ID 264

A new VOF interface reconstruction technique based on transported surface data

 

Abstract:

Simulating complex atomizing flows using volume of fluid (VOF) requires reconstructing the phase interface within each computational cell. Classical methods represent the liquid-gas interface as a piecewise planar surface (PLIC). To that end, a normal vector needs to be provided, which is typically found from some form of the VOF gradient. This increases the stencil size and results in premature break-up of structures due to poor local resolution. Moment of Fluid (MOF) remedies this issue by finding a plane normal vector to best match the phase barycenter. Unfortunately, realizability of the MOF-PLIC approach is not guaranteed. To address these shortcomings, a novel method of moments based interface reconstruction strategy is proposed that relies on transported interface surface data to perform the PLIC reconstruction. This method allows two planes to exist in a single cell, giving the ability to maintain arbitrarily thin sheets while phase volumes are discretely conserved. In order to determine the number of planes necessary to accurately reconstruct the interface, a k-means clustering algorithm is applied to the distribution of advected interface surface area vectors. The method can be used for interface representations with an arbitrary number of planes, however, the current work is limited to two planes. This work discusses the creation of the advected interface surface area vector distribution, the k-means clustering performed to find representative interface normal vectors, and the volume-conservative placement of two planes in a computational cell.