Fault diagnosis method, device, equipment, medium and product

Abstract

The invention provides a fault diagnosis method and device, equipment, a medium and a product, and belongs to the technical field of communication network fault detection, the method comprises the following steps: receiving multi-source real-time information of each functional layer in a communication network, the multi-source real-time information comprising a dynamic network topology relation and fault information of at least one functional layer; fusing the multi-source real-time information into a multi-modal feature vector; the multi-modal feature vectors are input into a large language model, a fault diagnosis report output by the large language model is obtained, and the fault diagnosis report comprises fault root cause nodes; wherein the large language model is a model obtained by performing reinforcement learning according to corpora in the communication field. Multi-source information in the communication network is fused, cross-layer fault reasoning is carried out by means of the large language model obtained based on corpus training in the communication field, fault root cause points and fault diagnosis reports with practical guiding significance are obtained, and efficient, high-precision and real-time fault diagnosis services are provided for the large-scale communication network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication network fault detection, and in particular to a fault diagnosis method, device, equipment, medium and product. BACKGROUND

[0002] With the rapid development of communication technology, the scale of communication networks is continuously expanding, the structure is becoming increasingly complex, and the types and frequency of network failures are also increasing. In a large-scale software-defined network (SDN) environment, fault diagnosis technology is a crucial component.

[0003] In current communication systems, fault detection needs to cope with complex network topologies and highly variable data traffic, and virtualization and network slicing increase the likelihood of faults. Traditional fault diagnosis methods rely on a single data source (such as hardware indicators or alarm events) for diagnosis, resulting in fragmented diagnosis information. For current large-scale, high-dimensional communication network data, it is necessary to consider multi-dimensional, multi-source information to obtain fault causes, and the fragmented diagnosis information obtained according to the prior art is not conducive to quickly handling faults and has poor practicality. SUMMARY

[0004] The present application provides a fault diagnosis method, device, equipment, medium and product, aiming to solve the problem that traditional fault diagnosis methods rely on a single data source for diagnosis, resulting in fragmented diagnosis information, which cannot be applied to current communication systems to achieve high-quality fault diagnosis.

[0005] In a first aspect, the present application provides a fault diagnosis method, comprising: receiving multi-source real-time information of each functional layer in a communication network, the multi-source real-time information including dynamic network topology relationships and fault information of at least one functional layer; fusing the multi-source real-time information into a multi-modal feature vector; inputting the multi-modal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, the fault diagnosis report including a fault root cause node; wherein the large language model is a model obtained by enhanced learning according to communication field corpus.

[0006] As one embodiment, the multi-source real-time information is fused into a multi-modal feature vector, specifically comprising: preprocessing the multi-source real-time information; weighting and splicing the preprocessed multi-source real-time information to obtain a multi-modal feature vector.

[0007] As one embodiment, the multi-source real-time information further includes performance indicators for network monitoring. The fault information includes fault alarm information and log text of the network device.

[0008] As an embodiment, the multi-source real-time information is preprocessed, specifically including: parsing the log text to generate a semantic vector; constructing a dynamic context graph structure using the dynamic network topology relationship; based on the context graph structure, aggregating the performance indicators of the neighbor nodes to form a topology-based key matrix; based on the context graph structure, aggregating the fault alarm information of the neighbor nodes to obtain a topology-based fault vector.

[0009] As an embodiment, the preprocessed multi-source real-time information is weighted and spliced to obtain a multi-modal feature vector, specifically including: based on the cross-attention mechanism, the speech vector, the topology-based key matrix, and the topology-based fault vector are weighted and spliced, and the formula is as follows: wherein, is a multi-modal feature vector, is a semantic vector, is a topology-based key matrix, is a topology-based fault vector, is a dimension scaling factor, is the transpose of the key matrix.

[0010] As an embodiment, the fault diagnosis method further includes: if the fault diagnosis report indicates the occurrence of a new type of fault, updating the dynamic knowledge base with the new type of fault; updating the large language model using the updated dynamic knowledge base.

