Research On The Method Of Causation Analysis Of Production Accident Based On Data Driving

Learning experience from historical accidents is of great significance to prevent the recurrence of similar accidents.However,with the increasing complexity of safety production system,the mechanism of safety production accidents is becoming more and more complex.Researchers and enterprise safety managers are unable to fully identify the potential information in historical accidents,which leads to the frequent occurrence of accidents.Therefore,enterprise safety management is faced with more severe challenges.How to fully and effectively learn experience from historical accidents has become the key problem to solve the frequent accidents.In recent years,with the rapid development of information technology in China's safety production enterprises and government departments,a large number of safety production data or accident data have been accumulated,which provides a way for more in-depth and effective accident learning.However,in the face of a large number of accident data with complex structure,the traditional data analysis theories and methods can not be used effectively,which leads to the waste of a lot of valuable accident information and restricts the in-depth understanding and research of accidents.Therefore,based on the above problems,this paper aims to build a data-driven mining process of safety production accident causes.According to different accident data structures,data mining methods suitable for their structural characteristics are selected to maximize the utilization of accident data.According to the sequence thinking of preliminary identification of accident information,discovery of safety problems and transformation of safety knowledge,the valuable information in historical accident data is fully mined.The main contents of this paper are as follows:(1)This paper discusses the current situation and characteristics of safety production big data and accident data,and selects text-based accident data and structured accident data as the main research object of this paper.(2)On the basis of defining the characteristics of text-based accident data and structured accident data,combined with the theory of accident cause and the currentrequirements of enterprise safety management,this paper selects data mining or machine learning methods that are suitable for different accident data structures,and constructs an accident cause mining process that is suitable for two different data structures.Based on this process,the original accident data is transformed into valuable accident information and safety knowledge.(3)Using the accident data of a chemical enterprise and Zhejiang expressway,this paper makes an empirical study on the accident cause mining process.Analyze the applicability,accuracy and efficiency of the mining process,and provide targeted suggestions for relevant enterprises and government departments based on the mining results.The results show that: the constructed accident cause mining process maximizes the utilization of accident data;compared with the traditional causative analysis method,the data-driven quantitative analysis process is more objective and efficient.The analysis results can be obtained from the perspectives of association rules,accident scenario reconstruction,causative factor centrality and result visualization to understand the accident mechanism.