Network positioning method and system based on topology correlation analysis

Abstract

The application discloses a network positioning method and system based on topological correlation analysis, relates to the technical field of communication network fault positioning, and comprises the following steps: acquiring a receiving optical power sequence, a forward error correction count sequence, a logical topological relationship and physical resource organization data of a network branch, calculating a branch drift sequence, and establishing a membership relationship between the network branch and a candidate physical positioning unit; generating a common mode drift sequence based on a nearest common ancestor, and performing common mode processing to calculate a selective synchronization coefficient; further calculating a selective interpretation degree of the candidate physical positioning unit, and combining a hierarchical search graph and a swarm intelligence algorithm to obtain an optimal physical positioning unit; finally, determining an affected branch set and a relative confidence degree, and outputting a positioning result. The application can realize fine positioning of a local physical abnormal position of a PON tree-shaped optical network.

Description

Technical Field

[0001] This invention relates to the field of communication network fault location technology, and in particular to a network location method and system based on topology correlation analysis. Background Technology

[0002] In scenarios such as building access, underground utility tunnel access, and fiber-to-the-building in industrial parks, PON tree-shaped optical networks typically consist of central office equipment, splitter nodes, junction boxes, fiber trays, corner locations, and user-side optical network units. At the logical level, the network presents a hierarchical tree structure, while at the physical level, there are local organizational locations such as fiber trays within junction boxes, tray transitions, and corner entry / exit points. Since different network branches may share the same splitter node upstream and may share the same junction box, fiber tray, and corner location downstream, when a local location experiences slight bending pressure, moisture disturbance, stress release, or changes in fiber tray condition, multiple network branches passing through that local location will exhibit a synchronous change in received optical power decreasing and forward error correction pressure increasing. This synchronous change does not necessarily extend to the entire common upstream subtree but often only occurs in a portion of network branches sharing the same local physical organizational location. Therefore, how to accurately locate this local selective synchronization anomaly in a PON tree-shaped optical network by combining logical topology and physical resource organization has become a critical technical problem that needs to be solved in network operation and maintenance.

[0003] Existing network location methods typically rely on common upstream nodes, common splitting paths, or overall alarm associations to perform unified backtracking analysis of synchronization anomalies across multiple network branches. While these methods can identify anomalies in upper-layer common links or unified anomalies within the entire subtree, they tend to push the anomaly location back to a common ancestor node or common splitting node for anomalies occurring only in local physical locations. This results in location results remaining at a coarse-grained logical level, unable to be further compressed to the actual anomaly locations such as junction boxes, fiber trays, and corners. Furthermore, existing methods often struggle to effectively distinguish between overall drift caused by common upstream paths and selective synchronization changes caused by local physical locations. This can easily lead to local anomalies being overwhelmed by the overall phenomenon, resulting in an excessively large location range, too many nodes to investigate, and low on-site processing efficiency. Consequently, these methods fail to meet the actual needs of PON tree-structured optical networks for refined identification of the actual locations of local anomalies. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies, such as inaccurate location of local physical anomalies, misjudging selective synchronization anomalies as common upstream node anomalies, coarse location granularity, and large on-site investigation scope. Therefore, this invention proposes a network location method and system based on topology association analysis.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: Network localization methods based on topology association analysis include: S1. Obtain the received optical power sequence, forward error correction count sequence, logical topology relationship and physical resource organization data of the network branch; calculate the branch drift sequence based on the received optical power sequence and forward error correction count sequence; and establish the membership relationship between the network branch and the candidate physical location unit based on the physical resource organization data. S2. Based on the logical topology, determine the nearest common ancestor between each pair of network branches, generate a common-mode drift sequence, perform de-common-mode processing on the branch drift sequence to obtain a residual sequence, and calculate the selective synchronization coefficient between each pair of network branches based on the residual sequence. S3. Calculate the selective interpretability of each candidate physical location unit based on membership, logical topology, and selective synchronization coefficient. S4. Organize each candidate physical location unit into a hierarchical search graph, and use a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit. S5. Determine the set of affected branches corresponding to the optimal physical location unit based on the membership relationship, calculate the relative confidence level corresponding to the optimal physical location unit based on the selective interpretation degree, and output the location result.

[0006] Preferably, the physical resource organization data is used to characterize the transit relationships between each network branch and the splitter node, junction box, fiber tray and corner position, and the candidate physical positioning unit is composed of a combination of junction box, fiber tray and corner position.

[0007] Preferably, calculating the branch drift sequence based on the received optical power sequence and the forward error correction counting sequence includes: The median and absolute median difference of the received optical power sequence corresponding to each network branch, as well as the median and absolute median difference of the forward error correction count sequence, are determined respectively. The received optical power sequence is normalized based on the median and absolute median difference to obtain a first drift amount characterizing the degree of optical power reduction. The forward error correction count sequence is normalized based on the median and absolute median difference to obtain a second drift amount that characterizes the degree of increase in error correction pressure. The first drift amount and the second drift amount are combined at each time point to obtain the branch drift sequence of the corresponding network branch.

[0008] Preferably, the nearest common ancestor between each pair of network branches is determined based on logical topological relationships, a common-mode drift sequence is generated, and the branch drift sequence is subjected to de-common mode processing to obtain a residual sequence, including: For any two network branches, find the nearest common ancestor of the two network branches based on the logical topological relationship; Extract all leaf branches under the nearest common ancestor and obtain the branch drift sequence corresponding to each leaf branch; Median aggregation is performed on the branch drift sequences corresponding to all leaf branches at the same time to generate common mode drift sequences corresponding to any two network branches; Subtract the common-mode drift sequence from the branch drift sequence corresponding to each of the two network branches to obtain the residual sequence corresponding to each of the two network branches.

[0009] Preferably, the selective synchronization coefficients between each pair of network branches are calculated based on the residual sequence, including: Determine the first-order transformation sequence of the residual sequence corresponding to each of any two network branches; Based on the degree of correlation consistency between the first-order change sequences of any two network branches, the selective synchronization coefficient between any two network branches is calculated.

[0010] Preferably, based on membership, logical topology, and selective synchronization coefficients, the selective interpretability of each candidate physical location unit is calculated, including: Extract the first branch set that passes through the target candidate physical location unit based on the aforementioned membership relationship; Based on the logical topological relationship, determine the smallest logical ancestor that covers the first branch set, and extract all leaf branches under the smallest logical ancestor to form the second branch set; Based on the selective synchronization coefficients between each pair of network branches within the first branch set and the number of corresponding branch pairs, the intra-group synchronization strength of the target candidate physical positioning unit is calculated. Based on the selective synchronization coefficient between network branches in the first branch set and network branches in the second branch set other than the first branch set, and the number of corresponding cross-set branch pairs, the out-of-group leakage intensity of the target candidate physical location unit is calculated. Based on the ratio of the number of the first branch set to the number of the second branch set, the selectivity coefficient of the target candidate physical location unit is calculated; Based on the intra-group synchronization strength, the inter-group leakage strength, and the selectivity coefficient, the selective interpretability of the target candidate physical location unit is obtained.

[0011] Preferably, the candidate physical location units are organized into a hierarchical search graph, including: Extract the junction box, fiber optic tray and corner combination information corresponding to each candidate physical positioning unit from the physical resource organization data; Establish parent-child relationships between the splitter node and the junction box, between the junction box and the fiber optic tray, and between the fiber optic tray and the corner unit; A hierarchical search graph for path search is constructed based on the parent-child relationship.

[0012] Preferably, the optimal physical location unit is obtained by using a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability, including: The ant colony algorithm is used to perform path search on the hierarchical search graph; Based on the selective interpretation degree of the candidate physical location units corresponding to each child node under the parent node, the heuristic value for the transfer from the parent node to each child node is determined, and the heuristic value is normalized within the range of all child nodes of the same parent node. The pheromone weight and heuristic weight corresponding to the parent node are determined based on the normalized heuristic value, and the probability of transition to each child node is calculated based on the pheromone weight, the heuristic weight, the path edge pheromone, and the heuristic value. During the ant colony algorithm iteration process, a corresponding transfer path is generated based on the transfer probability, and the pheromone of the path edge in the transfer path is updated based on the selective interpretability of the candidate physical location unit at the end reached by the ant. The optimal physical location unit is determined based on the selective interpretability of each candidate physical location unit at the end and the pheromone of the path edges in the corresponding transfer path.

