Synthesis of Near-Fault Ground Motions Based on the Wavelet Packet Transform and BP Neural Network

Synthesis of near-fault (NF) ground motions is an important task in earthquake engineering since they are rarely recorded in historical events. This paper mainly focuses on proposing a novel method of generating NF ground motions in the direction of the strongest pulse under given seismological parameters while appropriately considering the non-stationarity especially in the frequency domain. The synthesis model consists of two parts: a low-frequency velocity pulse component constructed using the modified Gabor wavelet function and a high-frequency acceleration component generated utilizing the wavelet packet transform (WPT). Based on the collected NF records, the parameters involved in the synthesis model are recognized and related to the seismological parameters pertaining to NF ground motions using the BP network. In the process, random errors induced by different seismic events and different records within the same event are appropriately taken into account. Thereby, the whole procedure is proposed and successfully applied to generate the NF ground motions under given seismological parameters. Finally, a comparative study is carried out to further demonstrate the advantage of the method proposed in this study. The results reveal that the WPT utilized in the method can more effectively simulate the non-stationarity of NF ground motions in the frequency domain compared to existing methods. Moreover, the use of BP neural networks not only enhances the accuracy but also significantly reduces the uncertainty. The method proposed in this study can be well applicable for related studies in performance-based earthquake engineering.