Abstract
Simulating Non-Ideal Compressible-Fluid Dynamics (NICFD) is computationally demanding, particularly when dealing with moving boundaries and complex thermodynamics. Traditional mesh adaptation methods rely on solution interpolation, which introduces numerical oscillations that can destabilize simulations near the vapor-liquid saturation curve. This paper presents a novel interpolation-free Arbitrary Lagrangian-Eulerian (ALE) scheme. By conceptualizing topological mesh changes as a sequence of fictitious continuous deformations, the proposed method maintains conservation and stability without the need for interpolation. The scheme is validated using the Peng-Robinson Equation of State in unsteady piston simulations, demonstrating high accuracy and robustness in the NICFD regime.
Introduction: The Challenge of Non-Ideal Fluids
In the regime of Non-Ideal Compressible-Fluid Dynamics (NICFD), fluids such as dense vapors or supercritical fluids exhibit behaviors that deviate significantly from the ideal gas law. Phenomena like non-monotonic sound speed variations and rarefaction shocks are common. Accurately simulating these flows, especially in systems with moving boundaries like oscillating pistons, requires dynamic mesh adaptation to resolve sharp features like shockwaves efficiently.
However, a critical bottleneck in standard mesh adaptation is the requirement to interpolate fluid states from an old grid to a new one. In NICFD, even minute numerical oscillations from interpolation can push the calculated thermodynamic state into the two-phase region (under the Vapor-Liquid Equilibrium curve), causing mathematical models to fail and simulations to crash. The authors address this by developing a method that avoids interpolation entirely.
Methodology and Algorithmic Workflow
To overcome the interpolation bottleneck, the authors implement an interpolation-free finite-volume scheme within an ALE framework. The solver handles the evolution of the fluid state and the mesh geometry simultaneously, ensuring that mass, momentum, and energy are conserved throughout the process.
As shown in the flowchart above, the process begins with mesh deformation based on boundary movement. This is followed by a prediction step and a dedicated mesh adaptation loop. The adaptation loop utilizes error estimators (based on Mach number and pressure gradients) and the Mmg library to perform local grid modifications. Finally, the solution is computed on the updated mesh, closing the time loop.
Interpolation-Free Mesh Adaptation
The core innovation lies in how the solver handles grid connectivity changes, such as edge splitting or node insertion. Instead of discrete jumps that require interpolation, these changes are treated as "fictitious continuous deformations" over a sub-time step.
The process follows three distinct phases:
- Collapse ((\tau = 0.0) to (0.5)): The elements targeted for modification are continuously collapsed toward an arbitrary point until their volume reaches zero.
- Connectivity Change ((\tau = 0.5)): At the moment of zero volume, the grid topology is updated (e.g., adding a node (j)). Since the volume is zero, no mass, momentum, or energy fluxes are exchanged, eliminating the need for interpolation.
- Expansion ((\tau = 0.5) to (1.0)): The newly formed elements are expanded from the collapse point to their final physical configuration.
This geometric approach ensures that the solver perfectly preserves conservation laws throughout the adaptation process without introducing spurious oscillations.
Results and Discussion
The solver was tested using the polytropic Peng-Robinson Equation of State (EoS) for the siloxane MD4M. Validation against analytical solutions for 2D and 3D shocks confirmed the baseline accuracy of the ALE framework.
In complex piston simulations, the interpolation-free method proved its worth. Even during violent oscillations where the fluid state approached the Vapor-Liquid Equilibrium (VLE) curve, the adaptation introduced zero spurious oscillations, maintaining total simulation stability. A key finding was the inadequacy of standard models: using a Polytropic Ideal Gas (PIG) model resulted in a 51% deviation in maximum pressure predictions compared to the Peng-Robinson model, highlighting that specialized NICFD solvers are essential for accurate engineering in these environments.
Conclusion
The proposed adaptive ALE scheme provides a robust solution for unsteady NICFD simulations. By replacing interpolation with geometric deformation, the method avoids the numerical instabilities that plague complex thermodynamic modeling. Future work will focus on extending this framework to more intricate geometries, such as those found in Organic Rankine Cycles (ORCs) and supercritical (CO_2) power systems, where non-ideal effects are a primary design consideration.
Summary updated