Metacomp at the 5th AIAA CFD High-Lift Prediction Workshop 2025

In 2024, Metacomp presented at the 5^th AIAA CFD High-Lift Prediction Workshop (HLPW-5) and offered three main contributions which are shown below in separate sections. The first was the application of advanced shielding functions in hybrid RANS/LES, which proved important for mesh convergence in the RANS boundary-layer regions on finer meshes. The second was the application of these hybrid RANS/LES methods to a mesh-convergence study over the high-lift CRM geometry build-up, results of which showed significantly improved predictions compared to the baseline RANS model. The third was a study of transition and the influence of the geometry progression on the transition onset over the main wing.

HLPW5 Case 1 was a clean-wing configuration at a single angle of attack (11°) designed to probe how hybrid RANS/LES behaves in mostly-attached flow as mesh resolution is increased. The key point in Figures 1 and 2 is that conventional DDES[1] can begin to under-predict total shear stress on very fine meshes in attached boundary layers (an effect often referred to as ‘modeled-stress depletion’), while enhanced shielding (due to that of Deck & Renard[2], denoted DDES‑DR) tends to recover behavior closer to that of the underlying RANS framework in these mostly-attached wing flows.

Figure 2: Case 1 HRLES mesh convergence: DDES vs DDES-DR, drag (left) and lift (right)

Figure 3: Case 1 surface oil flow (SP1.1 viewpoint): SA-QCR RANS (left) vs DDES (middle) vs DDES-DR (right)

As the mesh is refined, DDES’ departure in drag convergence is modest, but the effect is more visible in lift. This is consistent with a net loss of total shear stress when the model transitions toward LES mode inside an attached boundary layer without generating enough resolved turbulence to compensate for the loss of modeled stress. The DDES‑DR trend stays closer to the RANS reference, indicating that the added shielding is serving its intended purpose by preventing premature departure into LES mode in attached boundary-layer regions on fine grids in the absence of resolved-scale content. Figure 3 provides a qualitative flow-topology check from surface oil-flow patterns (the SP1.1 viewpoint along the wing, from the perspective of the fuselage). Conventional DDES results show an exaggerated trailing-edge separation compared with the RANS baseline, while the DDES‑DR predicted flow topology appears much more like that of the RANS solution. The conclusion is that on sufficiently fine meshes, shielding details matter, and enhanced protection can reduce grid‑induced degradation of the predicted boundary-layer flow.

The case 2.2 geometry (see Figure 4) introduced slats and flap-track fairings and was evaluated across an angle-of-attack sweep. The schematic in Figure 5 shows the location of the experimental pressure belts. Figure 6 shows the predicted lift coefficient (compared with the baseline RANS model on the same grid) and the pressure coefficient distribution at a representative outboard belt location at high angle of attack where RANS/HRLES differences are most pronounced.

Figure 6: Case 2.2 RANS vs DDES vs experiment: Lift (left) and Cp at 23.8° belt I (right)

The left of Figure 6 (lift vs AoA) shows that time‑averaged DDES tracks experimental lift more closely as the configuration approaches its high‑AoA regime, where separation and unsteady structures grow and steady RANS typically struggles. This is the canonical niche for HRLES: letting the solver resolve part of the unsteady separated/shear‑layer physics while retaining RANS behavior in attached regions. The right of Figure 6 clarifies why outboard belts are informative: these stations are sensitive to three-dimensional flow development and separation progression. The HRLES curve generally follows the measured suction and recovery trend more faithfully than RANS at this outboard location, consistent with HRLES better representing unsteady separation/reattachment and shear‑layer behavior that tends to be over‑diffused in steady RANS.

Case 2.3 added inboard and outboard flaps, increasing geometric complexity and introducing stronger flap-related separation/interaction. The geometry schematic is shown in Figure 7.  Figure 8 highlights the global lift-curve behavior and an outboard pressure-belt comparison at high AoA.

Figure 8: Case 2.3 RANS vs DDES vs experiment: Lift (left) and Cp at 23.5° belt I (right)

The left of Figure 8 shows that HRLES (DDES‑DR) predicts the peak-lift behavior (CL,max region) comparatively well, but the curves exhibit a small systematic offset over most of the angles of attack – these are apparent in both RANS and HRLES. This was a general tendency shared by all RANS contributions to this workshop. The reasons for this shared low‑AoA bias aren’t certain, but are not obviously indicative of an HRLES‑specific failure mode. There are possibilities of a broader issue such as boundary-layer state/transition, tunnel corrections, or other experimental-condition mismatch.

At the higher AoA condition, (right of Figure 8) the pressure-distribution prediction again favors HRLES over RANS at these outboard stations. This is consistent with HRLES providing a better overall prediction of the total turbulent stresses in separated-flow regions.

Case 2.4 added a nacelle, pylon, and chine (see schematic in Figure 9). Figure 10 again shows lift vs AoA and an outboard pressure belt at high AoA. 

