The deformation of deep excavation is a matter of nonlinear large deformation,which is closely related to construction process, loading and unloading ways. In this paper, the relationships of supporting plan of deep excavation with the many influence factors have been established by use of the nonlinear mapping function of artificial neural network, so that the very complicated problem of nonlinear large deformation of deep excavation is solved. A neural network model using for the optimized selection of supporting plan of deep excavation has been built by learning from a great deal of deep excavation cases. This model, tested by successful and unsuccessful deep excavation cases, has been proven that the model is feasible, and offers reference to the support engineering of future deep excavations.