3-D back-analysis of initial ground stress based on ANN-RBF

Most of the methods which used to estimate the initial ground stress are reviewed. A new approach integrating the theory of finite element method and theory of artificial neural network based on radial basis function (RBF-ANN) was suggested to simulated the initial ground stress. Over fitting could be avoided by using ANN-RBF during the back analysis process. The new approach based on the 3D-FEM integrated the advantage of other methods and took into account of the influence of diverse factors including fault on the initial ground stress. Therefore, back-analysis results are more reliability. Case study shows that the reasonability of initial ground stress simulated by using this approach. In addition, the faults have been considered in 3D-FEM model to evaluate the influence on the initial ground stress and it is helpful to understand the distributing law of initial ground stress.