segment.

[0008] The temperature and optical signal sequences of the entire optical fiber link are collected; based on the optical fiber topology model, the optical signal sequences are matched to the corresponding topology segments one by one according to the optical fiber segments;

[0009] Construct an AI time-series feature classification model to filter anomalous topological segments based on temperature and light signal sequences; locate fiber optic breakpoints within anomalous topological segments.

[0010] The mechanism by which fiber optic breaks affect optical signal propagation is analyzed, and a causal link is established. Based on the fiber optic topology model and the causal link, the scope of the abnormal impact caused by the breaks is deduced, and the processing priority of fiber optic breaks is determined.

[0011] In conjunction with the first aspect, in a first embodiment of the first aspect of this application, dividing the optical fiber into several segments according to the physical characteristics of the optical fiber includes:

[0012] The physical characteristics of optical fiber include material, core diameter and cladding size, attenuation coefficient, refractive index distribution, and curvature of the laying path; fiber segmentation constraints are set, including homogenization constraints and boundary adaptation constraints; the homogenization constraint specifically requires that the fluctuation range of all physical characteristics within the same segment be controlled within a preset threshold; the boundary adaptation constraint specifically requires that segment boundaries be aligned with the natural boundaries of the optical fiber to avoid the distance between adjacent boundaries being less than the optical fiber length corresponding to the temperature measurement accuracy;

[0013] The locations of abrupt changes in physical properties and the natural boundaries of the optical fiber are integrated into a set of candidate boundaries, and duplicate boundaries are removed. The entire optical fiber is divided into several initial segments using the candidate boundaries as separators. The physical property parameters of each segment are checked one by one to see if they meet the optical fiber segmentation constraints. Optical fiber segments that do not meet the constraints are further divided.

[0014] In conjunction with the first aspect, in a second embodiment of the first aspect of this application, the construction of the fiber optic topology model based on the divided fiber segments includes:

[0015] Using segment codes as unique IDs, and associating attributes including average physical characteristics, length, start and end boundary IDs, and service type, segment entities are constructed. Node IDs are generated based on segment boundary coordinates, and attributes are labeled according to boundary type. The connected segment IDs at the nodes are recorded, constructing node entities. Each segment's start and end boundary corresponds to a unique node, and a node can be associated with one or more segments. When two optical fiber segments share the same node, it is determined that there is a direct connection relationship, and the connection type, including series and parallel, is labeled synchronously.

[0016] Perform association mapping between nodes and segments, label the topology links in layers according to business priority, form a core topology layer by combining core business carrying segments and corresponding nodes, form a normal topology layer by combining non-core business segments, and form an idle layer by combining isolated segments, thus forming a layered topology structure and constructing an optical fiber topology model.

[0017] In conjunction with the first aspect, in the third embodiment of the first aspect of this application, the step of matching the optical signal sequence to the corresponding topology segment one by one according to the optical fiber topology model includes:

[0018] For each set of optical signal sequences, calculate its corresponding physical location on the optical fiber using the following formula: Where X is the signal position. The reference position for the data acquisition device. This is the initial reference optical speed for the corresponding link. This refers to the signal propagation time difference;

[0019] Traverse all fiber segments in the fiber topology model and construct segment codes and location intervals. The mapping table, where, The starting position coordinates, The coordinates are the termination position coordinates; for X, an interval determination algorithm is used to match one by one, and based on the matching results, the signal position is corrected by the reference light speed of the corresponding segment, and multi-feature redundancy verification is performed.

[0020] In conjunction with the first aspect, in the fourth embodiment of the first aspect of this application, for X, the interval determination algorithm is used for matching one by one, and for the matching result, multi-feature redundancy verification is performed in combination with the reference light velocity correction signal position of the corresponding segment, including:

[0021] The interval determination algorithm is as follows: when there exists a fiber segment that satisfies... When X exceeds all segment intervals or falls on a segment boundary, it is marked as a boundary signal and temporarily assigned to the upstream segment; thus, the matching result is obtained.

[0022] Based on the matching results, the reference optical propagation speed of the corresponding fiber segment is used. Correct signal position, This represents the precise reference optical velocity of the i-th segment; the correction formula is as follows: The matching results are then resubmitted into the interval determination algorithm to verify the matching results. For the corrected signal position; when If the match still falls within the original paragraph range, then the matching relationship is confirmed; when... When deviating from the original paragraph, calculate the error between the two positions. ,when When, retain the original match, when Then, re-traverse the topological paragraph matching. Calculate the error threshold for the location;

[0023] For a single optical signal sequence, the correlation coefficient r is calculated by combining the mean physical characteristics of the corresponding segment, using the following formula: Where S is the optical signal attenuation characteristic value, COV is the covariance, and VAR is the variance. The attenuation coefficient is used. When r is not less than the strong correlation threshold, the match is confirmed to be valid. When r is less than the strong correlation threshold, check whether it is signal interference or a matching error, and re-execute the position calculation and interval matching.

[0024] In conjunction with the first aspect, in the fifth embodiment of the first aspect of this application, the construction of the AI ​​temporal feature classification model to filter anomalous topological segments based on temperature and light signal sequences includes:

[0025] Based on temperature and light signal sequences, temporal features are extracted, including time domain features, trend domain features, and cross-feature correlation features; feature concatenation, redundant feature removal, and feature weight allocation are then performed.

