Research On Slope Displacement Prediction Based On Combination Model Of Intelligent Algorithm

Landslide disaster caused irreversible loss of life,property and ecological environment all over the world.How to construct an efficient landslide assessment and early warning and forecast system,make use of the appropriate prediction technology,take effective preventive measures to predict the landslide,reduce or avoid the impact of the landslide will be our first consideration.The thesis analyzes the monitoring data such as slope displacement data,and uses intelligent algorithms such as gray model,support vector regression model,wavelet neural network,particle swarm optimization to construct a scientific and reasonable prediction model.Combined with the actual slope engineering,study the mechanism of slope deformation,grasp the law of slope deformation,and provide guidance for the prevention and control of landslides.The main research contents and results are as follows:(1)The present situation and some problems of landslide monitoring are briefly analyzed.Through the study of the monitoring data,the monitoring data of slope displacement which can reflect the deformation characteristics of slope directly and easily obtained are selected for analysis,and the internal mechanism and variation law of slope deformation are explored,so as to realize the accurate prediction of landslide.By analyzing the modeling process of grey model and support vector regression model,the advantages and disadvantages of the model are summarized.Aiming at many problems in the model,optimize and improve it to meet the needs of actual engineering.The background value and initial conditions of the grey model are improved and optimized.Based on this,using PSO to determine the three-parameter variable weight buffer NGM(1,1,k,c),and considering the influence of different data intervals on the prediction results,a grey prediction model based on data fusion is established.With the improvement of the support vector regression model,the SVR model based on fuzzy information granulation,the PSO-SVR model and the WNN-SVR model are proposed.The traditional SVR model is improved in different aspects and the optimal parameters of the model are determined efficiently and quickly.the scope of application is wider,and the requirements of fitting and prediction can be met simultaneously.(2)In view of the complex uncertainty of slope deformation mechanism and the limitation of single model,the application of combined model in slope engineering is studied and extended.The combination model with higher precision and stronger applicability is constructed by using specific weight determination method,and it is successfully applied to slope engineering.The entropy weight method with more objective accuracy is used to weight the PSO-SVR model and the PSO-NGM model,and the PSO-SVR-NGM formed is better than the single prediction model in terms of deformation trend and fitting prediction accuracy.SVR-NGM-WNN optimal weighted combination model combines the PSO-SVR model,the PSO-NGM model and the WNN model through the optimal weighted combination method,which reduces the interference of the bad model and enables the model to fully absorb the advantages of each single model and make full use of the known information,and obtain more accurate prediction results.(3)Finally,based on the grey model,considering the effects of systematic errors and monitoring gross errors,a semi-parametric robust estimation model is established.And taking into account the influence factors such as temperature,rainfall,reservoir water level,etc.,a multi-factor semi-parametric NGM model based on robust estimation is constructed.Compared with the prediction model which only considers the influence factors of single displacement,its accuracy and rationality are higher and the scope of application is wider.Through the successful application in concrete slope engineering,the superiority of the model is verified,and the research direction of model optimization is expanded.