Browse/search for people

Publication - Dr Flavia De Luca

    Automated classification of near‐fault acceleration pulses using wavelet packets


    Chang, Z, De Luca, F & Goda, K, 2019, ‘Automated classification of near‐fault acceleration pulses using wavelet packets’. Computer-Aided Civil and Infrastructure Engineering, vol 34., pp. 569-585


    This study proposes a new algorithm for automatically classifying two types of velocity‐pulses that are integral of a distinct acceleration pulse (acc‐pulse) or a succession of high‐frequency one‐sided acceleration spikes (non‐acc‐pulse). For achieving this, wavelet packet transform is used to filter the high‐frequency content and to extract the coherent velocity‐pulse. Then, the pulse period is unequivocally derived through the peak point method. Following the determination of the pulse‐starting (ts) and pulse‐ending (te) time instants in the velocity time‐history, a local acceleration time‐history truncated by ts and te is obtained. The maximum relative energy of the pulse between two adjacent zero crossings is then employed as indicator for distinguishing the two types of velocity‐pulses. The criteria for identifying acc‐pulses and non‐acc‐pulses are calibrated using a training data set of manually classified ground motions from the Next Generation Attenuation West 2 project. Finally, significance of such a classification between velocity‐pulses of different characteristics is assessed through the comparison of elastic acceleration response spectra of the two categories of pulse‐like records.

    Full details in the University publications repository