Dans les écoulements de convection naturelle en canal, de nombreuses structures cohérentes se développent. Leur rôle dans les transferts de masse et de chaleurs reste aujourd’hui mal connu. Des études numériques basées sur des Simulations aux Grandes Echelles d’un canal vertical dont un des côtés est chauffé uniformément à flux constant ont donc été entreprises. Différentes étapes de la transition laminaire-turbulente sont identifiées et deux bandes de fréquences sont observées en proche paroi chauffée. Une méthode, basée sur la Proper Orthogonal Decomposition (POD), a aussi été utilisée et a permis une très nette séparation spatio-temporelle des modes les plus énergétiques de l’écoulement en fonction des fréquences précédemment identifiées. Il devient donc possible d’étudier la contribution de ces modes à l’écoulement. Une analyse POD des champs de vitesses expérimentaux permet l’identification de similarités avec les structures observées numériquement.Les variations du gradient de température extérieur au canal, ont été identifiées comme l’une des sources de ces différences et ont fait l’objet d’études expérimentales et numériques. Il est montré qu’un augmentation du gradient positif de température diminue le débit mais aussi déplace la transition plus bas dans le canal. Les études numériques permettent également d’étudier le cas de gradients de température faibles ou négatifs, gradients qui sont difficile à obtenir expérimentalement. Un modèle théorique de l’influence de la stratification extérieure sur le débit massique induit est aussi développé. L’accord entre les résultats prédits et les résultats expérimentaux et numériques est excellent. Ce dernier point souligne que la stratification thermique est un facteur clé dans ce type d’écoulement et que son influence est précisément quantifiée dans ce travail / In natural convective flows, complex coherent structures develop whose the role in heat and mass transfer are not well understood. A numerical study, based on Large-Eddy-Simulations of a vertical channel with one side uniformly was therefore carried out. Different stages of transitional flow development were identified numerically with two characteristic frequency bands being observed in the flow, near the heated wall. Methods derived from the Proper Orthogonal Decomposition (POD) was also used and allows the most energetic modes to be separated accordingly to two characteristic frequency bands found numerically. As result, the contribution of the two families of modes to the near wall turbulent heat transfer and velocity-temperature correlation has been evaluated. POD was also performed on experimental measurements showing similarities with the numerically observed structures.In this work, variations of the external thermal stratification have been identified as one possible source of these differences and its influence was therefore investigated experimentally and numerically. It is shown that the increase in the positive gradient of the external stratification not only decreases the mass flow rate but also displaces the transition height to a lower location in the channel. Numerical simulations also allow the study of cases of weak and negative thermal stratifications which are difficult to achieve in laboratories. A theoretical model of the influence of the external thermal stratification on the mass flow rate was also developed. There is an excellent agreement between the theoretical predictions and the experimentally and numerically obtained mass flow rates. This clearly highlights that external temperature distributions are key driving factors and their influence is accurately quantified in this work

Este trabalho propõe um modelo de transferência de calor e massa para prever a formação de gelo em um container preenchido com legumes congelados. O problema físico é modelado como um meio poroso composto pelo próprio produto e o ar em seu entorno. O regime de convecção natural é assumido dentro do container, o qual promove o transporte de massa. Como uma primeira validação, o modelo é simulado considerando diferentes temperaturas de ar externo, causadas por flutuações da vizinhança. Resultados para quatro ciclos de temperaturas foram comparados, variando separadamente a temperatura média do ar, amplitude e frequência de oscilação. De modo geral, é observado que a temperatura do produto se comporta assim como era esperado e este resultado é diretamente associado à formação de gelo dentro do container. A formação de gelo cresce com uma maior amplitude de oscilação, porém decresce com um aumento na frequência e na temperatura média. Os parâmetros do modelo foram obtidos para dois diferentes produtos: fatias de cenouras congeladas e vagens congeladas, ambos em meio ao ar. As definições de parâmetros são oriundas de revisão bibliográfica, medições experimentais e simulações numéricas. Os parâmetros encontrados para a caracterização desses meios porosos foram similares para ambos os produtos, mesmo eles possuindo diferentes geometrias. A validação experimental foi feita para as fatias de cenoura considerando dois ciclos de temperatura O modelo numérico é capaz de prever o campo de velocidades do ar, as temperaturas do produto e a formação de gelo local. Os resultados foram validados em relação a um grupo independente de resultados numéricos, tal comparação apresentou uma boa concordância. A circulação de ar encontrada é, de fato, devido à convecção natural. O comportamento da temperatura dos produtos simulados concorda com os valores medidos e os valores de temperaturas