|本期目录/Table of Contents|

[1]张凯,周清,商景林,等.基于大语言模型的生产安全（工矿商贸）事故报告数字化方法*[J].中国安全生产科学技术,2025,21(1):38-47.[doi:10.11731/j.issn.1673-193x.2025.01.005]
　ZHANG Kai,ZHOU Qing,SHANG Jinglin,et al.Digitization method of work safety (industry,mining,commerce and trade) accident reports based on Large Language Model[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2025,21(1):38-47.[doi:10.11731/j.issn.1673-193x.2025.01.005]
点击复制

基于大语言模型的生产安全（工矿商贸）事故报告数字化方法*
分享到：

《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]

卷:
21
期数:
2025年1期
页码:
38-47
栏目:
学术论著
出版日期:
2025-01-30

文章信息/Info

Title:
Digitization method of work safety (industry,mining,commerce and trade) accident reports based on Large Language Model
文章编号:
1673-193X(2025)-01-0038-10
作者:
张凯; 周清; 商景林; 李焱; 程嘉昇
（上海市安全生产科学研究所，上海 200233）
Author(s):
ZHANG Kai;  ZHOU Qing;  SHANG Jinglin;  LI Yan;  CHENG Jiasheng
(Shanghai Institute of Work Safety Science,Shanghai 200233,China)
关键词:
事故报告; 数字化; 大语言模型; 24Model; 信息抽取; 致因分类
Keywords:
accident report;  digitization;  Large Language Model;  24Model;  information extraction;  causation classification
分类号:
X928.01
DOI:
10.11731/j.issn.1673-193x.2025.01.005
文献标志码:
A
摘要:
为推动生产安全管理数字化转型、智能化提取事故报告价值数据，基于大语言模型并结合信息抽取与文本分类技术，提出1种面向工矿商贸行业领域的生产安全事故报告数字化方法。首先分析事故报告结构特点和编写规律，以正则化方法匹配提取事故报告关键段落，实现文本预处理，然后建立命名实体识别任务对事故报告基本特征进行文本抽取，并验证大语言模型微调前后的信息抽取效果，同时根据24Model设计事故原因模块化分类指标，通过模型参数微调实现事故致因特征层级分类预测，并作多场景和单一场景下致因分类对比实验。研究结果表明：大语言模型具备较强的泛化能力，通过少量数据标注与参数微调可快速适配事故报告信息抽取、致因分类任务，模型综合评价指标分别可达0.87和0.85，本文所构建的数字化方法可行有效。研究结果可为应急管理大数据底座的建设提供技术参考。
Abstract:
To promote the digital transformation of work safety management and the intelligent extraction of valuable data from accident reports,a digitization method of the work safety accident reports targeting the industry,mining,commerce and trade industries and fields was proposed based on the Large Language Model combined with information extraction and text classification technology.Firstly,the structural characteristics and writing rules of accident reports were analyzed,and the regularized expression method was used to match and extract the key paragraphs of accident reports to achieve the text preprocessing.Secondly,a named entity recognition task was established to conduct the test extraction on the basic features of accident reports,and the information extraction effect before and after fine-tuning of Large Language Model was verified.Finally,the modular classification indexes of accident causes were designed based on the 24Model,then the hierarchical classification prediction of accident causation features was achieved through the fine-tuning of model parameters,and the comparative experiments were conducted on the causation classification in multiple and single scenarios.The results show that the Large Language Model has strong generalization ability,and can quickly adapt to the information extraction and causation classification tasks of accident reports through a small amount of data annotation and parameter fine-tuning.The comprehensive evaluation indexes of the model can reach 0.87 and 0.85,respectively,and the digitization method is feasible and effective.The research results can provide technical reference for the construction of emergency management big data base.

参考文献/References:

