Search

Convergent-divergent (C-D) riblets are a type of bio-inspired surface pattern, and have begun to receive research attention in recent years, due to their potential in skin friction reduction and flow separation control. In this thesis, the effect of C-D riblets on the secondary flow, flow separation and drag characteristics in laminar wall-bounded flows is studied via numerical simulations. Firstly, a systematic investigation of the effect of riblet height, wavelength and yaw angle on the secondary flow in a laminar boundary layer developing over a C-D riblet section is undertaken. Large scale secondary flow is observed in cross-stream planes which displays downward/upward motions over the diverging/converging lines. The exact structure of the secondary flow depends on the relative size of riblet height and wavelength to the local boundary layer thickness, and three different patterns are observed. With the increase of wavelength, the average strength of the secondary flow per unit area exhibits a peak around a ratio between wavelength and local boundary layer thickness of 1. As the yaw angle increases, the strength of the secondary flow reaches to the peak value at a yaw angle of 45deg. Secondly, the effects of C-D riblets on momentum transfer enhancement and the extent of flow separation zone are examined by applying a section of C-D riblets upstream of a backward-facing rounded ramp in a laminar channel flow. In comparison with the baseline case, flow separation is delayed and the reattachment occurs earlier, leading to a smaller separation zone around the diverging line. The opposite phenomena occur around the converging line. A minimum riblet height of 3.75% of the channel height is required to produce a net reduction in the separation zone. As riblet spacing s increases with fixed riblet height h, a maximum strength of the secondary flow and a maximum reduction in the separation zone are obtained at s/h=4. Thirdly, the effect of C-D riblets on drag characteristics is studied by proposing an exact expression for the drag coefficient in laminar channel flows with wall roughness, whereby the drag is decomposed into contributions from different components of the velocity gradient tensor in the flow field. Furthermore, the triple decomposition technique is used to identify the contribution to drag production from the mean velocity field, the riblet- and wavelength-scale dispersive flow field. The normalized drag increment starts to rise when the Reynolds number is large enough to enable the secondary flow to alter the streamwise velocity across the span. While the normalized drag increment is predominantly caused by the mean and small-scale dispersive velocity at low Reynolds number, the contribution from the large-scale dispersive velocity field increases rapidly with the Reynolds number and gradually becomes dominant. Among C-D riblets with rectangular, triangular and sinusoidal cross-sectional shapes, the triangular riblet pattern is found to produce a secondary motion with a similar strength with less drag penalty. Finally, a theoretical derivation is presented to prove that drag reduction cannot be achieved by applying wall roughness structures onto the smooth inner walls of streamwise-periodic steady incompressible laminar channel/pipe flows at the same volume flow rate. It is shown that wall roughness produces a higher drag due to two factors: a) wall roughness induces other non-zero velocity gradient terms in addition to the wall-normal/radial gradient of streamwise velocity that exist in a smooth channel/pipe flow; b) the profile of streamwise velocity in the wall-normal/radial direction deviates from the parabolic profile that produces the minimum kinetic energy loss at the same volume flow rate.

In this thesis, both vortical structures and the secondary flow in boundary layers over convergent-divergent riblets (C-D riblets) are experimentally studied. The development of the laminar boundary layer over C-D riblets is studied using dye visualisation and mono-/stereoscopic particle image velocimetry (PIV). C-D riblets are observed to generate a spanwise flow from the diverging line towards the adjacent converging line, leading to a weak recirculating secondary flow in cross-stream planes across the boundary layer which creates a downwelling over the diverging region and an upwelling over the converging region. The fluid inside riblet valleys follows a helicoidal path and it also interacts with the crossflow boundary layer. The boundary layer development over the riblet section is divided into a developing stage followed by a developed stage. With a decreased riblet height at the converging line and a linear spanwise height variation, the intensity of the induced secondary flow over the converging region is significantly reduced, while the flow field characteristics over the diverging region are basically preserved. The turbulent boundary layers developing over C-D riblets with three different heights of h+=8, 14 and 20 are studied in the longitudinal plane and the cross-stream plane. Although a logarithmic region is observed in the velocity profiles, Townsend's outer-layer similarity hypothesis is not valid. The coherent structures over C-D riblets are revealed in three perspectives, including spanwise vortices, vortex packets and uniform momentum zones, which help to obtain new insights into the vortical activities at different scales. While an increased riblet height affects the entire turbulent boundary layer over the converging region, the impact on the diverging region is largely confined within the near-wall region. In the cross-stream plane, a riblet height increase results in a wider downwelling region, a stronger spanwise flow, a narrower upwelling region and a stronger decelerating effect. Overall, the higher C-D riblets generate a more intense secondary flow, and the mechanism of an increasing riblet height is attributed to the greater capability of deeper yawed microgrooves.

