Researches On Evaluation Of The Slope Stability Based On Artificial Neural Network Techlonogy

With the development of larg-scale construction projects.more and more high and steep slopes appear in all kinds of construction domains. And these slopes are always the important factors which restricts whether the project is economical and reasonable or not. So it is of great significance to fully understand how to design a slope project or how to analyse and evaluate the stability of natural slopes economically and securely.Based on reviews of evaluation methods of slope stability, the thesis adopts artificial neural network (ANN) and RBF model to sensitivity analysis of the factors thich affect the stability of slopes. Furthermore ,the thesis combines ANN with Fuzzy and applies it to plote stability evaluation. The thesis is aimed at establishing intelligent assessment model for slope stability,while making full use of the characters of ANN including self-study,self-adaptive, self-organization and non-liner dynamics, and featuns of Fuzzy including uncertain analysis and multi-factor analysis.This thesis consists of an introduction to the significance of the study on slope project and a brief discussion on problems and development direction of it. Meanwhile it adopts ANN and fuzzy to the study of slope project, and puts an emphasis on the structure and learning algorithm of ANN BP model and ANN RBF model, with a brief presentation of the target determination, target grade and subordination function of Fuzzy as well. Combining the perpendicularity list with theory of utility function and ANN ,the thesis fully analyzes the sensitive ANN which influences the stability of slopes, and sets up ANN assessment model to achieve sensibility analysis. Moreover it fully discusses feasibility and scientificity of the theory mentioned above in sensitivity analysis through a huge number of experiments. Compared with the outcome through traditional limit balance analysis, this outcome proves the validity and convenience of the methods introduced in the thesis. In the assessment for slope stability, by adopting a new method to combine fuzzy with ANN technology, the thesis sets up the assessment models of fuzzy neural network for slope stability. And with the testification of applicability and validity of the assessment models, we draw a significant and meaningful conclusion.This research has got the pronotion of national nature-science foundation (No.49902022) and Huo Ying Dong education foundation (No.71018).