Improvements on the application of direct-CFD in unsteady aeroelastic simulations

Improvements on the application of direct-CFD in unsteady aeroelastic simulations

Verdonck, Hendrik (TU Delft Aerospace Engineering)

van Zuijlen, A.H. (mentor)

Delft University of Technology

Aerospace Engineering

2019-02-25

Aircraft manufacturers have to prove a flutter free design for all operational cases within the complete flight envelope plus a safety margin. This certification process relies on validated flutter computations which have to be made for all flight conditions including variations in the aircraft's loading and failure cases. The aerodynamic component of these unsteady aeroelastic simulations is restricted to fast computational methods due to the large parameter space. Linear, inviscid models were the industry standard for this application. They had to be corrected by wind-tunnel or CFD data to cover transonic flow phenomena. Recently, high-fidelity aerodynamic models have been introduced which, in contrast to the previous methods, have an inherent quality to represent the important transonic flow phenomena for all flight conditions. However, the significant computational cost of these models restricts the computation of unsteady aerodynamics to a limited set of reference elastic modes. These unsteady aerodynamic reference results are subsequently mapped to all 'production' flutter computations with a least-squares method. This thesis report presents an investigation on the possibility of accuracy or robustness improvements in the implementation of this new, high-fidelity direct-CFD method in unsteady aeroelastic simulations. An error estimation study quantified the impact of approximation errors of the least-squares method on the frequency and damping curves of the 'production' computational case. Modal basis quality criteria are established and their performance is compared for a test case. In contrast to the proposed hypothesis, global mode assurance criteria are sufficient to predict the errors. Local or aerodynamically weighted quality criteria show similar performance and can therefore be considered redundant for the presented test case. In case of a non-satisfactory reference set, the modal basis can be enriched automatically and effectively in order to eliminate the approximation error. Additionally, a performance study of two reference selection methods on four computational test cases has been conducted. The application of the elastic modes of a nominal structural lay-out as reference is satisfactory for nominal structure flutter computations at different load distributions and for failure case simulations without strongly deviating mode shapes. However, this reference selection method can be insufficient regarding the approximation error for critical failure cases with strongly deviating mode shapes with respect to the nominal structure modes. Yet, for all test cases the error estimators are able to predict this approximation error, such that they can be eliminated by modal basis enrichment. On the other hand, a new method is proposed which uses the POD (Proper Orthogonal Decomposition) theory to decompose a wide range of modes for different failure and load cases into a reference set. This method performed satisfactory for all test cases, without any enrichment necessity. The prerequisite for a good performance of this reference selection method is a well-considered selection of the POD setup and the presence of the considered failure cases in the POD input.

Aeroelasticity
Flutter
Direct-CFD

http://resolver.tudelft.nl/uuid:b65e68ed-933e-4f6c-9e4a-305c44166569

2024-02-28

Student theses

master thesis

© 2019 Hendrik Verdonck