Open Collections

Kourehpaz, Pouria

University of British Columbia

This dissertation endeavors to advance seismic risk assessment methodologies for building structures through better characterization of the hazard and accounting for key sources of uncertainty within risk assessment models. The assessments are carried out at two scales, i.e., individual buildings and building portfolios, and focus on a range of modern reinforced concrete shear wall archetype buildings in Seattle, WA. At the individual building level, this dissertation incorporates the impacts of deep sedimentary basins into performance-based seismic assessments by quantifying earthquake-induced economic loss and downtime of buildings with different strengths, stiffness, and heights. These analyses can inform the engineering design community of the tradeoffs of adopting different design strategies when dealing with deep basin amplification. Considering the significant uncertainty in individual building risk assessment results, this dissertation conducts a probabilistic sensitivity analysis on loss and downtime estimates to highlight the significance of parameter choices (e.g., fragility functions) in risk outputs. At the building portfolio level, this dissertation employs advanced seismic fragility and machine learning-based models to enhance the accuracy of damage predictions and seismic losses. To this end, this dissertation develops multivariate building-taxonomy level fragility functions conditioned on advanced ground motion intensity measures (e.g., average spectral acceleration) to enhance the accuracy of seismic damage and loss estimates at a regional scale. The results indicate that the seismic loss prediction performance is significantly improved by employing multivariate fragility functions (even with only three variables) compared to univariate functions. Furthermore, this dissertation proposes a machine learning-based framework to predict a building’s anticipated earthquake-induced damage state by accounting for variability in building-to-building structural properties and ground motion shaking intensities. A separate predictive model is developed to conduct collapse risk assessments by integrating synthetic data samples into the original dataset. This framework shows great potential for enhancing building portfolio seismic risk assessments by leveraging building-specific input features and synthetic data samples for rare structural damage state instances. The outcomes of this dissertation can be used to enhance our understanding of the fundamental principles and procedures used in assessing seismic risk, which, in turn, can inform policy decisions to mitigate seismic risk and enhance resilience.

Text; Dataset

2024-01-15

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/87274

Civil Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/