Optical network fault location method, device and related equipment

Abstract

The present disclosure provides an optical network fault location method, apparatus, and related equipment, relating to the field of optical transport network communication technology. The method comprises: obtaining at least one alarm data of the optical network; extracting triple information of each alarm data, wherein the triple information contains at least two entities and the relationship between the two entities; constructing a corresponding alarm knowledge graph based on the triple information of each alarm data; inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and outputting the fault probability corresponding to each entity in the alarm knowledge graph; and determining the fault location in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph. The present disclosure can improve the accuracy and efficiency of optical network fault location.

Description

Technical Field

[0001] The present disclosure relates to the field of edge computing technology, and in particular to an optical network fault location method, apparatus, and related equipment. Background Art

[0002] With the rapid development of the communications industry, the Optical Transport Network (OTN) has become the backbone of modern communications networks, carrying large amounts of data and service signals. A single network failure (for example, a node or link failure) can lead to a decline in Quality of Service (QoS) or even service interruption. Therefore, once a network failure occurs, network operators need to accurately locate the source of the failure and repair it as quickly as possible. OTN alarms are closely related to failures. OTN alarms are indicators of abnormal conditions or failures occurring in the network and are typically generated by the alarm mechanisms of network devices. When a failure occurs in an OTN network, such as optical signal loss, signal quality degradation, or device hardware failure, the network management system receives a series of alarms reported by multiple devices. Even if these alarms contain location information, it is difficult to identify the root cause of the alarm storm, making fault location difficult. OTN alarms are a direct manifestation of the failure. By analyzing and processing alarm information, faults can be located and resolved, ensuring the stable operation of the OTN communication network. Existing OTN fault location methods mainly include the following:

[0003] 1) Signal flow analysis: Signal flow analysis is a commonly used OTN fault location method. This method first collects information about the OTN network topology, device configuration, and service signal flow to determine the flow direction within the OTN network. It then examines the status of each device, port, and link along the signal flow. By comparing and analyzing the status information at each point, combined with alarm information and performance data, the fault is located. This method has the advantage of intuitively displaying the signal transmission path, helping engineers quickly find the fault point. However, for large and complex OTN networks, signal flow analysis can become quite tedious, requiring considerable time and effort.

[0004] 2) OTN fault location method based on expert experience and rule matching: The rule-matching-based OTN fault location method first collects alarm information and device status data from the OTN network; then, it filters and matches the collected data according to preset rules and thresholds; finally, it determines the possible location of the fault based on expert experience. However, due to the complexity and dynamic nature of OTN networks, fixed rules and thresholds are difficult to cover all possible fault scenarios, resulting in low location accuracy and the possibility of false positives and missed positives.

[0005] 3) Machine Learning-Based OTN Fault Location Method: This method uses machine learning algorithms to learn from a large amount of historical fault data and establish a classification model to classify and locate new fault data. This method improves location accuracy to a certain extent, but still faces some challenges. Machine learning models are sensitive to the distribution and characteristics of data and fail to discover the correlation between historical fault data. The model input data is single, and when the network structure or service traffic changes, the model's low generalization leads to low fault location accuracy.

[0006] Any failure can lead to large-scale service interruptions and network congestion, causing huge economic losses to operators and users. Although existing OTN fault location methods have improved location accuracy to a certain extent, some problems still exist. For example, these methods often ignore the causal relationship between alarm events, which affects the accuracy of fault location in complex network environments.

[0007] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0008] The present disclosure provides an optical network fault location method, apparatus, and related equipment, which at least to a certain extent overcome the problem of low optical network fault location accuracy in related technologies.

[0009] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0010] According to one aspect of the present disclosure, a method for locating optical network faults is provided, comprising: obtaining at least one alarm data of an optical network; extracting triplet information of each alarm data, wherein the triplet information includes at least two entities and a relationship between the two entities; constructing a corresponding alarm knowledge graph based on the triplet information of each alarm data; inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and outputting the fault occurrence probability corresponding to each entity in the alarm knowledge graph; and determining the fault location in the optical network based on the fault occurrence probability corresponding to each entity in the alarm knowledge graph.

[0011] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, the entity type in the alarm data includes at least one of the following: alarm type, fault object, service, and equipment, and extracting the triplet information of each alarm data includes: setting the alarm type, the fault object, the service, and the equipment in each alarm data as entities in the triplet information; setting the relationship between the alarm type, the fault object, the service, and the equipment in each alarm data as the relationship between entities in the triplet information; and determining the triplet information of each alarm data based on the entities and the relationship between the entities.

[0012] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, a corresponding alarm knowledge graph is constructed according to the triple information of each alarm data, including: setting the two entities contained in the triple information of each alarm data as two nodes of the alarm knowledge graph to be constructed; setting the relationship between the two entities contained in the triple information of each alarm data as an edge of the alarm knowledge graph to be constructed; and constructing the corresponding alarm knowledge graph according to the two nodes and edges set in the triple information of each alarm data.

[0013] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, the alarm knowledge graph of each alarm data is input into a pre-trained gated graph neural network model, and the fault occurrence probability corresponding to each entity in the alarm knowledge graph is output, including: inputting a pre-constructed alarm knowledge graph into a pre-trained gated graph neural network model, and outputting the probability of pending fault occurrence corresponding to each entity in the alarm knowledge graph; using an attention mechanism to process the probability of pending fault occurrence corresponding to each entity in the alarm knowledge graph, and determining the probability of failure occurrence corresponding to each node in the alarm knowledge graph.

[0014] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, before inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model and outputting the fault occurrence probability corresponding to each entity in the alarm knowledge graph, the method also includes: using an attention mechanism to decompose the alarm knowledge graph input into the initialized gated graph neural network model into a causal attention graph and a trivial attention graph; constructing a target loss function based on the causal attention graph and the trivial attention graph; and optimizing the initialized gated graph neural network model through the target loss function to obtain a trained gated graph neural network model.

[0015] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, the triplet information also includes: entity attribute information, determining the fault location in the optical network according to the fault probability corresponding to each entity in the alarm knowledge graph, including: when the fault probability corresponding to any node in the alarm knowledge graph is greater than or / equal to a preset threshold, determining the node as the root fault node; determining the fault location in the optical network according to the entity attribute information corresponding to the root fault node.

[0016] In some exemplary embodiments of the present disclosure, based on the aforementioned scheme, after obtaining at least one alarm data of the optical network, the method further includes: performing standardization processing on the obtained at least one alarm data of the optical network, wherein the standardization processing includes at least one of the following: cleaning processing, integration processing, and normalization processing.

[0017] According to another aspect of the present disclosure, an optical network fault locating device is also provided, including: an alarm data acquisition module for acquiring at least one alarm data of the optical network; a triple information determination module for extracting triple information of each alarm data, wherein the triple information contains at least two entities and the relationship between the two entities; an alarm knowledge graph construction module for constructing a corresponding alarm knowledge graph based on the triple information of each alarm data; a fault probability output module for inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and outputting the fault probability corresponding to each entity in the alarm knowledge graph; a fault location determination module for determining the fault location in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph.

[0018] According to another aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform any of the above-mentioned optical network fault location methods by executing the executable instructions.

[0019] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, any one of the above-mentioned optical network fault location methods is implemented.

[0020] According to another aspect of the present disclosure, a computer program product is provided, including: a computer program or instructions, which implement any one of the above optical network fault location methods when executed by a processor.

[0021] An optical network fault location method, apparatus, and related equipment provided in the embodiments of the present disclosure combine an alarm knowledge graph with a graph neural network. An alarm knowledge graph is constructed by acquiring alarm data in the optical network, and the relationships between entities in the alarm data are correlated. The deep learning capability of the graph neural network is utilized to determine the probability of failure of each entity in the alarm knowledge graph. This method can not only determine the occurrence of potential faults, but also accurately locate the location of the fault based on the size of the fault probability corresponding to each entity.

