A knowledge graph-based wireless communication fault diagnosis method

Abstract

The application discloses a wireless communication fault diagnosis method based on a knowledge graph, and comprises the following steps: collecting wireless communication network data to generate a standardized communication operation data set; constructing a heterogeneous wireless communication knowledge graph based on the standardized communication operation data set; generating a structure locking subgraph set based on a topological dependence path and a closed loop subgraph structure; calculating a locking strength index, generating a locking strength evolution sequence, determining a locking stable interval, and generating a structure locking domain; calculating an unlocking trigger judgment quantity in the structure locking domain to generate an unlocking trigger edge set; extracting a minimum unlocking trigger substructure to generate a minimum unlocking trigger substructure set; and generating a wireless communication fault diagnosis result according to the minimum unlocking trigger substructure set. The application constructs a structure locking and minimum trigger extraction mechanism based on a knowledge graph, realizes accurate fault positioning and propagation judgment, and has the advantages of high positioning accuracy and low misjudgment rate.

Description

Technical Field

[0001] This invention relates to the field of wireless communication fault diagnosis, and in particular to a wireless communication fault diagnosis method based on knowledge graphs. Background Technology

[0002] As wireless communication networks continue to expand, their structures exhibit high heterogeneity and high service coupling. Complex multi-level relationships are formed between base stations, cells, wireless links, and service bearers. Existing wireless communication fault diagnosis methods are mostly based on alarm statistical analysis, threshold determination, or single topology path backtracking for location, mainly targeting node-level or link-level anomalies and focusing on detecting fluctuations in local indicators. They lack the ability to jointly analyze the overall stability of topology dependency structure and service bearer relationship.

[0003] In scenarios where multi-source data is dynamically changing, existing methods struggle to characterize the locking state and its evolution at the structural level, fail to identify the smallest triggering structure that leads to instability, and struggle to accurately extract the root cause and propagation path of faults in complex relationships. This can easily lead to misjudgments or overly coarse localization granularity, affecting the accuracy and reliability of wireless communication fault diagnosis results. Summary of the Invention

[0004] One objective of this invention is to propose a knowledge graph-based method for wireless communication fault diagnosis. This invention utilizes a knowledge graph-based structure locking and minimum trigger extraction mechanism to achieve accurate fault location and propagation determination, and has the advantages of high location accuracy and low false positive rate.

[0005] A knowledge graph-based wireless communication fault diagnosis method according to an embodiment of the present invention includes the following steps: Collect wireless communication network data and preprocess it to generate a standardized communication operation dataset; A heterogeneous wireless communication knowledge graph is constructed based on a standardized communication operation dataset, including a set of communication entity nodes and topological dependency edges, service carrying relationship edges, and causal propagation relationship edges. Operation indicator attributes are configured for the set of communication entity nodes, and weight attributes are configured for the relationship edges. In the knowledge graph of heterogeneous wireless communication, a set of structure-locked subgraphs is generated based on the topological dependency path formed by topological dependency edges and the closed-loop subgraph structure formed by service bearer relationship edges. For each structural locking subgraph in the set of structural locking subgraphs, the locking strength index is calculated based on the rate of change of the relation edge weight attribute and the offset of the operation index attribute, generating a locking strength evolution sequence. The locking stability interval is determined based on the locking strength evolution sequence, and the structural locking domain is generated. The heterogeneous wireless communication knowledge graph is updated in state based on the real-time updated standardized communication operation dataset. The unlock trigger decision quantity is calculated in the structure locking domain, and the unlock trigger edge set is generated. Based on the set of unlock trigger edges, extract the minimum unlock trigger substructure from the set of structure-locked subgraphs, and generate the set of minimum unlock trigger substructures. Based on the minimum unlock trigger substructure set, the root cause localization result and fault propagation path result of wireless communication fault are formed, and the wireless communication fault diagnosis result is generated.

[0006] Optionally, the wireless communication network data specifically includes base station operating parameters, wireless link quality indicators, service carrying status, topology connection relationships, and alarm event data. The preprocessing specifically includes data consistency verification, time alignment, multi-source synchronization, outlier handling, noise filtering, indicator normalization, entity identifier mapping, and time window construction.

[0007] Optionally, the generation of the heterogeneous wireless communication knowledge graph specifically includes: The set of communication entity nodes is extracted based on the standardized communication operation dataset, including base station nodes, cell nodes, wireless link nodes, service bearer nodes, and alarm event nodes; Configure operational indicator attributes for each communication entity node in the communication entity node set, including normalized values ​​of link quality indicators, normalized values ​​of service bearer status, normalized values ​​of base station operation parameters, normalized values ​​of alarm event counts, and normalized values ​​of topology connectivity. A set of topological dependency edges is generated based on the set of communication entity nodes. A topological dependency weight is configured for each topological dependency edge, which is determined by weighting the normalized aggregate value of the wireless link quality index and the normalized value of the topological connection strength within the corresponding time window. A set of service bearer relationship edges is generated based on the set of communication entity nodes. A service bearer weight is configured for each service bearer relationship edge, which is determined by weighting the normalized value of service load and the normalized value of user load within the corresponding time window. A set of causal propagation relationship edges is generated based on the association between alarm event nodes and communication entity nodes. A causal propagation weight is configured for each causal propagation relationship edge, which is determined according to the statistical association frequency between the preceding alarm event and the subsequent alarm event within the time lag window. The set of communication entity nodes, the set of topological dependency edges, the set of service carrying edges, and the set of causal propagation edges are merged to generate a heterogeneous wireless communication knowledge graph that includes node operation index attributes and relation edge weight attributes.

