Quantifying the effect of ground-motion nonstationarity on the nonlinear dynamic responses of an ultra-high arch dam

Abstract

Nonstationarity of a ground-motion time history refers to the changing frequency features over time. The frequency nonstationarity can greatly affect nonlinear structural responses, but is difficult to be captured realistically by current stochastic simulation methods. In this study, the ECSC-rho method is developed to generate synthetic earthquake ground motions through wavelet-packet characterization and optimization. Time and frequency properties of acceleration time histories are characterized using wavelet-packet analysis. The correlation between wavelet-packet coefficients in time and frequency domains is used as a simple proxy to represent nonstationarity in an acceleration time history. The basic idea is to adjust the wavelet-packet coefficients of a seed motion in both time and frequency domains to reconstruct an accelerogram, such that it matches a target spectrum, Arias intensity and nonstationarity feature. Compared with existing spectrum-matching techniques, the ECSC-rho method has distinctive advantages in simulating ground motions with prescribed nonstationarity, response spectra and energy build-up processes. The performance of the proposed ECSC-rho method is systematically validated using an ultra-high arch dam model, which incorporates highly nonlinear dam responses such as the opening of contraction joints and cumulative concrete damage. Extensive numerical simulations will highlight the effects of ground-motion nonstationarity on nonlinear dam responses, and provide practical guidelines for using spectrally-matched synthetic motions in dynamic dam analysis.