[0026] Specifically, for time-domain features, basic statistical features and abrupt change features are extracted at fixed time steps to capture the instantaneous change patterns of the signal, covering the number of sequence means, extreme values, peaks, and the number of abrupt changes where the difference between adjacent time steps exceeds a threshold. For trend-domain features, the signal change trend is analyzed through linear fitting, extracting the trend slope and the rate of change of the trend within the sliding window, focusing on capturing sudden changes in the signal trend caused by breakpoints and distinguishing between normal fluctuations and abnormal abrupt changes. For cross-feature correlation features, a linkage analysis of temperature and optical signals is established to capture the coordinated change patterns of the two, including the correlation between temperature and optical signal attenuation at the same time step, and the matching relationship between the rate of temperature change and the propagation characteristics of the optical signal.

[0027] Furthermore, the extracted three types of temporal features are concatenated with the physical characteristics of the corresponding fiber optic segments to form a multi-dimensional feature vector that adapts to the model input, achieving the fusion of temporal features and segment attributes. A feature filtering algorithm is employed to remove constant features with extremely low variance and highly redundant features, retaining core features that contribute significantly to anomaly detection, thus reducing model computation and avoiding overfitting. Feature weights are set based on business priority, increasing the weights of temperature trends and optical signal mutations in core business segments to improve the sensitivity and accuracy of anomaly identification in core segments.

[0028] A hybrid model architecture combining a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is used to construct an AI temporal feature classification model. The fused temporal feature vector is input into the LSTM layer to capture the long-term and short-term dependencies of temperature and light signals over time. The CNN layer is used to enhance the feature extraction capability for local signal mutations caused by breakpoints. The Sigmoid activation function is used in the output layer to output the anomaly probability value of a single paragraph. The closer the probability is to 1, the higher the confidence of judging it as an anomalous paragraph. The model is trained and regularized to filter out anomalous topological paragraphs.

[0029] In conjunction with the first aspect, in the sixth embodiment of the first aspect of this application, the location of the fiber optic breakpoint in the abnormal topology segment includes:

[0030] Optical signal data are grouped according to the coding of abnormal topology segments, retaining matching optical signals within the corresponding segments and eliminating cross-segment interference signals and low-correlation signals. A distance conversion algorithm is used to calculate the actual distance of the optical signal from the transmitter to the detection point by combining the reference optical propagation speed of the abnormal segment with the optical signal propagation time difference. Using the reference position of the acquisition equipment as a reference, the specific position of the detection point on the optical fiber is calculated, and several suspected breakpoint locations are identified. Signal groups with abrupt changes in propagation time difference and sudden drops in scattered light intensity are screened out. Through a clustering algorithm, suspected breakpoint locations with consistent signal characteristics are grouped into one category, duplicate results are merged, and the optical fiber breakpoint is located.

[0031] In conjunction with the first aspect, in the seventh embodiment of the first aspect of this application, the step of analyzing the influence mechanism of optical fiber breakpoints on optical signal propagation and establishing a causal link includes:

[0032] For completely blocked breakpoints, a signal attenuation analysis algorithm is used to determine the degree of interruption in the optical signal propagation path and to identify the mechanism by which the transmitter signal cannot penetrate the breakpoint. For partially damaged breakpoints, a propagation velocity fluctuation algorithm is used to analyze the mechanism of signal velocity shift or scattering intensity distortion caused by changes in the fiber structure at the breakpoint. Based on the identified impact mechanism, a feature correlation algorithm is used to establish a causal link.

[0033] Specifically, for completely blocking breakpoints, a signal attenuation analysis algorithm is used to compare the differences in optical signal attenuation at the upstream and downstream near ends of the breakpoint point by point. Combined with the normal attenuation threshold of the optical fiber, the degree of interruption of the optical signal propagation path is determined. Analysis shows that such breakpoints, due to complete breakage or severe damage to the physical structure of the optical fiber, form a signal propagation barrier. The core mechanism is clearly defined: the transmitting optical signal cannot penetrate this barrier, leading to a complete interruption of the downstream optical signal and the absence of effective timing feedback. At the same time, abrupt attenuation change nodes are recorded as evidence for mechanism verification.

[0034] For some damaged breakpoints, the propagation velocity fluctuation algorithm, combined with the reference light propagation velocity of the corresponding segment, is used to analyze the impact of structural changes such as fiber core diameter deformation and local material damage at the breakpoint on optical signal propagation. The derivation shows that fiber structural variations alter the light propagation path and constraints, leading to deviations in optical signal propagation velocity from the reference value, or causing irregular distortion of scattered light intensity and a decrease in signal-to-noise ratio. This clarifies the direct correlation between structural changes and signal anomalies, and identifies the root cause of signal timing disorder.

[0035] Furthermore, based on the aforementioned two types of clearly defined impact mechanisms of breakpoints, a feature correlation algorithm is used to hierarchically correlate breakpoint types, fiber structure change states, abnormal manifestations of core optical signal features, and propagation impact results.

[0036] In conjunction with the first aspect, in the eighth embodiment of the first aspect of this application, the step of extrapolating the scope of the abnormal impact caused by the breakpoint based on the fiber optic topology model and causal link, and determining the processing priority of the fiber optic breakpoint, includes:

[0037] The impact of the serial segment is propagated downstream along the causal link to determine the coverage area of ​​the interruption; the parallel segment uses a redundancy assessment algorithm to determine whether the backup link can compensate and to define the impact boundary; a multi-dimensional assessment algorithm is used to quantify the impact level of each interruption point by combining the size of the impact range, the proportion of core business, and the degree of signal distortion, and to distinguish between core and secondary impacts; based on the quantification results, a weighted sorting algorithm is used to prioritize interruptions without redundant links and those affected by core business, followed by interruptions with backup links and those affected by non-core business; the processing priority of fiber optic interruptions is determined.

