山东煤炭科技2025，Vol.43Issue(12)：204-208,5.DOI:10.3969/j.issn.1005-2801.2025.12.036
煤矿安全监控系统报警分类及合并研究
Research on Alarm Classification and Merger of Coal Mine Safety Monitoring System
田乐逍 1王新会 1张倩1
作者信息
- 1. 应急管理部信息研究院,北京 100029
- 折叠
摘要
Abstract
Aiming at the problems of high information redundancy and low classification efficiency in over-limit alarm of coal mine safety monitoring system,an intelligentization processing framework for gas and carbon monoxide alarms is constructed.Based on the Coal Mine Safety Regulations and industry expert knowledge base,a standardized classification system including 7 types of carbon monoxide alarms and 6 types of gas alarms is established,and a solution that integrates Word2Vec text feature extraction and neural network classification model is proposed.The experiment shows that the neural network model achieves a Micro-F1 value of 99.04%(with a decision tree benchmark of 91.34%)in carbon monoxide alarm classification,and that of the gas alarm classification reaches 92.51%(with a benchmark of 89.72%);The cosine similarity algorithm is adopted to implement duplicate alarms merger,achieving an accuracy rate of 92.73%for merging carbon monoxide redundant alarms and 92.36%for gas alarms;A coal mine alarm processing system integrating deep learning and text similarity calculation is constructed,a technical closed loop of"feature extraction-intelligent classification-redundancy decomposition"is formed,providing a systematic method for achieving real-time processing of 100 000 alarms per day for mine safety monitoring,and the efficiency is improved by over 85%compared to the traditional manual processing.关键词
煤矿报警分类/文本相似度/超限报警/决策树/神经网络Key words
coal mine alarm classification/text similarity/over-limit alarm/decision tree/neural network分类
矿业与冶金引用本文复制引用
田乐逍,王新会,张倩..煤矿安全监控系统报警分类及合并研究[J].山东煤炭科技,2025,43(12):204-208,5.
