Slope Displacement Prediction And Instability Characteristics Analysis Based On CEEMDAN-LSTM

In the world,China is one of the countries with the most serious geological disasters,especially landslide disaster.Every rainy season,the loss of life and property caused by landslides is huge.In recent years,with the wide application of Internet of Things,sensors and other technologies in landslide monitoring,automatic monitoring of landslide deformation has gradually developed.Greatly improve the monitoring efficiency.With the improvement of monitoring efficiency,the amount of data acquired also increases sharply.How to make full use of monitoring data to predict the displacement evolution of landslides in the future is of great significance for landslide early warning.Therefore,more and more scholars are applying neural network and artificial intelligence to automatic monitoring system of geological disaster.Through the collection and mining of geological data,high-precision geological disaster prediction model is used to predict the deformation of landslide,so as to achieve the purpose of landslide warning and evaluation.Finally,this paper proposes a new multi-variable joint prediction model for landslide displacement,The algorithm is applied in soil slope engineering,which makes full use of the previous monitoring data to predict the landslide displacement in the future,and has been successfully applied in a practical case of Huaixin Expressway.This paper mainly carries out relevant research work from the following aspects:(1)In this paper,the data collected from a slope monitoring point of Huaixin Expressway is excluded from abnormal data and interpolated with missing data,and the monitoring data is improved to ensure the accuracy of the input data in the subsequent model training process.(2)Based on the geological overview of the study area,the numerical calculation model is established by Using Solution software,and the slope seepage effect under different rainfall duration conditions is numerically simulated,and the results show that with the increase of rainfall time,the slope safety factor will gradually decrease,and the analysis believes that rainfall is the main cause of landslide failure.(3)A Complete Ensemble Empirical Mode Decomposition with Adaptive Noise is proposed.CEEMDAN and Long short-term Memory network(LSTM)algorithm.Firstly,the landslide cumulative displacement is decomposed into trend term,random term and periodic term by CEEMDAN,and then the trigger factor database is established.The selection of relevant features is determined by the analysis of grey correlation degree,and then the LSTM model is used to predict the landslide cumulative displacement,the final prediction model is obtained.(4)Finally,for the Huaixin high-speed slope,the monitoring data is used for landslide prediction,and the prediction accuracy is compared with the traditional LSTM training model.RMSE,MAE,MAPE and~2 of CEEMDAN-LSTM model proposed in this paper are 3.53mm,3.48mm,0.087 and 0.96 respectively.However,the RMSE,MAE,MAPE and~2of the conventional LSTM model are 16.77mm,12.66mm,0.237 and 0.80,so the accuracy of the comprehensive algorithm proposed in this paper is better than that of the LSTM prediction algorithm.In order to further verify the accuracy of the algorithm,the CEEMDAN-LSTM model proposed in this paper is highly accurate by comparing and analyzing the prediction models of different monitoring points and different research areas in the same research area.