diferem por menos de 12%. Com respeito à formação de gelo, o modelo é capaz de prevê-la corretamente nas regiões mais suscetíveis a este fenômeno. Porém, a quantidade de gelo formado prevista pelo modelo (1,56 g/semana) é menor do que a experimental (4,67 g/semana), apesar de serem de mesma ordem de magnitude. O efeito de cada parâmetro no modelo é estudado visando detectar maneiras de aprimorar o modelo. Foi encontrado que os parâmetros mais importantes para a formação de gelo total são a difusividade de massa efetiva e o coeficiente de transferência de calor convectivo dentro do container. Ajustando estes parâmetros duas vezes foi possível encontrar resultados melhores com respeito à formação de gelo (3,09 g/semana). / A model of heat and mass transfer is proposed in order to predict frost formation into a closed container filled with frozen vegetables. The physical problem is modeled as a macroporous media composed by the product itself and the surrounding air. Natural convection air flow is assumed into the container, who promotes water mass transport. As a first validation, the model is simulated for several exterior air temperatures, under environmental fluctuations (boundary conditions). Results of four temperature cycles were compared, varying average air temperature, amplitude and frequency of oscillation, one by one. As a general result, it is observed that the product temperature behavior is as expected, and it is directly associated with frost formation into the container. Frost formation increases with large amplitude of oscillation, but decreases with higher frequencies and higher mean temperatures. Model parameters were obtained for two assembling: frozen slices of carrots and air, and frozen extra thin green beans and air. Parameter definition and evaluation combines literature review, measurements and numerical simulation. In general, parameters which characterize these porous media were similar for both products, even though they display different geometries. The experimental validation is performed for carrot slices with two temperature cycles The numerical model is able to predict air velocity field, air and product temperatures, and local frost formation. Results are validated in respect to a set of independent experimental results that shown a good agreement. Air flow circulation is as expected due to natural convection. Product temperature simulated behavior agrees with measurements, and temperature values differ by less than 12%. Respect to frost formation predictions, the model predicts correctly the most susceptible regions to frost formation. However, the quantity of frost formed predicted by the model (1.56 g/ week)is lower than the experimental one (4.67g/week), despite being of the same order of magnitude. The effect of each parameter in the model is study in order to detect how to improve the model. The most important parameters affecting total frost formation are effective mass diffusivity and convective heat coefficient into the storage container. Adjusting these parameters to twice, better results in terms of frost formation could be obtained (3.09 g/ week).

Cette thèse présente les résultats de deux études : la première concerne la convection naturelle turbulente dans une cavité rectangulaire chauffée uniformément par le bas et remplie d’un nanofluide et la seconde concerne l’investigation du transfert de chaleur conjugué dans un échangeur de chaleur à tubes ailetés.L’enceinte de la première étude a un faible rapport d’aspect. Ses parois gauche, droite et supérieure sont maintenues à une température relativement basse. Le fluide de travail est un nanofluide constitué d’eau et de nanoparticules, soient d’alumine (Al2O3), ou de cuivre (Cu) ou d’oxyde de cuivre (CuO). L’influence des paramètres tels que le nombre de Rayleigh (basé sur la hauteur H de la cavité et la densité de flux de chaleur), le type de nanofluide et la fraction volumique des nanoparticules sur la performance de refroidissement est présentée. Les équations de Navier-Stokes et les équations de conservation de la masse et de l'énergie sont résolues pour une géométrie bidimensionnelle par la méthode numérique des volumes finis. L'algorithme SIMPLE est utilisé pour le couplage pression-vitesse. La discrétisation des termes convectifs est faite avec le schéma QUICK. Le modèle de turbulence k-epsilon standard est utilisé. Le maillage du domaine simulé est généré par le code Gambit. Les résultats montrent que pour toutes les valeurs de Ra, le nombre de Nusselt moyen augmente d’une façon linéaire et monotone avec l’augmentation de la concentration des nanoparticules dans le fluide de base. Le flux de chaleur moyen prend des valeurs qui diminuent en fonction de l’ordre suivant : Cu, CuO et Al2O3.La deuxième étude est une investigation numérique des caractéristiques dynamique et thermique d'un échangeur de chaleur. Les calculs supposent un transfert de chaleur et un écoulement en régime permanent. Le nombre de Nusselt et le coefficient de frottement qui caractérisent l'échangeur de chaleur sont déterminés pour différentes valeurs du nombre de Reynolds. L’équation de conservation de l'énergie dans le fluide et l’équation de conduction de la chaleur dans le solide en trois dimensions ont été résolues avec les équations de la conservation de la masse et de la quantité de mouvement afin de déterminer ces caractéristiques. Les deux régimes d'écoulement laminaire et turbulent sont considérés. L'effet de la modélisation de la turbulence a été étudié en utilisant trois modèles différents (Spalart-Allmaras