Metacomp at the 4th High-Lift Prediction Workshop

In 2022, Metacomp presented at the 4^th High-Lift Prediction Workshop (HLPW-4) and contributed to two main areas of study. The first was an investigation into the optimal initial conditions for hybrid RANS/LES over the angle-of-attack sweep, and the second was a comparison between (corrected) free-air simulations (case 2a) and in-situ simulations of the CRM model mounted in the QinetiQ wind tunnel (case 2b, see Figure 1). Hybrid RANS/LES solutions were computed with DDES[1] using SA-QCR[2] as a baseline RANS model.

One clear advantage to using a CFD solver that has both fast steady-state RANS and scale-resolving capabilities is a wider choice of initial conditions when restarting in hybrid RANS/LES (HRLES) mode. When the goal is to minimize the total cost of an expensive scale-resolving simulation, reducing the length of the non-physical start-up transient window becomes critical, and one obvious path toward that is to restart the solver in HRLES mode from an initial condition comprised of a reasonably-well converged RANS solution – especially one in which the RANS model matches the underlying RANS framework in the HRLES. Figure 2 shows three separate strategies that were tested on the free-air configuration ANSA C-level mesh. While, for this range of angle of attack, there are minimal differences between a warm start (starting each HRLES from its solution at the prior alpha) and a cold impulsive start (free-stream conditions for each alpha), one particular approach stands out as apparently inferior – that of starting the HRLES from the converged RANS solution at the same angle of attack. The issue here appears to be a numerical hysteresis; once the RANS model exhibits stall at a particular angle of attack, switching to HRLES mode simply perpetuates the stall, and the high-lift solution is never recovered, even after very long run times. This is atypical behavior of most RANS->HRLES hand-offs, but a useful cautionary tale in situations where the impact of inaccurate RANS initial conditions can have an adverse effect.

Figure 3 shows a mesh convergence study carried out on the in-tunnel geometry, with comparisons against the raw (uncorrected) experimental data. One interesting observation here is that the solutions on all these in-tunnel meshes (even the coarsest mesh-A solution) show better agreement in the linear-lift regime than were obtained (on any mesh) in the free-air simulations (Figure 2). The approach boundary layer in the tunnel, non-uniform inflow and necklace vortices from the model stand-off enable a closer agreement to the raw data, for mesh-converged lift and drag. Because of the reduced domain around the model, the in-tunnel (case 2b) meshes are of comparable size to those of the same-family free-air (case 2a) meshes. This highlights the potential benefit of modeling all geometric detail of the enclosing domain if/when available.

Figure 4 shows instantaneous normalized Q-criterion isosurfaces for the 19.98° in-tunnel (case 2b), showing most of the resolved-scale content arising from the large in-board separation. Figure 5 shows the location of the pressure belts from the experimental model. Figure 6 shows pressure coefficient at two representative pressure-belt locations – belt A inboard and belt H outboard. The significant improvement in time-averaged Cp profiles with SA-QCR-based DDES is apparent at both stations.

Figure 6. HLPW4 case 2b pressure distributions at inboard location A and outboard location H

Figure 7 shows the experimental surface oil-flow visualization of the fuselage and inboard wing section. This can be compared with the steady-state SA-QCR RANS solution in figure 8 and the time-averaged SA-QCR-based DDES solution in figure 9. The superior agreement with the experimental flow topology appears to be a feature of all scale-resolving treatments.

• A comparison was made of various approaches to initialize hybrid RANS/LES calculations of flows over the NASA high-lift CRM. The choice of initial conditions was found to strongly favor either cold-/impulsively-started flow or warm start (from the DDES solution at the prior angle of attack.
• A comparison was also made of free-air versus in-tunnel simulations, with the latter providing notably better agreement (with the raw tunnel data) than the former (using tunnel-corrected experimental data).
• Further details of this work are available in reference [3], with an overall summary of the workshops HRLES contributions given in reference [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. Mani, M., Babcock, D. A., Winkler, C. M., & Spalart, P. R. (2013). Predictions of a Supersonic Turbulent Flow in a Square Duct. AIAA Paper 2013-0860, 2013.
3. Ashton, N., Eberhardt, S., Batten, P. and Skaperdas, V., Towards best-practices for Hybrid RANS-LES simulations of high-lift aircraft geometries, AIAA paper 2022-3524, 2022. 
4. Ashton, A., Batten, P., Cary, A., Holst, K., Summary of the 4^th High-Lift Prediction Workshop Hybrid RANS/LES Technology Focus Group, Journal of Aircraft, 2023