[0011] In a second aspect, the present application also provides a fault diagnosis device, comprising a receiving module, a fusion module, and a diagnosis module; The receiving module is used to receive multi-source real-time information of each functional layer in the communication network, and the multi-source real-time information includes dynamic network topology relationship and fault information of at least one functional layer; The fusion module is used to fuse the multi-source real-time information into a multi-modal feature vector; The diagnosis module is used to input the multi-modal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, and the fault diagnosis report includes a fault root cause node; The large language model is a model obtained by enhanced learning according to a communication field corpus.

[0012] In a third aspect, the present application provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements any of the above fault diagnosis methods when executing the computer program.

[0013] In a fourth aspect, the present application provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program implements any of the above fault diagnosis methods when executed by a processor.

[0014] In a fifth aspect, the present application provides a computer program product, which comprises a computer program, and the computer program implements any of the above fault diagnosis methods when executed by a processor. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0016] Figure 1 is one of the flowcharts of the fault diagnosis method provided by the present application; Figure 2 is another flowchart of the fault diagnosis method provided by the present application; Figure 3 is one of the flowcharts of the pre-processing of multi-source real-time information provided by the present application; Figure 4 is one of the structural diagrams of the fault diagnosis device provided by the present application; Figure 5 is the structural diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present application more clear, the technical solutions of the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0018] It should be noted that in the description of the present application, the terms "comprising", "containing" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus including the element.

[0019] The terms "first", "second", and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are usually a class, not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" means at least one of the connected objects, and the character " / ", generally indicates that the objects before and after are in a "or" relationship.

[0020] The following will be described in conjunction with Figures 1 to 5 The fault diagnosis method, device, equipment, medium and product provided by the present application are described.

[0021] It should be noted that the fault diagnosis method provided by the embodiments of the present application is realized based on the fault diagnosis device, the fault diagnosis method fuses multi-source information from each functional layer of the communication network, realizes cross-layer fault reasoning, and realizes efficient and real-time fault diagnosis in large-scale network.

[0022] The embodiments of the present application take the fault diagnosis device as an example to describe the fault diagnosis method.

[0023] Figure 1 is one of the flowcharts of the fault diagnosis method provided by the present application. Figure 2 is the second flowchart of the fault diagnosis method provided by the present application.

[0024] As Figure 1 shown, the fault diagnosis method provided by the present application comprises: S110: receiving multi-source real-time information of each functional layer in the communication network.

[0025] S120: fusing the multi-source real-time information into a multi-modal feature vector.

[0026] It can be understood that the fusion of multi-source real-time information can be realized by splicing, addition, multiplication, attention mechanism, pyramid pooling, deconvolution, encoder-decoder and the like.

[0027] S130: inputting the multi-modal feature vector into a large language model (LLM) to obtain a fault diagnosis report output by the LLM, the fault diagnosis report including a fault root cause node.

[0028] The large language model is a deep learning model trained using a large amount of text data, which can generate natural language text or understand the meaning of language text. The large language model can provide in-depth knowledge about a topic and language production. In this application, the large language model is a model obtained by reinforcement learning based on communication field corpus (such as communication technology documents, communication equipment manuals, historical fault cases, etc.) to enhance the knowledge understanding of the communication field of the large language model and provide a basis for fault diagnosis. The input data of the large language model is a multi-modal feature vector, and the output data is a fault diagnosis report.

[0029] Specifically, please refer to Figure 2 In the process of running the communication system, the related devices, software, etc. of each functional layer (including physical layer, protocol layer, service layer, etc.) in the communication network may fail, therefore, in the embodiments of the present application, the fault diagnosis device receives real-time information obtained by each position or node of the communication network, including dynamic network topology relationship, basic information of each functional layer (such as configuration, performance index of device or software, etc.) and fault information (such as faults of link, device, controller, etc.) generated by each functional layer in the communication process, etc. to form multi-source real-time information. In the communication process, the same fault cause may cause one or more functional layers to fail, therefore, the multi-source real-time information includes fault information of at least one functional layer.

[0030] After receiving the multi-source real-time information, the fault diagnosis device fuses the multi-source real-time information into a multi-modal feature vector to comprehensively reflect the comprehensive situation in the network communication process.

