Bayesian modeling reveals how damage state improves predictive power in structural resilience assessments, indicating significant implications for disaster risk management.
Key Points
The research aims to develop fragility models that account for cumulative damage in buildings during multiple hazard events.
Developed damage state- and hazard-dependent fragility models for reinforced concrete buildings
Used Bayesian inference with Hamiltonian Monte Carlo for statistical modeling
Evaluated model accuracy based on predictive performance and information criteria
Incorporated causal directed acyclic graphs for explainability
Damage state significantly enhances predictive power of fragility models
Period elongation is identified as a key predictor of structural degradation
The type of hazard causing prior damage has minimal additional effect once period elongation is included
Models support scenario analysis for decision-making in disaster risk management