Abstract:Improved back-propagation(BP) neural network is used to predict the deformation of deep foundation pit. A real-time predication model is established for deformation of deep foundation pit;and an effective predicting method called multi-step scroll real-time prediction method based on time effect is put forward. Making use of the Windows system platform,under the environment of MATLAB 7.0,adopting visual object-oriented programming technique,the real-time prediction program of the deep foundation pit deformation is developed. The result indicates that the model proposed here has fast approximation and high precision;and the predicted values agree well with the measured ones. The proposed method is a useful tool for deformation prediction of deep foundation pit.