A supersonic ejector is used to pump a secondary gas using a supersonic primary gas flow by augmentation of momentum and energy in a variable area duct. The internal compressible flow through an ejector has many complex gas dynamic features, like compressible shear layers and associated shock interactions. In many practical applications, ejectors are operated in the choked flow regimes where higher operating pressure ratios and mass flow rates are encountered. On the other hand, rather low entrainment and subsonic secondary flow dynamics (referred as the mixed regime of operation) dominate the dilution and purging applications of ejectors. The fundamental understanding of the flow dynamics associated with gaseous mixing process in the ejector especially in the mixed operational regime is still unclear. Obtaining a comprehensive understanding of the flow through a supersonic ejector in the mixed regime through experimental investigations is the prime focus of the present study. A new supersonic ejector test facility is designed, fabricated and established in the laboratory during the course of this study. The effects of using different gases in the secondary flow have been investigated. Two novel methods to improve the ejector by enhancing mixing are also implemented. Optical diagnostic tools (Time-resolved Schlieren and laser scattering) and wall static pressure measurements are used to investigate the dynamics of mixing process inside the ejector. State of the art image processing codes are developed to determine the length in the ejector for which the primary and the secondary flows are separate, referred here as the non-mixed length from the results of the flow visualization studies. Exhaustive experiments are carried out on the two dimensional rectangular supersonic ejector by varying the mass flow rates of primary and secondary flows, primary stagnation pressure, for two locations of the nozzle in the ejector. The non-mixed length determined from quantitative flow visualization tools is found to lie within 4.5 to 5.2 times the height of the duct (20 mm). The non-mixed flow length determined from flow visualization studies corroborates well with the wall static pressure measurements. A significant reduction of non-mixed length of about 46.7% is caused by shock wave-boundary layer interactions in the supersonic nozzle at over-expanded conditions. Further, the effects of differences in molecular weight and ratio of specific heats on the performance are also studied using cylindrical supersonic ejector at low entrainment ratios (0.008 to 0.06). In these studies air is used as the primary fluid while argon and helium are used in the secondary flow segment of the ejector. The results indicate that Argon has better entrainment characteristics compared to helium. Two novel supersonic nozzles (the tip rig nozzle and Elliptic Sharp Tipped Shallow lobed nozzle) are also devel- oped to enhance mixing in the ejector. About 30% enhancement of entrainment ratio is observed with the newly designed nozzle geometries. Illustrative numerical simulations are also carried out to complement the experimental studies.

Abstract : This research project has as main objective to study in detail the internal flow features of single-phase supersonic ejectors for refrigeration applications, and the potential effects of injecting droplets on the performance of the device. To this end, a numerical approach is proposed which has been separated into two parts: First, a RANS modelling strategy for supersonic ejectors has been outlined combining the NIST real gas equations database [NIST, 2010] and the k − ω SST turbulence model in its low-Reynolds number formulation. The proposed approach agrees within 5% (resp. 2%) to the experimental entrainment ratio (resp. compression ratio) data of García del Valle et al. [2014], properly captures the main internal flow features and has a reasonable computational cost. This RANS model has been applied in the analysis of a supersonic R134a ejector for refrigeration purposes, showing in particular that the secondary flow is entrained by momentum transfer through the mixing shear layer, that the distance between the primary nozzle exit and the shock-waves in the constant area section varies between 9 and 16 times the primary nozzle exit diameter and that the important axial character of the flow limits mixing of both inlet flows until after the shock train. Furthermore, an exergy analysis through the device shows that the mixing and the oblique shock waves are responsible for between 50% and 70% of the generated losses, the latter might be attenuated through droplet injection in the constant area section. Moreover, it has been shown that drop-in replacement of the working fluid with HFOs R1234yf and R1234ze(E) leads to mild changes in the ejector performance but reduces the HDRC system COP (resp. cooling capacity) in average by 7.1% (resp. 23.3%). Lastly, a comparison of the model predictions with the thermodynamic model of Galanis and Sorin [2016] for an air ejector, shows that as the working fluid approaches the ideal gas behaviour, the flow can be adimensionalized in terms of the secondary inlet temperature and pressure, the motive nozzle throat and the entrainment and compression ratios. In the second part, the influence of droplets has been studied from a local perspective by extending the RANS model to include a discrete phase, which affects the main flow through exchanges of momentum and thermal energy, and from a global perspective by building a thermodynamic model, which predicts the entrainment and limiting compression ratio given a fixed geometry and operating conditions. Both approaches present very good agreement in terms of p, T and M a internal profiles. Results for a supersonic ejector with R134a as baseline working fluid and droplets injected at the constant area section show that the flow structure has perceptible changes only at the highest injection fraction considered 10%, which induces boundary layer detachment, reduces the shock intensity by 8% and diminishes