Thèse de doctorat : Science et ingénierie des matériaux : INPL : 2008. / Titre provenant de l'écran-titre.

Thesis (M.S.)--West Virginia University, 2002. / Title from document title page. Document formatted into pages; contains xi, 93 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 92-93).

Thesis (M.S. in Mechanical Engineering)--Naval Postgraduate School, September 2002. / Thesis advisor(s): Knox T. Millsaps, Jr. Includes bibliographical references (p. 48-56). Also available online.

Thesis (Ph. D. in Mechanical Engineering)--Naval Postgraduate School, June 2003. / Dissertation supervisor: Ashok Gopinath. Includes bibliographical references (p. 162-164). Also available online.

The finite element method has been used to calculate the stresses and creep deformations of flanged turbine casing models subjected to internal pressure and bolting forces. The finite element results have been compared with results from photoelastic and lead model turbine casings. An axisymmetric thin shell of revolution ring finite element has been developed to analyse casings subjected to pressure, thermal and creep loads. The thin shell of revolution ring finite element is shown to be extremely powerful and has been used to investigate the shell portions of the turbine casing away from the flange. The three-dimensional isoparametric finite elements have been used for more accurate idealisations of the turbine casing. A thick shell isoparametric finite element has also been developed which can be used with the more common hexahedral isoparametric finite elements. A solution algorithm based on a frontal technique has been developed to solve the large number of linear equations given by the finite element equations. This algorithm, which is fully automatic and uses fast access backing store, has a resolution facility which is used to recalculate subsequent creep solutions assuming that the stiffness of the structure remains constant. The creep algorithms are based on time marching techniques where the creep solutions are found for small time increments, the final solution being the sum of all the incrementa1 solutions. During each time increment the stresses are assumed to remain constant and the change in stress between time increments is kept within a preset ratio. The creep algorithms have been used to predict the creep deformation of simple structures to compare with published results. The agreement between the finite element and lead model creep results is limited. The finite clement programs have been written to be compatible with the PAFEC suite of finite element programs.

published_or_final_version / Dentistry / Doctoral / Doctor of Philosophy

Computational fluid dynamics or CFD isan importanttool thatis used at various stages in the design of highly complex turbomachinery such as compressorand turbine stages that are used in land and air based power generation units. The ability of CFD to predict the performance characteristics of a specific blade design is challenged by the need to use various turbulence models to simulate turbulent flows as well as transition models to simulate laminar to turbulent transition that can be observed in various turbomachinery designs. Moreover, CFD is based on numerically solving highly complex differential equations, which through the use of a grid to discretize the geometry introduces numerical errors. Allthese factors combine to challenge CFD’s role as a predictor of blade performance. It has been generallyfound that CFD in its current state of the art is best used to compare between various design points and not as a pure predictor of performances. In this study the capability of CFD, and turbulence modeling, in turbomachinery based geometry is assessed.Three different blade designs are tested, that include an advanced two-stage turbine blade design, a three stage 2D or cylindrical design and finally a three stage bowed stator and rotor design. Allcases were experimentally tested at the Texas A&Muniversity Turbomachinery Performance and Flow Research Laboratory (TPFL).In all cases CFD provided good insights into fundamental turbomachinery flow physics, showing the expected improvement from using 2D cylindrical blades to 3D bowed blade designs in abating the secondary flow effects which are dominant loss generators.However, comparing experimentally measured performance results to numerically predicted shows a clear deficiency, where the CFD overpredicts performance when compared to experimentallyobtained data, largely underestimating the various loss mechanisms. In a relative sense, CFD as a tool allows the user to calculate the impact a new feature or change can have on a baseline design. CFD will also provide insight into what are the dominant physics that explain why a change can provide an increase or decrease in performance. Additionally,as part of this study, one of the main factors that affect the performance of modern turbomachinery is transition from laminar to turbulent flow.Transition is an influential phenomena especially in high pressure turbines, and is sensitive to factors such asupstream incidentwake frequency and turbulence intensity.A model experimentally developed, is implemented into a CFD solver and compared to various test results showing greater capability in modeling the effects of reduced frequency on the transition point and transitional flow physics. This model is compared to industry standard models showing favorable prediction performance due to its abilityto account for upstream wake effects which most current model are unable to account for.

