Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
Granular media are omnipresent in the food industry, whether as raw materials, intermediate or final products. While a lot has been done already in modelling food processes dealing with liquids, solid and gases, very little exists until now in granular flow modelling. The present thesis takes up the challenge of producing realistic granular simulations via DEM and validating them experimentally. Indeed DEM is one of the most promising granular media simulation approaches, but its wide application is still restrained by the limited number of particles that can be considered and the small collection of comparisons with experimental data. Yet it will be shown in this thesis that DEM is indeed a powerful tool for realistic modelling. We focus on two granular systems relevant to the food industry, for which we carry out in parallel simulations and physical experiments. First we study the size segregation of cereals in a vertically vibrated cereal box. Three configurations are tested experimentally, and compared with simulations achieving a qualitative validation. The importance of convection rolls and the role of particle and wall friction coefficients is illustrated. We extend the study to the ordering behaviour of elongated grains under vertical vibrations. We show that this phenomenon is intrinsic to elongated grains and takes place also in the absence of help from sidewalls. We study the role of particle geometry and vibration acceleration and we provide a novel interpretation in terms of available kinetic energy and relevant potential energy barriers. Building on these results, we study and interpret the shape segregation of rods from spheres. Experiments and simulations fit together nicely. The second food application concerns the flow and dosing from vending machine canisters. Our three dimensional DEM simulation unveils a rich microcosm of particle interactions and provides in particular an unexpected explanation for the decreasing dose mass over time. To make this simulation possible we developed an algorithm to detect the collisions between spherical grains and the helix auger. A dosing experiment using glass beads is successfully compared with a simulation in terms of the evolution of both the surface shape and the dosed mass. Good agreement is reached. Further dosing experiments are performed using a cohesive beverage powder. The effect of the dosing screw design on the flow in the canister is quantified and a dosing coil improving the powder withdrawal across the canister is designed and successfully tested. The quantitative comparisons between experiments and simulations achieved in this work and the insight gained from such simulations show clearly the benefit for the food industry stemming from a coupled use of DEM simulations and experiments.
Andrei Variu, Cheng Zhao, Yu Yu, Hanyu Zhang