Packet transport network fault identification and positioning method and device, and storage medium

Abstract

The invention relates to the technical field of transmission network fault management, and provides a packet transmission network fault identification and positioning method and device and a storage medium, and the method comprises the steps: obtaining alarm information, performance information and topology connection information of a packet transmission network; acquiring a first topological sub-graph of the equipment node; acquiring image feature information; inputting the graph feature information into a fault recognition model to obtain fault recognition result information; obtaining time sequence sub-graph information associated with the target link; obtaining dynamic graph node coding information; and inputting the dynamic graph node coding information into the fault positioning model to obtain fault positioning result information. Fault recognition is realized from the perspective of a graph through comprehensive performance and alarm multi-dimensional information, and the accuracy of fault recognition is improved. Time sequence features and topological features are fused for fault positioning, fault root causes are positioned from multiple dimensions of information time sequence changes and topological connection changes, and the accuracy and reliability of fault positioning are improved.

Description

Technical Field

[0001] This invention relates to the technical field of transmission network fault management, and in particular to a method, device and storage medium for identifying and locating faults in packet transmission networks. Background Technology

[0002] Transmission networks, as crucial hubs connecting various communication devices and users, play a vital role in communication systems. Among them, packet transport networks (PTNs), due to their efficiency and flexibility, have become the mainstream trend in current network evolution and carry a large amount of service traffic. With the continuous increase in the scale and complexity of packet transport networks, fault management has become one of the key challenges in ensuring network operational stability and service quality.

[0003] In existing technologies, fault location methods for packet transmission networks can be broadly categorized into manual fault location and fault location through monitoring key performance indicators. However, existing fault location methods for packet transmission networks have several problems: Manual fault location is inefficient and unreliable, relying on manual experience to check alarm logs item by item, which requires traversing complex topologies and is time-consuming and labor-intensive; it requires a large number of professional operation and maintenance personnel, resulting in high labor costs, especially in scenarios with frequent faults where the operation and maintenance burden increases significantly; manual fault location is easily affected by subjective factors, such as insufficient experience or negligence, which may lead to misjudgment or omission of critical faults.

[0004] Key performance indicator monitoring: Monitoring a single performance indicator has significant limitations. Relying solely on fluctuations in key performance indicators to determine faults can easily overlook other factors that affect faults and can lead to false triggering. For example, a sudden surge in business volume during holidays can cause fluctuations in key performance indicators, leading to false fault location when the fault is not actually a device failure. At the same time, since faulty nodes can affect other nodes, it is difficult to distinguish between the source fault node and the affected node, resulting in difficulty in accurately locating the fault.

[0005] Therefore, how to achieve efficient and accurate fault identification and location in packet transmission networks is an urgent problem to be solved. Summary of the Invention

[0006] This invention provides a method, device, and storage medium for fault identification and location in packet transmission networks, which addresses the shortcomings of existing technologies in fault location for packet transmission networks, such as false triggering and inaccurate fault location.

[0007] This invention provides a method for fault identification and location in packet transmission networks, comprising: Obtain alarm information, performance information, and topology connection information of the packet transmission network; Based on the topology connection information, a first topology subgraph corresponding to the device nodes in the packet transmission network is obtained, and the first topology subgraph represents the connection relationship of the device nodes; Based on the first topological subgraph, the alarm information, and the performance information, graph feature information is obtained, and the graph feature information is determined by fusing topological features, alarm features, and performance features. The graph feature information is input into the fault identification model to obtain fault identification result information; Based on the fault identification results, the faulty link is selected as the target link. Based on the target link and a preset time window, obtain the time-series subgraph information associated with the target link, the time-series subgraph information including the current and historical first topology subgraphs; Based on the temporal subgraph information, dynamic graph node encoding information is obtained, which is determined by the fusion of topological features and temporal features; The dynamic graph node encoding information is input into the fault location model to obtain fault location result information.

[0008] According to a method for fault identification and location in a packet transmission network provided by the present invention, the step of obtaining a first topology subgraph corresponding to a device node in the packet transmission network based on the topology connection information includes: Based on the topology connection information, a global device connection topology map of the packet transmission network is constructed, and the global device connection topology map represents the topology of the packet transmission network; Based on the global device connection topology and the preset number of device influence layers, the first topology subgraph corresponding to each device node is obtained.

[0009] According to a method for fault identification and location in a packet transmission network provided by the present invention, the step of obtaining graph feature information based on the first topology subgraph, the alarm information, and the performance information includes: Based on the first topology subgraph and the alarm information, obtain alarm coding matrix information; based on the first topology subgraph and the performance information, obtain performance coding matrix information. Based on the alarm coding matrix information and the performance coding matrix information, obtain the graph coding matrix information; Based on the first topological subgraph, an adjacency matrix and a degree matrix are determined. The adjacency matrix represents the connection relationship between devices, and the degree matrix represents the number of device connections. The graph encoding matrix information, the adjacency matrix, and the degree matrix form the graph feature information; The alarm coding matrix information represents the number of alarms occurring for each device node in the first topology subgraph, and the performance coding matrix information represents the average performance of each device node in the first topology subgraph.