The necessary clearance at the tip of unshrouded rotors of axial turbines allows fluid to leak from the pressure to the suction side of the blade and produces an important component of loss that is ultimately responsible for approximately 25 % of the total turbine rotor losses. Leakage fluid can pass through the tip clearance gap with either high or low loss generation. It has been customary in turbine design to employ high loss designs since it is only by the creation of loss that the gap mass flow rate can be restricted. The present work, however examined the effect of streamlined tips that have low entropy generation within the tip and high leakage flows. An axial turbine followed by a second stage nozzle (ie one and a half stages) was designed, built and instrumented and used to evaluate performance with particular reference to the understanding of tip clearance effects in a real machine and possible benefits of streamlined low loss rotor tips. A radiused pressure edge was found to improve the performance of a single stage and of a one and a half stage turbine at the selected tip clearances. This was in contrast to previous cascade results where mixing losses reduced the benefits of such tips. Clearance gap flow appears to be similar to other turbine flow where the loss mechanism of separation must be avoided. Loss formation within and downstream of a rotor is more complex than previously realized and does not appear to obey the simple rules used to design for minimum tip clearance loss. For example, approximately 48 % of the tip leakage mass flow within a rotor may be a flat wall-jet rather than a vortex. Second stage nozzle efficiency was significantly higher than first stage nozzle efficiency, and even increased with tip clearance. This was a surprising result since it means that not only was there a reduction in secondary flow loss but also that rotor leakage and rotor secondary flows did not generate significant downstream mixing loss. The manner in which the second nozzle responds to the complex leakage flows presented to it and how it completes the formation of tip clearance loss for various rotor tip clearances was identified. The tangentially averaged relative rotor flow in the tip clearance region differed radically from that found in cascades which was seen to be underturned with a high axial velocity. There was evidence rather of overturning presumably caused by secondary flow. Axial velocity followed an almost normal endwall boundary layer pattern with almost no leakage jet effect. Cascade tip clearance models are therefore not accurate in predicting leakage flows of real rotors. The reduction in second stage nozzle loss was seen to occur near the hub and tip confirming a probable reduction in secondary flow loss. Nozzle exit loss contours showed that the leakage flow suppressed the formation of the classical secondary flow pattern and that a new tip clearance related loss phenomenon existed on the suction surface. The second stage nozzle reduced the hub endwall boundary layer below that of both the first nozzle and that behind the rotor. It also appeared to rectify the secondary and tip clearance flows to the extent that a second stage rotor would experience no greater flow distortion than the first stage rotor would. Radial flow angles behind the second stage nozzle were found to be much smaller than those measured in a previous study of low aspect ratio, untwisted blades. / Thesis (Ph.D.)-University of Natal, Durban, 1993.

On unshrouded axial flow turbine rotors, the tip clearance, required for thermal expansion and manufacturing limitations, allows fluid to leak from the pressure side to the suction side of the blade. This flow across the blade tip causes a large proportion of the overall rotor loss. In this work, the flow was visualized, microscopic static pressures taken and flow field measurements were done in the blade tip region to investigate the complex nature of tip clearance flows. An annular turbine cascade with a rotating outer casing was used to simulate the relative motion at the tip of an axial rotor. It was found that relative motion did not have a significant effect on the basic structure of the micro-flow, even though it reduced the leakage mass flow rate which is important as far as mixing loss formation is concerned. The existence of a narrow, very low pressure depression, caused by the flow remaining attached around the sharp pressure corner edge, was confirmed. The width and pressure of the separation bubble were found to be strongly dependent on gap size but the relationship was not linear. The point at which the separation bubble reattaches was seen to coincide with a slight rise in static pressure. The separation bubble which caused the majority of the internal gap loss, and which was thought to contribute to the mixing loss, was shown to disappear when the pressure corner was given a radius of 2,5 gap widths.A linear cascade was used to evaluate the performance of two blade tip shapes that substantially reduced internal gap loss and to compare them to a standard sharp or flat tip blade. A method whereby linear cascade data was analyzed as if it were a rotor with work transfer, was used to evaluate the performance of the various blade tip geometries. It was found that both modified tips increased the mixing loss due to the extra leakage mass flow rate. The first tip with the radiused pressure corner was seen to have a lower efficiency than the flat tip blade. A second tip that was contoured to shed flow in a radial direction and thus decrease the leakage mass flow rate through the gap was seen to significantly increase the overall efficiency. / Thesis (M.Sc.)-University of Natal, Durban, 1989.

No description available.