Computational & Technology Resources
an online resource for computational,
engineering & technology publications
engineering & technology publications
Civil-Comp Conferences
ISSN 2753-3239
ISSN 2753-3239
CCC: 3
PROCEEDINGS OF THE FOURTEENTH INTERNATIONAL CONFERENCE ON COMPUTATIONAL STRUCTURES TECHNOLOGY
PROCEEDINGS OF THE FOURTEENTH INTERNATIONAL CONFERENCE ON COMPUTATIONAL STRUCTURES TECHNOLOGY
Edited by: B.H.V. Topping and J. Kruis
Paper 22.2
CFD-enabled prediction of aerodynamic coefficients for long-span bridges using machine learning
S. Tinmitonde1 and X.H. He2
1School of Civil Engineering, Central South University, Changsha, China
2National Engineering Laboratory for High-Speed Railway Construction, Changsha, China
Full Bibliographic Reference for this paper
S. Tinmitonde, X.H. He, "CFD-enabled prediction of aerodynamic coefficients for long-span bridges using machine learning", in B.H.V. Topping, J. Kruis, (Editors), "Proceedings of the Fourteenth International Conference on Computational Structures Technology", Civil-Comp Press, Edinburgh, UK,
Online volume: CCC 3, Paper 22.2, 2022, doi:10.4203/ccc.3.22.2
Keywords: long-span bridge, wind-resistant design, flutter, CFD, machine learning, accuracy.
Abstract
Long-span bridges are very vulnerable to wind-induced vibrations. Amid wind-induced excitation on long-span bridges, flutter was the most dangerous after the well-known old Tacoma Narrow Bridge (TNB) collapsed in 1940. Currently, the aerodynamic performance of the long-span bridges can be appreciated after conducting the experimental wind tunnel tests or by the mean of computational fluid dynamics (CFD) simulations. However, the traditional wind tunnel tests or CFD are thought to be very cumbersome and costly, especially when there are many design samples to evaluate. This study proposed predicting the aerodynamic coefficients (drag, lift, and moment coefficients) of a streamlined bridge subjected to shape modifications using machine learning approaches based on the CFD simulations dataset. Six machine learning pipelines, including gradient boosting regression(GBR), random forest regression (RFR), Bayesian ridge(BR), AdaBoost Regression (AdaBoost), decision tree regression (DTR), light gradient boosting machine(lightgbm), were built. The results showed that the GBR exhibited the best predictive performance on the drag coefficients, whereas the lightgbm algorithm performed well in predicting the lift and moment coefficients. This study is essential to help bridge designers to make a fast decision at the earlier design stage of modern long-span bridges to meet the increasingly rapid requirement of such mega-structures. This study can also help reduce the number of models to be tested based on preliminary information obtained from the ML models before any in-depth study.
download the full-text of this paper (PDF, 9 pages, 822 Kb)
go to the previous paper
go to the next paper
return to the table of contents
return to the volume description