In this work, the lattice Boltzmann method (LBM) was validated for direct numerical simulation (DNS) of wall-bounded turbulent flows. The LBM is a discrete-particle-based method that numerically solves the Boltzmann equation as opposed to conventional DNS methods that are based on the Navier-Stokes (NS) equations. The advantages of the LBM are its simple implementation, its ability to handle complex geometries, and its scalability on modern high-performance computers. An LBM code was developed and used to simulate fully-developed turbulent channel flow. In order to validate the results, the turbulence statistics were compared to those calculated from a conventional NS-based finite difference (FD) simulation. In the present study, special care was taken to make sure the computational domains for LBM and FD simulations were the same. Similar validation studies in the literature have used LBM simulations with smaller computational domains in order to reduce the computational cost. However, reducing the size of the computational domain affects the turbulence statistics and confounds the results of the validation. The turbulence statistics calculated from the LBM and FD simulations were found to agree qualitatively; however, there were several significant deviations, particularly in the variance profiles. The largest discrepancy was in the variance of the pressure fluctuations, which differed by approximately 7%. Given that both the LBM and FD simulations resolved the full range of turbulent scales and no models were used, this error was deemed to be significant. The cause of the discrepancy in the pressure variance was found to be the compressibility of the LBM. The LBM allows the density to vary, while the FD method does not since it solves the incompressible form of the NS equations. The effect of the compressibility could be reduced by lowering the Mach number, but this would come at the cost of significantly increasing the computational cost. Therefore, the conclusion of this work is that, while the LBM is capable of producing accurate solutions for incompressible turbulent flows, it is significantly more expensive than conventional methods for simple wall-bounded turbulent flows. / Thesis (Ph.D, Mechanical and Materials Engineering) -- Queen's University, 2011-09-15 23:24:09.968

An immersed boundary spectral method is used to simulate laminar boundary layer flow over a periodic array of cylindrical surface roughness elements. Direct comparisons are made with experiments by using a roughness-based Reynolds number Re[subscript k] of 216 and a diameter to spanwise spacing ratio d/[lamda] of 1/3. Near-field differences between three similar studies are presented and addressed. The shear layer developed over the roughness element produces the downstream velocity deficit region while splitting of the vortex sheet shed the trailing edge forms its lateral modes. Additional geometrical configurations are simulated for comparisons with experimental results and future analysis by linear stability theory. Total disturbance energy E[subscript rms] is fairly consistent with experimental results while spanwise energy components vary significantly. Physical relaxation of the disturbance wake is found to remain a prominent issue for this simulation technique. / text

Flows over hydraulically smooth walls are predominant in turbulence studies whereas real surfaces in engineering applications are often rough. This is important because turbulent flows close to the two types of surface can exhibit large differences. Unfortunately, neither experimental studies nor theoretical studies based on conventional computational fluid dynamics (CFD) can give sufficiently accurate, detailed information about unsteady turbulent flow behaviour close to solid surfaces, even for smooth wall cases. In this thesis, therefore, use is made of a state of the art computational method “Direct Numerical Simulation (DNS)” to investigate the unsteady flows. An “in-house” DNS computer code is developed for the study reported in this thesis. Spatial discretization in the code is achieved using a second order, finite difference method. The semi-implicit (Runge-Kutta & Crank-Nicholson) time advancement is incorporated into the fractional-step method. A Fast Fourier Transform solver is used for solving the Poisson equation. An efficient immersed Boundary Method (IBM) is used for treating the roughness. The code is parallelized using a Message Passing Interface (MPI) and it is adopted for use on a distributed-memory computer cluster at University of Aberdeen as well as for use at the UK’s national high-performance computing service, HECToR. As one of the first DNS of accelerating/decelerating flows over smooth and rough walls, the study has produced detailed new information on turbulence behaviours which can be used for turbulence model development and validations. The detailed data have enabled better understanding of the flow physics to be developed. The results revealed strong non-equilibrium and anisotropic behaviours of turbulence dynamics in such flows. The preliminary results on the rough wall flow show the response of turbulence in the core and wall regions, and the relationship between the axial and the other components are significantly different from those in smooth wall flows.

