A Study On Damping Correction Factors For Subduction Interface Earthquakes

Under the excitation of earthquake ground motions,the response of a structure depends on its damping ratios.The design spectrum is an important parameter in the design process.Most code-design spectrum is based on a damping ratio of 5%but many types of structures have different damping ratios.In order to derive a design spectrum with different damping ratios from the 5%damped design spectrum,damping ratio correction factors(referred to as B_d here)derived from strong-motion records are usually used to scale the 5%damped design spectrum.This study aims to develop B_d from the strong-motion records generated by subduction inter-slab(interface)earthquakes in Japan.Because the frequency contents in the strong ground motions from different types of earthquakes differ significantly,the subduction interface earthquakes are used to develop a B_d model in this study.A total of 3111 records from76 earthquakes obtained from the K-NET and KiK-net are used.Site class based on site periods is used as the site parameter and the recording stations are divided into four site classes based on previous studies.Two sets of damping ratio correction factors for 14 damping ratios and 36spectral periods will be presented in this thesis.The first model is based on the average B_d values and can be used to scale a code design spectrum for a given site class.This model does not contain any parameters for earthquake source and path effects.The second model contains parameters for the source,path,and site effects;moment magnitude and fault depth are used to represent source terms and the closest distance from the recording station to the fault represents the path terms.All model parameters are smoothed with respect to these terms so that B_d will vary with these parameters reasonably smoothly for the convenience of engineering applications.First,statistical analyses were carried out to determine the necessity for each site class to have a separate model.A random effects regression method was used to construct the B_d model accounting for the source,path,and site effects.The total model residuals were separated into two parts:between-and within-event residuals;the total standard deviations were also separated into two corresponding parts,i.e.the between-and within-event standard deviations.The within-event residuals were separated into the between-and within-site residuals and the within-event standard deviations were separated into between-and within-site standard deviations.The distributions of residuals with respect to the corresponding model variable were used to confirm the model goodness of fit for each model parameter.The results for the residual analyses show that the residuals are evenly distributed in the full range of this model parameter and the slope of the linear trend line fitted to the residuals is statistically negligible for all types of residuals.At many spectral periods,between-event standard deviations are generally the smallest one among three types of standard deviations,suggesting that the modeling of source effects are better than the source and site effects.At many spectral periods,the within-site standard deviations are the largest among the three types of standard deviations,suggesting that the modeling of path effect are less well than the source and path effects are.However,the small standard deviations may also mean that the inherited errors from this type of effect are small.The model predictions show that B_d decreases with increasing earthquake magnitude at spectral periods over 0.3s and decreases with increasing earthquake fault-top depth in a period range of 0.03-0.3s.Source distance has a moderate effect and B_d increases with increasing distance at short periods up to 0.1s and decreases within spectral periods over 0.15s.The effect of volcanic distance is statistically significant but is small at nearly all spectral periods.