Modification of wavelet-based downsampling method to reduce computational error in nonlinear dynamic analysis

Abstract

One of the significant obstacles in conducting linear and non-linear time history analysis is its time-consuming nature. In this article, a new downsamlping method based on discrete wavelet transform (DWT) and smoothing is proposed to overcome this problem. In order to assess the precision of this approach, 50,000 linear and non-linear dynamic analyses of single degree of freedom (SDOF) systems and 300 nonlinear dynamic analyses of frame structures have been performed. One hundred Fema440 records were utilized to generate approximate waves up to the third level and the outcomes of this method were then contrasted with those of DWT. It has been demonstrated that the third-level approximate wave produced by DWT, previously considered dependable in other research, generates significant errors in results and the average error (absolute error percentage of the acceleration spectrum) of its third-level approximate wave is approximately 17.5%. On the other hand, the proposed method generated approximate waves with an average error of less than 4.5% across all behavior coefficients and periods and the error rate decreases as the period and behavior coefficient increase. Analysis of steel moment-resisting frames indicated that the lowest error in both methods is achieved for the base shear and across different engineering demand parameters, the average error rate for the proposed method was below 7.5%. Furthermore, caution must be exercised when employing the proposed method for structures with periods shorter than 0.5 s.

Keywords:

• Computational costs
• discrete wavelet transform
• downsampling
• nonlinear dynamic analysis
• smoothing

Disclosure statement

No potential conflict of interest was reported by the author(s).