I am working on a final-year CFD project and have been stuck on this for 2 months with no luck investigating a NACA 4412 airfoil at Re = 50,000 and AoA = 10° in ANSYS Fluent 2025 R2. I am trying to reproduce low-Re experimental data before moving on to vortex-generator simulations, but I am getting very different lift predictions depending on the transition model and boundary conditions.

My current setup is:

NACA 4412, chord = 0.1 m

3D structured C-type mesh

Span = 0.025 m

Spanwise faces are now translational periodic

Airfoil remains horizontal; AoA is imposed through the inlet velocity components

Turbulence intensity = 0.35%

Transient simulations

Meshes tested: approximately 2.5M and 3.5M cells

The main problem is the lift coefficient.

Using Transition SST, I generally obtain roughly:

The mean suction-side wall shear indicates separation around 0.3 and 0.35, with the flow remaining separated over much of the aft chord. Increasing the mesh from 2.5M to 3.5M has not produced a major change so far.

I also tested the Transition k-kl-w model. It initially reached 0.6, but later became strongly unsteady, with lift oscillating significantly rather than remaining at that value.

I am now testing SST k-w with the gamma transport-equation transition model. This is producing much higher lift, around 1.06, with cd = 0.9 - 1, which is much closer to the experimental values I am trying to reproduce. However, I am concerned that part of the improvement may be coming from my outer-boundary treatment rather than the transition model itself.

One issue I am particularly unsure about is the correct far-field boundary setup when the airfoil is horizontal and the 10 degree AoA is imposed using velocity components. I have experimented with wall, symmetry, and inlet/outlet conditions on the upper and lower outer boundaries, and the aerodynamic coefficients appear quite sensitive to this.

I would appreciate advice on:

What is the most appropriate outer-boundary setup for this low-speed oblique freestream?

Is the Gamma transport model preferable to Transition SST for a laminar separation bubble at re=50000?

Is a cl difference this large between Transition SST, k-kl-w, and Gamma expected at such a low Reynolds number?

What quantities would you use to determine which model is physically correct  cp, skin friction/separation and reattachment locations, intermittency, etc.?

Is a 25 mm periodic span sufficient for this baseline case, or should I first validate the problem in 2D?

I am trying not to simply choose the model that gives the experimental cl. I want to understand why the models are predicting such different separation/transition behaviour and establish a defensible baseline before adding vortex generators.

Any advice from anyone who has modelled low-Re airfoils or laminar separation bubbles in Fluent would be greatly appreciated.