TL;DR
This paper introduces a novel modification to the Ghost Fluid Method (GFM) to address its inherent conservation errors in compressible multiphase flows. By carefully tracking conservative variables near fluid interfaces, the proposed method ensures mass and momentum conservation without sacrificing the GFM's ability to produce sharp interfaces and avoid spurious oscillations.
What Problem Does This Paper Address?
Simulating compressible multiphase flows, where different fluids interact, presents significant challenges. Standard numerical methods often struggle with accurately capturing the sharp interfaces between these fluids, leading to nonphysical oscillations. The Ghost Fluid Method (GFM) was developed to tackle this by treating each fluid separately, effectively "ghosting" information across the interface. While GFM excels at producing sharp interfaces and eliminating oscillations, it suffers from a fundamental drawback: it's non-conservative. This means that quantities like mass and momentum can be lost or gained artificially during the simulation, especially as interfaces move. This paper aims to fix this conservation issue within the GFM framework.
Methods
The core of the proposed solution lies in a "conservative modification" to the standard GFM. Instead of simply replacing cell values with ghost fluid data when an interface moves out of a cell, the authors introduce a new approach:
- Tracking Conservative Variables: They track the "true" conservative variables (density, momentum, energy) in cells near the interface.
- "Fluid e" Concept: A unique numerical flux, termed "Fluid e," is computed using information from both fluids. This flux is intended to represent a more accurate, single-valued flux across the interface.
- Post-Interface-Movement Correction: When the interface has moved out of a cell (i.e., the cell now contains only one fluid, but its values were previously influenced by the interface), the algorithm uses the "Fluid e" values to update the cell's conservative variables. This correction is applied after the interface has left the cell, thus not interfering with the GFM's core interface-handling mechanism.
- Base Scheme: The method is designed to be compatible with any underlying GFM base scheme. In this paper, they employ a fifth-order finite difference WENO scheme for spatial discretization and a third-order TVD Runge-Kutta method for time integration. A level set method is used for interface tracking.
Datasets
Not discussed in the paper. The paper focuses on numerical methods and validates them using standard test cases.
Key Insights and Ideas
- Decoupling Interface Handling from Conservation: The most significant insight is the strategy of applying conservation corrections after the interface has moved out of a cell. This clever approach preserves the GFM's strengths (sharp interfaces, no oscillations) while addressing its weakness (non-conservation).
- "Fluid e" as a Bridge: The concept of "Fluid e" provides a mechanism to derive a more accurate, unified flux that accounts for the interaction of both fluids, even when the interface is no longer present in a given cell.
- Two-Part Conservation Error: The authors identify that the GFM's conservation error has two components: one from the non-unique flux at the interface itself, and another from the ghost fluid replacement. Their modification specifically targets the latter.
Key Results
The paper demonstrates that the proposed conservative modification:
- Achieves Mass and Momentum Conservation: Numerical experiments, particularly the Sod shock tube problem, show a significant reduction in conservation errors, with the modified method exhibiting near-zero mass conservation error over time.
- Maintains Resolution: The modified GFM generally maintains comparable numerical resolution to the original GFM.
- Improves Accuracy in Some Cases: In certain examples (e.g., Example 2 and 3), the fully conservative modification shows better agreement with exact solutions, particularly for shock propagation.
- Offers Flexibility: The modification can be applied partially (e.g., only to density and momentum, not energy) to balance conservation with stability, as shown in Example 4.
Limitations and Future Work
- Energy Conservation: While the method is shown to be conservative in mass and momentum, the paper notes that achieving full energy conservation can sometimes lead to stability issues. The "partially conservative modification" is presented as a way to mitigate this.
- Computational Cost: The additional steps for tracking conservative variables and computing "Fluid e" fluxes will inherently add some computational overhead compared to the original GFM.
- Future Work: The authors suggest that the modification procedure can be extended to higher dimensions straightforwardly, which is demonstrated in Example 6. Further exploration of the trade-offs between full conservation and stability for energy would be a natural next step.
Summary updated