Detection and extraction of velocity pulses of near-fault ground motions using asymmetric Gaussian chirplet model

Abstract

In this paper, a method which is based on the asymmetric Gaussian chirplet model (AGCM), adapted dictionary-free orthogonal matching pursuit (ADOMP) algorithm, and Newton method, is proposed for detection and extraction of velocity pulses. In the proposed method, the ADOMP algorithm and the Newton method find the most correlating AGCM atom at each iteration and optimize its amplitude over all of the already chosen atoms. The first chirplet atom detects the location of velocity pulse and its combination with adjacent atoms extracts the velocity pulse of near-fault ground motions. The proposed method simulates asymmetric waves, sharp and impulsive peaks of velocity pulses, successive positive or negative half cycles of velocity pulses, and multiple dominant frequency peaks of velocity pulses. Also because of optimized energy concentration of AGCM atoms, they extract high energy locations of velocity time histories accurately. The previous methods underestimate sharp peaks of velocity pulses, overestimate their smooth peaks, and consider a dominant frequency for each velocity pulse. According to the results, considering the dominant frequency of first chirplet atom as the pulse's frequency is not realistic. In the proposed model, the first chirplet atom extracts more than 70% of the energy of velocity pulses. So, it is sufficient to estimate the parameters of first chirplet atom as a function of seismological parameters to can generate synthetic velocity pulses for future seismic hazard scenarios.