[0038] Secondly, this application provides an artificial intelligence-based fiber optic temperature measurement data management system, including:

[0039] The reference light propagation speed calculation module includes: an optical fiber segmentation unit that divides the optical fiber into several segments based on its physical characteristics; a topology model construction unit that constructs an optical fiber topology model based on the segmented optical fibers; and a reference light propagation speed calculation unit that calculates the reference light propagation speed of each segment based on its physical characteristics.

[0040] Topology segment matching module: includes: a data acquisition unit that collects the temperature and optical signal sequences of the entire optical fiber link; and a topology segment matching unit that matches the optical signal sequences to the corresponding topology segments one by one according to the optical fiber topology model.

[0041] Fiber optic breakpoint location module: includes: an abnormal topology segment screening unit that constructs an AI time-series feature classification model to screen abnormal topology segments based on temperature and optical signal sequences; and a fiber optic breakpoint location unit that locates fiber optic breakpoints within abnormal topology segments.

[0042] The processing priority determination module includes: a causal link establishment unit that analyzes the impact mechanism of fiber optic breakpoints on optical signal propagation and establishes causal links; and a processing priority determination unit that, based on the fiber optic topology model and causal links, extrapolates the range of abnormal impacts caused by the breakpoints and determines the processing priority of the fiber optic breakpoints.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] 1. This invention relies on a topology model to achieve accurate matching of optical signal sequences and optical fiber segments. Combined with an AI time-series feature classification model, it deeply mines the time-series linkage features of temperature and optical signals, accurately distinguishes between real breakpoint anomalies and interference signals, significantly reduces the false positive and false negative rates, and especially improves the sensitivity of anomaly identification in core business segments.

[0045] 2. This invention combines the reference light propagation speed of a segment with optical signal parameters to derive the breakpoint location, verifies calibration errors through multiple dimensions, and ensures positioning accuracy that meets maintenance requirements; at the same time, the system analyzes the impact mechanism of the breakpoint on optical signal propagation, establishes a structured causal link, and clearly sorts out the logic of influence transmission.

[0046] 3. Based on topology models and causal links, this invention uses a link propagation algorithm to deduce the scope of anomaly impact and cover chain reactions; it employs weighted sorting and conflict coordination algorithms to quantify the degree of business impact and redundancy capacity, scientifically determine the priority of breakpoint handling, balance business continuity and operational efficiency, and solve the drawbacks of experience-based decision-making in existing technologies. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating the steps of an artificial intelligence-based fiber optic temperature measurement data management method according to the present invention.

[0048] Figure 2 This is a system structure diagram of an artificial intelligence-based fiber optic temperature measurement data management system according to the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Example: Figures 1-2 As shown, the present invention provides a technical solution:

[0051] like Figure 1 As shown, this application provides an artificial intelligence-based fiber optic temperature measurement data management method, including the following steps:

[0052] Step S100: Divide the optical fiber into several segments according to its physical characteristics; construct an optical fiber topology model based on the divided optical fiber segments; calculate the reference light propagation speed of each segment based on its physical characteristics.

[0053] Specifically, the physical characteristics of the optical fiber include material, core diameter and cladding size, attenuation coefficient, refractive index distribution, and curvature of the laying path; fiber segmentation constraints are set, including homogenization constraints and boundary adaptation constraints; the homogenization constraint specifically requires that the fluctuation range of all physical characteristics within the same segment be controlled within a preset threshold; the boundary adaptation constraint specifically requires that segment boundaries be aligned with the natural boundaries of the optical fiber to avoid the distance between adjacent boundaries being less than the optical fiber length corresponding to the temperature measurement accuracy;

[0054] The locations of abrupt changes in physical properties and the natural boundaries of the optical fiber are integrated into a set of candidate boundaries, and duplicate boundaries are removed. The entire optical fiber is divided into several initial segments using the candidate boundaries as separators. The physical property parameters of each segment are checked one by one to see if they meet the optical fiber segmentation constraints. Optical fiber segments that do not meet the constraints are further divided.

[0055] Furthermore, segment entities are constructed by using segment codes as unique IDs and associating attributes including average physical characteristics, length, start and end boundary IDs, and service type. Node IDs are generated based on the coordinates of segment boundary positions, attributes are labeled according to boundary type, and the connection segment IDs at the nodes are recorded to construct node entities. Each segment's start and end boundary corresponds to a unique node, and a node can be associated with one or more segments. When two optical fiber segments share the same node, it is determined that there is a direct connection relationship, and the connection type, including series and parallel connections, is labeled synchronously.

[0056] Perform association mapping between nodes and segments, label the topology links in layers according to business priority, form a core topology layer by combining core business carrying segments and corresponding nodes, form a normal topology layer by combining non-core business segments, and form an idle layer by combining isolated segments, thus forming a layered topology structure and constructing an optical fiber topology model.

[0057] In one specific embodiment, this embodiment uses silica optical fiber commonly used in industrial scenarios, with a total length of 1000 meters. The core physical characteristics are as follows: the material is germanium-doped silica, the core diameter is 9μm, and the cladding size is 125μm; the attenuation coefficient is 0.2dB / km (at a wavelength of 1550nm); the refractive index distribution is step-type, with a core refractive index of 1.468 and a cladding refractive index of 1.462; the curvature range of the laying path is 50mm-100mm, and the entire route includes 3 fusion splice points (located at 220m, 550m, and 780m respectively) and 1 branch point (located at 630m), all of which are natural boundaries of the optical fiber.

[0058] Homogenization constraint thresholds are set as follows: the material is uniformly germanium-doped quartz (without fluctuations), core diameter fluctuation ≤ ±0.2μm, attenuation coefficient fluctuation ≤ ±0.02dB / km, refractive index fluctuation ≤ ±0.001, and laying path curvature fluctuation ≤ ±5mm. Boundary adaptation constraints: segment boundaries are preferentially aligned with the above-mentioned fusion splices and branch points, and the distance between adjacent boundaries is not less than 0.5 meters (corresponding to the fiber length requirement of a temperature measurement accuracy of 0.1℃).

[0059] By scanning with an optical fiber characteristic analyzer, the locations of abrupt changes in physical properties were identified: a slight change in the refractive index gradient at 180m, a curvature inflection point at 410m (curvature decreases from 60mm to 52mm), and a slight increase in the attenuation coefficient at 890m (reaching 0.21dB / km). These locations were integrated with four natural boundary points to form a set of candidate boundaries. After eliminating non-overlapping boundaries, a total of seven candidate points were identified (180m, 220m, 410m, 550m, 630m, 780m, and 890m). The candidate boundary was divided into 8 initial segments, and the constraints were checked one by one: the 180m-220m segment (40m) met all homogeneous constraints, the 410m-550m segment (140m) had a refractive index fluctuation of 0.0008 (≤ threshold), and only the 890m-1000m segment (110m) had an attenuation coefficient fluctuation of 0.025dB / km (exceeding the threshold). It was further divided into two segments, 890m-950m and 950m-1000m, and finally 9 fiber segments were obtained, with a length range of 30m-140m.

[0060] Constructing segment entities: Using F01-F09 as segment codes (unique IDs), F01 (0m-180m) is associated with the following attributes: average physical characteristics (core diameter 8.9μm, attenuation coefficient 0.19dB / km), length 180 meters, start and end boundary IDs (N01, N02), and carries core power supply services; F02 (180m-220m) carries ordinary monitoring services, and the remaining segments are associated with corresponding attributes according to this rule. Constructing node entities: Generating node IDs (N01-N10) based on boundary position coordinates, N02 (180m) is labeled as "characteristic mutation node - refractive index gradient change", and N03 (220m) is labeled as "natural boundary node - fusion point", and the associated segments F02 (upstream) and F03 (downstream) of N03 are recorded simultaneously.

[0061] Establishing a correlation mapping: F01-F02 are connected in series via N02, and F05 (550m-630m) is derived from N06 (630m, branch point) to form F06 (parallel backup link, carrying redundant services), forming a structure with series connection as the main component and single-branch parallel connection. Layered topology labeling: F01, F04 (410m-550m), and F07 (630m-780m) form the core topology layer; F02, F03, F05, and F08 form the ordinary topology layer; and F06 (backup) and F09 form the idle layer, completing the fiber optic topology model construction.

[0062] The reference light propagation speed is calculated based on the average physical characteristics of each segment. Combined with the basic formula for the propagation speed of quartz optical fiber, the reference speed of segment F01, with a refractive index of 1.468, is approximately 2.04 × 10^8 m / s. The reference speed of segment F08 (890m-950m) is slightly reduced to 2.03 × 10^8 m / s due to its slightly higher attenuation coefficient. The reference speeds of the remaining segments are between 2.03 × 10^8 and 2.04 × 10^8 m / s, and are synchronously associated with the corresponding segment entities.

[0063] Step S200: Collect the temperature and optical signal sequences of the entire optical fiber link; based on the optical fiber topology model, match the optical signal sequences to the corresponding topology segments one by one according to the optical fiber segments;

[0064] Specifically, for each set of optical signal sequences, the physical location on the corresponding optical fiber is calculated using the following formula: Where X is the signal position. The reference position for the data acquisition device. This is the initial reference optical speed for the corresponding link. This refers to the signal propagation time difference;

[0065] Traverse all fiber segments in the fiber topology model and construct segment codes and location intervals. The mapping table, where, The starting position coordinates, The coordinates are the termination position coordinates; for X, an interval determination algorithm is used to match one by one, and based on the matching results, the signal position is corrected by the reference light speed of the corresponding segment, and multi-feature redundancy verification is performed.

[0066] Furthermore, the interval determination algorithm is as follows: when there exists a fiber segment that satisfies... When X exceeds all segment intervals or falls on a segment boundary, it is marked as a boundary signal and temporarily assigned to the upstream segment; thus, the matching result is obtained.

[0067] Based on the matching results, the reference optical propagation speed of the corresponding fiber segment is used. Correct signal position, This represents the precise reference optical velocity of the i-th segment; the correction formula is as follows: The matching results are then resubmitted into the interval determination algorithm to verify the matching results. For the corrected signal position; when If the match still falls within the original paragraph range, then the matching relationship is confirmed; when... When deviating from the original paragraph, calculate the error between the two positions. ,when When, retain the original match, when Then, re-traverse the topological paragraph matching. Calculate the error threshold for the location;

[0068] For a single optical signal sequence, the correlation coefficient r is calculated by combining the mean physical characteristics of the corresponding segment, using the following formula: Where S is the optical signal attenuation characteristic value, COV is the covariance, and VAR is the variance. The attenuation coefficient is used. When r is not less than the strong correlation threshold, the match is confirmed to be valid. When r is less than the strong correlation threshold, check whether it is signal interference or a matching error, and re-execute the position calculation and interval matching.

