Optimization of effective strain coefficient for nonlinear site response estimation by equivalent-linear simulation

In this study, new field investigations such as PS logging, formation density logging and soil tests were conducted at IBRH13 (KiK-net Takahagi station), where ground motion records with more than 500 Gal were observed three times during near source earthquakes with the magnitude of around 6. Based on new field investigation data, an applicability of three types of equivalent-linear simulations for nonlinear site response estimation was examined by using three ground motion records with more than 500 Gal. As a result, it was found that the observed site responses were relatively well explained by the equivalent-linear simulation with frequency-dependent effective strain for damping factors. However, for the largest record over 1,000 Gal, all types of equivalent-linear simulations with the conventional effective strain coefficient of 0.65 were not effective to explain the characteristics of observed site responses. Moreover, an inversion method to optimize effective strain coefficients was proposed to expand an application strain range of equivalent-linear simulations. In the inversion, observed Fourier spectral amplification and response spectral amplification between the surface and GL-100m were used as target site responses. Optimization of effective strain coefficients was carried out by minimizing differences between target site responses and the ones from equivalent-linear simulations with frequency-dependent effective strain for damping factors. As a result, the optimized effective strain coefficients were identified as less than the conventional effective strain coefficient of 0.65 in the shallower layers. For the largest record over 1,000 Gal, it was demonstrated that the target site responses and waveform at surface ground were well explained by the estimated ones from equivalent-linear simulation with the optimized effective strain coefficients. Also, it was shown that the characteristics of converged damping factors were similar to the bi-linear type frequency-dependent model from the equivalent-linear simulation with frequency-dependent effective strain for damping factor. Furthermore, outcrop bedrock ground motion response spectra were evaluated by equivalent-linear simulations with optimized and conventional effective strain coefficients using observed bedrock record. From the comparison of those estimated outcrop bedrock ground motion response spectra, it was found that both cases were almost consistent with each other. Consequently, we infer that bedrock ground motion is hardly influenced by reflected waves in the case of during nonlinear soil behavior caused by strong ground motion with over 1,000 Gal at the surface ground. Whereas, it was demonstrated that surface ground motion was underestimated by equivalent-linear simulations with conventional effective strain coefficients using bedrock record. This discrepancy is due to the inaccuracy of nonlinear site response estimation by equivalent-linear simulation using conventional effective strain coefficients. In the case of estimating bedrock ground motion by using surface record, the estimated bedrock ground motion is also strongly affected by the quality of estimated nonlinear site response by equivalent-linear simulation. Therefore, the optimization of effective strain coefficients is also important to estimate bedrock ground motion by using surface record. © 2019 Architectural Institute of Japan. All rights reserved.