The dispersion and mixing of passive scalars released from two concentrated sources into open-channel and homogeneous isotropic turbulent flows are studied using direct numerical simulation (DNS). The simulations are conducted using two fully-parallelized in-house codes developed using the FORTRAN 90/95 programming language. A comparative study has been conducted to investigate the effects of the source separation distance, Reynolds number, relative length scales of the plume and turbulent flow, and source elevation on the dispersion and mixing of two plumes. For both flow configurations, four distinct stages in the downwind development of the cross correlation between the fluctuating concentration fields have been identified which feature zero, destructive and constructive interferences and a complete mixing state. Differences between the exceedance probability of concentrations for the single and total plumes are highlighted and analyzed, and the effects of destructive and constructive interference on the exceedance probabilities for the total plume are used to explain these differences. It is found that the relationship between the third- and fourth-order concentration moments and the second-order concentration moment can be well predicted using a clipped-gamma model. This leads to an interesting conclusion that all the higher-order (third-order and above) moments of the total concentration can be inferred from a knowledge of only the first- and second-order concentration moments of each single plume and of the cross correlation coefficient. From a spectral analysis, it is observed that there exists a range of `leading scales' at which the rate of turbulent mixing of the two plumes becomes the most efficient and the coherency spectrum of the plumes approaches the asymptotic value of unity quicker than at any other scales. / October 2016

No description available.

In the present thesis, the sensitivity of flame propagation to the turbulent motion of burning gases is investigated. The long-standing issue of the 'bending effect' is focused upon, which refers to the experimentally-observed inhibition of flame propagation velocity at high intensities of turbulence. Plausible mechanisms for the bending effect are investigated by isolating systematically the effects of turbulence intensity. By providing a novel perspective on this topic, the thesis addresses the fundamental limits of turbulent burning. The investigation employs Direct Numerical Simulation (DNS), which enables the basic conditions of burning to be controlled directly. A parametric DNS dataset is designed and generated by increasing turbulence intensity over five separate simulations. Effects of turbulent motion are isolated in this manner, such that the bending effect is reproduced in the variation of flame propagation velocity recorded. Subsequently, the validity of Damköhler's hypotheses is investigated to ascertain the mechanism of bending. Analysis of the DNS dataset highlights the significance of kinematic flame response in determining turbulent flame propagation. Damköhler's first hypothesis is found to be valid throughout the dataset, suggesting that the bending effect may be a consequence of self-regulation of the flame surface. This contradicts the dominant belief that bending occurs as a result of flame surface disruption by the action of turbulence. Damköhler's second hypothesis is found to be valid in a relatively limited regime within the dataset, its validity governed by flame-induced effects on the prescribed turbulent flow field. Therefore, this thesis presents turbulent flame propagation and the bending effect as emergent from the dynamics of a flame surface that retains its internal thermo-chemical structure. Finally, experimental validation is sought for the proposed mechanisms of bending. Comparisons have been initiated with measurements in the Leeds explosion vessel, based on which the widely accepted mechanism of bending was hypothesized twenty-five years ago. Modifications to the DNS framework warranted by this comparison have aided the development of novel computationally-efficient algorithms. The ongoing work may yield insights into the key mechanism of the bending effect in turbulent flame propagation.

