A wide range of model calibration methods and formulas have been introduced in the literature for calibrating low-fidelity (LF) computer models against high-fidelity (HF) computer models or data. Although most existing model calibration techniques assume that LF and HF models have identical inputs, in some engineering applications, for example, wing aerodynamics computations, inputs to LF and HF models are often defined differently due to different levels of abstraction in modeling or simulation. For these problems, this paper proposes a new model calibration method that calibrates a mapping from HF model inputs to LF inputs by matching HF and LF model outputs. The method incorporates regularization to prevent overfitting and to allow calibration parameter selection. In the application to calibrating aerodynamic simulation models, three advantages of the proposed method are demonstrated. First, it achieves higher calibration accuracy than the traditional bias correction method when HF data are scarce. Second, it provides convenient and effective calibration parameter selection in the calibration process. Finally, it enables physical insights to be drawn from the calibrated input mapping. Specifically, for the test cases examined, the calibration adjusts wing twist angles to compensate for the neglect of thickness in the vortex lattice representation of the wing geometry.
ASJC Scopus subject areas
- Aerospace Engineering