| This dissertation is concerned with the generation of correlated space-time ground motions, the formulation of response analysis methodologies, and the study of the effects of spatial variation of ground motion on structural response. A methodology is presented for conditional simulation of spatially-correlated earthquake ground motion. Frequency-dependent spatial components of ground motion are generated using linear prediction estimators. The method's advantage over Kriging techniques is that it correctly reproduces the specified space-time covariance structure of the ground motion. Conditional simulations are used in a time history response analysis of multi-support systems to study the influence of spatial correlation and frequency content of ground motion on the structural response of a two-span continuous beam.; A random vibration methodology is formulated for the seismic response analysis of linear multi-support systems. It reduces the response analysis to that of a series of one-degree systems in a way that fully accounts for the multiple support input. It is shown that the dynamic response covariance can be expressed in terms of various spectral parameters associated with the one-degree system responses. This methodology is used to study the effect of local soil conditions, wave propagation, and incoherence on multi-support system response. The relative importance of the pseudo-static and dynamic response components is also evaluated. The validity of such common assumptions as full coherency and statistical independence of multiple input ground motions is analysed. |
