A novel displacement back analysis method considering the displacement loss for underground rock mass engineering

Most back analysis methods for geotechnical engineering are based on the measured displacement. However, before the monitoring sections are assembled, the displacement-termed the displacement loss-has already been induced; this displacement is difficult to determine, and thus, it is not considered in the back analysis. In the present study, a novel displacement back analysis method considering the displacement loss is developed, that can obtain not only the reasonable mechanical parameters of rock masses but also the displacement loss. To reduce the computational cost of back analysis, a new hybrid optimization algorithm based on the Gaussian process (GP) and particle swarm optimization (PSO) is presented. The GP is used as an inexpensive fitness evaluation surrogate to predict the global optimum solution and accelerate the local search of PSO, which is employed to determine the best mechanical parameters for the model. Combined with FLAC(3D) numerical analysis, a novel back analysis method called GP-PSO-FLAC(3D) is proposed. The results of a case study demonstrate that this method can effectively predict more reasonable mechanical parameters and displacement loss using the monitored displacement. An engineering application in the auxiliary tunnel of the Jinping II hydropower station indicates that the elastic deformation of the surrounding rock increases rapidly after excavation, especially for deep tunnels, thereby resulting in a large displacement loss. The back analysis results for the main powerhouse of the Taian pumped storage power station indicate that the displacement loss also exists in engineering processes involving ordinary geostress conditions. Therefore, the displacement loss of a surrounding rock mass cannot be ignored in the stability evaluation or back analysis of underground engineering, especially for deep underground rock engineering.