This document addresses problems encountered when establishing the very low air-flow speeds required for experimental investigations of the mechanisms of low-Reynolds-number boundary-layer turbulence. Small-scale motions in the near-wall region are important features of turbulent boundary-layer dynamics, and, if these features are to be resolved by measurements in air with conventionally-sized hot-wire probes, a well-behaved canonical turbulent boundary layer must be developed at free stream flow speeds no higher than 4 m/s. However, at such low speeds, the turbulent boundary layers developed on the walls of a wind tunnel are very susceptible to perturbation by non-turbulent time-dependent flow structures which originate upstream from the test section in the laminar flow at the inlet and in the contraction. Four different non-turbulent flow structures have been identified. The first is a result of quasi-two-dimensional separation of the laminar boundary-layer from the surfaces of the wind-tunnel contraction. Potential flow simulations show that susceptibility to this form of separation is reduced by increasing the degree of axisymmetry in the cross-section geometry and by decreasing the streamwise curvature of the concave surfaces. The second source of time-dependence in the laminar boundary-layer flow is an array of weak streamwise vortices produced by Goertler instability. The Goertler vortices can be removed by boundary-layer suction at the contraction exit. The third form of flow perturbation, revealed by visualisation experiments with streamers, is a weak large-scale forced-vortex swirl produced by random spatial fluctuations of temperature at the wind-tunnel inlet. This can be prevented by thorough mixing of the inlet flow; for example, a centrifugal blower installed at the inlet reduces the amplitude of temperature nonuniformity by a factor of about forty and so prevents buoyancy-driven swirl. When subjected to weak pressure gradients near the start of a wind-tunnel contraction, Goertler vortices in laminar wall layers can develop into three-dimensional separations with strong counter-rotating trailing vortices. These trailing vortices are the fourth source of unsteady flow in the test-section. They can be suppressed by a series of appropriately located screens which remove the low-speed-streak precursors of the three-dimensional separations. Elimination of the above four contaminating secondary flows permits the development of a steady uniform downstream flow and well-behaved turbulent wall layers. Measurements of velocity in the turbulent boundary layer of the test-section have been obtained by hot-wire anemometry. When a hot-wire probe is located within the viscous sublayer, heat transfer from the hot-wire filament to the wall produces significant errors in the measurements of both the mean and the fluctuating velocity components. This error is known as wall-proximity effect and two successful methods are developed for removing it from the hot-wire signal. The first method is based on the observation that, if all experimental parameters except flow speed and distance from the wall are fixed, the velocity error may be expressed nondimensionally as a function of only one parameter, in the form DeltaU^+=f(y^+). The second method, which also accommodates the effect of changing the hot-wire overheat ratio, is based on a dimensional analyis of heat transfer to the wall. Velocity measurements in the turbulent boundary layer at the mid-plane of a nearly square test-section duct have established that, when the boundary-layer thickness is less than one quarter of the duct height, mean-velocity characteristics are indistinguishable from those of a two-dimensional flat-plate boundary layer. In thicker mid-plane boundary layers, the mean-velocity characteristics are affected by stress-induced secondary flow and by lateral constriction of the boundary-layer wake region. A significant difference between flat-plate and duct boundary layers is also observed in momentum-balance calculations. The momentum-integral equation for a duct requires definitions of momentumd and displacement thickness which are different from those given for flat-plate boundary layers. Momentum-thickness growth rates predicted by the momentum-integral equation for a duct agree closely with measurements of the newly defined duct momentum thickness. Such agreement cannot be obtained in terms of standard flat-plate momentum thickness. In duct boundary layers with Reynolds numbers Re_theta between 400 and 2600, similarity in the wake-region distributions of streamwise turbulence statistics has been obtained by normalising distance from the wall with the flat-plate momentum thickness, theta_2. This result indicates that, in contrast with the mean velocity characteristics, the structure of mid-plane turbulence does not depend on the proportion of duct cross-section occupied by boundary layers and is essentially the same as in a flat-plate boundary layer. For Reynolds numbers less than 400, both wall-region and wake-region similarity fail because near-wall turbulence events interact strongly with the free stream flow and because large scale turbulence motions are directly influenced by the wall. In these conditions, which exist in both duct and flat-plate turbulent boundary layers, there is no distinct near-wall or wake region, and the behaviour of turbulence throughout the boundary layer depends on both wall variables and on outer region variables simultaneously. / Thesis (Ph.D.)--School of Mechanical Engineering, 1998.

