A combined numerical and experimental approach to investigations of body-freedom flutter on flying wings

• Main Author: Ang, Elijah Hao Wei
• Other Authors: Ng Bing Feng
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Aeroelasticity; Unmanned aerial vehicles; Body-freedom flutter; Wind tunnel; Gust load alleviation
• Online Access: https://hdl.handle.net/10356/173670
• Institution: Nanyang Technological University

With the recent rise in the development of aircrafts with high aspect-ratio wings adopting lightweight flexible structures, these aircrafts become increasingly vulnerable to aeroelastic phenomena. One such example is flutter, a self-excited instability resulting from the coupling between inertial, structural, and aerodynamic forces. Presently, investigations into flutter assumes a clamp-fixed boundary condition, resulting in a cantilever configuration. However, in flying wing configurations, strong interactions between rigid-body modes and elastic modes can be observed, leading to body-freedom flutter (BFF), where the fuselage of the aircraft undergoes significant pitching and plunging oscillations. If uncontrolled, the system can undergo diverging oscillations which can lead to catastrophic failure. In such scenarios, the fuselage cannot be assumed to be fixed, and rigid-body modes have to be considered. Hence, this brings about a need for a new methodology to flutter investigations, as conventional methods used in the past may not be applicable to flying wings undergoing BFF. This thesis investigates the BFF phenomenon of a flying wing aircraft by presenting both experimental and numerical approaches. Firstly, a wind tunnel test is performed to investigate the effects of several design parameters, namely wing sweep angles, tip weights and spanwise location of weights, on the flutter speed and frequency of a pseudo-free-flying unmanned aerial vehicle. A novel slider and rail setup is designed to enable rigid-body pitching and plunging degrees-of-freedom of the unmanned aerial vehicle model which are crucial for BFF to occur. Meanwhile, bending-torsion flutter (BTF) of a cantilevered wing is also performed to compare the difference in flutter characteristics. BTF and BFF experience different dominant modes while undergoing flutter, with BTF’s dominant mode being torsion and BFF being the rigid-body short period mode, resulting in lower BFF speeds when compared to BTF. In general, BTF speeds increase as weights are moved towards the wingtip or increasing tip weights, which increases the inertia of the system to enhance stability. The effect is opposite for BFF where an increase in tip weights or movement of weights outboard leads to a reduction in static margin that destabilises the system. Higher sweep angles are found to experience increased sensitivity of flutter speeds to changes in tip weights or location of weights along the wingspan. Subsequently, an aeroelastic framework is adopted for numerical studies of BFF. The aerodynamics are defined using the unsteady vortex lattice method, which is highly applicable to cases where large kinematic movements are involved. The structural dynamics are modelled using an Euler-Bernoulli beam which is tightly coupled with the rigid-body motions of the flying wing, and subsequently discretised using the finite-element method. Both the aerodynamics and structural dynamics are coupled and linearised about an initial reference configuration, and presented in a monolithic state-space form which allows for a computationally efficient asymptotic stability analysis and time-marching solution. For the entirety of this work, a flying wing based on the X-56A MUTT is chosen to demonstrate the numerical aeroelastic model, and numerical results are validated against the wind tunnel experiment. Asymptotic stability of the X-56A MUTT is investigated for varying stiffness and inertial properties to determine the flutter boundaries for the various modes of flutter. In general, two modes of BFFs are observed, the first being a plunge dominant mode, while the second is a pitch dominant mode. Changes in the stiffness properties cause a coalescence in different modes with similar frequencies, resulting in the transition from one mode to the other. It is observed that increasing bending stiffness of the wing leads to the plunge dominant flutter mode which is resulted from coalescence between the rigid-body plunge mode and the elastic bending mode, strengthening the coupling between the two. On the other hand, increasing torsional stiffness leads to the torsional dominant flutter mode due to enhanced coupling between the rigid-body pitch mode and the elastic torsional mode. While keeping the mass and the static margin of the entire flying wing constant, it is possible to investigate the effect of mass of flexible wing on the BFF modes and speed. It is observed that increasing the mass per unit length of the flexible wing causes a shift from plunge dominant to pitch dominant mode. This observation is caused by an increase in stability in the elastic bending degree-of-freedom. Finally, the monolithic state-space formulation for the aeroelastic model allows for efficient control synthesis for active gust load alleviation strategies. Due to the large number of aerodynamic states present, a model reduction technique based on balanced realisation is used to improve computational efficiency during controller synthesis. Open-loop responses of the flying wing due to discrete 1-cosine gust and continuous von Karman spectrum gusts demonstrate the significant rigid-body pitch and plunge oscillations experienced by the flying wing as it undergoes BFF. Two control strategies are implemented for gust load alleviation of lightweight flying wings undergoing body-freedom oscillations, namely a genetic linear quadratic Gaussian controller (LQG) and an artificial neural network (ANN) controller. In the genetic LQG, weights are optimised using a genetic algorithm which maximises a defined fitness function. In general, the genetic LQG controller is able to reduce the plunge displacements by up to 94.2% while effectively damping out wingtip displacements due to wing bending for both discrete and continuous gusts. Similarly, the ANN controller is able to regulate both the plunge displacements and wingtip displacements, including gust cases that are not presented during the neural network training phase. It is observed that the ANN controller is more effective in correcting the wingtip displacements during discrete gusts than the LQG controller, while the opposite is true for the continuous gust cases. The ANN controller offers several advantages over the LQG controller, including the elimination of the need for a Kalman filter for full state estimation and its non-linear nature as a controller.