| description abstract | Due to uncertainties in the geological evolution process, engineering geological conditions, and nonuniqueness of the solution for in situ stress regression analysis, stress field regression has always been a difficult issue in geotechnical engineering. In view of this problem, a refined back-analysis method for the stress field is proposed in this study based on in situ stress and disturbed stress. First, the in situ stress and disturbed stress in the unloading zone are measured by field tests. The process of geological evolution and excavation disturbance is numerically simulated. The theoretical values of in situ stress and disturbed stress are obtained. The nonlinear mapping relationship of the boundary condition and the stresses at the measuring points is established by the evolutionary neural network. The measured in situ stress and disturbed stress in the unloading zone are imported into the neural network to obtain the boundary conditions. Finally, the boundary condition obtained by back analysis is assigned to the numerical model for calculating the stress field at the project site. The method proposed in this study was applied to analyze the stress field of Mengku Iron Mine in Xinjiang, China. It was found that the accuracy of back-analysis results has been greatly improved because the contributions of both the in situ stress and the disturbed stress were considered in the proposed method. Meanwhile, according to the engineering analysis, the maximum horizontal principal stress in the open-pit slope was reduced by a certain degree, with a maximum reduction of 30%–40% after mining. However, the major principal stress at 0–350 m below the mine pit increased remarkably, with an increment of above 60% compared to the in situ stress. It indicated that excavation unloading has a great influence on the local stress field. | |
