Oil-air separation is a key function in aero engines with closed-loop oil systems. Aero-engine separators are employed to separate oil from air before being released overboard. Typically, these devices make use of a porous medium such as an open-cell metal foam, in order to enhance oil separation. Although quite scarce, there has been some research aimed at developing a suitable modelling framework for aero-engine separators. However, numerical modelling of the air/oil flow through the open-cell metal foams employed in aero-engine separators has never been properly addressed. This thesis presents the development of a pore-scale numerical modelling approach to determine the transport properties of fluid flow through open-cell metal foams. Micro-computer tomography scans were used to generate 3D digital representations of several commercial open-cell metal foams. A code was developed in Matlab to render the CT images into 3D volumes and perform morphological measurements on the samples. Subsequently, conventional finite volume simulations are carried out in order to obtain the airflow and compute the pressure gradient across the investigated samples. Simulations were performed for a wide range of Reynolds numbers and the feasibility of using Reynolds-averaged Navier-Stokes (RANS) turbulence models is investigated. Validation was done by comparing the pore-scale pressure gradient results against experimental measurements. Further simulations were carried out to isolate and analyse particular effects in more detail, such as wall and entrance effects, fluid compressibility, time-dependent flow features, anisotropy of the foam structure and the impact of porosity and surface area on the pressure gradient. The oil phase within aero-engine separators has the form of disperse droplets. Thus, the oil phase in the pore-scale simulations was modelled using a Lagrangian particle tracking approach. Lagrangian simulations were run in steady state and one-way coupled, due to the low mass fraction of oil normally present within aero-engine separators Converged airflow pore-scale solutions were employed as the base flow for the Lagrangian tracking approach. A simplified oil capture criterion assumed the droplet trajectory to be terminated upon collision against the foam solid ligaments. The focus of the present work was on separation of small droplets with a diameter smaller than 10 microns. Hence, a series of calculations were performed using a representative droplet diameter range, and multiple flow velocities. The outcome of such approach was a qualitative evaluation of the oil separation effectiveness for several commercial open-cell metal foams under a representative range of flow regimes. Furthermore, rotational effects which are experienced by the metal foams within aero-engine separators were modelled using a moving frame of reference (MRF) approach. Finally, a methodology for upscaling the results obtained by the detailed pore-scale simulations into a simple macroscopic porous medium model is described, showing promising results. One of the aims of this work was to develop a numerical modelling framework able to provide an accurate representation of the airflow and a qualitative assessment of the oil capture within aero-engine separators. The feasibility of using the current state-of-the-art modelling framework is assessed. The separator design and geometry are based on the oil separation test rig located at the Karlsruhe Institute of Technology (KIT). Experimental measurements of the overall pressure drop and oil capture performed at KIT are used to validate the simulations. The methodology presented here overcomes some limitations and simplifications present in previous similar studies. The upscaled macroscopic porous medium model was applied to full aero-engine separator CFD simulations. Experiments and simulations were conducted for three different separator configurations, one without a metal foam, and two with metal foams of different pore sizes. For each configuration, a variation of air flow, shaft rotational speed and droplet size was conducted. The focus was on the separation of droplets with a diameter smaller than 10 \textmu m. Single-phase air flow simulation results showed that overall pressure drop increases with both increased shaft speed and air flow, largely in agreement with the experiments. Oil capture results proved to be more difficult to be captured by the numerical model and indicate that droplet re-atomization might play a significant role in the oil separation phenomena. Re-atomization, droplet-droplet collisions and droplet breakup were not considered at the present stage, but could be subject of future work. The modelling framework described here should not be seen as a definite answer but as an improvement upon the current state-of-the-art methodology, providing important lessons and recommendations for future work on aero-engine separators.

