An iterative method to improve the calibration accuracy of flat-joint models: Catch-up penalty algorithm

Abstract

Parameter calibration is a critical step in accurately modeling using the discrete element method (DEM), but the time-consuming and complex calibration process limits the practical utilization of DEM. Herein, a catch-up penalty algorithm was proposed to simultaneously adjust multiple micro parameters of the flat-joint model through iterations. The effect of micro parameters on macro parameters was investigated by conducting 64 sets of orthogonal tests in PFC3D and analyzing the results by ANOVA. Regression analysis was used to establish the preliminary formulas for directly obtaining initial values of micro parameters and the trend equations for deriving iterative formulas. Based on the preliminary and iterative formulas, the calibration process for the algorithm was proposed, in which the micro parameters of each iteration can be calculated, thereby reducing researchers' dependence on the experience. The calibration capability of the algorithm was verified on four types of rocks, and the results showed that the average calibration error between the simulation results and the target values was reduced to within 5 % after six iterations, proving the reliability and applicability of the algorithm.