Monitoring and prediction of surface deformation along railways lines in karst areas using multi-source data − a case study of the Beijing-Guangzhou Railway in China

The Karst collapse seriously affects the safe operation of railways. Identifying the instability target areas along the railway and capturing its deformation trend is one of the most effective methods for controlling this hazard. However, the traditional methods of monitoring and analyzing deformation along the railway line are too single, resulting in the inability to carry out comprehensive identification of deformation and advance prediction. Hence, a method using multi-source data is proposed to monitor and predict surface deformation along railway lines in karst areas. Taking the Guangzhou part of the Beijing-Guangzhou Railway in China as an example, the hazards and adjacent human engineering activities along the railway are monitored by using multi-source means of “Space-Air-Ground” at first. Secondly, taking the time-series deformation values as the data basis, the wavelet transform (WT) algorithm is used to reduce the noise of the time-series deformation in the three historical subgrade karst collapse points. Five classical machine learning (ML) models are used for the prediction analysis of the deformation, and the optimal ML model is utilized to advance prediction. The results show: 1) There are two settlement target areas along the railway, and the settlement range is expanding and the cumulative settlement increases year by year in 2021–2023, with the maximum cumulative settlement of −118.01 mm. 2) The characteristic points in Area A continue to sink from 2019 to 2023, with a maximum cumulative settlement of −108.01 mm. The characteristic points in Area B continue to sink from 2021 to June 2023 due to disturbance from adjacent projects. 3) The prediction effect of the WT-LSTM is optimal through prediction analysis, and the scale of future deformation prediction is about one year based on the results of the advance prediction. The research findings can provide key technical support for the identification and early prevention of settlement hazards along railways.