On Capturing Laminar/Turbulent Regions Over a Wing Using WMLES

This study demonstrates that accurately capturing both laminar and turbulent regions over a wing using Wall-Modeled Large-Eddy Simulation (WMLES) requires a hybrid grid resolution strategy combined with the introduction of unsteady disturbances to trigger transition, as standard uniform grids fail to satisfy the conflicting near-wall resolution requirements of both flow regimes.

The Gist

The Big Picture: The "Goldilocks" Problem of Airflow

Imagine you are trying to predict how air flows over a wing of an airplane. This is crucial for designing efficient, fuel-saving aircraft.

Scientists have two main ways to do this:

1. The "Super-Detailed" Way (DNS/WRLES): They try to simulate every single tiny swirl of air, from the surface of the wing all the way out. This is like trying to count every single grain of sand on a beach to understand the tide. It's incredibly accurate, but it requires so much computer power that it's impossible for big, complex planes.
2. The "Rough Estimate" Way (RANS): They use a simplified math model that averages out the chaos. It's fast, but sometimes it misses the fine details, like how a storm actually forms.

The Solution: The paper explores a middle ground called WMLES (Wall-Modeled Large-Eddy Simulation). Think of this as a "hybrid" approach.

• The Outer Layer: We simulate the big, chaotic swirls of air (the storm clouds) in detail.
• The Inner Layer: Right next to the wing surface, where the air is smooth and sticky, we use a "rule of thumb" (a wall model) instead of simulating every tiny molecule.

The goal of this paper is to figure out how to build the perfect "grid" (a digital mesh) to make this hybrid method work for a wing that has both smooth (laminar) air and chaotic (turbulent) air.

---

The Challenge: The "One-Size-Fits-All" Trap

The researchers ran into a classic "Goldilocks" problem: How do you make a grid that is just right for two very different types of air?

1. The Smooth Zone (Laminar): Near the front of the wing, the air is like a calm, smooth river. To see this, you need a very fine, high-resolution grid (like using a magnifying glass).
2. The Chaotic Zone (Turbulent): Further back, the air becomes a raging white-water rapid. To simulate this efficiently, you need a coarser grid where the "rule of thumb" (the wall model) can take over.

The Experiment:
The team tried using the same grid for the whole wing, like wearing the same pair of shoes for both a marathon and a ballet.

• If the shoes were too big (Coarse Grid): They could handle the "white-water rapid" (turbulence) just fine, but they completely missed the details of the "calm river" (laminar flow). The computer thought the smooth air was turbulent too early.
• If the shoes were too small (Fine Grid): They could see the "calm river" perfectly, but they broke the "white-water rapid" simulation. The computer got confused because the grid was too detailed for the wall model to work, causing the turbulence to start way too late.

The Analogy: It's like trying to take a photo of a butterfly and a mountain in the same frame. If you zoom in to see the butterfly's wings, the mountain becomes a blurry blob. If you zoom out to see the mountain, the butterfly disappears.

---

The Breakthrough: The "Custom Tailor" Approach

The researchers realized they couldn't use a "one-size-fits-all" grid. They needed a custom-tailored grid that changes its density depending on where it is on the wing.

1. The Map: They first used a simpler, faster simulation (RANS) to create a "map" of how thick the air layer gets as it moves down the wing.
2. The Tailoring: They built a new grid that had:
   • Tight spacing (many points) where the air layer was thin (the front/laminar zone).
   • Loose spacing (fewer points) where the air layer was thick (the back/turbulent zone).

The Missing Piece: The "Trip"
Even with the perfect custom grid, the simulation still struggled to switch from smooth to chaotic air naturally. It was like a car stuck in neutral; it needed a push.

The researchers added artificial "trips" (tiny disturbances) near the front of the wing.

• Analogy: Imagine a smooth road that suddenly turns into a bumpy dirt track. If you drive slowly, you might not notice the bump. But if you hit a small speed bump (a trip) at the exact right spot, you force the car to shake and switch gears.
• By introducing these specific "shakes" (disturbances) at the right frequency, they forced the air to transition from smooth to turbulent exactly where it was supposed to.

The Results

By combining the custom-tailored grid with the artificial trip, they finally got it right:

• They could see the smooth air perfectly at the front.
• They could see the chaotic turbulence perfectly at the back.
• The transition happened at the exact right spot.

The Takeaway

This paper teaches us that to simulate complex airflow accurately without spending a fortune on computer time, you can't just use a standard grid. You have to:

1. Adapt your tools: Use a grid that gets finer or coarser depending on the local conditions (like a tailor making a suit).
2. Give it a nudge: Sometimes nature needs a little help to start a transition, so adding a small, calculated disturbance can make the simulation much more accurate.

This is a big step forward for designing better, more efficient airplanes that can handle real-world flight conditions where air behaves in both smooth and chaotic ways.

Technical Summary

1. Problem Statement

The paper addresses the challenge of accurately simulating flows over airfoils that contain both laminar and turbulent regions using Wall-Modeled Large-Eddy Simulation (WMLES).

• The Core Conflict: WMLES relies on wall models that typically require the first off-wall grid point to be located within the logarithmic layer (log-layer) of the boundary layer to function correctly. However, in natural transition scenarios, the laminar boundary layer is extremely thin near the leading edge.
• The Dilemma:
  • If the grid is coarse (first point in the log-layer), the thin laminar boundary layer is under-resolved, leading to inaccurate skin friction predictions in the laminar region.
  • If the grid is refined to resolve the thin laminar layer (first point inside the viscous sublayer/buffer layer), the wall model becomes invalid for the downstream turbulent region, often causing delayed transition or inaccurate turbulent predictions.
• Objective: To evaluate WMLES predictive capabilities for a NACA0012 airfoil at $Re_c = 3 \times 10^6$ and $0^\circ$ angle of attack, and to document grid resolution requirements to capture both laminar and turbulent regions simultaneously.

