Computations were run to study heat transfer coefficient augmentation with film cooling for a simulated gas turbine blade leading edge. The realizable k-[epsilon] turbulence model (RKE) and Shear Stress Transport k-[omega] turbulence model (SST) were used for the computational simulations. RKE computations completed at a unity density ratio were confirmed to be consistent with experimental measurements conducted by Yuki et al.(1998) and Johnston et al. (1999) whereas SST computations exhibited significant discrepancies. Moreover the effect of the density ratio on heat transfer coefficient augmentation was studied because experimental measurements of heat transfer coefficient augmentation with film cooling are generally constrained to unity density ratio tests. It was shown that heat transfer coefficient augmentation can be simulated using unity density ratio jets, but only when scaled with the momentum flux ratio of the coolant jets. / text

Cette thèse présente un framework d’optimisation pour la conception hydrodynamique de forme d’hydrofoils. L’optimisation d’hydrofoil par simulation implique des objectifs d’optimisation divergents et impose des compromis contraignants en raison du coût des simulations numériques et des budgets limités généralement alloués à la conception des navires. Le framework fait appel à l’échantillonnage séquentiel et aux modèles de substitution. Un modèle prédictif est construit en utilisant la Régression par Processus Gaussien (RPG) à partir des données issues de simulations fluides effectuées sur différentes géométries d’hydrofoils. Le modèle est ensuite combiné à d’autres critères dans une fonction d’acquisition qui est évaluée sur l’espace de conception afin de définir une nouvelle géométrie qui est testée et dont les paramètres et la réponse sont ajoutés au jeu de données, améliorant ainsi le modèle. Une nouvelle fonction d’acquisition a été développée, basée sur la variance RPG et la validation croisée des données. Un modeleur géométrique a également été développé afin de créer automatiquement les géométries d’hydrofoil a partir des paramètres déterminés par l’optimiseur. Pour compléter la boucle d’optimisation,FINE/Marine, un solveur fluide RANS, a été intégré dans le framework pour exécuter les simulations fluides. Les capacités d’optimisation ont été testées sur des cas tests analytiques montrant que la nouvelle fonction d’acquisition offre plus de robustesse que d’autres fonctions d’acquisition existantes. L’ensemble du framework a ensuite été testé sur des optimisations de sections 2Dd’hydrofoil ainsi que d’hydrofoil 3D avec surface libre. Dans les deux cas, le processus d’optimisation fonctionne, permettant d’optimiser les géométries d’hydrofoils et confirmant les performances obtenues sur les cas test analytiques. Les optima semblent cependant être assez sensibles aux conditions opérationnelles. / This thesis presents a practical hydrodynamic optimization framework for hydrofoil shape design. Automated simulation based optimization of hydrofoil is a challenging process. It may involve conflicting optimization objectives, but also impose a trade-off between the cost of numerical simulations and the limited budgets available for ship design. The optimization frameworkis based on sequential sampling and surrogate modeling. Gaussian Process Regression (GPR) is used to build a predictive model based on data issued from fluid simulations of selected hydrofoil geometries. The GPR model is then combined with other criteria into an acquisition function that isevaluated over the design space, to define new querypoints that are added to the data set in order to improve the model. A custom acquisition function is developed, based on GPR variance and cross validation of the data.A hydrofoil geometric modeler is also developed to automatically create the hydrofoil shapes based on the parameters determined by the optimizer. To complete the optimization loop, FINE/Marine, a RANS flow solver, is embedded into the framework to perform the fluid simulations. Optimization capabilities are tested on analytical test cases. The results show that the custom function is more robust than other existing acquisition functions when tested on difficult functions. The entire optimization framework is then tested on 2D hydrofoil sections and 3D hydrofoil optimization cases with free surface. In both cases, the optimization process performs well, resulting in optimized hydrofoil shapes and confirming the results obtained from the analytical test cases. However, the optimum is shown to be sensitive to operating conditions.

