The factors of surface settlement caused by shield construction are so much that leads to greater difficulties in calculating the surface settlement

and BP neural network can establish a nonlinear relationship between various factors and surface settlement better.In order to get a more accurate value of surface settlement

a BP neural network prediction software was compiled by using Miscrosoft Visual C # and Matlab program.Combining geometry parameters

formation parameters and shield construction parameters

a BP neural network model was established which is used to predict the surface subsidence caused by shield construction.Then the model was applied to the soft layer down laid on hard layer of Nanchang metro engineering

and the proportion of muddy siltstone of the working face was taken into account in this prediction model

at last the error between the predicted and measured values was compared and analyzed.The result of prediction is consistent with the actual surface subsidence

which indicate that the BP neural network model can be used to similar engineering practice.