Face-based Volume-of-Fluid interface positioning in arbitrary polyhedra

TL;DR

This paper presents a fast, robust algorithm for accurately positioning a plane in arbitrary polyhedra to achieve a specified volume fraction, improving efficiency over existing methods especially for unstructured meshes.

Contribution

The work introduces a novel recursive volume computation method using the Gaussian divergence theorem and an exact polynomial expression near the truncation point, enabling efficient VoF interface positioning.

Findings

Requires only 1-2 truncations on average, outperforming existing methods.

Applicable to convex and non-convex polyhedra with various topologies.

Efficient for unstructured computational meshes.

Abstract

We introduce a fast and robust algorithm for finding a plane $\Gamma$ with given normal $\vec{n}_\Gamma$, which truncates an arbitrary polyhedron $\mathcal{P}$ such that the remaining sub-polyhedron admits a given volume $\alpha|\mathcal{P}|$. In the literature, this is commonly referred to as Volume-of-Fluid (VoF) interface positioning problem. The novelty of our work is twofold: firstly, by recursive application of the Gaussian divergence theorem, the volume of a truncated polyhedron can be computed at high efficiency, based on summation over quantities associated to the faces of the polyhedron. One obtains a very convenient piecewise parametrization (within so-called brackets) in terms of the signed distance s to the plane $\Gamma$. As an implication, one can restrain from the costly necessity to establish topological connectivity, rendering the present approach most suitable for the application to unstructured computational meshes. Secondly, in the vicinity of the truncation position s, the volume can be expressed exactly, i.e. in terms of a cubic polynomial of the normal distance to the PLIC plane. The local knowledge of derivatives enables to construct a root-finding algorithm that pairs bracketing and higher-order approximation. The performance is assessed by conducting an extensive set of numerical experiments, considering convex and non-convex polyhedra of genus (i.e., number of holes) zero and one in combination with carefully selected volume fractions $\alpha$ (including $\alpha\approx0$ and $\alpha\approx1$) and normal orientations $\vec{n}_\Gamma$. For all configurations we obtain a significant reduction of the number of (computationally costly) truncations required for the positioning: on average, our algorithm requires between one and two polyhedron truncations to find the position of the plane $\Gamma$, outperforming existing methods.