5G支撑网告警数据的故障定位方法

Fault Location Method for Alarm Data of 5G Supporting Network

随着电信网规模不断扩大,业务支撑网发生故障的频率不断上升,运营商无法从大量冗余的告警数据中进行故障定位。为此提出一种基于人工智能的告警事件关联的故障定位方法,该方法采用深度学习的方法提取一系列告警事件的多维度语义特征;并结合增量式BP神经网络对模型的参数进行增量式更新;然后,挖掘告警事件与故障类型的动态相关性并基于告警事件快速确定故障类型;最后,结合序列模式挖掘,实现故障的精准定位。实验表明,本文的方法可以有效挖掘有价值的告警—故障关联规则,从而解决告警事件的关联性分析难的问题。网络维护人员可根据现有的告警数据快速定位故障位置,有效保障网络安全。

With the continuous expansion of the scale of the telecommunications network and the increasing frequency of failures in the service supporting network, operators cannot locate the failures from a large number of redundant alarm data. To this end, a fault location method is proposed with the alarm event correlation based on the artificial intelligence. This method uses deep learning to extract multi-dimensional semantic features from a series of alarm events, and an incremental BP neural network is combined to incrementally update the model parameters. Then, the dynamic correlation between alarm events and fault types is mined and the fault type can be quickly determined based on alarm events. Finally, an accurate fault location is realized by combining the sequential pattern mining. Experiments show that the proposed method can effectively mine valuable alarm-fault association rules and solve the problem of the difficulty in analyzing the alarm event correlation. Through feature association, this algorithm can mine valuable and practical alarm fault association rules. The network maintenance personnel can quickly find the fault locations according to the existing alarm data, so as to effectively support the network security.