Wavelet Analysis on Detecting Pulse-Like Earthquakes

A quantitative approach for identifying pulse-like ground motions is proposed herein. It is based on the use of the wavelet transform which has the peculiarity to detect sudden jumps in time histories by separating the contributions of different levels of frequency. Moreover, it has the advantage of low computational cost. Three different wavelet-based signal processing procedures are considered here in order to detect large pulses in near-fault ground motions. The first one is based on the direct decomposition of velocity time histories in frequency level and has been exploited elsewhere in the scientific literature. The other two are introduced here and take into account energy and power spectra. It is shown that wavelet analysis of the energy allows one to put in evidence even pulses that can be hardly recognized in the analysis of velocity time-histories. The proposed procedure permits also to distinguish the various energy contributions in different frequency ranges. By analyzing the wavelet coefficients, in fact, it is possible to verify if the mechanical energy release rate associated with a certain earthquake is due to a few severe events or to a series of 'small' events. It is also possible to evidence the frequency contents of a specific pulse (let say the one with highest amount of energy and corresponding power), isolating its analysis from the rest of the ground motion. © 2008 American Institute of Physics.