Fragility analysis of degraded structures considering soil-structure interaction
The fragility curve plays an important role in estimating the structural vulnerability under an earthquake disaster. The zone of the most interest is the fragility curve where failure is rarely occurs. The classical approach is limited by a considerable quantity of epistemic uncertainty, due to the lognormal hypothesis. A second approach, the nonparametric one overcomes this problem. However, it requires Monte Carlo Simulations to estimate the probability of exceedance of the limit state, which can be computationally intensive, to estimate the interest zone of the fragility curve. To overcome this drawback, this work aims to reduce the number of nonlinear response evaluations of the pushover analysis, to estimate the fragility curve by using surrogate modeling updated by adding new samples that best represent the system response, also known as Active Learning Approach. This paper proposes a methodology for calculating failure probabilities of the seismic fragility curve with a considerable reduction of the computational cost.