Discrete simulation of granular and particle-fluid flows: from fundamental study to engineering application

Abstract

Multiphase chemical reactors with characteristic multiscale structures are intrinsically discrete at the elemental scale. However, due to the lack of multiscale models and the limitation of computational capability, such reactors are traditionally treated as continua through straightforward averaging in engineering simulations or as completely discrete systems in theoretical studies. The continuum approach is advantageous in terms of the scale and speed of computation but does not always give good predictions, which is, in many cases, the strength of the discrete approach. On the other hand, however, the discrete approach is too computationally expensive for engineering applications. Developments in computing science and technologies and encouraging progress in multiscale modeling have enabled discrete simulations to be extended to engineering systems and represent a promising approach to virtual process engineering (VPE, or virtual reality in process engineering). In this review, we analyze this emerging trend and emphasize that multiscale discrete simulations (MSDS), that is, considering multiscale structures in discrete simulation through rational coarse-graining and coupling between discrete and continuum methods with the effect of mesoscale structures accounted in both cases, is an effective way forward, which can be complementary to the continuum approach that is being improved by multiscale modeling also. For this purpose, our review is not meant to be a complete summary to the literature on discrete simulation, but rather a demonstration of its feasibility for engineering applications. We therefore discuss the enabling methods and technologies for MSDS, taking granular and particle-fluid flows as typical examples in chemical engineering. We cover the spectrum of modeling, numerical methods, algorithms, software implementation and even hardware-software codesign. The structural consistency among these aspects is shown to be the pivot for the success of MSDS. We conclude that with these developments, MSDS could soon become, among others, a mainstream simulation approach in chemical engineering which enables VPE.

Virtual process engineering on a three‐dimensional circulating fluidized bed with multiscale parallel computation

“…The particle solver is very suitable for parallelization and has excellent scalability, while the computational cost of the fluid flow solver is still manageable as it operates at a scale much larger than the particles. More favorably, this distribution of the computational load among different numerical schemes fits well with the heterogeneous architecture of typical modern supercomputers . In fact, some early attempts to simulate the gas phase and solid particles with many‐core and general‐purpose multicore processors, respectively, have already achieved very encouraging performances.…”

“…The coupling of FVM and DEM consists of the following tasks: The movement of particles in DPM is governed by the Newton's second law, and most of the computational cost is for the particle‐particle and particle‐wall collisions and the fluid‐particle interactions. Because of the locality and additivity of these operations, the DEM part of DPM can be parallelized efficiently. The evolution of fluid is usually realized by solving the partial differential equations in iterative schemes, whose efficiency is greatly limited by the communications between subtasks in large‐scale parallel computing.…”

“…The particles are modeled by DEM and run on many‐core GPUs, with the codes developed based on our software DEMms (discrete element method for multiscale simulation) . It has already been demonstrated that the DEM simulation could be well accelerated by multiple GPUs because the “Single‐Instruction, Multiple‐Thread (SIMT)” architecture of the GPUs could be fully utilized as a result of the “locality and additivity” of the particle‐particle collisions . The main idea of accelerating DEM simulations using GPUs is to fully utilize the fine‐grained parallelism provided by multiple GPU threads.…”

An efficient and reliable predictive method for fluidized bed simulation

“…To further reduce the computation cost in CFD‐DEM, many real particles are lumped into a computation parcel, which has been shown to be efficient for the simulation of CFB riser, pneumatic conveying, cyclone separator, CFB riser, heat transfer, and chemical leaching process of rare earth elements . Also, this method has been reviewed in a recent publication and the influence of drag corrections, grid size, and parcel size are also analyzed . The parcel‐based method assumes that all the real particles inside a coarse grained particle move collectively and share the same forces.…”

Pore-scale direct numerical simulation of particle transport in porous media

“…Discrete element method, which was first introduced by Cundall and Strack as a useful tool to simulate particle interactions [12], has now been applied in many fields involving particle motions, especially in the study of fluidized beds in chemical engineering [13,14], energy industry and other industrial processes [15][16][17][18][19][20] In a DEM model, particle-particle contact forces and particle-wall contact forces, including elastic, viscous and sliding forces, are described using simple mechanical elements, such as springs, dash dots and sliders. The well-established Newton's second law is used to govern the motion of particles under the Lagrangian framework.…”