[0010] According to a method for fault identification and location in a packet transmission network provided by the present invention, the step of obtaining dynamic graph node encoding information based on the time-series subgraph information includes: Based on the time-series subgraph information, the corresponding time-series diffusion matrix information is obtained, wherein the time-series diffusion matrix information represents the connectivity between current and historical device nodes; Based on the time-series diffusion matrix information, determine the target device node information with the highest connectivity to the target link; Based on the time-series diffusion matrix information, the target device node information, and the time-series subgraph information, obtain the diffusion node connectivity level matrix, the diffusion distance encoding matrix, and the time encoding matrix; Based on the performance information, the target device node information, and the time encoding matrix, obtain the attribute encoding matrix; The dynamic graph node encoding information is obtained based on the diffusion node connectivity level matrix, the diffusion distance encoding matrix, the time encoding matrix, and the attribute encoding matrix.

[0011] According to a method for fault identification and location in a packet transmission network provided by the present invention, the step of obtaining the corresponding time series diffusion matrix information based on the time series subgraph information includes: Based on the temporal subgraph information, determine the current adjacency matrix and current degree matrix corresponding to the current first topological subgraph, and determine the historical adjacency matrix and historical degree matrix corresponding to the historical first topological subgraph. Based on the current adjacency matrix and the current degree matrix, obtain the current diffusion matrix; based on the historical adjacency matrix and the historical degree matrix, obtain the historical diffusion matrix. The time-series diffusion matrix information is formed based on the current diffusion matrix and the historical diffusion matrix.

[0012] According to a method for fault identification and location in a packet transmission network provided by the present invention, after obtaining the fault location result information, the method further includes: Based on the fault location results, the root cause information of the fault is determined; Based on the root cause information of the fault, retrieve the corresponding recommended solution information from the experience knowledge base; Based on the root cause information of the fault and the recommended solution information, a fault report is generated.

[0013] According to a method for fault identification and location in a packet transmission network provided by the present invention, the fault identification model is obtained in the following ways: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the second topology subgraph corresponding to the work order fault node; Based on the second topology subgraph and the historical alarm information and historical performance information corresponding to the work order fault node, obtain the fault feature coding information; Based on the work order information, determine the first training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal feature encoding information is obtained, and the fault feature encoding information and the normal feature encoding information form a first training dataset. Based on the first training dataset and the first training label, the graph convolutional neural network model is trained until training is complete, and the graph convolutional neural network model is used as the fault identification model.

[0014] According to a method for fault identification and location in a packet transmission network provided by the present invention, the fault location model is obtained in the following ways: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the fault sequence subgraph information corresponding to the fault node of the work order; Based on the fault timing subgraph information, obtain the fault dynamic graph node encoding information; Based on the work order information, the root cause of the fault is determined as the second training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal dynamic graph node encoding information is obtained, and the fault dynamic graph node encoding information and the normal dynamic graph node encoding information form a second training dataset. The graph attention network model is trained based on the second training dataset and the second training label until training is complete, and the graph attention network model is used as the fault localization model.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a packet transmission network fault identification and location method as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a packet transmission network fault identification and location method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements a packet transmission network fault identification and location method as described above.

[0018] This invention provides a method, device, and storage medium for fault identification and location in packet transmission networks, which has at least the following advantages: First, a first topological subgraph of device nodes is obtained based on first topological connection information. This first topological subgraph reflects the scope of impact when a device node fails. Based on the first topological subgraph, corresponding information is obtained by combining alarm information and performance information to acquire graph feature information, which characterizes the features of the first topological subgraph. The graph feature information of each device node is input into a fault identification model to obtain fault identification results. Thus, based on the network topology, fault identification is performed by combining performance information and alarm information. This approach comprehensively integrates performance and alarm information from a graph perspective to achieve fault identification, avoiding fault identification based solely on a single device node and single-dimensional information, thereby improving the accuracy of fault identification. Simultaneously, based on the fault identification results, when a fault is identified, the faulty link (i.e., the target link) and associated temporal subgraph information are further determined. Since temporal subgraph information reflects the dynamic temporal changes in the network topology, the topological features and temporal features are fused to obtain dynamic graph node encoding information, which reflects the dynamic characteristics of the device nodes. The encoding information of the first homomorphic node is input into the fault location model to obtain the fault location result information. In this way, temporal features and topological features are fused for fault location, which adapts to the characteristics of dynamic network changes. The root cause of the fault is located from multiple dimensions of information temporal changes and topological connection changes, which helps to improve the accuracy and reliability of fault location. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a method for identifying and locating faults in a packet transmission network provided by the present invention.

[0021] Figure 2 This is a schematic diagram of information processing for a packet transmission network fault identification and location method provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the first topology subgraph in one embodiment of a fault identification and location method for packet transmission networks provided by the present invention.