[0022] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0024] Figure 1 A schematic diagram illustrating an exemplary application system architecture of an optical network fault location method according to an embodiment of the present disclosure is shown;

[0025] Figure 2 A schematic diagram of a method for locating an optical network fault according to an embodiment of the present disclosure is shown;

[0026] Figure 3 A schematic diagram of a process for processing event alarms using a gated graph neural network model according to an embodiment of the present disclosure is shown;

[0027] Figure 4 A schematic diagram of an attention mechanism according to an embodiment of the present disclosure is shown;

[0028] Figure 5 A schematic diagram of an OTN service carrying relationship in an embodiment of the present disclosure is shown;

[0029] Figure 6 A schematic diagram of an optical network fault location workflow according to an embodiment of the present disclosure is shown;

[0030] Figure 7 A schematic diagram of alarm association representation in an embodiment of the present disclosure is shown;

[0031] Figure 8 A schematic diagram of a method for locating an optical network fault according to an embodiment of the present disclosure is shown;

[0032] Figure 9 A schematic diagram of the entire process of an optical network fault location method according to an embodiment of the present disclosure is shown;

[0033] Figure 10 A schematic diagram of an optical network fault locating device according to an embodiment of the present disclosure is shown;

[0034] Figure 11 A schematic diagram of an electronic device to which an optical network fault location method is applied in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0035] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0036] In addition, the described features, structures or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, well-known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present disclosure.

[0037] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0038] For ease of understanding, several terms involved in this disclosure are explained below:

[0039] Optical Transport Network (OTN): A network used for high-speed digital communications that uses optical signals to transmit data. OTN provides higher bandwidth and lower latency and supports multiple protocols and services.

[0040] Knowledge Graph (KG): A semantic network used to represent entities, concepts, and the relationships between them. In OTN fault diagnosis, the KG is used to store and represent OTN network topology, device fault status, alarm information, and more.

[0041] Graph Neural Network (GNN): A neural network model used to process graph-structured data. In OTN fault diagnosis, GNNs are used to learn the relationships between nodes and edges and predict possible faults based on the network's real-time status and graph data.

[0042] Causal Attention Mechanism (CAM): A special attention mechanism that emphasizes the causal relationship between nodes or events in sequence data or graph-structured data. In OTN fault location, the CAM can capture the causal chain between alarm events, helping the system to more accurately understand and predict fault propagation paths, thereby improving the efficiency and accuracy of fault location.

[0043] Figure 1 FIG. 1 shows an exemplary application system architecture diagram to which the optical network fault location method according to the embodiment of the present disclosure can be applied. Figure 1 As shown, the system architecture may include a terminal device 101 , a network 102 and a server 103 .

[0044] The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103 , and can be a wired network or a wireless network.

[0045] Optionally, the above-mentioned wireless network or wired network uses standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or any combination of a virtual private network). In some embodiments, technologies and / or formats including Hyper Text Mark-up Language (HTML), Extensible Markup Language (XML), etc. are used to represent data exchanged over the network. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0046] The terminal device 101 can be various electronic devices, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, wearable devices, augmented reality devices, virtual reality devices, etc.

[0047] Optionally, the client of the application installed in different terminal devices 101 is the same, or the client of the same type of application based on different operating systems. Based on the different terminal platforms, the specific form of the client of the application can also be different, for example, the application client can be a mobile phone client, a PC client, etc.

[0048] The server 103 may be a server that provides various services, such as a background management server that provides support for devices operated by users using the terminal device 301. The background management server may analyze and process received request data and feed back the processing results to the terminal device.

[0049] Optionally, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The terminal can be a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited to these. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application.

[0050] Those skilled in the art will know that Figure 1 The number of terminal devices, networks, and servers in the embodiment is merely illustrative, and any number of terminal devices, networks, and servers may be provided based on actual needs. This embodiment of the present disclosure does not limit this.

[0051] Under the above system architecture, an embodiment of the present disclosure provides an optical network fault location method, which can be executed by any electronic device with computing and processing capabilities.

[0052] In some embodiments, the optical network fault locating method provided in the embodiments of the present disclosure can be executed by the terminal device of the above-mentioned system architecture; in other embodiments, the optical network fault locating method provided in the embodiments of the present disclosure can be executed by the server in the above-mentioned system architecture; in other embodiments, the optical network fault locating method provided in the embodiments of the present disclosure can be implemented by the terminal device and the server in the above-mentioned system architecture through interaction.

[0053] First, in response to the above-mentioned problems, the embodiments of the present disclosure provide a method for locating faults in optical networks that can be applied but is not limited to OTN. It can also be applied to optical communication networks in other fields. By acquiring alarm data in the optical network to construct an alarm knowledge graph, the entities in the alarm data are correlated with each other, and the deep learning capabilities of the graph neural network are used to determine the probability of failure of each entity in the alarm knowledge graph. This method can not only determine the occurrence of potential faults, but also accurately locate the location of the fault based on the size of the fault probability corresponding to each entity.

[0054] Figure 2 A schematic diagram of a method for locating an optical network fault in an embodiment of the present disclosure is shown, and the method includes the following steps:

[0055] S202: Obtain at least one alarm data of the optical network.

[0056] It should be noted that the alarm data in the embodiment of the present disclosure is a vital information source obtained by the optical network in the event of a fault. The alarm data can be obtained from the OTN network management system, alarm log files, alarm databases, alarm association tables summarized by experts, and manufacturer product alarm instructions. For example, by calling the alarm interface of the OTN network management system, real-time alarm data is obtained. The alarm data includes: alarm information and device information, wherein the alarm information includes: alarm name, alarm ID, alarm level, alarm device, alarm layer rate, alarm belonging Business, etc., and equipment information includes: device name, device location, device status, and the name of the circuit carried by the device; the detailed information of the alarm event, such as alarm type, alarm level, alarm location, etc., can be extracted by parsing the alarm log file; the alarm ID, alarm type, alarm level, alarm frequency, alarm reason, etc. can also be collected by querying the alarm database; the alarm association table can also be summarized by experts to record alarm rules, device associations, business associations, and attribute associations, etc.; the manufacturer's product alarm manual can also be used to query the recorded alarm name, alarm business association, etc.

[0057] S204 , extracting triplet information of each alarm data, wherein the triplet information includes at least two entities and a relationship between the two entities.

[0058] It should be noted that the entity type in each alarm data in the embodiment of the present disclosure includes at least one of the following: alarm type, fault object, service, and equipment. The alarm type, fault object, service, and equipment in each alarm data are set as entities in the triplet information, and the relationship between the alarm type, fault object, service, and equipment in each alarm data is set as the relationship between entities in the triplet information. Based on the relationship between entities, the triplet information of each alarm data is determined and extracted. Specifically, the entity relationship in the embodiment of the present disclosure can be divided into the relationship between the alarm type and the fault object, the relationship between different alarm types, the relationship between the alarm type and the service, the relationship between the alarm type and the fault object, the relationship between different services, the relationship between different devices, and the relationship between the device and the alarm type. In more detail, the relationship between different alarm types and the relationship between alarm types and services in the embodiment of the present disclosure are derived from the alarm association table summarized by experts and the manufacturer's product alarm manual; the relationship between the alarm type and the fault object and different devices can be queried through the OTN network management system, alarm log file, and alarm database.

[0059] S206: Construct a corresponding alarm knowledge graph based on the triple information of each alarm data.

