Quantifying Flow separation for ellipse and von-Kármán Airfoil: A dataset of surface pressure and skin friction

This paper presents a dataset derived from steady-state RANS simulations using the $k\omega$ SST turbulence model in OpenFOAM, which quantifies flow separation characteristics—including surface pressure, skin friction, and separation points—for 2D flow around an ellipse and a von-Kármán-Trefftz airfoil across various angles of attack and Reynolds numbers to serve as a benchmark for extended potential flow models.

The Gist

Imagine you are trying to design a super-efficient boat or airplane wing. To do this, engineers usually use two types of "maps" to predict how water or air will flow around the object.

1. The Simple Map (Potential Flow): This is like a sketch drawn with a pencil. It's fast, easy to calculate, and great for smooth, calm flows. But it has a blind spot: it can't see when the flow gets messy and starts to peel away from the surface (this is called flow separation).
2. The Detailed Map (CFD/RANS): This is like a high-definition 3D simulation. It's incredibly accurate and can see every swirl and tear in the flow, but it takes a supercomputer and days of time to run.

The Problem: Engineers want the speed of the "Simple Map" but the accuracy of the "Detailed Map." They want to teach the simple map how to see the messy parts. But to teach it, they need a perfect "answer key" to check their work against.

The Solution: This paper provides that answer key.

The Experiment: The "Torture Test" for Shapes

The researchers decided to test two specific shapes that are notorious for causing flow separation:

• The Ellipse: Think of a flattened circle, like a smooth egg or a submarine hull.
• The von Kármán Airfoil: A classic, symmetrical airplane wing shape.

They put these shapes in a virtual wind tunnel (using software called OpenFOAM) and blew air over them at different speeds and angles. They didn't just look at the big picture (how much lift or drag the shape created); they looked at the micro-details:

• Pressure: How hard is the air pushing on the surface?
• Skin Friction: How much is the air "rubbing" against the surface?
• The Separation Point: Exactly where does the air stop sticking to the surface and start flying off?

The Analogy: The Sticky Tape

Imagine sticking a piece of tape to a wall.

• Attached Flow: The tape is stuck tight. The air is "sticking" to the wing.
• Flow Separation: If you blow too hard or at a weird angle, the tape starts to peel off. The point where the tape first lifts up is the separation point.

In this study, the researchers mapped out exactly where that "tape" lifts off for the ellipse and the airfoil at different speeds. They found that for the ellipse, the tape peels off almost immediately, no matter how you tilt it. For the airfoil, it holds on tighter until you tilt it too far.

Why This Matters

Usually, when scientists publish data, they only give you the final score (e.g., "The plane flew 10% better"). They rarely give you the "replay" of the game (the pressure and friction data).

This paper is special because it's an open-source dataset. It's like giving everyone the raw video footage and the play-by-play commentary, not just the final score.

Who is this for?

• Students and Researchers: They can use this data to train new, faster computer models. Instead of running a slow, expensive simulation every time, they can use this data to teach a "smart" simple model to predict separation accurately.
• Engineers: They can use these numbers to validate their own designs. If their simulation matches this "gold standard" data, they know their design is likely correct.

The "Secret Sauce" of the Study

The researchers were very careful to ensure their data was trustworthy:

1. Mesh Convergence: They checked that their virtual grid was fine enough to catch tiny details. If they made the grid coarser, the results changed slightly, so they picked the "just right" grid size.
2. Turbulence Check: They tested different levels of "wind chaos" (turbulence) to make sure their results didn't change just because the wind was a little bit bumpier. It turned out their results were stable and reliable.

The Bottom Line

This paper is a gift to the engineering community. It provides a high-quality, detailed record of how air behaves around simple but tricky shapes. By giving us the "ground truth" of where flow separates, it helps scientists build better, faster, and smarter tools to design the next generation of ships, planes, and wind turbines.

In short: They built a perfect reference library so that everyone else can stop guessing and start designing with confidence.

Technical Summary

1. Problem Statement

Potential flow models are widely used in engineering for their computational efficiency but struggle to accurately capture flow separation, particularly on blunt bodies like ellipses at moderate Reynolds numbers. While viscous-inviscid coupling methods (e.g., XFOIL) exist, they often rely on embedded assumptions that limit their accuracy.

There is a significant lack of high-fidelity, openly accessible benchmark datasets that provide detailed surface data (specifically skin friction and separation points) for simple geometries undergoing flow separation. Most available data focuses on integral coefficients (lift/drag) rather than local surface distributions required for calibrating and validating extended potential flow models.

2. Methodology

The authors conducted a systematic Computational Fluid Dynamics (CFD) study to generate a high-quality reference dataset.

