Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Nuclear Engineering, 1980. / MICROFICHE COPY AVAILABLE IN ARCHIVES AND SCIENCE / Includes bibliographical references. / by John Edward Kelly. / Ph.D.

京都大学 / 0048 / 新制・課程博士 / 博士(エネルギー科学) / 甲第21887号 / エネ博第388号 / 新制||エネ||75(附属図書館) / 京都大学大学院エネルギー科学研究科エネルギー基礎科学専攻 / (主査)教授 齊藤 泰司, 教授 横峯 健彦, 准教授 伊藤 啓 / 学位規則第4条第1項該当 / Doctor of Energy Science / Kyoto University / DFAM

<p>Heat dissipation requirements of next-generation power electronics in electric vehicles, high-performance computing, and radar systems will far exceed the capabilities of conventional heat sink technologies such as single-phase liquid cold plates and air-cooled heat sinks. The leading candidate technology that promises to meet these needs is microchannel flow boiling. Compared to conventional heat sink technologies, flow boiling provides some of the highest heat transfer coefficients available and can dissipate heat at a lower pumping power and with more uniform surface temperatures. However, there are unique challenges associated with flow boiling that currently prevent practical implementation of the technology, including limited modeling capabilities, inherent critical heat flux (CHF) limitations, and the presence of two-phase flow instabilities. This thesis is targeted primarily at addressing the impact of dynamic two-phase flow instabilities on heat transfer and CHF in microchannel heat sinks, in contrast with earlier literature that has focused on prediction and characterization of the flow dynamics.</p> <p><br></p> <p>Two dynamic instabilities of importance in microchannel heat sinks are pressure drop oscillations (PDO) and parallel channel instabilities, both resulting from an interaction between the inertia of a two-phase mixture within a heated channel and a source of compressibility outside of the channel. However, the individual impact of these instabilities on heat transfer performance has not been quantified. In this thesis, an experimental facility is developed to isolate the individual and combined impact of PDOs and parallel channel instabilities on surface temperature and CHF in single- and parallel-microchannels. This is achieved by introducing a measurable compressible volume directly upstream of the test section and isolating the test section from any unwanted compressibility within components throughout the rest of the system. Experiments are first performed targeting the investigation of PDOs in single channels and then targeting PDOs and parallel channel instabilities in multi-channel heat sinks. In the case of parallel channels, inlet restrictors are introduced to suppress channel-to-channel interactions and provide a baseline case of stable boiling. Throughout these experiments, only moderate increases in time-average surface temperature are observed (6 °C) and reduction of CHF is negligible, despite drastically different flow pattern observations when instabilities are present. These observations are in stark contrast with other cases in the literature, for which significant deterioration of surface temperatures and CHF have been attributed to the presence of PDOs. For example, significant temperature oscillations have been observed in the literature studying silicon-etched microchannel heat sinks experiencing PDOs. A predictive model is clearly required to understand and detect the conditions when dynamic instabilities should be considered in heat sink design.</p> <p><br></p> <p>To better understand the conditions when PDOs might have significant impact on heat transfer performance, an investigation of thermal capacitance is performed using a dynamic two-phase model and a targeted experimental approach in heat sinks having different thermal masses. The model reveals that, if thermal capacitance is low, PDOs become more severe, and the amplitude of temperature oscillations increase. These predictions are confirmed by experimental observations, and, in addition, premature CHF is observed in the heat sink with lower thermal mass. With sufficient thermal capacitance, the system recovers before triggering CHF, preventing deterioration of performance due to PDOs. Among the wide range of flow conditions considered in this thesis, the reduction of thermal mass resulted in the greatest impact on transient response of a heat sink during flow boiling instabilities. This reveals thermal capacitance as a critical parameter when determining if dynamic instabilities will deteriorate performance in a microchannel heat sink application. This allows engineers to make an informed judgement on whether adding features to suppress instabilities, at the cost of increased pumping power, is warranted. In order for the practical implementation of two-phase heat sinks to be realized, further development of dynamic modeling capabilities is required, and these models should be backed by systematic experimental investigations into conditions where instabilities should be considered.</p>

No description available.

No description available.

