The following papers have been accepted but have not been finalized (proof reading) yet.
«« Back

Comparison of Interface Description Methods Available in Commercial CFD Software
Author(s): Andres Barrall, Roberta Minussi, Marcus Alves
Keywords: Volume of Fluid, Numerical Diffusion, Interfacial Smearing
This work addresses the characteristics of the interpolation functions and interface reconstruction routines for the VOF – Volume of Fluid method available in the commercial CFD software ANSYS-FLUENT. This software was used because it has both implicit and explicit VOF approaches along with diverse interpolation functions. Some of these functions were compared from different points of view: the quality of the reconstructed interface; the ability to preserve the initial mass inside the system (numerical diffusion); and the computing time. As means of undertaking the qualitative and quantitative comparisons, a test problem which combines the classical dam break and slosh tank benchmark problems was used. There is no analytical solution available for this problem, in which the most interesting feature is a high interaction between the velocity field and volume fraction, thus making it ideal for addressing the issue of interfacial smearing. ANSYS-FLUENT permits the use of 5 interpolation functions for transient simulations: PLIC, CICSAM, HRIC (explicit and implicit) and the UPWIND scheme and four when performing steady state ones: BGM, modified HRIC, COMPRESSIVE and UPWIND schemes. Both transient and steady state solutions were analyzed in this work, using all the above schemes, with the exception the UPWIND one for steady state simulations. It was found that, for thinner grids, PLIC, CISAM and the explicit HRIC schemes had similar performances concerning the quality of the reconstructed interface and mass conservation. By the other hand, for coarser grids, PLIC shows the best results, being the only one to conserve mass for all tests. The computation time were similar for all transient simulation (within each grid). Concerning the steady state simulations, which are, in fact, distorted transient simulations, the BGM and the compressive schemes produced similar results but BGM consumed more computational time.