Site amplification prediction model of shallow bedrock sites based on machine learning models

Lee, Yong-Gook, Kim, Sang-Jin, Achmet, Zeinep, Kwon, Oh-Sung, Park, Duhee and Di Sarno, Luigi ORCID: 0000-0001-6244-3251
(2023) Site amplification prediction model of shallow bedrock sites based on machine learning models. Soil Dynamics and Earthquake Engineering, 166. p. 107772.

[img] Text
SDEE 2023.pdf - Author Accepted Manuscript

Download (2MB) | Preview


Prediction of the site amplification is of primary importance for a site-specific seismic hazard assessment. A large suite of both empirical and simulation-based site amplification models has been proposed. Because they are conditioned on a few simplified site proxies including time-averaged shear wave velocity up to a depth of 30 m (VS30) and site period (TG), they only provide approximate estimates of the site amplification. In this study, site amplification prediction models are developed using two machine learning algorithms, which are random forest (RF) and deep neural network (DNN). A comprehensive database of site response analysis outputs obtained from simulations performed on shallow bedrock profiles is used. Instead of simplified site proxies and ground motion intensity measures, matrix data which include the response spectrum of the input ground motion and shear wave velocity profile. Both machine learning based models provide exceptional prediction accuracies of both the linear and nonlinear amplifications compared with the regression-based model, producing accurate predictions of both binned mean and standard deviation of the site amplification. Among two machine learning techniques, DNN-based model is revealed to produce better predictions.

Item Type: Article
Uncontrolled Keywords: Machine learning, Random forest, Deep neural network, Site amplification, Site response analysis
Divisions: Faculty of Science and Engineering > School of Engineering
Depositing User: Symplectic Admin
Date Deposited: 01 Feb 2023 09:20
Last Modified: 13 Jan 2024 02:33
DOI: 10.1016/j.soildyn.2023.107772
Related URLs: