A hybrid methodology for the prediction of subway train-induced building vibrations based on the ground surface response

The numerical simulation and theoretical methods for the subway train-induced vibration of the shallow foundation buildings often suffer from high cost, unstable prediction accuracy, and lack of clarity of important parameters. Therefore, a hybrid prediction method based on the Z-vibration level at the ground surface was proposed to rapidly obtain the vibration characteristics of the shallow foundation building adjacent to the subway. The numerical simulation was first used to obtain the subway train-induced vibration of the shallow foundation building under different working conditions. Then, a hybrid model was established and retrained by combining the field measurement data of the soil and building vibration along the subway line. Finally, the prediction accuracy of the hybrid model with different numbers of measurement points as input layers was explored, and a case study was performed. The results show that the most noticeable effect on subway train-induced building vibration is the length of the building span among the shallow foundation building parameters. Three measurement points of the Z-vibration level at the ground surface are suggested as the training set data for the input layer of the hybrid model in consideration of computational efficiency and accuracy. The prediction accuracy of the hybrid model gradually increases as the number of data sets increases, and the fully trained hybrid model performs more stable across the frequency range compared to the traditional model, with the majority of its predictions in the 90% confidence interval, which provides the possibility of simplifying the analysis and fast prediction of subway train-induced building vibration.

• Record URL:
  • https://doi.org/10.1016/j.trgeo.2024.101330
• Record URL:
  • http://www.sciencedirect.com/science/article/pii/S221439122400151X
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/22143912
• Supplemental Notes:
  • © 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies. Abstract reprinted with permission of Elsevier.
• Authors:
  • Tu, Wenbo
  • Shen, Lunqiang
  • Zhang, Pengfei
  • Zhang, Xiaolei
  • Liu, Linya
  • Chen, Juan
• Publication Date: 2024-9

Language

• English

Media Info

• Media Type: Web
• Features: References;
• Pagination: 101330
• Serial:
  • Transportation Geotechnics
  • Volume: 48
  • Issue Number: 0
  • Publisher: Elsevier
  • ISSN: 2214-3912
  • Serial URL: http://www.sciencedirect.com/science/journal/22143912

Subject/Index Terms

• TRT Terms: Machine learning; Predictive models; Structures; Subways; Vibration
• Subject Areas: Data and Information Technology; Environment; Public Transportation; Railroads;

Filing Info

• Accession Number: 01930199
• Record Type: Publication
• Files: TRIS
• Created Date: Sep 13 2024 10:34AM