In modern civil aeroengines the fan is driven by the low pressure turbine. In order to match the low rotational speed of the fan, the low pressure turbine consists of large diameter components and several stages, up to six or seven. An individual stage of a low pressure turbine consists of a row of stationary blades (stators) followed by a row of rotating blades (rotors). This arrangement necessitates radial clearance gaps above the tips of the rotor blades and below the hub end of the stator blades. Therefore, a fraction of the gas does not pass through the blade rows, but instead leaks through the tip and hub cavities into the downstream blade row. The leakage flow over shrouded turbine rotor blades is driven by the pressure drop across the blade row. The leakage flow experiences very little change in tangential velocity as it moves through the rotor shroud cavity, hence, when the leakage flow re-enters the mainstream downstream of the rotor trailing edge, it has higher tangential velocity than that of the mainstream flow. The aim of this research was to improve the understanding of the effects of the rotor shroud leakage flow on the aerodynamics of the downstream stator. Based on the improved understanding, a new design for the stator geometry was produced which has a performance that is more robust to increases in the rotor shroud leakage mass flow rate, so that the efficiency of the turbine is better maintained over the in-service lifetime of the aeroengine. The experimental measurements for the present study have been undertaken on the Peregrine Turbine Facility in the Whittle Laboratory. Measured contours of streamwise vorticity at the exit of the downstream stator identified two positive streamwise vortex structures. A computational study revealed that one vortex was caused by a large separation at the leading edge of the stator. It is demonstrated that this vortex can be eliminated by the appropriate redesign of the stator. The second vortex is caused by the roll up of the inlet streamwise vorticity sheet, which results from the radial gradient of tangential velocity between the rotor shroud leakage flow and the mainstream flow at the stator inlet. This vortex is a fundamental consequence of the rotor shroud leakage flow and cannot be eliminated within the stator passage.

An investigation of the flow characteristics and sediment transport processes has been carried out in a two-stage meandering channel. Three phases of experiments have been conducted with various floodplain roughnesses. The dimensions of the flume are 13m long and 2.4m wide with a fixed valley slope of 11500. The meandering main channel has a sinuosity of 1.384 with top width of 0.4m. In each phase of the experiment, hydraulic data pertaining to stage-discharge, bed topography and sediment transport rate were measured at various overbank flow depths. Several flow depths were chosen to measure the three-dimensional velocities by means of Laser Doppler Anemometer and the morphological bedforms were recorded using the Photogrammetric technique. The boundary shear stresses were also measured by means of a Preston Tube and Vane Indicator. The experimental results showed that the presence of the energy losses due to momentum exchange and turbulence, bedforms roughness and floodplain roughness induced additional flow resistance to the main channel flow, particularly for shallow overbank flows. The combination of these losses affected a significant reduction in velocity and boundary shear stress in the main channel which, subsequently led to the reduction of sediment discharge at low relative depth for most tested cases. The reduction was more pronounced when the floodplain roughness increased. The examination of the three-dimensional velocity indicated that the formation of bedforms in the main channel is significantly affected by the flow structures, especially the secondary flow. A new method for predicting velocity and sediment transport rate has been introduced based on the two-dimensional equation (Spooner's) coupled with the self-calibrated empirical transport formula. The proposed method gave accurate prediction for depthaveraged velocity and sediment transport rate for two-stage meandering channel.

