Study of the Effects of Artificial Dissipation and Other Numerical Parameters on Shock Wave Resolution

This study investigates how five different finite-difference schemes and mesh geometry affect shock wave resolution in two-dimensional supersonic inviscid flows, identifying that while various methods introduce non-physical perturbations, specific AUSM+ formulations combined with techniques like flux limiting and artificial dissipation offer superior robustness and accuracy validated against experimental data.

The Gist

Imagine you are trying to take a high-speed photograph of a supersonic jet breaking the sound barrier. In the real world, the air in front of the jet compresses instantly, creating a sharp, invisible wall of pressure called a shock wave.

Now, imagine trying to recreate this scene on a computer. The computer doesn't "see" the air; it sees a grid of tiny squares (like a chessboard) and tries to calculate what happens in each square. The problem is that computers are clumsy at drawing sharp lines on a grid. Instead of a crisp shock wave, the computer often draws a wobbly, fuzzy line, or worse, it starts hallucinating weird ripples in the air where the air should be perfectly calm.

This paper is like a detective story where two researchers investigate five different "camera lenses" (mathematical methods) to see which one can take the clearest, most accurate picture of this shock wave without introducing these computer-induced glitches.

Here is the breakdown of their investigation using simple analogies:

1. The Setup: The Blunt Body

The researchers chose a simple test case: a round, blunt object (like the nose of a rocket) sitting in a supersonic wind tunnel.

• The Goal: To see how the air bends around the object and forms a "bow shock" (a curved shock wave) in front of it.
• The Challenge: The computer grid they used was curved to match the shape of the object. This curvature makes the math much harder, like trying to draw a straight line on a crumpled piece of paper.

2. The Suspects: Five Different Math Schemes

The team tested five different ways to do the math (numerical schemes). Think of these as five different chefs trying to bake the same cake:

• Chef Beam & Warming: Uses a very smooth, balanced approach (centered), but it's so smooth it gets wobbly. It needs extra "stabilizers" (artificial dissipation) to keep from falling apart.
• Chef Steger & Warming: A classic recipe that splits the wind into "left-moving" and "right-moving" parts. It's fast but tends to shake the cake (create oscillations) near the shock.
• Chef van Leer: A refined version of Steger & Warming. It's smoother and less shaky, but still has some wobbles.
• Chef Liou (AUSM+): A modern, two-part recipe. The researchers created two versions of this chef (Ap.1 and Ap.2) because they had to translate the recipe from a "volume" format to a "grid" format.

3. The Crime Scene: The "Ghost" Ripples

When they ran the simulations, they found a spooky problem. In the area before the shock wave (where the air should be perfectly calm and moving at a constant speed), the computer started showing ghost ripples.

• The Analogy: Imagine you are standing in a calm lake. Suddenly, the water starts rippling for no reason. That's what the computer was doing. It was inventing waves where there should be none.
• The Culprit: The researchers realized the "curved grid" itself was the problem. Because the grid lines were bent, the math got confused about how to measure distances, creating these fake ripples.

4. The Solution: "Freestream Subtraction" (The Magic Eraser)

To fix the ghost ripples, the researchers used a clever trick called Freestream Subtraction.

• The Analogy: Imagine you are trying to measure the height of a wave on a beach, but the tide is rising and falling, making your ruler inaccurate. Instead of measuring the absolute height, you measure the difference between the water level and the average tide level.
• The Result: By telling the computer, "Ignore the constant background wind; only calculate the changes," the ghost ripples vanished. It was like using a magic eraser to wipe the noise off the page. This worked for almost all the chefs, making the results much cleaner.

5. The Shock Wave: The "Upstream" Drift

Even after fixing the ripples, they noticed something else. Some of the math methods (specifically those that added extra "friction" or artificial dissipation to stop the shaking) pushed the shock wave too far forward.

• The Analogy: Imagine a crowd of people running. If you tell them to slow down too much (add too much friction), the front of the crowd stops earlier than it should. The shock wave was "drifting" upstream because the math was too heavy-handed.
• The Winner: The researchers found that using a Flux Limiter (a smart switch that turns the math from "smooth but shaky" to "sharp but stable" only when needed) was the best approach. It kept the shock wave in the right place without pushing it around.

6. The Verdict

After testing all the methods against real-world data and other scientists' results, here is what they concluded:

• The Best Chef: The AUSM+ (Ap.2) version was the star. It was so well-designed that it didn't even need the "magic eraser" (freestream subtraction) to avoid the ghost ripples. It was naturally robust.
• The Best Fix: For the other methods, using Freestream Subtraction was a must. It's a simple change that fixes huge problems.
• The Lesson: When simulating supersonic flight, you have to be very careful with how you draw your grid and how you handle the math. If you aren't careful, your computer will invent physics that don't exist!

In a nutshell: This paper teaches us that to get a perfect computer simulation of a supersonic shock wave, you need the right math recipe, a smart way to handle the grid, and a "magic eraser" to wipe out the computer's imagination. When done right, the computer can predict the behavior of high-speed flight almost perfectly.

Technical Summary

1. Problem Statement

The paper addresses the challenges associated with the accurate numerical simulation of two-dimensional supersonic inviscid flows around a blunt body (specifically a circular cylinder). The primary focus is on the resolution of the detached bow shock wave and the preservation of the freestream state upstream of the shock.

