Abstract:

In shearing strength theory for unsaturated soil, the suction friction angle is a key parameter, it is difficult to determine its value, which is limited by the test of equipment, technology and method. The influence factors for such as the soils′ basic physico-mechanical properties and the test method are analyzed detailedly. At the same time, by using qualitative analysis and quantitative calculation methods, it is considered that those factors have a certain degree of nonlinear relationship, which includes dry density, water content and internal friction angle. Based on artifical intelligence——the BP neural network, the 4:9:1 network structure is built to calculate and predict the suction internal friction angle for unsaturated soils. The validity of the model shows that the precisions of simulation and prediction are high. This model can be applied to forecast the index of the suction internal friction angle .

Key words: suction internal friction angle;BP neural network, TRAINGDX training function, shearing strength