Multi-fidelity aerodynamic modeling method of aerospace vehicles based on Gaussian process regression

A multi-fidelity aerodynamic modeling method of aerospace vehicles with shape configuration independent was proposed based on Gaussian process regression model, in order to satisfy the demand of full speed domain and large airspace of aerospace vehicle in the preliminary design stage. The traditional engineering estimation method and computational fluid dynamics (CFD) numerical simulation method were treated as the data sources of low-fidelity and high-fidelity aerodynamic characteristics, respectively. Specifically, a fast estimation model for aerodynamics of aerospace vehicles was established by using the panel method in the enigineering estimation method. Then, high-fidelity aerodynamic performance of aerospace vehicles was achieved based on the three-dimensional compressible Euler equations in the CFD numerical simulation. Moreover, the developed multi-fidelity aerodynamic modeling method was validated by the dual-parameter problems of FTB. The prediction accuracy and stability of the developed multi-fidelity aerodynamic model were better than that of the single-fidelity surrogate model with the same number of high-fidelity data points through comparison and analysis. Meanwhile, the relative error of prediction was less than 1%. The multi-fidelity aerodynamic model was used as the source of aerodynamic data for the reentry problem of the aerospace vehicle, and the influence of single fidelity and multi-fidelity modeling methods on the simulation of re-entry trajectory was compared and analyzed. Results show that the proposed multi-fidelity aerodynamic model can fast provide a high accurate aerodynamic data required from trajectory simulation. © 2023 Zhejiang University. All rights reserved.