This work proposes a multi-fidelity modeling approach for predicting aeroelastic flutter of airfoil and wing shapes. Using aerodynamic models based on the doublet lattice method (DLM) and time-accurate Euler equations, cokriging-based surrogates of the aerodynamic influence coefficients (AICs) are generated as functions of Mach number and reduced frequency. The surrogate-based AICs are then used in the p-k method to determine flow conditions with zero aeroelastic damping. To demonstrate the multi-fidelity process, a popular pitching and plunging airfoil case is considered. Verification of the approach is done by comparing with results from FUN3D's internal aeroelastic solver, as well as data from the literature. Results demonstrate the advantages of using multi-fidelity cokriging in relation to kriging models employing the same high-fidelity training data. These advantages are especially clear when using very few high-fidelity samples, where the cokriging approach matches the high-fidelity trend significantly better, while also improving more consistently as the number of training samples is increased. Challenges associated with the AIC nonlinearity are also addressed by examining flow conditions with and without a shock.
展开▼
