Laminar-turbulent transition has a strong impact on aerodynamic performance in many practical applications. Hence, there is a practical need for developing reliable and efficient transition prediction models, which form a critical element of the CFD process for aerospace vehicles across multiple flow regimes. This dissertation explores machine learning approaches to develop transition models using data from computations based on linear stability theory. Such data provide strong correlation with the underlying physics governed by linearized disturbance equations. In the proposed transition model, a convolutional neural network-based model encodes information from boundary layer profiles into integral quantities. Such automated feature extraction capability enables generalization of the proposed model to multiple instability mechanisms, even for those where physically defined shape factor parameters cannot be defined/determined in a consistent manner. Furthermore, sequence-to-sequence mapping is used to predict the transition location based on the mean boundary layer profiles. Such an end-to-end transition model provides a significantly simplified workflow. Although the proposed model has been analyzed for two-dimensional boundary layer flows, the embedded feature extraction capability enables their generalization to other flows as well. Neural network-based nonlinear functional approximation has also been presented in the context of transport equation-based closure models. Such models have been examined for their computational complexity and invariance properties based on the transport equation of a general scalar quantity. The data-driven approaches explored here demonstrate the potential for improved transition prediction models. / Doctor of Philosophy / Surface skin friction and aerodynamic heating caused by the flow over a body significantly increases due to the transition from laminar to turbulent flow. Hence, efficient and reliable prediction of transition onset location is a critical component of simulating fluid flows in engineering applications. Currently available transition prediction tools do not provide a good balance between computational efficiency and accuracy. This dissertation explores machine learning approach to develop efficient and reliable models for predicting transition in a significantly simplified manner. Convolutional neural network is used to extract features from the state of boundary layer flow at each location along the body. These extracted features are then processed sequentially using recurrent neural network to predict the amplification of instabilities in the flow, which is directly correlated to the onset of transition. Such an automated nature of feature extraction enables the generalization of this model to multiple transition mechanisms associated with different flow conditions and geometries. Furthermore, an end-to-end mapping from flow data to transition prediction requires no user expertise in stability theory and provides a significantly simplified workflow as compared to traditional stability-based computations. Another category of neural network-based models (known as neural operators) is also examined which can learn functional mapping from input variable field to output quantities. Such models can learn directly from data for complex set of problems, without the knowledge of underlying governing equations. Such attribute can be leveraged to develop a transition prediction model which can be integrated seamlessly in flow solvers. While further development is needed, such data-driven models demonstrate the potential for improved transition prediction models.

Direct Numerical Simulations (DNS) are performed to investigate laminar-turbulent transition in a boundary layer on a sharp cone at Mach 6. The main objective of this dissertation research is to explore which nonlinear breakdown mechanisms may be dominant in a broad--band "natural" disturbance environment and then use this knowledge to perform controlled transition simulations to investigate these mechanisms in great detail. Towards this end, a "natural" transition scenario was modeled and investigated by generating wave packet disturbances. The evolution of a three-dimensional wave packet in a boundary layer has typically been used as an idealized model for "natural" transition to turbulence, since it represents the impulse response of the boundary layer and, thus, includes the interactions between all frequencies and wave numbers. These wave packet simulations provided strong evidence for a possible presence of fundamental and subharmonic resonance mechanisms in the nonlinear transition regime. However, the fundamental resonance was much stronger than the subharmonic. In addition to these two resonance mechanisms, the wave packet simulations also indicated the possible presence of oblique breakdown mechanism. To gain more insight into the nonlinear mechanisms, controlled transition simulations were performed of these mechanisms. Several small and medium scale simulations were performed to scan the parameter space for fundamental and subharmonic resonance. These simulations confirmed the findings of the wave packet simulations, namely that, fundamental resonance is much stronger compared to the subharmonic resonance. Subsequently a set of highly resolved fundamental and oblique breakdown simulations were performed. In these DNS, remarkable streamwise arranged "hot'' streaks were observed for both fundamental and oblique breakdown. The streaks were a consequence of the large amplitude steady longitudinal vortex modes in the nonlinear régime. These simulations demonstrated that both second--mode fundamental breakdown and oblique breakdown may indeed be viable paths to complete breakdown to turbulence in hypersonic boundary layers at Mach 6.