This document addresses problems encountered when establishing the very low air-flow speeds required for experimental investigations of the mechanisms of low-Reynolds-number boundary-layer turbulence. Small-scale motions in the near-wall region are important features of turbulent boundary-layer dynamics, and, if these features are to be resolved by measurements in air with conventionally-sized hot-wire probes, a well-behaved canonical turbulent boundary layer must be developed at free stream flow speeds no higher than 4 m/s. However, at such low speeds, the turbulent boundary layers developed on the walls of a wind tunnel are very susceptible to perturbation by non-turbulent time-dependent flow structures which originate upstream from the test section in the laminar flow at the inlet and in the contraction. Four different non-turbulent flow structures have been identified. The first is a result of quasi-two-dimensional separation of the laminar boundary-layer from the surfaces of the wind-tunnel contraction. Potential flow simulations show that susceptibility to this form of separation is reduced by increasing the degree of axisymmetry in the cross-section geometry and by decreasing the streamwise curvature of the concave surfaces. The second source of time-dependence in the laminar boundary-layer flow is an array of weak streamwise vortices produced by Goertler instability. The Goertler vortices can be removed by boundary-layer suction at the contraction exit. The third form of flow perturbation, revealed by visualisation experiments with streamers, is a weak large-scale forced-vortex swirl produced by random spatial fluctuations of temperature at the wind-tunnel inlet. This can be prevented by thorough mixing of the inlet flow; for example, a centrifugal blower installed at the inlet reduces the amplitude of temperature nonuniformity by a factor of about forty and so prevents buoyancy-driven swirl. When subjected to weak pressure gradients near the start of a wind-tunnel contraction, Goertler vortices in laminar wall layers can develop into three-dimensional separations with strong counter-rotating trailing vortices. These trailing vortices are the fourth source of unsteady flow in the test-section. They can be suppressed by a series of appropriately located screens which remove the low-speed-streak precursors of the three-dimensional separations. Elimination of the above four contaminating secondary flows permits the development of a steady uniform downstream flow and well-behaved turbulent wall layers. Measurements of velocity in the turbulent boundary layer of the test-section have been obtained by hot-wire anemometry. When a hot-wire probe is located within the viscous sublayer, heat transfer from the hot-wire filament to the wall produces significant errors in the measurements of both the mean and the fluctuating velocity components. This error is known as wall-proximity effect and two successful methods are developed for removing it from the hot-wire signal. The first method is based on the observation that, if all experimental parameters except flow speed and distance from the wall are fixed, the velocity error may be expressed nondimensionally as a function of only one parameter, in the form DeltaU^+=f(y^+). The second method, which also accommodates the effect of changing the hot-wire overheat ratio, is based on a dimensional analyis of heat transfer to the wall. Velocity measurements in the turbulent boundary layer at the mid-plane of a nearly square test-section duct have established that, when the boundary-layer thickness is less than one quarter of the duct height, mean-velocity characteristics are indistinguishable from those of a two-dimensional flat-plate boundary layer. In thicker mid-plane boundary layers, the mean-velocity characteristics are affected by stress-induced secondary flow and by lateral constriction of the boundary-layer wake region. A significant difference between flat-plate and duct boundary layers is also observed in momentum-balance calculations. The momentum-integral equation for a duct requires definitions of momentumd and displacement thickness which are different from those given for flat-plate boundary layers. Momentum-thickness growth rates predicted by the momentum-integral equation for a duct agree closely with measurements of the newly defined duct momentum thickness. Such agreement cannot be obtained in terms of standard flat-plate momentum thickness. In duct boundary layers with Reynolds numbers Re_theta between 400 and 2600, similarity in the wake-region distributions of streamwise turbulence statistics has been obtained by normalising distance from the wall with the flat-plate momentum thickness, theta_2. This result indicates that, in contrast with the mean velocity characteristics, the structure of mid-plane turbulence does not depend on the proportion of duct cross-section occupied by boundary layers and is essentially the same as in a flat-plate boundary layer. For Reynolds numbers less than 400, both wall-region and wake-region similarity fail because near-wall turbulence events interact strongly with the free stream flow and because large scale turbulence motions are directly influenced by the wall. In these conditions, which exist in both duct and flat-plate turbulent boundary layers, there is no distinct near-wall or wake region, and the behaviour of turbulence throughout the boundary layer depends on both wall variables and on outer region variables simultaneously. / Thesis (Ph.D.)--School of Mechanical Engineering, 1998.

