Settlement prediction of existing metro induced by new metro construction with machine learning based on SHM data: a comparative study
Yang Ding,Da Hang,Yu-Jun Wei,Xiao-Long Zhang,Si-Yuan Ma,Zhi-Xiong Liu,Shuang-Xi Zhou,Zhen Han
DOI: https://doi.org/10.1007/s13349-023-00714-4
2023-06-21
Journal of Civil Structural Health Monitoring
Abstract:When the new metro line crosses the existing metro line, its construction stage will have an adverse impact on the existing metro line, that is, settlement. In order to ensure the safety of metro line operation, it is necessary to obtain the variation law of settlement. In this paper, the four different machine learning methods (Radial Basis Function (RBF) neural network, the Back Propagation Neural Network (BPNN), Generalized Regression Neural Network (GRNN) and Gaussian Prior (GP) regression model) are established for settlement prediction. In which, the settlement data set is shared in the https://pan.baidu.com/s/1UuGnobkUljndzeGutlfdlg (Password: 1029). Based on the structural health monitoring (SHM) data of Nanjing Metro, the prediction performance of the RBF is the worst; In addition, the robustness of the RBF and BPNN is very poor, that is, for the settlement of different monitoring points, the required prediction model structure is also different, which will lead to the RBF and BPNN cannot be widely used in settlement prediction. On the contrary, the GRNN and GP prediction models have good robustness, in which the GP prediction model has the best robustness, but the GRNN model has poor prediction performance. Therefore, the GP model is recommended to predict the settlement of Nanjing metro in this paper.
engineering, civil