Study On Landslide Time Series Of Phase Space Reconstruction Embedding Volterra Series Model

Landslides have had a serious impact on people's production and life.Landslide prediction and forecasting research is the core issue of landslide hazard analysis,starting in the 1960 s.In recent years,nonlinear data processing methods such as fuzzy mathematics,artificial neural networks,chaos theory,grey theory,and wavelet theory have been widely used in the study of landslide time series.Small sample and complex nonlinear landslide time series are the most difficult points of research.In this paper,the project aims to effectively predict the complex nonlinear landslide time series using small samples.The theory and specific engineering practice are combined to study the time series of Shuping landslide.The phase space reconstruction technique can maintain the internal dynamics of the landslide time series under the condition of topological equivalence,and can effectively extend the dimension of the original data and reconstruct the original time series to be equivalent to its dynamics.In high-dimensional phase space,data processing and prediction are performed in this space.The volterra series exhibits excellent nonlinear prediction performance in mathematical simulation,and can express the structure of the sum function explicitly,with both linear and nonlinear characteristics.Based on the theory of phase space reconstruction,the traditional method is to re-enter high-dimensional data reconstructed from phase space into various prediction models or various combined prediction models.This paper proposes a phase space reconstruction embedded in volterra level.The model of the number is used for the prediction of the landslide.The method mainly embeds the delay time and the embedded dimension obtained by reconstructing the phase space into the mathematical model of the volterra series.The phase space reconstruction can be used to expand the data dimension.And the volterra series can be nonlinearly predicted,and the traditional two steps are turned into one step,which improves the computational efficiency.According to the actual project situation,firstly,the GPS monitoring data of the surface displacement of the Shuping landslide in the Three Gorges reservoir area is determined by the phase diagram method and the largest Lyapunov exponent method,and then the chaotic landslide time series is directly input to the proposed The model is predicted and finally analyzed and compared by the experimental results.The results show that the phase spacereconstruction proposed in this paper is effective in embedding volterra series model,and the predictions of volterra series kernel function are optimized by LMS algorithm,particle swarm optimization(PSO)and improved particle swarm optimization(IPSO).The prediction accuracy of the model is gradually increased.The work of this paper provides a new idea for the prediction and prediction of landslides,and it can also be used for other predictions of small-scale nonlinear time series.