[0008] Optionally, the generation of the structure-locked subgraph set specifically includes: In the heterogeneous wireless communication knowledge graph, the set of communication entity nodes, the set of topological dependency edges, and the set of service bearer edges are extracted, and the topological dependency adjacency matrix and the service bearer adjacency matrix are constructed respectively. The communication entity nodes are traversed based on the topological dependency adjacency matrix to generate a set of topological dependency paths formed by continuous connection of topological dependency edges, and the corresponding node set is determined for each topological dependency path. Based on the adjacency matrix of service bearers, a closed node sequence composed of service bearer relationship edges is identified, a set of closed-loop subgraph structures is generated, and a corresponding node set is determined for each closed-loop subgraph. The intersection and union relationship of the set of nodes of topologically dependent paths and the set of nodes of closed-loop subgraphs is calculated to obtain the structural coupling degree. When the structural coupling degree reaches the structural coupling threshold, the corresponding path and closed loop are determined to form a candidate locking structure. The node set in the candidate locking structure and its corresponding topological dependency edges and business carrying relationship edges are integrated to generate a structure locking subgraph; All generated structure-locked subgraphs are combined into a structure-locked subgraph set.

[0009] Optionally, the generation of the structure locking domain specifically includes: Based on the set of structure-locked subgraphs, select one structure-locked subgraph, extract the set of communication entity nodes corresponding to the structure-locked subgraph, as well as the corresponding topological dependency edges and service bearing relationship edges, and generate structure-locked subgraph state data. Within the sliding time window sequence, the weight attributes of each topological dependency edge and service bearing edge in adjacent time windows are read based on the state data of the structure-locked subgraph, the weight change rate of the edge is calculated, and the operation index attributes of each communication entity node in adjacent time windows are read, the offset of the operation index attributes is calculated, and a set of change features of the structure-locked subgraph is generated. Based on the set of changes in the structural locking subgraph, the locking strength index of the structural locking subgraph at each time window is calculated, and a set of locking strength indices is generated. Arrange the set of locking strength indices according to the time window order to generate a locking strength evolution sequence; In the locking strength evolution sequence, select time intervals that satisfy the locking strength index not lower than the locking strength threshold, the difference of the locking strength index between adjacent time windows not exceeding the locking fluctuation threshold, and continuously reach the threshold of the number of consecutive time windows, and generate the locking stability interval corresponding to the locking subgraph of this structure. The locking stability intervals of each structural locking subgraph in the set of structural locking subgraphs are integrated to generate a structural locking domain.

[0010] Optionally, the generation of the unlock trigger edge set specifically includes: Write the standardized communication operation dataset updated in real time into the heterogeneous wireless communication knowledge graph according to the time window order, and update the operation index attributes of communication entity nodes in the structure locking domain as well as the weight attributes of topological dependency relationship edges and service carrying relationship edges. For each structural locking subgraph in the structural locking domain, the mean values ​​of the operational index attributes of the communication entity nodes and the mean values ​​of the relation edge weight attributes are statistically analyzed within the structural locking domain to determine the corresponding baseline values ​​of the operational index attributes and the baseline values ​​of the weight attributes. Within the current time window, calculate the difference between the current weight attribute and the baseline value of the corresponding weight attribute of each relation edge within the structure locking domain to obtain the relation edge weight offset. At the same time, calculate the difference between the current running indicator attribute and the baseline value of the corresponding running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the running indicator attribute offset. For each relation edge, the weight offset of the relation edge is weighted and synthesized with the offset of the running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the unlock trigger determination quantity of the relation edge. The edges whose unlock trigger determination is greater than or equal to the unlock trigger threshold are identified as unlock trigger edges, forming a set of unlock trigger edges.

[0011] Optionally, the generation of the minimum unlock trigger substructure set specifically includes: Based on the set of unlock trigger edges, extract the communication entity node corresponding to each unlock trigger edge and the unlock trigger determination quantity of that unlock trigger edge; Sort the unlock trigger edges in descending order according to the unlock trigger determination quantity to generate an unlock trigger edge sequence; Select the unlock trigger edge with the largest unlock trigger decision value from the unlock trigger edge sequence, extract the unlock trigger edge and the communication entity nodes at both ends of it to form a candidate unlock substructure; In the candidate unlock substructure, check whether there are other unlock trigger edges that share a communication entity node with the unlock trigger edge. The unlock trigger edges that meet the condition of sharing a communication entity node and their corresponding communication entity nodes are merged into the candidate unlock substructure to generate an extended unlock substructure. The number of communication entity nodes and the number of relation edges in the extended unlock substructure are counted. Any relation edge is removed in turn and it is determined whether the remaining structure still contains all unlock trigger edges. When the remaining structure no longer contains all unlock trigger edges after removing any relation edge, the current extended unlock substructure is determined to be the minimum unlock trigger substructure. All minimum unlock trigger substructures that meet the minimum condition are aggregated to generate a minimum unlock trigger substructure set.

