Impact of ground motion selection methods on the seismic assessment of steel special moment frames

Abstract

This study quantifies the impact of different ground motion selection methods on the seismic performance evaluation of steel special moment frames. Two methods are investigated: a "traditional" approach, herein referred to as the Pacific Earthquake Engineering Research (PEER) method, and a newer approach known as the Conditional Mean Spectrum (CMS) method. The PEER method selects ground motions using the Riskbased Maximum Considered Earthquake (MCER) as the target spectrum, while the CMS method uses the conditional mean spectrum that anchor to the MCER at multiple conditioning periods. Three special moment frames of 4-, 8-, and 16-stories are designed in accordance with ASCE/SEI 7-10 to represent archetype steel frame buildings as found in regions of high seismicity. The seismic performance of these frames is assessed with the nonlinear dynamic procedure prescribed in ASCE/SEI 41-13, using ground motions selected and scaled in accordance with both methods. The performance of the buildings is evaluated at the Collapse Prevention (CP) performance level for a far-field site located in Los Angeles, CA. The CMS method results in lower mean and median response in terms of demand-to-capacity ratios in the reduced beam sections and column hinges. Ground motions selected and scaled using CMS result in a smaller dispersion of the output parameters in most of the beam and column elements, if the conditioning period that results in the highest mean demand-to-capacity ratio is the fundamental period, 1. The results of this study show that the ground motion selection process can cause significant differences in structural response that may lead to different retrofitting decisions. These results provide

Effect of Common U.S. Ground Motion Selection Methods on the Structural Response of Steel Moment Frame Buildings

“…This sorted data approach assumes that the evaluated data reasonably fit a lognormal distribution. A series of Kolmogorov-Smirnov tests on the DCR data demonstrated that the lognormal distribution fits a majority of the DCR data more appropriately than a normal distribution (Uribe et al 2018). This is an expected result given that the data have only positive values and the distribution is skewed to the right (Shome and Cornell 1999).…”

“…Other studies have used methods similar to the TMCE method that match records to the MCE R spectrum while minimizing the error in a specified range for use in nonlinear analyses (e.g., Kalkan and Chopra 2010). In contrast, the CMS method uses the CMS as the target spectrum for scaling GMs to match the spectral acceleration (S a ) at a conditioning period (Baker 2011, Uribe et al 2017. For the simplest case in which only one conditioning period is used, the CMS method produces a record set with no dispersion at this period, and it can be considered a single target spectral acceleration approach (Adam et al 2017).…”

Assessment of Performance-Based Seismic Design Methods in ASCE 41 for New Steel Buildings: Special Moment Frames

“…The effects of ground motion selection and scaling can be important for the NDP, including the number of records adopted to achieve a reasonable level of statistical confidence and the method by which the records were selected (without a bias to achieve an unfairly beneficial outcome, i.e., "cherry-picking"). Uribe et al (2017) showed that ground motion selection using the Conditional Mean Spectrum approach may result in reduced DCR N values when compared with some traditional approaches. Moreover, some of the higher mode periods fall directly in localized high energy regions of the response spectrum, resulting in increased demands that cannot be captured efficiently in a linear analysis using a smooth, generalized spectrum.…”

A Holistic Review of GM/IM Selection Methods from a Structural Performance-Based Perspective

“…The authors showed that the CMS method estimates structural response with lower dispersion when the first-mode controls. Additionally, the expected values of predicted responses were different [142]. Dyanati et al compared the results from four generic GM suites (e.g., SAC, FEMA far-field records) with GMs selected based on CMS for a 6-story steel braced frame, and showed that responses obtained from CMS cover all damages states, whereas the generic records are suitable for performance assessment at either low or high damage states [143].…”