La propagation des incendies à l’échelle de locaux et de villes est un enjeu majeur. Elle est notamment conditionnée par l’inflammation des matériaux dans les locaux attenants au sinistre. Cette dernière résulte de l’allumage du mélange gazeux combustible issu de la décomposition thermique de la phase condensée.Deux types d’inflammation sont définis dans la littérature : l’inflammation pilotée par la présence d’une source d’allumage, et l’auto-inflammation, résultant de l’emballement de la réaction dans la phase gazeuse. L’auto-inflammation joue un rôle majeur dans le contexte d’une propagation de local à local. Toutefois, ce processus n’a été que très peu étudié expérimentalement du fait de sa complexité et seules des analyses théoriques sont aujourd’hui disponibles concernant les phénomènes en jeu.L’enjeu de la présente étude est de caractériser les régimes d’autoallumage en fonction de différentes typologies de solide (comportement thermique et cinétique), afin de mieux comprendre leurs processus et leurs conditions d’occurrence. Cette compréhension fine permet alors de développer des modèles plus globaux de propagation pour une considération déterministe du risque incendie à l’échelle urbaine.Le caractère bref et local de l’auto-inflammation impose le choix d’une méthode de résolution complète des écoulements, des transferts et de la chimie. La Simulation Numérique Directe (DNS) a donc été sélectionnée afin de capter ces phénomènes, avec l’introduction d’une cinétique fine et non infiniment rapide de la décomposition thermique et de la combustion. / Fire propagation on the scale of buildings and cities is a major stake. It is conditioned by the ignition of solid fuels in rooms adjacent to the one where the disaster originally takes place. The ignition is so piloted by the initiation of the combustion reaction of the gaseous mixture stemming from the thermal decomposition of the condensed phase induced by heat transfer.Two types of ignition are defined in the literature: piloted-ignition related to the presence of a hot spot and auto-ignition resulting from the thermal runaway within the gas phase. The auto-ignition plays a major role in the context of fire spread between rooms. However, this process has been very little experimentally studied, because of its complexity, and only theoretical analyses were lead concerning the phenomena which take place during solid fuels auto ignition.The aim of the present study is to characterize auto-ignition regimes according to various solid typologies (regarding to thermal and kinetic behaviour) in order to understand better their processes and their occurrence conditions. Thereby, this fine understanding allows to develop global models of fire spread for a deterministic consideration of the fire hazards at urban scale.The brief and local character of the auto-ignition requires the choice of a complete resolution for flows, transfers and chemistry. Thus, the Direct Numerical Simulation (DNS) was selected to capture the phenomena, with the introduction of a fine and non-infinitely fast chemistry of thermal decomposition and combustion.

Kyoto University (京都大学) / 0048 / 新制・課程博士 / 博士(工学) / 甲第14589号 / 工博第3057号 / 新制||工||1455(附属図書館) / 26941 / UT51-2009-D301 / 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 宮原 稔, 教授 大嶋 正裕 / 学位規則第4条第1項該当

Controlling ignition timing and flame stabilization is one of the most outstanding challenges limiting the development of modern, efficient and low-emission compression ignition engines (CIEs). In this study, the role of turbulence on two-stage ignition dynamics and subsequent flame stabilization at diesel engine conditions is assessed by performing direct numerical simulations in a simplified inflow-outflow premixed configuration. The thermochemical conditions are chosen to match those of the most reactive mixture in the Engine Combustion Network’s n-dodecane Spray A flame (temperature of 813 K, pressure of 60 atm, equivalence ratio of 1.3, and with 15% vol. O2 in the ambient gas). Inflow velocities 4 to 16 times larger than the laminar flame speed are considered. As a result, in the absence of turbulence, ignition and flame stabilization are controlled by advection and chemistry, diffusion being negligible. Ignition delays match those of the homogeneous reactor and both the cool flame, due to low-temperature chemistry (LTC), and the hot flame, due to high-temperature chemistry (HTC), are spontaneous ignition fronts. Turbulence alters this picture in two ways. First, the second-stage (HTC) ignition delay is increased considerably, in contrast with the first-stage (LTC) ignition delay, which remains virtually unaffected. Second, a sufficiently high turbulence intensity makes the cool spontaneous ignition front transition to a cool deflagration which moves upstream to the inlet, while the hot flame is pushed downstream, still stabilized by spontaneous ignition. The latter phenomenon is caused by the reduced reactivity of LTC products as the cool flame transitions from spontaneous ignition to deflagration. Further increasing the turbulence intensity leads to both cool and hot flames transitioning to deflagrations. For the hot flame, the mechanism governing this transition is the increase in magnitude of progress variable gradient under increased turbulence or reduced inflow velocity, while in cool flames it is mainly due to the reduction in chemical source terms. In addition to turbulence intensity, the role of inflow velocity, integral length scale, and oxygen concentration level on this transition is assessed and modeling challenges are discussed. Finally, a chemical explosive mode analysis is provided to further characterise the ignition and transition phenomena. The present results highlight important fundamental roles of turbulence expected to modulate CIE combustion dynamics.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第19240号 / 工博第4075号 / 新制||工||1628(附属図書館) / 32239 / 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 宮原 稔, 教授 松坂 修二 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM