京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第18306号 / 工博第3898号 / 新制||工||1598(附属図書館) / 31164 / 京都大学大学院工学研究科化学工学専攻 / (主査)教授 山本 量一, 教授 松坂 修二, 教授 古賀 毅 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23503号 / 工博第4915号 / 新制||工||1768(附属図書館) / 京都大学大学院工学研究科機械理工学専攻 / (主査)教授 黒瀬 良一, 教授 花崎 秀史, 教授 岩井 裕 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

The implementation of a hybrid spectral/finite-element discretization on the unsteady, incompressible, Navier-Stokes equations with a semi-implicit time-stepping method, an explicit treatment of the advective terms, and an implicit treatment of the pressure and viscous terms leads to an algorithm capable of calculating 3D flows over complex 2D geometries. This also results in multiple Fourier mode linear systems which must be solved at every timestep, which naturally leads to two parallelization approaches: Fourier space partitioning, where each processor individually and simultaneously solves a linear system, and physical space partitioning, where all processors collectively solve each linear system, sequentially advancing through Fourier modes. These two parallelization approaches are compared based upon computational cost using multiple solvers: direct sparse LU, smoothed aggregation AMG, and single-level ILUT preconditioned GMRES; and on two supercomputers of different memory architecture(distributed and shared memory). This study revealed Fourier space partitioning outperforms physical space partitioning in all problems analyzed, and scales more efficiently as well. These differences were more dramatic on the distributed memory platform than the shared memory platform. Another study compares the previously mentioned solvers along with one additional solver, pointwise AMG, in Fourier space partitioning without parallelization to better understand computational scaling for problems with large meshes. It was found that the direct sparse LU solver performed well in terms of computational time, scaled linearly, but had very high memory usage which scaled in a super-linear manner. The single-level ILUT preconditioned GMRES solver required the least amount of memory, which also scaled linearly, but only had acceptable performance in terms of computational time for coarse meshes. Both AMG methods scaled linearly in both memory usage and time, and were comparable to the direct sparse LU solver in terms of computational time. The results of these studies are particularly useful for implementation of this algorithm on challenging and complex flows, especially direct numerical and large-eddy simulations. Reducing computational cost allows the analysis and understanding of more flows of practical interest.

No description available.

No description available.

<p>Here we present direct numerical simulations of a shearing box which models the MHD turbulence in astrophysical systems with cylindrical geometries. The purpose of these simulations is to detect the source of the electromotive force - the driver of large scale magnetic field evolution. This electromotive force is responsible for the large scale dynamo action which builds and maintains the magnetic field against dissipation in plasmas. We compare the estimates of the electromotive force from the kinematic approximation of mean field theory - the most prevalent theory for astrophysical dynamos - with a modified version of mean field theory which restricts the electromotive force by the consideration of magnetic helicity conservation. We will show that in general the kinematic approximation overestimates the observed electromotive force for the majority of the simulation, while the term derived from the helicity conservation estimates the electromotive force very well. We will also illustrate the importance of the shear in the fluid to the growth and strength of the resulting large scale magnetic field. Too strong and the small scale dynamo does not grow enough to properly seed a strong large scale dynamo. Too weak, and no large scale magnetic field is observed after the small scale dynamo has saturated. Finally, we will find that in order to maintain the strength of the emerged large scale magnetic dynamo we require a magnetic Prandtl number ($Pr \equiv \nu/\eta$) that is at least an order of magnitude above unity.</p> / Master of Science (MSc)

This work is concerned with stability and receptivity analysis as well as studies on control of the laminar-turbulent transition in boundary-layer flows through direct numerical simulations. Various flow configurations are considered to address flow around straight and swept wings. The aim of this study is to contribute to a better understanding of stability characteristics and different means of transition control of such flows which are of great interest in aeronautical applications. Acoustic receptivity of flow over a finite-thickness flat plate with elliptic leading edge is considered. The objective is to compute receptivity coefficient defined as the relative amplitude of acoustic disturbances and TS wave. The existing results in the literature for this flow case plot a scattered image and are inconclusive. We have approached this problem in both compressible and incompressible frameworks and used high-order numerical methods. Our results have shown that the generally-accepted level of acoustic receptivity coefficient for this flow case is one order of magnitude too high. The continuous increase of computational power has enabled us to perform global stability analysis of three-dimensional boundary layers. A swept flat plate of FSC type boundary layer with surface roughness is considered. The aim is to determine the critical roughness height for which the flow becomes turbulent. Global stability characteristics of this flow have been addressed and sensitivity of such analysis to domain size and numerical parameters have been discussed. The last flow configuration studied here is infinite swept-wing flow. Two numerical set ups are considered which conform to wind-tunnel experiments where passive control of crossflow instabilities is investigated. Robustness of distributed roughness elements in the presence of acoustic waves have been studied. Moreover, ring-type plasma actuators are employed as virtual roughness elements to delay laminar-turbulent transition. / <p>QC 20161124</p>

La compréhension du processus de stabilisation des flammes Diesel constitue un défi majeur en raison de son effet sur les émissions de polluants. En effet, la relation étroite entre la distance de lift-off (distance entre la flamme et l’injecteur) et la production de suie est maintenant bien établie. Cependant, différents mécanismes de stabilisation ont été proposés mais sont toujours sujets à discussion. L'objectif de cette thèse est de fournir une contribution expérimentale et numérique pour identifier les mécanismes de stabilisation majeurs.La combustion d'un spray n-dodécane issu d'un injecteur mono-trou a été étudiée dans une cellule à volume constant en utilisant une combinaison de diagnostics optiques : mesures hautes cadences et simultanées de schlieren, LIF à 355 nm, chimiluminescence haute température ou de chimiluminescence OH *. Des expériences complémentaires sont effectuées au cours desquelles le mélange est allumé entre l’injecteur et le lift-off par plasma induit par laser. L’évolution du lift-off jusqu’à son retour à une position d’équilibre plus en aval est ensuite étudiée pour différentes conditions opératoires. L'analyse de l'évolution du lift-off sans allumage laser révèle deux types principaux de comportement : des sauts brusques en amont et un déplacement plus progressif en aval. Alors que le premier comportement est attribué à des événements d'auto-inflammation, le second est analysé grâce aux résultats obtenus par allumage laser. Il a été constaté que l'emplacement du formaldéhyde avait un impact important sur la vitesse de retour du lift-off.Une simulation numérique directe (DNS en anglais) bidimensionnelle d'une flamme liftée turbulente se développant spatialement dans les mêmes conditions opératoires que les expériences et reproduisant l'évolution temporelle de la distance de lift-off est proposée. Du fait que les expériences montrent que la flamme se stabilise en aval du spray liquide, la DNS ne couvre qu'une région en aval où l’écoulement est réduit à un jet gazeux. La chimie de l’n-dodécane est modélisée à l'aide d'un schéma cinétique (28 espèces transportées) prenant en compte les chemins réactionnels basse et haute température. Comme observé expérimentalement, la stabilisation de la flamme est intermittente : des auto-inflammations se produisent tout d'abord puis se font convecter en aval jusqu'à ce qu'une nouvelle auto-inflammation se produise. Le mécanisme principal de stabilisation est l'auto-inflammation. Toutefois, on observe également à la périphérie du jet diverses topologies de flammes, telles que des flammes triples, qui aident la flamme à se stabiliser en remplissant des réservoirs de gaz brûlés à haute température localisés à la périphérie, ce qui déclenche des auto-inflammations. Toutes ces observations sont résumées dans un modèle conceptuel décrivant la stabilisation de la flamme.Enfin, un modèle prédisant les fluctuations de la distance du lift-off autour de sa valeur moyenne temporelle est proposé. Ce modèle a été développé sur la base d’observations faites dans l’étude expérimentale et numérique : premièrement, le suivi temporel du lift-off a été décomposé en une succession d’auto-inflammations et d’évolutions en aval. Deuxièmement, la période entre deux auto-inflammations et la vitesse d'évolution en aval ont été modélisées à l'aide de corrélations expérimentales disponibles dans la littérature. Troisièmement, le modèle a été adapté afin de prendre en compte l’effet des réservoirs à haute température sur les fluctuations de la flamme. Et enfin, le modèle a été comparé aux données expérimentales, au cours desquelles des variations de la température ambiante, de la concentration en oxygène et de la pression d'injection ont été effectuées. Dès lors que le modèle a montré une bonne correspondance avec les données expérimentales, il peut être utilisé en complément du modèle prédisant la distance du lift-off moyen afin de mieux décrire la stabilisation d’une flamme Diesel. / The understanding of the stabilization process of Diesel spray flames is a key challenge because of its effect on pollutant emissions. In particular, the close relationship between lift-off length and soot production is now well established. However, different stabilization mechanisms have been proposed and are still under debate. The objective of this PhD is to provide an experimental and numerical contribution to the investigation of these governing mechanisms.Combustion of an n-dodecane spray issued from a single-hole nozzle was studied in a constant-volume precombustion vessel using a combination of optical diagnostic techniques. Simultaneous high frame rate schlieren, 355LIF (laser-induced fluorescence) and high-temperature chemiluminescence or OH* chemiluminescence are respectively used to follow the evolution of the gaseous jet envelope, formaldehyde location and lift-off position. Additional experiments are performed where the ignition of the mixture is forced at a location upstream of the natural lift-off position by laser-induced plasma ignition. The analysis of the evolution of the lift off position without laser ignition reveals two main types of behaviors: sudden jumps in the upstream direction and more progressive displacement towards the downstream direction. While the former is attributed to auto-ignition events, the latter is studied through the forced laser ignition results. It is found that the location of formaldehyde greatly impacts the return velocity of the lift-off position.A two-dimensional Direct Numerical Simulation (DNS) of a spatially developing turbulent lifted flame at the same operating conditions than the experiments and reproducing the temporal evolution of the lift-off length is proposed to provide a better understanding of the flame stabilization mechanisms. The DNS only covers a downstream region where the flow can be reduced to a gaseous jet, since experimental observations have shown that the flame stabilized downstream of the liquid spray. N-dodecane chemistry is modeled using a reduced chemical kinetics scheme (28 species transported) accounting for the low- and high temperature reaction pathways. Similar to what has been observed in the experiments, the flame stabilization is intermittent: flame elements first auto-ignite before being convected downstream until another sudden auto-ignition event occurs closer to the fuel injector. The flame topologies, associated to such events, are discussed in detail, using the DNS results, and a conceptual model summarizing the observations made is proposed. Results show that the main flame stabilization mechanism is auto-ignition. However, multiple reaction zone topologies, such as triple flames, are also observed at the jet periphery of the fuel jet helping the flame to stabilize by filling high-temperature burnt gases reservoirs localized at the periphery, which trigger in its turn auto-ignitions.Finally, a model predicting the fluctuations of the lift-off length around its time-averaged value is proposed. This model has been developed based on observations made in the experimental and numerical study: first, the lift-off length time-evolution was decomposed into a succession of auto-ignition events and downstream evolutions. Second, the period between two auto-ignition and the velocity of the downstream evolution was modeled using experimental correlations available in the literature. Third, the model has been adapted to take into account the effect of the high-temperature reservoirs on the flame fluctuations. Last, the model was compared to experimental data, where the ambient temperature, oxygen concentration and injection pressure were varied. Since the model showed good agreement with the experimental data, it can be used in addition to the model predicting the time-averaged lift-off length to better describe the Diesel flame stabilization.

Le ballottement dans les réservoirs d'ergols est une des perturbations les plus importantes de la stabilité d'un satellite en orbite. En considérant des manœuvres faiblement inertielles, il n'existe pas de modèle analytique et l'expérimentation nécessite de longues périodes de temps en micro-gravité. Nous nous proposons donc, dans cette thèse, de réaliser des simulations numériques de ce phénomène. L'étude est basée sur le solveur DIVA résolvant les équations de Navier-Stokes diphasiques avec les méthodes level-set et Ghost Fluid. Deux technologies de réservoirs sont à l'étude : les réservoirs classiques, ne contenant que l'ergol liquide et le gaz pressurisant, et les réservoirs à membrane, pour lesquels une membrane hyperélastique sépare les deux fluides. Dans le premier cas, une étude paramétrique complète sur les effets du ballottement lors d'une manœuvre de rotation est menée et les différents régimes d'écoulement obtenus sont décrits par rapport aux paramètres d'études. Par la suite, les données de l'expérience FLUIDICS, envoyée à bord de la Station Spatiale Internationale, sont comparées aux résultats numériques et montrent un très bon accord. Par rapport au second cas, un modèle d'interaction fluides-membrane est développé en s'inspirant de travaux sur la déformation de cellules biologiques. Les déformations et contraintes propres à la membrane sont suivies de façon Eulérienne, les efforts exercés par la membrane sur les fluides environnants en sont déduits et intégrés au solveur diphasique. Les résultats obtenus sont validés par comparaison à des cas-tests de la littérature. / Propellant sloshing in tanks is one of the most important disturbances of satellite stability in orbit. Considering low-inertial manoeuvres, there is no analytical model and experimental facilities require long time period of microgravity conditions. Thus, this PhD thesis aims to predict this phenomenon by numerical simulations. The study is based on the DIVA code which solves the Navier-Stokes equations for two-phase flows with the level-set method and the Ghost Fluid method. Two tank technologies are studied: simple tanks, which only contain the liquid propellant and the gas maintaining the pressure, and diaphragm tanks, for which a hyperelastic membrane separates both fluids. In the first case, a parametric study on the sloshing effects is done considering rotational manoeuvres and the different behaviours observed are described in relation to the study parameters. Thereafter, the data from the FLUIDICS experiment, sent to the International Space Station, are compared to the numerical results and exhibit good agreement. In the second case, a fluids-membrane interaction model inspired from works on the deformation of biological cells is developed. The membrane strains and stresses are computed in an Eulerian way, from which the force exerted by the membrane on the surrounding fluids is deduced and integrated in the two-phase flows solver. The numerical results are validated by comparison with benchmarks from the literature.

A numerical study of fundamental aspects of turbulent mixing has been performed,with emphasis on the behavior of passive scalars of low molecular diffusivity (high Schmidt number Sc). Direct Numerical Simulation is used to simulate incompressible, stationary and isotropic turbulence carried out at high grid resolution. Data analyses are carried out by separate parallel codes using up to 1024^3 grid points for Taylor-scale Reynolds number (R_lambda) up to 390 and Sc up to 1024.Schmidt number of order 1000 is simulated using a double-precision parallel code in a turbulent flow at a low Reynolds number of R_lambda 8 to reduce computational cost to achievable level. The results on the scalar spectrum at high Schmidt numbers appear to have a k^{-1} scaling range. In the presence of a uniform mean scalar gradient, statistics of scalar gradients are observed to deviate substantially from Kolmogorov's hypothesis of local isotropy, with a skewness factor remaining at order unity as the Reynolds number increases. However, this skewness decreases with Schmidt number suggesting that local isotropy for scalars at high Schmidt number is a better approximation. Intermittency exponents manifested by three types of two-point statistics of energy and scalar dissipation, i.e., the two-point correlator (chi(x)chi(x+r)), the second-order moment of local scalar dissipation (chi_r^2) and the variance of the logarithmic local scalar dissipation sigma^2_{lnchi_r} are discussed. Several basic issues in differential diffusion between two scalars of different molecular diffusivities transported by the same turbule nt flow, the physical process of scalar spectral transfer and subgrid-scale transfer are also briefly addressed.