[0013] Preferably, the relative confidence level corresponding to the optimal physical location unit is calculated based on the selective interpretation degree, and the location result is output, including: The relative confidence level of the optimal physical location unit is determined based on the selective interpretability of each candidate physical location unit. Based on the membership relationship, find all network branches that have a physical connection with the optimal physical location unit, and merge them to obtain the set of affected branches; The optimal physical location unit, the affected branch set, and the relative confidence level are combined to output the location result of the corresponding network physical anomaly location.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention acquires the received optical power sequence, forward error correction count sequence, logical topology relationship, and physical resource organization data of network branches. First, it constructs a branch drift sequence, and then generates a common-mode drift sequence based on the nearest common ancestor and performs de-common-mode processing. This removes the overall changes caused by common upstream paths, common splitting conditions, or overall link fluctuations, highlighting the remaining synchronization changes caused by local physical organization positions. Furthermore, by calculating the selective synchronization coefficients between network branches, this invention can identify synchronization drift relationships that only persist between some related branches. This enables accurate characterization of local anomaly impact patterns, avoids the masking effect of overall fluctuations on local anomaly identification, and improves the ability to identify selective synchronization anomalies.

[0015] 2. This invention further constructs intra-group synchronization strength, inter-group leakage strength, selectivity coefficient, and selective interpretability based on the membership relationship between network branches and candidate physical location units, the branch range under the least logical ancestor, and the selective synchronization coefficient. It then combines a hierarchical search graph and a swarm intelligence algorithm to search for candidate physical location units level by level, ultimately determining the optimal physical location unit and its corresponding set of affected branches and relative confidence. This allows the location results to be compressed from upper-level logical nodes to the local physical organizational locations corresponding to junction boxes, fiber optic trays, and corner positions, narrowing the scope of anomaly investigation, enhancing the consistency between the location results and the actual anomaly locations, and improving the precision and interpretability of network anomaly location. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a network localization method based on topology association analysis provided in an embodiment of the present invention. Figure 2 This is a functional block diagram of a network positioning system based on topology association analysis provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example: This example provides a network localization method based on topology association analysis. See [link to example]. Figure 1 Specifically, including: S1. Obtain the received optical power sequence, forward error correction count sequence, logical topology relationship and physical resource organization data of the network branch; calculate the branch drift sequence based on the received optical power sequence and forward error correction count sequence; and establish the membership relationship between the network branch and the candidate physical location unit based on the physical resource organization data. S2. Based on the logical topology, determine the nearest common ancestor between each pair of network branches, generate a common-mode drift sequence, perform de-common-mode processing on the branch drift sequence to obtain a residual sequence, and calculate the selective synchronization coefficient between each pair of network branches based on the residual sequence. S3. Calculate the selective interpretability of each candidate physical location unit based on membership, logical topology, and selective synchronization coefficient. S4. Organize each candidate physical location unit into a hierarchical search graph, and use a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit. S5. Determine the set of affected branches corresponding to the optimal physical location unit based on the membership relationship, calculate the relative confidence level corresponding to the optimal physical location unit based on the selective interpretation degree, and output the location result.

[0019] In one embodiment of the present invention, the process of acquiring the received optical power sequence, forward error correction count sequence, logical topology relationship, and physical resource organization data of a network branch, calculating the branch drift sequence based on the received optical power sequence and the forward error correction count sequence, and establishing the membership relationship between the network branch and candidate physical location units based on the physical resource organization data includes: In this embodiment, for each network branch in the PON tree optical network, compliant raw data within a continuous valid sampling period is obtained according to an equal sampling period matching the reporting capability of the PON network equipment. The raw data includes the downlink received optical power sequence and forward error correction count sequence on the optical network unit side corresponding to each network branch. Simultaneously, the logical topology and physical resource organization data corresponding to the PON tree optical network are obtained. The physical resource organization data characterizes the physical path relationships between each network branch and splitter nodes, junction boxes, fiber trays, and corner positions. The logical topology characterizes the parent-child hierarchical structure of each network branch in the PON tree network. The logical topology is constructed based on the hierarchical mapping between optical line terminal ports, splitter ports, and optical network unit identifiers. The physical resource organization data is constructed based on the fiber core full routing data of the ODN resource management system. The path relationship is determined by the physical nodes traversed by the branch fiber core throughout its entire journey. For each network branch i, its received optical power sequence is denoted as... Where t is the sampling time, The forward error correction counting sequence covers the optical power values ​​corresponding to all valid sampling times. The sampling time corresponds one-to-one with the sampling time of the received optical power sequence. The forward error correction counting sequence uses the error correction block count accumulated during the forward error correction process. Before sampling, the data validity is screened to remove invalid sampling data corresponding to the offline, fiber breakage, and power outage periods of network branches. At the same time, it is ensured that the number of valid sampling points meets the minimum number requirements for calculating the median and absolute median difference.

[0020] Determine the median and absolute median difference of the received optical power sequence for each network branch, as well as the median and absolute median difference of the forward error correction count sequence. For the received optical power sequence, calculate its corresponding median:

[0021] In the formula, med represents the median calculation operation. Let i be the received optical power sequence of network branch i. The median of this sequence is used instead of the arithmetic mean because the median is highly robust to anomalous jump values, avoiding interference from single bursts of optical power fluctuations on the baseline value and ensuring the stability of drift calculations. Further calculation of the absolute median difference corresponding to the received optical power sequence is as follows:

[0022] In the formula, The absolute median difference of the optical power sequence received by network branch i is used instead of the standard deviation because the absolute median difference also has strong anti-interference capabilities, can adapt to the non-Gaussian distribution characteristics of optical power data in PON networks, and avoids normalization scale distortion caused by extreme values. For boundary scenarios where the absolute median difference is 0, the minimum resolution of the sequence is used as a substitute scale value to ensure that the normalization calculation can be performed normally.

[0023] The received optical power sequence is normalized based on the median and absolute median difference to obtain the first drift, which characterizes the degree of optical power decrease. The formula for calculating the first drift is as follows:

[0024] In the formula, Let i be the first drift of network branch i at each sampling time. The molecule uses... reduce The reason is that when the received optical power is lower than the reference median, the calculated result is positive, which is in the same direction as the trend of optical power degradation and can directly reflect the magnitude of optical power degradation drift. The denominator is normalized using the absolute median difference, which can eliminate the difference in the reference values ​​of optical power between different network branches and achieve horizontal comparability of the degree of drift between different branches. For the forward error correction counting sequence, its corresponding median is calculated as follows:

[0025] In the formula, Let the median of the forward error correction count sequence of network branch i be used as the baseline value for error correction counting. Further calculation of the absolute median difference corresponding to the forward error correction count sequence is as follows:

[0026] In the formula, The absolute median difference of the forward error correction count sequence of network branch i is used as the normalization scale parameter. The median and absolute median difference are also used to calculate the baseline and scale values, which can ensure the anti-interference of error correction count drift calculation and avoid the impact of sudden bit error events on the baseline value. For the boundary scenario where the absolute median difference is 0, a fixed minimum scale value is used as a substitute to ensure that the normalization calculation can be executed normally.

[0027] The forward error correction count sequence is normalized based on the median and absolute median difference to obtain a second drift value characterizing the degree of increase in error correction pressure. The formula for calculating the second drift value is as follows:

[0028] In the formula, The molecule represents the second drift of network branch i at each sampling time. reduce The reason is that when the forward error correction count is higher than the baseline median, the calculation result is positive, which is in the same direction as the upward trend of error correction pressure and can directly reflect the degree of degradation of link transmission quality. The denominator is normalized using the absolute median difference, which can eliminate the difference in the baseline magnitude of error correction counts of different network branches and achieve same-scale fusion with optical power drift. The first drift and the second drift are added at each corresponding sampling time to obtain the branch drift sequence of the corresponding network branch. The calculation formula of the branch drift sequence is:

[0029] In the formula, The branch drift sequence values ​​of network branch i at each sampling time are summed and fused because both the decrease in optical power and the increase in error correction pressure are direct manifestations of link degradation. When the two are superimposed in the same direction, they can comprehensively reflect the overall degradation drift degree of the network branch. At the same time, the two normalized drift values ​​are in the same dimension, and there will be no deviation dominated by a single index after addition. They can reflect the synchronous changes of optical layer transmission performance and link error correction performance in a balanced way. After calculating the branch drift sequence of all network branches, the validity of the sequence is verified to confirm that the sequence can completely reflect the link degradation drift trend. The branch drift sequence that passes the verification can be used for subsequent common mode removal and selective synchronization coefficient calculation.

[0030] It should be noted that the PON tree optical network is used in building access, underground integrated pipe gallery access, and fiber-to-the-building scenarios in parks. It consists of central office equipment splitting nodes, junction boxes, fiber trays, corner entrances, and user-side optical network units. It uses passive optical branching devices to complete signal transmission and distribution. Logically, it presents a tree-like hierarchical structure with branches from top to bottom. Physically, it is a passive optical network that includes local physical organizational units such as junction boxes, fiber trays, and corner entrances. The network takes the central office optical line terminal as the root node and the user-side optical network units as the terminal leaf nodes, forming multi-level branch transmission paths based on the splitting nodes. A network branch refers to an actual transmission path unit in the PON tree optical network that extends downward from a common upstream node and eventually reaches one or more user-side terminals. It corresponds to a physical path of optical signal propagation along the optical fiber link, which includes both the independent downlink path after splitting and the physical organizational positions such as junction boxes, fiber trays, and corner entrances that the path passes through.

