With the development of digitalization, the scale of business systems is expanding, and the number of devices is growing, directly leading to a surge in the types and quantity of alarms. This has significantly increased the difficulty for operation and maintenance (O&M), affecting system fault recovery. How to automatically parse alarms, generate corresponding alarm disposal method, has attracted widespread attention. Based on historical alarm data, expert O&M experience, and other factors, using machine learning algorithms or large language models (LLMs) to explore the correlation between alarms and response schemes, and generate disposal methods for real-time alarms, are important directions for intelligent O&M. This report introduces a knowledge-driven alarm interpretation generation method. By extracting alarm knowledge to build a professional knowledge base in the O&M field, the real-time alarm conducts retrieval in the knowledge base. The retrieved relevant background knowledge is combined with the current processing content of the LLM to enhance the model's knowledge understanding and response capabilities. Based on the joint knowledge graph of network elements and businesses, the associated alarms of the current alarm are inferred, and the impact on real-time businesses is determined. Finally, alarm interpretation text is outputted in five dimensions: alarm description, impact, cause, handling steps, and alarm correlation analysis, which effectively supports alarm understanding and handling in cloud O&M scenarios.

 

张亚茹，中移动信息技术有限公司，工程师。2023年7月毕业于中国科学院数学与系统科学研究院并获博士学位。主要研究方向为人工智能与云计算，现负责私有云智能运维以及存储分级调度相关工作。作为核心成员，参与1项国家级项目，1项央企算网联合体项目，以及多项公司重点项目。近五年在自然语言处理、图学习等领域发表或录用高水平论文7篇，参与3项行标的撰写，申请专利7篇。近三年已发表（或正式接收）的学术论文：

Zhang Y R, Tang X J. Feature disentanglement and homogeneous second order feature propagation for recommendation. Journal of Intelligent Informatics and Smart Technology, 2022, 7.

Zhang Y R, Tang X J. Introducing trigger evolutionary graph and event segment for event prediction[C]//The 21st International Symposium on Knowledge and Systems Sciences (KSS2022). Springer, 2022: 186-201.

Zhang Y R, Tang X J. News event prediction by trigger evolution graph and event segment. Journal of Systems Engineering and Electronics, 2023, 34(3): 615-626. (SCI)

张亚茹, 唐锡晋, 等. 联合自监督时序图学习与标签选择的事件预测. 系统工程理论与实践, 2025.(已接收)