Displacement prediction of deep excavated expansive soil slopes with high groundwater level based on VDM-LSSVM

Slope failures in deep excavated expansive soil canals are common when there is a high groundwater level and continuous rainfall. Predicting displacements plays a crucial role in preventing and mitigating such failures. To study this issue, a section of a well-known canal in the Nanyang Basin, part of the South-to-North Water Transfer Project (SNWTP), was examined. This canal section is characterized by deep excavation and expansive soil, and it experiences a high groundwater level. To understand the relationship between groundwater level, rainfall, and slope displacement, a qualitative analysis based on time series data was conducted. An integrated approach for displacement prediction was proposed, which combines the variational mode decomposition (VDM) algorithm, the least squares support vector machine (LSSVM), and k-fold cross-validation. The VDM algorithm decomposed the accumulated displacement into a trend, periodic, and fluctuation component. It also decomposed the environmental factors such as groundwater level, rainfall, canal water level, and air temperature into high-frequency and low-frequency factors. The gray relational analysis was used to determine the correlation between the environmental factors and displacement. For predicting the periodic and fluctuating displacement components, the periodic and fluctuating components of the environmental factors were selected as input datasets for the LSSVM algorithm. Finally, the total displacement was obtained by combining the three predictive components. The results of the study demonstrated that the proposed model achieved satisfactory prediction accuracy. Therefore, it can be considered effective and practical for expansive soil slope engineering with high groundwater conditions.