Construction And Application Of Early Warning Model Based On Accident Risk Analysis

With the deepening of safety production informatization in petrochemical enterprises,how to improve the ability of safety production data analysis in order to intelligently predict the occurrence rules of early warning accidents and improve the ability of early warning decision-making has become a crucial research topic.However,there are few researches on specific accident analysis and early warning in China.Therefore,based on the risk data analysis of petrochemical enterprises,this paper makes a systematic and in-depth study on the quantitative prediction of safety risk and the cumulative effect analysis of risk objects,and forms the following research results,which will develop and innovate the safety risk early warning and risk identification of petrochemical enterprises in China.The main contents and conclusions are as follows :(1)For petrochemical companies with more risk sources and complex risk-causing objects,the TF-IDF algorithm and naive Bayes model are introduced to realize petrochemical safety risk prediction for risk analog identification,risk value prediction,and risk level Based on the data of 10,552 enterprise-level risk lists of a petrochemical company from 2018 to 2020,conduct empirical analysis and research on risk evaluation and early warning,and predict relevant information of existing risk sources based on the learning of historical data.The risk category has an intuitive understanding,which reduces the serious consequences caused by the uncertainty of the risk value.(2)When the identified object involves multiple risks and the cumulative effect of multiple risks needs to be considered,the overall risk is calculated by taking the three aspects of safety/health impact,property loss,and non-financial and social impacts.The risk value is relatively large.This paper establishes an evaluation index system for accident risk value of chemical enterprises,selects 10 second-level indicators and 36 third-level indicators,and introduces AHP-based group decision-making to calculate weights.The comparison of algorithms proves that this method can effectively reduce The inaccuracy of the estimated risk value.(3)Designed and produced an early warning model system based on accident risk analysis based on the research methods proposed in Chapters 3 and 4 of the paper.The current PHAMS platform has integrated and improved risk identification and risk evaluation,and weight-based cumulative frequency methods.Algorithms to support the risk warning platform’s efficient analysis of big data on hidden dangers of accidents,achieving intelligence and accuracy.