[0012] Optionally, the generation of the wireless communication fault diagnosis result specifically includes: Based on the set of minimum unlock trigger substructures, extract the set of communication entity nodes and the set of relation edges corresponding to each minimum unlock trigger substructure. Based on the set of nodes in the minimum unlock trigger substructure, the root cause score of each communication entity node is calculated. The root cause score is obtained by the unlock trigger determination quantity of the communication entity node and its connection relationship edge. Based on the root cause score, the communication entity node with the highest score is selected as the root cause node. Based on the root cause node, the fault object type is determined, including base station node, cell node, wireless link node, service bearer node and alarm event node, and the wireless communication fault root cause location result is generated. In the heterogeneous wireless communication knowledge graph, a set of directed edges composed of causal propagation relationship edges is extracted. Starting from the root cause node, a path search is performed along the set of directed edges to generate a set of fault propagation paths. The sum of the unlocking trigger judgment quantities of each relation edge in each path is calculated as the path cost of that path. Based on the path cost value, the path with the minimum path cost value is selected from the set of fault propagation paths to generate the fault propagation path result. The results of wireless communication fault root cause localization and fault propagation path are combined to generate the final wireless communication fault diagnosis results.

[0013] The beneficial effects of this invention are: This invention constructs a heterogeneous wireless communication knowledge graph comprising a set of communication entity nodes, topological dependency edges, service bearer edges, and causal propagation edges. Based on this, it introduces a modeling mechanism for structural locking subgraphs and structural locking domains, elevating traditional node-level or link-level anomaly detection methods to the structural-level stability analysis level. By coupling and filtering the topological dependency paths and service bearer closed-loop structures, a set of structural locking subgraphs is formed. A locking strength index is calculated based on the change rate of relation edge weight attributes and the offset of operational indicator attributes, constructing a locking strength evolution sequence and a locking stability interval. This enables a dynamic characterization of the structural stability state, thereby identifying potential instability trends at the structural level and providing stable baseline constraints for subsequent fault diagnosis.

[0014] This invention calculates the unlock trigger determination quantity within the structural locking domain, extracts the unlock trigger edge set, and generates the minimum unlock trigger substructure set through edge-by-edge sieving and structural necessity verification mechanisms. This enables the accurate extraction of the minimum trigger unit that leads to structural instability. Compared to existing diagnostic methods that rely solely on alarm statistics or single-path backtracking, this invention can accurately locate the root cause node of the fault from the perspective of structural coupling and dynamic evolution. It also generates fault propagation path results by combining causal propagation relationship edges, effectively reducing the false judgment rate and redundant interference, improving the accuracy of fault root cause location and propagation path determination, thereby significantly improving the reliability and stability of wireless communication fault diagnosis results. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a knowledge graph-based wireless communication fault diagnosis method proposed in this invention; Figure 2 This is a schematic diagram illustrating the structure-locked subgraph generation process of a knowledge graph-based wireless communication fault diagnosis method proposed in this invention. Figure 3 This is a schematic diagram illustrating the structure-locked domain formation process of a knowledge graph-based wireless communication fault diagnosis method proposed in this invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0017] refer to Figures 1-3 A knowledge graph-based method for wireless communication fault diagnosis includes the following steps: Collect wireless communication network data and preprocess it to generate a standardized communication operation dataset; A heterogeneous wireless communication knowledge graph is constructed based on a standardized communication operation dataset, including a set of communication entity nodes and topological dependency edges, service carrying relationship edges, and causal propagation relationship edges. Operation indicator attributes are configured for the set of communication entity nodes, and weight attributes are configured for the relationship edges. In the knowledge graph of heterogeneous wireless communication, a set of structure-locked subgraphs is generated based on the topological dependency path formed by topological dependency edges and the closed-loop subgraph structure formed by service bearer relationship edges. For each structural locking subgraph in the set of structural locking subgraphs, the locking strength index is calculated based on the rate of change of the relation edge weight attribute and the offset of the operation index attribute, generating a locking strength evolution sequence. The locking stability interval is determined based on the locking strength evolution sequence, and the structural locking domain is generated. The heterogeneous wireless communication knowledge graph is updated in state based on the real-time updated standardized communication operation dataset. The unlock trigger decision quantity is calculated in the structure locking domain, and the unlock trigger edge set is generated. Based on the set of unlock trigger edges, extract the minimum unlock trigger substructure from the set of structure-locked subgraphs, and generate the set of minimum unlock trigger substructures. Based on the minimum unlock trigger substructure set, the root cause localization result and fault propagation path result of wireless communication fault are formed, and the wireless communication fault diagnosis result is generated.

[0018] In this embodiment, the wireless communication network data specifically includes base station operating parameters, wireless link quality indicators, service carrying status, topology connection relationships, and alarm event data. The preprocessing specifically includes data consistency verification, time alignment, multi-source synchronization, outlier handling, noise filtering, indicator normalization, entity identifier mapping, and time window construction.

