A Surrogate Model Approach in 2-D Versus 3-D Flapping Wing Aerodynamic Analysis

Abstract

Flapping wing MAV aerodynamics is challenging to understand due to its complexities. Surrogate modeling offers an effective tool to predict information at off-design points, measure the sensitivity of design variables involved, and illustrate general trends in the data. We have investigated 2D and 3D hovering airfoil/wing aerodynamics considering three kinematic parameters, i.e. plunge amplitude, angular amplitude, and phase lag, at Re = 100. While leading edge vortex, a persistent downward jet, and wake capture are three noticeable fluid dynamics features in 2D and 3D, tip vortices in 3D and instantaneous AoA can substantially affect the relative importance of them. Surrogate models of the 2D and 3D wing with aspect ratio of 4 show that for i) power required, the magnitudes and global trends of the predicted response between 2D and 3D are similar, ii) lift, the phase lag changes from a non-linear response in 2D to a monotonic one in 3D. The overall consequence is that in 3D, kinematic combinations of higher lift with lower power requirements can be attained. Global sensitivity analysis shows that the lift is the most sensitive to the phase lag in 3D as opposed to the angular amplitude in 2D. The ensemble surrogate strategies performed employed performed quite well compared to independent test data.

Modeling and Analysis of Shock Impingements on Thermo-Mechanically Compliant Surface Panels

“…Also recently, several studies 40,42,43 have investigated kriging surrogates for model reduction of the aerothermodynamics. Results from [40] and [42], as well as similar work done in lower Mach number regimes, 41,44,45,49 illustrate that surrogate based approaches are promising for accurate and efficient modeling of complex fluid dynamic phenomena. In particular, some of this work suggests that surrogates produce more accurate models than POD.…”

“…Typical approaches for model reduction of unsteady, nonlinear flows include proper orthogonal decomposition (POD), [36][37][38] Volterra series, 38,39 and surrogates. [40][41][42][43][44][45] Each of these seek to identify the primary features of a system from a limited number of high-fidelity flow solutions. The generation of this dataset requires an initial computational investment.…”

Aeroelastic and Aerothermoelastic Analysis in Hypersonic Flow: Past, Present, and Future

“…Typical approaches for constructing aerodynamic ROMs are proper orthogonal decomposition (POD) [81][82][83][84], Volterra series [82,85], and surrogates [71,[86][87][88][89][90][91][92]. Each of these approaches seeks to identify the primary features of a system from a limited number of full-order flow solutions.…”

Robust Treatment of Temperature Feedback in Aerothermodynamic Models for Fluid-Thermal-Structural Analysis

“…Popular methods include proper orthogonal decomposition (POD), [24][25][26][27] Volterra series, 27,28 and surrogates. [29][30][31][32][33][34][35] A consistent requirement for each is the incorporation of desired model input parameters to the flow responses using a series of empirical observations. In the context of F-T-S analysis in hypersonic flows, the input parameter space consists of a wide range of operating conditions and potential thermo-structural responses.…”