A Study On Landslide Prediction Based On Fuzzy Mathematics

Landslides are one kind of major natural disasters, and threaten the life of human and the safety of possession. Our government has cured some slopes, which can lead to serous effects, such as Lianziya risky rock mass. Because of the great number of landslides, we have not enough economic and technical force to comprehensively control about all latent risky landslides. Under the circumstance, predicting the landslide is another way to prevent and control landslide hazards, can make people take measures against them timely so as to reduce the loss caused by landslides to the lowest degree as soon as possible.Considering fuzziness of rock and soil engineering the theory of fuzzy classification and the theory of fuzzy optimal selection are applied in this field. Displacement curves of monitoring points of landslides can reflect their inner creeping mechanism. After classifying these curves we can get more valuable information, and be easier to predict the landslide. However there is not certain but fuzzy relation between all curves. So the fuzzy classification method is just used to deal with this question. Firstly, fuzzy similar matrix that reflects the fuzziness between the points is setted up by many methods. Secondly, according to the principle of direct classification, all monitoring points are classified. Furthermore, the failure type of landslide mass is known. Finally, the reliability of this method is certified through the application to Huangci landslide.Even though there is only 30 years' history since studying landslides, experts around world have presented more than 50 models, which have different confined conditions. So choosing which model is very critical to forecast a certain landslide. Based on studying the theory of fuzzy optimization systematically, the paper puts forward a fuzzy optimal selection method of landslide medium-long term prediction.Wugongling landslide, in Guishi Reservoir, Guangxi Zhuang Autonomous Region, is predicted in this paper. Firstly, all the monitoring points on the landslide mass are merged into three classes. Then we can easily determine the main sliding direction. Secondly, the relative optimal model chose from the given models by fuzzy optimal selection method, include Logarithm model, Verhulst inverse function model, Modified Normal Function model, BP neural network and Exponential smoothing method, is used to predict representative points.The main conclusions of this dissertation are: 1. to improve on the knitted-net method of direct cluster methods;...