Laboratory of Civil and Mechanical Engineering
In order to complete the experimental study of flow of particles through a porous medium, I couple results with numerical simulations performed with softwares which were developped during my PhD. These softwares (PMLG3D and P3Ded) are based on Discret Element Methods: molecular dynamic of soft spheres and molecular dynamic of hard spheres (Event-Driven). The use of numerical simulation allows to modify the various physical parameters governing the transport process and to be able to measure some various physical properties which are not easily accessible experimentally.
To realize the disordered packings of spheres, used for modelling the porous structure and the packing of particles, I use an algorithm known under the name of Powell's algorithm. More informations about this algorithm can be found on this page. To avoid formation of locally ordered zones, we introduce a small dispersion on the diameter of spheres building up the porous medium. In order to avoid particles being trapped in the porous structure, all numerical simulations are done far away from the geometrical trapping threshold. More infomations on this threshold can be found on this page.
The dynamic part of the simulation is treated with two approaches. In a fisrt program, I use the event-driven method. By default, with this method, only binary and instantaneous collisions are considered. Multi-particle contacts can be handled with the improvement known under the name of the tc-model. There is no time integration in this method.
In a second program, I use the Molecular Dynamic of soft sphere method. This method is well suited to deal natively with multi-particle contacts. In this method, a time integration of particles postitions and velocities is required. This can lead in higher CPU time usage compare to the previous method explained above. In order to remedy to this, several optimizations are implemented in the program PMLG3D. For example it uses two grids of cells and doubly-linked lists to achieve desired efficiency during the seek of interaction partners. The program is also parallellized with the MPI (Message Passing Interface) library.
Simulations results are in full agreement with experimental ones and numerical modelization permits to extend the experimental study to the mixing of particles. Some pictures and movies taken from these simulations are visible in the gallery.