Optimal intensity measures for probabilistic seismic demand models of a cable-stayed bridge based on generalized linear regression models. Issue 131 (April 2020)

• Record Type: Journal Article
• Authors: Guo, Junjun; Alam, M. Shahria; Wang, Jingquan; Li, Shuai; Yuan, Wancheng
• Is Part Of: Soil dynamics and earthquake engineering. Issue 131(2020)
• Publication Date: 2020-04
• Subjects: Optimal intensity measures -- Probabilistic seismic demand models -- Generalized linear regression models -- Lasso regression -- Cable-stayed bridge; Soil dynamics -- Periodicals; Earthquake engineering -- Periodicals; Sols -- Dynamique -- Périodiques; Génie parasismique -- Périodiques; 624.176205
• DOI: 10.1016/j.soildyn.2019.106024
• Languages: English
• ISSNs: 0267-7261

Abstract

Abstract: Seismic intensity measures (IMs) play an important role in predicting the seismic responses of structures subjected to strong earthquakes. This paper proposes a general procedure to identify the optimal IMs for a long span cable-stayed bridge subjected to far-field and near-fault ground motions based on generalized linear regression models. Firstly, the generalized linear regression models, such as ordinary least squares (OLS), ridge regression and Lasso regression are presented. Secondly, the three dimensional numerical model of the bridge is generated in the OpenSees platform. Thirdly, 22 IMs are considered, and 160 ground motions from four site conditions are selected to excite the bridge in longitudinal and transverse directions separately. Then, the optimal IMs are determined by Lasso regression, which is an extended version of OLS, and the quadratic polynomial regression model is adopted to establish the probabilistic seismic demand models of the bridge. The numerical results reveal that peak ground velocity (PGV) can be selected as the optimal IM if only one IM is considered in the seismic demand models. However, PGV has a poor predictive ability for the seismic responses in the transverse direction. Hence, PGV, peak ground displacement (PGD), root-mean-square of velocity (VRMS ), specific energy density (SED), velocity spectrum intensity (VSI) and Fajfar intensity (FI) are selected as the optimal IMs by Lasso regression, and they are correlated with velocity except for PGD. The identified six IMs together can significantly increase the fitting ability of the models. Highlights: Based on Lasso regression, PGV, PGD, VRMS, SED, VSI and FI are selected as the optimal IMs from 22 IMs. Quadratic polynomial regression is used to establish the probabilistic seismic demand model. These identified 6 IMs can combinedly provide one IM that can significantly increase the fitting ability of the model.