[0023] Figure 4 This is a schematic diagram of the processing procedure of the fault location model in one embodiment of the fault identification and location method for packet transmission networks provided by the present invention.

[0024] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] The following is combined with Figure 1 and Figure 2 The present invention describes a method for fault identification and location in a packet transmission network, characterized in that it includes: S100: Obtain alarm information, performance information, and topology connection information of the packet transmission network; S110: Based on the topology connection information, obtain a first topology subgraph corresponding to the device nodes in the packet transmission network, wherein the first topology subgraph represents the connection relationship of the device nodes; S120: Obtain graph feature information based on the first topology subgraph, the alarm information, and the performance information; S130: Input the graph feature information into the fault identification model to obtain fault identification result information; S200: Based on the fault identification result information, the faulty link is selected as the target link; S210: Based on the target link and a preset time window, obtain the time-series subgraph information associated with the target link, wherein the time-series subgraph information includes the current and historical first topology subgraphs; S220: Based on the temporal subgraph information, obtain dynamic graph node encoding information, wherein the dynamic graph node encoding information is determined by fusing topological features and temporal features; S230: Input the dynamic graph node encoding information into the fault location model to obtain fault location result information.

[0027] A first topological subgraph of device nodes is obtained based on the first topological connection information. This first topological subgraph reflects the scope of impact when a device node fails. Based on this first topological subgraph, and combined with alarm and performance information, corresponding information is obtained to acquire graph feature information, which characterizes the features of the first topological subgraph. The graph feature information of each device node is input into a fault identification model to obtain the fault identification result. Therefore, fault identification based on the network topology, combined with performance and alarm information, can comprehensively identify faults from a graph perspective, integrating performance and alarm information from multiple dimensions. This avoids fault identification based solely on a single device node and single-dimensional information, thus improving the accuracy of fault identification.

[0028] Simultaneously, based on the fault identification results, when a fault is confirmed, the faulty link, i.e., the target link, and its associated temporal subgraph information are further determined. Since the temporal subgraph information reflects the dynamic temporal changes in the network topology, the topological and temporal features are fused to obtain dynamic graph node encoding information, which reflects the dynamic characteristics of device nodes. The first homomorphic node encoding information is input into the fault location model to obtain fault location results. In this way, by fusing temporal and topological features for fault location, the characteristics of dynamic network changes are adapted, and the root cause of the fault is located from multiple dimensions of information temporal changes and topological connection changes, which helps to improve the accuracy and reliability of fault location.

[0029] Fault identification refers to identifying the fault situation and the corresponding network range affected, while fault location refers to locating the specific cause of the fault. Based on the fault identification model and the fault location model, the process is divided into two stages: fault identification and fault location. When no fault occurs, only fault identification is performed, without the need for fault location. Compared to methods that do not distinguish between the two processes, this avoids the simultaneous operation of fault identification and location under normal circumstances, which is beneficial to improving the efficiency of processing resource utilization.

[0030] It is understandable that alarm information and performance information include alarm data generated by multiple device nodes in a packet transmission network, as well as operational performance data. Alarm information refers to alarms generated by device nodes during operation, which are not necessarily caused by device failures. They may also be caused by sensitive operations, abnormal transmission processes, etc. Alarm information serves as one of the important reference information when a failure occurs.

[0031] In some embodiments of the present invention, alarm information includes alarm time, alarm title, alarm device, etc.; performance information includes network element status, optical power, service layer traffic, etc.; topology connection information includes uplink and downlink relationships between devices, etc.

[0032] In some embodiments of the present invention, after acquiring alarm information, performance information, and topology connection information, the alarm information, performance information, and topology connection information are preprocessed: Based on the alarm information, fields such as alarm device, alarm title, alarm occurrence time, and city / region are extracted. Information lacking alarm device name, alarm occurrence time, or alarm title is removed, as are alarms whose names do not contain an alarm title. The filtered alarm titles are then coded. This process eliminates useless information, making alarm titles easier to process and identify.

[0033] Based on the performance information, null values ​​are filled with the mean and normalized.

[0034] Based on the topology connection information, extract fields such as source network element (local network element), destination network element (remote network element), and the city / prefecture to which it belongs.

[0035] In some embodiments of the present invention, the normalization of performance information can be achieved using the following formula: in, The normalized value. This represents the current performance value of the device. This is the same performance value as other devices.

[0036] refer to Figure 2 In some embodiments of the packet transmission network fault identification and location method of the present invention, after S230, the method further includes: Based on the fault location results, the root cause information of the fault is determined; Based on the root cause information of the fault, retrieve the corresponding recommended solution information from the experience knowledge base; Based on the root cause information of the fault and the recommended solution information, a fault report is generated.

[0037] After obtaining the fault location results, the root cause information is determined to identify the fundamental reason for the fault. Corresponding recommended solutions are retrieved from the experience knowledge base to determine the appropriate solutions. Finally, a fault report is generated based on the root cause information and the recommended solutions. This fault report enables maintenance personnel to quickly locate the root cause of the fault and provides recommended solutions, facilitating rapid fault handling and improving efficiency.