L’objectif de cette thèse est de faire progresser la modélisation de la transition de couche limite sur des aubes de turbines basse-pression fortement chargées. Cette modélisation repose sur l’utilisation du modèle de transition de Menter et Langtry utilisé pour des calculs RANS dans le code elsA. Une fois les limitations du modèle de transition clairement identifiées par une étude sur la mise en données des calculs, nous avons entrepris de modifier ce dernier. Pour cela, un processus d’optimisation a été développé afin de permettre la recalibration des fonctions de corrélation internes au modèle de transition. Cette nouvelle version du modèle nous permet d’obtenir des gains sur la modélisation d’environ 20 % sur les cas T106C du VKI en capturant mieux la transition au sein du bulbe de décollement. Ces précédents calculs correspondent à des cas idéaux, où l’on peut considérer les surfaces comme étant lisses. Cependant, nous avons aussi un besoin de se rapprocher de surfaces plus réalistes pour lesquelles les rugosités peuvent avoir un impact sur l’écoule- ment. En effet, les rugosités de surface peuvent notamment avoir un effet sur la transition. En particulier, si les rugosités entraînent le déclenchement de la transition en amont du point de décollement laminaire théorique en surface lisse, ce décollement sera supprimé. Vu nos efforts pour améliorer la prévision de la transition par bulbe de décollement par le modèle γ-Rθt, il parait intéressant que celui-ci puisse prendre en compte l’état des surfaces. Pour cela, nous avons implanté une méthode de prévision de la transition sur surfaces rugueuses développée par Stripf et al. au sein du modèle γ-Rθt. Enfin, l’utilisation du modèle de transition γ-Rθt a été étendue au modèle de turbulence k-l de Smith. / The goal of this thesis is to enhance laminar-turbulent transition modeling on high-lift low- pressure turbine blades. The presented transition modeling method relies on the Menter and Langtry transition model used in a RANS framework in the elsA solver. Once the model’s limits were clearly identified through a parametric study, we moved on to modification of the model. To do so, an optimization method was developed that allows recalibration of the model’s inner correlation functions. This new version of the model allows us to obtain modeling gains of about 20% on the VKI T106C cases through better capture of the separation-induced transition process. These previous computations correspond to ideal cases, for which surfaces may be considered as being smooth. However, we also have the need to consider more realistic surfaces for which roughness may influence the flow. Indeed, among those effects, is the potential influence of surface roughness on transition. In particular, if surface roughness induces transition up-stream of the smooth separation point, the separation bubble will be suppressed. Considering our efforts on modeling separation-induced transition with the γ-Rθt model, it seemed natural to add roughness-induced transition modeling capacities to it. To do so, we implemented in the γ-Rθt model a method developed by Stripf et al. to take into account surface roughness. Finally, the use of the γ-Rθt transition model was extended to the k-l of Smith tur- bulence model. Indeed, this turbulence model is widely used in turbomachinery. In order that our works on transition modeling over turbine blades be more widely usable, we have completed this thesis by proposing an evolution of the transition model so that it may be used alongside the k-l model.

Infrared transition location estimates can be fast and useful measurements in wind tunnel and flight tests. Because turbulent boundary layers have a much higher rate of convective heat transfer than laminar boundary layers, a difference in surface temperature can be observed between turbulent and laminar regions of an airfoil at a different temperature than the free stream air temperature. Various implementations of this technique are examined in a wind tunnel. These include using a heat lamp as an external source and circulating fluid inside of the airfoil. Furthermore, ABS plastic and aluminum airfoils are tested with and without coatings such as black paint and surface wraps. The results show that thermal conduction within the model and surface reflections are the driving issues in designing an IR system for detecting transition. Aluminum has a high thermal diffusivity so is a poor choice for this method. However, its performance can be improved using an insulating layer. Internal fluid circulation was far more successful than the heat lamp because it eliminates the reflected IR due to the heat lamp. However, using smooth surface wraps can mitigate reflection issues caused by the heat lamps by reducing the scatter within the reflection, producing an IR image with fewer contaminating reflections.

