Spectral shape-based assessment of SDOF nonlinear response to real, adjusted and artificial accelerograms

• Published in: Engineering structures 2010-09, Vol.32 (9), p.2776-2792
• Main Authors: Iervolino, Iunio, De Luca, Flavia, Cosenza, Edoardo
• Format: Article
• Language: English
• Subjects: Adjustment; Applied sciences; Artificial accelerograms; Assessments; Buildings. Public works; Demand; Evolutionary; Exact sciences and technology; Geotechnics; Marketing; Measurements. Technique of testing; Nonlinearity; Real records; RSPMatch 2005; Scaling; Seismic phenomena; Spectra; Spectral matching; Stresses. Safety; Structural analysis. Stresses; Structure-soil interaction
• ISSN: 0141-0296; 1873-7323
• DOI: 10.1016/j.engstruct.2010.04.047

The simple study discussed in this paper compared different procedures to obtain sets of spectral matching accelerograms for nonlinear dynamic analysis of structures in terms of inelastic seismic response. Six classes of records were considered: original (unscaled) real records, real records moderately linearly scaled, real records significantly linearly scaled, real records adjusted by wavelets, and artificial accelerograms generated by two different procedures. The study is spectral shape-based; that is, all the considered sets of records, generated or selected, match individually (artificial and adjusted) or on average (real records) the same design spectrum for a case-study site in Italy. This is because spectral compatibility is the main criterion required for seismic input by international codes. Three kinds of single degree of freedom (SDOF) system, non-degrading and non-evolutionary, non-degrading and evolutionary, and both degrading and evolutionary, were used to evaluate the nonlinear response to the compared records. Demand spectra in term of peak and cyclic responses were derived for different strength reduction factors. Results of the analysis show that artificial or adjusted accelerograms may underestimate, in some cases, and at high nonlinearity levels, the displacement response, if compared to original real records, which are considered as a benchmark herein. However, this conclusion does not seem to be statistically significant. Conversely, if the cyclic response is considered, artificial record classes show a significant overestimation of the demand, which does not show up for wavelet-adjusted records. The two classes of linearly scaled records do not show systematic bias with respect to those unscaled for both types of response considered, which seems to confirm that amplitude scaling is a legitimate practice.