An accurate calculation of aerodynamic force coe cients for a given geometry is of fundamental importance for aircraft design. High-order spectral/hp element methods, which use a discontinuous Galerkin discretisation of the compressible Navier-Stokes equations, are now increasingly being used to improve the accuracy of flow simulations and thus the force coe cients. To reduce error in the calculated force coe cients whilst keeping computational cost minimal, I propose a p-adaptation method where the degree of the approximating polynomial is locally increased in the regions of the flow where low resolution is identified using a goal-based error estimator. We initially calculate a steady-state solution to the governing equations using a low polynomial order and use a goal-based error indicator to identify parts of the computational domain that require improved solution accuracy and increase the approximation order there. We demonstrate the cost-effectiveness of our method across a range of polynomial orders by considering a number of examples in two- and three-dimensions and in subsonic and transonic flow regimes. Reductions in both the number of degrees of freedom required to resolve the force coe cients to a given error, as well as the computational cost, are both observed in using the p-adaptive technique. In addition to the adjoint-based p-adaptation strategy, I propose a mesh deformation strategy that relies on a thermo-elastic formulation. The thermal-elastic formulation is initially used to control mesh validity. Two mesh quality indicators are proposed and used to illustrate that by heating up (expanding) or cooling down (contracting) the appropriate elements, an improved robustness of the classical mesh deformation strategy is obtained. The idea is extended to perform shock wave r-adaptation (adaptation through redistribution) for high Mach number flows. The mesh deformation strategy keeps the mesh topology unchanged, contracts the elements that cover the shock wave, keeps the number of elements constant and the computation as e cient as the unrefined case. The suitability of r-adaptation for shock waves is illustrated using internal and external compressible flow problems.

No description available.

No description available.

In most practical flow situations the boundary layer is three-dimensional rather than two-dimensional, Three-dimensional boundary-layer flows, however, are not sufficiently understood and many questions regarding their behaviour remain unanswered. In this study several different aspects of three-dimensional boundary-layer flows are investigated. Initially, through the examination of specific problems, we demonstrate the existence of discontinuous solutions in unsteady and steady three-dimensional boundary-layer flows. These case studies demonstrate the existence of pseudo-shocks and examine different aspects of discontinuous solutions, such as how they form and how the existence of a wall affects the solution. For all the problems the proper shock conditions are derived and the existence of a valid shock structure is proven. Furthermore, several new three-dimensional interactive problems with no counterpart in two-dimensional flows are investigated. These case studies illuminate how the boundary layer behaves in the presence of an obstacle for a flat plate and a curved wall configuration. For all cases the linearised problem is solved. For one of the cases a similarity solution is shown to exist and the eigen solutions which allow for upstream influence are analytically derived. For the latter case the solution of the non-linear problem is also presented. Finally, the behaviour of a steady three-dimensional boundary-layer flow, which develops under the influence of a two-dimensional vortex, is investigated numerically. This type of flows, apart from their theoretical importance in understanding three-dimensional separation, also have a significant practical relevance in aerodynamics, as they constitute an idealised model of the flow over a helicopter blade. For the particular problem it is demonstrated that, for a certain set of values of the governing parameters, the flow field develops a singularity at a finite distance from the leading edge. This singularity indicates the impossibility of unseparated flow, which the obtained numerical results suggest is a case of ordinary separation with a vortex breaking away at the position of the singularity.

