An omnipresent conflict in aircraft design optimization is the need for fast yet accurate analysis tools. For a broad search in a design space, a trade off is required. One approach is to perform the broad search using low fidelity methods and to perform higher-fidelity calculations only at a few design points. The knowledge gained at these higher-fidelity data points can be formulated as correction factors which are to be applied to the low fidelity methods to approximate the higher-fidelity results. This paper explores the potential to transfer this knowledge across the entire design space considered, focusing on aerodynamic calculations and mass estimates of a scaled UAV wing. The paper presents curve fits which reveal relationships between the correction factors and specific aircraft parameters. On top of that, adaption-based multi-fidelity modeling with a successively increasing number of higher-fidelity samples is applied, in order to explore the potential for automatic, successive knowledge build-up and transfer. The results of the study indicate the feasibility of gradual knowledge build-up and transfer for the considered correction factors. Predictions seem to be possible with high accuracy already on the basis of a few higher-fidelity data points. Furthermore, the study addresses two issues in the context of the application of multi-fidelity methods to the problem at hand. The first issue is that, to date, there is a lack of designated semi-empiric wing mass estimation methods for the UAV class in question. The study uses semi-empiric formulas from other aircraft classes and demonstrates that the associated correction factors can be related to specific aircraft parameters. The second issue refers to the accuracy of the higher-fidelity model. Typically, multi-fidelity models try to approximate the higher-fidelity solution as accurate as possible. However, the "higher-fidelity" model itself is again only an estimation of the real behavior, and may yield physically inexplicable results. This paper presents an approach, which does not approximate the higher-fidelity aerodynamic data itself. Instead, it approximates a fit of the higher-fidelity aerodynamic polar, which incorporates a-priori knowledge about the expected shape of the polar.
«
An omnipresent conflict in aircraft design optimization is the need for fast yet accurate analysis tools. For a broad search in a design space, a trade off is required. One approach is to perform the broad search using low fidelity methods and to perform higher-fidelity calculations only at a few design points. The knowledge gained at these higher-fidelity data points can be formulated as correction factors which are to be applied to the low fidelity methods to approximate the higher-fidelity re...
»