[0019] In this embodiment, the generation of the heterogeneous wireless communication knowledge graph specifically includes: The set of communication entity nodes is extracted based on the standardized communication operation dataset, including base station nodes, cell nodes, wireless link nodes, service bearer nodes, and alarm event nodes; Configure operational indicator attributes for each communication entity node in the communication entity node set, including normalized values ​​of link quality indicators, normalized values ​​of service bearer status, normalized values ​​of base station operation parameters, normalized values ​​of alarm event counts, and normalized values ​​of topology connectivity. A set of topological dependency edges is generated based on the set of communication entity nodes. A topological dependency weight is configured for each topological dependency edge, which is determined by weighting the normalized aggregate value of the wireless link quality index and the normalized value of the topological connection strength within the corresponding time window. A set of service bearer relationship edges is generated based on the set of communication entity nodes. A service bearer weight is configured for each service bearer relationship edge, which is determined by weighting the normalized value of service load and the normalized value of user load within the corresponding time window. A set of causal propagation relationship edges is generated based on the association between alarm event nodes and communication entity nodes. A causal propagation weight is configured for each causal propagation relationship edge, which is determined according to the statistical association frequency between the preceding alarm event and the subsequent alarm event within the time lag window. The set of communication entity nodes, the set of topological dependency edges, the set of service carrying edges, and the set of causal propagation edges are merged to generate a heterogeneous wireless communication knowledge graph that includes node operation index attributes and relation edge weight attributes.

[0020] In this embodiment, the generation of the structure-locked subgraph set specifically includes: In the heterogeneous wireless communication knowledge graph, the set of communication entity nodes, the set of topological dependency edges, and the set of service bearer edges are extracted, and the topological dependency adjacency matrix and the service bearer adjacency matrix are constructed respectively. The communication entity nodes are traversed based on the topological dependency adjacency matrix to generate a set of topological dependency paths formed by continuous connection of topological dependency edges, and the corresponding node set is determined for each topological dependency path. Based on the adjacency matrix of service bearers, a closed node sequence composed of service bearer relationship edges is identified, a set of closed-loop subgraph structures is generated, and a corresponding node set is determined for each closed-loop subgraph. The intersection and union relationship of the set of nodes of topologically dependent paths and the set of nodes of closed-loop subgraphs is calculated to obtain the structural coupling degree. When the structural coupling degree reaches the structural coupling threshold, the corresponding path and closed loop are determined to form a candidate locking structure. The generation of candidate locking structures specifically includes: The communication entity nodes are extracted from the set of nodes of the topologically dependent path and the set of nodes of the closed-loop subgraph. The intersection of the two is calculated to obtain the structural coupling region. The nodes and relational edges in the structural coupling region reflect the connection between the topological path and the closed-loop structure. The number of nodes and relational edges in the structural coupling region is calculated to evaluate the coupling degree between the path and the closed-loop structure. If the number of nodes and edges in the coupling region exceeds the preset structural coupling threshold, the path and the closed-loop structure are considered to constitute a valid candidate locking structure. Based on the calculation results of the structural coupling degree, the paths and closed-loop structures with the coupling degree reaching the structural coupling threshold are selected and merged to form a candidate locking structure. Redundant parts of the candidate locking structure are removed, unnecessary paths and nodes are removed, and an optimized candidate locking structure is generated. The node set in the candidate locking structure and its corresponding topological dependency edges and business carrying relationship edges are integrated to generate a structure locking subgraph; The generation of the structure-locked subgraph specifically includes: Read the candidate locking structure, extract the node set corresponding to the candidate locking structure, and extract the topological dependency edge set and service bearing edge set associated with the node set; filter out any edge in the topological dependency edge set whose endpoint does not belong to the node set, and filter out any edge in the service bearing edge set whose endpoint does not belong to the node set, to obtain the effective topological dependency edge set and effective service bearing edge set of the candidate locking structure; merge the effective topological dependency edge set and effective service bearing edge set to form the relation edge set of the candidate locking structure, and perform deduplication processing on duplicate relation edges in the relation edge set; filter out isolated nodes in the node set that do not have any connected relation edges in the relation edge set, to obtain the node set of the structure locking subgraph; construct the structure locking subgraph using the node set, relation edge set, topological dependency edge, and service bearing relation edge of the structure locking subgraph. All generated structure-locked subgraphs are combined into a structure-locked subgraph set.

[0021] In this embodiment, the generation of the structural locking domain specifically includes: Based on the set of structure-locked subgraphs, select one structure-locked subgraph, extract the set of communication entity nodes corresponding to the structure-locked subgraph, as well as the corresponding topological dependency edges and service bearing relationship edges, and generate structure-locked subgraph state data. Within the sliding time window sequence, the weight attributes of each topological dependency edge and service bearing edge in adjacent time windows are read based on the state data of the structure-locked subgraph, the weight change rate of the edge is calculated, and the operation index attributes of each communication entity node in adjacent time windows are read, the offset of the operation index attributes is calculated, and a set of change features of the structure-locked subgraph is generated. Based on the set of changes in the structural locking subgraph, the locking strength index of the structural locking subgraph at each time window is calculated, and a set of locking strength indices is generated. The generation of the lock strength index set specifically includes: The change feature data of each structural locked subgraph is extracted from the change feature set of the structural locked subgraph, including the change rate of the operation index attribute of each communication entity node and the change rate of the weight of each relation edge. The locking strength index of each communication entity node is calculated based on its change feature data; the locking strength index is a weighted combination of the change rate of the node's operation index attribute and the change rate of the weight of its adjacent relation edges. The locking strength index of each relation edge is calculated based on its change feature data; the locking strength index is a weighted combination of the change rate of the relation edge's weight and the change rate of the operation index attribute of the connected communication entity nodes. The locking strength indices of all nodes and relation edges in all structural locked subgraphs are summarized to generate a locking strength index set, where the locking strength index of each structural locked subgraph in each time window includes the strength indices of all nodes and edges within that structural locked subgraph. Based on the locking strength index set, the locking strength index indices within each time window are normalized to ensure consistency in comparisons between different time windows, generating the final locking strength index set. Arrange the set of locking strength indices according to the time window order to generate a locking strength evolution sequence; In the locking strength evolution sequence, select time intervals that satisfy the locking strength index not lower than the locking strength threshold, the difference of the locking strength index between adjacent time windows not exceeding the locking fluctuation threshold, and continuously reach the threshold of the number of consecutive time windows, and generate the locking stability interval corresponding to the locking subgraph of this structure. The generation of the stable interval specifically includes: The locking strength index of each time window is extracted from the locking strength evolution sequence. The difference in locking strength index between adjacent time windows is calculated to obtain the locking strength difference of each pair of adjacent time windows. All time window pairs with locking strength differences less than or equal to the locking fluctuation threshold are selected to generate a set of continuous time window pairs that meet the conditions. In the set of continuous time window pairs, the time intervals that satisfy the condition that the locking strength index is not lower than the locking strength threshold and the number of continuous time windows reaches a preset threshold are identified to form the locking stability interval. All locking stability intervals that meet the conditions are merged to obtain the locking stability interval corresponding to the final structure locking subgraph. The locking stability intervals of each structural locking subgraph in the set of structural locking subgraphs are integrated to generate a structural locking domain.

