STUDY OF SEGREGATION IN A ROTATING DRUM FOR NONSPHERICAL PARTICLES

Abstract

In recent decades, there has been a growing interest in examining granular materials and their behavior when they move through different types of equipment. One particular aspect of interest is how these materials tend to separate from one another as they flow. This phenomenon, known as segregation, has captured the attention of researchers across various fields. To delve deeper into this complex topic, the thesis is structured into three main categories of investigation: experimental analysis, discrete element method (DEM) studies, and machine learning approaches. These distinct segments allow for a comprehensive exploration of granular behavior under different conditions and provide valuable insights into how segregation occurs and can be understood and managed. The overarching goal of the thesis is to untangle the complexities of segregation within granular materials, with particular emphasis on the behavior of non-spherical particles within a rotating drum. In the experimental analysis, the focus is on studying the influence of particle shape on the segregation of a bi-disperse mixture of particles in a rotating drum experimentally. Various particle parameters such as shape, size, density, and size ratio are investigated, along with system parameters like time and rotational speed. The results show that the shape of both coarse and fine particles influences mixing. For coarse particles, there is a decreasing trend in the extent of radial segregation as follows: sphere > oblate > prolate > elongated-needle. For fine particles, the trend follows as sphere > cube > prolate. These trends are strongly correlated with monodispersed random packing density and particle angularity. The results indicate that coarse particles with higher mono-dispersed random packing density show less segregation, whereas fine particles with greater angularity improve mixing. In the simulation work, the Discrete Element Method (DEM) is used to investigate the evolution of radial segregation and its dependence on shape by varying the particle’s aspect ratio of spherical particles (Type-I). A total of 37 different types of binary mixtures are created, where the aspect ratio of particles varies from 0.25 to 4.0. Granular material simulations have improved our understanding of granular flows, which are increasingly used in research and industry. The rise of the Discrete Element Method and various particle simulation software has played a key role in the development of the particle technology field, although there is still much room for progress. While it is possible to simulate the motion of particles with a high degree of precision, including the forces, velocities, and temperatures that affect them, many aspects of these simulations require further study. In this work, the open-source software package LIGGGHTS is used to investigate the segregation of granular particles in a horizontally rotating drum. The results show that in a binary mixture, the aspect ratio of both coarse and fine particles influences mixing. The initial well-mixed binary mixture of coarse and fine particles evolves into a radially segregated state, where the fine particles occupy a central core of the drum surrounded by coarse particles. Spherical particles, whether acting as coarse or fine in a binary mixture, promote segregation. Two characteristics of particles, i.e., Monodisperse random packing density and sphericity, are used to explain the segregation trend and its relationship to aspect ratio. The study further extended and significantly broadened its scope by including new types of non-spherical particles, specifically cubes (Type-II) and cylinders (Type-III), commonly found in industry. A range of particle shapes, including sphere, prolate, cube, cuboid, and cylinder, is considered in this study, with aspect ratios varying from 1.0 to 5.0. This study evaluates how particle shape affects the segregation behavior of binary mixtures in a rotating drum. The Discrete Element Method (DEM) is used to investigate radial segregation and its dependence on particle shape by varying the particle’s aspect ratio. Two scenarios are examined: one with the shape of fine particles held constant and varying the shape for coarse particles, and another with the shape of coarse particles held constant and varying the shape for fine particles. These scenarios help quantify the relative mixing observed in the system for the considered shapes. The segregation index (SI) trend in terms of aspect ratio (AR) for coarse particles is as follows: AR (5.0) > AR (2.5) > AR (1). In contrast, for fine particles, the trend is AR (5.0) > AR (1) > AR (2.5). The results indicate that coarse particles with lower packing density exhibit higher segregation, while fine particles with pronounced elongation facilitate mixing. Additionally, machine learning is applied to rotating drum systems to thoroughly investigate the impact of particle level parameters (size, density, their combination, mass fraction) and system parameters (filling %, rotational speed, and baffle) on the segregation index within rotating drum is first assessed using Discrete Element Method (DEM). Later, a Machine Learning (ML) model is applied in conjunction with DEM to expand and fill in the parameter space for particle-level parameters in a computationally efficient way, providing accurate predictions of segregation in less time. The DEM results are validated by comparing them with experimental data, ensuring their accuracy and reliability. The results show that optimal mixing is achieved when the total filling percent in a system is 36.3 % while maintaining an equal proportion of particles. The highest level of mixing occurs at around 60 rpm, with fine particles concentrating near the drum’s core and coarser particles distributed around the periphery. The presence of 3 to 4 baffles optimally enhances mixing performance within the rotating drum. Four ML models are trained using data from DEM simulations to predict SI. An error analysis is performed to pick the best model out of the four ML models used in this study are Linear regression, Polynomial regression, Support Vector regression, and Random Forest (RF) regression. The analysis reveals that the RF model accurately predicts the SI. Using the RF model, SI can be reliably predicted for any value of the 7 features studied using DEM. An example 3D surface plot is generated by considering two (out of seven) of the most important particle level parameters: size and density. It shows that while both particle size and density contribute to segregation, variations in particle size appear to have a more pronounced effect on the SI compared to particle density.