Shock wave/boundary-layer interactions (SBLIs) are ubiquitous to both the external vehicle body and internal propulsion flowpath. The external surfaces include for example the nose, wings, and tail, while internal surfaces include bounding walls of mixed compression inlets, diffusers, and isolators. To obtain predictions of these flowfields, Reynolds-averaged Navier-Stokes (RANS) computational fluid dynamics (CFD) solvers are preferred over computationally expensive scale-resolving methods like large-eddy simulation (LES), hybrid RANS/LES, and direct numerical simulation (DNS). However, RANS predictions are highly dependent on the choice of turbulence model. Reasons include the fundamental assumption (Boussinesq eddy-viscosity approximation) made in the development of turbulence models and the estimation of the terms that govern turbulence transport.In this work, key mechanisms in the exact Reynolds stress transport equation are examined with a view towards identifying important flow phenomena and improving prediction of flows with SBLIs. Upon linking the unsteady dynamics back to the Reynolds stress transport, significant mechanisms can be isolated, whose relative importance varies across the SBLI region. By discovering the importance of these mechanisms to SBLIs, current models can be improved and mechanisms generally assumed negligible due to the lack of experimental or high-fidelity data can be modeled.With this purpose in mind, wall-resolved LES (WRLES) of SBLI are considered at several Reynolds numbers. The flow conditions and geometry are based on the experiments performed at the Institut Universitaire des Systemes Thermiques Industriel (IUSTI) in Marseille, France at Mach 2.29 for an 8 degree deflection angle shock wave impinging on a turbulent boundary layer developing on the opposite surface. Due to a closely coupled relationship between the corner flow and centerline separation in small- to medium-aspect-ratio configurations, the full tunnel span that includes the two sidewalls and produce a three-dimensional SBLI is studied.Verification of the numerical strategy is achieved by investigating the dependence of the results on mesh resolution, choice of domain size, numerical dissipation of the WRLES framework, and the digital filter inflow generation method. This is followed by detailed validation against the particle image velocimetry data from the experiment to assess the accuracy of the simulations. A comparison between the periodic and full-span simulations confirms that the corner physics is integral to the prediction of the centerline separation. The compression waves, visualized by plotting isosurface of the pressure gradient magnitude, are shown to emanate from the blockage created by the corner separation and are responsible for the amplification of the centerline separation. Furthermore, the phenomenology within the interaction region is invariably linked to the perturbation of the incoming turbulent boundary layer by the reflected-shock foot. It leads to the formation of a band in which the turbulence statistics become prominent and consistent with the development of the shear layer and Kelvin-Helmholtz shedding. The interaction between the corner separation and secondary flow of the second kind, i.e., a pair of counter-rotating vortices symmetric about the corner bisector, significantly diminishes the size of the vortex pair. However, as the corner separation becomes sidewall biased, the clockwise-rotating vortex grows, while the counter clockwise-rotating vortex remains anchored near the corner origin.Unsteady dynamics at the centerline and corner locations are evaluated by computing the weighted power spectral densities (PSDs) of pressure and velocity fluctuations. By placing the probes at key locations, unsteadiness effects related to the reflected-shock foot (low frequency, St ≈ 0.015), shear-layer development (intermediate frequency, St ≈ 0.1), and Kelvin-Helmholtz shedding (high frequency, St ≈ 0.5) are identified. The streamwise and spanwise recirculation dynamics of the centerline separation bubble are explored by employing low-pass filtered PSDs of u′ and w′, where broadband energetic scales in the intermediate-frequency range are found. The same intermediate-frequency scales are also present in the PSDs of the probes located within the corner separation. The corner-centerline two-point correlations illustrate the dynamical coupling of the corner regions with the reflected-shock foot that persists for multiple forward and aft shock oscillation cycles. The corner-corner two-point correlations reveal that the opposing corners are either positively correlated, uncorrelated, or negatively correlated and the period of this behavior cycle is significantly longer than the reflected-shock foot unsteadiness.A rigorous analysis of the Reynolds stress transport budget is performed to evaluate the significance of production, diffusion, transport, redistribution, and dissipation mechanisms in regions of SBLI. The budget sum, computed to serve as a measure of error in the budget, is also a good indicator of non-equilibrium in the flowfield caused by the unsteadiness. The pressure-diffusion mechanism is found to be important at the reflected-shock foot location and in its downstream vicinity. Notably, the analysis of turbulent-diffusion and pressure-strain mechanisms corroborates their key roles in balancing production within the shear-layer development and Kelvin-Helmholtz shedding regions. The molecular diffusion contributes in balancing dissipation near the wall, but is inconsequential in the unsteady regions. Upon considering the behavior of three diffusion mechanisms, the evidence indicates that separate modeling of these mechanisms would be beneficial in RANS simulations of SBLIs. Although the turbulent mass flux mechanism was hypothesized to be significant for the entrainment and ejection of mass from the separation bubble, it is in fact negligible and not a modeling concern. Finally, the spanwise variation in budgets due to the flow three-dimensionality is studied by comparing the centerline, quarter-span, and corner bisector positions. The leading mechanisms in the corner region are convection, production, turbulent diffusion, and pressure strain, while the behavior of quarter-span budgets is dependent upon the streamwise location. Ultimately, these results lay the foundation for a more systematic procedure to close the RANS equations in the presence of three-dimensionality, pressure gradients, and other sources of mechanical non-equilibrium.

