Search

CN-121984847-A - Network switch intelligent operation and maintenance method and system applying AI and network switch

CN121984847ACN 121984847 ACN121984847 ACN 121984847ACN-121984847-A

Abstract

The invention discloses an intelligent operation and maintenance method and system for a network switch applying AI and the network switch, and belongs to the technical field of network switches. According to the invention, the AI model is configured for each network switch, so that the initial fault type of the corresponding network switch is actively identified, the rapid discovery of faults is realized relative to a deployment-center operation and maintenance network, the network switch corresponding to the faults is rapidly positioned after the AI model identifies the initial fault type, the operation and maintenance efficiency is improved, the final fault type is obtained through a plurality of fault data and the initial fault type, the fault judgment is performed by integrating a plurality of AI models, the sharing of the fault data is realized, the accuracy of the fault judgment is further improved, the operation and maintenance efficiency is further improved, the final fault type is identified through the AI model meeting the conditions, the corresponding operation and maintenance strategy is output, the accuracy of the fault diagnosis of the AI model is utilized, and the reliability of the operation and maintenance is further improved.

Inventors

  • WANG JIANJIAN

Assignees

  • 杭州程宇智能科技有限公司

Dates

Publication Date
20260505
Application Date
20260323

Claims (10)

  1. 1. An intelligent operation and maintenance method for a network switch applying AI, which is characterized by comprising the following steps: Configuring an AI model for each network switch, wherein the AI model actively identifies the initial fault type of the corresponding network switch; Setting a reference range corresponding to the network switch in real time, wherein the reference range comprises a plurality of network switches related to the network switch; acquiring fault data matched with the initial fault type from historical fault data of the plurality of network switches; forwarding the fault data and the initial fault type to an AI model meeting a condition; And the AI model identifies the final fault type according to the initial fault type and the fault data and outputs a corresponding operation and maintenance strategy.
  2. 2. The method of claim 1, wherein the final fault type includes hardware faults and non-hardware faults; the operation and maintenance strategy comprises the following steps: if the final fault type is hardware fault, starting the disaster recovery network switch, and screening maintainers and maintenance strategies which meet preset conditions; and if the final fault type is a non-hardware fault, starting the disaster recovery network switch, positioning a fault reason, and executing a fault source equipment operation parameter adjustment strategy corresponding to the fault reason.
  3. 3. The method of claim 2, wherein the step of determining the position of the substrate comprises, The AI model comprises a machine learning algorithm sub-model and a deep learning algorithm sub-model; the configuring the AI model for each network switch includes: presetting the machine learning algorithm sub-model, and training the machine learning algorithm sub-model through historical fault data of the network switch; Presetting the deep learning algorithm sub-model, and training the deep learning algorithm sub-model through the historical fault data of the plurality of network switches and the historical fault data of the network switches.
  4. 4. The method of claim 3, wherein the setting in real time a reference range corresponding to the network switch comprises: acquiring a plurality of first network switches which are in the same area as the network switch, and setting a first weight; acquiring a plurality of second network switches which transmit the same service data with the network switches, and setting second weights; acquiring a plurality of third network switches in a similar scene with the network switches, and setting third weights; And screening a plurality of fourth network switches which simultaneously meet the plurality of first network switches, the plurality of second network switches and the plurality of third network switches according to the corresponding first weights, second weights and third weights to obtain the plurality of network switches.
  5. 5. The method of claim 4, wherein the AI model actively identifying an initial failure type of the corresponding network switch comprises: identifying abnormal operation parameters of the network switch, first abnormal feedback parameters of the data source equipment and first abnormal feedback parameters of the data receiving equipment; and respectively inputting the abnormal operation parameters, the first abnormal feedback parameters and the second abnormal feedback parameters into the machine learning algorithm submodel to obtain the initial fault type.
  6. 6. The method of claim 5, wherein said obtaining fault data matching said initial fault type from historical fault data of said plurality of network switches comprises: Obtaining historical fault data which is the same as the initial fault type and meets error conditions with the abnormal operation parameters, the first abnormal feedback parameters and the second abnormal feedback parameters from the historical fault data of the plurality of network switches; The historical fault data at least comprises operation and maintenance time, operation and maintenance strategy, initial fault type, final fault type, abnormal operation parameters, first abnormal feedback parameters and second abnormal feedback parameters.
  7. 7. The method of claim 6, wherein the forwarding the fault data and the initial fault type to AI models that satisfy a condition comprises: screening the AI model meeting the condition according to the operation and maintenance time and the occurrence frequency of the fault data; and forwarding the fault data and the initial fault type to an AI model meeting a condition.
  8. 8. The method of claim 7, wherein the AI model identifying a final fault type from the initial fault type and the fault data comprises: Setting a confidence factor of each fault data; and inputting the initial fault type, the fault data and the confidence factors corresponding to the fault data into the machine learning algorithm submodel to obtain the final fault type.
  9. 9. An intelligent operation and maintenance system for a network switch applying AI, wherein the system comprises a plurality of network switches, each network switch is configured with an AI model, the AI model actively identifies an initial fault type of the corresponding network switch, and the network switch is used for: Setting a reference range corresponding to the network switch in real time, wherein the reference range comprises a plurality of network switches related to the network switch; acquiring fault data matched with the initial fault type from historical fault data of the plurality of network switches; forwarding the fault data and the initial fault type to an AI model meeting a condition; And the AI model identifies the final fault type according to the initial fault type and the fault data and outputs a corresponding operation and maintenance strategy.
  10. 10. A network switch, wherein the network switch configures at least an AI model comprising a machine learning algorithm sub-model and a deep learning algorithm sub-model, wherein: The machine learning algorithm sub-model is used for actively identifying the initial fault type of the corresponding network switch; the network switch further comprises a processing module for: Setting a reference range corresponding to the network switch in real time, wherein the reference range comprises a plurality of network switches related to the network switch; acquiring fault data matched with the initial fault type from historical fault data of the plurality of network switches; forwarding the fault data and the initial fault type to an AI model meeting a condition; the deep learning algorithm submodel is also used for: and inputting the initial fault type, the fault data and the confidence factors corresponding to the fault data into the machine learning algorithm submodel to obtain the final fault type.