[0069] In one specific embodiment, a distributed optical fiber thermometer was used to collect end-to-end data from a 1000-meter quartz optical fiber in S100. The reference position of the acquisition device was aligned with the fiber optic start point (X0=0m), the time step was 100ms, and a total of 100 sets of time-series data were collected. The temperature sequence ranged from 25.3℃ to 31.8℃, and the optical signal sequence included propagation time difference and attenuation characteristic value, where the propagation time difference ranged from 0.8μs to 4.9μs, and the attenuation characteristic value corresponded to 0.18dB / km to 0.22dB / km.

[0070] Taking three typical optical signals as examples, and using the average reference optical velocity of 2.035 × 10^8 m / s in the S100 as the initial velocity, the physical positions are calculated. Signal 1 has a propagation time difference of 0.8 μs, corresponding to a position of approximately 163.2 m; signal 2 has a time difference of 2.2 μs, corresponding to a position of approximately 447.7 m; and signal 3 has a time difference of 4.1 μs, corresponding to a position of approximately 834.4 μs. The attenuation characteristic values ​​of each signal are recorded simultaneously.

[0071] The segment-location interval mapping table of the S100 topology model (e.g., F01: 0-180m, F04: 410-550m, F07: 630-780m) is called, and the interval determination algorithm is used for matching. Signal 1 (163.2m) falls in the F01 interval and is directly matched with F01; signal 2 (447.7m) is matched with F04; signal 3 (834.4m) exceeds F07 (630-780m) and is temporarily assigned to the upstream F07 (boundary signal).

[0072] The position calculation error threshold ε = 0.1m was set, and the corresponding segment's precise reference light velocity was corrected. Signal 1 matched F01 (velocity 2.04 × 10^8 m / s), and the corrected position was 163.3m, still within the F01 range, confirming the match; Signal 3 was corrected to a position of 832.9m, deviating from F07, with an error of 0.15m > ε, and was rematched to F08 (780-890m).

[0073] A strong correlation threshold of 0.7 was set, and the correlation between the signal and the segment attenuation coefficient was calculated. Signal 1 corresponds to an attenuation coefficient of 0.19 dB / km (F01) with a correlation coefficient of 0.75, indicating a valid match. After correction, signal 3 corresponds to an attenuation coefficient of 0.21 dB / km (F08) with a coefficient of 0.72, also indicating a valid match. Another signal had a coefficient of 0.65, which was identified as electromagnetic interference. After recalculating the match, the coefficient met the requirements.

[0074] Ultimately, all 100 signal groups were matched, with 97 groups achieving valid matching on the first attempt and 3 groups meeting the criteria after correction or rematching. These corresponded to segments F01-F09 respectively and were synchronously associated with the corresponding segments in the topology model, forming mapping data between segments and signal sequences.

[0075] Step S300: Construct an AI time-series feature classification model to filter anomalous topological segments based on temperature and light signal sequences; locate fiber optic breakpoints in anomalous topological segments;

[0076] Specifically, based on temperature and light signal sequences, temporal features are extracted, including time-domain features, trend-domain features, and cross-feature correlation features; feature concatenation, redundant feature removal, and feature weight allocation are then performed.

[0077] A hybrid model architecture combining a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is used to construct an AI temporal feature classification model. The fused temporal feature vector is input into the LSTM layer to capture the long-term and short-term dependencies of temperature and light signals over time. The CNN layer is used to enhance the feature extraction capability for local signal mutations caused by breakpoints. The Sigmoid activation function is used in the output layer to output the anomaly probability value of a single paragraph. The closer the probability is to 1, the higher the confidence of judging it as an anomalous paragraph. The model is trained and regularized to filter out anomalous topological paragraphs.

[0078] Furthermore, the optical signal data is grouped according to the encoding of the abnormal topology segment, retaining the matching optical signals within the corresponding segment and eliminating cross-segment interference signals and low-correlation signals; a distance conversion algorithm is used to calculate the actual distance of the optical signal from the transmitter to the detection point by combining the reference optical propagation speed of the abnormal segment with the optical signal propagation time difference; using the reference position of the acquisition device as a reference, the specific position of the detection point on the optical fiber is calculated, and several suspected breakpoint locations are identified; signal groups with abrupt changes in propagation time difference and a sudden drop in scattered light intensity are screened out, and through a clustering algorithm, suspected breakpoint locations with consistent signal characteristics are grouped into one category, duplicate results are merged, and the optical fiber breakpoint is located.

[0079] In one specific embodiment, based on 100 sets of time-series data collected by the S200, three types of features are extracted: in the time domain, the average temperature is 28.5℃, the extreme values ​​of optical signal attenuation are 0.18-0.22dB / km, and the number of signal abrupt changes in each segment is 1-3; in the trend domain, the temperature trend slope is obtained through linear fitting, which is 0.02℃ / step, and the rate of change of optical signal attenuation is 0.001dB / (km·step); the correlation between temperature and attenuation is obtained through cross-feature correlation, which is 0.68. After splicing the features, constant features with variance less than 0.0001 are removed, and the weight of the signal abrupt change features in the core segments (F01, F04, F07) is increased by 20%, forming a multi-dimensional feature vector.

[0080] A hybrid architecture of "2-layer LSTM + 1-layer CNN" is adopted. The LSTM layer has 64 units to capture temporal dependencies, and the CNN layer uses 3×3 convolutional kernels to enhance the extraction of local mutations. The output layer uses a sigmoid function to output the anomaly probability. The dataset is divided into a 7:2:1 ratio. The Adam optimizer has an initial learning rate of 0.001 and a dropout rate of 0.2 to suppress overfitting. The model converges after 20 epochs of training, with a recall rate of 96% for core paragraph anomaly identification. The anomaly thresholds for core paragraphs are set at 0.6 and for ordinary paragraphs at 0.7.