<p>The effect of pressure transients on the onset of intermittent dryout in upward annular flow was experimentally investigated in order to resolve the conflict between the observations drawn from two major data sets in the literature. A delay in time to the onset of dryout at the test section exit relative to the time predicted based on steady-state data was observed in the R-12 experiments of Celata et al (1988; 1991). Steady-state prediction methods were sufficient to predict the upstream progression of a pre-existing dryout front in the water experiments of Lyons and Swinnerton (1983). Steady state and pressure transient dryout experiments were performed using water with outlet pressures of 2 to 6 MPa and mass fluxes of 1000 to 2500 kg/m2/s in an electrically heated 1.32 m long 4.6 mm ID vertical Inconel 600 tube with depressurisation rates of up to 1.0 MPa/s. Transient experiments were performed with a small margin to dryout and with post-dryout initial conditions in order to test the hypothesis that these initial conditions influenced the onset of dryout during transients. The results of a comparison between the steady dryout data and two dryout prediction methods---the Biasi et al (1967) correlation and the 2006 CHF look-up table (Groeneveld et al, 2007)---were used to develop correction factor correlations to reduce systematic error when these methods were used to predict the transient time to dryout. These modified methods yielded mean predicted dryout delays of -0.1 and 1.5 s respectively with standard deviations of approximately 3 s. There was no statistically significant variation between the pre- and post-dryout initial conditions. Based on this result it was concluded that the initial conditions did not affect the observed time to dryout. The mean wall temperature exhibited a discontinuous decrease as the heat flux approached 92 to 95% of the dryout value. It was postulated that this was caused by a heat transfer regime change from liquid film evaporation to droplet evaporation based on the observations of Hewitt (1970), Doroschuk et al (1970) and Groeneveld (2011). For the range of conditions of the present work the onset of intermittent dryout (Groeneveld, 1986) was caused by deterioration of droplet evaporation heat transfer. Celata et al (1988) noted that in their pressure transient experiments the decrease in saturation temperature drove a rapid increase in the heat flux to the fluid. This was caused by the release of stored thermal energy as the test section wall cooled. Celata et al (1991) stated that the systematic dryout delay was observed for depressurisation rates greater than 0.2 MPa/s. Using Celata et al's (1988) pressure transient data it was concluded that the stored thermal energy transient did not influence the onset of intermittent dryout when rho_w c_pw L_w *(dT_sat/dt)<0.3*q''_a.</p> / Doctor of Philosophy (PhD)

Constitutive relationships are developed to describe the water transport characteristics of the gas diffusion layer (GDL) of proton exchange membrane fuel cells (PEMFCs). Additionally, experimental fixtures and procedures for the determination of the constitutive relationships are presented. The water transport relationships are incorporated into analytical models that assess the impact of the water transport relations and that make PEMFC performance predictions. The predicted performance is then compared to experimental results. The new constitutive relationships are significantly different than the currently popular relationships used in PEMFC modeling because they are derived from experiments on actual PEMFC gas diffusion layer materials. In prior work, properties of the GDL materials such as absolute permeability, liquid water relative permeability, porosity, and capillary behavior are often assumed or used as adjustment parameters in PEMFC models to simplify the model or to achieve good fits with polarization data. In this work, the constitutive relations are not assumed but are determined via newly developed experimental techniques. The experimental fixtures and procedures were used to characterize common GDL materials including carbon papers and carbon cloths, and to investigate common treatments applied to these materials such as the bulk application of a hydrophobic polymer within the porous structure. A one-dimensional model is developed to contrast results based on the new constitutive relations with results based on commonly used relationships from the PEMFC literature. The comparison reveals that water transport relationships can have a substantial impact on predicted GDL saturation, and consequently a significant impact on cell performance. The discrepancy in saturation between cases can be nearly an order of magnitude. A two-dimensional model is also presented that includes the impact of the compressed GDL region under the shoulder of a bipolar plate. Results show that the compression due to the bipolar plate shoulder causes a significant increase in liquid saturation, and a significant reduction in oxygen concentration and current density for the paper GDL. In contrast, compression under the shoulder has a minimal impact on the cloth GDL. Experimental inputs to the 2-D model include: absolute permeability, liquid water relative permeability, the slope of the capillary pressure function with saturation, total porosity, GDL thickness, high frequency resistance, and appropriate Tafel parameters. Computational polarization curve results are compared to experimental polarization behavior and good agreement is achieved. / Ph. D.