［1］马坚泓,阎祖强,周宏，等.上海市安全生产条例释义［M］.上海:中国出版集团,2023:58-60.
［2］上海市人民政府办公厅.上海市人民政府办公厅关于印发《上海市应急管理“十四五”规划》的通知［EB/OL］.(2021-08-16)［2024-12-23］.https://www.shanghai.gov.cn/nw12344/20210816/7c35057f10ff46a1a47f1be37e1a01f9.html.
［3］况宇琦.基于数据挖掘的塔吊事故案例分析研究［D］.武汉：华中科技大学,2021.
［4］张萌,陈佳惠,孙然然，等.基于规则的城市轨道交通安全事件信息抽取及其知识元表示［J］.科学技术与工程,2021,21(15):6435-6440. ZHANG Meng,CHEN Jiahui,SUN Ranran,et al.Rule-based information extraction of urban rail transit safety cases and its common knowledge meta-model representation ［J］.Science Technology and Engineering,2021,21(15):6435-6440.
［5］梁建军,雷咸锐,吴斌，等.基于规则模式的瓦斯爆炸事故信息抽取技术［J］.煤矿安全,2023,54(2):239-245. LIANG Jianjun,LEI Xianrui,WU Bin,et al.Gas explosion accident information extraction technology based on regular model［J］.Safety in Coal Mines,2023,54(2):239-245.
［6］张鹏翔.多维字符特征表示的铁路设备事故信息抽取方法［J］.中国安全科学学报,2022,32(6):109-114. ZHANG Pengxiang.Information extraction method for railway equipment accidents based on multi-dimensional character feature representation［J］.China Safety Science Journal,2022,32(6):109-114.
［7］DO Q,LE T,LE C.Automated construction activities extraction from accident reports using deep neural network and natural language processing techniques［C］//International Conference on Construction Engineering and Project Management,2022:744-751.
［8］王明达,张榜,吴志生，等.基于强化学习的城镇燃气事故信息抽取方法［J］.中国安全生产科学技术,2023,19(3):39-45. WANG Mingda,ZHANG Bang,WU Zhisheng,et al.Information extraction method of urban gas accidents based on reinforcement learning［J］.Journal of Safety Science and Technology,2023,19(3):39-45.
［9］陶建华,聂帅,车飞虎.语言大模型的演进与启示［J］.中国科学基金,2023,37(5):767-775. TAO Jianhua,NIE Shuai,CHE Feihu.The evolution and inspiration of large language model technology［J］.Bulletin of National Natural Science Foundation of China,2023,37(5):767-775.
［10］车万翔,郭江,崔一鸣.自然语言处理——基于预训练模型的方法［M］.北京:电子工业出版社,2021.
［11］邵浩,刘一峰.预训练语言模型［M］.北京:电子工业出版社,2021.
［12］张钦彤,王昱超,王鹤羲,等.大语言模型微调技术的研究综述［J］.计算机工程与应用,2024,60(17):17-33. ZHANG Qintong,WANG Yuchao,WANG Hexi,et al.Comprehensive review of large language model fine-tuning［J］.Computer Engineering and Applications,2024,60(17):17-33.
［13］张奇,桂韬,黄萱菁.自然语言处理导论［M］.北京:电子工业出版社,2023.
［14］LU Y J,LIU Q,DAI D,et al.Unified structure generation for universal information extraction［C］//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics,2022.
［15］国汉君,江益,姚勇征,等.基于内-外因理论和Apriori算法的动火作业事故分析［J］.中国安全生产科学技术,2024,20(11):101-109. GUO Hanjun,JIANG Yi,YAO Yongzheng,et al.Analysis of hot work accidents based on internal causes and Apriori algorithm［J］.Journal of Safety Science and Technology,2024,20(11):101-109.
［16］HOLLNAGEL E.Barriers and accident prevention ［M］.Routledge,2016.
［17］刘晓露,严玉琼,张苏,等.IFAM与24Model的对比研究 ［J］.中国安全生产科学技术,2024,20 (9): 233-240. LIU Xiaolu,YAN Yuqiong,ZHANG Su,et al.Comparative study of IFAM and 24Model［J］.Journal of Safety Science and Technology,2024,20 (9): 233-240.
［18］吕千,傅贵.事故调查报告中事故原因分析的改进——基于企业事故案例学习视角 ［J］.中国安全生产科学技术,2021,17 (10): 172-178. LYU Qian,FU Gui.Improvement of cause analysis in accident investigation reports——A perspective of enterprise accident cases learning［J］.Journal of Safety Science and Technology,2021,17 (10): 172-178.
［19］FU G,WANG J,YAN M.Anatomy of Tianjin Port fire and explosion:process and causes［J］.Process Safety Progress,2016,35(3):216-220.
［20］ZHANG S,WANG J.Analysis of South Korea Sewol sunken ferry accident based on behavioral safety［J］.Journal of Coastal Research,2015,73(10073):611-613.
［21］张江石,胡馨月,侯轩,等.基于24Model的地铁内涝事故原因分析与评估 ［J］.安全与环境工程,2024,31 (6): 111-117. ZHANG Jiangshi,HU Xinyue,HOU Xuan,et al.Analysis and evaluation of the causes of subway waterlogging accidents based on 24Model［J］.Safety and Environmental Engineering,2024,31 (6): 111-117.
［22］刘晓露,梁志星,吴君安,等.24 Model在道路运输事故成因中的应用［J］.汽车安全与节能学报,2024,15(2):185-192. LIU Xiaolu,LIANG Zhixing,WU Jun’an,et al.Application of 24Model in the cause of road transportation accident［J］.Journal of Automotive Safety and Energy,2024,15(2):185-192.
［23］中华人民共和国中央人民政府.生产安全事故报告和调查处理条例［EB/OL］.（2007-04-09）［2024-12-26］.https://www.gov.cn/zwgk/2007-04/19/content_588577.htm.
［24］傅贵,王秀明,李亚.事故致因“2-4”模型及其事故原因因素编码研究［J］.安全与环境学报,2017,17(3):1003-1008. FU Gui,WANG Xiuming,LI Ya.On the 24Model and the application of its causative codes to the analysis of the related accidents［J］.Journal of Safety and Environment,2017,17(3):1003-1008.
［25］傅贵,吴琼,李亚,等.不安全动作与不安全物态的判定方法探讨［J］.安全与环境学报,2017,17(3):994-998. FU Gui,WU Qiong,LI Ya,et al.Discussion and justification of the assessment methods of the unsafe acts and the unsafe conditions［J］.Journal of Safety and Environment,2017,17(3):994-998.
［26］付汇琪,陈萍,傅贵.对安全学科中“事故”定义的探讨［J］.科技导报,2017,35(7):11.
［27］FU G,XIE X,JIA Q,et al.The development history of accident causation models in the past 100 years:24Model,a more modern accident causation model［J］.Process Safety and Environmental Protection,2020,134:47-82.
［28］朱远星,彭知军,杜书杰,等.基于24Model城镇燃气工程施工安全事故原因分析［J］.煤气与热力,2023,43(6):29-37. ZHU Yuanxing,PENG Zhijun,DU Shujie,et al.Cause analysis of safety accidents in urban gas engineering construction based on 24Model［J］.Gas & Heat,2023,43(6):29-37.
［29］STEWART J M.Managing for world class safety［M］.New York:A Wiley-Interscience Publication,2002:1-31.
［30］安全管理研究中心.安全文化测量［EB/OL］.(2013-02-24)［2024-12-26］.http://www.safetyscience.cnpageaqwhcl/122.php.
［31］傅贵,索晓,王春雪.24Model的系统特性研究［J］.系统工程理论与实践,2018,38(1):263-272. FU Gui,SUO Xiao,WANG Chunxue.Study on the systematic characteristics of 24Model［J］.Systems Engineering-Theory & Practice,2018,38(1):263-272.

