Search

With the rapid economic development and urbanization in China, subway systems have become the primary mode of urban rail transit. However, during subway operation, the tunnels may experience settlement and deformation due to various influencing factors. To guarantee safe operation of subway systems and eliminate potential safety hazards, tunnel settlement prediction has important significance. However, existing studies have seldom discussed the effects of weighting factors on subway tunnel settlement prediction. In addition, the optimization of support vector machine (SVM) using particle swarm optimization (PSO) often suffers from issues such as local optimization and premature convergence. To address these problems, grey relational analysis (GRA) and weighted particle swarm optimization (WPSO) SVM were combined, and a GRA-WPSO-SVM prediction model was constructed. This model was applied to predict subway tunnel settlement in the Sanyao Section of the Xi'an Exhibition Center in China. Prediction results from the GRAWPSO-SVM prediction model were compared with those from the PSO-SVM and SVM using root mean square error (RMSE), mean relative error (MRE), and correlation coefficient as evaluation metrics. Results demonstrate that, the RMSE and MRE of GRA-WPSO-SVM are 0.0008 m and 1.9707%, which are better than those of PSO-SVM and SVM. Moreover, prediction results of the GRA-WPSO-SVM exhibit a strong correlation with the measured data of tunnels, with a correlation coefficient of 0.93. Obviously, the GRA-WPSO-SVM is effective. The proposed method provides an important evidence for the prediction of subway tunnel settlement and deformation trends. [ABSTRACT FROM AUTHOR]