Improved cooking stoves can bring significant benefits to women and children in rural African situations, due to reduced fuel consumption and improved indoor air quality. This investigation focuses on the use of Computational Fluid Dynamics (CFD) and Genetic Algorithms (GAs) to optimise a stove for Eritrea. Initial work focussed on developing a model of wood combustion in a fixed bed. An experimental investigation was carried out on regular wood cribs to determine the burn rate and temperature field above a wood fire. The experimental data was used to develop a numerical model using CFD software Fluent 6.2 and user-defined functions for the fixed bed of fuel. The model assumed that pyrolysis was limited by heat transfer through the fuel, and that char combustion was limited by oxygen diffusion to the fuel surface. Simulation results yielded a mean and maximum error of 16% and 42% respectively in fuel burn rate. In the second phase of the investigation, the numerical model of wood combustion was used as part of a larger CFD model to capture the behaviour of a complete stove. The model was compared with experimental data for rocket type stoves with different geometries. The model correctly identified the trends of fuel burn rate and heat transfer in the experimental data, though agreement with experimental values was poor and the model exhibited significant errors when altering stove height and diameter. In the final phase of the investigation, the stove model was used in conjunction with a genetic algorithm to optimise the stove shape. Two methods of genetic coding were investigated. The resulting stove is expected to half fuel consumption compared to the classic mogogo stove, though this remains to be experimentally verified.

An experimental investigation has been undertaken in a wind tunnel to study the induced airflow and drag reduction capability of AC glow discharge plasma actuators. Plasma is the fourth state of matter whereby a medium, such as air, is ionized creating a system of electrons, ions and neutral particles. Surface glow discharge plasma actuators have recently become a topic for flow control due to their ability to exert a body force near the wall of an aerodynamic object which can create or alter a flow. The exact nature of this force is not well understood, although the current state of knowledge is that the phenomenon results from the presence of charged plasma particles in a highly non-uniform electric field. Such actuators are lightweight, fully electronic (needing no moving parts or complicated ducting), have high bandwidth and high energy density. The manufacture of plasma actuators is relatively cheap and they can be easily retrofitted to existing surfaces. The first part of this study aims at characterising the airflow induced by surface plasma actuators in initially static air. Ambient air temperature and velocity profiles are presented around a variety of actuators in order to understand the nature of the induced flow for various parameters such as applied voltage, frequency, actuator geometry and material. It is found that the plasma actuator creates a laminar wall jet along the surface of the material on which it is placed. The second part of the study aims at using plasma actuators to reduce skin-friction drag in a fully developed turbulent boundary layer. Actuators are designed to induce spanwise forcing near the wall, oscillating in time. Thermal anemometry measurements within the boundary layer are presented. These show that the surface plasma can cause a skin-friction drag reduction of up to 45% due to the creation of streamwise vortices which interact with, and disrupt the near-wall turbulence production cycle.

Within the eld of hydraulics there is a growing trend towards the use of computer based models, which have proven to be an invaluable tool in engineering. A range of commercial packages is available which encompass different mathematical models and a variety of solution strategies. A number of problems can be identified with the software currently available, and as a result, research continues into developing better numerical techniques for computational hydraulics. The issues most often addressed by researchers consider the application of faster and more accurate numerical methods, many of which were originally developed for gas dynamics problems. There has been a growing trend in favour of Riemann based methods constructed within the finite volume framework. Such methods are noted for their good conservation and shock capturing capabilities. However, the computational cost of employing theses algorithms can lead to excessively long run times, particularly when higher order mathematical models are used. This often is as a result of stability constraints placed upon explicit schemes, which require the smallest possible time step permitted throughout the grid, to be applied globally. One possibility for improving this situation is to use local time stepping, whereby individual cells are advanced by their own maximum allowable time steps. To incorporate this concept into a transient model requires the development of a suitable integration strategy, to ensure that the solution remains accurate in time. Two such strategies developed for the Euler equations are considered within this thesis for application to the Saint Venant equations of open channel flow. Both techniques have been demonstrated to reduce run times and improve the quality of solutions in the regions of discontinuities. The investigation considers the the first order scheme of Roe, together with a second order extension constructed using a ux limiter approach. he eects of using an upwind based source term treatment, specifically developed for Roe's scheme, are also considered, and the source term calculations are incorporated into the LTS framework. Results are presented for a series of steady state and transient test cases, which illustrate how local time stepping can lead to reduced run times and improved solution accuracy. The results also highlight the benets of using an upwind source term treatment, particularly when variations in the channel geometry occur.