Thesis (MScEng (Mechanical and Mechatronic Engineering))--University of Stellenbosch, 2011. / ENGLISH ABSTRACT: The feasibility of a closed loop thermosyphon for the Reactor Cavity Cooling System of the Pebble Bed Modular Reactor has been the subject of many research projects. Difficulties identified by previous studies include the hypothetical inaccuracies of heat transfer coefficient correlations available in literature. The aim of the research presented here is to develop inside-pipe heat transfer correlations that are specific to the current design of the RCCS. In order to achieve this, a literature review is performed which identifies reactors which employ closed loop thermosyphons and natural circulation. The literature review also explains the general one-dimensional two-fluid conservation equations that form the basis for numerical modelling of natural circulation loops. The literature review lastly discusses available heat transfer coefficient correlations with the aim of identifying over which ranges and under which circumstances these correlations are considered accurate. The review includes correlations commonly used in natural circulation modelling in the nuclear industry in aims of identifying correlations applicable to the modelling of the proposed RCCS. One of the objectives of this project is to design and build a one-third-height-scale model of the RCCS. Shortcomings of previous experimental models were assessed and, as far as possible, compensated for in the design of the model. Copper piping is used, eliminating material and surface property uncertainties. Several sight glasses are incorporated in the model, allowing for the visual identification of two-phase flow regimes. An orifice plate is used allowing for bidirectional flow measurement. The orifice plate, thermocouples and pipe-in-pipe heat exchangers are calibrated in-situ to minimize experimental error and aid repeatability. Twelve experiments are performed with data logging occurring every ten seconds. The results presented here are limited to selected single and two-phase flow operating mode results. Error analyses and repeatability of experimental measurements for single and two-phase operating modes as well as cooling water mass flow rates are performed, to show repeatability of experimental results. These results are used to mathematically determine the experimental inside-pipe heat transfer coefficients for both the evaporator and condenser sections. Trends in the heat transfer coefficient profiles are identified and the general behaviour of the profiles is thoroughly explained. The RCCS is modelled as a one-dimensional system. Correlations for the friction factor, heat transfer coefficient, void fraction and two-phase frictional multiplier are identified. The theoretical heat transfer coefficients are calculated using the mathematical model and correlations identified in the literature review. Fluid parameters are evaluated using experimentally determined temperatures and mass flow rates. The resulting heat transfer coefficient profiles are compared to experimentally determined profiles, to confirm the hypothesis that existing correlations do not accurately predict the inside-pipe heat transfer coefficients. The experimentally determined coefficients are correlated to 99% confidence intervals. These generated correlations, along with identified and established twophase heat transfer coefficient correlations, are used in a mathematical model to generate theoretical coefficient profiles. These are compared to the experimentally determined coefficients to show prediction accuracy. / AFRIKAANSE OPSOMMING: Die haalbaarheid van ‘n natuurlike sirkulasie geslote lus vir die Reaktor Holte Verkoeling Stelsel (RHVS) van die Korrelbed Modulêre Kern-Reaktor (KMKR) is die onderwerp van talle navorsings projekte. Probleme geïdentifiseer in vorige studies sluit in die hipotetiese onakkuraatheid van hitte-oordrag koëffisiënt korrelasies beskikbaar in literatuur. Die doel van die navorsing aangebied is om binne-pyp hitte-oordrag koëffisiënt korrelasies te ontwikkel spesifiek vir die huidige ontwerp van die RHVS. Ten einde dit te bereik, word ‘n literatuurstudie uitgevoer wat kern-reaktors identifiseer wat gebruik maak van natuurlike sirkulasie lusse. Die literatuurstudie verduidelik ook die algemene een-dimensionele twee-vloeistof behoud vergelykings wat die basis vorm vir numeriese modellering van natuurlike sirkulasie lusse. Die literatuurstudie bespreek laastens beskikbare hitte-oordrag koëffisiënt korrelasies met die doel om te identifiseer vir welke massavloei tempo waardes en onder watter omstandighede hierdie korrelasies as korrek beskou is. Die ontleding sluit korrelasies in wat algemeen gebruik word in die modellering van natuurlike sirkulasie in die kern industrie met die hoop om korrelasies vir gebruik in die modellering van die voorgestelde RHVS te identifiseer. Een van die doelwitte van die projek is om ‘n een-derde-hoogte-skaal model van die RHVS te ontwerp en te bou. Tekortkominge van vorige eksperimentele modelle is geidentifiseer en, so ver as moonlik, voor vergoed in die ontwerp van die model. Koper pype word gebruik wat die onsekerhede van materiaal en opperkvlak eindomme voorkom. Verkseie deursigtige polikarbonaat segmente is ingesluit wat visuele identifikasie van twee-fase vloei regimes toelaat. ‘n Opening plaat word gebruik om voorwaartse en terugwaartse vloeimeting toe te laat. Die opening plaat, termokoppels en hitte uitruilers is gekalibreer in plek om eksperimentele foute te verminder en om herhaalbaarheid te verseker. Twaalf eksperimente word uitgevoer en data word elke tien sekondes aangeteken. Die resultate wat hier aangebied word, is beperk tot geselekteerde enkel- en tweefase vloei meganismes van werking. Fout ontleding en herhaalbaarheid van eksperimentele metings, om die herhaalbaarheid van eksperimentele resultate te toon. Hierdie is gebruik om wiskundig te bepaal wat die eksperimentele binne-pyp hitte-oordrag koëffisiënte is vir beide die verdamper en kondenseerder afdelings. Tendense in die hitte-oordrag koëffisiënt profiele word geïdentifiseer en die algemene gedrag van die profiles is deeglik verduidelik. Die RHVS is gemodelleer as 'n een-dimensionele stelsel. Korrelasies vir die wrywing faktor, hitte-oordrag koëffisiënte, leegte-breuk en twee-fase wrywings vermenigvuldiger word geïdentifiseer. Die teoretiese hitte-oordrag koëffisiënte word bereken deur middle van die wiskundige model en korrelasies wat in literatuur geidentifiseer is. Vloeistof parameters is geëvalueer met eksperimenteel bepaalde temperature en massa-vloei tempos. Die gevolglike hitte-oordrag koëffisiënt profiles is vergelyk met eksperimentele profiele om die hipotese dat die bestaande korrelasies nie die binne-pyp hitte-oordrag koëffisiënte akkuraat voorspel nie, te bevestig. Die eksperimenteel bepaalde koëffisiënte is gekorreleer en die gegenereerde korrelasies, saam met geïdentifiseerde twee-fase hitte-oordrag koëffisiënt korrelasies, word gebruik in 'n wiskundige model om teoretiese koëffisiënt profiele te genereer. Dit word dan vergelyk met die eksperimenteel bepaalde hitteoordrag koëffisiënte om die akkuraatheid van voorspelling te toon. Tekortkominge in die teoretiese en eksperimentele model word geïdentifiseer en aanbevelings gemaak om hulle aan te spreek in die toekoms.

