Fire probability prediction of offshore platform based on Dynamic Bayesian Network

Abstract

By integrating Markov model with Bayesian Network, an improved Dynamic Bayesian Network (DBN) model is proposed in this paper. It could be used to predict the dynamic probability of offshore platform fire. It is a new stochastic model which could process dynamic data. Three innovations are achieved in building this model. Firstly, a modified Human Factor Analysis and Classification System (HFACS) model is used to provide guidance for the construction of BN in describing the sequence of causes for offshore platform fire. Secondly, the prior probability is determined by using Evidence Theory model based on historical data. Thirdly, the conditional probability table of basic events is calculated by integrating Fuzzy Analytic Hierarchy Process (AHP) with Hierarchical Node Distance Formula. Lastly, the transition probability of BN is obtained based on Markov model. There are three major functions of the new model. First of all, through forward reasoning, the dynamic probabilities of platform fire at different time are obtained. Second, based on the diagnosis reasoning, the impact of human and organizational factors on the fire is researched. Last, the sensitivity analysis and uncertainty analysis of human, organizational and equipment factors are also researched to identify the significant causes and to quantify the effect.