Seismic performance assessment of a retrofitted pile-supported wharf considering soil-cement uncertainties using artificial neural network

Abstract

Seismic fragility analysis is an effective method to evaluate the seismic performance of retrofitted wharf systems affected by the uncertainty of soil-cement strength. Nevertheless, fragility analysis usually consumes a large consumption of computational power. In this study, seismic fragility analysis using the artificial neural network (ANN) for the retrofitted wharf, considering the aleatory uncertainty of soil-cement strength and the epistemic uncertainty of the ANN, is carried out; On this basis, the fragility surface for two types of damage limit states considering the uncertainty of soil-cement strength is obtained. It was found that: (1) overall, the soil-cement strengthening strategy is effective for improving the seismic safety of wharf systems, however, the strengthening effect is limited, especially under strong earthquakes, will be further weakened; (2) ANN can effectively predict the maximum seismic response of retrofitted pile-supported wharves, so as to quickly carry out seismic fragility analysis. Examples show that the prediction method has good generalization; and (3) the fragility surface model considers the aleatory uncertainty of soil-cement strength and the epistemic uncertainty of the ANN, which makes the performance-based evaluation of retrofitted pile-supported wharves more comprehensive.