Rock Mechanical Parameters Based On Bp Neural Network Inverse Analysis

Mechanical parameters of rock mass of rock slope is of great significance in the study of slope problems such as the stability evaluation of sock slope. It's always an important issue for research workers about how to obtain the exact mechanical parameters of rock.This paper mainly studies how to obtain the mechanical parameters of rock mass by using the displacement back analysis method. The main contents include:in-situ landslide test, with FLAC3D software to simulate the excavation of the slope and with BP neural network for the analysis of the mechanical parameters of rock. Firstly, bury inclinometer tubes in specific locations of the area of in-situ landslide test and layout monitoring points of surface displacement to monitor the excavation of deep displacement and surface displacement in the digging process of slope. Secondly, simulate the digging process of slope by using FLAC3D software. As the model is complex, we choose modeling by ANSYS and then import into FLAC3D for numerical analysis. Based on structural characteristics of the slope, numerical analysis is done on constitutive model of transversely isotropic and the input mechanical parameters is modified by Uniform Design Experimentation to ensure that input parameters is uniformly distributed within the test and then the deep displacement of slope excavation which is calculated by different input parameters is obtained. Thirdly, on the basis of the input set of the slope deep displacement obtained by FLAC3D numerical analysis and the output sets of the back analysis of the corresponding mechanical parameters, the BP neural network is established and the network training and simulation is carried on. Finally, take the measured values of the slope deep displacement in in-situ landslide test as input set for back analysis, do back analysis on mechanical parameters and compare back analysis value of mechanical parameters with true value. The feasibility and reliability of using BP neural network to do back analysis of mechanical parameters of rock mass is verified, then a simple and efficient method for obtaining mechanics parameters of rock mass is provided.