[0031] Subsequently, the fault diagnosis apparatus inputs the multi-modal feature vector into the large language model. The large language model determines the root cause node of the fault based on the structural characteristics of the network topology and the multi-modal feature vector to infer the transmission path between and within each functional layer, thereby realizing cross-layer reasoning. Specifically, the physical layer is formed by network device connections. If a network device fails, the failure may affect the normal access of the business layer, causing a business layer failure. If the fault diagnosis apparatus receives the above-mentioned failure information of two functional layers, it can determine that the root cause is the failure of the network device in the physical layer rather than the failure of the software in the business layer by reasoning based on the network structure. As an example, packet loss in the physical layer may cause service interruption in the business layer. After the multi-source real-time information is fused and input into the large language model after the business interruption failure in the business layer, the large language model is used for reasoning, and the root cause node obtained is in the physical layer, and the root cause is packet loss in the physical layer, thereby realizing cross-layer reasoning.

[0032] The large language model also uses natural language to describe the fault analysis principle based on the root cause node, provides fault repair suggestions, forms a fault diagnosis report, and outputs it to the user, so that the user can quickly understand the cause of the fault and obtain a solution to the fault.

[0033] The embodiments of the present application fuse multi-source information in a communication network and use a large language model trained based on a communication field corpus to perform cross-layer fault reasoning, obtain a fault root cause point and a fault diagnosis report with practical guiding significance, and improve the cross-layer fault diagnosis accuracy to provide efficient, high-precision, and real-time fault diagnosis services for large-scale communication networks.

[0034] In one possible implementation, in step S120, the multi-source real-time information is fused into a multi-modal feature vector, specifically including: S1201: Preprocessing the multi-source real-time information.

[0035] Specifically, the preprocessing can include data cleaning, data integration, data transformation, etc.

[0036] The data cleaning includes removing noise, duplicate data, and error data in the data, for example, correcting or deleting abnormal format records in device logs, and performing deduplication processing on duplicate alarm information, etc.

[0037] The data transformation includes extracting key events (such as “interface packet loss rate > 5%”) in log text using regular expressions, etc.

[0038] S1202: Weighted splicing the preprocessed multi-source real-time information to obtain a multi-modal feature vector.

[0039] In the embodiments of the present application, the pre-processed multi-source real-time information is weighted and spliced to realize fusion, which can make the large language model pay more attention to the importance (i.e., weight) of different information to fault diagnosis, and provide guidance for accurate fault diagnosis.

[0040] In a possible embodiment, the present application supports structured data, unstructured data, topology data and other multi-source real-time data, and supports SNMP (Simple Network Management Protocol), NetFlow (network monitoring function), Syslog (industrial standard protocol) and other protocols.

[0041] Specifically, the multi-source real-time information includes: Structured data: fault alarm information of network equipment (using SNMP, the message belongs to fault information), performance indicators of network monitoring (using NetFlow protocol, such as traffic indicators, device temperature curves, etc.), and the like.

[0042] Unstructured data: log text (using Syslog protocol, at least part of the log text provides fault information), operation and maintenance work order text, and the like.

[0043] Topology data: dynamic network topology relationship.

[0044] The embodiments of the present application collect structured, unstructured and topology data to form multi-modal and multi-source data related to communication networks, provide a comprehensive data foundation for large language models, and provide a guarantee for the accuracy of fault diagnosis.

[0045] Based on the above, in a possible embodiment, as shown in Figure 3 In step S1201, the multi-source real-time information is pre-processed, specifically including: S310: parsing the log text to generate a semantic vector.

[0046] In a possible embodiment, the semantic of the log text (such as "BGP (Border Gateway Protocol) neighbor oscillation") is parsed by using a large language model to generate a corresponding semantic vector.

[0047] S320: constructing a dynamic context graph structure by using the dynamic network topology relationship.

[0048] Specifically, network topology visualization is performed based on network topology relations (for example, using open source tools such as Gephi or Graphviz), network topology relations are encoded into a context graph structure, nodes and edge weights in the context graph structure are updated in real time according to dynamic changes of network topology relations (such as addition and deletion of devices, switching of links, 5G network slice reconstruction, etc.), and a dynamically changing context graph structure is formed.