[0081] The feature vectors of each paragraph were input into the model. The abnormal probability of F04 (410-550m) was 0.78, and the abnormal probability of F08 (890-950m) was 0.82, both exceeding the corresponding thresholds, and were judged as abnormal paragraphs. The probabilities of the remaining paragraphs were 0.12-0.55, all of which were normal. The core abnormal features of F04 signal attenuation abrupt change and F08 propagation time difference disorder were recorded simultaneously.

[0082] Optical signals are grouped according to F04 and F08 encoding. F04 retains 12 matching signals and F08 retains 10 matching signals. Two cross-segment interference signals and one low-correlation (r=0.63) signal are removed to ensure that the signals involved in the positioning all come from the abnormal segment of the target.

[0083] The reference velocity for F04 is 2.038 × 10^8 m / s. Based on the propagation time difference, three potential locations were calculated: 432.1 m, 432.3 m, and 456.7 m. The reference velocity for F08 is 2.03 × 10^8 m / s, yielding two potential locations: 856.2 m and 856.4 m. Signal groups with abrupt changes in propagation time difference and sudden drops in scattering intensity were selected. Clustering algorithms were used to merge similar locations; F04 was merged into a single location at 432.2 m (1 breakpoint), and F08 was merged into a single location at 856.3 m (1 breakpoint).

[0084] Finally, one fiber optic breakpoint was located at 432.2m in segment F04 and 856.3m in segment F08. Both were located within the corresponding segment intervals, consistent with the abnormal signal characteristics and physical property fluctuation patterns, and were synchronously associated with the abnormal segment in the topology model.

[0085] Step S400: Analyze the impact mechanism of fiber optic breakpoints on optical signal propagation and establish a causal link; based on the fiber optic topology model and the causal link, deduce the scope of the abnormal impact caused by the breakpoint and determine the processing priority of the fiber optic breakpoint.

[0086] Specifically, for completely blocking breakpoints, a signal attenuation analysis algorithm is used to determine the degree of interruption in the optical signal propagation path and to identify the mechanism by which the transmitter signal cannot penetrate the breakpoint; for partially damaged breakpoints, a propagation speed fluctuation algorithm is used to analyze the signal speed shift or scattering intensity distortion mechanism caused by changes in the fiber structure at the breakpoint; based on the identified impact mechanism, a feature correlation algorithm is used to establish a causal link.

[0087] Furthermore, the impact of the serial segment is propagated downstream along the causal link to determine the interruption coverage area; the parallel segment uses a redundancy assessment algorithm to determine whether the backup link can compensate and define the impact boundary; a multi-dimensional assessment algorithm is adopted, combining the size of the impact range, the proportion of core business, and the degree of signal distortion, to quantify the impact level of each interruption point and distinguish between core and secondary impacts; based on the quantification results, a weighted sorting algorithm is used to prioritize interruptions without redundant links and those affected by core business, followed by interruptions with backup links and those affected by non-core business; the processing priority of fiber optic interruptions is determined.

[0088] In one specific embodiment, the S300 positioning results are invoked to identify one breakpoint each at 432.2m of segment F04 and 856.3m of segment F08. The physical characteristics, optical signal anomaly data, and S100 topology information of the two segments are extracted simultaneously. F04 is a core layer series segment (carrying power supply services), F08 is a general layer series segment (carrying monitoring services), and the backup link F06 can cover some services downstream of F04.

[0089] Signal attenuation analysis of the F04 breakpoint showed an upstream attenuation of 0.19 dB / km and a downstream attenuation that spiked to 0.8 dB / km, classifying it as a complete blockage breakpoint. The mechanism was that a fracture in the fiber core formed a physical barrier, preventing the transmitting signal from penetrating and causing downstream signal interruption. Propagation velocity fluctuation analysis of the F08 breakpoint showed a reference velocity of 2.03 × 10⁸ m / s, with the velocity shifting to 2.01 × 10⁸ m / s at the breakpoint, and a scattering intensity distortion rate of 15%, classifying it as a partially damaged breakpoint. The mechanism was that damage to the fiber sheath caused structural changes, resulting in signal velocity shift and scattering distortion.

[0090] Two links are constructed using a feature association algorithm: F04 (complete blocking type) → core layer fracture → signal attenuation spikes → downstream F05 segment signal interruption; F08 (partial damage type) → skin damage → velocity shift + scattering distortion → signal timing disorder, without affecting the basic transmission of the downstream F09 segment. The link labeling algorithm is verified to ensure logical closed loop.

[0091] In series segment transmission by link: the F04 breakpoint affects the entire downstream F05 segment (550-630m), covering a length of 80m; the F08 breakpoint only affects its own segment (890-950m), with no obvious abnormalities in the downstream F09 signal. Parallel redundancy assessment: the backup link F06 can compensate for 60% of the service in the F05 segment, while F08 has no backup link. Therefore, the final impact range of F04 is defined as 80m downstream of F04 (partially compensateable), and the impact range of F08 is 60m of itself (no compensation).

