When faults occur in optical networks, many derivative alarms are generated from the root alarms, because of the derived properties of optical network alarms. In a large and complex optical network, the management system will receive tremendous alarms even though there is only one actual fault. Therefore, accurate fault localization has significance for network operation and maintenance. The complex relationship between alarms and faults makes it difficult to locate faults, especially multiple faults. To solve this problem, this paper proposes a fault localization method based on knowledge graph and network topology information. Using the information association ability of the knowledge graph, the implicit relationship between the alarm data is mined, and the graph neural network is used to perform automatic fault reasoning on the alarm knowledge graph. Further, the network topology information is integrated into the fault localization, which increases the dimension of the knowledge graph. Simulation results show that the proposed method can achieve 97% accuracy in multiple fault localization scenarios.
|