• Geometries:
  • Ellipse: Major axis as chord length ($c$), semi-major axis 0.5, semi-minor axis 0.125.
  • von K´arm´an Airfoil: Symmetrical, thickness ratio $t/c = 0.1457$, trailing edge angle $0^\circ$, constructed via the K´arm´an-Trefftz transformation.
• Numerical Setup:
  • Solver: OpenFOAM v2506 using the simpleFoam (steady-state) solver.
  • Turbulence Model: RANS with the Menter $k-\omega$ SST model. This model was chosen for its general applicability and native Low-Re wall treatment.
  • Boundary Conditions:
    • Inlet velocity: Constant at 10 m/s.
    • Reynolds numbers: $Re = 10^6$ and $Re = 10^7$.
    • Turbulence parameters ($k, \omega, \nu_t$) were set based on Spalart et al. and Menter recommendations to ensure a fully turbulent boundary layer assumption, avoiding the complexity of transition modeling for reproducibility.
  • Mesh: 2D meshes with Low-Re wall treatment ($y^+ \approx 1$). Separate meshes were generated for each Reynolds number to satisfy $y^+$ requirements, while the same mesh was used for all angles of attack.
• Validation & Sensitivity:
  • Mesh Convergence: Performed on the ellipse at $Re=10^6, \alpha=10^\circ$. A "Medium" mesh (48,000 cells) was selected as sufficient, showing <1% error in drag coefficient compared to finer meshes.
  • Turbulence Sensitivity: Tested varying ambient turbulence levels. Results showed that force coefficients and separation points were insensitive to reasonable variations in ambient turbulence, validating the baseline settings.
• Parameters: Simulations covered angles of attack ($\alpha$) from $0^\circ$ to $20^\circ$.

3. Key Contributions

The primary contribution of this work is the creation of a comprehensive, open-access dataset designed specifically for the development and validation of reduced-order flow models.

• Data Availability: The dataset includes local distributions of:
  • Pressure Coefficient ($C_p$)
  • Skin Friction Coefficient ($C_f$)
  • Streamlines
• Integral & Point Data:
  • Lift ($C_l$) and Drag ($C_d$) coefficients.
  • Precise locations of Stagnation points and Separation points (both suction and pressure sides) expressed in terms of chord position ($x/c$) and angular position ($\phi$).
• Scope: Covers two distinct geometries (blunt ellipse and sharp-edged airfoil) across two Reynolds numbers and a wide range of angles of attack, including post-stall regimes.
• Supplementary Material: Raw data is provided in a "Results.zip" file, facilitating immediate use by the research community.

4. Key Results

• Flow Separation Behavior:
  • Ellipse: Exhibits flow separation at all tested angles of attack. Separation points move upstream (toward the leading edge) as the angle of attack increases.
  • Airfoil: Remains attached at low angles ($0^\circ$–$10^\circ$). At higher angles ($15^\circ$–$20^\circ$), significant separation occurs on the suction side, with the separation point moving drastically forward (e.g., to $x/c \approx 0.07$ at $20^\circ$ for $Re=10^6$).
• Reynolds Number Effects:
  • Higher Reynolds numbers ($10^7$) generally result in slightly delayed separation (separation points move slightly downstream) compared to $10^6$.
  • Skin friction coefficients are generally higher at lower Reynolds numbers.
  • Pressure distributions ($C_p$) are largely insensitive to Reynolds number unless separation points differ significantly.
• Force Coefficients:
  • The ellipse shows a linear increase in lift up to $15^\circ$, followed by a stall.
  • The airfoil demonstrates higher lift-to-drag performance but experiences a sharp drop in lift and spike in drag at high angles of attack due to massive separation.

5. Significance and Limitations

Significance:

• Model Development: This dataset fills a critical gap for researchers developing extended potential flow models that aim to predict separation without full RANS computations. The local $C_f$ and separation point data are essential for calibrating empirical corrections (e.g., Kutta conditions, separation criteria).
• Benchmarking: It serves as a rigorous benchmark for validating CFD solvers and turbulence models against high-fidelity steady-state RANS data.

Limitations:

• Steady-State Assumption: The study uses steady-state RANS, which suppresses transient phenomena like von K´arm´an vortex shedding. Consequently, the separation points represent a time-averaged position rather than the oscillating nature of real unsteady flow.
• No Transition Modeling: The $k-\omega$ SST model assumes a fully turbulent boundary layer. In reality, laminar-turbulent transition might occur, potentially shifting separation points slightly. However, the authors argue that at these high Reynolds numbers, separation occurs in the turbulent region, making this assumption acceptable for the intended purpose.

Conclusion

The paper successfully provides a high-fidelity, reproducible dataset of surface pressure and skin friction for an ellipse and a von K´arm´an airfoil. By rigorously verifying mesh convergence and turbulence sensitivity, the authors ensure the data is reliable for calibrating and validating reduced-order models intended to handle flow separation, thereby advancing the capability of computationally efficient aerodynamic simulations.