Title: Synthesis of representative wave of spectrum-fitting input motions based on iterative learning procedure

Authors: Takashi Miyamoto; Riki Honda

Addresses: Department of Civil and Environmental Engineering, Faculty of Engineering, University of Yamanashi, 4-3-11, Takeda, Kofu, Yamanashi, 400-8511, Japan ' Department of International Studies, Graduate School of Frontier Sciences, The University of Tokyo, 5-1-5, Kashiwanoha, Kashiwa, Chiba, 277-8563, Japan

Abstract: In the synthesis of input motions for seismic design based on the specified response spectra, Fourier amplitude is elaborately adjusted, while little attention is paid to the time characteristics, although differences in the phase characteristics can cause different nonlinear response of structures. To consider this problem, we propose a scheme to synthesise a wave that represents a set of collected spectrum-fitting ground motions, using an iterative learning scheme to integrate their respective characteristics. In the learning process, the characteristics of the wave are updated by adjusting its time-frequency characteristics using wavelet transform. The learning efficiency is monitored by comparing the characteristic effects of the waves on structures before and after the update, at which point the difference is quantified via Kullbuck-Leibler divergence. Numerical results show that synthesised waves cause comparatively large responses on various nonlinear structural members of a target structure, which indicates the effectiveness of the proposed scheme.

Keywords: seismic design; structure-specific design; spectrum-fitting ground motions; nonlinear response; dynamic response; data-driven approach; iterative learning; wavelet transform; Kullback-Leibler divergence; earthquakes; structural design.

DOI: 10.1504/IJEIE.2016.080037

International Journal of Earthquake and Impact Engineering, 2016 Vol.1 No.1/2, pp.159 - 173

Received: 02 Aug 2016
Accepted: 27 Aug 2016
Published online: 29 Oct 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article