Combination of LS-SVM algorithm and JC method for fragility analysis of deep-water high piers subjected to near-field ground motions

Abstract The objective of this paper is to propose a novel fragility analysis method pertinent to the combination of the least squares support vector machine (LS-SVM) algorithm and JC method for the vulnerability estimation of deep-water high piers located in reservoirs under near-field ground motions. First, the uncertainties associated with the design parameters are considered and the LS-SVM algorithm is adopted to predict the cross-sectional critical curvature values of each damage state of the example pier and determine the probability distributions. Then, a group of finite element models of the example pier is developed using the OpenSees platform for considering the uncertainties related to the design parameters and near-field seismic waves; the Morrison equation is employed to calculate the added mass to simulate the hydrodynamic pressure. Meanwhile, the dynamical responses of the example pier with seven different water depth conditions are investigated using increment dynamic nonlinear analysis along the longitudinal and transverse directions under seismic excitations. Finally, the fragility functions are derived based on the JC method to consider the probability distributions of the cross-sectional critical curvature values and dynamical responses simultaneously; these functions are applied for assessing the damage probability at different damage states of the example pier. It can be concluded that the hydrodynamic pressure plays an important role in influencing the natural vibration period and mode shape of the deep-water high pier. In the slight, moderate, and extensive damage states, the damage probability of the middle-upper and bottom areas of the example pier is relatively large under longitudinal near-field seismic excitation; nevertheless, only the bottom is easily damaged under transverse near-field seismic excitation.