Considering the gap in available information and the need of the industries such as oil and gas production, energy, and food processing, this study focuses on the two phase flows around bends in process pipe lines. The aim of this study is to investigate the influence of 90° bends on the gas-liquid two phase flow behaviour in vertical and horizontal orientations using advanced two-phase flow measuring techniques. An experimental study has been conducted using silicone oil with a viscosity of 5 mPa.s and air to examine the transitional flows around 90° bends of 68 mm internal diameter (ID) with different configurations. Experiments were conducted at ambient conditions in an open system which consists of a 68 mm ID riser of 4.5 m long, vertical upward 90° bend and two horizontal sections of a 9.2m and 5.5m long in series with a horizontal 90° bend in between. The experimental matrix comprises 60 combinations of gas and liquid superficial velocities, ranging from 0.045 m.s-1 to 3.21 m.s-1 and 0.15 m.s-1 to 0.53 m.s-1 respectively. The phase distributions within the pipes were measured using Electrical Capacitance Tomography (ECT) and Wire Mesh Sensors (WMS). The behaviour of the flow was examined qualitatively using high speed imaging. To study the flow development in the riser, both ECT and WMS were placed in series and moved along three axial locations downstream of the mixing section. During the experiments at the bends, the ECT was kept immediately upstream while the WMS was moved to different positions downstream of the bend. The cross-sectional void fraction time series from the ECT and the WMS were used to quantify the main hydrodynamic parameters of the flow including cross-sectional averaged void fraction, bubble size distribution, radial void fraction profiles, slug length, slug frequency, void fraction in liquid slugs, and the slug bubble velocity. Results were compared against the available slug flow correlations. The results show that the phases separate shortly after the vertical to horizontal bend leading to stratified or wavy stratified flow. Beyond a certain threshold of the gas flow rate and liquid level, onset of slugs can be observed at a certain distance downstream of the vertical upward bend. This work suggests that the formation of hydrodynamic slugs downstream of the vertical upward bend is independent of the inlet conditions upstream. The horizontal bend, unlike the vertical upward bend, has a minor influence on the flow evolution, particularly slug flow. This is due to the influence of gravitational force on phase separation and its subsequent effect on the change of momentum in the vertical bend. The flow structures, mainly slugs and disturbance waves, are slightly accelerated as they pass through the horizontal bend with minimum change to the structure frequency and gas holdup within liquid slugs. Most of the existing correlations do not predict the measured void fractions in this work as those correlations were essentially limited to the conditions they were developed for as the basis of them lies in the curve fitting. In this work, the higher viscosity and lower surface tension led to higher gas holdup in liquid slugs causing the discrepancy.

This work contained in this thesis is the result of an industrial and academic collaboration, designed to investigate and further the present knowledge of dense turbulent dispersions. Experiments were conducted to provide support and experimental validation to a CFD code being simultaneously developed, which was able to give insight into these types of flows. Additional to this support, the aim of this thesis was to also further knowledge of key topics in this field. The experimental methodology chosen was to use a mixture of Particle Image Velocimetry and Particle Tracking Velocimetry. To discriminate between particle and liquid phases, two approaches were adopted, depending upon the experiment. In one approach, fluorescent dyes were used to tag one phase, whilst optical filters were applied to the camera lenses. In the second approach, a size-based binary mask was applied to a single image, in order to remove phase information and produce two sets of images. A number of different analysis techniques were researched and developed as part of this thesis. The performance of particle tracking algorithms was assessed to ascertain their most suitable usage. A number of different algorithms, designed to characterise particle positions, were validated against known test cases. These included the Box Counting Method, a Voronoi analysis, and Radial Distribution Functions. A further technique, known as the Particle Potential method, was also developed to characterise local clustering. Two experiments were undertaken throughout this project, both of which were developed from scratch so that full control was assured over all experimental parameters. A vertical channel experiment was designed to assess the injections of particles into a rectangular channel. These experiments allowed for an ideal test case of highly concentrated particles, without the need to achieve optical visibility through a dense solution. The experiments also provided an early test of a Refractive Index Matching candidate pair; hydrogel particles and water. The second experiment was known as the Circulating Dispersion Rig, which was designed to pump a slurry in a continuous loop in a cylindrical pipe. These experiments, due to the geometry used and dense nature of the slurry, were reliant upon trying to achieve optimum optical visibility, and so hydrogel/water mixtures were tested in advance against other, more well-utilised pairings. The experiments conducted have provided some insight into the nature of particles in turbulent flows, in particular their clustering properties. Clustering was assessed under various concentrations. Key results included analysis of these clusters using a Voronoi diagram technique, which identified four key types of cluster structure, and the parameters under which these form. Collision probabilities of particle pairs were also assessed, using Particle Tracking data and computation of relative velocities. Such information is of importance for experimental validation of CFD codes relating to dispersed two-phase flows, where particle-particle coupling must be assessed in order to provide accurate solutions. The key drive towards the future, should further experiments be desirable, would be to investigate the improvement of optically matching liquids and solids, which was felt to be the limiting factor towards achieving measurements at even higher concentrations. However, these experiments show some progress can be made in making measurements of four-way coupled turbulent flows.

