Tubercle-Modified NACA 0012

3D RANS · Langtry–Menter Transition SST · ANSYS Fluent · 78 simulations

Humpback whales have bumps along the leading edge of their flippers, called tubercles. They create spanwise vortices and delay stall. Literature says that these can be applied to airfoils to better performance, but I noticed a gap in Reynolds number studies. So I swept three Reynolds numbers regimes to test the tubercle effect.

The results disagree on any Reynolds dependence. Some say there is none, some say it increases, some say it decreases. Additionally, the studies ignore a critical flight regime that is becoming all the more relevant with the expansion of UAV technology, the transitional regime.

Velocity pathlines over the tubercle-modified NACA 0012, showing spanwise flow structures forming behind each tubercle trough
Velocity pathlines over the tubercle-modified section at 12º angle of attack. The local velocity increases at the troughs, contributing to boundary layer attachment and the mechanism by which the tubercles delay stall. ANSYS Fluent 2025 R2 Student.

Methods

Two 0.25 m chord NACA 0012 sections, one baseline and one tubercle-modified, were simulated in ANSYS Fluent 2025 R2 Student using the four-equation Langtry–Menter Transition SST (γ–Reθ) model. The leading edge follows a cosine profile, f(x) = a cos(2πx/λ), with amplitude a = 0.03c (7.5 mm) and wavelength λ = 0.25c (62.5 mm), matching the geometry used by Bardera et al. (2024).

Each configuration ran at 13 angles of attack, 0° to 24° in 2° increments, across three Reynolds numbers: 1×105 (laminar, 5.843 m/s), 5×105 (transitional, 29.215 m/s), and 1.5×106 (turbulent, 87.644 m/s). 78 runs total. The domain extends 10c upstream as a semicircle and 20c downstream. Meshes were 923,646 elements (control) and 916,225 (tubercle), with a first-layer height of 0.000025 m and 15 inflation layers at 1.2 growth, yielding y+ = 0.172 and 0.184 respectively. Inlet turbulence intensity was 0.1%. A coupled solver ran 500 iterations per case. I plotted the lift curves against XFOIL as a trend-match, not a one to one comparison.

SolidWorks model of the tubercle-modified airfoil showing the sinusoidal leading edge built across eight construction planes
Geometry generation in SolidWorks. The sinusoidal leading edge is lofted across eight construction planes; profiles were generated from airfoiltools.com.

Results

The tubercle effect is non-monotonic. In the laminar case the tubercles hurt performance significantly. In the other two they help, and they help most at the highest Reynolds number. Values below are at 4° angle of attack, against the unmodified control. Once the angle of attack exceeds a certain threshold, the tubercle effect becomes negative again, in all cases. This is due to the tubercles increasing the local turbulence on the upper surface of the airfoil, which in turn increases the drag coefficient and lowers the L/D ratio.

Turbulence kinetic energy contour over the tubercle-modified wing, showing elevated turbulence in streamwise bands behind each trough
Turbulence kinetic energy on the upper surface. The elevated bands running back from each trough are the local turbulence that drives the drag penalty at higher angles of attack.
Re = 1×105 — laminar

CL −25.88% · CD +28.66% · efficiency −24.41%

Re = 5×105 — transitional

CL +8.56% · CD +3.31% · efficiency +5.08%

Re = 1.5×106 — turbulent

CL +12.06% · CD +3.55% · efficiency +8.23%

Stall behavior changed in every case. The control section stalls at 22°. The tubercle section holds lift through 22° and 24°.

So the Reynolds dependence is non-monotonic, and the benefit is specific to the regime. That may be part of why published results disagree. Studies reporting harm tend to sit below Re ≈ 3×105 and studies reporting benefit above Re ≈ 5×105.

Lift coefficient versus angle of attack at Reynolds number 1.5 million, comparing control, tubercle, and XFOIL results
Lift curve at Re = 1.5×106. XFOIL peaks near CL = 1.4 and the two RANS curves near 0.8. That gap is discussed under Limitations.
Lift-to-drag ratio versus angle of attack at Reynolds number 1.5 million, control versus tubercle
Lift-to-drag ratio at Re = 1.5×106. The tubercle advantage holds at low angle of attack and is gone past the crossover.

Limitations

This is the most important aspect of a purely computational study. The RANS solvers are, by nature, averages, and they minimize the effects of unsteady flow phenomena. Additionally, the constraints within Ansys Student limit the mesh fidelity considerably.

No wingtip vortices

The spanwise boundary conditions prevent wingtip vortices from forming. This isolates the tubercle and airfoil interaction, but it also means the results only apply to high aspect ratios and effectively infinite spans. They should not be carried over to a finite-wing UAV without a separate wingtip analysis.

Reynolds Averaged Navier Stokes (RANS)

Langtry Menter Transition SST underpredicts CL and overpredicts CD against XFOIL, and gets unreliable above 16°. RANS averages out unsteady behavior that matters once the flow separates. The same offset shows up in Bardera et al. (2024) and Ali et al. (2024), so I read it as a limit of the model rather than a mistake in my setup. It was reassuring to see that peer-reviewed literature observes the same phenomena.

Convergence

The 1×10-5 residual criterion was not met. Residuals stabilized with small oscillation after roughly 100 iterations, and reported values are averages over the final 200 iterations to reduce uncertainty. Laminar-to-turbulent transition on the upper surface is difficult to resolve at this mesh density and contributes to the coefficient underprediction.

Mesh budget capped at one million cells

The Fluent Student Edition caps the mesh at one million cells. Most of that budget goes to the inflation layer to keep y+ below 1, which leaves less for the far field. A refinement zone downstream helps with resolving the drag coefficient, but there are inherent diminishing returns in the refinement choices.

Generalizability

A single tubercle amplitude and wavelength were tested, so nothing here generalizes across tubercle geometries. The NACA 0012 airfoil was intentionally chosen to isolate the tubercle effect from camber effects, but this prevents generalizability.

Tools

ANSYS Fluent, SolidWorks, Autodesk Fusion 360, XFOIL.

Conducted as an AP Research project. AP Research award, 1st Annual Chaminade Research Symposium · Honors, Long Island Science and Engineering Fair · 3rd place, SAWAA Science Fair.