[0060] It should be noted that the triples in the embodiment of the present disclosure can reflect the relationship between events, and the alarm knowledge graph is represented and stored in the form of a graph, in which the nodes represent entities and the edges represent the relationship between entities, which can intuitively show the association and mutual influence between alarm events. Moreover, the alarm knowledge graph in the embodiment of the present disclosure will be continuously updated and optimized as new alarm data is generated and old data is updated, for example, new alarm data can be added, outdated alarm data can be deleted, and possible errors can be corrected.

[0061] S208: Input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and output the fault occurrence probability corresponding to each entity in the alarm knowledge graph.

[0062] It should be noted that the gated graph neural network model in the embodiment of the present disclosure is a model obtained by pre-training various artificial intelligence algorithm models (for example, neural network models) or their combined models through machine learning. The model can automatically identify the alarm knowledge graph of each alarm data and match it to obtain the fault probability corresponding to each entity in each alarm knowledge graph. The input data of the model is the alarm knowledge graph corresponding to each alarm data, and the output data is the fault probability corresponding to each entity in the knowledge graph. Among them, the relationship between alarms and faults in OTN, that is, most alarms are caused by failures of entities or relationships between entities, and the gated graph neural network is a special graph neural network that can capture structural information in the graph and the dependencies between nodes. Therefore, the gated graph neural network model in the embodiment of the present disclosure can be specifically obtained by training the gated graph neural network. The gated graph neural network can make full use of the information in the input alarm knowledge graph, learn the topology and characteristics of the network, and is very suitable for capturing alarms in the alarm knowledge graph, thereby locating faults.

[0063] S210, determining the location of the fault in the optical network according to the fault occurrence probability corresponding to each entity in the alarm knowledge graph.

[0064] It should be noted that the output result of the gated graph neural network model in the embodiment of the present disclosure can be a probability distribution, which represents the possibility of failure of each entity in the alarm knowledge graph. Based on this probability distribution, the node most likely to fail is determined, and further troubleshooting and processing are performed.

[0065] The optical network fault location method provided in the embodiments of the present disclosure first obtains at least one alarm data of the optical network; secondly, extracts triple information of each alarm data, wherein the triple information contains at least two entities and the relationship between the two entities; then, based on the triple information of each alarm data, constructs a corresponding alarm knowledge graph; thereafter, inputs the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and outputs the fault probability corresponding to each entity in the alarm knowledge graph; finally, determines the fault location in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph. Compared with the related art, when a fault occurs in the optical network, the network management system will receive a series of alarms reported by multiple devices. Even if these alarms contain location information, it is difficult to identify the root alarm from the alarm storm, so fault location becomes difficult. The embodiment of the present disclosure combines the alarm knowledge graph with the graph neural network. By obtaining the alarm data in the optical network to construct an alarm knowledge graph, the entities in the alarm data are correlated with each other, and the deep learning ability of the graph neural network is used to determine the probability of failure of each entity in the alarm knowledge graph. Not only can the occurrence of potential faults be determined, but also the location of the fault can be accurately located according to the size of the fault probability corresponding to each entity.

[0066] In some embodiments, the alarm data in the embodiments of the present disclosure may include:

[0067] 1) Alarm name: The name of each alarm, such as board dislocation, optical transmission unit signal loss alarm, client device management unit pointer loss, etc.

[0068] 2) Alarm ID: A number or code that uniquely identifies each alarm event, used to distinguish different alarm events.

[0069] 3) Alarm device: describes the nature of the alarm event, such as network element, board, port, protection group, device protection group subnet connection, etc.

[0070] 4) Alarm level: The level is divided according to the severity of the alarm event, such as severe, mild, general, etc.

[0071] 5) Alarm layer rate: The rate of the layer to which the alarm occurs. For example, the layer rate of LR_OCH_Data_Unit_1 is the ODU service layer rate; the layer rate of LR_OCH_Transport_Unit_1 is the OUT service layer rate.

[0072] 6) Alarm service: The name or identifier of the device where the alarm event occurs, such as the Optical Transport Network Unit (OTU), Optical Data Unit (ODU), Optical Multiplex Section layer (OMS), etc.

[0073] 7) Alarm location: The specific location in the OTN network topology, such as a node or link.

[0074] 8) Alarm description: Use words or codes to describe the specific content and phenomena of the alarm event.

[0075] 9) Alarm cause: Conduct preliminary analysis and judgment on the possible causes of the alarm event.

[0076] 10) Related alarms: other alarm events associated with the current alarm event. Attributes include root alarm code, sub-alarm code, root alarm name, sub-alarm name, root alarm signal level, and sub-alarm signal level.

[0077] 11) Alarm frequency: Statistics on the number and frequency of alarm events within a period of time.

[0078] 12) Circuit name: The name of the circuit where the alarm, fault, service, or equipment is located.

[0079] 13) Device Name: The specific name of the device, including the model and manufacturer of the device. For example, the network element model: OptiX OSN 1800 is a network element model of the OptiX OSN series of a certain mobile phone manufacturer.

[0080] 14) Equipment location: The specific installation location or deployment location of the equipment in the physical space. This location information is crucial for fault diagnosis and positioning, and helps to quickly determine the area and scope of the alarm or fault.

[0081] 15) Equipment status: The current operating status of the equipment, including normal, fault, alarm, and maintenance status.

[0082] 16) Device-carried circuit name: The device-carried circuit name refers to the identifier of the specific circuit or service channel carried or connected by each device in the OTN.

[0083] 17) Circuit ID: A code that uniquely identifies the circuit name. Each circuit has a unique code ID.

[0084] 18) Alarm analysis rules: Rules for analyzing alarm events, including signal flow analysis, correlation analysis between signal flow alarms and signal protection switching alarms, correlation analysis between hardware alarms and signal flow, correlation analysis between power supply alarms and signal flow, and correlation analysis between system alarms and signal flow.

[0085] 19) Alarm rules: There are two types: transmission and derivative.

[0086] 20) Alarm ID association: Records the unique identifier (ID) association between different alarm events. Through the alarm ID, you can track and locate specific alarm events and establish logical relationships between them.

[0087] 21) Time Correlation: Records the time sequence and relationship between alarm events. By analyzing the occurrence time of alarm events, the sequence of alarm events can be identified, and the propagation path and impact range of the fault can be inferred.

[0088] 22) Device Association: This records the relationships between devices involved in an alarm event. This includes the connections, dependencies, and interactions between devices. Device association allows analysis of the fault's propagation path and potential impact across different devices.

[0089] 23) Service Correlation: This records the impact of alarm events on the services carried by the OTN. This includes the correspondence between alarm events and specific services, as well as alarm indicators indicating service interruption or performance degradation. Service correlation allows for accurate assessment of the impact of alarm events on services, providing decision support for fault recovery.

[0090] In some embodiments, the entity types in the embodiments of the present disclosure include at least one of the following: alarm type, fault object, service, and equipment, wherein the alarm type is further divided into root alarm and sub-alarm; the fault object includes links and nodes; the OTN service entities include customer data, optical data units, optical transmission units, optical channels, optical multiplexing layers, and optical transmission layers; the equipment entities include: management entities, port entities, rack entities, slot entities, board entities, etc.

[0091] In some embodiments, each entity has its own attributes. The alarm type entity includes attributes such as alarm name, alarm ID, alarm object type, alarm level, alarm occurrence time, alarm layer rate, alarm code, alarm device, alarm location, alarm description, and alarm cause; the fault object entity includes attributes such as device name, fault name, fault occurrence time, and circuit name; the service entity includes circuit name, OTN service type, bearer relationship, and device type; and the device entity includes device name, device location, device status, and the name of the circuit carried by the device. The relationship between entities is the core of the alarm knowledge graph. In the alarm knowledge graph, it is necessary to define and extract the relationship between alarm and fault object, alarm and alarm, alarm and service, alarm and fault object, service and service, device and device, and device and alarm. Specifically, the relationship between entities includes:

[0092] 1) Alarm Types: When a device fails, the first alarm triggered is the root alarm. The root alarm can derive or transmit other sub-alarms. Sub-alarms and root alarms have a deriving and transmitting relationship.

