Inversion Analysis of Slope Engineering Parameters Using Back Propagation Neural Network Based on Strength Reduction Coefficients

Abstract

AbstractThis paper utilizes the back-propagation (BP) neural networks to simulate the inverse displacement analysis. The method is used to invert the slope geotechnical parameters (e.g., the internal friction angle and cohesion) based on finite element analysis. Based on field data from a slope support remediation project in Suzhou, Jiangsu province. The shear strength parameters of rock and soil were determined by back analysis method based on the strength reduction coefficient. The BP neural network was executed for training, testing, and performance evaluation of the numerical simulation data. A new feedforward BP neural network has been developed with three hidden neutrons. The stability of the slope body was evaluated, the safety factor of the slope and displacement under different parameter combinations were determined before and after treatment by using the strength reduction finite element method (SRFEM). The simulation prediction was used to obtain the calculated parameters to reflect engineering reality. The results show that the maximum error between the simulated displacement and the measured displacement was 3.90%, which was within the allowed limit.KeywordsStrength reductionFinite element method (SRFEM)Numerical simulationBack propagation neural network (BPNN)Safety factor