Risk Assessment Of Drilling Platform Emergency Evacuation Based On Bayesian Network Learning

As an important equipment for the development of offshore oil and gas,due to the bad environment,and there are a lot of leaks and flammable materials on the platform.In case of an ignition source,causing accidents such as fires and explores.Emergency evacuation plays an important role as the last important barrier in protecting the life and safety of personnel after the accident.In the process of emergency evacuation,there are many uncertainties factors,of which human and organizational factors play a decisive role.Therefore,this paper focuses on human factors,and uses dynamic Bayesian network to conduct probabilistic early warning research on emergency evacuation on drilling platform,and analyzes the influence of human and organizational factors on emergency evacuation process to prevent the emergency evacuation occurred in the process of secondary accidents.This paper takes a drilling platform as the research object,relying on the independent innovation research project plan: Based on the virtual simulation technology,the deep water drilling platform fire emergency decision support system(17CX02062),prediction of success probability for drilling platform emergency evacuation.First of all,this paper statistical analysis the details of the evacuation in fire and explosion accidents on drilling platform in the Gulf of Mexico area,from 2003 to1016.According to the statistical results,the factors affecting the emergency evacuation of the drilling platform are preliminarily screened.In order to build a more comprehensive index system of factors,we have initially screened out as many influencing factors as possible.However,there may be collinearity between the indicators.Too many indicators may cause redundancy.Therefore,the correlations between indicators were analyzed by SPSS analysis software.The unnecessary indexes were removed.We establish the index system more perfect.Second,the Bayesian network is usually constructed by using expert knowledge and subjective experience.In order to avoid the limitation of subjectivity and fuzziness of the constructed model,this paper uses the Bayesian network structure learning and combines the relevant data to build a prediction model.After constructing the model,the parameters in the model are learned by Bayesian network parameters learning.In this way,the dependence of the model parameters directly from the expert knowledge is avoided.To get model parameters quickly and accurately,the reasoning engine is also referenced.Finally,the probability prediction model of successful evacuation of drilling platform is established based on human factors and organizational factors by using Bayesian network structure learning and parameter learning.Bayesian network theory and Markoff model are used to establish a dynamic prediction model for the probability of emergency evacuation of drilling platform.Based on the causal reasoning of dynamic Bayesian network,the probability of successful evacuation of drilling platform at different time is obtained.Based on diagnosis reasoning and sensitivity analysis of dynamical Bayesian network,we studied the influence of human and organizational factors on emergency evacuation of offshore drilling platform.The most influential factors in the emergency evacuation process are determined.It provides a reasonable basis for the establishment of a more perfect emergency evacuation system on the offshore drilling platform.