Calibrating microparameters of DEM models by using CEM, DE, EFO, MFO, SSO algorithms and the optimal hyperparameters

• Published in: Computational particle mechanics Vol. 11; no. 2; pp. 839 - 852
• Main Authors: Wang, Min, Lu, Zhenxing, Zhao, Yanlin, Wan, Wen
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.04.2024
• Subjects: Classical and Continuum Physics; Computational Science and Engineering; Engineering; Theoretical and Applied Mechanics; Microparameter calibration; Particle Flow Code (PFC); Python script; DEM; Optimal hyperparameters
• ISSN: 2196-4378; 2196-4386
• DOI: 10.1007/s40571-023-00656-0

The Particle Flow Code is a typical DEM numerical software; however, the microparameters of DEM models need to be calibrated before numerical simulation. In most cases, the trial-and-error method is used; however, it takes a great deal of time and the results depend on the researchers’ experience. To address this issue, the cross-entropy method (CEM), differential evolution (DE), electromagnetic field optimization (EFO), moth–flame optimization (MFO) and salp swarm optimization (SSO) algorithm have been used for microparameters calibration. We provide a numerical simulation example to verify the validity of these microparameter calibration methods; it turns out that the number of iterations was large. To reduce the computational effort of obtaining suitable microparameters of the DEM model, we determine the optimal hyperparameters of the CEM, DE, EFO, MFO and SSO microparameter calibrating techniques. Through the analysis of the results, the number of iterations of these algorithms was markedly reduced. Considering the number of iterations, the number of hyperparameters and the results of numerical simulation, we suggest SSO algorithm for microparameter calibration. We also give another numerical simulation example to verify the validity of the proposed method. We found that the number of iterations for obtaining suitable microparameters was less than 100, and only 1 hyperparameter needed to be determined. Compared with previous studies, the number of iterations decreased remarkably.