Within the bearing chamber of a gas turbine aero-engine, lubrication of the shaft and other bearings is achieved by an oil film which may become significantly disturbed by interacting with a range of chamber geometries which protrude from the chamber wall. Minimizing these disturbances and preventing possible dry areas is crucial in optimizing a bearing chambers design. In addition, multiple obstructions may be located close to one another, resulting in a more complex disturbed film profile than by individual obstacles. Prediction of the disturbance of the film is an important aspect of bearing chamber design. For analysis of the film profile over or around a local obstacle, typical bearing chamber flows can be approximated as an incompressible thin film flow down an inclined wall driven by gravity. The Reynolds number of thin film flows is often small, and for the bulk of this thesis a Stokes flow assumption is implemented. In addition, thin films are often dominated by surface tension effects, which for accurate modelling require an accurate representation of the free surface profile. Numerical techniques such as the volume of fluid method fail to track the surface profile specifically, and inaccuracies will occur in applying surface tension in this approach. A numerical scheme based on the boundary element method tracks the free surface explicitly, alleviating this potential error source and is applied throughout this thesis. The evaluation of free surface quantities, such as unit normal and curvature is achieved by using a Hermitian radial basis function interpolation. This hermite interpolation can also be used to incorporate the far field boundary conditions and to enable contact line conditions to be satisfied for cases where the obstacle penetrates the free surface. Initial results consider a film flowing over an arbitrary hemispherical obstacle, fully submerged by the fluid for a range of flow configurations. Comparison is made with previously published papers that assume the obstacle is small and / or the free surface deflection and disturbance velocity is small. Free surface profiles for thin film flows over hemispherical obstacles that approach the film surface are also produced, and the effects of near point singularities considered. All free surface profiles indicate an upstream peak, followed by a trough downstream of the obstacle with the peak decaying in a “horseshoe” shaped surface deformation. Flow profiles are governed by the plane inclination, the Bond number and the obstacle geometry; effects of these key physical parameters on flow solutions are provided. The disturbed film profiles over multiple obstacles will differ from the use of a single obstacle analysis as their proximity decreases. An understanding of the local interaction of individual obstacles is an important aspect of bearing chamber design. In this thesis the single obstacle analysis is extended to the case of flow over multiple hemispheres. For obstacles that are separated by a sufficiently large distance the flow profiles are identical to those for a single obstacle. However, for flow over multiple obstacles with small separation, variations from single obstacle solutions maybe significant. For flow over two obstacles placed in-line with the incident flow, variations with flow parameters are provided. To identify the flexibility of this approach, flows over three obstacles are modelled. The calculation of flows around obstacles provides a greater challenge. Notably, a static contact line must be included such that the angle between the free surface and the obstacle is introduced as an extra flow parameter that will depend both on the fluid and the obstacle surface characteristics. The numerical models used for flow over hemispheres can be developed to consider film flow around circular cylinders. Numerical simulations are used to investigate flow parameters and boundary conditions. Solutions are obtained where steady flow profiles can be found both over and around a cylindrical obstacle raising the awareness of possible multiple solutions. Flow around multiple obstacles is also analyzed, with profiles produced for flow around two cylinders placed in various locations relative to one another. As for flow over two hemispheres, for sufficiently large separations the flow profiles are identical to a single obstacle analysis. For flow around two obstacles spaced in the direction of the flow, effects of altering the four governing parameters; plane inclination angle, Bond number, obstacle size, and static contact angle are examined. The analysis of flow around three cylinders in two configurations is finally considered. In addition, for two obstacles spaced in-line with the incident flow, the numerical approaches for flow over and flow around are combined to predict situations where flow passes over an upstream cylinder, and then around an identical downstream cylinder. The final section of this thesis removes the basic assumption of Stokes flow, through solving the full Navier-Stokes equations at low Reynolds number and so incorporating the need to solve nonlinear equations through the solution domain. An efficient numerical algorithm for including the inertia effects is developed and compared to more conventional methods, such as the dual reciprocity method and particular integral techniques for the case of a three-dimensional lid driven cavity. This approach is extended to enable calculation of low Reynolds number film profiles for both flow over and around a cylinder. Results are compared to the analysis from previous Stokes flow solutions for modest increases in the Reynolds number.

