A probabilistic-statistical hybrid model for dominant pulses in pulse-like ground motions (PLGMs)

Abstract

Pulse-like ground motions (PLGMs) have been shown to significantly amplify the nonlinear seismic demands of structures, critically impacting seismic design and risk assessment. The effective simulation of pulse characteristics in PLGMs is essential for achieving these purposes. In this study, a probabilistic-statistical hybrid model optimized by genetic algorithms is proposed to efficiently simulate the dominant pulses of PLGMs. A comprehensive database consisting of 194 PLGMs from 36 global earthquakes was established, enabling detailed analysis of the correlations between the simulation model parameters and the real pulse in records. By utilizing Mavroeidis’ model, probabilistic-statistical models for the pulse period (T_p) and pulse amplitude (V_p), along with probabilistic distribution models for the wave shape parameter (γ) and phase parameter (φ), were systematically developed. This provides a comprehensive framework for characterizing pulses in PLGMs. Nonlinear structural analysis further demonstrates that the synthetic records generated by the proposed hybrid model effectively reproduce structural responses, enhancing their applicability in seismic design and risk assessment within engineering applications.