[0031] It should be noted that the received optical power sequence refers to the data set formed in chronological order after measuring the intensity of the light energy received at the end of the network branch at multiple consecutive moments. It reflects the changes in the arriving energy of light after being affected by beam splitting, attenuation, bending force, splicing status, and local micro-bending disturbances during propagation. The forward error correction count sequence refers to the data set formed after the receiver counts the number of correctable errors in the received signal at multiple consecutive moments. It reflects the changes in the error correction pressure caused by noise, attenuation fluctuations, symbol distortion, or short-term quality fluctuations during the transmission of the optical signal.

[0032] It should be noted that logical topology refers to the upstream and downstream affiliation and branching relationships of each network branch at the network connection layer. It describes which upstream node each branch shares and at which layer the separation occurs. This relationship corresponds to the organizational structure of the signal transmission from the central office through the splitter to each terminal. Physical resource organization data refers to the data characterizing the correspondence between each network branch and the splitter node, junction box, fiber tray, corner position, and its path. It is used to explain which physical carriers and deployment locations each branch actually passes through, thereby reflecting whether different branches share the same local stress environment or the same deployment structure. Branch drift sequence refers to the comprehensive characterization sequence of time evolution based on the changes in received optical power and forward error correction. It is used to characterize the degree to which a single network branch deviates from its stable transmission state.

[0033] Based on the acquired physical resource organization data of the PON tree optical network and the completed candidate physical location unit set, the affiliation relationship between network branches and candidate physical location units is constructed. The physical resource organization data comes from the full routing ledger data of the ODN resource management system, including the hierarchical association information of all splitter nodes, junction boxes, fiber trays, and corner positions in the PON tree optical network, as well as the full fiber core routing information corresponding to each network branch. The full fiber core routing information completely records all physical nodes, splice points, fiber tray positions, and path directions traversed by a single network branch from the central office optical line terminal port to the user-side optical network unit. Corner positions are fixed corner positions on the fiber tray used for fiber core entry and exit winding. Each candidate physical location unit consists of a unique combination of junction boxes, fiber trays, and corner positions. Each combination of junction boxes, fiber trays, and corner positions with direct hierarchical association corresponds to an independent Candidate physical location units are directly hierarchically associated, meaning that corner units belong to corresponding fiber optic trays, and fiber optic trays belong to corresponding junction boxes, with no cross-level or cross-node combinations. All junction box, fiber optic tray, and corner unit combinations with complete direct hierarchical associations are extracted from the physical resource organization data. Invalid combinations without corresponding hierarchical associations are eliminated to form a set of candidate physical location units. Each candidate physical location unit in the set is assigned a unique identifier, completing the standardized construction of candidate physical location units. Simultaneously, the physical resource organization data is verified for integrity. The verification rule is that the entire fiber core routing information of a single network branch must completely cover all physical nodes from the optical line terminal port to the optical network unit. Invalid network branch data with missing routing information or unclear node relationships are eliminated, ensuring that the combination elements of each candidate physical location unit are complete and the routing information of each network branch is fully traceable.

[0034] For each verified network branch, extract all physical nodes and path information traversed by the entire fiber core route of the network branch from the physical resource organization data. Determine the relationship between the network branch and each candidate physical location unit. The determination rule for the relationship is that the fiber core of the network branch completes the fusion splicing or coiling operation in the corresponding junction box, and the fiber core passes through the corresponding coiling tray in the junction box throughout its entire journey. At the same time, the fiber core completes the entry and exit coiling at the corner corresponding to the coiling tray. When all three conditions are met, the network branch is determined to have passed through the corresponding candidate physical location unit. To achieve standardization and traceability in the subsequent calculation process, a binary method is used to define the membership relationship between the network branch and the candidate physical location unit. The formula for calculating the membership relationship is as follows:

[0035] In the formula, The membership value represents the relationship between network branch i and candidate physical location unit c. i represents any qualified network branch in the PON tree optical network, and c represents any candidate physical location unit in the candidate physical location unit set. The reason for using binary values ​​to define the membership relationship is that the physical association status between the network branch and the candidate physical location unit can be clearly and unambiguously determined, providing a standardized binary input for subsequent selective interpretability calculation, avoiding calculation deviations caused by fuzzy membership, and ensuring the accuracy and repeatability of branch set division. For boundary scenarios where a single network branch passes through multiple candidate physical location units, the membership value of each corresponding candidate physical location unit is set to 1. For network branches that do not match any candidate physical location units, all their membership values ​​are 0. For candidate physical location units that do not pass through any network branches, all their corresponding column membership values ​​are 0.

[0036] After determining and calculating the membership relationships between all network branches and all candidate physical location units, a complete membership relationship matrix is ​​formed. The rows of the membership relationship matrix correspond to all valid network branches, the columns correspond to all candidate physical location units, and the elements in the matrix are the membership relationship values ​​between the corresponding network branches and candidate physical location units.

[0037] It should be noted that a candidate physical location unit refers to a local physical organization location unit formed by the combination of junction boxes, fiber trays, and corner positions. This unit corresponds to the specific placement and transition position of the optical fiber in the local space. It is a local location where micro-bending, pressure, moisture disturbance, or stress concentration are more likely to actually occur and affect multiple related branches. The affiliation relationship refers to the correspondence between a certain network branch and a certain candidate physical location unit. This correspondence is used to indicate whether multiple branches share the same local physical path and the same local structural constraints.

[0038] In one embodiment of the present invention, the nearest common ancestor between each pair of network branches is determined based on logical topological relationships, a common-mode drift sequence is generated, the branch drift sequence is subjected to de-common mode processing to obtain a residual sequence, and the selective synchronization coefficient between each pair of network branches is calculated based on the residual sequence, including: Based on the obtained logical topology of the PON tree optical network and the calculated branch drift sequences of each network branch, common-mode drift sequence generation and residual sequence calculation are performed on any two network branches. The logical topology is a directed tree structure with the central office optical line terminal as the root node, and the hierarchy from top to bottom is optical line terminal, splitter node, and terminal network branch. The network branch is the terminal leaf node of the tree structure, corresponding to the link branch where the user-side optical network unit is located. First, for any two network branches, the lowest common ancestor of the two network branches is found based on the logical topology. The lowest common ancestor is defined as the common ancestor in the tree topology. The search process takes the two target network branches as the nodes of the descendant nodes that are farthest from the root node. First, it obtains the full-level ancestor node chain from its own node to the root node for each of the two target network branches. Then, it traverses upward from the end leaf node position of the two ancestor node chains to find the first common node, which is the target's nearest common ancestor. During the traversal, the search result is checked simultaneously to ensure that the obtained nearest common ancestor is the deepest level node that meets the conditions, avoiding tracing upward to higher-level ancestor nodes, which would lead to inaccurate common mode elimination. At the same time, in abnormal scenarios where the two target network branches are the same branch, the calculation of this group is terminated directly.

[0039] After finding the nearest common ancestor, all leaf branches under the nearest common ancestor are extracted. All leaf branches are defined as all terminal network branches within the subtree rooted at the nearest common ancestor. During extraction, set verification is performed simultaneously to ensure that both target network branches are included in the leaf branch set. Validity verification is performed on the network branches within the set. The verification rule is that the branch drift sequence of the network branch has a sampling time sequence consistent with the target sequence. For missing sampling points, median interpolation of adjacent time points is used to fill in the gaps. Branches with a missing sampling point ratio exceeding the valid sampling requirement are considered invalid and removed. This ensures that the sampling times of the branch drift sequences of all valid network branches within the set are aligned. The branch drift sequences corresponding to each of all valid leaf branches are obtained synchronously. Then, median aggregation is performed on the branch drift sequences corresponding to all valid leaf branches at the same time to generate the common-mode drift sequence corresponding to the two network branches. The formula for median aggregation is:

[0040] In the formula, Let be the common-mode drift sequence values ​​corresponding to network branch i and network branch j at time t. When the number of samples involved in the calculation is even, the average of the two middle values ​​is taken as the median result. Let be the branch drift sequence value of the k-th network branch within the leaf branch set at time t. The lowest common ancestor of network branch i and network branch j The set of all valid leaf branches under its jurisdiction, where t is the sampling time aligned with the branch drift sequence, uses the median for aggregation because the median is highly robust to sudden outliers in the branch drift sequence, preventing distortion of the common-mode drift sequence caused by sudden fluctuations in individual branches. It also accurately characterizes the overall common-mode drift trend of all branches within the nearest common ancestor subtree, achieving complete extraction of the impact on upper-layer common links. The reason for calculating a dedicated common-mode drift sequence for each pair of network branches is that different network branches have different nearest common ancestors. Calculating the common-mode sequence using only the subtree data of the corresponding nearest common ancestor accurately removes the common-mode impact of upper-layer common links directly related to that pair of network branches, avoiding excessive cancellation of local synchronization signals caused by unified common-mode removal across the entire network.

