A back-analysis method using an intelligent multi-objective optimization for predicting slope deformation induced by excavation

摘要

A multi-objective inverse analysis method for slope excavation was proposed, in which orthogonal design, numerical simulation, back propagation neural network (BPNN) and elitist non-dominated sorting genetic algorithm (NSGA-II) were integrated. The multi-objective model is constructed by minimizing a set of multi-objective error functions between the time series of observations and corresponding calculated values. Compared with the back-analysis methods that uses traditional algorithms, the proposed method is validated by a numerical example to be more effective for multi-objective optimization. The methodology is also applied to the excavation of a right bank slope at the Dagangshan hydropower station located in the Sichuan Province, China. The obtained inversion parameters are used in forward analysis to predict displacements. In this case application, three types of field observations are used simultaneously in the back-analysis, which include displacements in the Dadu River water flow (y-) direction, transverse (x-) direction and vertical (z-) direction. Compared with the field displacement data, the trend of the predicted displacements agrees well with the measurements. The results indicate that the proposed method can more precisely and reliably predict the slope deformation induced by excavation.