Flow through the turbine section of gas turbine engines is inherently unsteady due to a variety of factors, such as the relative motion of rotors and stators. In low pressure turbines, periodic wake passing has been shown to impact boundary layer separation, blade surface pressure distribution, and loss generation. The effect of periodic disturbances on the endwall flow is less understood. Endwall flow in a low-pressure turbine occurs in the boundary layer region of the flow through the blade passage where the blade attaches to the hub in the turbine. The response of an endwall vortical structure, the passage vortex, to various upstream disturbances is considered in this investigation. The passage vortex is a three-dimensional unsteady flow feature which generates aerodynamic losses as it interacts with the flow along the blade suction surface. High-speed velocimetry and numerical simulations have shown that the vortex intermittently loses coherence and varies in strength and position over time. The intermittent loss of coherence of the passage vortex is believed to be related to the leading-edge junction flow dynamics. An array of pneumatic devices was installed upstream of a linear cascade of low-pressure turbine blades to produce periodic disturbances that impact the blade leading edge region. A small disturbance and a large disturbance were created and characterized by their maximum velocity deficit and nondimensionalized solenoid valve on time using a plane of particle image velocimetry. A plane of high-speed stereoscopic particle image velocimetry data was collected inside the blade passage to examine how the disturbances impacted the vortex. Surface-mounted hot-film data was collected near the leading edge and in passage region to help relate flow behavior in both locations. The size and frequency of the disturbances had a nonlinear impact on the vortex size and strength. Fourier analysis revealed that the actuation frequency caused a harmonic response, and a change in the temporal behavior of the passage vortex. Each actuation frequency caused a different response from the vortex, but the vortex dynamics did not lock-on to the disturbance frequency.

