An aerodynamic variable-fidelity modelling framework for a low-observable UCAV

摘要

The extensive amount of aerodynamic data needed for the creation of a Stability & Control data set requires efficient and reliable tools. For this reason, the task of producing a large and dense data set in general relies on ground-facility testing and is conducted after the product definition phase, where, nevertheless, already plenty of aerodynamic data is generated by use of simple low fidelity aerodynamic methods. However, these methods are not capable of predicting complex nonlinear aerodynamic phenomena, which complicates a reliable assessment of new designs and demands more accurate predictions early in the design process. Variable-fidelity surrogate modelling offers an opportunity to include high-fidelity simulation tools at affordable effort by enriching a low-fidelity data set with just a few selected highly accurate results. In this paper, an automated aerodynamic variable fidelity modelling process is proposed to create such an aerodynamic data set at product definition phase. It is demonstrated on an industrial-relevant, agile, low-observable unmanned combat aerial vehicle. An adaptive sampling strategy is applied which proves to be both efficient and accurate. This is supplemented by a model selection algorithm. Moreover, a comparison with a single-fidelity surrogate modelling process shows the variable-fidelity modelling strategy to be superior. (c) 2020 Elsevier Masson SAS. All rights reserved.