Optimal operation of process equipment requires a thorough understanding of the transport phenomena involved, including hydrodynamics. In this field of engineering, fluid flow simulation, i.e. CFD (Computational Fluid Dynamics), is an indispensable tool. Industrial flows, however, are a harsh environment for CFD. The flow geometry can be very complex, and the equipment is, for its vast majority of applications, operated in the turbulent regime. The classical CFD approach is based on the spatial (and for transient flows also temporal) discretization of the Navier-Stokes equations. In case of turbulent flows, transport equations for the Reynolds stresses, which include closure models, are added.
A novel approach to CFD was introduced by Frisch, Hasslacher, and Pomeau [1]. In their lattice-gas automata, a many-particle system conserving mass, momentum, and energy mimics fluid flow. Lattice-Boltzmann schemes, see e.g. [2], are closely related to lattice-gas automata. They have obviated some of the noise problems connected to lattice-gases. The numerical operations involved in lattice-Boltzmann schemes are simple, and more importantly, local. As a result, parallelization of the algorithms by means of domain decomposition can be achieved very efficiently. Furthermore, the efficiency of the scheme is not hampered by the complexity of the flow geometry.
Lattice-gas and lattice-Boltzmann schemes have been successfully applied to complex, low-Reynolds number flows (e.g. the flow in porous media, hydrodynamic interaction in suspensions). In combination with subgrid-scale turbulence modeling (i.e. Large Eddy Simulations), the schemes are also well applicable to turbulent, industrial flow systems [3,4]. In this paper, we present results of large-scale simulations (i.e. three-dimensional, time dependent simulations on typically 5*106 grid points) of flows in stirred tanks and cyclones. The former are widely used in industry as mixing devices, the latter as separators. To represent the rotating impeller in stirred tank flow, an adaptive force field algorithm has been developed [4].
For maximum portability, the SPMD computer code was designed to run on distributed memory platforms, e.g. workstation clusters. PVM was used as a message passing tool. The major results, however, were achieved on a shared memory system: a HP-Convex S-Class machine equipped with four PE's. The speed-up of the computer code on this system is almost ideal.
The flow field results that will be presented, give a very detailed view of the physical phenomena involved. In stirred tank flow, the vortex system behind the impeller blades, which is associated with high turbulent activity, and for a large part responsible for the mixing capacity of the device, has been accurately resolved. Furthermore, the three-dimensional, helix-like vortex core, characteristic for cyclone flow, has been very well reproduced by the simulations.
References
1. Frisch, U., Hasslacher, B., and Pomeau, Y., "Lattice-gas automata for the Navier-Stokes equation," Phys. Rev. Lett., 56(14), 1505 (1986).
2. Somers, J.A., "Direct simulation of fluid flow with cellular automata and the lattice Boltzmann equation," Appl. Sc. Res., 51, 127 (1993).
3. Eggels J.G.M., "Direct and large-eddy simulations of turbulent fluid flow using the lattice Boltzmann scheme," Int. J. Heat and Fluid Flow, 17, 307 (1996).
4. Derksen, J.J., and Van den Akker, H.E.A., "Parallel simulation of turbulent fluid flow in a mixing tank," Lecture Notes in Computer Science, 1401, 96 (1998).