[0041] After generating the common-mode drift sequence, subtract the common-mode drift sequence from the branch drift sequences corresponding to the two network branches respectively to obtain the residual sequences corresponding to the two network branches. The formula for calculating the residual sequences is as follows:

[0042]

[0043] In the formula, Let i be the residual sequence value corresponding to network branch i at time t. Let j be the residual sequence value corresponding to network branch j at time t. Let i be the branch drift sequence value of network branch i at time t. The branch drift sequence value of network branch j at time t is used. The reason for subtracting the common-mode drift sequence from the branch drift sequence is that it can eliminate the common-mode drift influence caused by the common ancestor node and the upper-layer common link between the two network branches. The remaining residual sequence only retains the individual drift components of the two network branches that deviate from the overall trend of the whole subtree, thereby making the synchronization drift caused by local physical common factors other than the common ancestor explicit, and providing accurate input data for subsequent selective synchronization coefficient calculation. For the scenario where the nearest common ancestor is the root node optical line terminal and the boundary scenario where the effective leaf branch set under the nearest common ancestor only contains the two target network branches, the above median aggregation rule is still used to generate the common-mode drift sequence to ensure the uniformity and repeatability of the calculation rules.

[0044] It should be noted that the nearest common ancestor (LCA) refers to the common node that is simultaneously upstream of both network branches and closest to both branches in the logical topology. This node corresponds to the last upstream organization position that the two branches traverse together in the actual transmission path. After this point, the two branches begin to transmit in different directions. All leaf branches refer to the set of all terminal branches that extend downwards from the LCA and no longer branch, ultimately leading to each terminal. This set reflects all actual downlink transmission units covered by the common upstream node. Branch drift sequence refers to the data sequence formed over time representing the degree of deviation of a single network branch from its stable transmission state at multiple consecutive moments, used to characterize... The branch experiences overall state fluctuations under the combined effects of optical power attenuation fluctuations and error correction pressure changes; the common-mode drift sequence refers to the time series composed of the common changes of all the leaf branches at various times, reflecting the overall drift caused by sharing the same upstream transmission path, common splitting conditions, or common environmental disturbances; the residual sequence refers to the remaining change sequence obtained by subtracting the common-mode drift sequence from the branch drift sequences corresponding to any two network branches, reflecting the local deviation components still retained by each network branch after removing the common upstream influence, which can better characterize the differential changes caused by local physical organization location, local stress state, or local micro-bending disturbances.

[0045] Based on the calculated, time-complete, and validated residual sequences corresponding to any two network branches, first-order change sequence calculation and selective synchronization coefficient calculation are performed. A preliminary validity check is then performed on the residual sequences of the two target network branches. The check rules are that the sampling times of the two residual sequences are aligned and the sequence lengths are consistent. After passing the check, the first-order change sequence of the residual sequences corresponding to each of the two network branches is determined. The first-order change sequence is the difference sequence of the residual sequences at adjacent sampling times, used to characterize the instantaneous change trend of the residual sequences. The calculation formula for the first-order change sequence is:

[0046]

[0047] In the formula, Let be the first-order change sequence value corresponding to network branch i at time t. Let j be the first-order change sequence value corresponding to network branch j at time t. Let i be the residual sequence value corresponding to network branch i at time t. Let i be the residual sequence value corresponding to network branch i at time t-1. Let j be the residual sequence value corresponding to network branch j at time t. Let be the residual sequence value corresponding to network branch j at time t-1, and t be the sampling time aligned with the residual sequence. The value of t ranges from the second sampling time to the last sampling time, ensuring a uniform sampling time range for the first-order change sequence. For boundary scenarios where there is no preceding time in the first sampling time, the first-order change sequence is calculated directly from the second sampling time, ensuring the consistency and repeatability of the calculation rules. The reason for using the difference between adjacent time points to calculate the first-order change sequence is that it can eliminate the fixed baseline offset and long-term slow drift effects in the residual sequence, retaining only the instantaneous change trend of adjacent time points. The first-order change sequence can accurately remove irrelevant baseline interference, amplify the synchronous change characteristics between the two branches, and remove the common mode trend residue brought by the upper-layer common link, providing accurate input data for subsequent synchronization coefficient calculation.

[0048] After calculating the first-order change sequences of the two network branches, a validity check is performed on the first-order change sequences. The check rules are that the time lengths of the first-order change sequences are consistent, the number of valid sampling points meets the statistical requirements, and invalid sequences with zero variance are removed. After the check passes, the selective synchronization coefficient between the two network branches is calculated based on the correlation consistency between the first-order change sequences of the two network branches. The selective synchronization coefficient is quantified using the Pearson linear correlation coefficient, and the calculation formula is as follows:

[0049] In the formula, Let be the selective synchronization coefficient between network branch i and network branch j. This represents the summation of values ​​at all valid sampling times of the first-order change sequence. Let be the first-order change sequence value corresponding to network branch i at time t. Let t be the first-order change sequence value corresponding to network branch j at time t. The Pearson linear correlation coefficient is used as the selective synchronization coefficient because it can accurately characterize the degree of synchronization of the linear change trends between two first-order change sequences. Its value is fixed between -1 and +1. The closer the coefficient value is to +1, the stronger the synchronicity of the residual change trends of the two branches, indicating that the two branches exhibit synchronous drift away from the influence of the common mode of their common ancestor. The numerator is the sum of the products of the values ​​of the two sequences at the same time, which can quantify the degree of superposition of the change trends of the two sequences in the same direction. The denominator is the square root product of the sum of the squares of the two sequences. Normalization can eliminate the impact of differences in the numerical magnitudes of two sequences, ensuring that the synchronization coefficients between different branch pairs are horizontally comparable. For boundary scenarios where the variance of any first-order change sequence is 0 or both first-order change sequences are all zero sequences, the selective synchronization coefficient is directly determined to be zero, avoiding calculation errors where the denominator is zero. For scenarios where the calculated selective synchronization coefficient is negative, it is determined that the two branches have no synchronous drift in the same direction, and the synchronization coefficient is treated as zero. The selective synchronization coefficient can then be used to construct the selective synchronization matrix, providing core input data for the subsequent calculation of the selective interpretation degree of candidate physical positioning units.

[0050] It should be noted that the selective synchronization coefficient is a quantitative indicator used to characterize whether the remaining changes of any two network branches still maintain synchronous fluctuations after removing the common upstream influence. This indicator does not reflect whether the original states of the two branches change simultaneously, but rather whether there are still coordinated changes caused by more local common physical locations after removing the overall consistent changes caused by common ancestors, common trunk paths, or common spectral conditions. When two network branches pass through the same junction box, the same fiber tray, and the same corner, if the local location experiences slight bending and pressure, moisture disturbance, or stress release, the two branches may still exhibit residual fluctuations with consistent direction and similar speed after removing the common upstream changes. In this case, the selective synchronization coefficient is high, indicating that there is a selective synchronization relationship between the two branches caused by local shared physical paths. Conversely, if the two branches only share the common upstream path and do not share the downstream local physical location, their remaining changes usually no longer have obvious synchronicity after removing the common upstream changes. In this case, the selective synchronization coefficient is low. Therefore, this coefficient can be used to distinguish between overall synchronization caused by common ancestors and selective synchronization caused by local physical co-location.

[0051] In one embodiment of the present invention, the selective interpretability of each candidate physical location unit is calculated based on membership, logical topology, and selective synchronization coefficient, including: Based on the established membership matrix of network branches and candidate physical location units, the logical topology of the PON tree optical network, and the calculated selective synchronization coefficient matrix of all branch pairs, selective interpretability calculation is performed on the target candidate physical location unit. The first branch set passing through the target candidate physical location unit is extracted according to the membership relationship. The first branch set is a set of all network branches whose membership relationship with the target candidate physical location unit is 1. During the extraction process, the validity of the branches is simultaneously checked, and branches with incomplete selective synchronization coefficients and invalid time-series data are removed to ensure that all network branches in the first branch set have complete data required for calculation. For boundary scenarios where the first branch set is empty, the selective interpretability of the target candidate physical location unit is directly determined to be 0, and subsequent calculations are terminated.

[0052] After extracting and validating the first branch set, the minimum logical ancestor covering the first branch set is determined based on the logical topology. The minimum logical ancestor is defined as the node in the tree topology that is farthest from the root node while having all network branches in the first branch set as descendants. This is consistent with the definition of the nearest common ancestor in the previous steps. The search process first obtains the full-level ancestor node chain from its own node to the root node for each network branch in the first branch set. Then, the deepest common node in all ancestor node chains is found to be the target minimum logical ancestor. For boundary scenarios where the first branch set contains only a single network branch, the parent node of that network branch is directly determined as the minimum logical ancestor. After the search is completed, a verification is performed to ensure that the minimum logical ancestor can cover all network branches in the first branch set and that there are no deeper-level common ancestor nodes. This avoids tracing back to higher-level nodes, which could lead to distortion in the subsequent selection coefficient calculation. Then, all leaf branches under the minimum logical ancestor are extracted to form the second branch set. All leaf branches under the minimum logical ancestor are defined as all terminal network branches within the subtree range rooted at the minimum logical ancestor. After extraction, a set verification is performed to ensure that the first branch set is a complete subset of the second branch set, with no missing or mismatched branches. At the same time, it is verified that all network branches in the second branch set have complete selection synchronization coefficient data, and invalid branches are removed.

[0053] After constructing and verifying the two branch sets, the intra-group synchronization strength of the target candidate physical location unit is calculated based on the selective synchronization coefficients between each pair of network branches within the first branch set and the corresponding number of branch pairs. The formula for calculating the intra-group synchronization strength is as follows:

[0054] In the formula, The intra-group synchronization strength of the target candidate physical location unit c. This is the first branch set corresponding to the target candidate physical location unit c. This represents the number of network branches within the first branch set. Let be the selective synchronization coefficient between network branch i and network branch j. The summation term is the sum of the selective synchronization coefficients of all pairwise different network branch pairs within the first branch set. During the calculation, negative selective synchronization coefficients are uniformly treated as 0 in the summation calculation. For boundary scenarios where the number of network branches within the first branch set is less than 2, the intra-group synchronization strength is directly determined to be 0. The reason for using this formula to calculate the intra-group synchronization strength is that by taking the average of the selective synchronization coefficients of all intra-group branch pairs, the influence of the difference in the number of branches within the first branch set can be eliminated, achieving horizontal comparability of intra-group synchronization strength between different candidate physical positioning units. At the same time, it accurately quantifies the degree of synchronization drift consistency between the branches covered by the target candidate physical positioning unit. The higher the intra-group synchronization strength value, the more significant the synchronization drift characteristics of the branches within the unit.

[0055] After completing the intra-group synchronization strength calculation, based on the selective synchronization coefficients between network branches in the first branch set and network branches in the second branch set excluding the first branch set, and the corresponding number of cross-set branch pairs, the inter-group leakage strength of the target candidate physical location unit is calculated. The formula for calculating the inter-group leakage strength is as follows:

[0056] In the formula, The external leakage intensity of the target candidate physical location unit c. The smallest logical ancestor corresponding to the candidate physical location unit c of the target The subordinate second branch set, The set of network branches remaining after removing the first branch set from the second branch set. The remaining set contains the number of network branches. The summation term is the sum of the selective synchronization coefficients of all branches in the first branch set and all branches in the remaining set that form cross-set branch pairs. During the calculation, all negative selective synchronization coefficients are treated as 0 in the summation. For boundary scenarios where the second branch set is the same as the first branch set and there are no remaining branches, the out-of-group leakage intensity is directly determined to be 0. The reason for using this formula to calculate the out-of-group leakage intensity is that by taking the average of the selective synchronization coefficients of all cross-set branch pairs, it is possible to accurately quantify whether the synchronization drift characteristics of the branches in the target candidate physical location unit are leaked to other branches with the same ancestor. The lower the out-of-group leakage intensity, the more it indicates that the synchronization drift characteristics are only concentrated in the branches covered by the target unit, rather than a unified anomaly of the entire subtree. This can effectively distinguish between local physical colocation anomalies and upper-layer common link anomalies.

[0057] After completing the external leakage intensity calculation, based on the quantitative ratio between the first branch set and the second branch set, the selectivity coefficient of the target candidate physical location unit is calculated. The formula for calculating the selectivity coefficient is:

[0058] In the formula, denoted as the selectivity coefficient of the target candidate physical location unit c. The number of network branches in the first branch set. The number of network branches in the second branch set is given. For boundary scenarios where the second branch set is the same as the first branch set, the selectivity coefficient is directly determined to be 0. The reason for using this formula to calculate the selectivity coefficient is that it can quantify the proportion of the branches covered by the target candidate physical location unit in its smallest logical ancestor subtree. The smaller the proportion, the closer the selectivity coefficient is to 1, indicating that the synchronization drift feature only appears in a small selective subset in the subtree. It can effectively distinguish between selective subset anomalies and whole subtree unified anomalies, and avoid misjudging whole tree common anomalies as local physical anomalies.

[0059] After calculating the selectivity coefficient, the selectivity interpretation degree of the target candidate physical location unit is obtained based on the intra-group synchronization strength, inter-group leakage strength, and selectivity coefficient. The formula for calculating the selectivity interpretation degree is as follows:

[0060] In the formula, The selective explanatory power is denoted as c. For scenarios where the calculated selective explanatory power is negative, it is uniformly treated as 0. The reason for using this formula to calculate the selective explanatory power is that the difference between the synchronization strength within the group and the leakage strength outside the group quantifies the exclusive explanatory power of the target candidate physical location unit for synchronization drift characteristics. A positive difference indicates that the synchronization characteristics within the unit are significantly stronger than those outside the group. Then, by weighting with a selectivity coefficient, the explanatory power for anomalies in selective subsets is highlighted, while the interference of uniform anomalies in the whole subtree is filtered out. The higher the final selective explanatory power value, the more accurately the candidate physical location unit can explain the selective synchronization micro-bending drift phenomenon and the closer it is to the actual abnormal physical location.

[0061] It should be noted that the target candidate physical location unit refers to the local physical organization location unit currently selected as the analysis object. It is composed of a specific combination of junction boxes, fiber trays, and corner positions, and is used to represent the location where multiple network branches may pass through in the local space and be affected by similar bending constraints, stress states, or environmental disturbances. The first branch set refers to the set formed by all network branches that actually pass through the target candidate physical location unit. This set represents the range of branches that share the same local physical path and the same local structural constraints. The minimum logical ancestor refers to the common upstream node in the logical topology that can simultaneously cover all network branches in the first branch set and is closest to these network branches. This node corresponds to the minimum common convergence position of these branches in the logical connection relationship. The second branch set refers to the set formed by all leaf branches covered downward by the minimum logical ancestor. This set represents the range of all terminal branches that share the same minimum common upstream node with the first branch set.

[0062] It should be noted that intra-group synchronization strength refers to the degree of consistency in synchronous changes among network branches within the first branch set after removing common upstream influences, used to characterize whether there are significant local coordinated fluctuations between branches passing through the same target candidate physical location unit; extra-group leakage strength refers to the degree of synchronous changes exhibited between network branches within the first branch set and other network branches in the second branch set excluding the first branch set, used to characterize whether the abnormal association corresponding to the target candidate physical location unit spreads to other branches under the same common ancestor that have not passed through that physical location; the selectivity coefficient refers to the coefficient based on the first branch set in the second branch set. The quantitative coefficient, formed by the proportion of branches in the set, is used to characterize whether the branches affected by the target candidate physical location unit only account for a portion of the common ancestor, thereby reflecting whether the anomaly pattern has local selectivity rather than overall universality; the selective explanatory power refers to the evaluation quantity formed by the combination of intra-group synchronization strength, inter-group leakage strength and selective coefficient, used to measure the explanatory power of the target candidate physical location unit for the selective synchronization change phenomenon of the current network branches. The more its value reflects that the internal synchronization of the branches passing through the unit is obvious, the external diffusion is weak and it only affects a portion of the branches under the common ancestor, the more likely the target candidate physical location unit is to be the location of the real local anomaly.

