The purpose of this work is to outline a framework for assessing the resilience of a petrochemical storage plant, through the construction of a dynamic hierarchical Bayesian network. The BN approach allows keeping memory of the states, in order to manage the actual safety and reliability evidences during the petrol transfer operation from storage tank to trucks in a repository of oil products. The proposed framework aims at assessing risk in process plants by analysing continuous process hazard data from a Bayesian point of view. A sequence of hazard functions derived for the FTAs, is modelled with a hidden Markov chain. The capability of the model implemented by means of Markov Chain Monte Carlo methods are tested at a real scale plant.
Keywords: data driven model, hidden Markov models, resilience, semi-supervised learning,.