Prediction Of Landslide Based On Theory Of Reconstruction Of Phase Space

Landslide is a disaster geological phenomena of the rock of the surface of the earth, and multi-happened as well. China is a country where landslide disaster occurred frequently and the disaster loss is very serious. Landslide calamity bring out great loss to the people's life and asset. Under this condition, if the deformation of landslide can be predicted timely and accurately, we may take measures and countermeasures against natural disasters as soon as possible so as to reduce the loss caused by these kinds of disaster to the lowest degree.This thesis has launched research to two key problems existing while landslide forecasting. One is how to effectively cull the pretended noise that affect the accuracy of landslide forecasting while landslide forecasting. Another is how to improve the forecasting accuracy of deformation landslide. In the course of studying, author introduce the theory of reconstruction of phase space and operation of blind signal, and apply FastICA denoising method to denoising for landslide displacement monitor curve and set up the immune support vector machine forecast model.The thesis discuss these problem:(1) Improve the method of simple average denoising and apply it to deal with landslide displacement monitor data.(2) Using the algorithm of FastICA to denoising on landslide displacement monitor data first time.(3) Combine immune clonal selection algorithm to optimize parameter of support vector machine. Construct the model of immune support vector machine regression and apply it to predict landslide displacement.At last, Danba landslide data are used to check up and analyze the effect of denoising methods and the immune support vector machine forecast model.