modèle de turbulence à une équation, le modèle k-epsilon ; standard et le modèle RSM). La validation du modèle a été effectuée en comparant les facteurs de frottement, f, et le facteur, j, de Colburn avec les données expérimentales trouvées dans la littérature. Les résultats tracés ont montré un bon accord qualitatif entre les résultats numériques et les données expérimentales. Les résultats montrent également que le plus simple des trois modèles de turbulence testés (à savoir. Spalart-Allmaras) donne les valeurs les plus proches des données expérimentales. / This thesis presents the results of two studies: the first concerns natural turbulent convection in a rectangular cavity heated from the bottom wall and filled with a nanofluid and the second relates to the investigation of conjugate heat transfer in a fin-and-tube heat exchanger.The cavity of the first study is tall and has a heat source embedded on its bottom wall, while its left, right and top walls are maintained at a relatively low temperature. The working fluid is a water based nanofluid having three nanoparticle types: alumina, copper and copper oxide. The influence of pertinent parameters such as the Rayleigh number, the type of nanofluid and solid volume fraction of nanoparticles on the cooling performance is studied. Steady forms of twodimensional Reynolds-Averaged-Navier-Stokes equations and conservation equations of mass and energy, coupled with the Boussinesq approximation, are solved by the volume control based on the discretisation method employing the SIMPLE algorithm for pressure-velocity coupling. Turbulence is modeled by using the standard k-epsilon model. The Rayleigh number, Ra, is varied from 2.49xE09 to 2.49xE11. The volume fractions of nanoparticles where varied too. Stream lines, isotherms, velocity profiles and temperature profiles are presented for various combinations of Ra, the type of nanofluid and solid volume fraction of nanoparticles. The results are reported in the form of average Nusselt number on the heated wall. It is shown that for all values of Ra, the average heat transfer rate from the heat source increases almost linearly and monotonically as the solid volume fraction increases. Finally the average heat transfer rate takes on values that decrease according to the ordering Cu, CuO and Al2O3.In the second study We determined the heat transfer and friction characteristics of a realistic fin-and-tube heat exchanger. The computations assume steady-state heat transfer and fluid flow. Nusselt number and friction factor characteristics of the heat exchanger are presented for various values of Reynolds numbers. The energy conservation and the heat conduction equations in 3 dimensions have been solved in the fluid and the solid respectivelyalong with the mass and momentum conservation equations in order to determine these characteristics. Both laminar and turbulent flow regimes are considered. The effect of turbulence modeling was investigated using three different models (the one equation Spalart-Allmaras turbulence model, the standard k-epsilon; model and the RSM model). The computations allowed the determination of the dynamic and thermal fields. Model validation was carried out by comparing the calculated friction factor f and Colburn j-factor to experimental results found in the literature. The plotted results showed a qualitatively good agreement between numerical results and experimental data. The results obtained also showed that the simplest of the three turbulence models tested(i.e. Spalart-Allmaras) gives the closest values to the experimental data.

Cette thèse présente une étude numérique d’un séchoir solaire indirect à convection naturelle destiné à sécher les produits agricoles (les figues). La première partie analyse un séchoir solaire contenant une chambre de séchage couplée à un absorbeur inversé et une cheminée solaire. Les simulations ont été faites afin de déterminer les champs dynamique et thermique sous l’influence de la variation de la configuration de la cheminée solaire et la taille de l’ouverture d’admission. Les équations de conservations basées sur le modèle de turbulence k-ε standard sont résolues par la méthode des volumes finis à l’aide du code commercial ANSYS-Fluent. La deuxième partie analyse l’effet d’ajouter un stockage thermique sous forme d’un lit en gravier dans la chambre de séchage. Le lit est modélisé comme un milieu poreux. En plus, cette étude propose l’utilisation d’une deuxième entrée d'air dans la chambre de séchage afin d’assurer une distribution thermique plus homogène au niveau des claies et de garantir un séchage plus uniforme. / This thesis presents a numerical study of an indirect natural convection solar dryer for drying agricultural products (Figs). The first part analyzes a solar dryer containing a drying chamber coupled to a reversed absorber and a solar chimney. Simulations were made to determine the dynamic and thermal fields under the influence of the variation of the solar chimney configuration and the size of the inlet opening. The governing equations based on the standard k-ε turbulence model are solved by the finite volume method using the ANSYS-Fluent commercial code. The second part analyzes the effect of adding a thermal storage in the form of a gravel bed in the drying chamber. The bed is modeled as a porous medium. Furthermore, this study proposes the use of a second air inlet in the drying chamber in order to ensure a more homogeneous thermal distribution at the level of the racks and to guarantee a more uniform drying.

