Menü des aktuellen Bereichs:




Efficient Resolved Open-Source Simulation of Fluid-Particle Systems

During the last year, huge progress was made in the creation of a new tool called LBDEMcoupling for resolved simulation of fluid particle systems. This tool uses the lattice Boltzmann (LB) method instead of classical finite-volume CFD (FV-CFD) to simulate fluid flow. The LB method is very well suited to flows in complex, moving geometries because it does not operate on a complicated mesh but on a regular grid. LBDEMcoupling creates a coupling between two well-established open-source platforms: the DEM code LIGGGHTS [1] for particle simulation and the LB library Palabos [2].

Thorough testing of this coupled approach was performed. First, we tested the evaluation of drag and torque on a single fixed particle in several flow configurations and compared the results to values obtained by very finely resolved FV-CFD simulations. We found excellent agreement between the two approaches. Then, we considered a test case by Ten Cate et al. [3]: a single sedimenting sphere in a tank of fluid. We compared our simulations to measurements published in [3] and found good agreement (see figure 1).

We also tested the parallel efficiency of this new tool. For this, we set up a case of approximately 10000 spheres sedimenting in a tank of fluid resolved by a grid of 300x300x600 = 54mio cells (see figure 2). We found this rather large case to be feasible in approximately 30h on 32 cores of our cluster. Also, parallel efficiency was tested (figure 3) and it was found that the scalability of the coupling is comparable to the scalability of a pure LB simulation of comparable size.

First real-world applications of this tool will be: simulation of the behaviour of suspensions in narrow channels, and sediment transport in open channel flow.

[3] Ten Cate et al., Phys.Fluids 14, 4012 (2002)

Fig. 1: Velocity over time for a single sedimenting sphere: Lines are simulation results, markers show measurements from [3]. Deviations at the final stage of sedimentation are due to lubrication forces that were not included in our model.

Fig. 2: 10000 sedimenting spheres in fluid. Computational time on 32 cores of our cluster was about a day.

Fig. 3: Scalability of LBDEMcoupling compared to Palabos on our cluster.

(Philippe Seil, Supervision: Stefan Pirker)