Adjoint quasi-three-dimensional aerodynamic solver for multi-fidelity wing aerodynamic shape optimization

Abstract

A quasi-three-dimensional method for wing aerodynamic analysis for drag prediction is presented. This method can predict the wing drag with a level of accuracy similar to higher fidelity three-dimensional CFD analysis, with a much lower computational cost. A tool has been developed based on the proposed method and the outputs of the tool have been validated using a higher fidelity CFD tool. Another advantage of the mentioned method (and the tool developed based on that) is to compute the derivatives of any function of interest, such as the wing drag, lift, or pitching moment, with respect to the design variables, mainly the wing geometry, using analytical methods. The tool uses a combination of the Adjoint method, the chain rule for differentiation, and the automatic differentiation to compute the sensitivities. The quasi-three-dimensional aerodynamic solver is used for a multi-fidelity wing aerodynamic shape optimization. A trust region algorithm is used to connect the low fidelity aerodynamic solver to a high fidelity CFD tool for wing drag prediction. The derivatives of the objective function are computed using the low fidelity solver, and the high fidelity solver is used to calibrate the results of the low fidelity one.