摘要
Abstract
To address challenges such as complex architectures,heterogeneous equipment,over-reliance on manual expertise,and lagging responses in military information systems,this paper proposes a construction scheme for an Artificial Intelligence for IT Operations(AIOps)system based on Large Language Models(LLMs)and multi-agent technology.The system achieves unified perception of massive logs and time-series metrics through multi-source cross-modal data fusion;it employs Supervised Fine-Tuning(SFT)to construct a domain-specific O&M large model,enhancing its semantic understanding of military O&M contexts.By integrating Knowledge Graphs and GraphRAG technology,the system combines topological correlation with deep reasoning to achieve precise root cause analysis(RCA)for complex failures.Furthermore,this paper designs a collaborative workflow consisting of five types of agents-task command,data collection,anomaly detection,root cause diagnosis,and decision-making/repair-forming an autonomous"Perception-Decision-Execution"O&M closed loop.The system effectively mitigates"alert storms,"transforms passive response into proactive early warning,reduces dependence on human experts,and significantly enhances the mission assurance capabilities and resilience of military information systems.关键词
军事信息系统/智能运维/大语言模型/多智能体/知识图谱Key words
military information system/Artificial Intelligence for IT Operations(AIOps)/Large Language Model(LLM)/multi-agent/knowledge graph分类
信息技术与安全科学
