ABSTRACT
Due to the potential of strong aftershocks to cause further damage to mainshock-damaged structures, accurate evaluation of the structural performance during a real seismic sequence is essential. Because of the deficiency of such real data, the necessity for developing artificial aftershock accelerograms consistent with the real mainshock accelerogram is indisputable. This article presents a new method for generating aftershock spectrum, using an artificial neural network. In this procedure, a multilayer perceptron neural network is used to create a network, whose input is response acceleration spectrum of the mainshock and the output is response acceleration spectrum of the aftershock corresponds to the mainshock with equal hazard level. To train this network, 126 seismic sequences recorded on soil type C are used. Results show good agreement between the response acceleration spectrum derived from the artificial neural network and the real aftershock spectrum. In the next step, a simple approximate logarithmic equation is proposed to calculate the aftershock strong-motion duration in terms of the mainshock strong-motion duration, and the ratio of mainshock-aftershock intensity. Finally, with the presence of the response acceleration spectrum and the aftershock strong-motion duration, accelerograms consistent with this aftershock artificial spectrum is extracted using the wavelet method.