Study On Deformation Evolution And Landslide Displacement Prediction Model Of Step-type Reservoir Bank Landslide In Three Gorges

The Three Gorges Reservoir area is a typical geological disaster prone area in China,which is located in the upper and middle reaches of the Yangtze River,with complex geological structure and strong new tectonic movement.Since the impoundment of the Three Gorges Project in 2003,the rock and soil of the landslide body and the reservoir bank slope have been repeatedly affected by the periodic large fluctuation of reservoir water level and rainfall.This process has a great negative effect on the geological environment conditions of the surrounding areas,resulting in the original stable reservoir bank prone to deformation,even instability and collapse,seriously affecting the economic and social development of the reservoir area,threatening the safety of people ’ s lives and property along the coast and the safety of the Yangtze River shipping.In this paper,the step-type reservoir bank landslide in the Three Gorges Reservoir area is taken as the research object.Based on the relevant monitoring data,the deformation evolution characteristics and the influence of different factors on the landslide are explored.On this basis,the influencing factors of landslide deformation are reasonably selected;finally,the displacement prediction of step-type reservoir bank landslide is carried out.The main research contents and results of this paper are as follows :Firstly,according to the monitoring data obtained by the surface displacement monitoring system in Baishuihe landslide area,the deformation characteristics of Baishuihe landslide are further explored.The results show that the deformation of Baishuihe landslide mainly occurs from May to August every year,which is concentrated in the early warning area on the eastern side of the landslide.The landslide displacement curve shows a step growth.The deformation degree of the leading edge of the landslide is much larger than that of other locations,which is a traction landslide.Secondly,the influence of various factors on the deformation of Baishuihe landslide is analyzed,considering the different influence of three stages of reservoir water storage on landslide deformation,and then the quantitative analysis of external inducing factors is carried out.The results show that the dominant factors of landslide deformation are different at different water storage stages.Landslide deformation is often accompanied by rapid decline of reservoir water level and the response of landslide displacement to reservoir water level change has a certain time lag;Baishuihe landslide deformation is significantly correlated with rainfall,and the response of landslide deformation to rainfall is lagging.Through quantitative analysis,it is found that the rainfall in the previous month and the average reservoir water level in the previous month are the two influencing factors that have the greatest correlation with landslide deformation in different sets.The influence of reservoir water level factors is mainly reflected in the period of reservoir water level decline.The water level difference between the inside and outside of the slope increases gradually,resulting in the seepage pressure towards the outside of the landslide body,which affects the stability of the landslide.At the same time,the landslide body is soaked in river water,which weakens the shear strength of the landslide front and increases its own weight,and also seriously affects the stability of the landslide.The influence of rainfall factors is reflected in the period of heavy rainfall in the flood season.The overall water content of the landslide increases,which will be accompanied by the increase of pore permeability pressure and the decrease of shear strength parameters,and ultimately lead to the deterioration of landslide stability conditions.Human engineering activities will destroy the original surface vegetation,change the topography of the slope,and induce local deformation and failure of the landslide.The influence of internal geological factors is reflected in the effective free surface of Baishuihe landslide which is easy to instability,and there are weak interlayers in the rock strata with low shear strength and poor anti-sliding ability,which are conducive to landslide deformation.Joints on the slope surface are more conducive to rainwater infiltration into the slope,which is not conducive to landslide stability.Thirdly,the grey correlation analysis and correlation analysis are used to quantitatively study each alternative influencing factor and determine the influencing factor of the final input model.Fourthly,a GRU-GAN landslide displacement prediction model based on wavelet transform is proposed and verified.Through analysis,it is found that the existing landslide displacement prediction methods are faced with some research challenges.First of all,most of the current popular landslide prediction models are static prediction models,which are difficult to capture the dynamic information in the time series data of landslide displacement,and thus cannot well describe the dynamic change characteristics of landslide.Secondly,the shallow machine learning method adopted by some scholars is difficult to explore the potential law between influencing factors and landslide displacement.Finally,a single deep learning model has more or less certain limitations.The landslide evolution process is the result of the combined action of various internal and external influencing factors,which has high uncertainty,nonlinearity and complexity,making the single neural network model not necessarily applicable.In order to cope with the above challenges,this paper proposes a landslide displacement prediction model based on wavelet transform combined with Gated Recurrent Unit(GRU)and Generative Adversarial Network(GAN),which is called GRU-GAN model.Firstly,the landslide time series is studied and analyzed based on the monitoring data.The landslide displacement data is decomposed into low frequency component and high frequency component by wavelet transform method which can highlight the fluctuation information of data.Secondly,the Long Short-Term Memory(LSTM)model and GRU-GAN model are constructed to predict the displacement of trend term and periodic term respectively,and then the predicted cumulative displacement of landslide is reconstructed.Finally,the model is verified,and the results show that compared with other models,the model used in this paper has higher prediction accuracy.