Computationally efficient boundary representation for the particle-scale simulation of a centrifugal filter

  • Serper, Damla (Aalto University)
  • Hanley, Kevin J (University of Edinburgh)
  • Oinas, Pekka (Aalto University)

Please login to view abstract download link

Centrifugal filtration is a commonly applied separation method in industry. In this method, a suspension is fed into a rotating perforated basket. Consequentially solids move towards the walls and form a porous layer (cake) on the walls due to the effect of rotation, while liquid (filtrate) permeates through the cake. Cake formation in centrifugal filtration is a multi-scale phenomenon. The micro-scale behaviour of the granular material affects the macroscale cake formed, although the relationship between these scales is poorly understood [1]. Discrete element method (DEM) is a particle-scale simulation tool which is capable of providing information that is hard to obtain experimentally, such as details of the process of cake formation in centrifugal filters. It is commonly used in conjunction with computational fluid dynamics (CFD) to simulate multi-phase flow systems. Unfortunately, modelling cake formation with DEM presents computational challenges. A major challenge is the computationally efficient representation of a highly porous boundary. Conventionally these boundaries could be represented by a mesh, but at great computational cost. In this work, we have developed a novel boundary representation method to overcome this [2]. The mesh geometry is replaced with a solid cylinder within the selected DEM platform, LIGGGHTS. When a particle comes into contact with a region that has been designated as a pore, the contact force between the particle and the wall is disabled, allowing the particle to pass through. In this work, we tested different scenarios with the conventional and novel boundary representation methods on a supercomputer. The computation time was reduced up to 40% by adopting the novel boundary representation method [3]. This scale of time reduction could make DEM a more feasible option for investigating filtration processes in the future.