Modeling the evolution of industrial accidents triggered by natural disasters using dynamic graphs: A case study of typhoon-induced domino accidents in storage tank areas

Technological accidents triggered by natural disasters (Natech) have become a significant threat to the safety of coastal energy infrastructure. The domino accidents involved could propagate rapidly in a short duration and amplify an accident exponentially. In previous studies of Natech-related domino accidents, the scope of natural disasters was mainly focused on the phase of primary events, and the subsequent accident cascades were entirely driven by the domino effect. This simplification essentially weakens the real-time intervention of ongoing disasters in the evolution of accidents, making it difficult to accurately model the evolutionary patterns of domino accidents. Accordingly, a Natech-related domino evolution graph (NT-DEG) is proposed in this paper for dynamic modeling of accident evolution under the real-time disturbance of natural disasters. In addition, a stochastic process and a deterministic process are innovatively used to control the generation of primary events and the update of the evolution network at each timestamp. Moreover, using a dynamic community detection algorithm and graph metrics, an identification method for critical units in dynamic domino accidents is proposed. The application of NT-DEG to typhoon-related Natech indicates that continuous-onset natural disasters may cause new failure units that could participate in the original domino accident at any moment. This real-time intervention noticeably changes the propagation path and scale of the original accident and accelerates accident evolution. In addition, the validation based on the integrated simulation reveals that the proposed identification method can accurately identify the critical units, and the safety measures against them can noticeably reduce the scale of accident cascades.