A Parallel CFD-DEM Coupling Model And Numerical Simulation Of Dense Particulate Two-phase Flows

Dense particulate flow is one important and frontier research subject of the gas-solid two-phase flow, whose main features are frequent inter-particle collision, high local volume concentration, and strong interphase coupling interaction. As the computational capacity is rapidly growing, numerical modeling has become a powerful tool to study the important properties of dense particulate flows. The most widely used simulating models are Eulerian-Eulerian model and CFD-DEM coupling model. The CFD-DEM coupling model which can track every individual particle and take full account of the inter-particle, particle-wall and particle-gas interaction, has been received a lot of research attention.The parallel technique for pseudo-3d Computational fluid dynamics-Discrete element method (CFD-DEM) coupling model is implemented by User Defined functions (UDFs) based on the MPICH2 of FLUENT, and the validation is conducted by use of experimental results available in the literature. Subsequently, the numerical simulation of the fluidized beds with single immersed tube and with multiple immersed tubes are carried out by the proposed model.Firstly, the multi-side strategies of implementing the CFD-DEM coupling model are presented.(1) The governing equations of the gas phase mainly follow Model A that is based on pressure gradient force model. By constructing the voidage scalar field and reorganizing the governing equations of the gas phase, Model A can be obtained via modifying the single-phase model and the convergence is improved.(2) The drag force is calculated based on the balance of the drag force and the coupling between the two phases in the TFM, in which the interphase momentum transfer coefficient is obtained according to the Huilin-Gidaspow correlation. The gas-solid two-phase interaction is in accordance with the Newton’s third law.(3) The voidage is determined though the transformation of the pseudo-3D voidage obtained by the area occupied by the particles in the cell proposed by Hoomans [38].(4) In the module of particle collision, the detection of collision events is conducted only for the particles located in the neighboring cells, witch has the efficiency of O(NT).(5) The CFD-DEM coupling model on unstructured quadrilateral mesh is developed. Wherein the linked list and self-definition structure are used to store the particle physical information, memory addresses of potential collision partners and the memory addresses of particles in the specified grid. This CFD-DEM coupling model is flexible and can be used to simulate dense particulate flows taking place in more complex geometry without modifying the code.Secondly, the one-dimensional domain decomposition technique is employed to parallelize the DEM algorithm and several indicators are introduced to evaluate the CFD-DEM model. Finally, the collision module, the particle motion module and the information passing module is tested.(1) In order to implemente the parallel CFD-DEM coupling model in the massively cluster systems, the object of developing parallel CFD-DEM coupling model under MPICH2 of FLUENT is established.(2) The speed-up ratio and the scalability are chosen to evaluate the CFD-DEM coupling model.(3) The one-dimensional domain decomposition technique is employed to parallelize the DEM algorithm. In the parallel CFD-DEM coupling algorithm, the information transfer of Euler parameters between neighboring sub-domains is realized by the standard module of FLUENT, while the information transfer of particle information is completed by the message passing macros.(4) Simulation results show that the collision module, the solid motion module and the information passing module are reliable.Subsequently, numerical simulation of the gas-solid two-phase flow, bubbling and particle mixing behavior is conducted by the developed parallel CFD-DEM coupling model. The parallel simulations predict almost the same results compared to the serial simulations with good speed-up characteristic and scalability.(1) The continuously bubbling fluidized bedAs the superficial gas velocity increases, the pressure drop first increases and then decreases with increase in the superficial gas velocity. The minimum fluidization velocity and the pressure drop corresponding to the minimum fluidization velocity agree well with the experimental values. The time-averaged particle velocity vectors obtained by our simulation agree well with those obtained by Tsuji [40].When Uf=2.6 m/s, the fluctuation of solid flow can be observed and the period is 420-460 ms, slightly lower than the experimental value of 480 ms.(2) The conical-base spouted bedTwo distinct regions can be identified:the start-up and stable fluidization stages. In the stable fluidization stages, an oscillation cycle period can be observed. The characteristics of the time-averaged particle velocity are obtained on the basis of the instantaneous particle velocity field:① The spout width increases with increase in the height and the difference at the upper end of the spout between the simulation result and the experimental result is 12mm. ② The particles entrained from the annulus rapidly accelerate in a short distance and then slowly accelerate to the maximum vertical velocity. Subsequently the particles enter into the fountain region. Particles reach the maximum altitude of 180 mm, slightly higher than the experiment value of 155 mm.③ In the spout, the particle vertical velocity reaches the maximum at the axis and then decreases along the radial direction. ④ In the annulus, with the increase of the radial distance, the particle vertical velocity in 1 cm rapidly increases to then maximum value, and then gradually decreases, but particles do not stagnate near wall.(3) The pulsed jet fluidized bedIn the bubbling process, the bubble continuously grows and small-scale vortices on both sides of the mainstream gradually evolved into two main vortices as the gas enters into the bed. When t= 150 ms, the bubble shape obtained by the simulation is "chunky" type, and the height where the particle vertical velocity is 0 is 9 mm, slightly lower than the experimental value of 10 mm. After a single bubble had passed through the bed, more particles in the bottom zone give a triangular shape to the top of the peak. The maximum rising altitude is slightly higher than the experiment value, but similar to the result obtained by Bokkers et al. t51^, mainly due to the overestimated drag force.Finally, numerical simulation of the particle mixing processes in the fluidized bed with immersed tube and the fluidized beds with immersed tubes are carried out by the developed parallel CFD-DEM coupling model through dividing the particles into two groups. The gas-solid hydrodynamics and particle mixing mechanism were explored, and the erosion properties of the immersed tube were surveyed.(1) The fluidized bed with immersed tubeThe immersed tube causes bubble coalescence and breakage. The preferential path of bubble motion mainly lies around the tube rather than near the walls. The radial distribution of bubble phase becomes more homogeneous with increase in the superficial gas velocity. The radial heterogeneities of the solid flow and distribution are exhibited:a core dilute region with upward flowing solids and a high concentration of solids flowing downward near the walls. Both of the existence of immersed tube and the increase in superficial gas velocity are helpful for particle mixing and for decrease in the time for full mixing status. The erosion quantity mainly depends on the impact frequency and particle velocity.(2) The fluidized beds with immersed tubesThe tube bundle causes bubble coalescence and breakage and the preferential path of bubble motions mainly lies in the central region of the bed rather than near the walls. The radial heterogeneities of the solid flow and distribution are exhibited:a core dilute region with upward flowing solids and a high concentration of solids flowing downward near the walls. The superficial gas velocity has a significant effect on the fluidization behavior and the particle mixing characteristic. As the superficial gas velocity increases, the radial distribution of the bubble phase becomes more homogeneous and the time need for full mixing status decreases. Tube configuration also has a significant effect on the fluidization behavior and the particle mixing characteristic:bubbles in inline tube bundle are incline to be enlarged and rise vertically through the lane between inline tube columns and the stagnant zone for bubble phase is more easily formed at the back of the inline tubes; the time required for the bed with staggered tube bundle to reach the dynamic equilibrium mixing stage is longer because the staggered immersed tubes further hinder the vertical movement of solids. Different erosion patterns appear for the tubes locating in different regions of the systems and the erosion rate quantity mainly depends on particle collision.