煤矿安全风险分析的文本数据模型与集成分析平台  

Text data model and integrated analysis platform for coal mine safety risk analysis

作　　者：王启飞 王俊龙 刘昊霖 赵逸涵 孙英峰 李蓓 WANG Qifei;WANG Junlong;LIU Haolin;ZHAO Yihan;SUN Yingfeng;LI Bei(School of Mechanical-Electronic and Automobile Engineering,Beijing University of Civil Engineering and Architecture,Beijing 102616,China;Beijing District Heating Group Co.,Ltd.,Beijing 100026,China;Research Institute of Macro-Safety Science,University of Science and Technology Beijing,Beijing 100083,China;Ningxia Hui Autonomous Region Bureau of Coal Geology,Yinchuan 750004,China)

机构地区：[1]北京建筑大学机电与车辆工程学院,北京102616 [2]北京市热力集团有限责任公司,北京100026 [3]北京科技大学大安全科学研究院,北京100083 [4]宁夏回族自治区煤炭地质局,银川750004

出　　处：《安全与环境学报》2024年第11期4358-4365,共8页Journal of Safety and Environment

基　　金：国家自然科学基金项目(U23A20285);北京市自然科学基金项目(2023MH202);煤与煤层气共采国家重点实验室开放基金项目(2022KF21);宁夏自然科学基金项目(2023AAC03778);北京市教育委员会科学研究计划项目(KM202410016004)。

摘　　要：在煤矿安全领域,事故的预防至关重要。为了对煤矿开采风险进行深入分析,提出了一种基于文本数据的煤矿安全事故智能分析模型及集成分析平台。首先,采用融合数据增强技术的卷积神经网络文本分类(Text-Convolutional Neural Network,Text-CNN)方法构建煤矿安全事故分析模型,对大量非结构化事故文本进行精准的分类筛选;然后,利用自然语言处理(Natural Language Processing,NLP)技术建立煤矿事故简报集成分析系统,通过该系统对煤矿事故报告进行事故统计分析、风险分析等,总结出不同地区煤矿事故的死亡情况与类型差异,明确了煤矿安全事故之间的潜在模式。研究表明,通过集合事故简报分析模型的集成分析平台可以实现对煤矿安全事故信息的获取再利用,分析事故潜在规律和风险大小,有助于提升煤矿的风险管理水平,提高事故预防能力。In the realm of coal mine construction safety,accident prevention is critical.To facilitate a thorough analysis of coal mining risks,this paper proposes a coal mine safety risk model and an integrated analysis platform that leverages text data.First,the model gathered extensive unstructured text data from online platforms,including safety accident briefings from various industries.It utilized text classification methods to categorize and filter accident information,enabling the effective reuse of data within the coal mining sector.During the text classification process,various classification models were employed to categorize accidents within the same dataset.The effectiveness of these models was evaluated using accuracy,recall,and F1-score derived from the confusion matrix,allowing for a comprehensive analysis of their performance from multiple perspectives.Ultimately,the Text-Convolutional Neural Network(Text-CNN)classification model,enhanced with additional data,was identified as the most effective.Subsequently,an accident briefing analysis system was established using Natural Language Processing(NLP)technology to conduct an in-depth analysis of the selected coal mine accident information.The integrated analysis platform for accident briefings comprised several subsystems,each dedicated to specific tasks,including the screening of accident information,statistical analysis of incidents,and risk assessment of accidents.By conducting a comprehensive analysis of coal mine safety accidents and fatalities across various regions,we gained valuable insights into the different types of accidents prevalent in each area.This analysis also helped to identify potential patterns among coal mine safety incidents.The results indicate that the generalization capability of different models,along with the completeness of the dataset,significantly affects the accuracy of accident information classification and screening.Additionally,the integrated analysis platform,incorporating the accident briefing analysis model,can examine th

关 键 词：安全工程 煤矿事故 卷积神经网络文本分类(Text-CNN) 自然语言处理(NLP) 事故预防 

分 类 号：X936[环境科学与工程—安全科学]