[0049] S330: Based on the context graph structure, the performance indicators of the neighbor nodes are aggregated to form a topology-based key matrix.

[0050] In a possible embodiment, the performance indicators (such as packet loss rate, delay, etc.) of the neighbor nodes are aggregated based on GraphSAGE (Graph Sample and Aggregated, a graph neural network model based on inductive learning framework) to generate a 128-dimensional vector as the topology-based key matrix.

[0051] S340: Based on the context graph structure, the fault alarm information of the neighbor nodes is aggregated to obtain a topology-based fault vector.

[0052] In a possible embodiment, the fault alarm information of the neighbor nodes is aggregated based on GraphSAGE to generate a 128-dimensional vector as the topology-based fault vector.

[0053] The embodiments of the present application convert multi-source real-time information into vectors, which helps to improve the fusion quality of multi-source information.

[0054] In a possible embodiment, since the fault alarm information of the network device belongs to formatted data, data transformation is not required for fusion.

[0055] On the basis described above, in a possible embodiment, in step S1202, the preprocessed multi-source real-time information is weighted and spliced based on a cross-attention mechanism to obtain a multi-modal feature vector.

[0056] Specifically, the semantic vector, the topology-based key matrix, and the topology-based fault vector are weighted and spliced, and the formula is as follows: ; wherein, is the multi-modal feature vector, is the semantic vector, is the topology-based key matrix, is the transpose of the key matrix, is the topology-based fault vector, is a dimension scaling factor.

[0057] The embodiment of the application realizes end-to-end fusion and semantic association of multi-modal data of a communication network through a cross-attention mechanism. On the one hand, the cross-attention mechanism can help a large language model better understand and utilize information from different data sources. On the other hand, the cross-attention mechanism allows one sequence to focus on information in another sequence through an attention mechanism, which is conducive to processing matching between multi-modal data. In addition, by dynamically fusing features of different modalities, the cross-attention mechanism enables the large language model to have better adaptability and robustness when facing new and unseen data, thereby improving the generalization ability of the model.

[0058] The prior art adopts a rule-based method for fault diagnosis, which relies on a pre-set rule base to diagnose faults. However, when a new type of fault occurs in a communication network, the rule base is difficult to update quickly, resulting in low diagnostic efficiency. Based on such consideration, in one possible embodiment, the fault diagnosis method further comprises: If the fault diagnosis report indicates the occurrence of a new type of fault, the dynamic knowledge base is updated with the new type of fault; The large language model is updated with the dynamic knowledge base to update the rule base of the large language model.

[0059] Specifically, an online learning pipeline is deployed for the large language model. When a new type of fault case is detected, the model automatically extracts knowledge from the new type of fault case, updates the dynamic knowledge base, and automatically triggers the Low-Rank Adaptation of Large Language Models (Lora) to update the model parameters of the large language model, optimizes the diagnostic logic, and maintains the diagnostic accuracy.

[0060] The embodiment of the application builds a dynamic knowledge base and dynamically optimizes the parameters of the large language model using the dynamic knowledge base, so that the large language model quickly adapts to new fault modes through a lightweight incremental learning mechanism, reducing the model update cost.

[0061] Based on the above, the application further provides a fault diagnosis apparatus. The fault diagnosis apparatus can correspond to the fault diagnosis method described above.

[0062] As an example, as shown in Figure 4 The fault diagnosis apparatus includes a receiving module 410, a fusion module 420, and a diagnosis module 430.

[0063] The receiving module 410 is configured to receive multi-source real-time information of each functional layer in a communication network, the multi-source real-time information including dynamic network topology relationships and fault information of at least one functional layer.

[0064] The fusion module 420 is configured to fuse the multi-source real-time information into a multi-modal feature vector.

[0065] The diagnostic module 430 is configured to input the multi-modal feature vector into the large language model, obtain a fault diagnosis report output by the large language model, and the fault diagnosis report includes a fault root cause node.