Les études expérimentales et numériques suivantes visent à la compréhension profonde de l’écoulement se développant dans le compresseur haute-vitesse axial de 3.5 étages CREATE, étudié sur un banc d’essai de 2 MW au Laboratoire de Mécanique des Fluides et Acoustique (LMFA) à Lyon, France. Ce travail a trois objectifs principaux : D’abord, une description globale de l’écoulement avec une identification des limites aux méthodes d’exploration utilisées ; Ensuite, la caractérisation de l’effet du clocking stator-stator dans un compresseur à haute-vitesse ; Troisièmement, l’identification des instabilités à faibles débits pour confirmer les études sur les compresseurs à basse-vitesse et contribuer à plus de compréhension.Il est montré qu’une mauvaise interprétation des données de performance stationnaire se fait facilement en raison des contraintes de mesure et des coefficients de correction sont proposés. À certains endroits dans le compresseur, des limites aux méthodes d’exploration (expérimentales et numériques) de l’écoulement sont identifiées. Cette identification va permettre la poursuite du développement des méthodes. Les principales erreurs de prédiction des simulations concernent la surestimation du blocage induit par l’écoulement de jeu et l’augmentation de pression. En outre, les mesures fournies par les sondes de pression pneumatique surestiment la pression statique en amont des stators. Cette erreur est probablement provoquée par l’interaction entre le champ potentiel du stator et la sonde elle-même. De plus, l’anémométrie Doppler laser surestime la vitesse en aval des stators. Le transport des sillages du rotor à travers des stators n’est pas correctement capturé avec les particules d’ensemencement.Le clocking a seulement un petit effet global dans la bande d’incertitude de mesure dans ce compresseur. Plusieurs contributions à ce faible effet de clocking sont identifiées par l’analyse du transport des structures d’écoulement : Le mélange circonférentiel du sillage de stator et la déformation des sillages le long de leur trajet dans l’écoulement. L’effet local du clocking dépend de la hauteur de veine en raison de la variation de la forme des aubages et du transport des sillages. Des effets positifs et négatifs sont présentés, qui globalement se compensent dans ce compresseur. Les instabilités dans ce compresseur dépendent du point de fonctionnement et des méthodes d’exploration de l’écoulement. Aux points de fonctionnement stables et à la vitesse nominale du compresseur, les résultats numériques montrent une perturbation tournante dans les rotors 2 et 3, alors que les mesures montrent une perturbation tournante que dans le premier rotor et seulement à basse vitesse du compresseur. Dans les deux cas, les perturbations montrent des caractéristiques semblables. Une étude numérique permet d’exclure l’influence des interactions rotor-stator sur la perturbation tournante et met en évidence sa source. Des nouvelles connaissances sur le comportement stable et la périodicité du rotating instability (mesuré) sont dérivées contrairement au comportement instable suggéré par la dénomination et la littérature. Il est montré que cette perturbation évolue en cellule de décrochage tournante à l’approche de la limite de stabilité. A la vitesse nominale du compresseur, une entrée en instabilités de type spike est identifiée expérimentalement. Une description précise de l’apparition brutale du spike et sa différence par rapport à une cellule de décollement tournant sont présentées. / The following experimental and numerical investigations aim at the deep understanding of the flow field in the 3.5 stages high-speed axial compressor CREATE, studied on a 2 MW test rig at the Laboratory of Fluid Mechanics and Acoustics (LMFA) in Lyon, France. This work focuses on three major objectives: Firstly, a global description of the flow field with an identification of limitations to the used exploration methods; Secondly, the characterization of the effect of stator-stator clocking in a high-speed compressor; Thirdly, the identification of instabilities arising at low mass flow rates for confirming studies on low-speed compressors and giving new insights.This work demonstrates that a mis-interpretation of steady performance data occurs easily due to measurement constraints and correction coefficients are proposed. At certain locations in the compressor, the flow field exploration (experimental and numerical) methods are identified to be challenged. This identification will initiate further development of the methods. The main mis-predictions of the simulations concern the over-prediction of the blockage induced by the tip leakage flow and eventually an over-predicted pressure rise. Furthermore, the measurements provided by the pneumatic pressure probes over-estimate the static pressure upstream of the stators. This error is induced by the interaction between the stator potential field and the probe it-self. In addition, the laser Doppler anemometry method over-estimates the velocity downstream the stators. The transport of the rotor wakes through the stators might not be correctly captured with the seeding particles in this high-speed compressor.The investigation of the stator clocking reveals only a small global effect within the measurement uncertainty band. Several contributions to the weak effect of clocking are identified by analysis of the flow structure transport, namely the time-mean mixing out of the stator wakes and the deformation of wakes along their flow path. The local effect of clocking depends on the span-height because of the variation of the circumferential position of the stator wakes and the stator blade shape over the span-height. Local possible positive and negative effects of clocking are identified and are shown to be almost in balance in this compressor. Furthermore, this work demonstrates that the unsteadiness in the flow field is not linked conclusively to the stator clocking.In this compressor, the arising instabilities depend on the operating point and flow field exploration methods. At stable operating points and nominal compressor speed, the numerical results reveal a rotating disturbance in the rotors 2 and 3, whereas the measurements show a rotating disturbance only in the first rotor and only at part speed. In both cases the disturbance exhibits rotating instability like characteristics. An exhaustive numerical study allows to exclude the commonly assumed influence of rotor-stator interactions on the rotating disturbance and pinpoints its source. New insights into the stable behavior and periodicity of the measured rotating instability are derived contrary to the unstable behavior suggested by the naming and literature. This disturbance is shown to evolve into rotating stall cells when approaching the stability limit. At nominal compressor speed, a spike type surge inception is identified I n the measured field. A precise description of the abrupt onset of the spike cell and its difference to a rotating stall cell are presented.