[0093] 2) Alarm type and fault object: The "caused by..." relationship between the alarm and the faulty device. When a device fails, the alarm mechanism is triggered and the alarm description and cause are recorded.

[0094] 3) Device-to-device: There is a "connection" relationship or "belongs to" relationship between two devices. For example, network element A and network element B are connected through a link, and port A belongs to board A.

[0095] 4) Service-to-service: There is a bearer-carried relationship. For example, in OTN services, OTUs carry ODUs, and ODUs are carried.

[0096] 5) Alarm type and service: The alarm "occurs" on a certain service, for example, the optical transport network unit signal loss alarm occurs on the OTU service.

[0097] 6) Device and alarm type: Changes in the device status may trigger an alarm, which will generate alarm information and notify the management system. The alarm can be associated with the device entity through the alarm device.

[0098] In some embodiments, the embodiments of the present disclosure construct a corresponding alarm knowledge graph based on the triple information of each alarm data, including: setting the two entities contained in the triple information of each alarm data as two nodes of the alarm knowledge graph to be constructed; setting the relationship between the two entities contained in the triple information of each alarm data as the edge of the alarm knowledge graph to be constructed; and constructing the corresponding alarm knowledge graph based on the two nodes and edges set in the triple information of each alarm data. Specifically, the nodes in the alarm knowledge graph in the embodiments of the present disclosure represent entities, and the edges represent the relationships between these entities. The alarm knowledge graph can be used to uniformly process the acquired alarm data to form a structured knowledge representation system, which provides rich data support for fault location, thereby achieving more accurate positioning of the fault location.

[0099] In some embodiments, the embodiments of the present disclosure input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and output the probability of fault occurrence corresponding to each entity in the alarm knowledge graph, including: inputting the pre-built alarm knowledge graph into the pre-trained gated graph neural network model, and outputting the probability of failure occurrence to be processed corresponding to each entity in the alarm knowledge graph; using the attention mechanism to process the probability of failure occurrence to be processed corresponding to each entity in the alarm knowledge graph, and determining the probability of failure occurrence corresponding to each node in the alarm knowledge graph. Specifically, the alarm knowledge graph can uniformly model various alarm data sources in the optical network to form a structured knowledge representation system, thereby achieving more accurate positioning of the fault location; using the gated graph neural network model to learn the temporal dependency of the relationship between entities in the alarm knowledge graph, greatly improving the real-time performance of fault location. In communication networks, time management is crucial for service recovery. This advantage in the embodiments of the present disclosure can significantly reduce the communication interruption time caused by faults.

[0100] In some embodiments, as Figure 3 As shown, the embodiment of the present disclosure initializes the input alarm knowledge graph through the gated graph neural network model. Specifically, the alarm knowledge graph is used as the input of the gated graph neural network model. Formally, the input graph model is defined as G = {V, A}, where V is a node set and A is an adjacency matrix representing the graph structure. For each node v∈V, there is a hidden state h at time step t. v (t), when t = 0, each node (i.e., alarm event) has an initial state vector, and then updates its state through iteration, and initializes the hidden state of entity v in the alarm knowledge graph according to the alarm knowledge graph in, Represents the features of each entity node v in the alarm knowledge graph at time T.

[0101] In some embodiments, the disclosed embodiments update the node status through message passing and gating mechanisms between nodes. Each node updates its own representation based on the information of neighboring nodes in each iteration, thereby capturing the complex associations between alarm events. In each iteration of the gated graph neural network model, the node status is updated based on the status of its neighboring nodes and the relationship between them. This update process is usually implemented by formula (1):

[0102]

[0103] Among them, a v (t) represents the state of node v in the alarm knowledge graph after iteration; represents a two-column matrix corresponding to node v; Indicates the state of node 1 at time t-1; represents the state of node |v| at time t-1; b represents the bias term.

[0104] In some embodiments, b in formula (1) of the present disclosure is generally a vector used to add a fixed offset when the state is updated.

[0105] In some embodiments, before inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model and outputting the fault occurrence probability corresponding to each entity in the alarm knowledge graph, the optical network fault location method in the embodiment of the present disclosure also includes: using an attention mechanism to decompose the alarm knowledge graph input into the initialized gated graph neural network model into a causal attention graph and a trivial attention graph; constructing a target loss function based on the causal attention graph and the trivial attention graph; optimizing the initialized gated graph neural network model through the target loss function to obtain a trained gated graph neural network model. Specifically, the embodiment of the present disclosure introduces an attention mechanism to divide the alarm knowledge graph into a causal attention graph and a trivial attention graph, and constructs an optical network fault location method, which can effectively avoid the influence of confounding factors in the reasoning graph on the reasoning results, accurately capture the causal relationship between alarms, and provide a new solution for optical network fault location.

[0106] In some embodiments, in order to avoid the non-causal features in the alarm knowledge graph from being used as shortcut features and having a negative impact on the prediction results, the present disclosure introduces an attention mechanism into the gated graph neural network model: by calculating the causal probability between alarms, the causal relationship between alarms is accurately captured, and the impact of confounding factors on the result prediction is reduced, such as Figure 4As shown, the embodiment of the present disclosure introduces an attention mechanism to decompose the alarm knowledge graph into a causal attention graph and a trivial attention graph. Fault classification is performed through the causal attention graph, while the trivial attention graph is regarded as a graph without causal feature information. The attention mechanism can learn the corresponding attention scores based on the feature information of the nodes and edges in the alarm knowledge graph. For example, for node v i And the edge (v i ,v j ), the embodiment of the present disclosure can calculate the node v by formula (2) and (3) i And the edge (v i ,v j )’s attention score:

[0107] αci,αti=σ(MLPnode(h i )) (2)

[0108] βci j, βti j=σ(MLPedge(h i ||h j )) (3)

[0109] Among them, αci is the attention score of node v_i in the causal attention graph; αti is the attention score of node v i The attention score in the trivial attention map; σ is the softmax function; MLPnode is a multi-layer perceptron; h i is node v i The state vector of βci j is the state vector of the causal attention graph connecting the nodes v i and v j The attention score of the edge of βti j is the edge of the trivial attention graph connecting node v i and v j The attention score of the edge; MLPedge is a multi-layer perceptron; h j is node v j The state vector of node v i and v j The state vector h i and h j Stitched together.

[0110] In more detail, αci is the i ) is calculated by the softmax function σ, and MLPnode accepts node v i The state vector h i As input, it produces a scalar output, which becomes a probability value after passing through the softmax function σ, indicating v iThe relative importance in the causal attention graph; βci j is given by MLPedge(h i ||h j ) is calculated by the softmax function σ, where || represents the concatenation operation, that is, the node v i and v j The state vector h i and h j Spliced ​​together as the input of MLPedge, similarly, βci j is also a probability value, indicating that the edge (v i ,v j ) in the causal attention graph; αti is determined by MLPnode(h i ) is calculated by the softmax function σ; βti j is obtained by MLPedge(h i ||h j ) is calculated by the softmax function σ.

