Estimation method of site amplification factor based on interpretable random forest
The site amplification factor(Fpga)is a key parameter to reflect the difference of ground shaking peaks in different areas and to assess the impact range of ground shaking.The estimation of Fpga generally a-dopts a single parameter and is selected by experience,which makes the estimation results subjective.For this reason,40 stations of the KiK-net network are selected,and a comprehensive multi-parameter Fpga estimation method is proposed based on the Random Forest algorithm,which can explain the influence of each parameter on the estimation of Fpga.The method takes the magnitude,epicenter distance,epicenter depth,hypocentral dis-tance,overburden thickness h,site period T,and equivalent shear wave speeds VSE,VS30,VS20,and VS10 as the inputs,and Fpga as the output.Test results show that the correlation coefficient of the predicted Fpga is 0.84,and the mean value of the error is 0.01;the epicenter distance,epicenter depth,and hypocentral distance are posi-tively correlated with the estimated Fpga,and the magnitude,h,T,VSE,VS30,VS20,and VS10 are negatively correla-ted,with T and h contributing most to the model.The model can effectively predict the site amplification effect.
万方数据
维普
