Characterization and Representation of Near-Fault Ground Motions Using Cumulative Pulse Extraction with Wavelet Analysis

This paper investigates the presence and characteristics of multiple pulses (with dominant period T-P between 0.5 and 12 s) in historical near-fault ground-motion records. An iterative method for extracting multiple strong pulses embedded in a ground motion is presented. The method results in a representation of the pulse-like part of the ground-motion record composed of the sum of the extracted pulses. Each of the pulses is identified via wavelet analysis using the M&P wavelet. The method is used to extract multiple strong velocity pulses from the fault-normal horizontal component of 40 pulse-like ground-motion records from 17 historical earthquakes, with magnitudes ranging from M-w 6.3 to 7.9, recorded at a distance less than 10 km from the fault rupture with a peak ground velocity greater than 0: 6 m/s. Thirty-five records are found to have more than one predominant pulse, and 36 records have at least one pulse with 1 s <= T-P <= 4 s. The pulse representations (using one or more pulses) of the ground-motion records are evaluated by comparing the linear single-degree-of-freedom displacement spectra for periods ranging between 0.5 and 10 s. The effect of the following parameters of the pulse extraction method on the extracted pulses and resulting representation of the records are studied: (a) the number of extracted pulses (varying from one to three) used in the representation of the record; (b) the type of time domain (velocity or acceleration time history of the motion) at which the analysis is conducted; and (c) the three weighting functions used in the wavelet analysis. Finally, relationships between the dominant period of the extracted pulses, associated amplitudes, and earthquake magnitude are presented, indicating that the amplitude of the strongest pulses (with 1.5 s <= T-P <= 5 s) does not depend significantly on the earthquake magnitude.