[0111] By calculating the above attention scores, the embodiment of the present disclosure can divide the alarm knowledge graph into a causal attention graph and a trivial attention graph represented by formulas (4) and (5):

[0112] G c ={A⊙Ma,X⊙Mx} (4)

[0113] G t ={A⊙Ma,X⊙Mx} (5)

[0114] Among them, G c Represents the causal attention graph; A represents the adjacency matrix of the alarm knowledge graph; Ma represents the mask matrix, which is used to filter out edges related to causal relationships; ⊙ represents element-wise multiplication, multiplying the elements of corresponding positions of two matrices of the same shape; X represents the node feature matrix; Mx represents the mask matrix, which is used to filter out node features related to causal relationships; G t represents the trivial attention map.

[0115] In some embodiments, the causal attention graph in the embodiments of the present disclosure is composed of the result of element-wise multiplication of the original adjacency matrix A and the causal mask matrix Ma, and the result of element-wise multiplication of the node feature matrix X and the causal feature mask matrix Mx; the trivial attention graph is also composed of the result of element-wise multiplication of the original adjacency matrix A and the trivial mask matrix Ma, and the result of element-wise multiplication of the node feature matrix X and the trivial feature mask matrix Mx. In this way, two subgraphs can be constructed separately, one emphasizing the causal relationship and the other may emphasize the details outside the causal relationship. These subgraphs can be used for further analysis, such as anomaly detection, fault location and other tasks.

[0116] The graph representation is obtained through two Graph Neural Network (GNN) layers, and the aggregation (readout) function and classifier are used for prediction. The prediction results of the causal attention graph and the trivial attention graph are obtained through formulas (6) and (7):

[0117] hG c =freadout(GConvc(A⊙Ma,X⊙Mx)),zG c =Φ c (hG c ) (6)

[0118] hG t =freadout(GConvc(A⊙Ma,X⊙Mx)),zG t =Φ t (hG t ) (7)

[0119] Among them, hG c Represents the causal attention graph G c The final representation vector of the causal attention graph; freadout represents the aggregation of the node representations of the entire causal attention graph into a global graph representation; GConvc represents the graph convolution layer, which is used to update the state of each node in the graph; zG c Represents the classifier Φ c The representation vector hG of the causal attention graph c The result of the prediction; Φ c represents the classifier; hG t Represents the trivial attention graph G t The final representation vector of zG t Represents the classifier Φ t The representation vector hG of the trivial attention map t The result of the prediction; Φ t Represents a classifier.

[0120] The causal attention graph is used to extract information for classification. According to the prediction results of the causal attention graph, the first loss function L is constructed by formula (8): SUP Specifically, the embodiment of the present disclosure may use the cross entropy loss function:

[0121]

[0122] Among them, |D| represents the number of graphs in the dataset; G represents a single graph; D represents the dataset; y G Represents the true label vector of graph G.

[0123] In some embodiments, the disclosed embodiments calculate the predicted probability zG of the causal attention graph c and the true label y G The cross entropy loss between is designed to allow the model to better classify on the causal attention graph.

[0124] For trivial attention graphs, KL divergence is used as the second loss function. KL divergence is used to measure the similarity of two distributions. The output of KL divergence is closer to the uniform distribution. The second loss function L is obtained by formula (9). unif :

[0125] L unif =1∣D∣∑G∈DKL(yunif,zG t ) (9)

[0126] Where KL represents the Kullback-Leibler divergence, which is used to measure the difference between two probability distributions; yunif represents a uniformly distributed vector.

[0127] In some embodiments, the disclosed embodiments measure the predicted probability zG of the trivial attention graph by KL divergence. t The difference between the uniform distribution yunif is intended to make the prediction distribution of trivial attention maps more uniform, thereby reducing the impact of trivial information on the model.

[0128] The causal attention graph and the trivial attention graph are randomly matched to form an intervention graph. According to the intervention graph, the third loss function L is determined by formulas (10) and (11): caus :

[0129] zG′=Φ(hG c +hG t ) (9)

[0130]

[0131] Where zG′ represents the new prediction probability vector obtained by adding the representation vectors of the causal attention graph and the randomly matched trivial attention graph, and then passing it through the classifier Φ; |T| represents the number of graphs in the set of trivial attention graphs; T represents the set of trivial attention graphs; t′ represents the trivial attention graph.

[0132] In some embodiments, the disclosed embodiments combine the representation vectors of the causal attention graph and the randomly matched trivial attention graph to obtain a new prediction probability zG′, and calculate the new prediction probability zG′ and the true label y G The cross entropy loss between is designed to evaluate the performance of the model when causal information is mixed with trivial information, thereby strengthening the model's understanding of causal relationships.

[0133] Add the first loss function, the second loss function, and the third loss function, and use formula (11) to get the target loss function L:

[0134] L=L SUP +λ1L unif +λ2L caus (11)

[0135] Among them, λ1 represents the hyperparameter of the disentanglement intervention strength; λ2 represents the hyperparameter of the causal intervention strength.

[0136] In some embodiments, after a single fault triggers an alarm in the embodiment of the present disclosure, no matter how the alarm propagates in the optical network, each group of training alarms corresponds to a fault label, and the possibility of label distribution is obtained through a simple fully connected neural network. In this way, after the model outputs the alarm data corresponding to the fault object entity, the alarm entity connected to the fault object entity in the alarm knowledge graph is compared with the alarm data reported on the network to obtain the root alarm. The "alarm description" and "alarm cause" in the attribute fields of the root alarm entity record the detailed location and cause of the fault. The embodiment of the present disclosure uses the cross entropy loss function to optimize the model. When the model converges to within the threshold, it is considered that the gated graph neural network model has reached the optimal level.

[0137] In some embodiments, the triplet information in the disclosed embodiments also includes: entity attribute information, and determining the location of the fault in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph, including: determining that a node is a root fault node when the fault probability corresponding to any node in the alarm knowledge graph is greater than or / equal to a preset threshold; and determining the location of the fault in the optical network based on the entity attribute information corresponding to the root fault node. Specifically, the gated graph neural network model outputs the fault probability corresponding to each entity, thereby determining the root fault node, achieving rapid and accurate positioning of faults in the optical network, and improving the operation and maintenance efficiency and service quality of the communication network.

[0138] In some embodiments, when a new alarm event occurs in the optical network, the alarm interface of the optical network management system is called to obtain real-time alarm information, including: alarm type, fault object, equipment, network topology, etc., and an alarm knowledge graph is constructed with the alarm event as the entity, and the relationship matrix of the current node and edge is defined. The trained gated graph neural network model is used for reasoning. The gated graph neural network model can infer potential fault nodes related to the alarm event by iteratively updating the status of the node. These fault nodes may represent faulty equipment or links. Finally, the specific fault location is determined according to the output result of the gated graph neural network model. The output result can be a probability distribution, indicating the possibility of failure of each node. According to this probability distribution, the node most likely to fail is determined, and further troubleshooting and processing are performed.

[0139] In some embodiments, as the network environment and business needs change, the optical network will also change dynamically. The disclosed embodiment updates the alarm knowledge graph and graph neural network model in real time, and through statistical analysis of the causal attention subgraphs of multiple samples, it can update the expert knowledge base such as the "expert summary alarm association table", improve the effectiveness of the alarm knowledge graph construction, quickly adapt to changes, and ensure the accuracy and efficiency of fault location. Moreover, traditional optical network fault diagnosis often relies on manual experience or fixed rules, with low diagnostic efficiency and prone to errors. The disclosed embodiment automatically constructs an alarm knowledge graph based on the alarm data obtained in real time, and automatically trains the constructed alarm knowledge graph as the input of the gated graph neural network model to determine the location of the fault. This not only improves the speed of diagnosis, but also reduces errors caused by human factors.

