Scaling Bias and Record Selection for Quantifying Seismic Structural Demand

Abstract

Lack of sufficient earthquake records for a given site and hazard level is a frequent challenge in seismic response assessment. This paper considers the response of several simple nonlinear systems to “amplitude-scaled” ground motions suites to determine bias in the response obtained compared with unscaled records. The dependence of this bias ratio on properties of the nonlinear system (e.g., elastic period and yield strength) and the parameter used to scale the ground motions [e.g., peak ground acceleration (PGA), peak ground velocity (PGV), and elastic spectral displacement] are investigated. Results showed that the amount of the bias is considerable and dependent on the properties of the nonlinear system, scaling parameter, and amplitude, and the bias can significantly affect the results of performance-based analyses. The origin of the bias was investigated and, consequently, a new metric is proposed to predict the bias induced by ground motion scaling without nonlinear analysis. Results demonstrated that using the predictor as a scaling parameter can significantly reduce the bias for various nonlinear single- and multi-degree-of-freedom structures.