[0063] In embodiments of the present invention, each candidate physical location unit is organized into a hierarchical search graph, and a swarm intelligence algorithm is used to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit, including: Based on the aforementioned physical resource organization data of the PON tree optical network and the completed candidate physical location unit set, a hierarchical search graph is constructed. Junction boxes, fiber trays, and corner locations corresponding to each candidate physical location unit are extracted from the physical resource organization data. During extraction, the integrity of the combination information is simultaneously verified to ensure that each candidate physical location unit's corresponding junction box, fiber tray, and corner location has matching superior attribution information. Simultaneously, complete node information for all splitter nodes directly associated with junction boxes within candidate physical location units is extracted, and invalid node data without corresponding candidate physical location units is removed. This ensures the completeness and relevance of subsequent hierarchical association construction. Subsequently, parent-child relationships are established between splitter nodes and junction boxes, between junction boxes and fiber trays, and between fiber trays and corner locations. The parent-child relationship definition is as follows. This establishes a one-way hierarchical relationship between upper-level nodes and their unique, directly subordinate lower-level nodes. Each lower-level node can only correspond to one direct parent node, and cross-level parent-child relationships are prohibited. Specifically, the parent-child relationship between a splitter node and a junction box is determined by the fact that the lower-level output port of the splitter node is directly connected to the input port of the corresponding junction box via the ODN fiber core, ensuring an end-to-end direct physical link connection without intermediate splitter nodes or other physical nodes. Similarly, the parent-child relationship between a junction box and a fiber optic tray is determined by the fact that the fiber optic tray is physically fixed inside the corresponding junction box and is its dedicated fiber optic unit, with no cross-junction relationship. Finally, the parent-child relationship between a fiber optic tray and a corner unit is determined by the fact that the corner unit is a dedicated fixed corner structure on the corresponding fiber optic tray used for fiber core winding and belongs to the corresponding fiber optic tray, with no cross-tray relationship. This system ensures standardization and traceability for subsequent path searching.

[0064] An adjacency matrix is ​​used to standardize and store all parent-child relationships. The formula for calculating the adjacency matrix is ​​as follows:

[0065] In the formula, The adjacency matrix represents the associated values ​​of row node u and column node v, where u and v are any two nodes from all valid physical nodes. Node types include optical splitters, junction boxes, fiber trays, and corner nodes. The rows and columns of the adjacency matrix follow a consistent node sorting rule, with the sorting order from top to bottom according to the hierarchy of optical splitters, junction boxes, fiber trays, and corner nodes. The matrix dimension equals the total number of valid physical nodes. The reason for using an adjacency matrix to store parent-child relationships is that it can clearly and unambiguously represent the unidirectional hierarchical relationship between all nodes, providing standardized structured data for the subsequent construction of the hierarchical search graph. At the same time, it can directly adapt to the path search logic of the ant colony algorithm, quickly locate the relationship between nodes, and improve the computational efficiency and accuracy of the search process.

[0066] After establishing and standardizing the storage of parent-child relationships, a validity check of these relationships is performed. Duplicate relationships, cross-level indirect relationships, multiple parent node affiliations, and reverse relationships are eliminated to ensure that all parent-child relationships are direct, unidirectional, top-down, and conform to the hierarchical type constraints. Subsequently, a hierarchical search graph for pathfinding is constructed based on these parent-child relationships. This hierarchical search graph is a directed acyclic graph (DAG). Its node set consists of four types of physical nodes that have passed validity checks: optical splitters, junction boxes, fiber optic trays, and corner nodes. Its edge set consists of directed edges between all nodes with valid parent-child relationships. The direction of the directed edges is fixed from the parent node to the child node, conforming to the hierarchical order from optical splitter to junction box to fiber optic tray to corner node. The mathematical definition of the hierarchical search graph is:

[0067] In the formula, To construct the complete hierarchical search graph, The set of nodes consisting of all valid physical nodes. The set of edges is the directed edges corresponding to all valid parent-child relationships. The reason for using a directed acyclic graph to construct the hierarchical search graph is that it can match the top-down tree-like physical deployment hierarchy of the PON network ODN system, limiting the search path of the ant colony algorithm to only proceed along the direction from the parent node to the child node, avoiding invalid search results caused by reverse search, cross-level search, and cyclic paths, thus improving the efficiency and accuracy of path search. At the same time, it ensures that the end node of the complete path obtained by the search corresponds to the corner node of the candidate physical positioning unit, and the complete path can directly match the combination structure of junction box, fiber tray and corner.

[0068] After the initial construction of the hierarchical search graph is completed, the validity of the graph structure is verified. The verification rules are that the hierarchical search graph is an acyclic directed graph, and the cycle detection is completed by depth-first traversal. All corner nodes corresponding to candidate physical location units are included in the node set, all directed edges conform to the hierarchical order from top to bottom, there are no isolated nodes and intermediate nodes without end paths. For nodes that cannot form a complete four-layer path from the splitter node to the junction box to the fiber tray to the corner, they are directly removed from the node set, and the corresponding invalid edges are removed at the same time. This ensures that all complete paths in the hierarchical search graph directly correspond to valid candidate physical location units. The hierarchical search graph that passes the verification can be directly used for the subsequent ant colony algorithm path search and optimal physical location unit search stages.

[0069] It should be noted that the hierarchical search graph refers to a hierarchical path structure constructed based on the actual subordinate and connection relationships between splitter nodes, junction boxes, fiber trays, and corners in a PON tree-shaped optical network. It is used to transform the traversal sequence and nesting relationship of network branches in local physical organization locations into an organizational model that can be traversed and searched layer by layer. The upper-level nodes correspond to a larger range of common physical organization locations, while the lower-level nodes correspond to more specific local physical organization locations. The parent-child relationship represents the hierarchical entry relationship of the fiber optic path in the physical organization. Therefore, through the hierarchical search graph, it can be clearly determined which splitter node a network branch passes through first, then which junction box it enters, then which fiber tray it falls into, and finally which corner it passes through. This allows candidate physical location units corresponding to local micro-bending pressure, moisture disturbance, or stress concentration to be placed in a clear hierarchical search path, which facilitates the directional location of the most likely real anomaly location by combining the synchronous change characteristics of the branches corresponding to each layer of nodes.

[0070] Based on the constructed hierarchical search graph and the calculated selective interpretability of each candidate physical location unit, an ant colony algorithm is used to perform path search and determine the optimal physical location unit. First, the ant colony algorithm is initialized, including setting the total number of ants, the maximum number of iterations, the initial pheromone value for all path edges, the pheromone evaporation coefficient, the upper and lower pheromone thresholds, and the termination round for consecutive unchanged pheromone levels. Specifically, the total number of ants is set to twice the total number of corner nodes in the hierarchical search graph, the maximum number of iterations is set to 50, the initial pheromone value for all path edges is set to 1.0, the pheromone evaporation coefficient is set to 0.7, the upper pheromone threshold is set to 100.0, the lower pheromone threshold is set to 0.001, and the termination round for consecutive unchanged pheromone levels is set to 10. These parameter settings can be adjusted according to the hierarchical search graph. The node size is adjusted proportionally. After initialization, all ants are evenly placed at the root node (optical splitter) of the hierarchical search graph, and the iterative search process is started. In each iteration, a single ant starts from its current parent node and moves to its child nodes level by level until it reaches the terminal node or there are no available child nodes. First, based on the selective interpretability of the candidate physical location units corresponding to each child node under the parent node, the heuristic value for the transfer from the parent node to each child node is determined. When the child node is a terminal corner node, the selective interpretability of the unique candidate physical location unit to which the corner node belongs is directly used as the corresponding value. When the child node is a junction box or fiber tray node at an intermediate level, the maximum selective interpretability of all terminal candidate physical location units in the subtree rooted at that node is used as the corresponding value. The formula for calculating the heuristic value is:

[0071] In the formula, This is the heuristic value for the ant's movement from parent node u to child node v. This represents the selective explanatory power value corresponding to child node v. Let u be the set of all direct child nodes of the parent node u. The minimum value of the selective explanatory power of all child nodes under parent node u is used to calculate the heuristic value. The reason for using this formula to calculate the heuristic value is that by subtracting the minimum value and adding 1, all heuristic values ​​are ensured to be non-zero positive values, avoiding errors in the calculation of transition probability caused by zero values. At the same time, it can amplify the differences in selective explanatory power between different child nodes and guide ants to move to child nodes with stronger explanatory power for anomalies.

[0072] After calculating the heuristic value, the heuristic value is normalized across all child nodes of the same parent node. The normalization formula is as follows:

[0073] In the formula, The normalized heuristic value, The normalization process is used to sum the heuristic values ​​of all child nodes under parent node u. The reason for this normalization is to map the heuristic values ​​of all child nodes under the same parent node to the range of 0 to 1, and the sum is 1. This eliminates the influence of the numerical magnitude caused by the difference in the number of child nodes under different parent nodes, and provides standardized input data for subsequent adaptive weight and transition probability calculation. For boundary scenarios where the parent node contains only a single child node, the normalized heuristic value is directly set to 1.