This thesis project aims to advance the modelling of pressure gradient turbulent boundary layers (PG TBLs) and offer new insights into TBLs modelling. Previous analytical studies have explored various mathematical models, but this research introduces an extended unstacked Deep Operator Networks (DeepONets) architecture with double outputs and five branch parameters. The objective is to capture the mean velocity and Reynolds stress of turbulent boundary layers under pressure gradients. Numerical and experimental datasets of PG TBLs were accessed and utilized to train the DeepONets models. These models successfully predicted the mean velocity and Reynolds stress profiles using outer-scaled parameters. The DeepONets effectively learned the operator that describes the desired profiles based on input parameters, which correspond to the development of boundary layer thickness and pressure gradients. To identify the model with the best prediction performance, error statistics and distribution were examined across different configurations and dimensions. Furthermore, the individual and global sensitivity analyses revealed the relationship between input parameters and their influence on modelling PG TBLs with DeepONets.

Turbulent boundary layers flowing over a smooth surface were studied to understand the influence of varying pressure gradients and flow Reynolds number on the boundary layer growth and mean turbulent properties. The test was conducted in the Virginia Tech Stability Wind Tunnel with a 0.914 m chord length, NACA 0012 Airfoil in the test section. This airfoil was rotated to different angles of attack to induce varying pressure gradients on the boundary layer developing on the test section walls. Mean pressure measurements, boundary layer pressure measurements, and time-resolved, wall-normal, stereoscopic particle image velocimetry (TR-PIV) measurements were made. The TR-PIV data was acquired at a chord-based Reynolds number of 1.2 million, 2 million, and 3.5 million, at a sampling rate of 1 kHz, in two different camera configurations. The boundary layer pressure measurements were acquired at different flow Reynolds numbers ranging between 0.76 million and 3.5 million. Both adverse and favorable pressure gradients of varying intensities were imposed on the boundary layer by rotating a 0.914 m chord NACA 0012 airfoil to angles of attacks between -{10}^o and {12}^o. Measurements at varying streamwise locations enabled the study of boundary layer flow development under changing pressure gradients. The pressure gradient influences were observed in the boundary layer characteristic properties, on the mean velocities, and on the Reynolds stresses present in the flow. The pressure gradient influences were found to be consistent at varying Reynolds numbers, but the intensity of their effects was influenced by the flow Reynolds number. Moreover, the influence of pressure gradients and flow Reynolds numbers was evident in both outer and inner scales. The test data acquired was also validated with previous works. / M.S. / The interaction of turbulent boundary layers and smooth surfaces is prevalent in our world. It plays a vital role in various phenomena, such as, aircraft stall, cabin noise, and structural vibrations. Varying flow conditions influence the behavior of boundary layers and the extent of their implications. The effects of pressure gradients and the level of turbulence, described by the Reynolds numbers, on turbulent boundary layer flow was studied. This was done through an experiment conducted at the Virginia Tech Stability Wind Tunnel facility. The test data was acquired through boundary layer pressure measurements and Time-Resolved, Stereoscopic Particle Image Velocimetry (TR-PIV) at varying streamwise locations in the test section. A 0.914 m chord, NACA 0012 airfoil was placed in the test section and its angle of attack was varied to -{10}^o,0^o,\ \ and\ {12}^o to induce a favorable, minimum, and an adverse pressure gradient, respectively. The TR-PIV measurements were acquired at a sampling rate of 1 kHz and in two different camera configurations. The flow Reynolds number was based on the airfoil chord length (Re_c) and was varied to 1.2 million, 2 million, and 3.5 million for the TR-PIV tests. The boundary layer pressure measurements were acquired using an array of 30 Pitot probes placed in the boundary layer of the flow. The flow Reynolds number for these test runs ranged between 0.76 million and 3.5 million. The acquired data was used to analyze the mean statistical properties of turbulent boundary layers primarily focusing on the mean velocities, boundary characteristic parameters, Reynolds normal stresses, and Reynolds shear stresses. The results showed that the nature of pressure gradient influences on the mean properties of turbulent boundary layers remained consistent regardless of the flow Reynolds number. However, the intensity of the pressure gradient effects was influenced by the flow Reynolds number. These observations were made at various streamwise data acquisition locations through which the evolution of the flow was also studied. Lastly, the results obtained in this experiment were validated with previous works.

