Influence of different fully non-stationary artificial time histories generation methods on the seismic response of frequency-dependent structures

Advances in computational power and numerical optimization routines enable the application of rigorous simulation and optimization techniques for the design and assessment of structural and geotechnical (S&G) systems. Nowadays, the state of practice in the seismic design of high-importance structures has progressively moved toward the use of time-history dynamic analysis that are highly sensitive to the way of modelling the dynamic action. To be reliable, this type of analysis requires efficient methodologies for the selection of the seismic inputs used as ground excitations. In this paper, a new stochastic approach, based on the use of the wavelet transform is proposed to generate an arbitrary number of seismic records having the same characteristics of a target accelerogram. The choice of the amplitude and number of bands in which to partition the frequency domain is a key source of variability to consider for the generation of samples with the desired time-varying amplitude and frequency content. To evaluate the influence of the use of an alternative fully non-stationary artificial accelerograms generation methods, a comparison between the proposed method and the one recently proposed by the first two authors is also presented in this paper. Specifically, in order to quantify the influence of accelerograms models on the seismic response, a structural system with viscoelastic damping is analysed, representative of a broad range of frequencydependent S&G assets.