June 5, 2025: International Conference on Structural Safety and Reliability

Dr. Kenawy presented her research on adaptive sequential experimental design of regional-scale earthquake simulations at the 2025 International Conference on Structural Safety and Reliability.

Adaptive Optimal Design of Regional-Scale Physics-Based Earthquake Scenario Simulations

Regional-scale physics-based earthquake simulations have undergone significant advances over the past decade, creating new opportunities for engineering assessments of the consequences of large and rare earthquake events. However, quantifying the uncertainties associated with the seismic risks estimated using high-fidelity regional-scale simulations is challenging due to their significant computational costs, which limit the number of feasible simulation runs. As a result, identifying representative seismic scenarios is typically determined by expert judgement and does not guarantee reliable estimation of engineering risk measures (for example, structural damage states to infrastructure portfolios). Therefore, engineers and stakeholders need rigorous methods to optimize the acquisition of data from expensive regional-scale simulations and reduce the uncertainty in the predicted engineering outcomes, i.e., to determine the most “informative” regional-scale simulations to conduct. This study explores the use of modern experimental design techniques to optimize the design of computational earthquake simulation-based experiments, for the purpose of maximizing the information gain from a budget-constrained number of regional-scale physics-based simulations. An adaptive sampling technique is integrated with a surrogate modeling approach to iteratively determine the “the next most informative simulation” for a given computational budget and target engineering demand parameter. The proposed technique is used to perform a preliminary analysis of the uncertainty in the seismic risks to infrastructure across the San Francisco Bay Area in California due to a magnitude 7 rupture on the Hayward fault. The study demonstrates that optimal experimental design based on adaptive sampling may offer an effective tool to maximize the predictive power of physics-based regional-scale risk simulations, and optimize the allocation of computational resources.