EFFICIENT INTEGRATION OF NONLINEAR SITE RESPONSE

Dominic Assimaki-Professor
School of Civil and Environmental Engineering - Georgia TechABSTRACT:
While the quantification of site effects is of great significance in seismic hazard mitigation, there currently exists a large degree of uncertainty concerning the mathematical model to be employed for the efficient evaluation of these effects, and the site investigation program for evaluation of the associated input parameters. Towards the development of a comprehensive framework for credible and efficient integration of site response predictions in rupture scenaria simulations that addresses these issues, we here combine downhole observations and broadband ground motion synthetics for characteristic profiles in the Los Angeles Basin, and investigate the variability in ground motion introduced by the site response assessment methodology and uncertainty in nonlinear model parameter description. Regional velocity and attenuation structures are initially compiled using geotechnical data and the crustal velocity structure at three sites in Southern California. Broadband ground motion simulations are next conducted for scenaria of weak, medium and large magnitude events, and three component seismograms are computed on a surface station grid at distances 1km-75km from the surface projection of the fault. Elastic, equivalent linear and nonlinear site response simulations at multiple levels of complexity are then evaluated, and the coefficient of variation (COV) of site amplification factors is evaluated, defined as the ratio of the predicted peak ground acceleration (PGA) and spectral acceleration (SA) at short and long periods to the corresponding ground motion intensity measure on rock-outcrop. A frequency index is developed to identify the site conditions and ground motion characteristics where the high COV of free-field response implies that incremental nonlinear analyses should be employed in lieu of approximate methodologies. Next, the parametric uncertainty of nonlinear models is assessed by defining fixed- and free-parameters for each model and systematically randomizing the latter. Results show that site nonlinearity susceptibility and optimal nonlinear model complexity may be quantified by the proposed intensity-frequency index pair representation of site conditions and ground motion characteristics, while development of the target generalized computational framework for site response analyses is conditioned on the collection a statistically significant number of ground motion observations and synthetics at a wide spectrum of site conditions.

Location: Kaprielian Hall (KAP) - 209

Audiences: Everyone Is Invited

Contact: Evangeline Reyes