TL;DR
This paper presents a computational fluid dynamics (CFD) simulation of airflow over two Micro Air Vehicle (MAV) wing designs at low Reynolds numbers. The study validates their in-house CFD code against experimental data, offering insights into the aerodynamic performance and flow physics crucial for MAV design.
What Problem Does This Paper Address?
Designing Micro Air Vehicles (MAVs) presents unique challenges, primarily due to their operation at low Reynolds numbers (Re). At these low Re regimes (typically 10⁴ to 5 × 10⁵), aerodynamic performance is highly sensitive to wing geometry, and traditional aerodynamic principles may not directly apply. Furthermore, the small physical dimensions of MAVs can lead to unfavorable scaling characteristics. Understanding the complex flow phenomena, such as tip vortices and stall behavior, is critical for developing efficient and controllable MAVs for applications like surveillance in hostile environments. This paper aims to address this by simulating and analyzing the aerodynamic characteristics of two MAV wing configurations at low Re, comparing their performance to experimental data.
Methods
The researchers employed a computational fluid dynamics (CFD) approach using their in-house RANS3D code. This code solves the Unsteady Reynolds-Averaged Navier-Stokes (URANS) equations in a non-orthogonal curvilinear coordinate system with a finite volume solver. For turbulence modeling, they utilized different approaches:
- Black Kite Wing: The standard k-ε turbulence model was used.
- Golden Hawk Wing: The paper investigates the influence of three turbulence models: standard k-ε, Spalart-Allmaras (SA), and Shear Stress Transport (SST).
The simulations were performed on multi-block structured grids with a resolution of 193 × 63 × 69 points, ensuring a y⁺ value of approximately 35 at the first near-wall point. The QUICK scheme was used for discretizing convective fluxes, and a second-order central difference scheme for pressure and viscous terms. The numerical solution employed an iterative decoupled approach similar to the SIMPLE algorithm, modified for collocated variable arrangements.
Datasets
The study uses experimental data from CSIR-NAL wind tunnel measurements for validation. This data pertains to the aerodynamic coefficients (lift, drag, pitching moment) of the full MAV configurations (including fuselage and tail) for both the Black Kite and Golden Hawk wings. Additionally, they compare their results with Fluent simulations performed by ADE, Bangalore for the Black Kite wing.
Key Insights and Ideas
- Validation of CFD Code: The in-house RANS3D code, when coupled with appropriate turbulence models, demonstrates reasonable agreement with experimental data for MAV wings at low Reynolds numbers. This validates the code's capability for such analyses.
- Turbulence Model Impact: For the Golden Hawk wing, the choice of turbulence model (k-ε, SA, SST) influences the prediction of aerodynamic coefficients, particularly at higher angles of attack. The SA and SST models predicted higher lift and drag beyond a certain angle compared to the standard k-ε model.
- Stall Behavior Differences: A notable difference was observed in stall behavior between the two wings. The Golden Hawk wing exhibited an earlier stall in experimental measurements (around 16-17°) compared to the CFD predictions (34-36°), while the Black Kite wing's stall was not clearly indicated in measurements up to 25°.
- Tip Vortex Importance: The simulations successfully captured the formation of tip vortices, especially at higher angles of attack. These vortices are a direct consequence of the pressure difference across the wing and significantly influence aerodynamic performance, particularly the lift-to-drag ratio.
- Full Configuration vs. Wing-Only: Discrepancies between CFD (wing-only) and experimental (full configuration) data are attributed to the influence of the fuselage and tail, highlighting the need for full-configuration simulations or careful comparison with wing-only experimental data.
Key Results
- Black Kite Wing: The RANS3D simulation showed good agreement with experimental data for lift and drag coefficients up to an angle of attack of 15°. The predicted maximum lift-to-drag ratio was 9.72, significantly higher than the experimental value of 4.7 and Fluent's 6.12. The pitching moment also showed good agreement at lower angles.
- Golden Hawk Wing: The study compared three turbulence models. The SA and SST models predicted higher lift and drag beyond 18° and 26° respectively, compared to the standard k-ε model. The lift-to-drag ratio predictions varied, with the standard k-ε and SST models showing better agreement with experimental data than the SA model. The SA model over-predicted the maximum L/D significantly. All models captured the trend of the pitching moment, though experimental data showed a kink at the stall angle not replicated by the simulations.
- Flow Visualization: Surface pressure contours and cross-flow patterns for the Golden Hawk wing revealed realistic pressure distributions and the generation of complex vortices at higher angles of attack. Particle traces clearly illustrated the formation and diffusion of tip vortices.
Limitations and Future Work
The authors acknowledge several limitations and suggest future work:
- Discrepancies with Experiments: Discrepancies between CFD and experimental results, particularly for the Golden Hawk wing, need further investigation. This includes refining grid resolution, exploring different grid topologies, and modeling the effects of flow transition, which is crucial at low Reynolds numbers.
- Wing Tip Closure: The influence of wing tip closure on aerodynamic performance needs to be investigated.
- Full Configuration Simulation: While the current study focused on wing-only configurations for comparison, future efforts may involve simulating the full MAV configuration to better match experimental conditions.
- Transition Modeling: The paper highlights the importance of transition modeling at low Reynolds numbers and suggests it as a key area for future research to improve prediction accuracy.
Summary updated