Submitted to the Department of Civil and Environmental Engineering.
Thesis (Ph.D.)--Stanford University, 2013.
Quantification of the seismic performance of structures is a critical step in the design and analysis of our built environment. This is sometimes accomplished using response history analysis, which requires ground motions resulting from (or attempting to simulate) real earthquakes. This thesis describes statistical studies of structural analysis results obtained from ground motions developed using response spectrum compatibilization (a.k.a. ``spectrum matching''), in order to evaluate whether the compatibilization approach produces ground motions that induce lower levels of demand on structures relative to those developed using other methods. Ground motions strong enough to cause significant damage are rare yet inevitable, motivating research on how to select, modify, or synthesize appropriate records for analysis. The spectrum matching process is a modification procedure used to provide such ground motions through nonlinear modification of spectral shape, and may be useful either to reduce spectral variability among a suite of records, or to change unsuitable spectral shapes to make them suitable. Response history analysis can be quite time-consuming, and a reduction in spectral variability generally leads to a reduction in nonlinear structural response variability, implying fewer ground motions (and fewer hours) are required to obtain an estimate of mean response. A drawback of using matched ground motions is that there is not yet consensus as to whether they will produce biased structural response results relative to results obtained from ground motions that were not matched. This thesis demonstrates the existence of a statistically significant and unconservative response bias for a variety of structural models and response parameters when ground motions are matched to reduce spectral dispersion. The cause for this bias is fully attributed to the impact of spectral shape variability, as characterized within a suite of ground motions by conditional spectral dispersion. A modified variable-target matching methodology is then proposed to provide rare, intense ground motions that produce unbiased response estimates by maintaining appropriate spectral shape variability. Finally, general recommendations are provided on the use of spectrum matching for design and analysis, and both the limitations of this study and future opportunities are discussed.