[0038] In some embodiments of the present invention, the experience knowledge base may be constructed based on the results of historical work orders, by statistically analyzing the causes of failures and the corresponding handling measures. The information in the experience knowledge base may include items such as the root cause category of the failure, alarm title, description of the cause of the failure, and recommended handling measures.

[0039] refer to Figure 3 In some embodiments of the packet transmission network fault identification and location method of the present invention, step S110 includes: Based on the topology connection information, a global device connection topology map of the packet transmission network is constructed, and the global device connection topology map represents the topology of the packet transmission network; Based on the global device connection topology and the preset number of device influence layers, the first topology subgraph corresponding to each device node is obtained.

[0040] A global graph, or global device connection topology graph, is constructed based on topology connectivity information. This graph reflects the connection relationships between various device nodes in a packet transmission network. Figure 3 As shown, a preset number of device impact layers is used to estimate the scope of the impact of a faulty device. Based on the preset number of device impact layers and the global device connection topology graph, a first topology subgraph corresponding to each device node is obtained. The first topology subgraph reflects the scope of the impact when a device node fails. Therefore, further dividing the global device connection topology graph into the first topology subgraph helps reduce the amount of data processed, eliminating the need to process the global device connection topology graph simultaneously, reducing the computational power requirements, and facilitating processing by dividing the graph into multiple subtasks based on the first topology subgraph.

[0041] Based on the first topology subgraph, from the perspective of graph information, the overall fault impact range is considered, rather than analyzing a single device node. Fault identification based on the first topology subgraph is beneficial to improving the accuracy of identification and providing a foundation for subsequent fault location.

[0042] refer to Figure 3 In some embodiments of the present invention, in the global device connection topology graph, the set of device nodes can be represented by {Nodej}, where j=1, 2, ..., Z is the number of the device node, Z is the total number of device nodes, and there are no undirected edges.<A,B> This represents the connection relationship between device node A and device node B. The preset influence layer hop is set to 2. The first topology subgraph corresponding to device node C is determined with device node C as the initial node and based on hop as the influence radius. Figure 3 As shown, the first topological subgraph can also be understood as the alarm propagation subgraph.

[0043] refer to Figure 2 In some embodiments of the packet transmission network fault identification and location method of the present invention, step S120 includes: Based on the first topology subgraph and the alarm information, obtain alarm coding matrix information; based on the first topology subgraph and the performance information, obtain performance coding matrix information. Based on the alarm coding matrix information and the performance coding matrix information, obtain the graph coding matrix information; Based on the first topological subgraph, an adjacency matrix and a degree matrix are determined. The adjacency matrix represents the connection relationship between devices, and the degree matrix represents the number of device connections. The graph encoding matrix information, the adjacency matrix, and the degree matrix form the graph feature information; The alarm coding matrix information represents the number of alarms occurring for each device node in the first topology subgraph, and the performance coding matrix information represents the average performance of each device node in the first topology subgraph.

[0044] Based on the device nodes included in the first topology subgraph, information corresponding to the device nodes is obtained from alarm information and processed to form alarm coding matrix information, reflecting the alarm characteristics corresponding to the first topology subgraph. Similarly, information corresponding to the device nodes is obtained from performance information and processed to form performance coding matrix information, reflecting the performance characteristics corresponding to the first topology subgraph. Based on the alarm coding matrix information and performance coding matrix information, the alarm characteristics and performance characteristics are fused to obtain graph coding matrix information, reflecting the characteristics of the first topology subgraph and providing an information foundation for subsequent fault identification. Furthermore, based on the first topology subgraph, adjacency matrices representing the connection relationships between devices and the number of other devices connected to a single device are obtained; these adjacency matrices and degree matrices reflect the topological characteristics of the devices. Thus, graph feature information is formed based on the graph coding matrix information, adjacency matrix, and degree matrix, facilitating fault identification from the topological features, performance characteristics, and alarm characteristics reflected in the graph. This multi-dimensional comprehensive fault identification improves the accuracy of fault identification.

[0045] In some embodiments of the present invention, based on the first topological subgraph Construct an adjacency matrix ,in For the first topological subgraph The total number of device nodes in the middle, denoted as The element in the i-th row and j-th column of the adjacency matrix. =1 indicates that device node i and device node j have a data transmission relationship; otherwise... =0. In some embodiments, the adjacency matrix... As shown below: Matrix adjacency matrix The corresponding degree matrix As shown below: The element in the i-th row and i-th column represents the number of connected devices to device node i.