A primary source of periodic unsteadiness in low-pressure turbines is the wakes shed from upstream blade rows due to the relative motion between adjacent stators and rotors. These periodic perturbations can affect boundary layer transition, secondary flow, and loss generation. In particular, for high-lift front-loaded blades, the secondary flowfield is characterized by strong three-dimensional vortical structures. It is important to understand how these flow features respond to periodic disturbances. A novel approach was taken to generate periodic unsteadiness which captures some of the physics of turbomachinery wakes. Using stationary pneumatic devices, pulsed jets were used to generate disturbances characterized by velocity deficit, elevated turbulence, and spanwise vorticity. Prior to application in a turbine flow environment, the concept was explored in a small developmental wind tunnel using a single device. The disturbance flowfield for different input settings was measured using hot-film anemometry and Particle Image Velocimetry. Insight was also garnered on how to improve later design iterations. With an array of devices installed upstream of a linear cascade of high-lift front-loaded turbine blades, settings were found which produced similar disturbances at varying frequencies that periodically impinged upon the leading-edge region. These settings were used to conduct an in-passage secondary flow study using high-speed Stereoscopic Particle Image Velocimetry. Results demonstrated the application of the periodic unsteadiness generator but found minor changes to the passage vortex. The vortex rotational strength decreased, and migration increased with increased perturbation frequency. Fourier analyses found the PV to be responsive at the actuation frequency with phase-locked ensemble-averaged data revealing that the disturbance periodically caused the PV to lose rotational strength. However, at the tested discrete frequencies, the vortex did not become locked-in and was defined by an erratic time-character similar to without the disturbances.

As a contribution to advancing turbofan engine ground test technology in support of propulsion system integration in modern conceptual aircraft, a novel inlet distortion generator (ScreenVaneTM) was invented. The device simultaneously reproduces combined inlet total pressure and swirl distortion elements in a tailored profile intended to match a defined turbofan engine inlet distortion profile. The device design methodology was intended to be sufficiently generic to be utilized in support of any arbitrary inlet distortion profile yet adequately specific to generate high-fidelity inlet distortion profile simulation. For the current investigation, a specific inlet distortion profile was defined using computational analysis of a conceptual boundary layer ingesting S-duct turbofan engine inlet. The resulting inlet distortion profile, consisting of both total pressure and swirl distortion elements, was used as the objective profile to be matched by the ScreenVane in a turbofan engine ground test facility. A ScreenVane combined inlet total pressure and swirl distortion generator was designed, computationally analyzed, and experimentally validated. The design process involved specifying a total pressure loss screen pattern and organizing a unique arrangement of swirl inducing turning vanes. Computational results indicated that the ScreenVane manufactured distortion profile matched the predicted S-duct turbofan engine inlet manufactured distortion profile with excellent agreement in pattern shape, extent, and intensity. Computational full-field total pressure recovery and swirl angle profiles matched within approximately 1% and 2.5° (RMSD), respectively. Experimental turbofan engine ground test results indicated that the ScreenVane manufactured distortion profile matched the predicted S-duct turbofan engine inlet manufactured distortion profile with excellent agreement in pattern shape, extent, and intensity. Experimental full-field total pressure recovery and swirl angle profiles matched within approximately 1.25% and 3.0° (RMSD), respectively. Following the successful reproduction of the S-duct turbofan engine inlet manufactured distortion profile, a turbofan engine response evaluation was conducted using the validated ScreenVane inlet distortion generator. Flow measurements collected at discrete planes immediately upstream and downstream of the fan rotor isolated the component for performance analysis. Based on the results of this particular engine and distortion investigation, the adiabatic fan efficiency was negligibly altered while operating with distorted inflow conditions when compared to nominal inflow conditions. Fuel flow measurements indicated that turbofan engine inlet air mass flow specific fuel consumption increased by approximately 5% in the presence of distortion. While a single, specific turbofan engine inlet distortion profile was studied in this investigation, the ScreenVane methodology, design practices, analysis approaches, manufacturing techniques, and experimental procedures are applicable to any arbitrary, realistic combined inlet total pressure and swirl distortion.