[0140] In some embodiments, after obtaining at least one alarm data of the optical network, the optical network fault location method in the embodiment of the present disclosure further includes: performing standardization processing on the obtained at least one alarm data of the optical network, wherein the standardization processing includes at least one of the following: cleaning processing, integration processing, and normalization processing. Specifically, in the embodiment of the present disclosure, the cleaning processing of the alarm data is to remove duplicates based on the alarm unique identification ID; for missing alarm information, it is possible to choose to delete relevant records or fill in with default values ​​according to actual conditions, and identify and filter obviously invalid or erroneous alarm data according to alarm rules; the integration processing of the alarm data is to merge multiple related alarms into one record to reduce redundant information, associate the alarm data with other data sources such as device information, network topology, and business data, and enrich the alarm context information; the normalization processing of the alarm data is to ensure that all alarm data follows a unified format and naming convention to facilitate subsequent processing and analysis, and certain fields in the alarm data can be standardized, for example, converting the timestamp into a unified format, or converting the alarm level into numerical data.

[0141] In some embodiments, after constructing the alarm knowledge graph and the gated graph neural network model, the present disclosure can further achieve sustainable improvement in model accuracy through the following steps:

[0142] 1) Gated Graph Neural Networks Combined with Operation and Maintenance Expert Experience: During the fault reasoning stage, the gated graph neural network is combined with the experience of operation and maintenance experts. For example, the expert summary of the IN_PWR_LOW low-light alarm -> optical transmission unit signal loss is added to the causal reasoning graph. For example, in an already trained causal reasoning graph, if the model has not learned the above expert rule experience, the expert rule experience is added to the model and its conditional probability is set to a high level (such as 0.9). This implementation can better avoid problems such as insufficient generalization in the model prediction process and further improve the reasoning accuracy.

[0143] 2) Construct an expert-summarized alarm association table: Through statistical analysis of the causal attention subgraphs of multiple samples, an expert knowledge base such as the "Causal Association Statistical Analysis Table" can be constructed. Through statistical analysis, the causal relationships in the alarm knowledge graph with a higher probability of occurrence are automatically sorted out. When the number of samples reaches a certain scale, the causal relationships summarized in the "Causal Association Statistical Analysis Table" are confirmed by experts, and some causal relationships can be marked as expert experience, thereby improving the effectiveness of the alarm knowledge graph construction and achieving the effect of the model becoming more accurate with use.

[0144] In some embodiments, the embodiments of the present disclosure can also visualize the diagnostic results on the user interface, specifically to certain fault points or fault links, and provide operation and maintenance personnel with detailed fault information and solution suggestions to facilitate fault handling and network optimization operations.

[0145] In some embodiments, the embodiments of the present disclosure are combined with a certain operator's existing network OTN fault and alarm data to conduct technical research. The OTN six-layer service bearing model is specifically as follows: Figure 5As shown in the figure, the optical layer consists of the optical transmission section layer (OTS), the optical multiplexing layer and the optical channel (OCH). The optical channel spans a group of equipment boards, which span one or more optical multiplexing layers. An optical multiplexing layer contains multiple optical transmission layers in series. Due to the hierarchical relationship between the OTS, OMS and OCH, any fault affecting the bottom layer will be propagated to the upper layer. OTN provides rich monitoring signals in the optical layer overhead. When the single-board sensor detects a fault, a strict alarm mechanism is introduced on each single board. For example, the optical fiber interface unit board will identify the upstream optical fiber cut fault and report the alarm to the network management system. Another example is that when any abnormality related to the receiving optical path is sensed, the network management system will receive an alarm reported by the optical transport network unit board.

[0146] In some embodiments, as Figure 6 As shown, when a unidirectional link of a certain province's OTN circuit fails, the optical network fault location method in the embodiment of the present disclosure is used. First, the optical network alarm data is obtained and preprocessed to construct an alarm knowledge graph; the gated graph neural network model is trained and optimized using the alarm knowledge graph; the trained gated graph neural network model is integrated into the operator's operation system; a series of alarm knowledge graphs related to the failure of a certain link in the backbone network circuit of the certain province are constructed and sent to the gated graph neural network model of the system for fault location. Finally, the location of the link fault point is visualized on the interface, and detailed fault information and solution suggestions are given. The specific embodiment process is as follows:

[0147] 1) Use the operator's operating system to call the northbound interface of the manufacturer's network management system to obtain alarm data (alarm data records the alarm name, alarm ID, alarm object type, alarm level, alarm location, alarm occurrence time, alarm code, alarm level, alarm device, alarm location, alarm description, alarm reason, etc.), collect alarm log files (record the alarms and reasons, alarm types, alarm levels, and alarm locations generated by each network topology layer device under the OTN service) and alarm association tables summarized by operator industry experts (derivation and transmission rule tables between alarms, device associations, service associations, and attribute associations, some of which are listed in the table). Figure 7 ), obtain alarm-related data from the alarm description documents of each manufacturer's products (alarm name, alarm business association), and pre-process the alarm data, including data cleaning, data integration, and data normalization.

[0148] 2) Define and extract entities from processed alarm data: the four types of entities are alarm type, fault object, service, and equipment, among which the alarm type is further divided into root alarm and sub-alarm; fault objects include links and nodes; OTN service entities include customer data, optical data units, optical transmission units, optical channels, optical multiplexing layers, and optical transmission layers; equipment entities include: management entities, port entities, rack entities, slot entities, board entities, etc.; ② Relationship definition and extraction: each entity has its own attributes. The alarm entity includes attributes such as alarm name, alarm ID, alarm object type, alarm level, alarm occurrence time, alarm layer rate, alarm code, alarm device, alarm location, alarm description, and alarm cause; the fault object entity includes attributes such as device name, fault name, fault occurrence time, and circuit name; the service entity The entity includes circuit name, OTN service type, bearer relationship, and device type; the device entity includes device name, device location, device status, and the name of the circuit carried by the device. In the OTN alarm knowledge graph, the relationships between alarms and faulty devices, alarms and alarms, alarms and services, devices and devices, and services and services are defined and extracted. The relationships between alarms and services are derived from the alarm association table summarized by experts and the manufacturer's product alarm manuals. The relationships between alarms and faulty devices, and between devices are derived from the OTN network management system, alarm log files, and alarm database. A graph database (Neo4j) is selected to store and query the alarm knowledge graph. As new alarm data is generated and old data is updated, the alarm knowledge graph is continuously updated and optimized.

[0149] 3) Use a deep learning framework (such as PyTorch) to build a gated graph neural network model and initialize the hidden state h of entity v in the alarm knowledge graph. v (0), after a single fault triggers an alarm, no matter how the alarm is propagated in the OTN, each set of training alarms corresponds to a fault label. The possibility of label distribution is obtained through a simple fully connected neural network. After the model outputs the alarm data corresponding to the fault entity, the alarm entity connected to the fault entity in the alarm knowledge graph is compared with the alarm data reported by the network to obtain the root alarm. The cross entropy loss function is used to optimize the model, and the loss function threshold is set. When the model converges within the threshold, it is considered to be optimal.