[0046] Alarm and performance information corresponding to the first topology subgraph can be retrieved according to a preset time window, such as a 30-minute time window. For the first topology subgraph... , obtain The alarms of each device node are collected and classified according to alarm type, and an alarm coding matrix is ​​constructed. Where K is the total number of alarm categories, For the first topological subgraph The total number of device nodes in the system. Representative alarm coding matrix The element in the k-th row and j-th column represents the number of times the k-th type of alarm occurred on device node j within 30 minutes. Based on the first topology subgraph. The performance information of each device node over a 30-minute period is obtained and normalized to facilitate the analysis of different performance indicators and to construct a performance coding matrix. Where P is the total number of performance categories, For the first topological subgraph The total number of device nodes in the middle. Representative matrix The p-th row and j-th column represents the average performance p of device node j over 30 minutes. This is used in constructing the alarm coding matrix. and performance coding matrix Then, the two are stacked to form a graph-coded matrix information. .

[0047] In some embodiments of the packet transmission network fault identification and location method of the present invention, step S220 includes: Based on the time-series subgraph information, the corresponding time-series diffusion matrix information is obtained, wherein the time-series diffusion matrix information represents the connectivity between current and historical device nodes; Based on the time-series diffusion matrix information, determine the target device node information with the highest connectivity to the target link; Based on the time-series diffusion matrix information, the target device node information, and the time-series subgraph information, obtain the diffusion node connectivity level matrix, the diffusion distance encoding matrix, and the time encoding matrix; Based on the performance information, the target device node information, and the time encoding matrix, obtain the attribute encoding matrix; The dynamic graph node encoding information is obtained based on the diffusion node connectivity level matrix, the diffusion distance encoding matrix, the time encoding matrix, and the attribute encoding matrix.

[0048] Because the connections between devices in a packet transmission network change dynamically, meaning the topology connection information is dynamic, the time-series subgraph information, including the current first topology subgraph and historical first topology subgraphs, reflects changes in device node connections. Based on the time-series subgraph information, the time-series diffusion matrix information is obtained, thereby identifying the target device node with the highest connectivity to the faulty target link. This highest connectivity target device node is the potential cause of the fault. Based on the target device node information, combined with the time-series diffusion matrix information and the time-series subgraph information, the diffusion node connectivity level matrix, diffusion distance encoding matrix, and time encoding matrix are obtained to characterize the connectivity, diffusion distance, and timestamp of each target device node. Based on the target node information, combined with performance information and the time encoding matrix, an attribute encoding matrix is ​​obtained to characterize the attribute encoding of each target node. Combining the diffusion node connectivity level matrix, diffusion distance encoding matrix, time encoding matrix, and attribute encoding matrix, dynamic graph node encoding information is obtained. This achieves the integration of device node connectivity, diffusion distance, and device attributes, considering temporal changes, thus achieving the effect of fusing topology and temporal features.

[0049] Therefore, the dynamic graph node encoding feature integrates topological features and temporal features, which makes it easier to consider the temporal changes in connectivity during fault location, avoids fault location failure or inaccurate fault location due to changes in connectivity, and helps to locate the cause of the fault more accurately.

[0050] In some embodiments of the packet transmission network fault identification and location method of the present invention, the step of obtaining the corresponding time series graph diffusion matrix information based on the time series subgraph information includes: Based on the temporal subgraph information, determine the current adjacency matrix and current degree matrix corresponding to the current first topological subgraph, and determine the historical adjacency matrix and historical degree matrix corresponding to the historical first topological subgraph. Based on the current adjacency matrix and the current degree matrix, obtain the current diffusion matrix; based on the historical adjacency matrix and the historical degree matrix, obtain the historical diffusion matrix. The time-series diffusion matrix information is formed based on the current diffusion matrix and the historical diffusion matrix.

[0051] The temporal subgraph information includes the current and historical first topological subgraphs. Based on the current and historical first topological subgraphs, the current adjacency matrix and current degree matrix, as well as the historical adjacency matrix and historical degree matrix, are obtained respectively, and then the current diffusion matrix and historical diffusion matrix are obtained. The current diffusion matrix and historical diffusion matrix are combined to reflect the temporal changes of the diffusion matrix, forming the temporal graph diffusion matrix.

[0052] In some embodiments of the present invention, the graph diffusion matrix can be represented as: Where S is the graph diffusion matrix; T is the state transition matrix obtained by normalizing the adjacency matrix A, and the eigenvalues ​​of T are... θ is the weighting coefficient, satisfying ; k is the device node number. The i-th row of the graph diffusion matrix S represents the connectivity between device node i and each device node k.

[0053] In some embodiments of the present invention, the graph diffusion matrix can be obtained based on the PageRank algorithm: Where α is the damping coefficient; In is the n-order identity matrix; D is the degree matrix; and A is the adjacency matrix.

[0054] Based on the above method of obtaining the graph diffusion matrix, the temporal subgraph information includes the current first topological subgraph and the historical first topological subgraph, and the corresponding current graph diffusion matrix is ​​obtained. and historical diffusion matrix This is combined to form a time-series diffusion matrix.

[0055] Based on the time-series diffusion matrix, for example, the graph edges corresponding to the target link of the fault are... Then select the first link in the time series diffusion matrix corresponding to the target link. row and number Okay, remove The elements are selected by choosing the top n largest numbers. In some embodiments, n is 5, in order to determine the top n device nodes with the greatest connectivity to i and j, and to obtain the target device node information.

