Abstract: Affected by stress distribution，groundwater，opening mode，etc.，the surrounding rock deformation of the deep buried tunnels show nonlinearity in displacement series height.Thus，the non-linear combination model based on the deformation information fusion was proposed.In this model，based on the individual forecast data of GM (1，1) and RBF，and with the use of the neural network，the weight of each model among the combined model was optimized to build a nonlinear combined forecast model.Then，the non-linear combination forecast model was applied to predict surrounding rock deformation in deep buried tunnels.Meanwhile，non-linear combination predictions were contrasted with simple average weighting combination and the optimal linear weighted combination.The results showed that compared with the traditional weighting method，the non-linear combination forecast model has a higher and more reliable precision，and owns a more stable prediction result.It is of a certain theoretical and practical significance in surrounding rock deformation prediction for deep buried tunnel.

Key words: Deep buried tunnels, Surrounding rock deformation, Non-linear combination, Weight, Prediction