The condition of fluid flow has been proven to have a significant influence on a wide variety of material processes. In electromagnetic levitation (EML) experiments, the internal flow is driven primarily by electromagnetic forces. In 1-g, the positioning forces are very strong and the internal flows are turbulent. To reduce the flows driven by the levitation field, experiments may be performed in reduced gravity and parabolic flights experiments have been adopted as the support in advance. Tracer particles on the surface of levitated droplets in EML experiment performed by SUPOS have been used to investigate the transition from laminar to turbulent flow. A sample of NiAl3 was electromagnetically levitated in parabolic flight and the laminar-turbulent transition observed from the case was studied in this work. For the sample with clearly visible tracer patterns, the fluid flow has been numerical evaluated with magnetohydrodynamic models and the laminar-turbulent transition happened during the acceleration of the flow, instead of steady state. The Reynolds number at transition was estimated approximately as 860 by the experiment record. The predicted time to transition obtained from the results of simulation showed significant difference (~ up to 300 times) compared with the time obtained from the experiment—0.37s. The discrepancy between numerical and experimental results could not be explained by the proposed hypotheses: geometry, boundary conditions or solid core. The simulations predict that the flow would become turbulent almost instantaneously after the droplet was fully molten. There are important physics shown by the simulation which were not captured.

The focus of this thesis is on the numerical study of subcritical transition to turbulence in the asymptotic suction boundary layer (ASBL). Applying constant homogeneous suction prevents the spatial growth of the boundary layer, granting access to the asymptotic dynamics. This enables research approaches which are not feasible in the spatially growing case. In a first part, the laminar–turbulent separatrix of the ASBL is investigated numerically by means of an edge-tracking algorithm. The consideration of spanwise-extended domains allows for the robust localisation of the attracting flow structures on this separatrix. The active part of the identified edge states consists of a pair of low- and high-speed streaks, which experience calm phases followed by high energy bursts. During these bursts the structure is destroyed and re-created with a shift in the spanwise direction. Depending on the streamwise extent of the domain, these shifts are either regular in direction and distance, and periodic in time, or irregular in space and erratic in time. In all cases, the same clear regeneration mechanism of streaks and vor- tices is identified, bearing strong similarities with the classical self-sustaining cycle in near-wall turbulence. Bifurcations from periodic to chaotic regimes are studied by varying the streamwise length of the (periodic) domain. The resulting bifurcation diagram contains a number of phenomena, e.g. multistability, intermittency and period doubling, usually investigated in the context of low-dimensional systems. The second part is concerned with spatio–temporal aspects of turbulent ASBL in large domains near the onset of sustained turbulence. Adiabatically decreasing the Reynolds number, starting from a fully turbulent state, we study low-Re turbulence and events leading to laminarisation. Furthermore, a robust quantitative estimate for the lowest Reynolds number at which turbulence is sustained is obtained at Re <img src="http://www.diva-portal.org/cgi-bin/mimetex.cgi?%5Capprox" /> 270. / <p>QC 20140213</p>

