摘要
为有效检查煤矿建筑设计的合规性,从事故致因角度出发,利用词相似度计算(Word2Vec)模型和数据库技术,整理数量庞大、排列冗杂的规范数据,使其标准化表达。首先,运用Word2Vec模型对2009—2021年225起由煤矿建筑设计不合规引起的事故分析报告进行词向量训练,发掘出与各类事故发生最为相关的煤矿建筑设计风险要素;其次,从已获取的建筑设计风险要素出发,收集与筛选相关现行的煤矿建设规范,通过分析具体规范条款,构建规范标准化表达框架;最后,将其整理到结构化查询语言(SQL)数据库中,构建标准化规范数据库。结果表明:利用Word2Vec模型,可根据词深层次特征发掘出事故分析报告中与事故发生的建筑风险要素,将其作为检索中心词初步检索和过滤规范信息,构建基于事故致因的标准化煤矿建筑规范库,打破规范标准间的壁垒。
In order to effectively check the compliance of coal mine building design,this paper used word similarity calculation(Word2Vec)model and database technology to organize and standardize a large number of regulations data from the perspective of accident causation.Firstly,the Word2Vec model was used to train word vectors on 225 accident reports from 2009 to 2021 caused by non-compliance of coal mine building design,so as to find out the risk elements of coal mine building design which were most associated with the occurrence of various accidents.Then,the relevant existing coal mine construction regulations were collected and screened from the acquired building design risk factors,a standardized expression framework was constructed for the regulations by analyzing specific regulations provisions.Finally,the regulations were organized into a structured query language(SQL)database to build a standardized regulation database.Using the Word2Vec model,the construction risk elements associated with the occurrence of accidents in accident analysis reports can be uncovered based on the deep-level characteristics of words.The construction risk elements can be used as the central word for the initial search and filtering of specification information,thus building the standardized library of coal mine construction regulations,which breaks the barriers between regulations and standards.
作者
刘全龙
张晓霖
张悦倩
邱尊相
LIU Quanlong;ZHANG Xiaolin;ZHANG Yueqian;QIU Zunxiang(School of Economics and Management,China University of Mining and Technology,Xuzhou Jiangsu 221116,China)
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2023年第5期128-133,共6页
China Safety Science Journal
基金
江苏省教育厅哲学社会科学研究重大项目(2020SJZDA085)。