Artificial Ground Motions using Hibert-Huang Transform and Artificial Neural Network

In this paper, an artificial ground motions generation method is proposed based on the radial basis function neural network (RBFNN). The inputs of the RBFNN are the pseudospectral velocities (PSVs) of the ground motions and the outputs are the frequency information and amplitude information obtained using the Hibert-Huang transform (HHT). Because the frequency and amplitude are predicted independently, the prediction is more accurate and efficient. In order to evaluate the accuracy and efficiency of these artificial ground motions, a five-storey steel frame is modeled and the seismic responses of the structure are obtained. The results show that the artificial ground motions predict the same seismic responses as the original ones..