This experiment investigated the effects of a non-axisymmetric endwall contour and upstream purge flow on the secondary flow of an inlet guide vane. Three cases were tested in a transonic wind tunnel with an exit Mach number of 0.93-a flat endwall with no upstream purge flow, the same flat endwall with upstream purge flow, and a 3D contoured endwall with upstream purge flow. All cases had a backward-facing step upstream of the vanes. Five-hole probe measurements were taken 0.2, 0.4, and 0.6 Cx downstream of the vane row trailing edge, and were used to calculate loss coefficient, secondary velocity, and secondary kinetic energy. Additionally, surface static pressure measurements were taken to determine the vane loading at 4% spanwise position. Surface oil flow visualizations were performed to analyze the flow qualitatively. No statistically significant differences were found between the three cases in mass averaged downstream measurements. The contoured endwall redistributed losses, rather than making an improvement distinguishable beyond experimental uncertainty. Flow visualization found that the passage vortex penetrated further in the spanwise direction into the passage for the contoured endwall (compared to the flat endwall), and stayed closer to the endwall with a blowing ratio of 1.5 with a flat endwall (compared to no blowing with flat endwall). This was corroborated by the five hole probe results.

In many regions of the high-pressure gas turbine, film cooling flows are used to protect the turbine components from the combustor exit hot gases. Endwalls are challenging to cool because of the complex system of secondary flows that disturb surface film coolant coverage. The secondary flow vortices wash the film coolant from the surface into the mainstream significantly decreasing cooling effectiveness. In addition to being effected by secondary flow structures, film cooling flow can also affect these structures by virtue of their momentum exchange. In addition, many studies in the literature have shown that endwall contouring affects the strength of passage secondary flows. Therefore, to develop better endwall cooling schemes, a good understanding of passage aerodynamics and heat transfer as affected by interactions of film cooling flows with secondary flows is required. This experimental and computational study presents results from a linear, stationary, two-passage cascade representing the first stage nozzle guide vane of a high-pressure gas turbine with an axisymmetrically contoured endwall. The sources of film cooling flows are upstream combustor liner coolant and endwall slot film coolant injected immediately upstream of the cascade passage inlet. The operating conditions simulate combustor exit flow features, with a high Reynolds number of 390,000 and approach flow turbulence intensity of 11% with an integral length scale of 21% of the chord length. Measurements are performed with varying slot film cooling mass flow to mainstream flow rate ratios (MFR). Aerodynamic effects are documented with five-hole probe measurements at the exit plane. Heat transfer is documented through recovery temperature measurements with a thermocouple. General secondary flow features are observed. Total pressure loss measurements show that varying the slot film cooling MFR has some effects on passage loss. Velocity vectors and vorticity distributions show a very thin, yet intense, cross-pitch flow on the contoured endwall side. Endwall adiabatic effectiveness values and coolant distribution thermal fields show minimal effects of varying slot film coolant MFR. This suggests the dominant effects of combustor liner coolant. show dominant effects of combustor liner coolant on cooling the endwall. A coolant vorticity correlation presenting the advective mixing of the coolant due to secondary flow vorticity at the exit plane is also discussed.

Next generation aircraft design focuses on highly integrated airframe/engine architectures that exploit advantages in system level efficiency and performance. One such design concept incorporates boundary layer ingestion which locates the turbofan engine inlet near enough to the lifting surface of the aircraft skin that the boundary layer is ingested and reenergized. This process reduces overall aircraft drag and associated required thrust, resulting in fuel savings and decreased emissions; however, boundary layer ingestion also creates unique challenges for the turbofan engines operating in less than optimal inlet flow conditions. The engine inlet flow profiles predicted from boundary layer ingesting aircraft architectures contain complex distortions that affect the engine operability, durability, efficiency, and performance. One component of these complex distortion profiles is off-axial secondary flow, commonly referred to as swirl. As a means to investigate the interactions of swirl distortion with turbofan engines, an experiment was designed to measure distorted flow profiles in an operating turbofan research engine. Three-dimensional flow properties were measured at discrete planes immediately upstream and immediately downstream of the fan rotor, isolating the component for analysis. Constant speed tests were conducted under clean and distorted test conditions. For clean tests, a straight cylindrical inlet duct was attached to the fan case; for distorted tests, a StreamVane swirl distortion generator was inserted into the inlet duct. The StreamVane was designed to induce a swirl distortion matching results of computation fluid dynamics models of a conceptual blended wing body aircraft at a plane upstream of the fan. The swirl distortion was then free to develop naturally within the inlet duct before being ingested by the engine. Results from the investigation revealed that the generated swirl profile developed, mixed, and dissipated in the inlet duct upstream of the fan. Measurements immediately upstream of the fan rotor leading edge revealed 50% reduction in measured flow angle magnitudes along with evidence of fanwise vortex convection when compared to the StreamVane design profile. The upstream measurements also indicated large amounts of secondary flow entered the fan rotor. Measurements immediately downstream of the fan rotor trailing edge demonstrated that the fan processed the distortion and further reduced the intensity of the swirl; however, non-uniform secondary flow persisted at this plane. The downstream measurements confirmed that off-design conditions entered the fan exit guide vanes, likely contributing to cascading performance deficiencies in downstream components and reducing the performance of the propulsor system.