This thesis describes the development and validation of a new method for the generation of synthetic inlet conditions, referred to as the Synthetic Eddy Method (SEM), for Large-Eddy Simulation (LES). The motivation for this work is the growing interest of the engineering community in hybrid methods coupling Reynolds-averaged Navier-Stokes (RANS) approaches in regions of the flow that are at equilibrium (where RANS can be trusted), with LES approaches elsewhere. The focus of this thesis is on the RANS-to-LES interface inside attached turbulent boundary layers, where an unsteady LES content has to be explicitly generated from a steady RANS solution. The SEM is a stochastic algorithm that generates instantaneous velocity fluctuations from input statistical quantities that are typically available from a RANS solution. The method is based on the classical view of turbulence as a superposition of eddies. The signal is expressed as a sum of synthetic eddies with random position and intensity, and which are subsequently convected through the LES domain inlet. The method generates stochastic signals with prescribed mean velocity, Reynolds stresses, and length and time scale distributions. The SEM is implemented into the unstructured finite volume code, Code Saturne, and used to generate inflow data for the LES of plane channel flow. It is shown that the robustness of the SEM depends strongly on the correct specification of its input parameters (i.e. mean velocity, turbulent kinetic energy, and integral length and time scales). Equations to compute all the input parameters of the SEM from simple RANS statistics are then derived, and the SEM is used to couple an upstream RANS simulation with a LES in the case of simple wall-bounded flows (i.e. channel, boundary layer and duct flows). Generally with synthetic turbulence, some distance is required downstream of the inlet for realistic fluctuations to develop. With the SEM, this development length is found to be approximately 3,000 wall units for all the cases simulated - much shorter than other comparable methods of generation of synthetic turbulence. Hybrid simulations of more complicated turbulent flows involving separation and reattachment (i.e. the flow over a backward facing step and over an airfoil trailing edge) are then performed. The SEM is compared to (and found to perform better than) other existing methods of generation of synthetic turbulence. It is shown that when the SEM is used, the RANS-to-LES interface can be placed only 1−2 boundary layer thicknesses upstream of the region of interest (where the flow separates) without significantly alterating the results.

Due to the growing transport needs in emerging economies and recent success of the low-cost airlines, the demand for short/medium-haul aeroplanes is increasing. Within the next twenty years, the existing single-aisle aircraft are likely to be replaced by new models mounting new propulsion systems. One promising con- figuration being considered is the open-rotor, which is a revision of the propfan. However, further progress has to be done in order to transform propfan engines, whose technology dates back to the 1980s, into viable and feasible open-rotor con- cepts. Among the aspects yet to be investigated in su ficient depth is the de finition of a methodology for the open-rotor nacelle design. The aim of the present research is to help enhance the knowledge in this area. Even if there are a number of important fields of investigation for open-rotor designs, this work is limited to the analysis of the pusher architecture with no exhaust impingement through rotors. The research is initially performed combining both a graphical and a compu- tational approach, investigating the mathematical and physical aspects involved in the de finition of appropriate nacelle pro files, boundary conditions for the CFD analysis and simplifi ed rotor modelling. The first simulations are mainly focused on a typical propfan nacelle, which is taken as a reference model: the computations provide useful results for evaluating its aerodynamic features ... [cont.].

Engineering methods that couple multi-objective optimisation (MOO) techniques with high fidelity computational tools are expected to minimise the environmental impact of aviation while increasing the growth, with the potential to reveal innovative solutions. In order to mitigate the compromise between computational efficiency and fidelity, these methods can be accelerated by harnessing the computational efficiency of Graphic Processor Units (GPUs). The aim of the research is to develop a family of engineering methods to support research in aviation with respect to the environmental and economic aspects. In order to reveal the non-dominated trade-o_, also known as Pareto Front(PF), among conflicting objectives, a MOO algorithm, called Multi-Objective Tabu Search 2 (MOTS2), is developed, benchmarked relative to state-of-the-art methods and accelerated by using GPUs. A prototype fluid solver based on GPU is also developed, so as to simulate the mixing capability of a microreactor that could potentially be used in fuel-saving technologies in aviation. By using the aforementioned methods, optimal aircraft trajectories in terms of flight time, fuel consumption and emissions are generated, and alternative designs of a microreactor are suggested, so as to assess the trade-offs between pressure losses and the micro-mixing capability. As a key contribution to knowledge, with reference to competitive optimisers and previous cases, the capabilities of the proposed methodology are illustrated in prototype applications of aircraft trajectory optimisation (ATO) and micromixing optimisation with 2 and 3 objectives, under operational and geometrical constraints, respectively. In the short-term, ATO ought to be applied to existing aircraft. In the long-term, improving the micro-mixing capability of a microreactor is expected to enable the use of hydrogen-based fuel. This methodology is also benchmarked and assessed relative to state-of-the-art techniques in ATO and micro-mixing optimisation with known and unknown trade-offs, whereas the former could only optimise 2 objectives and the latter could not exploit the computational efficiency of GPUs. The impact of deploying on GPUs a micro-mixing _ow solver, which accelerates the generation of trade-off against a reference study, and MOTS2, which illustrates the scalability potential, is assessed. With regard to standard analytical function test cases and verification cases in MOO, MOTS2 can handle the multi-modality of the trade-o_ of ZDT4, which is a MOO benchmark function with many local optima that presents a challenge for a state-of-the-art genetic algorithm for ATO, called NSGAMO, based on case studies in the public domain. However, MOTS2 demonstrated worse performance on ZDT3, which is a MOO benchmark function with a discontinuous trade-o_, for which NSGAMO successfully captured the target PF. Comparing their overall performance, if the shape of the PF is known, MOTS2 should be preferred in problems with multi-modal trade-offs, whereas NSGAMO should be employed in discontinuous PFs. The shape of the trade-o_ between the objectives in airfoil shape optimisation, ATO and micro-mixing optimisation was continuous. The weakness of MOTS2 to sufficiently capture the discontinuous PF of ZDT3 was not critical in the studied examples … [cont.].