2. Methodology

• Solver: The study utilized FUN3D, a node-based, unstructured-grid finite-volume solver developed by NASA Langley.
• Flow Conditions: NACA0012 airfoil, $M = 0.3$, $Re_c = 3 \times 10^6$, $\alpha = 0^\circ$.
• Wall Modeling: An equilibrium wall model based on the Spalding log-law was used, with the first off-wall grid point serving as the exchange location for wall shear stress.
• Simulation Scenarios:
  1. Fully Turbulent Case: The entire surface was assumed turbulent.
  2. Laminar-Turbulent Case: The flow was assumed laminar up to a transition point ($x_{tr} \approx 0.40$) and turbulent downstream.
• Transition Prediction: Linear stability theory and the N-Factor method were used to predict the transition onset ($x \approx 0.36$) and the most amplified disturbance frequency ($F_0 \approx 6.5 \times 10^{-5}$).
• Grid Strategies Tested:
  • Uniform Grid: A structured C-grid with constant wall-normal spacing (tested at various heights: $3\times10^{-4}$ to $3\times10^{-5}$ chord).
  • Adaptive Grid: An unstructured grid generated based on precursor RANS simulations, maintaining a fixed number of points per boundary layer thickness (20 ppd) to accommodate the varying thickness from laminar to turbulent regions.
• Tripping: In the adaptive grid scenario, unsteady harmonic body forces (mimicking the most amplified instability waves) were introduced upstream of the neutral point to trigger transition.

3. Key Results

A. Fully Turbulent Case

• WMLES successfully captured the turbulent flow, showing good agreement with Wall-Resolved LES (WRLES) and RANS for pressure coefficients and skin friction trends.
• Observation: Turbulence structures triggered very close to the leading edge, even without explicit tripping, likely due to numerical noise or grid-induced tripping.
• Discrepancy: Skin friction was slightly underpredicted near the leading edge compared to RANS, attributed to the coarse grid resolution in that region.

B. Laminar-Turbulent Case (Uniform Grids)

• Coarse Grid ($\Delta z_n \approx 3 \times 10^{-4}$):
  • Turbulent Region: Captured well; skin friction matched RANS.
  • Laminar Region: Poorly captured; skin friction was underpredicted because the grid was too coarse to resolve the thin laminar boundary layer.
• Fine Grid ($\Delta z_n \approx 3 \times 10^{-5}$):
  • Laminar Region: Captured accurately with ~10 points inside the boundary layer.
  • Turbulent Region: Failed. The first grid point fell below the buffer layer, invalidating the wall model. Consequently, the transition to turbulence was significantly delayed (occurring at $x \approx 0.6$ instead of $0.4$).
• Conclusion: A single uniform grid spacing cannot simultaneously satisfy the resolution requirements for both the thin laminar layer and the wall-model requirements for the turbulent layer.

C. Laminar-Turbulent Case (Adaptive Grid + Tripping)

• Grid: An unstructured grid with ~20 points per boundary layer thickness (based on RANS thickness variations) was used.
• Tripping: External disturbances with the most amplified frequency were applied at $x=0.05$.
• Outcome:
  • Laminar Region: Skin friction was captured with excellent agreement to RANS due to sufficient grid points in the wall-normal direction.
  • Transition: The explicit tripping successfully triggered the breakdown of 2D waves into 3D turbulence near the predicted transition location ($x \approx 0.4$).
  • Turbulent Region: The flow remained turbulent downstream with accurate skin friction profiles.
• Velocity Profiles: Mean velocity profiles at $x=0.8$ showed good agreement between the adaptive WMLES, RANS, and WRLES.

4. Key Contributions

1. Resolution of the Grid Dilemma: The study demonstrates that a variable wall-normal grid distribution (based on local boundary layer thickness) is necessary to resolve both laminar and turbulent regions in WMLES.
2. Necessity of Explicit Tripping: It highlights that for WMLES with coarse grids (designed for the outer layer), natural transition mechanisms (instability waves) are often not captured spontaneously. Explicit tripping (via body forces or disturbances) is essential to trigger transition at the correct location when using such grids.
3. Validation of Hybrid Approach: The paper validates a workflow combining precursor RANS (for grid generation and transition location) and WMLES (for unsteady flow physics) as a viable strategy for high-Reynolds number airfoil simulations.
4. Sensor Limitations: The authors note that sensor-based methods (which detect fluctuations to switch models) may fail in WMLES because the coarse grids cannot resolve the initial instability waves, leading to delayed detection of transition.

5. Significance

This work provides critical guidelines for the aerospace community regarding the application of WMLES to realistic configurations where laminar-turbulent transition is significant (e.g., high-lift devices, natural laminar flow airfoils).

• Practical Implication: It suggests that relying solely on a single uniform grid or sensor-based switching is insufficient. Instead, a physics-informed grid generation strategy (using RANS to estimate boundary layer growth) combined with explicit tripping is the most robust approach for capturing mixed flow regimes.
• Future Outlook: The findings support the development of more complex 3D configurations where accurate prediction of transition and viscous drag is crucial for aerodynamic efficiency.

Note on Publication Status: The manuscript was rejected by AIAA Journal and Computers & Fluids (as noted in the header), with reviewers suggesting it was better suited as a conference paper or technical report due to its applied nature and lack of novel theoretical breakthroughs, though it offers significant practical engineering insights.