The influence of geometrical parameters on the development of intermittent flow is studied in this thesis. The geometrical parameters considered are the diameter of the pipe, the angle of inclination of the pipe, and the distribution of the area of the gas injection. Intermittent flow in gas-liquid two-phase flows occurs when, from a fixed point, a gas dominated structure followed by a liquid dominated structure seems to repeat at a certain mean frequency. It is mainly slug flow but churn and cap bubble flow also fall into this broad category. Intermittent gas-liquid two-phase flow was investigated in a 67 mm diameter, 6 m long rig and also in a 127 mm diameter, 12 m long rig. The test section of the 67 mm rig was mounted in a steel frame supported by a pivot that allowed changing the inclination of the pipe from vertical to horizontal in steps of 15°. The 127 mm rig can only be operated in the upwards vertical position. The fluids utilised were air and silicon oil of viscosity = 5 cP and density = 0.912 kg/m3. The interfacial surface tension was measured at 0.02 N/m. The facilities were both operated at atmospheric pressure. The gas superficial velocity (Ugs) was varied from 0.17 to 2.9 m/s and liquid superficial velocity (Uls) from 0.023 to 0.47 m/s. The void fraction generated by each set of conditions was captured for 60 seconds using a Wire Mesh Sensor and a twin plane Electrical Capacitance Tomography probe. The effect of the diameter and the angle of inclination of the pipe under different gas and liquid superficial velocities was reported. The main findings can be summarised as that the velocity of the periodic structures was found to be higher in large diameter pipes and increases with increasing the angle of inclination reaching a maximum around 50° then decreases. In addition, the frequency of the gas structures was found to be higher in small diameter pipes and increases with increasing the inclination of the pipe for all the gas and liquid superficial velocities investigated. Additionally, two correlations to predict the velocity and the frequency of the periodic gas structures as a function of the diameter, the inclination of the pipe, the gas superficial velocity and the liquid superficial velocity were developed. The proposed correlations were found to not only be in excellent agreement with the present experimental results (less than 20% difference), but also in good agreement with data published by other researchers. This include data produced using different fluids, different diameters of pipe and different gas and liquid superficial velocities to the ones investigated in this work. It was also found that the gas injection area, modified using different gas-liquid mixers, do not have an influence on the development of the intermittent two-phase flows at 75 diameters axial length from the mixing point.

Annular flow is one of the most common two-phase flow regimes observed in industrial applications. In annular flow, the liquid flows partly as a thin film along the pipe wall and partly as droplets entrained in the turbulent gas core. Most of the previous studies about the characteristics of annular flow and the developed correlations were conducted using an air/water system. This thesis reports an investigation about the characteristics of the annular flow regime and a development of liquid film thickness measurement using an ultrasonic technique in air/water and air/silicone oil systems. Experiments were carried on an upward vertical annular flow test facility with 34.5 mm inner diameter (ID) using air/water and air/silicone oil two-phase systems. Time-varying of total pressure drop, liquid film thickness and wall shear stress were measured. The total pressure drop was measured using a remote seal differential pressure transducer and the wall shear stress was measured using a glue-on hot film sensor. An ultrasonic technique was developed to measure the liquid film thickness. It was evaluated using static and dynamic measurements. For static measurements, it was compared with the liquid film thickness calculated based on knowledge of liquid volume and area of the test rig. For dynamic measurements, it was compared with two well-known conductance measurement techniques (Multi Pin Film Sensor and concentric probe) in falling film and upward vertical annular flow test facilities respectively. The relative error between the ultrasonic technique and the other two techniques was within ±5%. A new processing method for ultrasonic measurement called Baseline removal method was developed for measuring liquid film thickness less than 0.5 mm. The influence of gas and liquid superficial velocities, viscosity and surface tension on the measured parameters was studied using both systems. Both systems showed similar trend behavior with increasing gas and liquid superficial velocities even there was a difference in fluid properties. The results were also compared with the existing correlations developed using an air/water system to predict each one of the measured parameters. Most of the tested correlations predicted the total pressure drop, liquid film thickness and wall shear stress with relative deviation of ±50% or even higher in some cases.

