A Study On Nonlinear Site Amplification Model And Ground-motion Attenuation Relations

Ground-motion prediction equation（GMPE）could be used to quantitatively estimate a potential ground-motion intensity in future for an area and it is very valuable for engineering application.As an important part of ground-motion prediction equation,the site term presents a predicted site amplification ratio and the standard deviation of the model would evaluate its variability.When developing a site term for a ground-motion prediction equation,the accurate estimation of nonlinear site response and selection of the site parameter are key issues,which definitely would affect the predicted results and its variability.Unfortunately,the ground-motion observation work in our country starts late and detailed site information for the observation stations is still unvailable.The research work about the above issues is not common and few results is accessible in current stage.Because of this,detailed and in-depth research about nonlinear site response prediction,development of ground-motion prediction equation and the effect of site parameter on them was carried out in this dissertation.The work includes:（1）Using the borehole soil information and strong-motion records from KiK-net and K-NET stations in Japan,we derived a set of empirical site amplification ratios based on a GMPE for rock-site.Then the characteristic of the empirical site amplification ratios was presented and analyzed.The results showed that the resonant effect is significant for linear site response,whereas the number of records with strong nonlineat site response is small in the empirical dataset.According to site response analysis theory and characteristics of the empirical site amplification ratios,we proposed using a simplified transfer function for a single-degree-of-freedom system（SDOF）as basic function form for the linear site term of site amplification model,which could fully present site resonant effect.（2）Accounting for the fact that the number of strong-motion records with nonlinear site response is small,we compiled a dataset of one dimensional（1D）nonlinear site amplification ratios using site response analysis method.The borehole profiles from KiK-net and K-NET stations were used to account for the uncertainty of soil shear-wave velocity profile and nonlinear dynamic parameters of the soil layers,and acceleration time histories from rock-sites with balanced distributions among moment magnitude,source distance,and peak ground acceleration were selected for 1D site response analysis.Then,a large number of nonlinear site amplification ratios were derived for the dataset.We presented the basic characteristics of the 1D nonlinear site amplification ratios and made comparison with the empirical site amplification ratios derived from strong-motion records.The results showed that the 1D amplification ratios not only contains resonant effect,but impedance ratio effect and nonlinear site effect.The difference between 1D amplification ratios and empirical ratios is small at PGA,while it is large at intermediate long and long period.（3）We developed three 1D nonlinear site amplification models using the 1D site amplification ratio dataset.The 1D models used different site parameters,one used site period,T_s,four times the travel time of vertical propagating shear wave from bedrock to the ground surface,one used V_s30,the time-averaged shear-wave velocity in top 30 meters(being equivalent to T_vs30 in this article),and one used T_{vs H},a combined parameter between bedrock depth and V_s30.These models have similar function forms and each of them has a linear site term and a nonlinear site term.A comparison on the predicted 1D site amplification ratios and their uncertainties among these three 1D site amplification models showed that the predicted1D amplification ratios from the 1D models are close for stiff soil sites,while they are so different for soft soil sites at intermediate long and long periods.The standard deviations indicate that the predicted site response from the 1D T_s model is best.We also proposed an adjustment method to implement the 1D models into ground-motion prediction equations and to improve its applicability.For a very small number of strong-motion records with nonlinear site response from Japan,the trend in the variation of amplification ratios from the 1D model is generally similar to those derived from the ratios between the soil site records and predicted rock motions by the GMPEs developed in this study.Comparing with the nonlinear site terms in the NGA-West2 attenuation models,the general characteristics of the predicted amplification ratios appear to be reasonable.（4）A set of ground-motion prediction equations with site period T_s,V_s30,T_{vs H} as site parameter,respectively,were developed using a large dataset,in which 5957 high-quality strong-motion records were observed from the KiK-net and K-NET sites,for shallow crustal and upper mantle earthquakes in Japan.We emphasized our study on site terms and its variability when developing the GMPEs and referred to the source term and path term in GMPE by Zhao et al.（2016a）.The 1D nonlinear site amplification models developed in this thesis were used to constraint nonlinear site response of the GMPEs.Therefore,a quantitative evaluation to linear site response and nonlinear site response has come true.Compared with the GMPE using site class as the site effect proxy,the GMPE using T_s as site parameter predicts continuously varying site amplification ratios and response spectrum versus site effect parameter.Also,the goodness-of-fit for the GMPE using T_s as the site effect proxy is significantly improved and the between-site standard deviations significantly reduced at long periods.The predominant period of the predicted site amplification ratios for the GMPE using V_s30 as site effect proxy is close to that using T_s as site effect proxy for stiff soil sites,whereas they are different for soft soil sites.At intermediate and long periods,between-site standard deviations of the GMPE using V_s30 as site parameter are significantly larger than those using T_s as site parameter,while those using T_{vs H} are close to those using T_s as site parameter.（5）The performance of site parameters was assessed by comparing its impact on the predicted amplification ratios and the standard deviation of the developed 1D models and GMPEs.Accounting for the engineering test difficulty of the site parameters,we conclude that T_s is the best parameter for site amplification model and GMPEs,and V_s30 is the worst,and the T_{vs H} is similar to T_s.The method and developed models in this study may provide a good guidance for site amplification model and GMPEs development,and provide a good reference for the works of seismic zonation in China.