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.

Convergent-Divergent (C-D) nozzles are used in rocket engines to produce thrust as a reaction to the acceleration of hot combustion chamber gases in the opposite direction. To maximize the engine performance at high altitudes, large area ratio, bell-shaped or contoured nozzles are used. At lower altitudes, the exit pressure of these nozzles is lower than the ambient pressure. During this over-expanded condition, the nozzle-internal flow adapts to the ambient pressure through an oblique shock. But the boundary layer inside the divergent portion of the nozzle is unable to withstand the pressure rise associated with the shock, and consequently flow separation is induced. Numerical simulation of separated flows in rocket nozzles is challenging because the existing turbulence models are unable to correctly predict shock-induced flow separation. The present thesis addresses this problem. Axisymmetric, steady-state, Reynolds-Averaged Navier-Stokes (RANS) simulations of a conical nozzle and three sub-scale contoured nozzles were carried out to numerically predict flow separation in over-expanded rocket nozzles at different nozzle pressure ratios (NPR). The conical nozzle is the JPL 45◦-15◦ and the contoured nozzles are the VAC-S1, the DLR-PAR and the VAC-S6-short. The commercial CFD code ANSYS FLUENT 13 was first validated for simulation of separated cold gas flows in the VAC-S1 nozzle. Some modeling issues in the numerical simulations of flow separation in rocket nozzles were determined. It is recognized that compressibility correction, nozzle-lip thickness and upstream-extension of the external domain are the sources of uncertainty, besides turbulence modeling. In high-speed turbulent flows, compressibility is known to affect dissipation rate of turbulence kinetic energy. As a consequence, a reduction in the spreading rate of supersonic mixing layers occurs. Whereas, the standard turbulence models are developed and calibrated for incompressible flows and hence, do not account for this effect. ANSYS FLUENT uses the compressibility correction proposed by Wilcox [1] which modifies the turbulence dissipation terms based on turbulent Mach number. This, as shown in this thesis, may not be appropriate to the prediction of flow separation in rocket nozzles. Simulation results of the standard SST model, with and without the compressibility correction, are compared with the experimental data at NPR=22 for the DLR-PAR nozzle. Compressibility correction is found to cause under-prediction of separation location and hence its use in the prediction of flow separation is not recommended. In the literature, computational domains for the simulation of DLR subscale nozzles have thick nozzle-lips whereas for the VAC subscale nozzles they have no nozzle-lip. Effect of nozzle-lip thickness on flow separation is studied in the DLR-PAR nozzle by varying its nozzle-lip thickness. It is found that nozzle-lip thickness significantly influences both separation location and post-separation pressure recovery by means of the recirculation bubbles formed at the nozzle-lip. Usually, experimental values of free stream turbulence are unknown. So conventionally, to minimize solution dependence on the boundary conditions specified for the ambient flow, the computational domain external to the nozzle is extended in the upstream direction. Its effect on flow separation is studied in the DLR-PAR nozzle through simulations conducted with and without this domain extension. No considerable effect on separation location and pressure recovery is found. The two eddy-viscosity based turbulence models, Spalart-Allmaras (SA) model and Shear Stress Transport (SST) model, are well known to predict separation location better than other eddy-viscosity models, but with moderate success. Their performances, in terms of predicting separation location and post-separation wall pressure distribution, were compared with each other and evaluated against experimental data for the conical and two contoured nozzles. It is found that they fail to predict the separation location correctly, exhibiting sensitivity to the range of NPRs and to the type of nozzle. Depending on NPR, the SST model either under-predicts or over-predicts Free Shock Separation (FSS). Moreover, it also fails to capture Restricted Shock Separation (RSS). With compressibility correction, it under-predicts separation at all NPRs to a greater extent. Even though RSS is captured by using compressibility correction, the transition from FSS to RSS is over-predicted [2]. Early efforts by few researchers to improve predictions of nozzle flow separation by realizability corrections to turbulence models have not been successful, especially in terms of capturing both the separation types. Therefore, causes of turbulence modeling failure in predicting nozzle flow separation correctly were further investigated. It is learnt that limiting of the shear stress inside boundary layer, due to Bradshaw’s assumption, and over-prediction of jet spreading rate are the causes of SST model’s failure in predicting nozzle flow separation correctly. Based on this physical reasoning, values of the a 1 parameter and the two diffusion coefficients σk,2 and σω,2 were empirically modified to match the predicted wall pressure distributions with experimental data of the DLR-PAR and the VAC-S6-short nozzles. The results confirm that accurate prediction of flow separation in rocket nozzles indeed depends on the correct prediction of spreading rate of the supersonic separation-jet. It is demonstrated that accurate RANS simulation of flow separation in rocket nozzles over a wide range of NPRs is feasible by modified values of the diffusion coefficients in turbulence model.