The interaction of a boundary with turbulence is a defining feature of many turbulent flows, resulting in a turbulent boundary layer which plays a prominent role in the production and dissipation of turbulence. Commonly, this interaction is dominated by the effects of mean shear. However, more subtle aspects of the interaction, such as effects associated with turbulent motions impinging onto the boundary, are still thought to play a key role in giving rise to the boundary layer structure. Unfortunately, these aspects of the interaction are currently poorly understood. A better understanding of these aspects of the interaction may be derived by isolating them from the effects of mean shear through the study of zero-mean-shear turbulence interacting with a boundary. This study reports experimental work investigating the interaction between oscillating-grid turbulence (OGT) and a solid impermeable boundary. OGT is a commonly used experimental tool that produces a turbulent flow which is approximately homogeneous and isotropic in planes parallel to the oscillating grid but which is inhomogeneous in planes perpendicular to the oscillating grid. Throughout this study, instantaneous velocity measurements of the flow are obtained by applying two-dimensional particle imaging velocimetry to the vertical plane through the centre of the oscillating grid. A detailed preliminary study to characterise the flow generated by the OGT apparatus is initially performed. Visualisation of the flow close to the oscillating grid indicates that large-scale circulations are induced in OGT by the merging of grid-induced jets close to the tank walls. The installation of an open-ended cuboidal `inner box' below the grid is shown to inhibit the merging of these jets, thereby resulting in a more regular jet structure close to the oscillating grid and a corresponding reduction in mean flow within the inner box. It is also found that, contrary to assumptions in the literature, this amendment to the standard OGT apparatus is most effective when the top of the inner box is located close to the oscillating grid. The reduction in mean flow intensity that results from the use of a correctly installed inner box brings about a turbulent flow in which the mean flow velocity components are small compared to velocity fluctuations, thereby enabling a meaningful comparison to be made with zero-mean-shear turbulence. Consequently, the interaction between OGT and a solid impermeable boundary is studied to derive insight into the mechanisms governing the interaction of zero-mean-shear turbulence with boundaries. Results indicate that a critical aspect of the interaction is the blocking of a boundary-normal flux of turbulent kinetic energy across the boundary-affected region, which acts to increase the magnitude of the boundary-tangential turbulent velocity components, relative to the far-field trend, but not the boundary-normal turbulent velocity component. This feature arises as a result of the anisotropic nature of the flow produced by OGT, whereby the turbulent fluctuations decay with distance normal to and away from the oscillating grid, and would not be present in a turbulent flow that was otherwise homogeneous above the boundary-affected region of the flow. This observation provides new insight into the validity of well-established models of the interaction of zero-mean-shear turbulence and a solid impermeable boundary and provides a physical mechanism that explains the disparity in previously reported measurements relating to this problem. The results reported are also in support of intercomponent energy transfer mechanisms previously proposed to govern the interaction of zero-mean-shear turbulence with boundaries, including viscous and `return-to-isotropy' mechanisms. That is, within a thin region adjacent to the boundary, approximately equal in thickness to the viscous sublayer, the data indicate that turbulent motions incident towards the boundary are more energetic than motions away, which are characteristics of an intercomponent energy transfer primarily driven by the viscous dissipation of turbulent kinetic energy. In addition, at the edge of the boundary-affected region, where the magnitude of the boundary-tangential turbulent velocity components exceeds the magnitude of the boundary-normal turbulent velocity component, results indicate that an intercomponent energy transfer occurs from the boundary-tangential turbulent velocity components to the boundary-normal turbulent velocity component in a so-called `return-to-isotropy' energy transfer. However, the data also indicate the presence of an additional intercomponent energy transfer, from the boundary-normal turbulent velocity component to the boundary-tangential turbulent velocity components over a thin region outside the viscous sublayer. Comparison to previously published results of related studies indicates that this mechanism is also prevalent in previous work, but is not captured within existing models of intercomponent energy transfer at the boundary. Results further indicate that the intercomponent energy transfer mechanisms are not independent of the blocking of the boundary-normal turbulent kinetic energy flux. That is, the blocking of the boundary-normal turbulent kinetic flux promotes anisotropy within the boundary-affected region of the flow and thereby induces a stronger `return-to-isotropy' energy transfer mechanism. Hence, the effect of a solid impermeable boundary on turbulent velocity components in zero-mean-shear turbulence depends critically on the nature of the original turbulent field (i.e. homogeneous or inhomogeneous turbulence).

