哈尔滨工程大学学报2026，Vol.47Issue(7)：1418-1426,9.DOI:10.11990/jheu.202602004
防人因失误的智能化分析与辅助工具研究进展
Research progress on intelligent analysis and support tools for preventing human errors
摘要
Abstract
To address safety challenges arising from the digital transformation of nuclear power plant control rooms,a structured analytical framework for operator error prevention is developed.Rather than enumerating individual technolo-gies,the paper traces the technical evolution of intelligent human reliability analysis(HRA)and clarifies three major development stages.The first stage focuses on knowledge-graph-based semantic modeling of human-machine interfaces to maintain cognitive consistency between procedures and interface operations.The second stage introduces machine learning and cognitive modeling techniques to estimate operator states in real time,thereby improving transparency of performance dynamics.The third stage explores the use of large language models to support proactive human-machine co-intervention for risk interruption.For each stage,the underlying technical mechanisms are examined and compared with traditional HRA approaches.The results indicate that intelligent methods offer improved data utilization,finer-grained assessment,and enhanced responsiveness in dynamic operational contexts.At the same time,issues related to model reliability and interpretability remain critical and require further attention.By integrating these developments,the study outlines a technical system for intelligent human factor analysis in digital main control rooms and provides a foun-dation for future engineering implementation and methodological refinement of intelligent HRA.关键词
核电安全/人因可靠性/认知模型/行为监测/多模态数据/智能决策支持/语义推理/数字化主控室Key words
nuclear power safety/human reliability/cognitive modeling/behavioral monitoring/multimodal data/intelligent decision support/semantic reasoning/digital main control room分类
能源科技引用本文复制引用
萧星宇,齐奔,梁金刚,赵军,林凌,沈祺,林育东,童节娟,王海涛..防人因失误的智能化分析与辅助工具研究进展[J].哈尔滨工程大学学报,2026,47(7):1418-1426,9.基金项目
领创科研和青年英才科研项目(04151001025). （04151001025）