[0074] After normalization, the pheromone weight and heuristic weight corresponding to the parent node are determined based on the normalized heuristic values. The normalized entropy corresponding to the parent node is then calculated. The formula for calculating the normalized entropy is as follows:

[0075] In the formula, The normalized entropy corresponding to the parent node u. Let be the number of child nodes of parent node u, and ln be the natural logarithm calculation operation. During the calculation, [the following is considered:] ... In scenarios where the value is 0, it is treated as 0 to avoid errors in calculating the logarithm to zero. The reason for using normalized entropy is that it can quantify the uniformity of the distribution of heuristic values ​​among child nodes under the same parent node. When the entropy value is closer to 1, it indicates that the difference between the heuristic values ​​of each child node is smaller, and more reliance on pheromones is needed to guide the global search to avoid getting trapped in local optima. When the entropy value is closer to 0, it indicates that the difference between the heuristic values ​​of each child node is larger, and more reliance on heuristic values ​​is needed to guide local development to speed up algorithm convergence. For boundary scenarios where the parent node contains only a single child node, the normalized entropy is directly set to 0.

[0076] The pheromone weight and heuristic weight are adaptively set based on the normalized entropy value, and the calculation formula is as follows:

[0077]

[0078] In the formula, The pheromone weight corresponding to the parent node u. The heuristic weights corresponding to the parent node u are set using adaptive weights. This allows for dynamic adjustment of the search strategy based on the distribution of heuristic values ​​among the child nodes under the current node, avoiding premature convergence or low search efficiency caused by fixed weights. It balances the algorithm's global exploration capability with its local development capability, ensuring accurate identification of globally optimal candidate physical location units. After setting the weights, the probability of transitioning to each child node is calculated based on the pheromone weights, heuristic weights, path edge pheromones, and heuristic values. The formula for calculating the transition probability is:

[0079] In the formula, Let be the probability that the k-th ant moves from its parent node u to its child node v. The current pheromone value of the path edge from parent node u to child node v is used to calculate the transition probability. This formula combines the global experience accumulated from historical pheromone levels with the local guidance of the current node's heuristic value, dynamically adjusting the influence ratio of both with adaptive weights. This ensures that the ant's transition direction conforms to both the experience of historical optimal paths and the anomaly interpretation capability of the current candidate physical location unit. For boundary scenarios where the parent node contains only a single child node, the transition probability is directly set to 1. During the ant colony algorithm iteration, a single ant completes node-by-node transitions based on the transition probability corresponding to each parent node at each level until it reaches the corner node at the end of the hierarchical search graph, generating a complete transition path. The candidate physical location unit to which the end corner node of the transfer path belongs is the end candidate physical location unit reached by the ant. After each ant completes path generation, path validity verification is immediately performed. The verification rule is that the path must be a complete four-layer path from the optical splitter node to the junction box to the fiber tray to the corner, and all nodes and edges in the path must exist in the hierarchical search graph. Invalid paths that fail the verification do not participate in subsequent pheromone updates. After all ants have completed path generation and validity verification in each iteration, the pheromone of the path edges in the transfer path is updated based on the selective interpretability of the end candidate physical location unit corresponding to each valid path. First, the pheromone evaporation operation is performed. The calculation formula for the evaporation operation is:

[0080] In the formula, ρ is the pheromone evaporation coefficient, which is fixed at 0.7. The reason for using the evaporation operation is to gradually reduce the pheromone accumulation of historical paths, weaken the influence of early invalid paths, avoid the algorithm getting stuck in local optima, and improve the global search capability.

[0081] After the evaporation operation is completed, an incremental pheromone update is performed. The formula for calculating the incremental update is:

[0082]

[0083] In the formula, Let ν be the pheromone increment of the k-th ant on the path edge uv. The candidate physical location unit for the end reached by the k-th ant The selective interpretability is K, where K is the total number of ants. The reason for adopting this incremental update rule is to allow the paths corresponding to candidate physical location units with higher selective interpretability to obtain higher pheromone increments, guiding ants in subsequent iterations to move to paths with greater anomaly interpretability, strengthening the pheromone accumulation of the optimal path, and accelerating the convergence speed of the algorithm. After completing the incremental update, pheromone out-of-bounds processing is performed, limiting the pheromone values ​​of all path edges to the preset upper and lower pheromone threshold range. Values ​​exceeding the upper limit are assigned the upper limit value, and values ​​below the lower limit are assigned the lower limit value, avoiding search failure caused by pheromone saturation or disappearance. After completing the full pheromone update, the global optimal candidate physical location unit of the current iteration round is recorded, and it is determined whether the maximum number of iterations has been reached, or whether the global optimal result of consecutive preset rounds has not changed. If the termination condition is not met, the next iteration round is started; if the termination condition is met, the iteration process is terminated.

[0084] After the iteration terminates, the optimal physical location unit is determined based on the selective interpretation degree of each candidate physical location unit and the pheromone of the path edges in the corresponding transfer path. The formula for calculating the optimal physical location unit is as follows:

[0085] In the formula, This is the optimal physical positioning unit. This represents the complete and valid path from the beam splitter node to the end corner node corresponding to candidate physical positioning unit c. The optimal physical location unit is determined by the product of pheromones of all edges of the complete path. This formula combines the explanatory power of the candidate physical location unit for the anomaly with the global pheromone accumulation of the path during the algorithm iteration process. It considers both the unit's specific explanatory power for selective synchronous drift and the algorithm's global search experience for the optimal path, ensuring that the obtained optimal physical location unit is the result that best matches the actual anomaly location.

[0086] It should be noted that the optimal physical location unit refers to the local physical organizational location unit that has the highest explanatory power for the current network branch anomaly among all candidate physical location units and is ultimately determined as the most likely true anomaly location. This unit is usually composed of a specific combination of junction boxes, fiber trays, and corner positions, representing the location where the optical fiber actually passes through in the local space and is more likely to be affected by common micro-bending pressure, moisture disturbance, or stress concentration. The optimal physical location unit does not simply refer to the unit with the highest selective explanatory power alone, but rather to the target unit finally determined from the path search results after comprehensively considering the explanatory power of the terminal candidate physical location unit itself for the selective synchronous change of the target branch and the strength of the pheromone accumulated on each path edge in the entire transfer path to reach the unit, under the layer-by-layer path constraints of the hierarchical search graph. Therefore, it reflects both the local physical location's local explanatory power for the anomaly pattern and the degree to which the location is continuously pointed to and reinforced during the hierarchical search process from the upper-level common organizational location to the lower-level specific organizational location, thus serving as the final location result of the network physical anomaly.

[0087] Based on the aforementioned calculated optimal physical location unit, the constructed network branch and candidate physical location unit membership matrix, and the selective interpretability values ​​corresponding to all candidate physical location units, the localization results are generated and output. First, the set of affected branches corresponding to the optimal physical location unit is determined based on membership. Then, all network branches with a physical transit relationship to the optimal physical location unit are searched based on membership. The criterion for determining a physical transit relationship is consistent with the definition of membership; that is, a physical transit relationship is determined when the membership value between a network branch and the optimal physical location unit is 1. All network branches meeting the criteria are deduplicated and merged to obtain the set of affected branches. The mathematical definition of the set of affected branches is:

[0088] In the formula, For optimal physical location unit The corresponding set of affected branches, For any valid network branch in a PON tree optical network, For network branch i and the optimal physical location unit The corresponding membership values; the reason for using this method to define the set of affected branches is that it can unambiguously filter out all network branches that pass through abnormal physical locations based on a standardized membership matrix, and accurately delineate the scope of the abnormal impact.

[0089] After initially constructing the affected branch set, the relative confidence level of the optimal physical location unit is determined based on the selective interpretability of each candidate physical location unit. The relative confidence level is calculated using softmax normalization. Before calculation, the selective interpretability of all candidate physical location units is centered to avoid overflow in the natural exponent calculation. The formula for calculating the relative confidence level is as follows:

[0090] In the formula, For optimal physical location unit The corresponding relative confidence level, For optimal physical location unit The corresponding selective explanatory power, This represents the maximum selective interpretation value among all candidate physical location units. The set consisting of all candidate physical location units. For set For any candidate physical location unit in the algorithm, exp represents the natural exponent calculation operation. The reason for using centralized processing and softmax normalization to calculate the relative confidence is that centralized processing eliminates computational overflow caused by excessively large values, while the natural exponent function amplifies the differences in selective interpretability between different candidate physical location units. The interpretability of all candidate units is mapped to a relative probability value between 0 and 1, and the sum of the relative confidence of all candidate units is 1. This allows for a direct quantification of the confidence level of the optimal physical location unit compared to other candidate units, providing a clear decision-making reference for on-site maintenance and troubleshooting. For boundary scenarios where the selective interpretability of all candidate physical location units is 0, the optimal physical location unit's confidence level is directly set. The relative confidence level is 0. For scenarios where multiple candidate physical location units have the same selective interpretability, the relative confidence level of all candidate units is calculated synchronously according to the same rules. After the relative confidence level calculation is completed, the optimal physical location unit, the set of affected branches, and the relative confidence level are combined to output the location result of the corresponding network physical anomaly. The output location result fully includes three core dimensions: the specific physical location where the anomaly occurred, the list of all affected network branches, and the relative confidence level of the location result. This can directly support maintenance personnel in carrying out on-site troubleshooting and fault repair work. After the location result is output, the selective interpretability and relative confidence level ranking list of all candidate physical location units are output synchronously to provide auxiliary reference information for maintenance personnel.

