Speaker: Mojtaba Harati, Ph.D.
When: Friday, October 17, 9:30AM
Where: KH A510 (Lincoln) and PKI 160 (Omaha)
Zoom*: 937 3450 0302
*students enrolled in CIVE 990R need permission to attend via Zoom
Title: The g Function: A Machine Learning Framework for Synthesizing Multi-Hazard Fragility Surfaces
Abstract: This study introduces the g function—a data-driven machine learning framework developed to synthesize multi-hazard fragility surfaces representing the joint vulnerability of structures to sequential or interacting hazards such as earthquakes and tsunamis. Traditional fragility modeling approaches rely on extensive nonlinear simulations for each hazard combination, which are computationally prohibitive. The proposed g function maps the interdependencies between boundary fragility curves and their corresponding damage probabilities across multi-dimensional intensity measure (IM) spaces, enabling efficient reconstruction of complete fragility surfaces without retraining or re-simulation. The framework is validated using high-fidelity structural analyses for reinforced concrete frames subjected to earthquake–tsunami sequences and further extended to other hazard and event combinations, including mainshock–aftershock sequences and fire-following-earthquake scenarios in urban wildfire contexts. More recently, it has also been applied in industry for wind–surge joint modeling under hurricanes. Results show excellent agreement (R² > 0.95) with direct simulation benchmarks. Beyond computational efficiency, the g function serves as a versatile synthesis tool for applications ranging from structural damage assessment and loss estimation at the community scale to integrated recovery and resilience analyses within multi-hazard environments.