Description

Network switch intelligent operation and maintenance method and system applying AI and network switch Technical Field The invention relates to the technical field of network switches, in particular to an intelligent operation and maintenance method and system for a network switch applying AI and the network switch. Background The network switch is used as a core device for data transmission and control, and the running stability of the network switch directly influences the communication efficiency and service continuity of the whole system. The traditional network switch state operation and maintenance mode mainly depends on a threshold judgment mode of single parameters such as port rate, power consumption, CPU occupancy rate and the like, and in the deployment process of an operation and maintenance network, the state judgment is carried out on an abnormality detection system by a centralized deployment network based on a static threshold or a linear model, abnormal data is sent to a central server, and the central server carries out fault judgment and operation and maintenance; The method cannot effectively cope with complex coupling relation between a dynamically-changed running environment and multidimensional parameters, when faults occur, hysteresis and untimely response often exist in data feedback and fault judgment and operation and maintenance of a central server, in practical application, faults of a network switch often appear different, namely after the network switch is deployed, the type of the faults often appear the specificity corresponding to the area, the scene and the transmitted data of the network switch, and the traditional method can cover the fault type only by carrying out great amount of training and maintenance on a diagnosis model, so that the deployment cost is increased, and the reliability of operation and maintenance is reduced. Disclosure of Invention In order to solve the problems in the prior art, the embodiment of the invention provides a network switch intelligent operation and maintenance method, a system and a network switch applying AI, comprising: In one aspect, there is provided a network switch intelligent operation and maintenance method applying AI, the method comprising: Configuring an AI model for each network switch, wherein the AI model actively identifies the initial fault type of the corresponding network switch; Setting a reference range corresponding to the network switch in real time, wherein the reference range comprises a plurality of network switches related to the network switch; acquiring fault data matched with the initial fault type from historical fault data of the plurality of network switches; forwarding the fault data and the initial fault type to an AI model meeting a condition; And the AI model identifies the final fault type according to the initial fault type and the fault data and outputs a corresponding operation and maintenance strategy. Optionally, the final fault type includes hardware faults and non-hardware faults; the operation and maintenance strategy comprises the following steps: if the final fault type is hardware fault, starting the disaster recovery network switch, and screening maintainers and maintenance strategies which meet preset conditions; and if the final fault type is a non-hardware fault, starting the disaster recovery network switch, positioning a fault reason, and executing a fault source equipment operation parameter adjustment strategy corresponding to the fault reason. Alternatively to this, the method may comprise, The AI model comprises a machine learning algorithm sub-model and a deep learning algorithm sub-model; the configuring the AI model for each network switch includes: presetting the machine learning algorithm sub-model, and training the machine learning algorithm sub-model through historical fault data of the network switch; Presetting the deep learning algorithm sub-model, and training the deep learning algorithm sub-model through the historical fault data of the plurality of network switches and the historical fault data of the network switches. Optionally, the setting, in real time, a reference range corresponding to the network switch includes: acquiring a plurality of first network switches which are in the same area as the network switch, and setting a first weight; acquiring a plurality of second network switches which transmit the same service data with the network switches, and setting second weights; acquiring a plurality of third network switches in a similar scene with the network switches, and setting third weights; And screening a plurality of fourth network switches which simultaneously meet the plurality of first network switches, the plurality of second network switches and the plurality of third network switches according to the corresponding first weights, second weights and third weights to obtain the plurality of network switches. Optionally, the AI model actively identifying