Generation of Multiple Earthquake Accelerograms Compatible with Spectrum Via the Wavelet Packet Transform and Stochastic Neural Networks

The principal purpose of this article is to present a novel methodology based on wavelet packet transform techniques and stochastic neural networks to generate more artificial earthquake accelerograms from available data, which are compatible with specified response spectra or the design spectra. The proposed method uses the decomposing capabilities of wavelet packet transform on earthquake accelerograms, and the learning abilities of stochastic neural network to expand the knowledge of the inverse mapping from response spectrum to coefficients of wavelet packet transform of earthquake accelerogram. This methodology results in a stochastic ensemble of wavelet packet transform coefficients of earthquake accelerograms and, they are used to the generate accelerograms applying the inverse wavelet packet transform. Finally, an interpretive example is presented which uses an ensemble of recorded accelerograms to train and test the neural network, aiming at the demonstration of the method effectiveness.     

Seismic Behavior of RC Bridge Piers under Bidirectional Excitations: Implications of Site Effects

Seismic demand of structures may be potentially magnified due to bidirectional shaking. This may also be strongly influenced by the amplification and/or attenuation of motions due to local site conditions. Current investigation examines the implications of these two aspects collectively. A code-designed reinforced concrete pier has been analyzed under recorded accelerograms and their derivatives from 1-D site analysis. Magnification in peak and cumulative responses due to bidirectional shaking may be sensitive to record-to-record variability. However, such magnification remains relatively stable across site characteristics. Combination rules are shown adequate to estimate peak deformation and deficient for cumulative demand irrespective of the site conditions.