An evolutionary stochastic ground-motion model defined by a seismological scenario and local site conditions

This paper is concerned with modeling earthquake-induced ground accelerations and the simulation of the dynamic response of linear structures through the principles of stochastic dynamics. A fully evolutionary approach, with nonstationarity both in amplitude and in frequency content, is proposed in order to define the seismic action, based on seismological information in the form of a small number of input parameters commonly available in deterministic or probabilistic seismic design situations. The signal is obtained by filtering a Gaussian white-noise. The finite duration and time-varying amplitude properties are obtained by using a suitable envelope function. By utilizing a subset of the records from the PEER-NGA strong-motion database, and time-series analysis tools extended to nonstationary processes, the key transfer-function properties, in terms of circular frequency, damping ratio and spectral intensity factor, are identified. A regression analysis is conducted for practical and flexible application of this model, in order to empirically relate the identified time-varying parameters of the filter to the characteristics defining earthquake scenarios such as magnitude, rupture distance and soil type. A validation study and a parametric investigation using elastic response spectra is also included. Results show that the final seismic model can reproduce, with satisfactory accuracy, the characteristics of acceleration records in a region, over a broad range of response periods.