Investigation of particle clustering in discrete element method simulations
Abstract
Granular flows in the chemical industry often have problems because the observed particle-phase phenomena that occur in them are not well understood. One example is that of particle clustering where, in the flow, particles temporarily group together and then come apart. Clustering, as observed in riser flow, has been studied using two-phase Computational Fluid Dynamics (CFD) models. However, many CFD models cannot properly model the behavior of a system where clustering occurs. Attempts have been made to incorporate the effects of clustering into such models. It is difficult to develop particle-phase closure relations for such models. Discrete Element Method (DEM) simulations represent a means by which to develop such closure relations because these simulation methods track every individual particle. Therefore, particle-phase stress results from them will incorporate the effects of both particle-particle collisions and of any microstructure formation that might occur. The remaining challenge is to determine how large of a system is required in order to accurately simulate asymptotically large systems. To investigate relatively large systems, a computationally efficient DEM hard-sphere simulation has been developed. Furthermore, simulations over a range of coefficients of restitutions, solids volume fractions and system sizes were performed. Results from these simulations show that the particle-phase stress increases at large system sizes after which it was initially thought to reach a plateau. Furthermore, for system sizes of 300,000 particles, the particle-phase stress still appears to be increasing. Band-like clusters form in systems where the additional increase in particle-phase stress occurs. Similar simulations are also conducted for bidisperse systems with similar results. Finally, after a certain system size, additional particle-phase phenomena are observed. First, the particle-phase stress appears to fluctuate during the course of the simulation. Second, for two randomly generated simulations with the same overall conditions, the particle-phase stress appears to fluctuate around different stress values.
Degree
Ph.D.
Advisors
Curtis, Purdue University.
Subject Area
Chemical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.