Figure 10: Case 2.4 RANS vs DDES-DR vs experiment: Lift (left) and Cp at 23.6° belt I (right)

The left of Figure 10 shows HRLES with improved CL,max prediction relative to the underlying RANS model, but the low‑AoA lift deficit again persists in both RANS and HRLES approaches, mirroring the pattern already seen in Case 2.3. This persistence across mesh levels and modeling approaches (as described in references [3] and [4]) hints at a systematic condition/physics mismatch rather than a pure turbulence-model deficiency.

The right of Figure 10 again shows HRLES aligning better with experimental pressure trends at the outboard station. For the specific Belt I cut shown here, HRLES retains the overall advantage in matching the measured distribution, reinforcing that the largest payoffs occur in outboard/high‑AoA regions where 3‑D separation and unsteady shear layers matter most.

Figures 11 and 12 show, respectively, surface oil-flow patterns at low and high AoA, illustrating how separation migrates from the flaps (with attached flow over the main element) at lower angles of attack, to the main element (with attached flow over the flaps) at higher angles of attack. These figures provide some evidence of why the low‑AoA regime can still be challenging: at low AoA the main element remains largely attached while separation is present on the flaps (Figure 11), whereas at high AoA separation shifts more strongly onto the main element and the overall flow topology changes (Figure 12). These distinct separation patterns imply that accurate prediction depends strongly on the incoming boundary-layer state at the flap and on how the solver represents the flap shear layer – both of which can be sensitive to transition and to HRLES mode switching.

Figure 13 shows predictions from the Langtry–Menter transition model, which was used here in conjunction with virtual trip wires set at the appropriate locations on the unprotected elements (slats and unprotected main-element surfaces were not tripped in the experiment). This figure shows skin-friction maps for Case 2.2 (left) and Case 2.4 (right). These provide some direct evidence that boundary-layer transition onset over the main wing differs across configurations and may not match the fully-turbulent assumption used in the workshop’s contributed RANS and HRLES solutions.

Figure 13: Langtry–Menter skin friction: Case 2.2 at 6° (left) and Case 2.4 at 7.6° (right)

The left of Figure 13 (Case 2.2 at 6°) shows the transition solution producing laminar regions on the slat (expected because the slat is untripped) but with the main element behaving more like a fully turbulent leading-edge in this configuration. In contrast, the right of Figure 13 (Case 2.4 at 7.6°) indicates laminar persistence not only on the slat but also across appreciable portions of the upstream main-element leading edge. Reference [3] attributes this to an adverse pressure gradient that delays transition on the main element when flaps are present.

This matters because a laminar upstream boundary layer changes the downstream separation extent and pressure recovery, directly affecting lift and moment. If the wind-tunnel flow has delayed transition in Cases 2.3/2.4 while a CFD solution assumes fully turbulent flow from the leading edges, a systematic low‑AoA lift bias can persist even as meshes are refined and even when switching from RANS to HRLES with the same underlying RANS model. This delayed transition provides a plausible explanation for why the low‑AoA discrepancy is shared by RANS and HRLES in the lift curves (Figures 8 and 10).

• In attached-flow regions on fine meshes (Case 1), standard DDES can show modeled-stress depletion; enhanced shielding (DDES‑DR) tends to keep HRLES closer to the underlying RANS (Figures 1–3).
• For high‑AoA conditions and outboard stations, HRLES improves agreement with measurements in both forces and pressure distributions (Figures 6, 8, and 10).
• Oil-flow patterns highlight that ‘easy’ low‑AoA conditions still include separated flap flow, while high‑AoA conditions shift separation onto the main element, changing the areas of the flow-field that the turbulence treatments are expected to get right (Figures 11 and 12).
• Transition-prediction skin-friction maps (Figure 13) suggest a delayed transition on the main element when flaps are present, providing a plausible explanation for the persistent low‑AoA lift deficit in Cases 2.3 and 2.4.

1. Spalart, P. R., Deck, S., Shur, M. L., Squires, K. D., Strelets, M., & Travin, A.  A New Version of Detached-Eddy Simulation, Resistant to Ambiguous Grid Densities. Theoretical and Computational Fluid Dynamics, 20(3), 181–195, 2006.
2. Deck, S., & Renard, N. Towards an Enhanced Protection of Attached Boundary Layers in Hybrid RANS/LES Methods. Journal of Computational Physics, 400, 108970, 2020.
3. Ashton, N., Batten, A., Skaperdas, V. & Holst, K., High-Lift Prediction Workshop 5: Hybrid RANS/LES Technology Focus Group Summary. AIAA 2025-0048. AIAA SCITECH 2025 Forum. January 2025
4. Batten, P., Ashton, N., & Skaperdas, V. Hybrid RANS/LES Computations of the HLPW-5 Test Cases Using CFD++. AIAA 2025-0685, 2025.