Improvements upon current gas turbine sealing technology performance are essential for decreasing specific fuel consumption to meet stringent future efficiency targets. The clearances between rotating and static components of a gas turbine, which need to be sealed, vary over a flight cycle. Hence, a seal which can passively maintain an optimum clearance, whilst preventing contact between itself and the rotor, is extremely desirable. Various configurations of a Rolls Royce (RR) seal concept, the Large Axial Movement Face Seal (LAMFS), use static pressure forces to locate face seals. Prototypes were tested experimentally at the Osney Thermofluids Laboratory, Oxford. Firstly a proof-of concept rig (simulating a 2-D seal cross-section) manufactured by RR was re-commissioned. The simplest configuration using parallel seal faces induced an unstable seal housing behaviour. The author used this result, CFD, and analytical methods to improve the design and provide a self-centring ability. A fully annular test rig of this new seal concept was then manufactured to simulate a 3D engine representative seal. The full annulus eliminated leakage paths unavoidable in the simpler rig. A parametric program of experiments was designed to identify geometries and conditions which favoured best-practice design. The new seal design is in the process of being patented by Rolls Royce. A 'fluidic' seal was also investigated, showing very promising results. A test rig was manufactured so that a row of jets could be directed across a leakage cross-flow. An experimental program identified parameters which could achieve a combined lower leakage mass flow rate compared with the original leakage. Influence of jet spanwise spacing, injection angle, jet to mainstream pressure ratio, mainstream pressure difference and channel height were analysed. It is hoped this thesis can be used as a tool to further improve these seal concepts from the parametric trends which were identified experimentally.

Although pre-swirlers play a determinant role in the transport of air from stationary parts to rotating holes, knowledge about their actual performance is limited. Therefore, this paper aims to relate how the pre-swirler pressure drop affects the performance of different pre-swirlers in terms of discharge coefficient, adiabatic pre-swirl effectiveness, and swirl ratio. The results are extracted from numerical simulations carried out on three different designs, two guide vanes, and a nozzle. When available, the results are compared to experimental data. The guide vanes have shown similar responses to the pressure drop variations. Their discharge coefficients remain relatively insensitive with an average value of 97%. The swirl ratio range from 0.704 to 1.013 and 0.703 to 1.023 respectively for a pressure drop varying from 3 to 7 bars. The adiabatic pre-swirl effectiveness is of 96% and 94%, respectively, under steady state operation.The nozzle design has shown inferior performance as compared to the guide vane designs. Its discharge coefficient remains around 91% and the swirl ratio varies between 0.678 and 1.121 for a pressure drop ranging from 3 to 10 bars. Under steady state operation, the adiabatic pre-swirl effectiveness is 1.22. The influence of through-flows on the aforementioned parameters was also analyzed. It was observed that the through-flow deteriorates the performance of the pre-swirlers, whether in terms of dimensionless pre-swirl effectiveness, or swirl ratio. The discharge coefficient was however not affected.

There is a constant struggle to increase the efficiency in gas turbines, where one method is to have a higher inlet temperature to the turbine. Often, this results in temperatures higher than the critical temperature of the materials, which makes cooling of the components an important part of the turbine. The cooling air is tapped from the compressor, and has hence required work while being compressed, but since it is removed from the thermodynamic cycle it will not provide any work in the turbine stages. Therefore, it is important to understand the losses in the cooling system to be able to use the smallest amount of cooling air possible, while still cool sufficiently to not decrease the turbine’s lifetime. The pressure losses in the cooling and secondary air systems are due to either friction or minor losses; contractions, expansions and bends. The losses can be described by a discharge coefficient, ; a rate of how close the actual mass flow is to the ideal mass flow, or a pressure loss coefficient, ; a rate of the pressure drop. In the cooling and secondary air systems there are orifices and cooling geometries. These can have different geometrical properties depending on application, and thereby have different heat transfer performances and causing a higher or lower pressure drop. At Siemens Industrial Turbomachinery AB, SIT AB, a one-dimensional in-house program named C3D is used for thermal calculations and calculations of flow properties of internal cooling flow networks. The program uses hydraulic networks consisting of nodes and branches to simulate the flow inside the components. Correlations used for describing pressure losses have been collected and divided depending on their valid ranges, with the aim to make pressure loss calculations easier. A MATLAB code have been developed, which, depending on input parameters, separates the correlations and returns a plot with the correlations that can be used. In order to make the code as useful as possible, a few assumptions were made; curve fitting of correlations which were only available as plots and interpolation to get larger valid ranges for some cases. These assumptions will influence the results, but the code will still be able to give an indication of which correlation to use, and hence, the objective is fulfilled. Simulations in one dimension are commonly used, since it is less time consuming than three-dimensional modelling. Therefore, with focus on the pressure losses, a one-dimensional model of a blade in the in-house program C3D has been evaluated using a three-dimensional model in the CFD program Ansys CFX. Also, two new models were created in C3D; both with geometrical properties and pressure loss coefficients adjusted to the CFX model, but the first model is using the same hydraulic network as in the evaluated, reference, model while the second is using a new network, built according to the streamlines in CFX. The resulting mass flows in the C3D models were compared to the mass flows in the CFX model, which ended in the conclusion that it is hard for the one-dimensional models to understand the complex, three-dimensional flow situations, even when adjusting them to the CFX model. Anyhow, the adjustments made the model somewhat closer to the three-dimensional case, and hence CFX should be used in an earlier stage when developing C3D models.

