AIR-HPCResearch Group · UPCT
AIR-HPC / AIR Toolkits / AIRgVOF

Integration middleware and host-solver adapters

AIRgVOF

Non-intrusive middleware and host-solver adapters for coupling the Fortran-based gVOF library with external CFD solvers.

Active developmentOpenFOAM v2606Source release planned

Overview

AIRgVOF is a middleware architecture for connecting gVOF’s Fortran numerical kernels with external multiphase-flow solvers. The current implementation couples a development version of gVOF with OpenFOAM v2606 while keeping the geometric volume-of-fluid algorithms independent of the host solver.

The project explores a non-intrusive integration strategy: OpenFOAM provides the flow solution and MPI domain decomposition, while gVOF provides interface initialization, reconstruction and geometric advection with native OpenMP parallelism. The current results demonstrate the viability of this hybrid architecture.

Current capabilities

  • Exchange of mesh, phase-fraction and flux data between OpenFOAM and gVOF.
  • Volume-fraction initialization through a hiVOF-based routine in the development version of gVOF.
  • Geometric PLIC reconstruction and unsplit interface advection on arbitrary polyhedral meshes.
  • Hybrid execution combining OpenFOAM’s MPI parallelism with gVOF’s OpenMP parallelism.
  • Consistent, conservative liquid-flux treatment across decomposed-domain boundaries.
Scope of this public description. Implementation details are intentionally limited while the methodology is under active development.

Software characteristics

Host solver
OpenFOAM v2606
Languages
C++ host adapter and Fortran kernels
Parallel model
MPI × OpenMP
Platform tested
Linux
Compiler family
GNU toolchain
Build
gfortran, wclean and wmake

Development dependencies: unpublished development versions of gVOF, VOFTools, isoap and hiVOF. macOS support is being investigated but is not currently claimed.

Preliminary result

Three-dimensional rising bubble on a polyhedral mesh

This simulation provides a compact demonstration of the complete integration path on a demanding three-dimensional two-phase problem. A bubble with initial diameter D0 = 10 mm rises in a 0.04 × 0.04 × 0.08 m³ domain discretized by a BCC-derived polyhedral mesh.

Reconstructed dimpled bubble interface at 0.30 seconds, displayed against a central crinkle-clipped section of the BCC-derived polyhedral mesh.
Reconstructed interface at t = 0.30 s, shown against a central crinkle-clipped section of the BCC-derived polyhedral mesh.
Domain0.04 × 0.04 × 0.08 m³
Initial bubble diameter10 mm
Mesh2,011,489 BCC-derived polyhedral cells
Nominal BCC resolution20 spacings per initial diameter; hBCC = 0.5 mm
Volume-equivalent resolutionheq = 0.3992 mm; D0/heq ≈ 25
Parallel configuration4 MPI ranks × 8 OpenMP threads
Time integration8,151 steps to 0.30 s; maxCo = maxAlphaCo = 0.02
BoundednessVolume fraction remained in [0, 1]
Volume conservation+0.005318% drift from 0.01 to 0.30 s
Late-time Reynolds numberMean Re = 1.519 from 0.20 to 0.30 s; experiment: 1.5
Evolution of the reconstructed interface in the BCC20 rising-bubble case.
Side, bottom and central-section views of the reconstructed rising bubble and polyhedral mesh.
Complementary side, bottom and central-section views of the reconstructed interface and BCC-derived polyhedral mesh.

Quantitative context

The late-time mean Reynolds number is within 1.26% of the available experimental value of 1.5. Published numerical values for related formulations include 1.70 for PLIC, 1.43 for SCIR and 1.57 for iPPIC. These values provide literature context only: they are not presented as a controlled head-to-head comparison because meshes, implementations and numerical settings differ.

The reference discussion is available in J. López (2026), Unsplit geometric volume-of-fluid method with iterative piecewise-paraboloid interface reconstruction on arbitrary three-dimensional grids, Journal of Computational Physics, 553, 114714. https://doi.org/10.1016/j.jcp.2026.114714.

Interpretation. This single case is intended to demonstrate integration feasibility, bounded and conservative interface transport, execution on an arbitrary polyhedral mesh and hybrid MPI–OpenMP operation. It is not a formal validation, mesh-convergence or parallel-scalability study.

Development status

AIRgVOF is in active development. Its source code has not yet been released; a public source release is planned after the integration and accompanying scientific documentation are complete.

For research or software enquiries, contact Joaquín López at the Universidad Politécnica de Cartagena.