Wing stiffness parameterisation for surrogate models

Bennett Leong*, Simon Coggon, Jonathan Cooper

*Corresponding author for this work

Research output: Contribution to journalArticle (Academic Journal)peer-review

Abstract

Surrogate models are key enablers for robust, multidisciplinary design in aerospace. However, a critical question for their efficient use remains in terms of how the aircraft should be parameterised into design variables. This study investigates the parameterisation of wing stiffness for a long-range, twin-engine transport aircraft by comparing the predictions of surrogates with different input parameterisations to the outputs of a physics-based aircraft load analysis. The stiffness parameters were defined to be percent changes in wing stiffness versus a baseline aircraft applied on discrete spanwise sections of the wing box. The number of stiffness parameters was varied between 6 and 64 to see if some of the wing box sections could be grouped together by averaging the stiffness changes across the wing section to reduce the dimensionality of the problem. In addition, a sensitivity analysis using Sobol indices was performed which demonstrated which stiffness parameters contributed most to the variability in wing loads during various manoeuver and static and dynamic gust cases. It was found that geometry-based stiffness parameters show poor consistency between parameterisations and thus the stiffness changes applied over discrete sections of the wing cannot be combined into an average stiffness change without a significant sacrifice in accuracy. However, good consistency was shown in the sensitivity of load responses between different parameterisations.

Original languageEnglish
JournalCEAS Aeronautical Journal
Volume2020
DOIs
Publication statusPublished - 6 Feb 2020

Keywords

  • Multidisciplinary optimisation
  • Surrogate models
  • Uncertainty quantification and management

Fingerprint Dive into the research topics of 'Wing stiffness parameterisation for surrogate models'. Together they form a unique fingerprint.

Cite this