Research on Risk Early Warning Method Based on Big Data of Hidden Danger

Author:Wang Xin Hao

Supervisor:luo yun


Degree Year:2019





In recent years,with the deepening of the informatization,the state and society have gradually discovered that big data analysis capabilities are crucial to curbing the occurrence of serious accidents and ensuring production safety.However,the unbalanced development of data storage and analysis capability is aggravating,which has gradually evolved into the main contradiction in the field of safety production in China.Safety production data has the characteristics of huge quantity,complex content and low value density.It is difficult to efficiently excavate the important value by general system safety analysis method.It will be a crucial research topic to study analytical methods suitable for safe production of big data to guide risk early warning.This thesis aims to establish a data-driven risk early warning method.This method takes the large-scale hidden danger text report as the analysis object,and integrates risk management theory and big data analysis method to realize risk identification,risk quantitative evaluation and risk warning.The main contents and innovation contributions of this thesis are as follows.(1)Intelligent risk identification method for large-scale hidden danger text is proposed.In this method,risk identification method and text mining technology are organically combined.Many technical methods,such as system security analysis,text feature extraction and knowledge atlas theory,have been applied to analyze the sentence structure of hidden danger text and extract the risk information contained in the big data of hidden danger.Ultimately,unstructured accident risk data are transformed into structured risk basic information.(2)At this stage,basic risk information is used as the data base.The single-risk and compound-risk analysis methods based on equivalent class transformation algorithm are designed respectively to mine the potential correlation between risks.The identification algorithm and mining process of risk change pattern are designed to analyze the trend of risk change.Conditional probability and change trend of risk are integrated to construct quantitative risk assessment model and visualization scheme.Ultimately,basic risk information is transformed into potential risk information.(3)Risk warning entities and functions are clearly defined in the big data environment.The basic method and implementation process of risk warning are designed.Combined with the enterprise’s risk warning requirements,the risk warning function and process framework for big data analysis is built.The risk warning system logical architecture and technical architecture are designed to support big data analysis.Ultimately,the potential risk information is translated into a risk warning directive.(4)HSE management system audit data and safety production standardization review data are used as empirical application objects.The practicability of early warning mode,the accuracy of early warning method and the adaptability of data analysis are analyzed.The positive effect of this method in improving the utilization efficiency of large data of potential accidents and the efficiency of risk early warning has been proved.