Quantitative classification of near-fault ground motions using wavelet analysis

Abstract A method is described for quantitatively identifying ground motions containing strong velocity pulses, such as those caused by near-fault directivity. The approach uses wavelet analysis to extract the largest velocity pulse from a given ground motion. The size of the extracted pulse relative to the original ground motion is used to develop a quantitative criterion for classifying a ground motion as “pulse-like. ” The criterion is calibrated by using a training data set of manually classified ground motions. To identify the subset of these pulselike records of greatest engi-neering interest, two additional criteria are applied: the pulse arrives early in the ground motion and the absolute amplitude of the velocity pulse is large. The period of the velocity pulse (a quantity of interest to engineers) is easily determined as part of the procedure, using the pseudoperiods of the basis wavelets. This classification approach is useful for a variety of seismology and engineering topics where pulselike ground motions are of interest, such as probabilistic seismic hazard analysis, ground-motion prediction (“attenuation”) models, and nonlinear dynamic analysis of struc-tures. The Next Generation Attenuation (NGA) project ground motion library was processed using this approach, and 91 large-velocity pulses were found in the fault-normal components of the approximately 3500 strong ground motion recordings considered. It is believed that many of the identified pulses are caused by near-fault directivity effects. The procedure can be used as a stand-alone classification criterion or as a filter to identify ground motions deserving more careful study