水电工程施工安全隐患文本智能类推研究  

Research on text intelligent analogy of potential safety hazards in hydropower engineering construction

作　　者：郑霞忠[1,2] 汪珂 陈云[2] 晋良海[2] ZHENG Xiazhong;WANG Ke;CHEN Yun;JIN Lianghai(Hubei Key Laboratory of Construction and Management in Hydropower Engineering,China Three Gorges University,Yichang 443002,Hubei,China;College of Hydraulic&Environmental Engineering,China Three Gorges University,Yichang 443002,Hubei,China)

机构地区：[1]三峡大学水电工程施工与管理湖北省重点实验室,湖北宜昌443002 [2]三峡大学水利与环境学院,湖北宜昌443002

出　　处：《安全与环境学报》2023年第12期4449-4456,共8页Journal of Safety and Environment

基　　金：国家自然科学基金项目(52209163,51878385)。

摘　　要：水电工程施工安全隐患治理不断向信息化与智能化转型。为高效挖掘大规模非结构化的安全隐患文本数据,提出融合案例推理与深度学习的水电工程施工安全隐患文本智能类推方法,辅助隐患治理方案的制订,提高隐患治理效率。首先,柔性化处理案例数据,用框架法表示案例并建立隐患案例库;其次,从案例推理技术视角出发,构建隐患治理方案类推框架;最后,融合Word2vec模型优化检索过程,将隐患文本转化为词向量并计算相似度。以某水电站2016—2020年记录的3160条安全隐患信息为数据源,经实例验证,该类推方法综合准确率达0.867,表明隐患文本智能类推方法有助于管理人员及时对隐患进行整改,能够有效指导水电工程安全施工及安全管理。With the development of emerging information technology,the management of potential safety hazards in hydropower project construction is constantly transforming to informatization and intelligence.To efficiently mine large-scale unstructured text data of potential safety hazards,an intelligent analogy method for the text of potential safety hazards in hydropower project construction integrating case reasoning and in-depth learning is proposed to assist in the formulation of potential risk management plans and improve the efficiency of potential risk management.Artificial intelligence technology is used to overcome subjectivity and experience dependence on manual troubleshooting of hidden dangers,and effectively reduce the incidence of accidents.Based on hidden danger text big data,an intelligent text analogy method for hydropower project construction hidden dangers is proposed,which integrates case reasoning and in-depth learning.First of all,flexibly process case data to ensure that case information is not omitted or repeated.Frame method represents cases and establishes hidden danger case library.Secondly,from the perspective of case-based reasoning technology,we build an analogy framework for hidden danger management schemes and use historical information to derive current required knowledge.At last,the Word2vec model is integrated to optimize the retrieval process,the hidden text is converted into a word vector,and the similarity is calculated.In the case retrieval stage,determine the case set of the target case by calculating the Jaccard coefficient,narrow the case retrieval scope,and reduce the analogy time.Word2vec calculates the similarity of hidden trouble descriptions based on context logic and short text semantic features.The results show that the word vector visualization is realized based on the Embedding projector with 3 160 hidden danger information recorded in 2016-2020 of the hydropower station as the data source.On this basis,the target case analogy results are obtained,and the comprehensive acc

关 键 词：安全社会工程 案例推理 隐患治理 深度学习 文本类推 Word2vec模型 

分 类 号：X947[环境科学与工程—安全科学]