[0150] 4) Integrate the trained gated graph neural network model into the operator's operation system to realize the automatic construction of the alarm knowledge graph, the automatic training of the graph neural network model, and the automatic execution of OTN fault location, such as Figure 8When a link in a provincial backbone network fails, the operator system calls the northbound alarm interface of the manufacturer's network management system to obtain real-time alarm information, including the alarm fault device, alarm details and network topology. The link failure causes the device to report MUT_LOS. Since MUT_LOS derives and transmits a series of alarms such as R_LOS, R_LOF, AIS, OTUk_TIM, OTUk_LOFLOM, SSF, ODU_RDI, LINK_ERR along the service and signal flow ( Figure 8 Only some alarm examples are given, and the actual network management collects more alarms). An alarm knowledge graph is constructed based on the alarm derivation transmission relationship, alarm product manual and business association relationship, and the relationship matrix between the current node and the edge is defined. The trained gated graph neural network model is used for reasoning. The gated graph neural network model can infer the potential fault nodes related to the alarm event by iteratively updating the status of the node. After locating the root alarm, the alarm fault range will be shortened according to the alarm description phenomenon, and the diagnostic phenomenon and results will be reported at the front end. Finally, the link (SNC) between the network element 7 port 1 of a certain A computer room and the network element 5 port 3 of a certain B computer room is located. The fault is located at a certain position. The embodiment of the present disclosure can visualize the diagnostic results on the user interface and provide solution suggestions, specifically Figure 8 The SNC fault location is shown in the figure. OTN operation and maintenance personnel perform fault handling and network optimization operations based on the detailed fault information and solution suggestions. Through the embodiments of the present disclosure, faults in the OTN network can be quickly and accurately located, improving the operation and maintenance efficiency and service quality of the communication network. At the same time, the embodiments of the present disclosure also have broad application prospects and can be extended to other types of communication networks for fault location.

[0151] In some embodiments, the disclosed embodiments can integrate the gated graph neural network model into a unified system to realize the automatic construction of the alarm knowledge graph, the automatic training of the graph neural network model, and the automated execution of fault location. By simulating different fault scenarios and alarm data, a complete optical network fault location system is tested, designed and implemented. The system can collect and process alarm information in the optical network in real time, automatically construct an alarm knowledge graph, and use the gated graph neural network model for automatic fault location.

[0152] In some embodiments, as Figure 9 As shown, the embodiments of the present disclosure solve the problems of low accuracy, lack of dynamic adaptability, and limited fault location capabilities of traditional optical network fault location methods by introducing alarm knowledge graphs and graph neural network technology, and provide a new and effective method for optical network fault location, which specifically includes the following steps:

[0153] S902, obtaining alarm data in the optical network and performing standardized processing on the alarm data. The alarm data can be obtained through the OTN network management system, alarm log files, alarm database, alarm correlation table summarized by experts, and manufacturer's product alarm manual.

[0154] S904: Construct and store an alarm knowledge graph based on the acquired alarm data. Entity types in the alarm data include at least one of the following: alarm type, fault object, service, and device. By defining and extracting entities from the alarm data and defining and extracting relationships between entities, an alarm knowledge graph is constructed based on the extracted entities and relationships between them.

[0155] S906: Design a gated graph neural network model and train it to obtain a trained gated graph neural network model. This model incorporates an attention mechanism, uses the constructed alarm knowledge graph as input, and outputs the fault probability corresponding to each entity in the alarm knowledge graph.

[0156] S908 integrates the trained gated graph neural network model into a unified system to achieve automatic construction of the alarm knowledge graph, automatic training of the graph neural network model, and automated execution of fault location.

[0157] In some embodiments, the embodiments disclosed herein combine the two technologies of alarm knowledge graph and graph neural network for the first time and apply them to optical network fault diagnosis. This innovative combination makes fault location more accurate and efficient. Unlike traditional fault diagnosis methods, the optical network fault location method in the embodiments disclosed herein can continuously optimize the diagnostic capabilities of the model, realize automated and intelligent diagnosis of optical network faults, adapt to the development needs of optical networks, and can also be extended to optical communication networks in other fields, and has broad application prospects.

[0158] Based on the same inventive concept, the present disclosure also provides an optical network fault location device, such as the following embodiment. Since the principle of solving the problem in the device embodiment is similar to that in the above method embodiment, the implementation of the device embodiment can refer to the implementation of the above method embodiment, and the repeated parts will not be repeated.

[0159] Figure 10 A schematic diagram of an optical network fault location device according to an embodiment of the present disclosure is shown, wherein the device includes:

[0160] The alarm data acquisition module 1001 is used to obtain at least one alarm data of the optical network;

[0161] A triplet information determination module 1002 is configured to extract triplet information of each alarm data, wherein the triplet information includes at least two entities and a relationship between the two entities;

[0162] The alarm knowledge graph construction module 1003 is used to construct a corresponding alarm knowledge graph based on the triple information of each alarm data;

[0163] The fault probability output module 1004 is used to input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model and output the fault probability corresponding to each entity in the alarm knowledge graph;

[0164] The fault location determination module 1005 is used to determine the fault location in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph.

[0165] An optical network fault locating device provided in an embodiment of the present disclosure obtains at least one alarm data of an optical network through an alarm data acquisition module; extracts triple information of each alarm data through a triple information determination module, wherein the triple information contains at least two entities and the relationship between the two entities; constructs a corresponding alarm knowledge graph based on the triple information of each alarm data through an alarm knowledge graph construction module; inputs the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model through a fault probability output module, and outputs the fault probability corresponding to each entity in the alarm knowledge graph; and determines the fault location in the optical network through a fault location determination module based on the fault probability corresponding to each entity in the alarm knowledge graph. Compared with the related art, when a fault occurs in the optical network, the network management system will receive a series of alarms reported by multiple devices. Even if these alarms contain location information, it is difficult to identify the root alarm from the alarm storm, so fault location becomes difficult. The embodiment of the present disclosure combines the alarm knowledge graph with the graph neural network. By obtaining the alarm data in the optical network to construct an alarm knowledge graph, the entities in the alarm data are correlated with each other, and the deep learning ability of the graph neural network is used to determine the probability of failure of each entity in the alarm knowledge graph. Not only can the occurrence of potential faults be determined, but also the location of the fault can be accurately located according to the size of the fault probability corresponding to each entity.

[0166] In some embodiments, the entity type in the alarm data in the embodiment of the present disclosure includes at least one of the following: alarm type, fault object, service, and equipment. The triple information determination module in the embodiment of the present disclosure is also used to set the alarm type, fault object, service, and equipment in each alarm data as entities in the triple information; set the relationship between the alarm type, fault object, service, and equipment in each alarm data as the relationship between entities in the triple information; and determine the triple information of each alarm data based on the entities and the relationship between entities.

[0167] In some embodiments, the alarm knowledge graph construction module in the embodiments of the present disclosure is also used to set the two entities contained in the triple information of each alarm data as two nodes of the alarm knowledge graph to be constructed; set the relationship between the two entities contained in the triple information of each alarm data as the edge of the alarm knowledge graph to be constructed; and construct the corresponding alarm knowledge graph according to the two nodes and edges set in the triple information of each alarm data.

[0168] In some embodiments, the fault probability output module in the embodiments of the present disclosure is also used to input a pre-built alarm knowledge graph into a pre-trained gated graph neural network model, and output the probability of occurrence of pending faults corresponding to each entity in the alarm knowledge graph; use an attention mechanism to process the probability of occurrence of pending faults corresponding to each entity in the alarm knowledge graph, and determine the probability of occurrence of faults corresponding to each node in the alarm knowledge graph.

[0169] In some embodiments, the optical network fault location device in the embodiments of the present disclosure further includes: an alarm knowledge graph decomposition module, which is used to input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and before outputting the fault occurrence probability corresponding to each entity in the alarm knowledge graph, use an attention mechanism to decompose the alarm knowledge graph input into the initialized gated graph neural network model into a causal attention graph and a trivial attention graph; a target loss function construction module, which is used to construct a target loss function based on the causal attention graph and the trivial attention graph; and a model optimization module, which is used to optimize the initialized gated graph neural network model through the target loss function to obtain a trained gated graph neural network model.