Key issues investigated include:

• Numerical Oscillations: Non-physical disturbances appearing in the upstream freestream region, particularly near the domain's symmetry line.
• Shock Profile Distortion: Deformations in the shape and location of the bow shock caused by numerical artifacts.
• Metric Term Inconsistencies: Errors arising from the discretization of domain transformation metrics in general curvilinear coordinates, which violate freestream invariance.
• Stability vs. Accuracy: The trade-off between stabilizing second-order schemes (via artificial dissipation or limiters) and the resulting shift in shock location or loss of convergence precision.

2. Methodology

The authors utilized an in-house CFD code solving the compressible 2-D Euler equations in general curvilinear coordinates. The study compared five different finite-difference schemes:

1. Beam and Warming (Implicit): A centered scheme using approximate factorization (ADI) with implicit Euler time marching. It requires explicit nonlinear artificial dissipation (Pulliam's model) for stability.
2. Steger and Warming (Explicit/Implicit): A flux vector splitting (FVS) upwind scheme.
3. van Leer (Explicit/Implicit): An improved FVS upwind scheme ensuring continuous differentiability of eigenvalues across Mach regimes.
4. AUSM+ (Ap.1): A reinterpretation of Liou's Advection Upstream Splitting Method Plus, adapted to finite differences where metric terms accompanying pressure are reconstructed using a continuous polynomial function (similar to the pressure splitting function).
5. AUSM+ (Ap.2): A second reinterpretation of AUSM+ where metric terms and the Jacobian at interfaces are reconstructed via simple arithmetic averaging of adjacent nodal values.

Test Case & Setup:

• Geometry: A 76.2 mm diameter circular blunt body.
• Flow Conditions: Supersonic freestream ($M_\infty = 8.0$), cold gas ($\gamma = 1.4$).
• Mesh: A structured $100 \times 100$ grid with clustering near the wall, intentionally designed with high curvature to expose numerical issues.
• Techniques Tested: The study evaluated the impact of freestream subtraction (modifying fluxes to subtract the freestream state), flux limiters (minmod limiter for 2nd-order schemes), and explicit artificial dissipation.

3. Key Contributions

• Identification of Metric-Induced Oscillations: The authors demonstrated that non-physical oscillations in the freestream are primarily caused by inaccuracies in the discrete treatment of metric terms (Jacobian and transformation derivatives) in curvilinear coordinates, rather than the schemes themselves.
• AUSM+ Reinterpretation: The paper presents two novel finite-difference interpretations of the AUSM+ scheme. It identifies AUSM+ (Ap.2) (using simple averaging for metrics) as superior because it naturally suppresses freestream oscillations without requiring freestream subtraction.
• Freestream Subtraction Efficacy: The study validates freestream subtraction as a highly effective, low-cost fix to reduce freestream oscillation amplitudes by at least 6 orders of magnitude for centered and standard upwind schemes.
• Artificial Dissipation Impact: The paper highlights a critical side effect of explicit artificial dissipation models (like Pulliam's): they systematically shift the shock wave upstream compared to literature benchmarks, whereas limiter-based stabilization preserves the correct shock location.

4. Results and Discussion

Freestream Oscillations

• Observation: Without freestream subtraction, the Steger-Warming, van Leer, and AUSM+ (Ap.1) schemes all produced significant oscillations in the upstream Mach number (deviations up to 4.4% for Steger-Warming).
• Cause: These oscillations stem from the discrete violation of domain transformation invariants due to high mesh curvature.
• Solution:
  • Freestream Subtraction: Reduced oscillations significantly for all schemes except AUSM+ (Ap.2).
  • AUSM+ (Ap.2): This formulation eliminated oscillations entirely (down to machine precision) even without freestream subtraction, making it the most robust formulation for this specific context.

Shock Wave Resolution and Stability

• First-Order Schemes: Generally stable but diffusive. With freestream subtraction, they produced smooth shock profiles.
• Second-Order Schemes: Pure second-order upwind schemes diverged due to lack of dissipation.
  • Stabilization Methods: Stability was achieved either by adding explicit artificial dissipation or applying a flux limiter.
  • Shock Location Discrepancy:
    • Schemes using explicit artificial dissipation shifted the shock wave significantly upstream.
    • Schemes using flux limiters (minmod) maintained shock locations consistent with literature data (Peery & Imlay, Lin) but suffered from slower convergence (residue reduction limited to 3–4 orders of magnitude vs. 11 orders for other methods).
• Pressure Coefficient ($C_p$): All stabilized schemes showed good agreement with experimental data ($C_p$ distribution) on the cylinder surface, except for the centered Beam-Warming scheme, which showed deviations near the stagnation point due to its upstream-shifted shock.

5. Significance and Conclusion

The paper provides critical insights for CFD practitioners simulating high-speed flows with strong shocks on curved grids:

1. Metric Terms are Critical: The geometry of the mesh (curvature) directly influences numerical stability and accuracy in curvilinear coordinates.
2. Recommended Practices:
   • Freestream Subtraction is highly recommended for centered and standard upwind schemes to eliminate non-physical freestream disturbances.
   • Flux Limiters are preferred over explicit artificial dissipation for second-order schemes to avoid artificial shock shifting, despite the penalty in convergence speed.
   • AUSM+ (Ap.2) is identified as the most robust formulation for finite-difference implementations on curvilinear grids, as it inherently handles metric reconstruction without inducing oscillations.
3. Validation: The study confirms that with proper numerical handling (specifically avoiding artificial dissipation-induced shifts), inviscid Euler simulations can accurately predict shock locations and surface pressure distributions comparable to experimental and high-fidelity Navier-Stokes results.

In summary, the authors successfully isolated the sources of numerical error in blunt body shock simulations and proposed specific, practical modifications (subtraction, limiter usage, and specific metric reconstruction) to achieve high-fidelity results.