[0066] The large language model is a model obtained by enhanced learning according to a communication field corpus.

[0067] The embodiments of the present application fuse multi-source information in a communication network, and perform cross-layer fault reasoning by means of a large language model trained based on a communication field corpus, to obtain a fault root cause point and a fault diagnosis report with actual guiding significance, thereby providing efficient, high-precision, and real-time fault diagnosis services for large-scale communication networks.

[0068] In a possible implementation manner, as shown in Figure 4 The fusion module 420 includes a preprocessing module 4201 and a splicing module 4202.

[0069] The preprocessing module 4201 is configured to preprocess the multi-source real-time information.

[0070] The splicing module 4202 is configured to splice the preprocessed multi-source real-time information by weighting, to obtain a multi-modal feature vector.

[0071] In the embodiments of the present application, the preprocessed multi-source real-time information is spliced by weighting to achieve fusion, which can enable the large language model to pay more attention to the importance of different information for fault diagnosis, thereby providing guidance for accurate fault diagnosis.

[0072] In a possible implementation manner, the multi-source real-time information further includes performance indicators for network monitoring; and the fault information includes fault alarm information and log texts of network devices.

[0073] The embodiments of the present application collect structured, unstructured, and topological data, form multi-modal and multi-source data related to a communication network, provide a comprehensive data basis for a large language model, and provide a guarantee for the accuracy of fault diagnosis.

[0074] In a possible implementation manner, the preprocessing module 4201 is specifically configured to: parse the log texts to generate semantic vectors; construct a dynamic context graph structure by using a dynamic network topology relationship; aggregate performance indicators of neighbor nodes based on the context graph structure, to form a topology-based key matrix; aggregate fault alarm information of neighbor nodes based on the context graph structure, to obtain a topology-based fault vector.

[0075] The embodiments of the present application convert the multi-source real-time information into vectors, which is helpful to improve the fusion quality of multi-source information.

[0076] In a possible implementation, the splicing module 4202 is specifically configured to: The semantic vector, the topology-based key matrix and the topology-based fault vector are weighted and spliced based on the cross-attention mechanism, and the formula is as follows: ; Wherein, is a multi-modal feature vector, is a semantic vector, is a topology-based key matrix, is a topology-based fault vector, is a dimension scaling factor, is a transpose of the key matrix.

[0077] The embodiment of the application realizes end-to-end fusion and semantic association of multi-modal data of a communication network through a cross-attention mechanism. On the one hand, the cross-attention mechanism can help a large language model better understand and utilize information from different data sources. On the other hand, the cross-attention mechanism allows one sequence to focus on information in another sequence through an attention mechanism, which is conducive to processing matching between multi-modal data. On the other hand, by dynamically fusing features of different modalities, the cross-attention mechanism enables the large language model to have better adaptability and robustness when facing new and unseen data, thereby improving the generalization ability of the model.

[0078] In a possible implementation, the fault diagnosis apparatus further comprises a knowledge base updating module and a model updating module. The knowledge base updating module is configured to update the dynamic knowledge base using a new type of fault when the fault diagnosis report indicates that the new type of fault occurs.

[0079] The model updating module is configured to update the large language model using the updated dynamic knowledge base.

[0080] The embodiment of the application builds a dynamic knowledge base and dynamically optimizes parameters of the large language model using the dynamic knowledge base, so that the large language model quickly adapts to a new type of fault mode through a lightweight incremental learning mechanism, thereby reducing the model updating cost.

[0081] Figure 5 is a structural schematic diagram of an electronic device provided by the application, such as Figure 5As shown, the electronic device can include a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 complete mutual communication through the communications bus 540. The processor 510 can invoke a logical instruction in the memory 530 to execute a fault diagnosis method, which includes: receiving multi-source real-time information of each function layer in a communication network, the multi-source real-time information including dynamic network topology relationships and fault information of at least one function layer; fusing the multi-source real-time information into a multi-modal feature vector; inputting the multi-modal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, the fault diagnosis report including a fault root cause node; and wherein the large language model is a model obtained through enhanced learning according to a communication field corpus.