This thesis focuses on fundamental aspects of microfluidic systems and applies relevant findings to innovative designs for advanced particle manipulation applications. Computational Fluid Dynamics (CFD) is adopted for fluid modeling, based on the Finite Volume method. The accuracy of the solutions obtained is confirmed by grid sensitivity analysis and by comparisons with experimental work. Curved microchannel features and the induced Dean flow are studied through a parametric space exploration and simulations. The Lagrange-Euler coupling method – Surface Marker Point methodology – is applied to simulate large-size particles (of comparable size to the channel). Through this simulation approach, all the forces on such particles are directly derived through solving the governing equations and the influence of these particles on the flow is considered in a fully coupled manner. A new approach – the Frozen Flow & Flow Correction Coefficient method – is developed, making trans-relaxation-time simulations possible and improving computational efficiency significantly, for 3D simulations of arbitrary shape and size microparticles in complicated microfluidic channels. Detailed comparisons between simulation results and experiments involving particle sedimentation and particle equilibrium position have been conducted for methodology validation. Mechanisms of hydrodynamic particle manipulation are then studied, including hydrodynamic focusing and separation. It is found that the Tubular Pinch effect, Dean flow and the Radial Pressure Gradient effect interact to yield two distinct particle separation mechanisms. For advanced applications, particle focusing, non-magnetic and magnetic separation for neutrally buoyant particles are proposed, based on newly gained insight on the above-mentioned mechanisms. Appropriate channel designs have been proposed both for particle focusing and size-based particle separation, while the vertical-magnetic-Dean separation scheme is highlighted for magnetic separation. Finally, a new integrated system is proposed, that combines the above novel designs into a device-like ensemble. It promises to offer functionality for biomaterial separation and detection, including different types of cells, antigens and biomarkers.

