Abstract
High-fidelity fluid simulations require a delicate balance between capturing sharp discontinuities and resolving fine-scale turbulence. Traditional methods often suffer from excessive numerical dissipation or instability near shocks. This paper presents the TENO-M framework, a novel approach that integrates nonlinear limiters into the Targeted ENO (TENO) paradigm. By filtering nonsmooth candidate stencils rather than discarding them, TENO-M provides flexible dissipation control, enabling robust shock-capturing and high-resolution turbulence modeling.
Introduction
In Computational Fluid Dynamics (CFD), researchers face a fundamental trade-off. Simulating shock waves, such as those from supersonic aircraft, requires numerical dissipation to prevent non-physical oscillations. Conversely, simulating fine-scale turbulence requires low dissipation to preserve small-scale eddies. Existing methods like Weighted Essentially Non-Oscillatory (WENO) schemes often introduce too much dissipation, blurring fine details. While Targeted ENO (TENO) schemes improved upon this by using optimal linear weights in smooth regions, they traditionally discard "nonsmooth" candidate stencils entirely. This binary selection can lead to numerical artifacts or insufficient stability near extreme discontinuities. The TENO-M framework addresses this by introducing flexible nonlinear dissipation control.
Methods
The TENO-M framework innovates on the standard TENO approach through a three-step process:
Stencil Selection
The algorithm first measures the smoothness of several candidate stencils. As shown in Figure 1, these stencils ($S_0$ to $S_5$) represent local groups of data points used to calculate the numerical flux at cell interfaces.
The Filtering Step
Unlike traditional TENO, which discards stencils flagged as nonsmooth, TENO-M passes these stencils through a nonlinear limiter. This allows the framework to retain useful information while suppressing oscillations. Three types of limiters were evaluated:
- Monotonicity-Preserving (MP): Optimized for maintaining local extrema in turbulent flows.
- Total Variation Diminishing (TVD): Provides robust oscillation suppression for strong shocks.
- Van Albada (VA): A standard second-order limiter for general stability.
Linear Assembly
Because the candidate stencils are pre-filtered to be oscillation-free, the final numerical flux can be assembled using optimal linear weights. This eliminates the complex weight renormalization required in original TENO schemes. The coefficients for these linear combinations are detailed in the table below.
Results and Discussion
The TENO-M framework was validated against a suite of rigorous mathematical benchmarks:
- Accuracy Preservation: The schemes successfully maintained theoretical 6th- and 8th-order accuracy in smooth regions.
- Turbulence Resolution: In the Taylor-Green Vortex benchmark, the TENO8A-M-MP variant faithfully reproduced the Kolmogorov $-5/3$ energy spectrum, demonstrating its effectiveness for Implicit Large Eddy Simulation (ILES).
- Robustness in Extremes: The framework successfully handled extreme cases like the Noh problem (infinitely strong shock) and the Le Blanc problem (pressure ratio of $10^9$), where standard TENO8A often fails or produces non-physical density overshoots.
- Computational Efficiency: Despite the additional filtering step, TENO-M is computationally efficient. In 2D Double Mach reflection tests, TENO8A-M-TVD5 was approximately 1.5% faster than standard TENO8A due to the simplified assembly logic.
Conclusion
TENO-M provides a versatile framework for CFD, allowing users to tune numerical dissipation as a "knob" by selecting appropriate limiters. This hybridization occurs seamlessly within the stencil selection process, avoiding the need for separate "troubled cell" indicators. Future work will explore applications in Magnetohydrodynamics (MHD) and multi-phase flows, as well as adaptation for unstructured meshes in industrial settings.
Summary updated