(723b) Establishing Continuum Model Segregation Parameters for Practical Particle Mixtures

Recent developments in advection-diffusion based continuum models of particle size or density segregation could make it possible to accurately predict segregation across scales, from bench top to plant scale. In these models, the degree of segregation in a flowing granular material results from a competition between advection, segregation, and diffusion. Previously, discrete element method (DEM) simulations were used to determine values for the segregation and diffusion parameters (segregation length scale S and diffusion coefficient D) and for the flow kinematics (in particular the flowing layer thickness δ) necessary to solve the continuum model in simple geometries, but only for mixtures of relatively large particles resembling spherical glass beads. Here, we extend the applicability of the continuum model to actual physical particle mixtures by implementing a method to directly determine the continuum model parameters from a small scale experiment. In this method, a heap is formed in a quasi-2D container by adding the particle mixture of interest at a controlled flow rate, concentration profiles of the material deposited on the heap are measured, and continuum model parameters (S, D, and δ) specific to the mixture being tested are determined by minimizing the error between the experimental concentration profiles and those predicted by the continuum model. To validate this parameter estimation method, particle concentration fields from DEM simulations in a similar geometry were used with the same minimization approach to determine the model parameters, which were then compared to the parameters measured directly from the DEM simulations. Because there can exist nearly identical solutions to the continuum model with different combinations of the three model parameters (S, D, and δ), it is necessary to provide one of the parameters as an input. Since it is relatively easy to determine the flowing layer thickness δ experimentally, we specify δ and then determine S and D by parameter fitting. Using this approach, values of S and D calculated using the minimization approach are unique and match measured values from the DEM simulations. In combination with geometric specific flow models, S and D can then be used in large scale continuum simulations to predict segregation of the given particle mixture in various plant-scale processes. Funded by The Procter & Gamble Company.