Network fault analysis method and system for complex network

Abstract

The invention discloses a network fault analysis method and system for a complex network, and belongs to the field of network fault analysis. The method comprises the following steps: decomposing a complex network topological graph into a plurality of closely adjacent sub-network topological graphs according to the degree and connection relationship of nodes; combining the serialized matrix of each sub-network topological graph and the fault information of each time of the sub-network topological graph into a fault matrix; inputting all the fault matrixes of each subnet topological graph intoa convolutional neural network model for training to obtain the fault diagnosis model of the subnet topological graph; and segmenting the complex network topology to be tested to obtain each sub-network topology of the complex network topology, and introducing the sub-network topologies into the trained fault diagnosis model of the plurality of sub-network topological graphs in combination with the alarm information for comprehensive analysis to obtain a network fault diagnosis result. According to the method, the complex network topology is divided into a plurality of sub-networks with characteristics, each sub-network is of the same composition, isomorphism and a connection relationship are kept, and feasibility of large-area network coverage network fault analysis is realized.

Description

technical field [0001] The invention belongs to the field of network fault analysis, and more specifically relates to a network fault analysis method and system for complex networks. Background technique [0002] The communication network network is connected in turn by multiple routes, devices, etc., and the connection subjects (routes, devices, etc.) are regarded as nodes to form a ring network, such as a power communication network. When a network fails, it is often caused by one or more nodes. It is necessary to analyze the entire network to find the failed node and the type of failure. Professional network management collects network fault alarm information data, including: fault type, fault level, alarm name, alarm source, location information, occurrence time and other fields. Alarm information data is derivative and disseminated, which adds difficulty to the real fault alarm judgment. [0003] With the development of science and technology, the scale of the network...

AI Technical Summary

Problems solved by technology

It is difficult for the existing communication network network fault analysis to cover all networks and devices, and it is difficult to cope with the increasing network scale, especially for network faults that have never occurred

In addition, because the traditional network fault analysis is to number the network topology, once the device number changes, the network representation will also change. Therefore, the system cannot intelligently analyze networks with the same topology but different numbers

Requires significant model training time and results in repetitive training of the same or similar networks

Images