Influence of constitutive model parameter uncertainty on fragility estimates: Application to a Chilean highway bridge

Uncertainty quantification and propagation are central components of Performance-Based Earthquake Engineering (PBEE). Although seismic hazard is widely recognized as a dominant source of variability in structural response, comparatively less attention has been paid to how constitutive model parameter uncertainty propagates through demand, capacity, and limit-state definitions to the fragility of individual structures. This work investigates the propagation of constitutive model parameter uncertainty within the PBEE framework using a representative Chilean highway bridge as a case study. The study employs, nonlinear finite element modeling with uncertain constitutive parameters, and cloud fragility analysis with a suite of 211 unscaled ground motions selected following site-specific probabilistic seismic hazard analysis spanning return periods from 30 to 5000 years. Results show that constitutive parameter uncertainty, propagated through both structural demand and component capacity estimates, can substantially affect component fragility predictions. Extreme cases show a range between 20% and 90% probability of exceeding a certain limit state, at a given intensity. Finally, global sensitivity analysis is performed, leveraging efficient simulation with Gaussian process surrogate models to compute total Sobol indices. The surrogate models show a R 2 score over 0.9 for all components, while contributing less than 5.5% to the total variance in fragility estimates due to posterior uncertainty. These indices provide insight into the composition of uncertainty in fragility estimates in terms of the relative contribution of individual model parameters and their interactions. For the case-study bridge, only a small subset of parameters is found to dominate the induced uncertainty at higher intensity levels, with elastomeric bearings governing component interactions, thus offering guidance for prioritizing future experimental data collection campaigns.