Calibration and intelligent optimization for DEM numerical parameters in heterogeneous rock mass

Abstract

Calibrating meso parameters is a crucial step in modeling rock materials using the Distinct Element Method (DEM). Improved calibration methods integrating machine learning and optimization algorithms have been proposed to revise the time-consuming conventional trial-and-error approach. This study introduces an innovative calibration method for heterogeneous DEM numerical parameters, optimized with the improved DBO algorithm (IDBO) utilizing Latin Hypercube Sampling combined with Gaussian Processes (GP-LHS) for initialization and hybrid strategies for iteration. Findings suggest the proposed enhanced heterogeneous DEM calibration method can accurately reproduce laboratory results and similar material experiments. Additionally, this method performs better than the DBO-DEM, PSO-DEM, and GA-DEM calibration methods in heterogeneous DEM parameters calibration speed with the same accuracy requirement during unconfined compression tests. The findings affirm the potential of the method for heterogeneous DEM numerical parameters calibration.