关键词: 路基形变预测, 沙地猫群算法, SVR, 多源气象数据, PS-InSAR

Abstract: Aiming at the problems that the deformation of railway subgrade in seasonal freezing area is easily affected by environment and the accuracy of traditional single variable deformation prediction model is insufficient, a TSCSO-SVR model combining multi-source meteorological data is proposed in this paper. Firstly, PS-InSAR technology is used to monitor the deformation of roadbed, and the correlation between meteorological factors and roadbed deformation is analyzed. Then, the improved sand cat swarm algorithm (TSCSO) is obtained by combining nonlinear decline, dynamic disturbance and spiral search, and the TSCSO-SVR subgrade settlement prediction model is constructed. Finally, combined with the measured data of a section of railway in Shihezi, Xinjiang. The results show that the prediction effect of multivariate model is generally better than that of univariate model. Compared with other models, TSCSO-SVR prediction model has the highest prediction accuracy and has good application value.

Key words: subgrade deformation prediction, sand cat swarm algorithm, SVR, multi-source meteorological data, PS-InSAR