[0022] In this embodiment, the generation of the unlock trigger edge set specifically includes: Write the standardized communication operation dataset updated in real time into the heterogeneous wireless communication knowledge graph according to the time window order, and update the operation index attributes of communication entity nodes in the structure locking domain as well as the weight attributes of topological dependency relationship edges and service carrying relationship edges. For each structural locking subgraph in the structural locking domain, the mean values ​​of the operational index attributes of the communication entity nodes and the mean values ​​of the relation edge weight attributes are statistically analyzed within the structural locking domain to determine the corresponding baseline values ​​of the operational index attributes and the baseline values ​​of the weight attributes. Within the current time window, calculate the difference between the current weight attribute and the baseline value of the corresponding weight attribute of each relation edge within the structure locking domain to obtain the relation edge weight offset. At the same time, calculate the difference between the current running indicator attribute and the baseline value of the corresponding running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the running indicator attribute offset. For each relation edge, the weight offset of the relation edge is weighted and synthesized with the offset of the running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the unlock trigger determination quantity of the relation edge. The edges whose unlock trigger determination is greater than or equal to the unlock trigger threshold are identified as unlock trigger edges, forming a set of unlock trigger edges.

[0023] In this embodiment, the generation of the minimum unlock trigger substructure set specifically includes: Based on the set of unlock trigger edges, extract the communication entity node corresponding to each unlock trigger edge and the unlock trigger determination quantity of that unlock trigger edge; Sort the unlock trigger edges in descending order according to the unlock trigger determination quantity to generate an unlock trigger edge sequence; Select the unlock trigger edge with the largest unlock trigger decision value from the unlock trigger edge sequence, extract the unlock trigger edge and the communication entity nodes at both ends of it to form a candidate unlock substructure; In the candidate unlock substructure, check whether there are other unlock trigger edges that share a communication entity node with the unlock trigger edge. The unlock trigger edges that meet the condition of sharing a communication entity node and their corresponding communication entity nodes are merged into the candidate unlock substructure to generate an extended unlock substructure. The number of communication entity nodes and the number of relation edges in the extended unlock substructure are counted. Any relation edge is removed in turn and it is determined whether the remaining structure still contains all unlock trigger edges. When the remaining structure no longer contains all unlock trigger edges after removing any relation edge, the current extended unlock substructure is determined to be the minimum unlock trigger substructure. The generation of the minimum unlock trigger substructure specifically includes: Extract the set of communication entity nodes and the set of relational edges corresponding to the extended unlocking substructure, and identify the relational edges belonging to the unlocking trigger edge set as core unlocking trigger edges; construct the connectivity structure of the extended unlocking substructure, and record the connection relationship between each core unlocking trigger edge and its two end communication entity nodes; under the premise that all core unlocking trigger edges are included, perform the removal operation on the non-core unlocking trigger edges in the relational edge set in turn, and after each relational edge is removed, re-determine whether there is still a connected path in the remaining structure that covers all core unlocking trigger edges; when the remaining structure still covers all core unlocking trigger edges after removing a non-core unlocking trigger edge, confirm that the relational edge is a redundant relational edge and delete it permanently; when the remaining structure no longer covers all core unlocking trigger edges after removal, restore the relational edge; after completing the screening of all non-core unlocking trigger edges, perform necessity verification on the remaining relational edge set, try to remove each core unlocking trigger edge one by one and determine whether the remaining structure still contains all core unlocking trigger edges; when the remaining structure no longer meets the coverage condition after removing any core unlocking trigger edge, determine that the current structure is the minimum unlocking trigger substructure; All minimum unlock trigger substructures that meet the minimum condition are aggregated to generate a minimum unlock trigger substructure set.