Liquid hold-up is seen to increase as liquid viscosity and fraction of gas taken off increases suggesting corresponding increase in partial separation of phases. However, effect of liquid viscosity does not become significant until a threshold is exceeded when fraction of gas taken off equals 0.40. In all cases examined, periodicity of flow structures is observed to increase as liquid viscosity increases. Considering the results of the three investigations carried out, it can be concluded that periodicity of two-phase flow structure increases as liquid viscosity increases and transition to co-current annular flow occurs at gas superficial velocity of 21 m/s.

Air quality in urban and industrial complexes is of great importance owing to the many implications on human and environmental health. Air pollution in built-up areas is typically associated with traffic exhaust emissions. High pedestrian level concentrations are the result of a non trivial combination of pollutant sources, climate and city morphological configurations. The increase of urbanisation puts a strain on city resources, resulting in increased use of transport and a denser and more compact urban fabric. The consequence of such a change in city morphology exacerbates current human air pollution exposure. There have been several Computational Fluid Dynamics (CFD) studies on air pollution problems in urban areas, which have largely centred on employing the conventional Reynolds-Averaged Navier-Stokes (RANS) approach, and in all of these investigations, the RANS models have been reported to numerically overpredict pollutant concentration levels when compared to wind tunnel (WT) measurements. In addition, the majority of previous investigations have failed to account for the aerodynamic effects of trees, which can occupy a significant portion of typical urban street canyons. The presence of trees aggravates the pollutant concentration at pedestrian level by altering the air circulation and ventilation. Trees act as obstacles to the airflow, reducing wind velocity within street canyons and restricting air exchange with the above-roof flow. As a result fewer pollutants are dispersed out of the canyon. To address shortcomings of previous numerical investigations, the work undertaken in this project has two main objectives. The study first aims to implement Large Eddy Simulation (LES) to improve the flow and concentration predictions, and second to demonstrate the aerodynamic impacts of trees. A wall y+ approach is used to determine the computational grid configuration and corresponding RANS turbulence model. The approach is evaluated in the present numerical study and is found to be exceptionally useful in resolving flow structures near shear zones, particularly in tree-lined canyons. This allows for the appropriate mesh resolution to be selected, when taking into account a compromise between prediction accuracy and computational cost. In part one of the project, the prediction accuracy of the pollutant dispersion within tree-free urban street canyons of width to height ratios W/H = 1 and W/H=2, are examined using two steady-state RANS turbulence closure models - the standard k-ε and Reynolds Stress Model (RSM) and LES coupled with the advection-diffusion method for species transport. The numerical results, which include the statistical properties of pollutant dispersion, e.g. the mean concentration distributions, the time-evolution and three-dimensional spreads of the pollutant, are then compared to WT measurements available from the online database (CODASC, 2008) www.codasc.de. The accuracy and computational cost of both numerical approaches are compared. The time-evolution of the pollutant concentration (for LES only) and the mean values are presented. It is observed that amongst the two RANS models, RSM performs better than standard k-ε except at the centreline of the canyon walls. However, LES, although more computationally expensive, does better than RANS. A supplementary investigation is performed to illustrate that unsteady RANS (URANS) is not a suitable replacement for LES when wishing to resolve the internally induced fluctuations of flow and concentration fields. URANS fails to capture the transient mixing process. Part two of the research extends the study from the tree-free street canyons by investigating the aerodynamic influence of tree plantings. Configurations of W/H=1 with single row of trees and W/H=2 with two rows of trees are simulated. In both cases, two tree crown porosities are studied, one for a loosely (Pvol = 97.5%) and another for a densely (Pvol = 96%) packed tree crown, corresponding to pressure loss coefficients λ = 80 m-1 and λ = 200 m-1, respectively. Results of the tree-lined cases are then compared to the tree-free street canyons from the previous investigation. It is observed that the presence of trees reduces the in-canyon circulation and air exchange, thus increasing the overall concentration levels. Similar to the tree-free cases, LES performs better than RANS. In addition, it is shown that a wider street W/H = 2 with two rows of trees promotes better air ventilation and circulation with lower pollutant accumulation at pedestrian level, as opposed to a narrow street W/H = 1 with one row of trees. This is also true for tree-free cases.

