Research On Safety Risk Assessment Of Petrochemical Enterprise In Industrial Big Data Environment

The process of petrochemical enterprises is complicated,and its sources of dangerous exist extensively.Once accidents occur,it could cause substantial economic losses or even significant casualties.Thus,risk assessment is of great importance in accident prevention and it has become a research hotspot to scientifically and reasonably carry out the assessments.The popularization of IOT technology in petrochemical enterprises nowadays has witnessed an explosion of data collected from petrochemical production,which results in so-called multi-level petrochemical-industry big data with space-time characteristics.Industrial big data has been applied in the safety domains so far,but the corresponding theory is not yet mature.Further studies shall be conducted to investigate how industrial big data can be integrated to promote risk assessment.The research contents and main contributions of this paper are as follows:(1)In order to study how to manage safety data efficiently,data definition is first discussed,and a structured approach for safety big data of petrochemical enterprises is proposed.First,the connotation and classification of safety big data are elaborated,the characteristics of all kinds of safety data are investigated and analyzed,and the structural description form and data table definition of various safety data are put forward.Then,the classic TE process is extended to a virtual factory-TE Smart Plant.Taking TE Smart Plant as an example,the process of obtaining and instantiating various types of safety data is analyzed.This chapter aims to provide ideas for petrochemical safety data management.(2)Because traditional risk analysis technology is unable to capture the dynamic risk characteristics of the process system,a dynamic risk analysis method is proposed based on Bayesian network.The method fully combines the advantages of traditional risk analysis methods and Bayesian networks,utilizing the powerful reasoning ability of Bayesian network to identify the key risk factors.Meanwhile,it can introduce industrial site safety data for risk iterative updating.Therefore,continuous learning and dynamic risk analysis are achieved.(3)A method of early warning for fire accidents based on multi-source safety data fusion is projected,including the following processes:taking the change of ambient temperature and humidity in case of fire into account,applying CUSUM algorithm to detect potential changes in environmental data distribution and generating alarm by means of feature-level fusion.With the fire probability extracted from the video surveillance data,D-S evidence theory is applied in the decision-level fusion to infer the final fire probabilities which is according to the above-mentioned multi-sourced safety information.The method can effectively reduce false alarm rate.