[0024] In this embodiment, the generation of wireless communication fault diagnosis results specifically includes: Based on the set of minimum unlock trigger substructures, extract the set of communication entity nodes and the set of relation edges corresponding to each minimum unlock trigger substructure. Based on the set of nodes in the minimum unlock trigger substructure, the root cause score of each communication entity node is calculated. The root cause score is obtained by the unlock trigger determination quantity of the communication entity node and its connection relationship edge. Based on the root cause score, the communication entity node with the highest score is selected as the root cause node. Based on the root cause node, the fault object type is determined, including base station node, cell node, wireless link node, service bearer node and alarm event node, and the wireless communication fault root cause location result is generated. In the heterogeneous wireless communication knowledge graph, a set of directed edges composed of causal propagation relationship edges is extracted. Starting from the root cause node, a path search is performed along the set of directed edges to generate a set of fault propagation paths. The sum of the unlocking trigger judgment quantities of each relation edge in each path is calculated as the path cost of that path. Based on the path cost value, the path with the minimum path cost value is selected from the set of fault propagation paths to generate the fault propagation path result. The results of wireless communication fault root cause localization and fault propagation path are combined to generate the final wireless communication fault diagnosis results.

[0025] Example 1: To verify the feasibility of the present invention in practice, the present invention was applied to a wireless communication network in a large city. The network consists of multiple base stations, wireless links and different communication services. The network provides high-frequency, high-bandwidth communication services in several important areas of the city, including residential areas, commercial areas and industrial areas. Due to the dense layout of base stations and cells and the different types of wireless links and service carrying requirements, the operation of the communication system is affected by multiple factors, which often leads to problems such as degraded communication quality, unstable signals and data packet loss in the network.

[0026] In this application scenario, wireless communication faults typically manifest as communication link failures, data transmission delays, and service interruptions. Existing traditional fault diagnosis methods are mostly based on alarm information or simple analysis of base stations and links, which makes it difficult to comprehensively and accurately identify potential fault sources in the communication system, nor can they determine the actual impact of fault propagation paths. Faced with this problem, the limitations of traditional methods often result in inaccurate fault location and solutions or untimely responses, leading to longer fault recovery times and higher network maintenance costs.

[0027] To address this issue, we applied the wireless communication fault diagnosis method of this invention. In implementation, we first collected real-time operational data from the wireless communication network, including base station operating parameters, wireless link quality indicators, service bearer status information, and related alarm event data, to establish a standardized communication operation dataset. Since this data comes from a wide range of sources and is diverse, we employed advanced preprocessing techniques to ensure data consistency and accuracy. These included time alignment, multi-source synchronization, outlier handling, and noise filtering, ultimately generating a standardized communication operation dataset. The construction of this dataset provides a precise foundation for subsequent analysis.

[0028] Based on a standardized communication operation dataset, we constructed a heterogeneous wireless communication knowledge graph. This graph not only includes a set of communication entity nodes, but also topological dependency edges, service carrying relationship edges, and causal propagation relationship edges. Each communication entity node is configured with operation indicator attributes according to its position and function in the network, and each relationship edge is also configured with weight attributes according to actual communication performance and service load. In this way, the wireless communication knowledge graph we constructed not only presents the basic characteristics of the network structure, but also reflects the specific performance data of each node and edge, providing a comprehensive perspective for subsequent fault analysis.

[0029] Using this heterogeneous wireless communication knowledge graph, we further conduct core analysis for fault diagnosis, namely, generating a set of structure-locked subgraphs. We generate a set of topology-dependent paths based on topology-dependent edges and a set of closed-loop subgraph structures based on service-bearing edges. By calculating the intersection and union relationships of the set of topology-dependent path nodes and the set of closed-loop subgraph nodes, we obtain the structural coupling degree. When the structural coupling degree reaches a preset threshold, the path and the closed-loop subgraph are considered to constitute candidate locked structures. These candidate locked structures reflect the key areas in the network that may lead to communication instability, providing a strong basis for subsequent fault localization.

[0030] To further improve the accuracy of fault diagnosis, we analyze the stability of the network by calculating the locking strength index of each structural locking subgraph. Based on the locking strength evolution sequence, we determine the locking stability interval of each structural locking subgraph, thereby accurately identifying the stable and unstable regions in the communication network. This mechanism enables us to identify which parts of the network are in a stable state at different time periods and which parts may fail, thus effectively predicting the fault propagation path.

[0031] Based on this, and combined with the standardized communication operation data updated in real time, we further calculated the unlock trigger determination quantity, generated the unlock trigger edge set, and identified the smallest fault trigger unit from it through the extraction mechanism of the smallest unlock trigger substructure. These smallest unlock trigger substructures accurately located the source of the fault in the network, making the root cause location of the fault more accurate. In this process, we applied the set of smallest unlock trigger substructures to generate the root cause location result of wireless communication faults and their propagation path, and finally generated the wireless communication fault diagnosis result.

[0032] Table 1 Performance Comparison of Wireless Communication Fault Diagnosis Methods

[0033] Based on Table 1, we compared the performance of the method of this invention and the traditional method in wireless communication fault diagnosis. The results show a significant performance improvement. First, in terms of fault location accuracy, the accuracy of the method of this invention is 85%, which is 10% higher than the 75% of the traditional method. This improvement is mainly due to the fact that the present invention adopts a more comprehensive knowledge graph construction and structure locking analysis method, which can diagnose the network from multiple dimensions, thereby more accurately identifying the fault source, rather than relying solely on the analysis of a single node or link.

[0034] Regarding fault recovery time, the fault recovery time of the method of this invention is 18 minutes, which is 40% less than the 30 minutes of the traditional method. This invention can quickly locate the fault source and identify the most likely fault propagation path through real-time fault root cause localization and analysis based on structural locking subgraph, thereby accelerating the fault recovery speed. In contrast, the traditional method requires a longer time for fault tracing and manual intervention, resulting in a longer recovery time.