the superheat at the ejector outlet by 15 ◦C. Nonetheless, ejector performance metrics are severely affected as the limiting compression ratio, Elbel efficiency and exergy performance reduce respectively by 5%, 11% and 15%, due mainly to the additional entropy generated through droplet injection and mixing with the main flow. / Ce projet de recherche a pour objectif principal d’étudier en détail les caractéristiques de l’écoulement interne dans des éjecteurs supersoniques monophasiques pour des applications en réfrigération, et les effets potentiels de l’injection de gouttelettes sur les performances de l’appareil. A cette fin, une approche numérique est proposée et a été séparée en deux parties. Tout d’abord, une stratégie de modélisation RANS pour les éjecteurs supersoniques a été décrite en combinant la base de données pour les gaz réels NIST [NIST, 2010] et le modèle de turbulence k − ω SST dans sa formulation à bas nombre de Reynolds. L’approche proposée prédit avec un accord d’environ 5% (resp. 2%) le rapport d’entraînement (resp. rapport de compression) avec les données expérimentales de García del Valle et al. [2014]. Il capte également correctement les principales caractéristiques de l’écoulement interne et a un coût de calcul raisonnable. Ce modèle RANS a été appliqué à l’analyse d’un éjecteur supersonique au R134a utilisé à des fins de réfrigération, montrant en particulier que le flux secondaire est entraîné par un transfert d’impulsion à travers la couche de cisaillement, que la position de départ des ondes de choc dans la section constante se situe dans une plage de 9 à 16 fois le diamètre de sortie de la buse primaire et que l’important caractère axial du flux limite le mélange des deux écoulements d’entrée au-delà du train d’ondes de choc. De plus, une analyse exergétique à travers le dispositif montre que le mélange et les ondes de choc obliques sont responsables de 50% et 70% des pertes générées, ces dernières pouvant être atténuées par injection de gouttelettes dans la section à zone constante. De plus, il a été démontré que le remplacement direct du fluide de travail par les HFO R1234yf et R1234ze(E) entraîne de légers changements dans la performance de l’éjecteur mais réduit en moyenne le COP du système HDRC (resp. la capacité de refroidissement) de 7.1% (resp. 23.3%). Enfin, une comparaison des prédictions du modèle avec le modèle thermodynamique de Galanis and Sorin [2016] pour un éjecteur à air montre que lorsque le fluide de travail se rapproche du comportement de gaz idéal, l’écoulement peut être normalisé en fonction de la température et de la pression à l’entrée secondaire, la gorge de la tuyère principale et les rapports d’entraînement et de compression. Dans la seconde partie, l’influence des gouttelettes a été étudiée d’un point de vue local en étendant le modèle RANS à une phase discrète qui affecte le flux principal par des échanges de quantité de mouvement et d’énergie thermique, et d’un point de vue global en construisant un modèle thermodynamique qui prédit l’entraînement et le rapport de compression limitant étant donné une géométrie fixe et les conditions de fonctionnement. Les deux approches présentent un très bon accord en termes de profils internes de p, T et Ma. Les résultats pour un éjecteur supersonique au R134a comme fluide de base, avec des gouttelettes injectées à mi-chemin dans la section de la zone constante, montrent que la structure d’écoulement dans cette région présente des changements perceptibles seulement à la fraction d’injection la plus élevée, 10%, en diminuant l’intensité du choc de 8% et la surchauffe à la sortie de l’éjecteur de 15 ◦C. Néanmoins, la performance de l’éjecteur est sévèrement affectée vu que le rapport de compression, l’efficacité d’Elbel et le performance exergétique sont réduites respectivement de 5%, 11% et 15%, principalement en raison de l’entropie supplémentaire générée par l’injection de gouttelettes et le mélange avec le flux principal.

A hot-flow axisymmetric Air Augmented Rocket (AAR) test apparatus was constructed to test various mixing duct configurations at static conditions. Primary flow for the AAR was provided through a liquid methanol-gaseous oxygen bipropellant rocket. Experimental thrust measurements were recorded and propellant mass flow rates and chamber conditions were calculated using an iterative solver dependant on recorded propellant line stagnation pressures. Primary rocket flow produced thrust ranging from 14 to 17.9lbf. Primary mass flow rate through testing ranged from 0.071 to 0.085lbm/s with calculated chamber pressures between 298-362psia. Calculated primary flow velocity ranged from 6,600ft/s to 8,000ft/s depending on propellant pressure inputs and calculated chamber conditions. The AAR test apparatus was capable of testing various mixing duct geometries and measuring the axial thrust of the mixing ducts separately from the total thrust of the system. Two mixing duct geometries, a straight wall mixing duct and diverging wall mixing duct, with identical exterior dimensions and inlet geometry were tested for a range of air/fuel mixture ratios from 0.82 to 2.2 spanning the stoichometric mixture ratio of 1.5. Mixing duct thrust did not vary greatly with primary flow characteristics. Straight mixing duct thrust averaged 0.97lbf and diverging mixing duct thrust averaged 0.18lbf. Total system thrust decreased by an average of 0.62lbf with a straight mixing duct and 0.74lbf with a diverging mixing duct. Decreases in total thrust are attributed to low pressure flow interaction between the mixing duct and the primary rocket assembly. Visual flow comparison between mixing duct configurations and fuel ratio cases were carried out using high definition video recording with a grid reference for comparison. The diverging mixing duct produced the greatest variation in visible flow when compared to a straight mixing duct and no mixing duct configuration. This indicated that the diverging mixing duct had a greater influence on primary and secondary flow field mixing than the straight mixing duct.