The commercially available Computational Fluid Dynamics (CFD) software ANSYS-CFX version 11 (2008) is employed to predict the distribution of the bed and sidewall shear stresses in trapezoidal open channels. The investigation includes computation of wall shear stress (1) directly, using CFD for a range of channels layouts (straight, turning, with ridges), and (2), building on the division line concept initially formulated by Leighly in 1932 and later by Einstein in 1942, through the evaluation of the Guo and Julien (2005) equations, as proposed by Cacqueray et al. (2009). These equations include the two complex integral terms II and III, pertaining to viscous and secondary current effects, have been analysed for each cross section, for straight channels. The CFD predictions are validated first against the experimental results of Tominaga et al. (1989) to ensure that the models used are appropriate. Once this is done, the impact of (1) the variation of the slant angle of the sidewalls, (2) the channel aspect ratio and (3) the composite roughness on the shear stress distribution in straight prismatic channels is analysed directly based on the CFD predictions. In wide open channels the lines of boil, consisting of low speed streaks, periodically in the transverse direction, is believed to be due to the initiation of sand ridges on the bed; in other words due to the coupled interaction between moving bed and flow. A numerical analysis on the flow structures created and the distribution of shear stresses on the bed and sidewalls of channel sections having ridges on its bed is therefore carried out to clarify this point and assess the potential consequences on our predictions. Finally, because of obvious practical relevance, as most rivers follow a winding course, numerical simulations on the distribution of shear on the boundaries inside a channel bend is also presented.

Long, thin, flexible cylindrical elements of large scale structures are heavily influenced by the fluid flow around them. Equally, their movement has an appreciable effect on the fluid flow. This two-way interaction leads to complex dynamic behaviour that can cause fatigue and thus reduce operational lifetime. As demand for longer span bridges and drilling in deeper marine environments increases, research into the best modelling practice of this scenario gains importance. The work described in this thesis establishes a suitable method to model in CFD aero/hydro-elastic behaviour of slender cylindrical elements in large scale structures. In order to achieve this outcome, the author has: modelled the drag crisis on a static cylindrical element; developed a suitable FSI coupling program; combined the drag crisis model with the FSI coupling program and validate against published experimental data. The turbulence formulation used was carefully chosen taking into account the flow features that are important to the onset of the drag crisis. An LES formulation capable of adapting the model constant of the SGS model according to local shear conditions was identied as the best candidate to achieve this aim. The fluid and structural solvers used were loosely coupled by an explicit method that achieved a balance of kinetic energy as well as matching displacement at the moving fluid/solid interface. The integration method and implementation of this coupling strategy was verified by running a test case at low Reynolds number that produced a regular sinusoidal lift function on the cylinder that was kept stationary. The displacement, velocity, and acceleration response produced by the structural solver was compared against a closed solution and found to match with an acceptable level of error. A number of FSI simulations with the cylinder free to move in the cross-flow direction only was carried out. The displacement response was compared against published numerical and experimental data and the importance of having a sufficient spanwise dimension of flow domain was highlighted. Simulations with the cylinder free to move in the along-flow direction aswell as cross-flow direction were carried out. In some simulations where lock-in was observed, the effect of the drag crisis was clearly seen. Energy entered into the system as a result of low drag on the upstream motion of the cylinder caused by the drag crisis. More simulations at different velocities are recommended to define a displacement response curve and make further new observations.

Elongated bubbles which are constrained by the walls of a pipe are commonly known as Taylor bubbles. Taylor bubbles are prevalent in industrial gas-liquid flow, where they are commonly found in buoyancy driven fermenters, the production and transportation of hydrocarbons in the oil and gas industry, the boiling and condensing process in thermal power plants, and the emergency cooling of nuclear reactors. These bubbles also exist in the natural world, and are the driving force behind certain types of volcanic eruption. An analysis of the literature identified a paucity of experimental or numerical studies investigating the rise of Taylor bubbles in pipes with a diameter in excess of 0.12 m or in pipes which contain a change in geometry. The aim of this thesis was to gain a better understanding of the behaviour of Taylor bubbles in flow conditions which have not previously been studied. To achieve this, a CFD model was used to simulate the rise of single Taylor bubbles and a set of experiments conducted. The CFD model was validated against the results of published experimental studies, empirical correlations and theoretical predictions. Further validation was conducted using the results of the experimental study which investigated the rise of Taylor bubbles in a pipe of diameter 0.29 m. These experiments confirmed that the theoretically predicted stability and rise rate of the bubble were correct. Bubbles were also shown to exhibit oscillatory behaviour. Sets of parametric simulations replicated the behaviour observed in the experiments and predicted by theoretical models for a wide range of conditions. The qualitative and quantitative experimental behaviour of a Taylor bubble rising through an expansion in pipe geometry was replicated by the CFD model. Bubbles of sufficient length were observed split as they rose through the expansion in diameter, which produced pressure oscillations. The effects of a variation in a number of parameters, including the angle of expansion, the ratio of the upper to lower pipe diameters and the liquid viscosity, were explored.

The wide range of applications of bubbly flows in the industry makes its accurate modelling strongly demanded. The modelling of bubbly flow, however, is not straightforward because it consists of multi-scale structures in both time and space. Furthermore, the experimental verifications of the theoretical and the Direct Numerical Simulations, DNS, available are noticeably scarce especially for bubbles at high Reynolds numbers. Hence, this research aims to make its contribution to addressing the basic roots that make the modelling process difficult. These roots are represented, but not limited, by the bubble dynamics and bubble-bubble interaction. Tracking of single and dual air bubbles in quiescent water is experimentally carried out using a high-speed camera. For the numerical simulations, using ANSYS-FLUENT software with the VOF model, a structured adaptive mesh technique was developed here that is used to achieve a desirable level of refinement of the mesh around the rising bubbles. It has been found that there is a relationship between the lateral migration of an ellipsoidal bubble and the shape oscillation. This relationship, however, has not been observed for the numerical approach. It also overestimated the experimental findings of bubble kinetics and shape oscillation by 30%. Interestingly, this research contributes to awakening the small details in the underlying physics of the interaction between a pair of rising bubbles. It has been found that a slight deviation in the size of a trailing bubble plays an important role in the state of the trailing bubble whether it approaches the leading bubble or separates from at large separation distances. This is considered due to the greater rise velocity that a smaller bubble has in the ellipsoidal regime compared to the larger bubble. Furthermore, when the separation distance between the rising bubbles is decreased, the appreciable acceleration that the trailing bubble owns has led it to approach the leading bubble and coalesce with it. This behaviour is supported by experiments and by the good agreement that the numerical approach showed. An empirical model that predicts the coalescence rate based on the deviation in the size ratio is presented. Finally, the spatial boundaries over which the coalescence of bubble pairs might occur has been numerically and experimentally presented.