An Investigation on the Accuracy of Wavelet-Based Down-Sampling in Seismic Structural Analysis

Wavelet transforms have been used for 20 years to reduce dynamic analysis costs, with studies suggesting the use of this method for three levels of decomposition. In addition, in these studies, the time step of the solution is assumed to be equal to the time step of the earthquake record and therefore the stability of the numerical method is essential in such cases. Also, in previous studies, the effects of the distance from the fault on the accuracy of the approximate waves have not been evaluated. This paper examines the accuracy of wavelet-based down-sampling methods for 50 far-field, near-field, and pulse-like earthquake records and 24 wavelet functions. Also, two numerical methods, Newmark's average acceleration and the interpolation-based method are studied. The results of this research show that Newmark's average acceleration method is suitable for wavelet-based dynamic analysis. Also, in this study, it is shown that by choosing the appropriate wavelet function, even up to four levels of decomposition may be used with an error of less than 2%. Therefore, with this work, the cost of calculations may be reduced up to 16 times.