A 2-D Air Augmented Rocket, the Cal Poly Air Augmented Rocket (CPAAR) Test Apparatus operating as a mixer-ejector was tested to investigate high stagnation pressure ratio and transient flow fields of an ejector. The primary rocket ejector was supplied with high pressure nitrogen at a maximum chamber pressure of 1758 psia and a maximum mass flow rate of 1.4 lb/s. The secondary flow air was entrained from a fixed volume plenum chamber producing pressures as low as 3.3 psia. The maximum total pressure ratio achieved was 221. The original CPAAR apparatus was rebuilt re-instrumented and capability expanded. A fixed volume plenum was attached to the secondary ducts through a constant area square section to mimic the cross section of the secondary ducts with a bell mouth inlet. The mixing duct length was increased from 8 in. to 18 in. An investigation of the mixing duct flow-field was done with data from pressure and temperature instrumentation. A study of the transient operation of the rocket was compared with results from former research to qualify the quasi-steady assumption of the flow-field. The CPAAR produced Fabri-choked operation, the startup transient observed caused the secondary flow to become established during Fabri-choke mode operation. The supersonic saturated mode was not observed during quasi-steady operation. The quasi-steady operation was defined based on characteristics from previous quasi-steady models of transient operation of supersonic ejectors. The measurement of the data during testing resulted in a 2.96% experimental uncertainty in the entrainment ratio calculation. The smallest entrainment ratio observed was 0.05 at a total pressure ratio of 220. The location of the Fabri-choke point was shown through the interpretation of the primary and secondary flow as a result of the pressure and temperature measurements. The experimental evidence showed the location of the secondary choke point has a logarithmic relationship with the total pressure ratio. At a total pressure ratio of 220, the area of the aerodynamic throat of the secondary flow is 0.26 in2 and the location occurs 6 inches downstream from the nozzle exit. The secondary flow un-choke is related to the breakdown of the shock structure of the primary flow and produces a flow-field asymmetry which blocks the right duct flow. The CPSE simulation was unable to accurately predict AAR performance when the inputs are changed from the original CPAAR configuration. At high pressure ratios (PR=220), the error in the prediction is 90%.

Le travail mené dans le cadre de cette thèse porte sur l’étude expérimentale et numérique de l’écoulement au sein d’un éjecteur supersonique. Le régime d’écoulement qui s’installe dans ces appareils est très complexe du fait des phénomènes physiques qui les caractérisent comme la turbulence et les ondes de choc. Les méthodes expérimentales utilisées sont la mesure de la pression le long de l’axe de l’éjecteur `a l’aide d’une sonde développée à cet effet, la visualisation de l’écoulement par tomographie laser et la mesure de vitesse par PIV. Les simulations numériques sont réalisées à l’aide du code Ansys-Fluent en 2D axisymétrique et en 3D. Dans un premier temps, une étude de sensibilité du modèle numérique portant sur les paramètres de simulations et les modèles de turbulence est menée sur l’éjecteur fonctionnant sans flux induit. La validation des simulations repose sur une comparaison des résultats numériques avec des mesures de vitesse par PIV. Un modèle 3D s’est avéré incontournable pour l’étude de l’écoulement dans l’éjecteur avec flux induit à cause de sa géométrie complexe. Les outils expérimentaux et numériques développés permettent d’analyser finement l’interaction des flux moteur et induit, en particulier les processus de recompression par chocs obliques et de mélange. Une tentative de modélisation par LES des instabilités de l’écoulement détectées expérimentalement est également abordée. / The work reported in this thesis relates to the experimental and numerical studies of the flow within a supersonic ejector. The flow pattern which occurs in these apparatuses is very complex because of the flow phenomena encountered like flow turbulence and shock waves. The experimental methods used are the measurement of the pressure along the axis of the ejector using a specific probe developed for this purpose, the flow visualization by laser tomography and the velocity measurement by PIV. The numerical simulations are carried out using the Ansys-Fluent code with 2D axisymmetric and 3D models. First, a study of sensitivity to the numerical parameters of simulation and to the turbulence models is carried out on the ejector operating without induced flow. The validation of the simulations is achieved by a comparison between the numerical results and velocity measurements by PIV. A 3D model is necessary for the simulation of the flow in the ejector operating with induced flow because of the complex ejector geometry. The experimental techniques and the numericalmodels developed make it possible to analyze the interaction of the primary and secondary flows, in particular the process of recompression by oblique shocks and the mixing process. An attempt at modeling by LES simulation the flow instabilities detected during experiments is also approached.