Analysis Of Causes Of Tunnel Engineering Accidents Based On Text Mining

As a large-scale underground project with large-scale construction organization,long construction period and high risk of geological disasters,various risk events occurred in the process of project construction often cause life and property losses to on-site construction personnel.In order to ensure the construction safety of tunnel engineering and effectively control the hidden safety hazards in the process of engineering construction,how to reduce the occurrence of safety accidents is an urgent problem to be solved.The idea of knowledge management and text mining technology can realize the conversion of data to knowledge,that is,to convert the historical data in the process of tunnel construction into corresponding risk management knowledge.Based on this,this paper uses the investigation reports formed by the accidents in the tunnel construction process as a corpus,and builds a professional thesaurus to improve the text segmentation effect.A vocabulary that characterizes safety risk factors is developed to form a list of accident causes,and to achieve the purpose of structured processing of relevant report texts.The main work of this paper is as follows:(1)Introduce the LDA topic clustering method.Through the LDA topic clustering method,the topic of the tunnel engineering accident text is excavated,and the text data is cleaned according to the characteristics of the tunnel construction.Determine the optimal number of topics through perplexity calculation and consistency calculation,and then perform topic mining on the processed text,realize text visualization based on py LDAavis,and analyze the visualized text.(2)On the basis of the extracted event text,optimize and improve HFACS.Since HFACS contains two levels of accident causes,the parent category and the sub-category,SVM is used to divide the accident report into two levels according to the cause of the accident,and construct a A structured collection of event data.The causes of accidents are classified according to different levels of personnel and organizations,and the preconditions for unsafe behavior are subdivided into three levels: task,environment,and individual.Each level is then analyzed according to the characteristics of the accident text.Refinement to form more specific accident causes.(3)For the cause of the parent type of accident in the constructed accident data set,the construction safety risk control measures are proposed to reduce the probability of risk and ensure the safety of construction personnel.