[0035] In terms of false positive rate, the method of the present invention shows a significant advantage, with a false positive rate of only 5%, which is much lower than the 12% of the traditional method, a reduction of 7%. This is because the present invention can not only accurately identify the fault source, but also remove redundant information through structural locking analysis, thus avoiding false positives caused by complex network topology, interference signals and noise in the traditional method.

[0036] Regarding the accuracy of fault propagation paths, this invention also demonstrates a significant advantage, achieving an accuracy of 80%, which is 10% higher than the 70% of traditional methods. This is because this invention can add causal propagation relationship edges to the constructed knowledge graph, forming a more comprehensive propagation path analysis model, making the tracking of fault propagation paths more accurate and reliable.

[0037] Regarding fault detection time, the fault detection time of this invention is 40 seconds, which is 43% less than the 70 seconds of the traditional method. This invention can detect and identify faults more efficiently through real-time updated standardized communication operation datasets and structure locking analysis, while traditional methods often rely on slow alarm propagation and manual analysis.

[0038] In summary, this invention demonstrates significant improvements in multiple key performance indicators. By innovatively combining wireless communication knowledge graphs and structure-locked subgraph analysis, this invention enables faster and more accurate fault diagnosis, significantly improving the operation and maintenance efficiency of wireless communication networks and reducing fault recovery time and misjudgment rate.

[0039] The above are merely preferred embodiments 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 knowledge graph-based method for wireless communication fault diagnosis, characterized in that, Includes the following steps: Collect wireless communication network data and preprocess it to generate a standardized communication operation dataset; A heterogeneous wireless communication knowledge graph is constructed based on a standardized communication operation dataset, including a set of communication entity nodes and topological dependency edges, service carrying relationship edges, and causal propagation relationship edges. Operation indicator attributes are configured for the set of communication entity nodes, and weight attributes are configured for the relationship edges. In the knowledge graph of heterogeneous wireless communication, a set of structure-locked subgraphs is generated based on the topological dependency path formed by topological dependency edges and the closed-loop subgraph structure formed by service bearer relationship edges. For each structural locking subgraph in the set of structural locking subgraphs, the locking strength index is calculated based on the rate of change of the relation edge weight attribute and the offset of the operation index attribute, generating a locking strength evolution sequence. The locking stability interval is determined based on the locking strength evolution sequence, and the structural locking domain is generated. The heterogeneous wireless communication knowledge graph is updated in state based on the real-time updated standardized communication operation dataset. The unlock trigger decision quantity is calculated in the structure locking domain, and the unlock trigger edge set is generated. Based on the set of unlock trigger edges, extract the minimum unlock trigger substructure from the set of structure-locked subgraphs, and generate the set of minimum unlock trigger substructures. Based on the minimum unlock trigger substructure set, the root cause localization result and fault propagation path result of wireless communication fault are formed, and the wireless communication fault diagnosis result is generated.

2. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The wireless communication network data specifically includes base station operating parameters, wireless link quality indicators, service carrying status, topology connection relationships, and alarm event data. The preprocessing specifically includes data consistency verification, time alignment, multi-source synchronization, outlier handling, noise filtering, indicator normalization, entity identifier mapping, and time window construction.

3. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the heterogeneous wireless communication knowledge graph specifically includes: The set of communication entity nodes is extracted based on the standardized communication operation dataset, including base station nodes, cell nodes, wireless link nodes, service bearer nodes, and alarm event nodes; Configure operational indicator attributes for each communication entity node in the communication entity node set, including normalized values ​​of link quality indicators, normalized values ​​of service bearer status, normalized values ​​of base station operation parameters, normalized values ​​of alarm event counts, and normalized values ​​of topology connectivity. A set of topological dependency edges is generated based on the set of communication entity nodes. A topological dependency weight is configured for each topological dependency edge, which is determined by weighting the normalized aggregate value of the wireless link quality index and the normalized value of the topological connection strength within the corresponding time window. A set of service bearer relationship edges is generated based on the set of communication entity nodes. A service bearer weight is configured for each service bearer relationship edge, which is determined by weighting the normalized value of service load and the normalized value of user load within the corresponding time window. A set of causal propagation relationship edges is generated based on the association between alarm event nodes and communication entity nodes. A causal propagation weight is configured for each causal propagation relationship edge, which is determined according to the statistical association frequency between the preceding alarm event and the subsequent alarm event within the time lag window. The set of communication entity nodes, the set of topological dependency edges, the set of service carrying edges, and the set of causal propagation edges are merged to generate a heterogeneous wireless communication knowledge graph that includes node operation index attributes and relation edge weight attributes.

4. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the structure-locked subgraph set specifically includes: In the heterogeneous wireless communication knowledge graph, the set of communication entity nodes, the set of topological dependency edges, and the set of service bearer edges are extracted, and the topological dependency adjacency matrix and the service bearer adjacency matrix are constructed respectively. The communication entity nodes are traversed based on the topological dependency adjacency matrix to generate a set of topological dependency paths formed by continuous connection of topological dependency edges, and the corresponding node set is determined for each topological dependency path. Based on the adjacency matrix of service bearers, a closed node sequence composed of service bearer relationship edges is identified, a set of closed-loop subgraph structures is generated, and a corresponding node set is determined for each closed-loop subgraph. The intersection and union relationship of the set of nodes of topologically dependent paths and the set of nodes of closed-loop subgraphs is calculated to obtain the structural coupling degree. When the structural coupling degree reaches the structural coupling threshold, the corresponding path and closed loop are determined to form a candidate locking structure. The node set in the candidate locking structure and its corresponding topological dependency edges and business carrying relationship edges are integrated to generate a structure locking subgraph; All generated structure-locked subgraphs are combined into a structure-locked subgraph set.

5. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the structure-locked domain specifically includes: Based on the set of structure-locked subgraphs, select one structure-locked subgraph, extract the set of communication entity nodes corresponding to the structure-locked subgraph, as well as the corresponding topological dependency edges and service bearing relationship edges, and generate structure-locked subgraph state data. Within the sliding time window sequence, the weight attributes of each topological dependency edge and service bearing edge in adjacent time windows are read based on the state data of the structure-locked subgraph, the weight change rate of the edge is calculated, and the operation index attributes of each communication entity node in adjacent time windows are read, the offset of the operation index attributes is calculated, and a set of change features of the structure-locked subgraph is generated. Based on the set of changes in the structural locking subgraph, the locking strength index of the structural locking subgraph at each time window is calculated, and a set of locking strength indices is generated. Arrange the set of locking strength indices according to the time window order to generate a locking strength evolution sequence; In the locking strength evolution sequence, select time intervals that satisfy the locking strength index not lower than the locking strength threshold, the difference of the locking strength index between adjacent time windows not exceeding the locking fluctuation threshold, and continuously reach the threshold of the number of consecutive time windows, and generate the locking stability interval corresponding to the locking subgraph of this structure. The locking stability intervals of each structural locking subgraph in the set of structural locking subgraphs are integrated to generate a structural locking domain.

6. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the unlock trigger edge set specifically includes: Write the standardized communication operation dataset updated in real time into the heterogeneous wireless communication knowledge graph according to the time window order, and update the operation index attributes of communication entity nodes in the structure locking domain as well as the weight attributes of topological dependency relationship edges and service carrying relationship edges. For each structural locking subgraph in the structural locking domain, the mean values ​​of the operational index attributes of the communication entity nodes and the mean values ​​of the relation edge weight attributes are statistically analyzed within the structural locking domain to determine the corresponding baseline values ​​of the operational index attributes and the baseline values ​​of the weight attributes. Within the current time window, calculate the difference between the current weight attribute and the baseline value of the corresponding weight attribute of each relation edge within the structure locking domain to obtain the relation edge weight offset. At the same time, calculate the difference between the current running indicator attribute and the baseline value of the corresponding running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the running indicator attribute offset. For each relation edge, the weight offset of the relation edge is weighted and synthesized with the offset of the running indicator attribute of the communication entity nodes at both ends of the relation edge to obtain the unlock trigger determination quantity of the relation edge. The edges whose unlock trigger determination is greater than or equal to the unlock trigger threshold are identified as unlock trigger edges, forming a set of unlock trigger edges.

7. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the minimum unlock trigger substructure set specifically includes: Based on the set of unlock trigger edges, extract the communication entity node corresponding to each unlock trigger edge and the unlock trigger determination quantity of that unlock trigger edge; Sort the unlock trigger edges in descending order according to the unlock trigger determination quantity to generate an unlock trigger edge sequence; Select the unlock trigger edge with the largest unlock trigger decision value from the unlock trigger edge sequence, extract the unlock trigger edge and the communication entity nodes at both ends of it to form a candidate unlock substructure; In the candidate unlock substructure, check whether there are other unlock trigger edges that share a communication entity node with the unlock trigger edge. The unlock trigger edges that meet the condition of sharing a communication entity node and their corresponding communication entity nodes are merged into the candidate unlock substructure to generate an extended unlock substructure. The number of communication entity nodes and the number of relation edges in the extended unlock substructure are counted. Any relation edge is removed in turn and it is determined whether the remaining structure still contains all unlock trigger edges. When the remaining structure no longer contains all unlock trigger edges after removing any relation edge, the current extended unlock substructure is determined to be the minimum unlock trigger substructure. All minimum unlock trigger substructures that meet the minimum condition are aggregated to generate a minimum unlock trigger substructure set.

8. The wireless communication fault diagnosis method based on knowledge graphs according to claim 1, characterized in that, The generation of the wireless communication fault diagnosis results specifically includes: Based on the set of minimum unlock trigger substructures, extract the set of communication entity nodes and the set of relation edges corresponding to each minimum unlock trigger substructure. Based on the set of nodes in the minimum unlock trigger substructure, the root cause score of each communication entity node is calculated. The root cause score is obtained by the unlock trigger determination quantity of the communication entity node and its connection relationship edge. Based on the root cause score, the communication entity node with the highest score is selected as the root cause node. Based on the root cause node, the fault object type is determined, including base station node, cell node, wireless link node, service bearer node and alarm event node, and the wireless communication fault root cause location result is generated. In the heterogeneous wireless communication knowledge graph, a set of directed edges composed of causal propagation relationship edges is extracted. Starting from the root cause node, a path search is performed along the set of directed edges to generate a set of fault propagation paths. The sum of the unlocking trigger judgment quantities of each relation edge in each path is calculated as the path cost of that path. Based on the path cost value, the path with the minimum path cost value is selected from the set of fault propagation paths to generate the fault propagation path result. The results of wireless communication fault root cause localization and fault propagation path are combined to generate the final wireless communication fault diagnosis results.

Images