A novel laminar flow control based on generation of spanwise mean velocity gradients (SVG) in a flat plate boundary layer is investigated where disturbances of different types are introduced in the wall-bounded shear layer. The experimental investigations are aimed at; (i) generating stable and steady streamwise streaks in the boundary layer which set up spanwise gradients in the mean flow, and (ii) attenuating disturbance energy growth in the streaky boundary layers and hence delaying the onset of turbulence transition. The streamwise streaks generated by four different methods are investigated, which are spanwise arrays of triangular/rectangular miniature vortex generators (MVGs) and roughness elements, non-linear pair of oblique waves, and spanwise-periodic finite discrete suction. For all the investigated methods the boundary layer is modulated into regions of high- and low speed streaks through formation of pairs of counter-rotating streamwise vortices. For the streaky boundary layers generated by the MVGs a parameter study on a wide range of MVG configurations is performed in order to investigate the transient growth of the streaks. A general scaling of the streak amplitudes is found based on empiricism where an integral amplitude definition is proposed for the streaks. The disturbances are introduced as single- and broad band frequency twodimensional Tollmien–Schlichting (TS) waves, and three-dimensional single and a pair of oblique waves. In an attempt to obtain a more realistic configuration compared to previous investigations the disturbances are introduced upstream of the location were streaks are generated. It is shown that the SVG method is efficient in attenuating the growth of disturbance amplitudes in the linear regime for a wide range of frequencies although the disturbances have an initial amplitude response to the generation of the streaks. The attenuation rate of the disturbance amplitude is found to be optimized for an integral streak amplitude of 30% of the free-stream velocity which takes into account the periodic wavelength of the streaky base flow. The stabilizing effect of the streamwise streaks can be extended to the nonlinear regime of disturbances which in turn results in transition to turbulence delay. This results in significant drag reduction when comparing the skin friction coefficient of a laminar- to a turbulent boundary layer. It is also shown that consecutive turbulence transition delay can be obtained by reinforcing the streaky boundary layer in the streamwise direction. For the streaky boundary layer generated by pair of oblique waves their forcing frequency sets the upper limit for the frequency of disturbances beyond which the control fails. / <p>QC 20160208</p>

<p>Transition from laminar to turbulent flow has beentraditionally studied in terms of exponentially growingeigensolutions to the linearized disturbance equations.However, experimental findings show that transition may occuralso for parameters combinations such that these eigensolutionsare damped. An alternative non-modal growth mechanism has beenrecently identified, also based on the linear approximation.This consists of the transient growth of streamwise elongateddisturbances, mainly in the streamwise velocity component,called streaks. If the streak amplitude reaches a thresholdvalue, secondary instabilities can take place and provoketransition. This scenario is most likely to occur in boundarylayer flows subject to high levels of free-stream turbulenceand is the object of this thesis. Different stages of theprocess are isolated and studied with different approaches,considering the boundary layer flow over a flat plate. Thereceptivity to free-stream disturbances has been studiedthrough a weakly non-linear model which allows to disentanglethe features involved in the generation of streaks. It is shownthat the non-linear interaction of oblique waves in thefree-stream is able to induce strong streamwise vortices insidethe boundary layer, which, in turn, generate streaks by thelift-up effect. The growth of steady streaks is followed bymeans of Direct Numerical Simulation. After the streaks havereached a finite amplitude, they saturate and a new laminarflow, characterized by a strong spanwise modulation isestablished. Using Floquet theory, the instability of thesestreaks is studied to determine the features of theirbreakdown. The streak critical amplitude, beyond which unstablewaves are excited, is 26% of the free-stream velocity. Theinstability appears as spanwise (sinuous-type) oscillations ofthe streak. The late stages of the transition, originating fromthis type of secondary instability, are also studied. We foundthat the main structures observed during the transition processconsist of elongated quasi-streamwise vortices located on theflanks of the low speed streak. Vortices of alternating signare overlapping in the streamwise direction in a staggeredpattern.</p><p><strong>Descriptors:</strong>Fluid mechanics, laminar-turbulenttransition, boundary layer flow, transient growth, streamwisestreaks, lift-up effect, receptivity, free-stream turbulence,nonlinear mechanism, streak instability, secondary instability,Direct Numerical Simulation.</p> / QC 20100518

