Aerodynamic models can be constructed from multiple types of data (i.e., flight test, experiment, numerical simulation, etc.), but the cost and the accuracy of data is different. This work tries to reduced the total cost for aerodynamic modeling by incorporating different types of data (a few high-fidelity data + a low-fidelity model), and achieves high-fidelity aerodynamic prediction. Main research topics involve:

• Developing data-driven frameworks to model the difference between the low-fidelity and the high-fidelity models.

• Using reduced-order models as the low-fidelity model to accelerate the multi-fidelity aerodynamic prediction.

• Reconstructing high-fidelity experimental results by the proposed multi-fidelity modeling approach and numerical simulation with very low experimental cost or limited data (e.g., a few pressure sensors).