In the context of global warming and sea-level rise, typhoons and multiple typhoon-induced disasters are gradually evolving into important threats to coastal infrastructure, energy facilities, and lifeline systems. Typhoon-induced floods, strong winds, and heavy rainfall will further trigger industrial accidents, non-linearly amplify the consequences of accidents, and bring unprecedented disasters to life, property, and the natural environment, especially as coastal industrial areas are concerned. The current research on typhoon focusing on technical disasters (belonging to the category of Natech events) is still in its infancy. The mechanism associating primary disasters to secondary (or derived) multiple hazards and the mechanism behind multiple hazards that trigger industrial equipment catastrophes are unclear, the disaster chain effect is difficult to quantify, and there is a lack of a systematic risk assessment framework that includes the transformation among various hazards. Accordingly, this thesis developed a numerical synthesis model of typhoon-induced multi-hazard scenarios to reveal the coupling pattern between floods, strong winds, and heavy rainfall, focusing on the Natech events triggered by the typhoons. The critical conditions of multiple hazards triggering the catastrophes of industrial accidents were further explored based on the synthesized data, and a network-based model was built to systematically elaborate the mutual transformation mechanism among hazards. Finally, this thesis developed a quantitative risk assessment method for typhoon-triggered Natech events based on the coupling patterns of natural disasters, the mechanisms behind Natech events, the evolution and escalation model of technological disasters, and the disaster chain effect of Natech events. The research content and results of the study are described below.

(1) This study synthesized multi-hazard scenarios induced by typhoons and built a joint pattern of the multi-hazard system. Thus, a numerical synthesis model of global-scale tropical cyclones and their induced multi-hazards was developed by joining the statistical dynamic typhoon model with the ADICRC storm surge model. This model integrated typhoon origin and secondary-derived disaster prediction and provided a database with high statistical consistency to study typhoon-induced Natech events. The statistical analysis of synthetic multi-hazard scenarios showed a high nonlinear dependence among the multiple typhoon-induced hazards. The Vine copula approach can accurately capture the joint distribution of multi-hazard systems, providing theoretical support to calculate the frequency and intensity of typhoon-induced multi-hazards.

(2) This study quantified the multi-hazard risk of industrial equipment and designs, and how they relate to typhoon-induced multi-hazard events. The critical conditions for multi-hazards to trigger industrial cataclysms were revealed by deriving the mechanical equilibrium and the control equations of limit equilibrium states of the industrial equipment under wind-flood composite loads. At the same time, a probabilistic risk model for typhoon-triggered failures of industrial equipment, which provides a probability basis for the risk quantification of Natech events, was developed considering the joint distribution of multi-hazards and the fragility of industrial equipment to multi-hazards. In addition, a risk-driven design framework of a multi-hazard event was established to provide a reliable reference index of multi-hazard joint intensity to prevent the disaster and mitigate its effects on coastal industrial areas.

(3) This study built an evolution and escalation model for typhoon-triggered industrial accidents. The module of the natural disaster scenario is expanded to systematically describe the transformation and energy transfer that occurs among hazards in various links of typhoon-triggered industrial cascade accidents (domino effect) in the traditional Natech event assessment framework. The Natech-related domino-effect network model of accident was defined, and the network learning technology was used to reveal the evolution path and escalation mode of the initial accident within the industrial cluster. Therefore, the network-based safety analysis provided strong support to interrupt accident propagation and escalation.

(4) This study evaluated the disaster chain effect of typhoon-induced landslide industrial accidents. The chain risk of typhoon-derived landslide-triggered industrial accidents was decomposed into three sub-modules: slope instability, landslide-triggered industrial accidents, and active safety measures. A hybrid Bayesian network was used to connect the risk factors in each link, conducting the step-by-step spread of the Natechs chain risk. The Bayesian network model pointed to rainfall, shallow soil depth, slope angle, and distance from the plants as the key factors in determining the risk of a disaster chain. In addition, the Bayesian-based reverse inference optimizes the design parameters of the drainage system for the peripheral slope of the industrial cluster.