| description abstract | During the excavation of large-scale underground caverns, in which dynamic feedback analysis is required, the efficiency and accuracy in determining mechanical parameters of surrounding rock masses have significant influences on the safety and effectiveness of construction. In this study, a novel intelligent displacement back-analysis method is proposed to determine the geomechanical parameters. In this method, the parameter determination is transformed into a global optimization problem, which treats the error between in situ measured displacements and numerically calculated displacements as an objective function and regards geomechanical parameters as decision variables. To solve this optimization problem featuring high nonlinearity, multiple peak values, and high computation cost, an intelligent optimization algorithm combining the particle swarm optimization (PSO) technique and the Gaussian process machine learning (GP) theory is developed, and then, the algorithm is used to cooperate with the finite difference method (FDM) to form the method called PSO-GP-FDM for displacement back-analysis. Subsequently, the PSO-GP-FDM method is applied to the back-analysis of rock mass parameters for the Tai'an Pumped Storage Power Station. With the obtained mechanical properties of rock masses, the FDM-based numerical modeling can reproduce very well the in situ measured displacements in this hydropower station after excavation, demonstrating that the PSO-GP-FDM method is feasible to obtain reasonable mechanical parameters of surrounding rock masses. With excellent global optimization ability and high computational efficiency, the proposed method is suggested for displacement back-analysis of geomechanical parameters of underground caverns. | |
