Numerical Analysis, Neural Network Forecast And Artificial Intelligence Repair Of Landslide Problem

Landslide is a main geological hazard and one of the important geological and engineering problems that have been under research over the world. Because its occurring conditions, factors and mechanism are variable and complex, it's difficult to predict it. And the repair cost is also very expensive. In recent 20 years, especially during the "International Decade of Natural Disaster Reduction", the research in the field has long-run development. On the basis of Bishop, FEM (Finite Element Method), DEM (Disturbing Energy Method), ANN (Artificial Neural Network) and AI (Artificial Intelligence), the thesis has study in details about the stability analysis, condition forecast, repair mode selection of the landslide, and obtain some significative conclusion.The basic process of FEM and the principle of nonlinear element method are simply presented. The analysis modes used by now are summarized, and the Duncan E-B model and the Cambridge model are introduced particularly. The thesis also introduces the non linear equations resolution methods.On the basis of predecessor work, the thesis developed BISl.O for complicated soil, SFEA1.0 based on E-B model,FRAL1.0 and BESL1.0 that are assistant programs for SFEA1.0,DIS1.0 based on DEM.A slope is studied according to Bishop, FEM and DEM in the thesis. The reasons that result in the different results by Bishop and FEM are discussed. The results of FEM are analyzed , and the action of the strain isoline ,stress level isoline, energy isoline and safety factor isoline is applied. In addition, the destroy process of slope is simulated by element regulation and repeat calculation. The last conclusion indicates that FEM is feasible, and DEM, which points out a new approach for the slope stability analysis, is reasonable.The slope condition forecast based on ANN is studied deeply. At first, the basic concept is introduced about the ANN. Secondly, the BP formula is proved, and the program structure is given. On the basis of BP model, the thesis designed a slope stability forecast model. And ANN1.0 is compiled for the model application. The model is practiced by many examples, and then it is tested by other examples. In a word, thefact show that ANN is a kind of new method for slope stability forecast, which can reflect a lot of nonlinear factors. The merit and demerit of BP model are also discussed in the thesis.The selection of slope repair method is studied on the foundation of AI. In the thesis, the corrective knowledge is presented. And the necessity and the feasibility of building expert system are talked over. As an adhibition, a minitype expert system named as SLOPE RE PAIR1.0 is built. The test example shows that it is logical.Based on modern program design theory, such as object-oriented, a visual slope analysis software is developed by Visual Basic6.0, AutoCAD, and PowerStation4.0. At last, the thesis point out that making full use of modern program design theory is very necessary for engineering software to convert into commodity software.