Mieux comprendre les spécificités de la dynamique des couches limites à grand nombre de Reynolds malgré les contraintes métrologiques et son coût de simulation numérique est crucial. A titre d'exemple, cette dynamique peut déterminer plus de la moitié de la traînée d'un avion en croisière. Décrire la turbulence pariétale peut guider la résolution numérique d'une partie des fluctuations à un coût maîtrisé par des stratégies WMLES (simulation des grandes échelles avec modèle de paroi). Les présentes analyses physiques de couches limites turbulentes incompressibles à gradient de pression nul et à grand nombre de Reynolds s'appuient sur des simulations numériques avancées. Après validation d'une base de données, le frottement moyen pariétal est décomposé selon l'identité FIK (Fukagata et al. (2002)), dont l'application malgré le développement spatial est discutée. Une analyse spectrale montre que les grandes échelles (\lambda_x > \delta) contribuent à environ la moitié du frottement vers Re_\theta = 10^4. Les limitations de l'identité FIK motivent la dérivation d'une décomposition physique de la génération du frottement dont le comportement asymptotique est alors relié à la production d'énergie cinétique turbulente dans la zone logarithmique. Pour mieux reconstruire les spectres spatiaux, une nouvelle méthode d'estimation de la vitesse de convection turbulente en fonction de la longueur d'onde des fluctuations, adaptée au développement spatial et à des signaux temporels de durée finie, est dérivée, interprétée et évaluée à Re_\theta = 13000. Certaines des conclusions éclairent des modifications d'une stratégie WMLES, le mode III de la méthode ZDES. / Better understanding the specificities of the dynamics of high-Reynolds number boundary layers despite metrological constraints and its numerical simulation cost is crucial. For instance, this dynamics can determine more than half of the drag of a cruising aircraft. Describing wall turbulence can guide the numerical resolution of some of the fluctuations at a limited cost by WMLES strategies (wall-modelled large eddy simulation). The present physical analyses of zero-pressure gradient incompressible turbulent boundary layers at high Reynolds number rely on advanced numerical simulations. After validating a database, mean skin friction is decomposed by means of the FIK identity (Fukagata et al. (2002)), whose application despite the spatial growth is discussed. A spectral analysis shows that the large scales (\lambda_x > \delta) contribute approximately half of the friction near Re_\theta = 10^4. The limitations of the FIK identity motivate the derivation of a physical decomposition of the generation of friction whose asymptotic behaviour is then related to turbulent kinetic energy production in the logarithmic layer. In order to better reconstruct spatial spectra, a new method to estimate the turbulent convection velocity as a function of the wavelength of the fluctuations, adapted to spatial growth and to temporal signals of finite duration, is derived, interpreted, and assessed at Re_\theta = 13000. Some of the conclusions enlighten modifications to a WMLES strategy, mode III of the ZDES method.

A key component of research in the aerospace industry constitutes hypersonic flights (M>5) which includes the design of commercial high-speed aircrafts and development of rockets. Computational analysis becomes more important due to the difficulty in performing experiments and reliability of its results at these harsh operating conditions. There is an increasing demand from the industry for the accurate prediction of wall-shear and heat transfer with a low computational cost. Direct Numerical Simulations (DNS) create the standard for accuracy, but its practical usage is difficult and limited because of its high cost of computation. The usage of Reynold's Averaged Navier Stokes (RANS) simulations provide an affordable gateway for industry to capitalize its lower computational time for practical applications. However, the presence of existing RANS turbulence closure models and associated wall functions result in poor prediction of wall fluxes and inaccurate solutions in comparison with high fidelity DNS data. In recent years, machine learning emerged as a new approach for physical modeling. This thesis explores the potential of employing Machine Learning (ML) to improve the predictions of wall fluxes for hypersonic turbulent boundary layers. Fine-grid RANS simulations are used as training data to construct a suitable machine learning model to improve the solutions and predictions of wall quantities for coarser meshes. This strategy eliminates the usage of wall models and extends the range of applicability of grid sizes without a significant drop in accuracy of solutions. Random forest methodology coupled with a bagged aggregation algorithm helps in modeling a correction factor for the velocity gradient at the first grid points. The training data set for the ML model extracted from fine-grid RANS, includes neighbor cell information to address the memory effect of turbulence, and an optimal set of parameters to model the gradient correction factor. The successful demonstration of accurate predictions of wall-shear for coarse grids using this methodology, provides the confidence to build machine learning models to use DNS or high-fidelity modeling results as training data for reduced-order turbulence model development. This paves the way to integrate machine learning with RANS to produce accurate solutions with significantly lesser computational costs for hypersonic boundary layer problems.