Probability Analysis Of Offshore Fire Based On Dynamic Bayesian Network

The offshore oil and gas industry is worldwide known for its greatest risk.As foundational facilities in the exploration and exploitation of oil and gas,the offshore platform is prone to fire accident with a high probability for harsh working environment,narrow working space,concentrating operating equipment,especially,for large of flammable and explosive substances.According to statistics,the offshore fire accident is one of the most reported accident types.Offshore fire accident poses great threat to economic,human safety and environmental protection.Therefore,the study of offshore fire probability and factors affecting fire probability plays a vital role in reducing offshore fire probability and improving the whole security of platform.Depending on the National Natural Science Foundation of China(Grant no.51409260): Offshore fire and explosion dynamic risk assessment method research based on human and organizational factors,the prediction of offshore platform fire probability is carried out.Statistical analysis of offshore platform fire accidents in the Gulf marine,from 1996 to 2015,has been carried out,including the main activities,working units,ignition types,human and organizational factor related to offshore platform fire accidents.Seventy-eighty percent of offshore platform accident are caused by human error.Based on HFACS(human factors and classification system)model,combined with the feature of offshore platform and fire accident statistical analysis,the offshore HFACS model is proposed.Dynamic Bayesian network is the most effective probability reasoning method,this paper apply dynamic Bayesian network for research of offshore platform fire probability.Conditional probability is the main parameters of dynamic Bayesian network,in this paper,a calculation method based on triangle fuzzy function and level node distance formula is proposed to calculate the conditional probability.Transition probability is the main parameters of time series in dynamic Bayesian network.The study of transition probability is carried out based on Markov model.Finally,the calculation formula of transition probability is obtained.Oil and gas processing units and manifold unit of production platform have the largest fire hazard,based on the Swiss model and HFACS model,the barrier failure model of fire accident in oil and gas processing unit and blowout fire accident are established.Furthermore,the offshore fire probability model considering coupling influence of human factors,organizational factor and technical factors is established based on dynamic Bayesian network.Through dynamic Bayesian networks forward reasoning,temporal variation law of the offshore platform fire probability is obtained.Based on diagnosis reasoning of dynamic Bayesian network,the impact of human and organizational factors on the fire probability is studied,and finally identified the sensitivity factors of human factors,organizational factors and equipment technical factors.