Advanced rotating shaft seals have the potential to significantly increase the efficiency and performance of steam and gas turbines. Two such seals, brush and leaf seals, rely on the use of thousands of flexible filaments to close clearances between rotating components and their static casings. The current life of the components is poor compared to the rest of the gas turbine, limiting the seals' deployment, particularly in the jet engine at high temperature and pressure. Poor understanding of the seal installation response to frictional heat generated at the point of filament-rotor contact during operation has limited the ability to predict engine closures and hence seal behaviour and life. The resulting temperature rises may compromise the mechanical integrity of the engine rotor in extremis leading to a shaft failure. This thesis considers the heat transfer mechanisms that govern frictional heating, of both the fluid and solid components in the vicinity of such seals, characterising the process both experimentally and using numerical models. Through the identification of key features of the heat transfer a simple numerical methodology is shown to predict the thermal behaviour of the seal installation sufficiently accurately for engine design purposes. A low order heat transfer model, using a simple electrical analogy for heat transfer is used to investigate frictional heat generation. When contact occurs between the rotor surface and the seal filaments, mechanical energy is dissipated as heat at the interface. This is conducted into the rotor and the seal filaments in proportions that depend on the heat transfer characteristics of both contacting bodies (thermal resistances). To calculate the heat partition ratio and the resulting contact temperature, the thermal resistances of both rotor and seal need to be known. To that end, a new test facility, the Seal Static Thermal Test Facility (SSTTF), is developed. This is first used to study the convective heat transfer occurring in the vicinity of the seal; heat transfer coefficients based on appropriate, scalable, gas reference temperatures are reported. Importantly the results show a larger area on the rotor surface affected by the presence of the seal than was assumed by previous workers. The test rig is further modified to generate heating in a static test rig equivalent to the frictional heating at the filament tips. The test rig allows the contact temperature between rotor and seal, a critical previously unknown parameter to be measured in a well-conditioned environment. The presence of many thousands of vanishingly small flow passages in filament seals makes their explicit modelling unfeasible for engine design purposes. Thus the results from the experimental campaign are used to develop a simple computational fluid dynamic model of the seal, including empirically derived frictional heating, and seal porosity models, to achieve similar leakage and surface heat transfer to the rotor as was seen in the static experiments. The low order CFD methodology presented in the thesis is finally employed to model the transient operation of a brush seal under engine representative rotor surface speeds and differential pressures. Experimental data were generated in the Oxford Engine Seal Test Facility for a typical brush seal rubbing against a high growth rotor. These experiments were modelled using CFD and finite element analysis using parameters derived from static tests for the porous modelling of the seal leakage. Comparison of results shows that, without further tuning, the thermal behaviour is captured well with a moderate conservative overestimation of rotor heating with increased differential pressure across the seal allowing the strategy to be used as an engine design tool.

Une méthode non intrusive de construction d’un modèle de remplacement de l’écoulement dans une cavité de système d’air secondaire de turboréacteur est recherchée. Le modèle réduit doit pouvoir être intégré dans un modèle de l’ensemble du moteur et couplé à la thermique de la structure pour simuler son comportement thermique sur une mission complète sous aile. Pour cela, il doit prendre un grand nombre de paramètres en entrée, retourner autant de sorties et être utilisable sur des intervalles de variations étendus de ces paramètres. Plusieurs approches sont envisagées et implémentées, puis appliquées à la modélisation d’une cavité sous turbine fictive :— Création de surfaces de réponse des termes de la décomposition ANOVA des flux pariétaux.— Création de surfaces de réponse des flux pariétaux combinée avec une méthode de raffinement adaptative exploitant la trajectoire dans l’espace d’entrée issue du couplage modèle réduit - modèle de structure.— Réduction de dimension des champs d’interface échangés à partir de résultats des itérations du couplage des modèles thermiques de l’écoulement et de la structure, puis création de surfaces de réponse des coordonnées réduites.Cette dernière voie permets d'obtenir des résultats encourageants sur le cas test proposé d'abord dans le cas à conditions limites d'entrée de l'écoulement fixées, puis en incluant des variations de certaines d'entre-elles. / A non intrusive method to create surrogate models describing the flow in jet engines’ secondary air system is desired. The resulting model must be integrated in a thermal model describing the whole engine during a complete mission under the wing. This requires the model to use a high number of input and output parameters and to be valid on a broad domain of variation of its parameters. Several approches are explored in this thesis and applied to a simplified turbine cavity :— Surrogate modeling of terms of the ANOVA decomposition of wall fluxes.— Surrogate modeling of wall fluxes combined with an adaptive refinement method exploiting the trajectory followed by the input parameters during the coupling between the metamodel and the structural model.— Dimensionality reduction of the interface data exchanged during the coupling between flow and structure thermal model and surrogate modeling of the resulting reduced coordinate.This last approach leads to good results on the test case considered in this thesis with fixed inlet boundary conditions and then with variations of some of the inlet parameters.