A systematic evaluation of DEM input parameters measurement and calibration

Abstract

Rotary drums are widely used in industry, however, determining granular dynamics within them is still a complex task, thus remaining a focal point of research over the years. In this context, experimental techniques require numerical support to obtain data that is not easily acquired experimentally. One well-established numerical technique for studying particulate systems is the Lagrangian approach embed with Discrete Element Method (DEM). Although, DEM has limitations related to the definition of input parameters. It is crucial that DEM input parameters be measured and/or calibrated to reliably represent the granular behavior. These parameters are related to particle properties such as density, Young’s modulus, Poisson’s ratio, among others, and particle interactions, such as coefficients of restitution, static friction, and rolling friction. This study explores the experimental determination of interaction parameters preceding the acquisition of DEM input parameters in a rotary drum. Statistical analysis highlights the importance of each parameter and their combined interactions with the static angle of repose. The granular materials used were: soybeans, glass beads, polyacetal, 6 mm steel, and 4 mm steel. The coefficient of restitution was determined via free-fall methodology using equipment that verifies particle trajectory, static friction coefficient through equilibrium between weight and resistance forces, rolling friction coefficient using a launching device on surfaces of different roughness, and static angle of repose in a rectangular box with sections where particles rested. After removing the partition, angles were measured at both top and bottom sections. Regression techniques using a neural network created in Python were employed for both measurements. Experimental investigations and numerical simulations using DEM indicated that the coefficient of restitution depends on the thickness and type of material of the flat surfaces, showing an asymptotic behavior with increasing surface thickness. The static friction coefficient increased on rougher surfaces and in particles with greater imperfections. Rolling friction coefficient also increased on rough surfaces. Analysis of the angle of repose revealed that granular properties influence the response, such as static and rolling friction coefficients, whereas the coefficient of restitution had no significant influence. Neural network correlation coefficients exceeded 0.95, indicating a strong association between the variables. Our findings revealed that mathematical models can be used in conjunction with static repose angle experiments to calibrate DEM parameters for more robust and complex simulations. This research also quantifies and provides insights into the interactions between DEM parameters in direct measurement experiments, promoting more efficient calibration for future research.