[0056] It should be noted that the time-series subgraph information contains at least one historical first topological subgraph, and can be implemented with multiple historical first topological subgraphs. Correspondingly, the historical graph diffusion matrix can also have multiple components, such as... , In some embodiments, a timing window can be set to adjust the scope of timing considerations. For example, if the timing window τ=2, the timing subgraph information includes the current first topological subgraph and the previous historical first topological subgraph.

[0057] After determining the target device node information, the connectivity level of the target device node is extracted from the diffusion matrix of the time sequence diagram, and a diffusion node connectivity level matrix is ​​constructed. ,in, , m represents the number of the top n nodes with the highest connectivity and the number of edge nodes in the time window τ. The number of edges represents the number of target links. Based on the target device node information, the diffusion distance between the target device nodes and the target links is extracted, and a diffusion distance encoding matrix is ​​constructed. Based on the target device node information, timestamps are extracted to indicate the recording time of the information, and a time encoding matrix is ​​constructed. Based on the target device node information, combined with performance information and a time encoding matrix, the performance metrics of the target device node at the timestamp are processed, such as by inputting them into an LSTM encoding matrix to obtain the attribute encoding matrix. The encoding information of a dynamic graph node can be represented as: The processing of dynamic graph node encoding information is as follows: Figure 4 As shown.

[0058] In some embodiments of the packet transmission network fault identification and location method of the present invention, the fault identification model is obtained in the following ways: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the second topology subgraph corresponding to the work order fault node; Based on the second topology subgraph and the historical alarm information and historical performance information corresponding to the work order fault node, obtain the fault feature coding information; Based on the work order information, determine the first training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal feature encoding information is obtained, and the fault feature encoding information and the normal feature encoding information form a first training dataset. Based on the first training dataset and the first training label, the graph convolutional neural network model is trained until training is complete, and the graph convolutional neural network model is used as the fault identification model.

[0059] Work order information is information recorded by maintenance personnel during the troubleshooting process. Based on work order information, historical alarm information, historical performance information, and historical topology connection information, a second topology subgraph is constructed for the faulty work order device nodes. Then, combined with historical alarm and performance information, fault feature encoding information is obtained. Additionally, based on historical alarm, performance, and topology connection information, a third topology subgraph is constructed for normally operating device nodes. Then, combined with the corresponding historical alarm and performance information, normal feature encoding information is obtained. The fault and normal feature encoding information form the first training dataset. Simultaneously, based on the work order information, the first training label is determined. The graph convolutional neural network model is trained until convergence and the required number of training epochs are met. The trained graph convolutional neural network model is then used as the fault identification model. Therefore, based on the graph convolutional neural network, the first topology subgraph, adjacency matrix, degree matrix, and encoded performance and alarm features are fused for model training to obtain a fault identification model. This fault identification model can accurately identify whether a fault has occurred and the network topology region where the fault occurs based on the input feature encoding information.

[0060] The work order information includes the name of the faulty device, the time of the fault, the type of fault cause, and the city / prefecture where the fault occurred. In some embodiments of the present invention, after obtaining the work order information, historical alarm information, historical performance information, and historical topology connection information, preprocessing can be performed first. For the work order information, missing information such as the name of the faulty device, the time of the fault, and the type of fault cause is removed. The root cause of the fault is extracted from the fault cause category, and the extracted root cause is encoded. The preprocessing of historical alarm information, historical performance information, and historical topology connection information can be the same as the preprocessing of alarm information, performance information, and topology connection information described above, and will not be repeated here.

[0061] In some embodiments of the present invention, the construction method of the second topology subgraph can be the same as that of the first topology subgraph, that is, first constructing a global device connection topology graph and then using the work order device node as the initial node, and determining the second topology subgraph according to the number of device influence layers; in some embodiments, since the work order device node has been determined to be a fault node, the second topology subgraph can be constructed directly according to the historical topology connection information and the number of elements and device influence layers.

[0062] In some embodiments of the present invention, in a Graph Convolution Network (GCN), the convolution module can be represented as: in, It is the first Layer model node representation; It is the graph encoding matrix information of the topological subgraph i. ; It is the adjacency matrix of the topological subgraph i. ; It is an adjacency matrix The corresponding degree matrix; yes The weight matrix to be learned for the layer. ; It is the first The input feature dimension of the layer; It is the first The output feature dimension of the layer P represents the total number of performance categories, and K represents the total number of alarm categories.

[0063] In some embodiments of the packet transmission network fault identification and location method of the present invention, the fault location model is obtained in the following ways: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the fault sequence subgraph information corresponding to the fault node of the work order; Based on the fault timing subgraph information, obtain the fault dynamic graph node encoding information; Based on the work order information, the root cause of the fault is determined as the second training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal dynamic graph node encoding information is obtained, and the fault dynamic graph node encoding information and the normal dynamic graph node encoding information form a second training dataset. The graph attention network model is trained based on the second training dataset and the second training label until training is complete, and the graph attention network model is used as the fault localization model.