Diversas aplicações industriais utilizam escoamentos de fluidos viscoelásticos, e em muitos casos é necessário saber se os escoamentos propagam-se no estado laminar ou no turbulento. Embora a hidrodinâmica de fluidos viscoelásticos seja fortemente afetada pelo balanço entre forças inerciais e elásticas no escoamento, o efeito da elasticidade sobre a estabilidade de escoamentos inerciais não foi completamente estabelecido. Neste trabalho, estuda-se o que ocorre durante a transição laminar-turbulenta, investigando a convecção de ondas de Tollmien-Schlichting para o escoamento incompressível, para um fluido viscoelástico, entre placas paralelas, utilizando a equação constitutiva Giesekus. Para isto, adotou-se a simulação numérica direta para verificar a estabilidade dos escoamentos à perturbações não estacionárias deste fluido. Experimentos computacionais para verificação do código foram realizados. Com os resultados numéricos obtidos, foi possível verificar e analizar a estabilidade de escoamentos utilizando-se o modelo não newtoniano Giesekus. / Several industrial applications use viscoelastic fluid flows, and it is necessary to know if the flows propagate in the laminar or turbulent state. Although the hydrodynamics of viscoelastic fluids is strongly affected by the balance between inertial and elastic forces in the flow, the effect of elasticity on the stability of inertial flows has not been completely established. In this work we study what happens during the laminar-turbulent transition, investigating the convection of Tollmien-Schlichting waves for the incompressible flow, for a viscoelastic fluid, between parallel plates, using the constitutive equation Giesekus. For this, the direct numerical simulation was used to verify the stability of the flows to the non-stationary perturbations of this fluid. Computational experiments to verify the code were performed. With the numerical results obtained, it was possible to verify and analyze the stability of flows modelled by Giesekus non-newtonian model.

Nos escoamentos em turbomáquinas temos como principais características a tridimensionalidade, possível ocorrência de separação da camada limite, relaminarização, transição laminar-turbulenta, dentre outros efeitos físicos. De acordo com alguns estudos experimentais em turbinas observouse que a transição laminar-turbulenta pode se estender por até 60% da corda de uma pá de turbina. Uma boa estimativa para se prever corretamente o local da transição é indispensável para que seja obtida uma melhoria na eficiência das turbinas. Escoamentos sobre superfícies côncavas estão sujeitos à instabilidade centrífuga, podendo dar origem a vórtices longitudinais, conhecidos como vórtices de Görtler. Esses vórtices são responsáveis por gerar distorções fortes nos perfis de velocidade e consequentemente nos perfis de temperatura. O presente estudo tem por objetivo estudar a influência da variação da curvatura de uma superfície côncava, e os efeitos do comprimento de onda transversal no processo de transição, e sua influência nas taxas de transferência de calor. Para tal, um código de simulação numérica paralelizado, com alta ordem de precisão, foi utilizado para resolver numericamente as equações de Navier-Stokes. Este código é validado através de comparações entre resultados obtidos com uso da teoria de estabilidade linear, e com resultados de simulações numéricas não lineares. Resultados obtidos evidenciam a influência da variação da curvatura, e os efeitos causados pelo comprimento de onda transversal nas instabilidades de Görtler, e secundária. Tais evidências comprovam que a variação da curvatura pode ser útil no controle do processo de transição laminar-turbulenta, e que as taxas de transferência de calor de um escoamento de Görtler desenvolvido em superfícies de curvatura variável podem ser intensificadas, atingindo valores superiores aos obtidos em escoamentos turbulentos. / Some characteristics of flows in turbomachinery are the three-dimensionality, possible occurrence of separation of the boundary layer, relaminarization, laminar-turbulent transition, among other physical effects. According to some experimental observations in turbines, it has been observed that the laminar-turbulent transition can extend over 60% chord of a turbine blade. A good estimate to correctly predict the location of the transition is essential for an improvement in the efficiency of turbines. Flow over concave surfaces is subjected to centrifugal instability, which may lead to formation of longitudinal vortices, known as the Görtler vortices. These vortices are responsible for generating strong distortions in the velocity profiles and hence the temperature profiles. The current goal aims to study the influence of the curvature variation of a concave surface and the effects of spanwise wavelength on the transition process and its influence on the heat transfer rates. For this, a parallel numerical simulation code, with a high order of precision, was used to numerically solve the Navier-Stokes equations. This code is validated through comparisons between results obtained using linear stability theory, and nonlinear numerical simulations results. Results obtained show the influence of the curvature variation, and the effects caused by the spanwise wavelength on the Görtler and secondary instabilities. This evidence proves that the curvature variation can be useful in the control of the laminar-turbulent transition process, and that heat transfer rates of a Görtler flow developed on variable curvature surfaces can be intensified, and reach values higher than these achieved in turbulent flows.