New advances in large scale industrial DEM modeling towards energy efficient processes
Patrick Pizette  1@  , Nicolin Govender  1, 2@  , Daniel N. Wilke  3@  , Nor Edine Abriak  1@  
1 : IMT Lille Douai, Univ. Lille, EA 4515 - LGCgE – Laboratoire de Génie Civil et géoEnvironnement, département Génie Civil & Environnemental  -  Site web
Ecole nationale supérieure Mines-Télécom Lille Douai
F-59000 Lille, France -  France
2 : Center for High Performance Computing, CSIR Pretoria
Pretoria 0001, South Africa -  Afrique du Sud
3 : Department of Mechanical and Aeronautical Engineering, University of Pretoria

Granular material processing is crucial to a number of industries such as pharmaceuticals, construction, mining, geology and primary utilities. The handling and processing of granular materials represents roughly 10% of the annual energy consumption [1]. A recent study indicated that in the US alone, current energy requirements across Coal, Metal and Mineral Mining amounts to 1246 TBtu/yr, whereas the practical minimum energy consumption is estimated to be 579 TBtu/yr, while the theoretical limited is estimated around 184 TBtu/yr [2]. It is evident that design modification allowing for process optimization can play a significant role in realizing a more energy efficient industry sector that can have significant implications on the annual global energy demands.

The status quo in industry when facing the complex physics governing granular materials, is that current industry developed strategies to handle granular materials remain overly conservative and often energy-wasteful to prevent or reduce industrial-related bulk material handling problems like segregation, arching formation, insufficient handleability. Granular scale approaches have also been developed to both understand the fundamental physics governing granular flow and to study industrial applications, especially to improve the understanding and estimation of energy dissipation and energy efficiency of granular flow processes. The Discrete Element Method (DEM) proposed by Cundall and Strack [3] is starting to mature and evolve into a systematic approach to estimate and predict the response of granular systems. However, DEM is computationally intensive and is limited by the number of particles that can be considered realistically are limited to hundreds of thousands or low millions. However, before DEM can be practically considered for industrial applications the number of particles need be increased to tens of millions particles for a sufficient amount of processing time.

This study discusses new advances and perspectives made possible by the Graphical Processor Unit (GPU) when simulating discrete element models, specifically for granular industrial applications. Attention is specifically focussed on the newly developed BlazeDEM3D-GPU framework for an industrial flow investigation [4]. Note that BlazeDEM3D-GPU is an open-source DEM code developed by Govender et al. [5] that has been validated for industrial ball mill simulations and hopper discharge applications using ten of millions of particles using a single NVIDIA GPU card on a desktop computer [4, 6].

The industrial granular flow investigation considered in this study is of the storage silos located at the industrial concrete central in France. The typical silo diameter is 8m with a height of around 17m. Three dimensional DEM studies were been performed to investigate the influence of particle sizes and inter-particle cohesion on the bulk flow rate and induced shear stresses for various hopper designs located at concrete central. As required for this industrially relevant application, up to 32 million particles were required to be simulated within a reasonable computing time. These simulations were performed within these requirements but only made possible by the utilization of GPUs. These results show that the GPU computing allows for realistically relevant number of simulated particles for the 3D DEM applications within a reasonable time frame. This makes large-scale analysis practically relevant but more importantly allows for a number of analyses to be conducted to steer granular processing solutions towards an increased efficiency in energy utilization.


Personnes connectées : 1