DEM parameter calibration strategy applied to the strain-softening characteristics of sliding zone soil with the support of GA-BP

Abstract

The mesoscopic parameters of the discrete element model play a vital role in the precise simulation of the shear mechanical behavior exhibited by sliding zone soil. Presently, discrete element simulations of shear behavior in slip belt soils are mainly conducted via triaxial compression tests for the calibration of fine-scale parameters. It has not been found that the parameters are calibrated directly from the results of ring shear tests which are becoming increasingly widely used to study shear mechanical behavior. This study introduces a novel approach for the calibration of discrete element parameters, employing Genetic Algorithm- Back Propagation to effectively simulate the strain softening behavior of sliding zone soil in ring shear tests. The samples utilized in this research are generated through orthogonal design and three-dimensional particle flow code. The Genetic Algorithm- Back Propagation neural network training data were used to establish a nonlinear mapping relationship between the macro and fine mechanical parameters of slip belt soils, and the genetic algorithm was used to find the fine optimal parameters. The indoor ring shear tests were conducted and then compared with the Genetic Algorithm- Back Propagation model inversion results. The findings indicate that the calibration method proposed in this study is capable of rapidly and accurately inverting the meso-mechanical parameters. Building on this, it has been demonstrated that this method is capable of effectively representing the strain softening behavior of rock and soil, thereby enhancing both the efficiency and accuracy of parameter calibration.