Computational multi-scale modeling of super-dispersed multiphase flows

Computational multi-scale modeling of super-dispersed multiphase flows


Funded by: IMPACT
Postdoc: Thomas Weinhart
Supervisor: Onno Bokhove / S. Luding
Collaboration: Dr. A.R. Thornton


Dry granular avalanche flows are a common occurrence in both the natural geophysical environment and industry, and occur across many  orders of magnitude. Common examples range from: rock slides,  

containing upwards of 1000m^3 of material; to the flow of  sinter,  pellets and coke into a blast furnace for iron-ore melting; down to  the flow of sand in an hour-glass.

The dynamics of these flows are complicated by many factors; for  examples: polydisperity, variation in density, non-uniform shape,  surface contact properties, flow obstacles and constrictions, etc...    

Molecular Dynamics (DPM) algorithms are an extremely powerful tool to  investigate the effects of these and other factors and with the rapid  recent improvement in computational power the full simulation of the  

flow in a small hour glass is now obtainable. However the full DPM  simulation of real geophysical mass flow, will probably, never be  possible.

Continuum models are able to simulate the volume of real geophysical  flows, but have to make averaging approximations reducing the  properties of over a trillion individual particles to a handful of  averaged quantities.  Once these average parameters have been tuned  via experiment or historical data these models can be surprisingly  accurate; but, a model tuned for one flow configuration often has no  prediction power for other setup. DPM can be used to obtain the  mapping between the microscopic and macroscopic parameters allowing  determination of the macroscopic data without the need for a-piori  knowledge. In simple situations it is possible to pre-compute the  translation between the particle and continuum; but, in more  complicated situations heterogeneous multiscale modelling is required  (HMM). In HMM continuum and micro models are dynamically coupled with  two way feedback between the models.  The coupling is done in  

selective regions in space and time, thus reducing computational  expense and allowing simulation of the complex granular flows.

We start by consider the flow of granular material down an inclined  chute. The flow can be described by a macroscopic model with the  exception of the basal friction coefficient, which requires  microscopic modeling with a shorter spatial and temporal step size.

For the HMM the macroscopic behaviour is described by a Discontinuous  Galerkin discretization of the shallow water equations with unknown  bottom friction coefficient. A Discrete Particle Model is used to  compute the undetermined basal friction coefficient and hence close  the model. The microscopic model requires a short time scale, but is  assumed to relax rapidly in time. The model is implemented in hpGEM.

The model is tested against the Pouliquen-Jenkins flow rule for rapid  granular flow along an inclined chute with rough base. We simulate  granular flows through a contraction and show speed comparisons with  the microscopic model to demonstrate the effectiveness of the algorithm.

The DPM code developed as part of this project is available for public  use, for more information please via the codes website at



Conference items: