Engineering Hazard-Resilient Infrastructure through Physics-based Simulation and Machine Learning

This seminar presents recent research advances from the Hazard Mitigation & Infrastructure Resilience (HMIR) Lab at McGill. The topics covered include aftershock hazard evolution, regional seismic risk assessment, and machine learning applications in natural hazards engineering. First, using the Cascadia Subduction Zone megathrust earthquake as a case study, aftershock hazards are simulated to explicitly account for the temporal dependence and stochastic nature of mainshock–aftershock sequences. A Markov chain model is further employed to characterize the evolution of structural damage under successive aftershocks. Second, leveraging second-generation component-level seismic fragility models for California bridges, a refined framework is established to estimate component-level repair costs and functional recovery. The framework is utilized for high-resolution seismic loss assessment of the Los Angeles bridge network. The presentation further showcases recent developments in statistical and deep learning surrogate models for innovative applications in earthquake and multi-hazards engineering. Collectively, these advances enhance modelling fidelity, resolution, computational efficiency, and uncertainty quantification towards next-generation assessment and design of multi-hazard resilient infrastructure.