Oil and water separation processes in primary separators and transportation of these fluids through pipelines for further processing, is very vital but has often proved problematic due to changes in composition of the fluids together with build-up of sand or asphaltenes, in many petroleum industries. These separator vessels are large and cost effective to install together with safety implications due to equipment failure. Hence an understanding of the two-phase flow dynamics and motivation to improve upon their design and performance is necessary. Therefore, the main aim of this research programme is to investigate the coalescence efficiencies of droplet size in both stirred tank and horizontal pipe line under turbulence in oil-water two phase systems and also to determine the possibility of using a compact sudden pipe expansion as a phase separator for converting dispersed flow to segregated flow. For the purpose of experimental investigations, a sudden pipe expansion rig was designed, constructed and commissioned in L3 main laboratory of the department of Chemical and Environmental Engineering, University of Nottingham, UK. The fluids used are tap water ( ; ) and silicone oil 5cp, ( ; ) at operating conditions for mixture velocities between 0.20 ms-1 – 3.50 ms-1 and input oil volume fraction from 20% OVF to 80% OVF at different pipe inclinations (+40, +20, 00, -20,-40). Ring conductance probes were used to obtain phase layer distribution information of the oil-water flow together with the backscatter Lasentec FBRM M500P laser for drop size measurements and Phantom PCC 2.7 high speed camera for imaging. In both horizontal and upward inclined flows, the input oil volume fraction and mixture velocity strongly influenced the formation of drop sizes downstream of the expansion, due to coalescence mechanism. Coalescence of both oil and water dispersed phase droplets and segregation was found to start at short distance from the inlet section and progresses gradually downstream of the pipe expansion. The water droplets coalesce and settles faster than dispersed oil droplets in water continuous phase flow system and also the bulk oil layer was observed to flow faster at the top than the bottom water layer for the horizontal and upward flow conditions. Therefore the results reveals a strong influence of mixture velocity, input oil volume fraction and pipe angle inclination on drop break up and coalesce mechanisms in oil-water flow systems. The results obtained satisfactorily agree well with the existing conventional flow patterns and flow pattern maps. The flow patterns identified includes, stratified flow (ST), stratified with mixed interface (ST& MI), oil-water intermittent flow, stratified wavy flow and dispersed flow in all pipe orientations. Additional flow patterns were also observed such as Raleigh Taylor ‘plume shaped’ flow in the horizontal flow and plug wavy flow (Caterpillar like waves) in both upward and downward inclination, which have not been reported in large diameter pipes. There was evidence of stratification for fully developed ST regime at short distance from inlet (10D) at mixture velocities below 0.30ms-1 in both horizontal and upward inclinations. Therefore, in order to achieve stratified flow in oil-water two-phase flow system, a sudden pipe expansion was used successfully to convert a dispersed flow to segregated flow within a 6m long test pipe distance. The best orientation was that of the horizontal flow position for dispersed flows to separate quickly as both the upward and downward inclined pipe flow conditions tend to hamper the evolution process by increasing the mixed layer region.

