The observed variability is very large among natural earthquake records. The nonlinear dynamic analyses are used to determine seismic behavior of structures under input ground motions (GMs). In the current practice, the input variability is minimized with ground motion modifications due to the cost and the duration of the analyses, yet without clear indications of the consequences on the output structural responses. The study, herein, aims at quantifying the impact of permitting ground motion variability on the distribution of engineering demand parameters (EDPs). The following questions are discussed in this thesis:
What is the level of variability in natural and modified GMs?
What is the impact of input variability on the EDPs of various structural types?
For a given deterministic earthquake scenario, we use a magnitude-distance bin (M7.0R40) to collect unscaled earthquake records. A variety of ground motion prediction equations (GMPEs) is used to define the target spectra, which are necessary for the GM selection and modification. The GM modifications are applied over the unscaled earthquake records to collect (1) linearly scaled real GMs, (2) loosely spectrum-matched GMs, and (3) tightly spectrum-matched GMs. Nonlinear dynamic analyses of simple and complex structural models are then performed with the GM families to determine the EDPs, namely, roof displacements, base shear forces, interstory drift ratios, and global damage index. The changes triggered by the GM modifications are evaluated relative to the unscaled earthquake records through record-to-record comparison.
The response spectrum compatible selection is then performed to select five GMs from the whole GMs, i.e., a GM set. Two types of set variability are considered: intraset variability relates to the dispersion in a given set, and interset variability relates to the existence of multiple sets compatible with a target spectrum. Accordingly, the EDP distributions of the modified GMs are gathered with set variability and are compared to the observed EDP distributions of the unscaled earthquake records through set comparison.
This thesis demonstrates that a single GM set, commonly used in the practice, is not sufficient to obtain an assuring level of the EDPs regardless of the GM selection and modification methods, which is due to the record and set variability. The unscaled real records compatible with the scenario are discussed to be the most realistic option to use in the nonlinear dynamic analyses, and the ‘best’ GM selection and modification method is demonstrated to be specific to the EDP, the objective of the seismic analysis, and the structural model. It is pointed out that the choice of a GMPE can provoke significant differences in the GM characteristics and the EDPs, and it can even overshadow such differences imposed by the GM modifications.