相似文献/References:

[1]曾明荣,王兴.中欧事故报告与调查处理比较研究[J].中国安全生产科学技术,2015,11(5):148.[doi:10.11731/j.issn.1673-193x.2015.05.023]
　ZENG Ming-rong,WANG Xing.Comparative research on reporting, investigation and disposal of work safety accidents in China and EU[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2015,11(1):148.[doi:10.11731/j.issn.1673-193x.2015.05.023]
[2]贾虎军,王立娟,靳晓,等.基于无人机航测的尾矿库三维空间数据获取与风险分析[J].中国安全生产科学技术,2018,14(7):115.[doi:10.11731/j.issn.1673-193x.2018.07.017]
　JIA Hujun,WANG Lijuan,JIN Xiao,et al.Threedimensional spatial data acquisition and risk analysis of tailings pond based on UAV aerial survey[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2018,14(1):115.[doi:10.11731/j.issn.1673-193x.2018.07.017]
[3]林麟,李群.基于结构化与数字化技术的轨道交通建设工程应急预案体系设计与构建方法研究[J].中国安全生产科学技术,2019,15(12):135.[doi:10.11731/j.issn.1673-193x.2019.12.022]
　LIN Lin,LI Qun.Research on design and construction method of emergency plan system for metro construction project based on structuration and digitization technologies[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2019,15(1):135.[doi:10.11731/j.issn.1673-193x.2019.12.022]
[4]王伟,刘晓雅,卢菊花,等.“数字化集中仓”模式下危化品仓储安全监管的演化博弈分析*[J].中国安全生产科学技术,2023,19(1):156.[doi:10.11731/j.issn.1673-193x.2023.01.023]
　WANG Wei,LIU Xiaoya,LU Juhua,et al.Evolutionary game analysis of safety supervision for hazardous chemicals warehousing under “digital centralized warehouse” mode[J].JOURNAL OF SAFETY SCIENCE AND TECHNOLOGY,2023,19(1):156.[doi:10.11731/j.issn.1673-193x.2023.01.023]

备注/Memo

备注/Memo:
收稿日期： 2024-08-13
* 基金项目： 上海市安全生产科学研究所科研项目(SFDF-24010);国家标准化管理委员会标准制修订项目（20242804-Q-450）
作者简介： 张凯，硕士，高级工程师，主要研究方向为人工智能、应急数字化。
更新日期/Last Update: 2025-01-26