TY - GEN
T1 - Application of XGBoost Model for Predicting the Dynamic Response of High-Speed Railway Bridges
AU - Le Nguyen, Khuong
N1 - Funding Information:
This research is funded by University of Transport Technology (UTT) under grant number ÐTTÐ2021-28.
Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2022/10/28
Y1 - 2022/10/28
N2 - The dynamic response at high speed affects both the vehicles and the structures in a complex manner, especially in the railway infrastructure problems. In this study, we developed a new KD-Railway tool for analyzing the dynamic behavior of high-speed railways by using the finite element method. Then, extreme gradient boosting (XGBoost) was used to predict and better understand the dynamic response of high-speed railway bridges. The model was trained and tested using a dataset including properties and dynamic responses of 10,000 bridges generated by KD-Railway. The input variables were the bridge span length, the flexural rigidity, mass per length of the bridge, the cross-section area of bridge decks, the train speed, the damping ratio, and the HSLM train models. On the other hand, maximum vertical deflection and maximum acceleration were considered as the output parameters. The coefficients of determination (R2) for these two outputs were (0.996, 0.931, 0.977) and (0.987, 0.901, 0.962) for the training, testing, and entire dataset, respectively. The sensitivity analyses were also conducted to evaluate the importance of each input variable on the outcomes.
AB - The dynamic response at high speed affects both the vehicles and the structures in a complex manner, especially in the railway infrastructure problems. In this study, we developed a new KD-Railway tool for analyzing the dynamic behavior of high-speed railways by using the finite element method. Then, extreme gradient boosting (XGBoost) was used to predict and better understand the dynamic response of high-speed railway bridges. The model was trained and tested using a dataset including properties and dynamic responses of 10,000 bridges generated by KD-Railway. The input variables were the bridge span length, the flexural rigidity, mass per length of the bridge, the cross-section area of bridge decks, the train speed, the damping ratio, and the HSLM train models. On the other hand, maximum vertical deflection and maximum acceleration were considered as the output parameters. The coefficients of determination (R2) for these two outputs were (0.996, 0.931, 0.977) and (0.987, 0.901, 0.962) for the training, testing, and entire dataset, respectively. The sensitivity analyses were also conducted to evaluate the importance of each input variable on the outcomes.
KW - Dynamic response
KW - Extreme gradient boosting (XGBoost)
KW - High-speed railway bridges
KW - HSLM train models
UR - http://www.scopus.com/inward/record.url?scp=85119407995&partnerID=8YFLogxK
UR - https://cigos2021.sciencesconf.org/program
U2 - 10.1007/978-981-16-7160-9_178
DO - 10.1007/978-981-16-7160-9_178
M3 - Conference contribution
AN - SCOPUS:85119407995
SN - 9789811671593
T3 - Lecture Notes in Civil Engineering
SP - 1765
EP - 1773
BT - CIGOS 2021, Emerging Technologies and Applications for Green Infrastructure - Proceedings of the 6th International Conference on Geotechnics, Civil Engineering and Structures
A2 - Ha-Minh, Cuong
A2 - Tang, Anh Minh
A2 - Bui, Tinh Quoc
A2 - Vu, Xuan Hong
A2 - Huynh, Dat Vu Khoa
PB - Springer
CY - Singapore
T2 - 6th International Conference on Geotechnics, Civil Engineering and Structures, CIGOS 2021
Y2 - 28 October 2021 through 29 October 2021
ER -