[0092] F04 has a core business impact of 100%, an impact range of 80m, and a signal distortion rate of 100%, with a quantification score of 9. F08 has a core business impact of 0%, an impact range of 60m, and a signal distortion rate of 40%, with a quantification score of 4. After weighted sorting, the F04 breakpoint (without full compensation and core business affected) has a higher priority than the F08 breakpoint. Therefore, the processing order is determined to be F04 breakpoint first, and F08 breakpoint processed later.

[0093] The breakpoints at 432.2m in paragraph F04 (first priority) and 856.3m in paragraph F08 (second priority) are recorded synchronously, including their impact mechanism, scope, and priority basis, and are linked to the topology model and breakpoint dataset.

[0094] like Figure 2 As shown, this application provides an artificial intelligence-based fiber optic temperature measurement data management system, including:

[0095] The reference light propagation speed calculation module includes: an optical fiber segmentation unit that divides the optical fiber into several segments based on its physical characteristics; a topology model construction unit that constructs an optical fiber topology model based on the segmented optical fibers; and a reference light propagation speed calculation unit that calculates the reference light propagation speed of each segment based on its physical characteristics.

[0096] Topology segment matching module: includes: a data acquisition unit that collects the temperature and optical signal sequences of the entire optical fiber link; and a topology segment matching unit that matches the optical signal sequences to the corresponding topology segments one by one according to the optical fiber topology model.

[0097] Fiber optic breakpoint location module: includes: an abnormal topology segment screening unit that constructs an AI time-series feature classification model to screen abnormal topology segments based on temperature and optical signal sequences; and a fiber optic breakpoint location unit that locates fiber optic breakpoints within abnormal topology segments.

[0098] The processing priority determination module includes: a causal link establishment unit that analyzes the impact mechanism of fiber optic breakpoints on optical signal propagation and establishes causal links; and a processing priority determination unit that, based on the fiber optic topology model and causal links, extrapolates the range of abnormal impacts caused by the breakpoints and determines the processing priority of the fiber optic breakpoints.

[0099] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A fiber optic temperature measurement data management method based on artificial intelligence, characterized in that, Includes the following steps: Based on the physical properties of optical fiber, the optical fiber is divided into several segments; an optical fiber topology model is constructed based on the divided optical fiber segments; and the reference light propagation speed of each segment is calculated based on the physical properties of each segment. The temperature and optical signal sequences of the entire optical fiber link are collected; based on the optical fiber topology model, the optical signal sequences are matched to the corresponding topology segments one by one according to the optical fiber segments; Construct an AI time-series feature classification model to filter anomalous topological segments based on temperature and light signal sequences; locate fiber optic breakpoints within anomalous topological segments. Analyze the impact mechanism of fiber optic breakpoints on optical signal propagation and establish a causal link; Based on the fiber optic topology model and causal links, the scope of the abnormal impact caused by the breakpoint is deduced, and the processing priority of the fiber optic breakpoint is determined.

2. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The process of dividing the optical fiber into several segments based on its physical properties includes: The physical characteristics of optical fiber include material, core diameter and cladding size, attenuation coefficient, refractive index distribution, and curvature of the laying path; fiber segmentation constraints are set, including homogenization constraints and boundary adaptation constraints; the homogenization constraint specifically requires that the fluctuation range of all physical characteristics within the same segment be controlled within a preset threshold; the boundary adaptation constraint specifically requires that segment boundaries be aligned with the natural boundaries of the optical fiber to avoid the distance between adjacent boundaries being less than the optical fiber length corresponding to the temperature measurement accuracy; The locations of abrupt changes in physical properties and the natural boundaries of the optical fiber are integrated into a set of candidate boundaries, and duplicate boundaries are removed. The entire optical fiber is divided into several initial segments using the candidate boundaries as separators. The physical property parameters of each segment are checked one by one to see if they meet the optical fiber segmentation constraints. Optical fiber segments that do not meet the constraints are further divided.

3. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The fiber optic topology model constructed based on the partitioned fiber segments includes: Using segment codes as unique IDs, and associating attributes including average physical characteristics, length, start and end boundary IDs, and service type, segment entities are constructed. Node IDs are generated based on segment boundary coordinates, and attributes are labeled according to boundary type. The connected segment IDs at the nodes are recorded, constructing node entities. Each segment's start and end boundary corresponds to a unique node, and a node can be associated with one or more segments. When two optical fiber segments share the same node, it is determined that there is a direct connection relationship, and the connection type, including series and parallel, is labeled synchronously. Perform association mapping between nodes and segments, label the topology links in layers according to business priority, form a core topology layer by combining core business carrying segments and corresponding nodes, form a normal topology layer by combining non-core business segments, and form an idle layer by combining isolated segments, thus forming a layered topology structure and constructing an optical fiber topology model.

4. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The process of matching the optical signal sequence to the corresponding topology segment one by one according to the fiber optic topology model includes: For each set of optical signal sequences, calculate its corresponding physical location on the optical fiber using the following formula: Where X is the signal position. The reference position for the data acquisition device. This is the initial reference optical speed for the corresponding link. This refers to the signal propagation time difference; Traverse all fiber segments in the fiber topology model and construct segment codes and location intervals. The mapping table, where, The starting position coordinates, The coordinates are the termination position coordinates; for X, an interval determination algorithm is used to match one by one, and based on the matching results, the signal position is corrected by the reference light speed of the corresponding segment, and multi-feature redundancy verification is performed.

5. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 4, characterized in that, For X, an interval determination algorithm is used for match-by-match. Based on the match results, the signal position is corrected according to the reference light velocity of the corresponding segment, and multi-feature redundancy verification is performed, including: The interval determination algorithm is as follows: when there exists a fiber segment that satisfies... When X exceeds all segment intervals or falls on a segment boundary, it is marked as a boundary signal and temporarily assigned to the upstream segment; thus, the matching result is obtained. Based on the matching results, the reference optical propagation speed of the corresponding fiber segment is used. Correct signal position, This represents the precise reference optical velocity of the i-th segment; the correction formula is as follows: The matching results are then resubmitted into the interval determination algorithm to verify the matching results. For the corrected signal position; when If the match still falls within the original paragraph range, then the matching relationship is confirmed; when... When deviating from the original paragraph, calculate the error between the two positions. ,when When, retain the original match, when Then, re-traverse the topological paragraph matching. Calculate the error threshold for the location; For a single optical signal sequence, the correlation coefficient r is calculated by combining the mean physical characteristics of the corresponding segment, using the following formula: Where S is the optical signal attenuation characteristic value, COV is the covariance, and VAR is the variance. The attenuation coefficient is used. When r is not less than the strong correlation threshold, the match is confirmed to be valid. When r is less than the strong correlation threshold, check whether it is signal interference or a matching error, and re-execute the position calculation and interval matching.

6. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The construction of the AI ​​time-series feature classification model, which filters anomalous topological segments based on temperature and light signal sequences, includes: Based on temperature and light signal sequences, temporal features are extracted, including time domain features, trend domain features, and cross-feature correlation features; feature concatenation, redundant feature removal, and feature weight allocation are then performed. A hybrid model architecture combining a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN) is used to construct an AI temporal feature classification model. The fused temporal feature vector is input into the LSTM layer to capture the long-term and short-term dependencies of temperature and light signals over time. The CNN layer is used to enhance the feature extraction capability for local signal mutations caused by breakpoints. The Sigmoid activation function is used in the output layer to output the anomaly probability value of a single paragraph. The closer the probability is to 1, the higher the confidence of judging it as an anomalous paragraph. The model is trained and regularized to filter out anomalous topological paragraphs.

7. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The fiber optic breakpoints in the location anomaly topology segment include: Optical signal data are grouped according to the coding of abnormal topology segments, retaining matching optical signals within the corresponding segments and eliminating cross-segment interference signals and low-correlation signals. A distance conversion algorithm is used to calculate the actual distance of the optical signal from the transmitter to the detection point by combining the reference optical propagation speed of the abnormal segment with the optical signal propagation time difference. Using the reference position of the acquisition equipment as a reference, the specific position of the detection point on the optical fiber is calculated, and several suspected breakpoint locations are identified. Signal groups with abrupt changes in propagation time difference and sudden drops in scattered light intensity are screened out. Through a clustering algorithm, suspected breakpoint locations with consistent signal characteristics are grouped into one category, duplicate results are merged, and the optical fiber breakpoint is located.

8. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The analysis of the impact mechanism of fiber optic breakpoints on optical signal propagation establishes a causal link, including: For completely blocked breakpoints, a signal attenuation analysis algorithm is used to determine the degree of interruption in the optical signal propagation path and to identify the mechanism by which the transmitter signal cannot penetrate the breakpoint. For partially damaged breakpoints, a propagation velocity fluctuation algorithm is used to analyze the mechanism of signal velocity shift or scattering intensity distortion caused by changes in the fiber structure at the breakpoint. Based on the identified impact mechanism, a feature correlation algorithm is used to establish a causal link.

9. The fiber optic temperature measurement data management method based on artificial intelligence according to claim 1, characterized in that, The process of extrapolating the scope of anomalies caused by fiber optic breakpoints based on fiber optic topology models and causal links, and determining the processing priority of fiber optic breakpoints includes: The impact of the serial segment is propagated downstream along the causal link to determine the coverage area of ​​the interruption; the parallel segment uses a redundancy assessment algorithm to determine whether the backup link can compensate and to define the impact boundary; a multi-dimensional assessment algorithm is used to quantify the impact level of each interruption point by combining the size of the impact range, the proportion of core business, and the degree of signal distortion, and to distinguish between core and secondary impacts; based on the quantification results, a weighted sorting algorithm is used to prioritize interruptions without redundant links and those affected by core business, followed by interruptions with backup links and those affected by non-core business; the processing priority of fiber optic interruptions is determined.

10. An artificial intelligence-based fiber optic temperature measurement data management system, using the artificial intelligence-based fiber optic temperature measurement data management method according to any one of claims 1-9, characterized in that, include: The reference light propagation speed calculation module includes: an optical fiber segmentation unit that divides the optical fiber into several segments based on its physical characteristics; a topology model construction unit that constructs an optical fiber topology model based on the segmented optical fibers; and a reference light propagation speed calculation unit that calculates the reference light propagation speed of each segment based on its physical characteristics. Topology segment matching module: includes: a data acquisition unit that collects the temperature and optical signal sequences of the entire optical fiber link; and a topology segment matching unit that matches the optical signal sequences to the corresponding topology segments one by one according to the optical fiber topology model. Fiber optic breakpoint location module: includes: an abnormal topology segment screening unit that constructs an AI time-series feature classification model to screen abnormal topology segments based on temperature and optical signal sequences; and a fiber optic breakpoint location unit that locates fiber optic breakpoints within abnormal topology segments. The processing priority determination module includes: a causal link establishment unit that analyzes the impact mechanism of fiber optic breakpoints on optical signal propagation and establishes causal links; and a processing priority determination unit that, based on the fiber optic topology model and causal links, extrapolates the range of abnormal impacts caused by the breakpoints and determines the processing priority of the fiber optic breakpoints.

Images