Abstract: The stability of surrounding rock of tunnel can be effectively identified by deformation prediction, which is a crucial basis for informatized tunnel construction. An improved prediction method combining local weights and global weights is proposed to improve traditional methods. The proposed method is simple, practical, and also reflects the impacts of sample data and node length on combinatorial weights. Meanwhile, ideas of cumulative superposition and multiplicative superposition are put forward in consideration of the influences of local and global factors on the combinatorial weights. Engineering case study verifies that the proposed combinatorial forecasting model is of higher prediction accuracy than traditional forecasting model, with the relative error less than 2%; in addition, multiplicative superposition is of better effect, higher accuracy and stability than cumulative superposition.

Key words: railway tunnel, local weight, global weight, combinatorial forecasting, deformation prediction