On s'intéresse dans ce travail à la modélisation d'un écoulement diphasique gaz-liquide co-courant descendant dans les réacteurs à lit fixe de particules, procédé largement utilisé dans le domaine industriel. En raison de la complexité de l'écoulement, induite par les nombreuses configurations multiphasiques pouvant coexister au sein du lit, les modèles développés directement à l'échelle du réacteur sont généralement issus d'approches semi-empiriques, en considérant l'écoulement homogène. Or, il a été observé que des hétérogénéités locales, géométrique et hydrodynamique, telle qu'une mal-distribution de la phase liquide, entrainaient une diminution du taux de réaction et conduisait les modèles existants à surestimer la productivité d'un réacteur. La nécessité de prendre en compte les phénomènes microscopiques dans un modèle macroscopique à l'échelle du réacteur rend l'utilisation d'approches multi-échelles indispensable. L'écoulement étant cependant d'une nature complexe, le changement d'échelle ne peut se faire de façon directe et nécessite donc la mise en place d'outils de modélisation adaptés à une échelle intermédiaire. Dans une première étape, la méthode de simulation numérique directe ``Volume-Of-Fluid'' (VOF) est validée dans le cas d'un film ruisselant dans un tube capillaire. Cette méthode est ensuite utilisée, à l'échelle microscopique, afin de proposer et de valider des relations de fermeture pour un modèle de type ``réseau de pores'' pouvant être utilisé à une échelle intermédiaire, celle du Volume Elémentaire Représentatif. Ce changement d'échelle est tout d'abord effectué dans le cas d'un lit fixe en deux dimensions, c'est-à-dire un empilement de cylindres entre deux plaques. Cette configuration permet la mise en place d'un dispositif expérimental qui, couplé à des simulations VOF 2D à plus grande échelle, valide l'approche de type "réseau de pores" adoptée. Le modèle réseau est ensuite étendu au cas d'un lit fixe réel, c'est-à-dire en trois dimensions, dont la géométrie est obtenue par micro-tomographie. Les lois de comportement locales sont redéfinies à l'aide de simulations numériques directes à l'échelle microscopique. Les résultats provenant de simulations de type « réseaux de pore » sont ensuite confrontés, dans le cas d'une répartition homogène des phases, aux modèles 1D habituellement utilisés pour les écoulements diphasiques en lit fixe. Enfin, une campagne expérimentale est menée afin d'observer, par imagerie scanner, l'étalement d'un jet de liquide sur un empilement de grains. Une comparaison qualitative est ensuite effectuée entre les observations expérimentales et les simulations numériques réseaux dans le cas spécifique de l'étalement d'un jet de liquide / We study in this work the modelling of two-phase cocurrent downflows in fixed bed reactors, a process widely used in industry. Due to the flow complexity, i.e., the presence of different interface configurations and, therefore, different phase interactions, most models have been developed using empirical approaches, with the assumption of a homogeneous flow in the reactor. However, several studies showed that local heterogeneities, geometric and hydrodynamic, such as the liquid distribution, could have a great influence on the flow at the reactor-scale and, therefore, on the reactor performance. Consider the microscopic phenomena in a macroscopic model require the use of multi-scale approaches. However, due to the flow complexity, the upscaling cannot be done directly and requires the development of modelling tools suitable for an intermediate scale. In a first step, the direct numerical method \ Volume-Of-Fluid" (VOF) is validated in the case of a two-phase flow in a capillary tube with the presence of a thin film. Then, this method is used, at a microscopic level to propose and validate closure laws for a pore-network model which will be used to simulate the flow at the intermediate scale. This upscaling approach is first tested in a two-dimensional case,i.e., an array of cylinders between two walls. This configuration allows the set up of an experimental approach, coupled with 2D VOF simulations at the intermediate scale, in order to validate the pore-network approach. The pore-network approach is then extended to a real fixed bed, i.e. in three dimensions, whose geometry is obtained by micro-tomography. Local laws of the pore-network model are redefined using direct numerical simulations at a microscopic scale. Pore-network simulations are then compared, for a homogenous phase distribution, with 1D models typically used for two-phase flow in fixed beds. Finally, an experimental campaign was set up to observe, by imaging scanner, the spreading of a liquid jet on a fixed bed pilot. A qualitative comparison is then performed between experimental observations and pore-network simulations in the specific case of the spreading of a liquid jet