TY - JOUR
T1 - Prediction of long-term extreme response due to non-Gaussian wind on a HSR cable-stayed bridge by a hybrid approach
AU - Xu, Zhiwei
AU - Dai, Gonglian
AU - Chen, Y. Frank
AU - Rao, Huiming
N1 - Funding Information:
This research was financially supported by the Project of Science and Technology Research and Development Program of China Railway Corporation (No. 2017G006-N , NO. K2018G017 ) and the Fundamental Research Funds for the Central Universities of Central South University (No. 2020zzts156 ), to which the authors are very grateful. The authors also appreciate the computational assistance provided by the National Supercomputing Centre of Singapore.
Funding Information:
Another critical issue is how to analyze the long-term EVD efficiently, which is a problem of methodology. Among the analytical approaches mentioned before, the FLM is the most robust but rather computationally demanding (Lystad et al., 2021); and the IFORM and ECM developed by an iterative scheme may result in the inefficient and inaccurate computational convergence for high-dimensional extreme evaluation problems (Luo et al., 2022). These deficiencies will be aggregated when analyzing a long-span bridge with a large number of degrees of freedom. More recently, the potential of machine learning algorithms in predicting structural long-term extreme responses has been recognized by several researchers (Leong and Bahuguni, 2020; Monsalve-Giraldo et al., 2018; Xu et al., 2020). Generally, the machine leaning is used to approximate the wind-load-extreme response chain by a surrogate model like support vector machine (SVM) (Samui, 2008), extreme gradient boosting (XGBoost) algorithm (Chen and Guestrin, 2016), artificial neural network (ANN) (Abbas et al., 2020), etc. The computational burden has been shown to be significantly improved when dealing with the high-dimensional variables problems in long-term extreme response evaluation because of their inherent “learning capacity” and the ability to account for various error sources (Rizzo and Caracoglia, 2021). In this study, the strategy for predicting the long-term extreme load effects by the machine learning algorithm is improved to significantly reduce the training time.As stated above, the full long-term method is employed in this section to verify the long-term EVD predicted by the proposed method in this study. A 100 m simply-supported HSR bridge with the cross-sections of box girder and pier shown in Fig. 20 is used in the verification study for reducing the computational cost and amplifying the extreme response. The track structure is the Chinese HSR double-block ballastless track. These components are discretized as spatial beam elements and linear damper-spring elements in the finite element model. The static aerodynamic coefficients CD, CL, and CM of the girder under crosswinds are 1.37, 0.65, and 0.17 respectively, obtained from the wind tunnel test.This research was financially supported by the Project of Science and Technology Research and Development Program of China Railway Corporation (No. 2017G006-N, NO. K2018G017) and the Fundamental Research Funds for the Central Universities of Central South University (No. 2020zzts156), to which the authors are very grateful. The authors also appreciate the computational assistance provided by the National Supercomputing Centre of Singapore.
Publisher Copyright:
© 2022 Elsevier Ltd
PY - 2022/12
Y1 - 2022/12
N2 - The effect of non-Gaussian inflows on structural long-term extreme buffeting responses has been little investigated. In this study, the sensitivity of long-term extreme value distribution (EVD) of a high-speed railway cable-stayed bridge to the non-Gaussian intensity is studied first. The turbulence skewness and kurtosis are then taken as the environmental variables to investigate their single and combined effects on bridge's long-term EVDs based on a proposed hybrid approach that combines the machine learning algorithm and virtual process method. The 2.5-year measured turbulence wind and 40-year annual extreme wind speed recorded near the bridge site are utilized to describe the probability distributions of the skewness and kurtosis of turbulence wind and 10-min mean wind speed. The research results reveal that: (1) the long-term EVD of torsional angle is more sensitive to non-Gaussian turbulence wind than vertical and lateral extreme responses; (2) the single effect of turbulence skewness is detrimental but limited, and the combined effect of skewness and kurtosis of turbulence u (w) is also weak within the considered MRIs (1–100 years). Lastly, the virtual process method is shown to be applicable to predict structural long-term EVDs; and it is efficient without losing significant prediction accuracy.
AB - The effect of non-Gaussian inflows on structural long-term extreme buffeting responses has been little investigated. In this study, the sensitivity of long-term extreme value distribution (EVD) of a high-speed railway cable-stayed bridge to the non-Gaussian intensity is studied first. The turbulence skewness and kurtosis are then taken as the environmental variables to investigate their single and combined effects on bridge's long-term EVDs based on a proposed hybrid approach that combines the machine learning algorithm and virtual process method. The 2.5-year measured turbulence wind and 40-year annual extreme wind speed recorded near the bridge site are utilized to describe the probability distributions of the skewness and kurtosis of turbulence wind and 10-min mean wind speed. The research results reveal that: (1) the long-term EVD of torsional angle is more sensitive to non-Gaussian turbulence wind than vertical and lateral extreme responses; (2) the single effect of turbulence skewness is detrimental but limited, and the combined effect of skewness and kurtosis of turbulence u (w) is also weak within the considered MRIs (1–100 years). Lastly, the virtual process method is shown to be applicable to predict structural long-term EVDs; and it is efficient without losing significant prediction accuracy.
UR - http://www.scopus.com/inward/record.url?scp=85140738011&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85140738011&partnerID=8YFLogxK
U2 - 10.1016/j.jweia.2022.105217
DO - 10.1016/j.jweia.2022.105217
M3 - Article
AN - SCOPUS:85140738011
SN - 0167-6105
VL - 231
JO - Journal of Wind Engineering and Industrial Aerodynamics
JF - Journal of Wind Engineering and Industrial Aerodynamics
M1 - 105217
ER -