Simultaneous flow of gas-liquid in pipes presents considerable challenges and difficulties due to the complexity of the two-flow mixture. Oil-gas industries need to handle highly viscous liquids, hence studying the effect of changing the fluid viscosity becomes imperative as this is typically encountered in deeper offshore exploration. This work looks at the effect of liquid viscosity on gas-liquid flows. The work was carried out using two different pipes of 67mm and 127mm internal diameter. For the experiments carried out on the 67mm diameter pipe, air and three different liquids were used with viscosities 1, 42 and 152cp. With these experiments, the effect of viscosity on the entrainment process from the Taylor bubble in a vertical tube was investigated with the Taylor bubble being held stationary in a downward liquid flow with the use of three different gas injection methods. Taylor bubble length, the gas flow rate and the liquid flow rate approaching the stationary bubble were varied. In addition, the wake length below the stationary bubble was measured at different conditions of gas and liquid superficial velocities and comparison was made with the work by previous authors. Videos were taken with high speed camera to validate the measurement taken on wake lengths. A Wire Mesh Sensor system was placed at two different positions below the air injection point on the 67mm diameter pipe of the stationary bubble facility whose data acquisition provided time and cross-sectionally resolved information about spatial distribution. This information was used to generate time averaged void fraction, bubble size distribution and contour plots of the two-phase flow structure. A Probability Density Function (PDF) of void fraction can be obtained from the former, with PDFs of the wake section of the stationary bubbles showing that the flows are in the bubbly region while the PDF for the entire slug unit assumed that for a typical twin-peaked slug flow. The interpretation of this is that holding a bubble stationary can simulate real slug flow. Results on the bubble length measurement and gas loss into a bubble wake have shown good agreement with existing work by other authors. Experiments on the 127 mm diameter pipe were carried out because most published work on gas/liquid flow were on smaller diameter pipes with air and water, yet many of the industrial applications of such flows in vertical pipes are in larger diameter pipes and with liquids which are much more viscous than water. Another important parameter considered in the study is pressure because of its effect on gas density. This part of the research goes some way to rectify this lack and presents void fraction and pressure gradient data for sulphur hexafluoride with gas densities of 28 and 45 kg/m3 and oil (viscosity 35 times water). The gas and liquid superficial velocities were varied in the ranges 0.1-3 and 0.1-1 m/s respectively. The void fraction was also measured with a Wire Mesh Sensor system. Flow patterns were identified from the signatures of the Probability Density Function of cross-sectionally averaged void fraction. These showed the single peak shapes associated with bubbly and churn flow but not the twin-peaked shape usually seen in slug flow. This confirms previous work in larger diameter pipes but with less viscous liquids. For the bubble to churn flows investigated, the pressure gradients decreased with increasing superficial gas velocity. The change in pressure ultimately affects the density of gas in the two-phase flow mixture. Though there was little effect of pressure on void fraction below certain transitional flow rates, the effect became significant beyond these values. Different statistical analysis techniques such as power spectral density, probability density function, mean, standard deviation and time series of the acquired data have been used which also show the significant effect of pressure on void fraction at high gas density which have not been measured previously.

