Displacement prediction of reservoir landslides considering strength deterioration and step-like deformation characteristics under wetting-drying cycles

The periodic fluctuation of reservoir water level not only causes the dynamic change of seepage pressure in landslides， but also accelerates the strength deterioration of landslide materials in the hydro-fluctuation belt， resulting in the step-like deformation of reservoir landslides. To enhance the displacement prediction accuracy of reservoir landslides， this paper combines complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and gated recurrent unit (GRU)， and proposes a new displacement prediction model for reservoir landslides considering strength deterioration and step-like deformation characteristics under wetting-drying cycles. Firstly， based on the geological conditions and deformation monitoring data of the Baishuihe landslide in the Three Gorges Reservoir Area (TGRA)， we analyze the deterioration law of shear strength parameters in the hydro-fluctuation belt and the step-like deformation characteristics under wetting-drying cycles. CEEMDAN is then used to decompose the cumulative displacement into trend displacement and periodic displacement， and the GRU dynamic prediction model is constructed to predict them. Meanwhile， the deterioration of shear strength parameters， rainfall， reservoir water level and historical displacement are considered as influencing factors of periodic displacement through gray correlation analysis method. The hybrid model is finally used to predict the step-like deformation of the Baishuihe landslide. The results indicate that the proposed model can effectively predict the step-like deformation characteristics of reservoir landslides under periodic fluctuations of reservoir water level. The gradual weakening of step-like deformation is mainly associated with the progressive deterioration of landslide materials in the hydro-fluctuation zone over the reservoir operation period.