[0091] like Figure 2 The diagram shown is a functional block diagram of a network positioning system based on topology association analysis provided in an embodiment of the present invention.

[0092] In this embodiment, the functions of each module / unit are as follows: The data mapping module is used to acquire the received optical power sequence, forward error correction count sequence, logical topology relationship and physical resource organization data of the network branch, calculate the branch drift sequence based on the received optical power sequence and forward error correction count sequence, and establish the membership relationship between the network branch and the candidate physical location unit based on the physical resource organization data. The common mode analysis module is used to determine the nearest common ancestor between any two network branches based on logical topological relationships, generate a common mode drift sequence, perform de-common mode processing on the branch drift sequence to obtain a residual sequence, and calculate the selective synchronization coefficient between any two network branches based on the residual sequence. The indicator evaluation module is used to calculate the selective interpretability of each candidate physical location unit based on membership, logical topology, and selective synchronization coefficient. The path optimization module is used to organize each candidate physical location unit into a hierarchical search graph, and use a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit. The location parsing module is used to determine the set of affected branches corresponding to the optimal physical location unit based on the membership relationship, calculate the relative confidence of the optimal physical location unit based on the selective interpretation degree, and output the location result.

[0093] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A network localization method based on topological association analysis, characterized in that, include: S1. Obtain the received optical power sequence, forward error correction count sequence, logical topology relationship and physical resource organization data of the network branch; calculate the branch drift sequence based on the received optical power sequence and forward error correction count sequence; and establish the membership relationship between the network branch and the candidate physical location unit based on the physical resource organization data. S2. Based on the logical topology, determine the nearest common ancestor between each pair of network branches, generate a common-mode drift sequence, perform de-common-mode processing on the branch drift sequence to obtain a residual sequence, and calculate the selective synchronization coefficient between each pair of network branches based on the residual sequence. S3. Calculate the selective interpretability of each candidate physical location unit based on membership, logical topology, and selective synchronization coefficient. S4. Organize each candidate physical location unit into a hierarchical search graph, and use a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit. S5. Determine the set of affected branches corresponding to the optimal physical location unit based on the membership relationship, calculate the relative confidence level corresponding to the optimal physical location unit based on the selective interpretation degree, and output the location result.

2. The network localization method based on topological association analysis according to claim 1, characterized in that, The physical resource organization data is used to characterize the transit relationships between each network branch and the splitter node, junction box, fiber tray and corner position. The candidate physical positioning unit is composed of a combination of junction box, fiber tray and corner position.

3. The network localization method based on topological association analysis according to claim 1, characterized in that, The branch drift sequence is calculated based on the received optical power sequence and the forward error correction count sequence, including: The median and absolute median difference of the received optical power sequence corresponding to each network branch, as well as the median and absolute median difference of the forward error correction count sequence, are determined respectively. The received optical power sequence is normalized based on the median and absolute median difference to obtain a first drift amount characterizing the degree of optical power reduction. The forward error correction count sequence is normalized based on the median and absolute median difference to obtain a second drift amount that characterizes the degree of increase in error correction pressure. The first drift amount and the second drift amount are combined at each time point to obtain the branch drift sequence of the corresponding network branch.

4. The network localization method based on topological association analysis according to claim 1, characterized in that, Based on logical topological relationships, the nearest common ancestor between each pair of network branches is determined, a common-mode drift sequence is generated, and the branch drift sequence is processed to obtain a residual sequence, including: For any two network branches, find the nearest common ancestor of the two network branches based on the logical topological relationship; Extract all leaf branches under the nearest common ancestor and obtain the branch drift sequence corresponding to each leaf branch; Median aggregation is performed on the branch drift sequences corresponding to all leaf branches at the same time to generate common mode drift sequences corresponding to any two network branches; Subtract the common-mode drift sequence from the branch drift sequence corresponding to each of the two network branches to obtain the residual sequence corresponding to each of the two network branches.

5. The network localization method based on topological association analysis according to claim 4, characterized in that, The selective synchronization coefficients between each pair of network branches are calculated based on the residual sequence, including: Determine the first-order transformation sequence of the residual sequence corresponding to each of any two network branches; Based on the degree of correlation consistency between the first-order change sequences of any two network branches, the selective synchronization coefficient between any two network branches is calculated.

6. The network localization method based on topological association analysis according to claim 1, characterized in that, Based on membership, logical topology, and selective synchronization coefficients, the selective interpretability of each candidate physical location unit is calculated, including: Extract the first branch set that passes through the target candidate physical location unit based on the aforementioned membership relationship; Based on the logical topological relationship, determine the smallest logical ancestor that covers the first branch set, and extract all leaf branches under the smallest logical ancestor to form the second branch set; Based on the selective synchronization coefficients between each pair of network branches within the first branch set and the number of corresponding branch pairs, the intra-group synchronization strength of the target candidate physical positioning unit is calculated. Based on the selective synchronization coefficient between network branches in the first branch set and network branches in the second branch set other than the first branch set, and the number of corresponding cross-set branch pairs, the out-of-group leakage intensity of the target candidate physical location unit is calculated. Based on the ratio of the number of the first branch set to the number of the second branch set, the selectivity coefficient of the target candidate physical location unit is calculated; Based on the intra-group synchronization strength, the inter-group leakage strength, and the selectivity coefficient, the selective interpretability of the target candidate physical location unit is obtained.

7. The network localization method based on topological association analysis according to claim 2, characterized in that, The candidate physical location units are organized into a hierarchical search graph, including: Extract the junction box, fiber optic tray and corner combination information corresponding to each candidate physical positioning unit from the physical resource organization data; Establish parent-child relationships between the splitter node and the junction box, between the junction box and the fiber optic tray, and between the fiber optic tray and the corner unit; A hierarchical search graph for path search is constructed based on the parent-child relationship.

8. The network localization method based on topological association analysis according to claim 7, characterized in that, The optimal physical location unit is obtained by searching the hierarchical search graph using a swarm intelligence algorithm based on selective interpretability, including: The ant colony algorithm is used to perform path search on the hierarchical search graph; Based on the selective interpretation degree of the candidate physical location units corresponding to each child node under the parent node, the heuristic value for the transfer from the parent node to each child node is determined, and the heuristic value is normalized within the range of all child nodes of the same parent node. The pheromone weight and heuristic weight corresponding to the parent node are determined based on the normalized heuristic value, and the probability of transition to each child node is calculated based on the pheromone weight, the heuristic weight, the path edge pheromone, and the heuristic value. During the ant colony algorithm iteration process, a corresponding transfer path is generated based on the transfer probability, and the pheromone of the path edge in the transfer path is updated based on the selective interpretability of the candidate physical location unit at the end reached by the ant. The optimal physical location unit is determined based on the selective interpretability of each candidate physical location unit at the end and the pheromone of the path edges in the corresponding transfer path.

9. The network localization method based on topological association analysis according to claim 1, characterized in that, The relative confidence level corresponding to the optimal physical location unit is calculated based on the selective interpretation degree, and the location results are output, including: The relative confidence level of the optimal physical location unit is determined based on the selective interpretability of each candidate physical location unit. Based on the membership relationship, find all network branches that have a physical connection with the optimal physical location unit, and merge them to obtain the set of affected branches; The optimal physical location unit, the affected branch set, and the relative confidence level are combined to output the location result of the corresponding network physical anomaly location.

10. A network positioning system based on topology correlation analysis, applied in the network positioning method based on topology correlation analysis according to any one of claims 1-9, characterized in that, include: The data mapping module is used to acquire the received optical power sequence, forward error correction count sequence, logical topology relationship and physical resource organization data of the network branch, calculate the branch drift sequence based on the received optical power sequence and forward error correction count sequence, and establish the membership relationship between the network branch and the candidate physical location unit based on the physical resource organization data. The common mode analysis module is used to determine the nearest common ancestor between any two network branches based on logical topological relationships, generate a common mode drift sequence, perform de-common mode processing on the branch drift sequence to obtain a residual sequence, and calculate the selective synchronization coefficient between any two network branches based on the residual sequence. The indicator evaluation module is used to calculate the selective interpretability of each candidate physical location unit based on membership, logical topology, and selective synchronization coefficient. The path optimization module is used to organize each candidate physical location unit into a hierarchical search graph, and use a swarm intelligence algorithm to search the hierarchical search graph based on selective interpretability to obtain the optimal physical location unit. The location parsing module is used to determine the set of affected branches corresponding to the optimal physical location unit based on the membership relationship, calculate the relative confidence of the optimal physical location unit based on the selective interpretation degree, and output the location result.

Images