The transport and separation of oil and water is an essential process to the oil and chemical industries. Although transporting the mixtures is often necessary due to few reasons, it is generally beneficial to separate out the phases in order to reduce installation and maintenance costs, at the same time, avoiding safety problems. Thus, separation of liquid-liquid flows is a necessary part of many industrial processes. Hence, knowledge of two-phase flow dynamics is important for the design optimisation of separators. Therefore, the aim of this research is to investigate the feasibility of a sudden pipe expansion to be used as phase separator because it compact in design and capable for converting dispersed flow to stratified flow. In the test section, spatial distribution of the liquid-liquid phases in a dynamics flow system was visualised for the first time for by means of capacitance Wire Mesh Sensor (CapWMS), providing instantaneous information about the interface shapes, waves and phase layer evolution of oil-water flow. Visual assessment and analysis of the WMS data showed three distinct layers: an oil layer at the pipe top; a water layer at the pipe bottom and a mixed layer between them. The interfaces that form between the separated phases (oil or water) and the mixed layer were classified as oil interface or water interface. Results showed interface shapes were initially concave or convex near to the inlet of the test section and became flat further downstream the expansion, especially for water interfaces. There were no waves observed for horizontal and downward pipe orientations at all flow conditions and axial position downstream of the expansion. As for the upward inclined pipe orientation, waves were found, and they formed at position close to the inlet at all input oil volume fraction except at 0.2 OVF. The amplitude of the waves was: ~ 0.29D for 0.8 OVF; ~ 0.22D for 0.6 OVF and ~ 0.26D for 0.4 OVF. The higher the input oil volume fraction, the larger the waves become. In conclusion, the WMS results demonstrated that spatial distributions are strongly dependent on the mixture velocity, input oil fraction and inclination angles for the far position. In this present work, droplets were found to be larger near the interface. Drops were large nearer to the interface at the near position (10D) for all pipe orientations and throughout the test section for horizontal flow. The drops size decreased when the distance from the interface increased for these pipe configurations. As for the furthest position from the expansion for upward and downward inclined pipe orientation, larger droplets could also be seen at distance away from the interface and vice versa. The gravity or buoyant force is one of the contributing factors to the settling of the droplets. These forces are acting simultaneously on the droplets i.e. if the buoyant force which tends to spread the droplets throughout the pipe cross-section, is not large enough to overcome the settling tendency of gravity settling of the droplets occurs. Hence, the droplets that are non-uniformly scattered within the continuous phase begin to coalesce as they flow further downstream the pipe, producing larger drops. In addition, as the distance from expansion increased, the mixed layer becomes narrow and more drops begin to coalescence to form large drop due to increased droplet-droplet collision. Owing to these factors, results indicate that the mechanisms of coalescence occurred faster at the bottom, for water droplets and at the top, for oil droplets than the other locations in a pipe cross-section. For a better separation design, the coalescence process should occur at the aforementioned (bottom for water and top for oil) locations within the expansion pipe. However, at higher mixture velocities the mixed layer would be responsible for the smaller droplet size for horizontal and both inclinations of pipe orientation. The mixed layer dominated almost entirely in the pipe cross-section.

Lorentz-forcing spanwise travelling wave actuation in the turbulent boundary layer has been studied in a water channel at various experimental conditions (St = 139.2, 186 and 232; T+ = 17, 42 and 82). At the Reynolds number of Reτ = 388, a maximum skin friction drag reduction of 30% is achieved in some cases, while up to 22.8% of viscous drag increase is also observed. The results of the turbulent boundary layer profiles show that the turbulence intensities for both the drag-reducing and the drag-increasing cases are reduced. The higher moments of turbulence statistics such as the skewness and the kurtosis increase near the wall when T+ = 42, St = 232 in the drag-reducing case. For the drag-increasing case (T+ = 17, St = 232), the skewness and the kurtosis are decreased when very close to the wall (y+ < 6), while they are increased for y+ > 6, similar to the drag-reducing case. The reduction in the turbulent intensities as well as the changes in VITA velocity profiles suggest that the drag changes are due to the modified near-wall activities by the Lorentz forcing. Flow visualisation shows that the low-speed streaks are twisted into the spanwise directions in both the drag-reducing and the drag-increasing cases. For the drag-reducing case, the low-speed streaks are clustered together to form a wide low-speed region similar to what Du et al (2002) have found. This low-speed region seems to act as the ‘storage’ of low-speed fluid to help reduce the skin friction drag. To achieve the drag reduction, the spanwise displacement of low-speed streaks must be greater than 115 wall units in the present configuration, which compares well with the average spacing of low-speed streaks in the turbulent boundary layer. When the drag increase occurs, only pseudo-local spanwise oscillation is observed without a formation of a wide low-speed region. The pseudo-local spanwise oscillation appears to produce converging and diverging motions around the forcing-activation area. The induced streamwise vorticity layers are believed to enhance the effect of the sweep motion, which results in the increasing skin-friction drag.