[0170] In some embodiments, the triplet information in the embodiment of the present disclosure also includes: entity attribute information. The fault location determination module in the embodiment of the present disclosure is also used to determine that a node is a root fault node when the probability of a fault corresponding to any node in the alarm knowledge graph is greater than or / equal to a preset threshold; and determine the location of the fault in the optical network based on the entity attribute information corresponding to the root fault node.

[0171] In some embodiments, the optical network fault locating device in the embodiments of the present disclosure further includes: a data processing module for performing standardized processing on the at least one alarm data of the target optical network after obtaining the at least one alarm data of the target optical network, wherein the standardized processing includes at least one of the following: cleaning processing, integration processing, and normalization processing.

[0172] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods, or program products. Therefore, various aspects of the present disclosure may be implemented in the following forms: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits," "modules," or "systems."

[0173] Based on the same inventive concept, an embodiment of the present disclosure further provides an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any of the aforementioned optical network fault location methods by executing the executable instructions. Because the principles for solving the problems in this electronic device embodiment are similar to those in the aforementioned method embodiment, the implementation of this electronic device embodiment can refer to the implementation of the aforementioned method embodiment, and any repetitive details will not be repeated.

[0174] Refer to the following Figure 11 1100 according to this embodiment of the present disclosure will be described. Figure 11 The electronic device 1100 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0175] like Figure 11 As shown, electronic device 1100 is implemented as a general-purpose computing device. Components of electronic device 1100 may include, but are not limited to, the aforementioned at least one processing unit 1101, the aforementioned at least one storage unit 1102, and a bus 1103 connecting various system components (including storage unit 1102 and processing unit 1101).

[0176] The storage unit stores program codes, which can be executed by the processing unit 1101, so that the processing unit 1101 executes the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification.

[0177] In some embodiments, when the electronic device is used to control, for example, the optical network fault location method disclosed above, the processing unit 1101 may perform the following steps of the above method embodiment:

[0178] Acquire at least one alarm data of the optical network; extract triple information of each alarm data, wherein the triple information contains at least two entities and the relationship between the two entities; construct a corresponding alarm knowledge graph based on the triple information of each alarm data; input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and output the fault probability corresponding to each entity in the alarm knowledge graph; determine the fault location in the optical network based on the fault probability corresponding to each entity in the alarm knowledge graph.

[0179] The storage unit 1102 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 11021 and / or a cache memory unit 11022 , and may further include a read-only memory unit (ROM) 11023 .

[0180] The storage unit 1102 may also include a program / utility 11024 having a set (at least one) of program modules 11025, such program modules 11025 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0181] The bus 1103 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0182] The electronic device 1100 can also communicate with one or more external devices 1104 (e.g., a keyboard, a pointing device, a Bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 1100, and / or any device that enables the electronic device 1100 to communicate with one or more other computing devices (e.g., a router, a modem, etc.). Such communication can occur via an input / output (I / O) interface 1105. Furthermore, the electronic device 1100 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 1106. As shown, the network adapter 1106 communicates with other modules of the electronic device 1100 via a bus 1103. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 1100, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0183] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0184] Based on the same inventive concept, embodiments of the present disclosure also provide a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements any of the aforementioned optical network fault location methods. Because the principles underlying the problem solved by this computer-readable storage medium embodiment are similar to those of the aforementioned method embodiment, the implementation of this computer-readable storage medium embodiment can be referenced to the implementation of the aforementioned method embodiment, and any repetitions will be omitted.

[0185] More specific examples of computer-readable storage media in the present disclosure may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0186] In the present disclosure, a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0187] Alternatively, the program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0188] In a specific implementation, the program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, and the like, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0189] Based on the same inventive concept, embodiments of the present disclosure also provide a computer program product, including a computer program or instructions. When executed by a processor, the computer program or instructions implement the optical network fault locating method described in any of the aforementioned method embodiments. Because the principles for solving the problems described in this computer program product embodiment are similar to those described in the aforementioned method embodiments, the implementation of this computer program product embodiment can be referenced to the implementation of the aforementioned method embodiments, and any repetitive details will not be repeated.

[0190] It should be noted that although several modules or units of the device for action execution are mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0191] Furthermore, although the steps of the method of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in this particular order, or that all steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0192] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0193] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

Claims

1. A method for locating an optical network fault, characterized in that: include: Obtain at least one alarm data of the optical network; Extracting triplet information of each alarm data, wherein the triplet information includes at least two entities and a relationship between the two entities; Based on the triple information of each alarm data, a corresponding alarm knowledge graph is constructed; Input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and output the fault occurrence probability corresponding to each entity in the alarm knowledge graph; Determining a fault location in the optical network based on a fault probability corresponding to each entity in the alarm knowledge graph; Before inputting the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model and outputting the fault occurrence probability corresponding to each entity in the alarm knowledge graph, the method also includes: using an attention mechanism to decompose the alarm knowledge graph input into the initialized gated graph neural network model into a causal attention graph and a trivial attention graph; constructing a target loss function based on the causal attention graph and the trivial attention graph; and optimizing the initialized gated graph neural network model through the target loss function to obtain a trained gated graph neural network model.

2. The optical network fault location method according to claim 1, characterized in that: The entity type in the alarm data includes at least one of the following: alarm type, fault object, service, and equipment. Extracting triplet information of each alarm data includes: Setting the alarm type, the fault object, the service, and the device in each alarm data as entities in triplet information; Setting the relationship between the alarm type, the fault object, the service, and the device in each alarm data as the relationship between entities in the triplet information; According to the entities and the relationships between the entities, triplet information of each alarm data is determined.

3. The optical network fault location method according to claim 1, characterized in that: Based on the triple information of each alarm data, the corresponding alarm knowledge graph is constructed, including: The two entities contained in the triple information of each alarm data are respectively set as two nodes of the alarm knowledge graph to be constructed; Setting the relationship between two entities contained in the triple information of each alarm data as an edge of the alarm knowledge graph to be constructed; According to the two nodes and edges set in the triple information of each alarm data, the corresponding alarm knowledge graph is constructed.

4. The optical network fault location method according to claim 1, characterized in that: Input the alarm knowledge graph of each alarm data into a pre-trained gated graph neural network model, and output the fault probability corresponding to each entity in the alarm knowledge graph, including: Input the pre-built alarm knowledge graph into the pre-trained gated graph neural network model, and output the probability of occurrence of the pending fault corresponding to each entity in the alarm knowledge graph; An attention mechanism is used to process the probability of occurrence of pending faults corresponding to each entity in the alarm knowledge graph, and the probability of occurrence of faults corresponding to each node in the alarm knowledge graph is determined.

5. The optical network fault location method according to claim 3, characterized in that: The triplet information further includes: entity attribute information, and determining the fault location in the optical network according to the fault probability corresponding to each entity in the alarm knowledge graph, including: When the probability of a fault occurring corresponding to any node in the alarm knowledge graph is greater than or equal to a preset threshold, the node is determined to be a root fault node; The fault location in the optical network is determined according to the entity attribute information corresponding to the root fault node.

6. The optical network fault location method according to claim 1, characterized in that: After obtaining at least one alarm data of the optical network, the method further includes: Performing standardization processing on the acquired at least one alarm data of the optical network, wherein the standardization processing includes at least one of the following: cleaning processing, integration processing, and normalization processing.

7. An optical network fault location device, characterized in that: include: The optical network fault locating device is used to execute the optical network fault locating method according to any one of claims 1 to 6.

8. An electronic device, characterized in that: include: processor; as well as a memory for storing executable instructions of the processor; The processor is configured to execute the optical network fault location method according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the optical network fault locating method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising: A computer program or instruction, characterized in that when the computer program or instruction is executed by a processor, it implements the optical network fault location method according to any one of claims 1 to 6.

Images