Alternate algorithm for characterization of strong velocity pulse in ground motions

Abstract

The identification of pulse-like ground motion is essential in seismic hazard analysis and structural design. The approaches for pulse characterization can be classified into two categories: signal processing based (e.g., wavelet transform, Hilbert Huang transform, variational mode decomposition, empirical Fourier decomposition) and purely time series based without any transformation. A classification boundary can be established using a set of pulse-like and non-pulse-like ground motions. Alternatively, a subjective but reproducible definition of pulse-like ground motions based on relative energy of the pulse can be explored. This paper presents an energy-based algorithm that operates in three stages: (1) detection of the strongest pulse orientation, (2) extraction of the pulse through iterative fitting of a trigonometric function; and (3) characterization of the pulse-like ground motion based on relative energy and location of the pulse. The proposed algorithm does not specify classification of thresholds and rather allows for subjective user specific thresholds depending on the application. Further, a methodology analogous to the wavelet-based approach is proposed in the time domain to estimate the pulse period. The framework is illustrated for three ground motions from the NGA West-2 database and demonstrates the characterization of non-pulse-like and pulse-like ground motions. The ambiguous ground motions are observed to be sensitive of chosen thresholds and their characterization is subject to user discretion based on the specific application.