[0082] In addition, the logical instruction in the memory 530 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0083] On the other hand, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, when the program instructions are executed by a computer, the computer can execute the fault diagnosis method provided by the above-mentioned embodiments, which includes: receiving multi-source real-time information of each function layer in a communication network, the multi-source real-time information including dynamic network topology relationships and fault information of at least one function layer; fusing the multi-source real-time information into a multi-modal feature vector; inputting the multi-modal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, the fault diagnosis report including a fault root cause node; and wherein the large language model is a model obtained through enhanced learning according to a communication field corpus.

[0084] In yet another aspect, the present application also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the fault diagnosis method provided by any of the above embodiments, and the method comprises: receiving multi-source real-time information of each functional layer in a communication network, the multi-source real-time information comprising dynamic network topology relationships and fault information of at least one functional layer; fusing the multi-source real-time information into a multi-modal feature vector; inputting the multi-modal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, the fault diagnosis report comprising a fault root cause node; and wherein the large language model is a model obtained through enhanced learning according to a communication field corpus.

[0085] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0086] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions essentially or in other words the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a Read-Only Memory (ROM) / Random Access Memory (RAM), a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0087] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions essentially or in other words the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a Read-Only Memory (ROM) / Random Access Memory (RAM), a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing examples, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A fault diagnosis method, characterized in that: include: Receiving multi-source real-time information of each functional layer in the communication network, wherein the multi-source real-time information includes a dynamic network topology relationship and fault information of at least one functional layer; fusing the multi-source real-time information into a multimodal feature vector; Inputting the multimodal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, wherein the fault diagnosis report includes a fault root cause node; The large language model is a model obtained by performing reinforcement learning based on corpus in the communication field.

2. The fault diagnosis method according to claim 1, characterized in that: The multi-source real-time information is fused into a multi-modal feature vector, specifically including: Preprocessing the multi-source real-time information; The pre-processed multi-source real-time information is weightedly spliced ​​to obtain the multimodal feature vector.

3. The fault diagnosis method according to claim 2, characterized in that: The multi-source real-time information also includes performance indicators for network monitoring; The fault information includes fault alarm information and log text of the network device.

4. The fault diagnosis method according to claim 3, characterized in that: Preprocessing the multi-source real-time information specifically includes: Parsing the log text to generate a semantic vector; Constructing a dynamic context graph structure using the dynamic network topology relationship; Based on the context graph structure, performance indicators of neighbor nodes are aggregated to form a topology-based bond matrix; Based on the context graph structure, fault alarm information of neighboring nodes is aggregated to obtain a topology-based fault vector.

5. The fault diagnosis method according to claim 4, characterized in that: The pre-processed multi-source real-time information is weightedly spliced ​​to obtain the multimodal feature vector, specifically including: The semantic vector, the topology-based key matrix, and the topology-based fault vector are weightedly concatenated based on the cross-attention mechanism, and the formula is as follows: ； in, is the multimodal feature vector, is the semantic vector, is the topology-based bond matrix, is the topology-based fault vector, is the dimension scaling factor, is the transpose of the key matrix.

6. The fault diagnosis method according to claim 1, characterized in that: The fault diagnosis method further includes: If the fault diagnosis report indicates a new type of fault, the dynamic knowledge base is updated using the new type of fault; Update the large language model using the updated dynamic knowledge base.

7. A fault diagnosis device, characterized in that: It includes receiving module, fusion module and diagnosis module; The receiving module is used to receive multi-source real-time information of each functional layer in the communication network, wherein the multi-source real-time information includes a dynamic network topology relationship and fault information of at least one functional layer; The fusion module is used to fuse the multi-source real-time information into a multimodal feature vector; The diagnosis module is used to input the multimodal feature vector into a large language model to obtain a fault diagnosis report output by the large language model, wherein the fault diagnosis report includes a fault root cause node; The large language model is a model obtained by performing reinforcement learning based on corpus in the communication field.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the fault diagnosis method according to any one of claims 1 to 6 is implemented.

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

10. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, the fault diagnosis method according to any one of claims 1 to 6 is implemented.

Images