Adaptive Decomposition and Instantaneous Frequency-Based Identification of Multipulse Sequences in Near-Fault Ground Motions

초록

Velocity pulses significantly influence structural responses. Although several algorithms have been developed to extract a single pulse from a record, the extraction of multipulse sequences has not been fully investigated. This article presents a method that integrates adaptive chirp-mode pursuit for extracting nonlinear components along with ridge detection algorithms to determine the locations and durations of multiple pulses from near-fault ground motions. The approach emphasizes narrowband frequency content for each pulse, successfully distinguishing low-and high-frequency components. This method does not impose any assumptions about waveform symmetry, number of cycles, or frequency shape. Instead, it identifies energy ridges based on local maxima and continuity across the time axis. The results indicate that the proposed algorithm can successfully separate colocated velocity pulses and identify those at different times. Visual inspection of multipulse records leads to the introduction of an initial metric for identifying secondary pulses. To assess the significance of frequency content, a frequency-domain pulse indicator is developed for clustering multipulse records, with its threshold optimized based on visual clustering assessments. This pulse indicator is highly sensitive to pulse intensity and strongly correlated with time-domain features. A statistical analysis shows a direct correlation with moment magnitude, alongside an inverse relationship with the main pulse’s contribution to the total energy of the record. Prediction equations for the periods of main and secondary pulses, as functions of moment magnitude, were developed, revealing that the slope of the prediction line for the second pulse is 14% less steep than that for the main pulse. The proposed prediction model outperforms most existing models in extracting long-period components. On average, the second pulse contributes 50% less to the energy and 35% less to peak ground velocity than the main pulse. The proposed algorithm can enhance understanding of pulse sequences and improve estimates of seismic demand in near-fault regions. © Seismological Society of America.

키워드