No description available.

Direct numerical simulations (DNS) of turbulent pipe flows are carried out to investigate the suppression of previously-identified internal noise sources with an acoustic liner using a time-domain acoustic liner model developed by Tam and Auriault (AIAA Journal, 34(5), 917-923, 1996). The liner model is implemented and tested in an in-house DNS code. Validation tests are conducted to show its correct implementation in the DNS solver. In order to study the liner model capability a number of tests are carried out with different liner parameters and flow Mach Numbers. To understand the effect of the liner on the acoustic and turbulent components of the unsteady wall pressure, an azimuthal/axial Fourier transform is applied and the acoustic and turbulent wavenumber regimes are clearly identified. It is found that the spectral component occupying the turbulent wavenumber range is unaffected by the liner, whereas the acoustic wavenumber components are strongly attenuated, with individual radial modes evident as each cuts on with increasing Strouhal number. The acoustic wavenumber analysis shows that the acoustic component of the wall pressure prevails over the hydrodynamic wall pressure. This allows the acoustic liner model to dissipate the acoustic field only, leaving the hydrodynamic component statistically unchanged. Furthermore, a DNS of a pipe/jet configuration is computed to study the effects of the acoustic liner model on the far-field noise. Noise prediction is performed using the Ffowcs Williams-Hawkings (FWH) method. The FWH method has also been tested to identify the best configuration of the FWH surface. A conical-shaped surface proved to be a better surface. Furthermore, results show far-field noise reduction when the liner model is present.

In this work the effect of isolated surface roughness on the behaviour of a hypersonic boundary layer is investigated, with a particular focus on the effect of the three-dimensional roughness shape on the instability of the roughness wake and the subsequent transition process. The analysis is performed computationally using direct numerical simulations, which solve the compressible Navier-Stokes equations, and a new code, developed in the scope of the current work, to analyse the linear stability of these equations. The full three-stage roughness-induced transition process has been investigated: firstly, the receptivity process and generation of boundary layer instabilities from freestream disturbances; secondly, the generation of a roughness wake and its initial linear instability; and finally the non-linear breakdown to turbulence of the roughness wake. In particular the effect of the three-dimensional roughness shape on these processes has been studied, looking at the roughness height, frontal profile, planform shape and upward/downward ramps. Also the effect of freestream disturbance amplitude andwall cooling has been investigated. It has been found that the roughness height and frontal profile have a large influence on the stability characteristics of the resulting wake and the subsequent transition. The roughness planform shape has a marginal effect, although cylindrical and diamond-shaped elements yield more unstable wakes than a square roughness element. Bi-local stability analysis can be used in most cases to predict the most unstable wake mode, but it under-predicts the instability growth rates due to non-parallel effects. The roughness shape has been observed to affect the transition onset location. The criteria commonly used to predict roughness-induced transition, do not take into account the three-dimensional shape, and an alternative transition prediction, based on the amplitude of the roughness-induced streamwise streak, is considered.