Dynamic risk assessment and sensitivity analysis of lightning-induced Natech tank fire based on probabilistic chain model and Bayesian network

Abstract

Lightning-induced Natech events are the key triggers of chemical storage tank fires, which require comprehensive consideration of the multi-factor dynamic coupling of the environment, equipment and hazard chain for risk assessment. In this study, a risk assessment procedure integrating lightning probabilistic chain model and Bayesian network is proposed. Firstly, the whole-process probabilistic chain of “lightning-failure-ignition-propagation” is constructed to quantify the risk of fire accidents based on the IEC 62305 lightning strike model, tank vulnerability analysis and domino effects. Secondly, a Bayesian network model of lightning-induced Natech event is established, and the key sensitive parameters are identified by variance decomposition and univariate perturbation method, which reveals the dominant roles of “Primary tank failure” and “Lightning strike on tank”. In addition, the layered integral cylinder flame model established can be well adapted to the calculation of heat radiation in the terrain difference scenario. The study shows that well-equipped lightning protection can reduce the probability of lightning strike by 22.5 %, and increasing the thickness of tank wall can reduce the failure probability by up to 4.2 %. Finally, risk visualization of critical accident stages based on human vulnerability models and risk acceptability criteria, supports emergency management and optimization of multi-level protection for tank farms in complex terrain.