Nonlinear displacement back-analysis for deep excavation based on radial basis neural network

Based on the new earth pressure computation model the elastic subgrade reaction method was put forward to consider the nonlinear interaction between the supporting structure and soil. Then the radial basis neural network was introduced into the displacement back-analysis for deep excavation, the principle and methods of the displacement back-analysis for the sections chosen according to the behaviorl and spatial effect of excavation were studied, and the computer program was developed, which can be used to back-analyze the measured displacements at some places of the supporting structure in different construction stages. So the soil parameters gained by back-analysis embody the influence of the space-time effects and nonlinear interaction. The results of the practical application of this program to an engieering project show that the calculated and measured displacements of retaining wall agree with each other very well.