[0064] Based on work order information, historical alarm information, historical performance information, and historical topology connection information, a fault time-series subgraph of the faulty work order device nodes is constructed. Then, combined with historical performance information, fault dynamic graph node encoding information is obtained. Additionally, based on historical alarm information, historical performance information, and historical topology connection information, normal time-series subgraph information corresponding to normally operating device nodes is obtained, and normal dynamic graph node encoding information is obtained. The fault dynamic graph node encoding information and the normal dynamic graph node encoding information form a second training dataset. Simultaneously, based on the work order information, the root cause of the fault is determined as a second training label. The graph attention network model is trained accordingly until the model converges and reaches the required number of training epochs, thus completing the training. The trained graph attention network model is then used as the fault localization model. In this way, the fault localization model can accurately locate the root cause of the fault based on the input dynamic graph node encoding information.

[0065] In some embodiments of the present invention, the processing procedure of the fault location model is as follows: Figure 4 As shown, the fault location model includes a Transformer module, and the processing of the Transformer module can be represented as follows: in, It is the first Layer output; It is the first Layer input; It is the first Layer embedding dimension; τ is the timing window; n is the number of device nodes with the highest connectivity. . The contribution of device node coding was calculated. Map the input to a new feature space.

[0066] The loss function used in the fault location model is cross-entropy loss.

[0067] In some embodiments of the present invention, for cases with a small number of fault labels, two methods can be used to increase the number of abnormal labels: one is to use negative sampling processing, by replacing normal nodes and constructing abnormal data that differs significantly from the original data; the other is to use data augmentation to generate abnormal label data that is similar to the original abnormality.

[0068] The following describes a packet transmission network fault identification and location device provided by the present invention. The packet transmission network fault identification and location device described below and the packet transmission network fault identification and location method described above can be referred to in correspondence.

[0069] The present invention also provides a fault identification and location device for packet transmission networks, comprising: The acquisition module is used to acquire alarm information, performance information, and topology connection information of the packet transmission network. The graph building module is used to build topological subgraphs; The graph feature fusion module is used to obtain graph feature information based on topological subgraphs, alarm information, and performance information. The fault identification module is used to obtain fault identification results based on graph feature information. The timing diagram module is used to obtain timing sub-diagram information associated with the target link based on the target link of the fault and a preset time window; The temporal feature fusion module is used to obtain the dynamic graph node encoding information based on the temporal subgraph information; The fault location module is used to obtain fault location results based on the node encoding information of the dynamic graph. The reporting module is used to determine the root cause of the fault based on the fault location results, retrieve the corresponding recommended solutions from the experience knowledge base, and generate a fault report.

[0070] The first topological subgraph of device nodes is obtained based on the first topological connection information. This first topological subgraph reflects the scope of impact when a device node fails. Based on this first topological subgraph, combined with alarm and performance information, corresponding information is obtained to acquire graph feature information, which characterizes the features of the first topological subgraph. The graph feature information of each device node is input into the fault identification model to obtain the fault identification result. Thus, based on the network topology, combined with performance and alarm information, fault identification can be performed from a graph perspective, comprehensively considering multi-dimensional performance and alarm information to achieve fault identification, avoiding fault identification based solely on a single device node and single-dimensional information, thereby improving the accuracy of fault identification. Simultaneously, based on the fault identification result, when a fault is identified, the faulty link (i.e., the target link) and associated temporal subgraph information are further determined. Since temporal subgraph information reflects the dynamic temporal changes in the network topology, the topological and temporal features are fused to obtain dynamic graph node encoding information, reflecting the dynamic characteristics of device nodes. The first homomorphic node encoding information is input into the fault location model to obtain the fault location result information. Therefore, by integrating temporal and topological features for fault location, we can adapt to the dynamic changes in the network and locate the root cause of the fault from multiple dimensions of information temporal changes and topological connection changes, which is conducive to improving the accuracy and reliability of fault location.

[0071] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions stored in the memory 830 to execute the aforementioned packet transmission network fault identification and location method.

[0072] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute a packet transmission network fault identification and location method provided by the above methods.

[0074] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a packet transmission network fault identification and location method provided by the methods described above.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0077] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0078] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fault identification and location in a packet transmission network, characterized in that, include: Obtain alarm information, performance information, and topology connection information of the packet transmission network; Based on the topology connection information, a first topology subgraph corresponding to the device nodes in the packet transmission network is obtained, and the first topology subgraph represents the connection relationship of the device nodes; Based on the first topological subgraph, the alarm information, and the performance information, graph feature information is obtained, and the graph feature information is determined by fusing topological features, alarm features, and performance features. The graph feature information is input into the fault identification model to obtain fault identification result information; Based on the fault identification results, the faulty link is selected as the target link. Based on the target link and a preset time window, obtain the time-series subgraph information associated with the target link, the time-series subgraph information including the current and historical first topology subgraphs; Based on the temporal subgraph information, dynamic graph node encoding information is obtained, which is determined by the fusion of topological features and temporal features; The dynamic graph node encoding information is input into the fault location model to obtain fault location result information.