The present study investigates mixed out aerodynamic loss coefficient measurements for a high turning, contoured endwall passage under transonic operating conditions in presence of upstream purge slot and mateface gap. The upstream purge slot represents the gap between stator-rotor interface and the mateface gap simulates the assembly feature between adjacent airfoils in an actual high pressure turbine stage. While the performance of the mateface and upstream slot has been studied for lower Mach number, no studies exist in literature for transonic flow conditions. Experiments were performed at the Virginia Tech's linear, transonic blow down cascade facility. Measurements were carried out at design conditions (isentropic exit Mach number of 0.87, design incidence) without and with coolant blowing. Upstream leakage flow of 1.0% coolant to mainstream mass flow ratio (MFR) was considered with the presence of mateface gap. There was no coolant blowing through the mateface gap itself. Cascade exit pressure measurements were carried out using a 5-hole probe traverse at a plane 1.0Cax downstream of the trailing edge for a planar geometry and two contoured endwalls. Spanwise measurements were performed to complete the entire 2D loss plane from endwall to midspan, which were used to plot pitchwise averaged losses for different span locations and loss contours for the passage. Results reveal significant reduction in aerodynamic losses using the contoured endwalls due to the modification of flow physics compared to a non contoured planar endwall.

A subsonic wind tunnel facility was designed and built to test and optimize various diffuser-collector box geometries at the one-twelfth scale.  The facility was designed to run continuously at an inlet Mach number of 0.42 and an inlet hydraulic diameter Reynolds number of 340,000. Different combinations of diffusers, hubs, and exhaust collector boxes were designed and evaluated for overall optimum performance. Both 3-hole and 5-hole probes were traversed into the flow to generate multiple diffuser inlet and collector exit performance profile plots. Surface oil flow visualization was performed to gain an understanding of the complex 3D flow structures inside the diffuser-collector subsystem. The cutback radial hardware was found to increase the subsystem pressure recovery by over 10% from baseline resulting in an approximate 1% increase in gas turbine power output.

A three-component laser Doppler anemometer was used to measure the fully developed, turbulent flow in a closed, symmetric, smooth-wall compound channel. Measurements were made across one quadrant of the cross-section since the flow was assumed symmetric. Measurements were made for a single channel Reynolds number. All mean velocity components were calculated and are reported. The mean velocity field results are in good agreement with results reported for similar geometries. The vector plots and the axial vorticity distribution reveal the existence of secondary flow cells in both the main channel and the flood plain. The maximum values of the secondary velocities are at the comer region, on the interface between the main channel and the flood plain. In this region the mean velocity gradients are large, indicating that this might be an area of high turbulence production. The distributions of all Reynolds stresses across the cross-section are reported. The Reynolds stress distributions show peak values near the interface corner region and small values near the center-line and on the axes of symn1etry of the channel. The turbulence kinetic energy distribution verifies the existence of high turbulence energy fluid in the comer region.