28.05.2014 Views

r - The Hong Kong Polytechnic University

r - The Hong Kong Polytechnic University

r - The Hong Kong Polytechnic University

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>The</strong> 5th Cross-strait Conference on Structural and Geotechnical Engineering (SGE-5)<br />

<strong>Hong</strong> <strong>Kong</strong>, China, 13-15 July 2011<br />

UNCERTAINTY IN NONLINEAR SEISMIC GROUND RESPONSE ANAYLYSES<br />

O. L. A. Kwok<br />

Department of Civil Engineering,<br />

National Taiwan <strong>University</strong>, Taipei, Taiwan. Email: annieonleikwok@ntu.edu.tw<br />

ABSTRACT<br />

One-dimensional seismic ground response analyses are seldom performed using nonlinear procedures because<br />

parameter selection and code usage protocols are poorly documented and understood, the effect of parametric<br />

variability on the analysis results is generally unknown, and the benefits of nonlinear analysis relative to the<br />

widely-used equivalent-linear analysis are un-quantified and unclear. In this paper, effect of parametric<br />

variability on predictions using nonlinear ground response analysis will be discussed. Sources of variability<br />

considered include material properties such as dynamic material curves and velocity profiles. Various vertical<br />

array sites are used in the study. It is found that the standard deviation of predictions is dominated by material<br />

curve variability at small periods and velocity variability at medium to large periods respectively.<br />

KEYWORDS<br />

nonlinear, ground response analysis, one-dimensional, uncertainty.<br />

INTRODUCTION<br />

<strong>The</strong>oretical modeling of one-dimensional site response can generally be accomplished using equivalent linear or<br />

nonlinear analysis. Equivalent-linear ground response modeling is commonly performed in practice as it<br />

requires the specification of well-understood and physically meaningful input parameters (shear-wave velocity,<br />

unit weight, modulus reduction and damping). Nonlinear ground response analyses provide a more accurate<br />

characterization of the true nonlinear soil behavior, but their implementation in practice has been limited, which<br />

is principally a result of poorly documented and unclear parameter selection and code usage protocols.<br />

Moreover, previous studies have thoroughly investigated the sensitivity of site response results to the input<br />

parameters in equivalent-linear analyses (e.g. Roblee et al., 1996), but this level of understanding is not<br />

available for the input parameters used in nonlinear analyses. <strong>The</strong> objective of current study is to evaluate the<br />

uncertainties in predictions by nonlinear ground response modeling due to various sources of variability,<br />

including material properties and modeling schemes. Vertical array sites are used in this study as the uncertainty<br />

due to input ground motion can be eliminated and predictions from nonlinear analyses can be compared directly<br />

to the data (recordings).<br />

SOURCES OF VARIABILITY<br />

Modeling Schemes<br />

<strong>The</strong> ground response analyses are performed by using various nonlinear codes including DEEPSOIL (Hashash<br />

and Park 2001, 2002; Park and Hashash 2004), D-MOD_2 (Matasovic 2006), a ground response module in the<br />

OpenSees simulation platform (Ragheb 1994; Parra 1996; Yang 2000; McKenna and Fenves 2001;<br />

opensees.berkeley.edu), SUMDES (Li et al. 1992), and TESS (Pyke 2000). All of these codes are time-domain<br />

analysis methods which allow soil properties within a given layer to change with time as the strains in that layer<br />

change. Like a structure, the layered soil column is idealized either as a multiple degree of freedom lumped<br />

mass system or a continuum discretized into elements with distributed mass with appropriate boundary<br />

conditions. Soil material models employed range from relatively simple cyclic stress-strain relationships<br />

(DEEPSOIL, D-MOD_2 and TESS) to advanced plasticity models (OpenSees and SUMDES).<br />

Material Properties<br />

<strong>The</strong> four vertical array sites considered in this study include KGWH02 (from the Japanese Kiknet network of<br />

strong motion stations), Lotung (Taiwan), Turkey Flat and La Cienega (both of them are in California of the<br />

-323-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!