Domino effects are typically high impact low probability (HILP) accidents, whereby escalation effects triggered by fires are most frequent. The evolution of fire-related domino effects depends on synergistic effects and the performance of safety barriers, but those factors usually are time-dependent. In the present study, a methodology is developed to provide more accurate probabilities related to domino effects, by considering the temporal evolution of escalation vectors caused by time-dependent factors. The Dynamic Bayesian Network (DBN) approach is applied both to model the spatial-temporal propagation pattern of domino effects and to estimate the dynamic probabilities of domino chains. The methodology is illustrated with a case study to determine the dynamic aspect of the probabilities of domino effects considering the impact of add-on (active and passive) safety barriers and taking into account synergistic effects. The critical units for facilitating domino propagation have been identified by the analysis of posterior probabilities, and further validated using graph theory. The methodology will be helpful for risk management and emergency decision-making of any chemical industrial area.

A major chemical accident has the characteristics of being destructive, and potentially provoking a great loss of lives and property damage in any Chemical Industrial Park (CIP). Emergency rescue and evacuation are essential parts of emergency decision-making for enhancing the capacity and effectiveness of emergency handling and reducing the potential loss of accidents. Most of current literature concentrates on one-way route planning of emergency rescue and evacuation, and applies different models, optimization objectives and algorithms. However, when applying the one-way route planning model in a CIP, a road conflict is possible due to the inherent weak traffic capacity. Therefore, a new method of two-way route planning of emergency rescue and emergency evacuation which considers intelligent obstacle avoidance, is proposed in the paper. The method we developed integrates three modeling components: (i) a dynamic grid environment model to simulate the interaction between the road network and the time-varying location of emergency rescue and evacuation. (ii) a two-way route planning model to simultaneously optimize routes of emergency rescue and routes of emergency evacuation. (iii) an intelligent obstacle avoidance model to prevent potential road conflicts. The results illustrate that the proposed model is able to generate a set of two-way optimum routes and overcomes possible road conflicts successfully.