2. The method for fault identification and location in a packet transmission network according to claim 1, characterized in that, The step of obtaining the first topology subgraph corresponding to the device nodes in the packet transmission network based on the topology connection information includes: Based on the topology connection information, a global device connection topology map of the packet transmission network is constructed, and the global device connection topology map represents the topology of the packet transmission network; Based on the global device connection topology and the preset number of device influence layers, the first topology subgraph corresponding to each device node is obtained.

3. The method for fault identification and location in a packet transmission network according to claim 1, characterized in that, The step of obtaining graph feature information based on the first topological subgraph, the alarm information, and the performance information includes: Based on the first topology subgraph and the alarm information, obtain alarm coding matrix information; based on the first topology subgraph and the performance information, obtain performance coding matrix information. Based on the alarm coding matrix information and the performance coding matrix information, obtain the graph coding matrix information; Based on the first topological subgraph, an adjacency matrix and a degree matrix are determined. The adjacency matrix represents the connection relationship between devices, and the degree matrix represents the number of device connections. The graph encoding matrix information, the adjacency matrix, and the degree matrix form the graph feature information; The alarm coding matrix information represents the number of alarms occurring for each device node in the first topology subgraph, and the performance coding matrix information represents the average performance of each device node in the first topology subgraph.

4. The method for fault identification and location in a packet transmission network according to claim 1, characterized in that, The step of obtaining dynamic graph node encoding information based on the temporal subgraph information includes: Based on the time-series subgraph information, the corresponding time-series diffusion matrix information is obtained, wherein the time-series diffusion matrix information represents the connectivity between current and historical device nodes; Based on the time-series diffusion matrix information, determine the target device node information with the highest connectivity to the target link; Based on the time-series diffusion matrix information, the target device node information, and the time-series subgraph information, obtain the diffusion node connectivity level matrix, the diffusion distance encoding matrix, and the time encoding matrix; Based on the performance information, the target device node information, and the time encoding matrix, obtain the attribute encoding matrix; The dynamic graph node encoding information is obtained based on the diffusion node connectivity level matrix, the diffusion distance encoding matrix, the time encoding matrix, and the attribute encoding matrix.

5. The method for fault identification and location in a packet transmission network according to claim 4, characterized in that, The step of obtaining the corresponding time series diffusion matrix information based on the time series subgraph information includes: Based on the temporal subgraph information, determine the current adjacency matrix and current degree matrix corresponding to the current first topological subgraph, and determine the historical adjacency matrix and historical degree matrix corresponding to the historical first topological subgraph. Based on the current adjacency matrix and the current degree matrix, obtain the current diffusion matrix; based on the historical adjacency matrix and the historical degree matrix, obtain the historical diffusion matrix. The time-series diffusion matrix information is formed based on the current diffusion matrix and the historical diffusion matrix.

6. The method for fault identification in a packet transmission network according to claim 1, characterized in that, After obtaining the fault location result information, the following is also included: Based on the fault location results, the root cause information of the fault is determined; Based on the root cause information of the fault, retrieve the corresponding recommended solution information from the experience knowledge base; Based on the root cause information of the fault and the recommended solution information, a fault report is generated.

7. The method for fault identification and location in a packet transmission network according to claim 1, characterized in that, The fault identification model is obtained through the following methods: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the second topology subgraph corresponding to the work order fault node; Based on the second topology subgraph and the historical alarm information and historical performance information corresponding to the work order fault node, obtain the fault feature coding information; Based on the work order information, determine the first training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal feature encoding information is obtained, and the fault feature encoding information and the normal feature encoding information form a first training dataset. Based on the first training dataset and the first training label, the graph convolutional neural network model is trained until training is complete, and the graph convolutional neural network model is used as the fault identification model.

8. The method for fault identification and location in a packet transmission network according to claim 1, characterized in that, The fault location model is obtained through the following methods: Obtain work order information, historical alarm information, historical performance information, and historical topology connection information; Based on the work order information and the historical topology connection information, obtain the fault sequence subgraph information corresponding to the fault node of the work order; Based on the fault timing subgraph information, obtain the fault dynamic graph node encoding information; Based on the work order information, the root cause of the fault is determined as the second training label; Based on the historical alarm information, the historical performance information, and the historical topology connection information, normal dynamic graph node encoding information is obtained, and the fault dynamic graph node encoding information and the normal dynamic graph node encoding information form a second training dataset. The graph attention network model is trained based on the second training dataset and the second training label until training is complete, and the graph attention network model is used as the fault localization model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a packet transmission network fault identification and location method as described in any one of claims 1 to 8.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a packet transmission network fault identification and location method as described in any one of claims 1 to 8.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a packet transmission network fault identification and location method as described in any one of claims 1 to 8.

Images