Calibration of DEM direct shear box test model with hulled millet and viscoelastic contact model using a genetic algorithm

ABSTRACT One of the principal challenges currently facing those who use simulations based on the discrete element method is to calibrate the micromechanical parameters of a contact model with the necessary degree of accuracy. This process may be conducted through a trial-and-error approach, through manual or parameter sensitivity calibration, or even through the utilisation of a calibration algorithm. The latter method is becoming increasingly popular among simulation professionals, as it greatly reduces the time required for calibration and produces more accurate micromechanical parameters. One disadvantage is that it is frequently time-consuming and labour-intensive to link the calibration algorithm to the simulation method. Furthermore, multiple correct micromechanical parameter settings can be produced, leading to parameter ambiguity. In this research, a genetic algorithm was used to calibrate the micromechanical parameters of hulled millet using laboratory direct shear box test. Discrete element simulations were conducted using a developed contact model that incorporates linear springs, viscous damping, and numerical particle shape distortions. Validation of the simulations was conducted through comparison with direct shear box test measurement results for three preloads. The objective function of the genetic algorithm considered the shape of the increasing slope of the shear force curve, thereby enabling calibration of the model not only to the maximum shear force value. The algorithm was employed to calibrate two distinct combinations of micromechanical parameters. In the initial case, the particle elasticity modulus (19.1 MPa), the particle-particle friction angle (23.0°), the viscous damping coefficients considered equal (0.0531) and the resistance coefficients treated as equal (0.3862) were calibrated. In the second case, the particle-particle friction angle (20.9°), the normal (0.7452) and tangential (0.6059) viscous damping coefficients and the resistance coefficients treated as equal (0.4866) were calibrated. The two calibration processes produced two distinct sets of micromechanical parameters, leading to the issue of parameter ambiguity. Consequently, in addition to the quantitative analysis, qualitative analysis was conducted to evaluate the results and select the correct parameter set. The utilisation of the fabric tensor proved instrumental in differentiating between the parameter ambiguity results.