This thesis is devoted to the investigation of passivesafety and inherent features of subcritical nucleartransmutation systems - accelerator-driven systems. The generalobjective of this research has been to improve the safetyperformance and avoid elevated coolant temperatures inworst-case scenarios like unprotected loss-of-ow accidents,loss-of-heat-sink accidents, and a combination of both theseaccident initiators. The specific topics covered are emergencydecay heat removal by reactor vessel auxiliary cooling systems,beam shut-off by a melt-rupture disc, safety aspects fromlocating heat-exchangers in the riser of a pool-type reactorsystem, and reduction of pressure resistance in the primarycircuit by employing bypass routes. The initial part of the research was focused on reactorvessel auxiliary cooling systems. It was shown that an 80 MWthPb/Bi-cooled accelerator-driven system of 8 m height and 6 mdiameter vessel can be well cooled in the case of loss-of-owaccidents in which the accelerator proton beam is not switchedoff. After a loss-of-heat-sink accident the proton beam has tobe interrupted within 40 minutes in order to avoid fast creepof the vessel. If a melt-rupture disc is included in the wallof the beam pipe, which breaks at 150 K above the normal coreoutlet temperature, the grace period until the beam has to beshut off is increased to 6 hours. For the same vessel geometry,but an operating power of 250 MWth the structural materials canstill avoid fast creep in case the proton beam is shut offimmediately. If beam shut-off is delayed, additional coolingmethods are needed to increase the heat removal. Investigationswere made on the filling of the gap between the guard and thereactor vessel with liquid metal coolant and using water spraycooling on the guard vessel surface. The second part of the thesis presents examinationsregarding an accelerator-driven system also cooled with Pb/Bibut with heat-exchangers located in the risers of the reactorvessel. For a pool type design, this approach has advantages inthe case of heat-exchanger tube failures, particularly if wateris used as the secondary uid. This is because a leakage ofwater from the secondary circuit into the Pb/Bi-cooled primarycircuit leads to upward sweeping of steam bubbles, which wouldcollect in the gas plenum. In the case of heatexchangers in thedowncomer steam bubbles may be dragged into the ADS core andadd reactivity. Bypass routes are employed to increase the owspeed in loss-of-ow events for this design. It is shown thatthe 200 MWth accelerator-driven system with heat-exchangers inthe riser copes reasonably well with both a loss-of-ow accidentwith the beam on and an unprotected loss-of-heat-sink accident.For a total-loss-of-power (station blackout) and an immediatebeam-stop the core outlet temperature peaks at 680 K. After acombined loss-of-ow and loss-of-heat-sink accident the beamshould be shut off within 4 minutes to avoid exceeding the ASMElevel D of 977 K, and within 8 minutes to avoid fast creep.Assuming the same core inlet temperature, both the reactordesign with heat-exchanger in the risers and the downcomershave similar temperature evolutions after a total-loss-ofpoweraccident. A large accelerator-driven system of 800 MWth with a 17 mtall vessel may eventually become a standard size. For thishigher power ADS, the location of the heat-exchangers hasgreater impact on the natural convection capability. This isdue to that larger heatexchangers have more inuence on thedistance between the thermal centers during a lossof- owaccident. The design with heat-exchangers in the downcomers,the long-term vessel temperature peaks at 996 K during aloss-of-ow accident with the beam on. This does not pose athreat of creep rupture for the vessel. However, the locationof the heat-exchangers in the downcomers will probably requiresecondary coolant other than water, like for example oil (fortemperatures not higher than 673 K) or Pb/Bi coolant.

In this thesis, we consider a projection-based stabilization method for solving buoyancy driven flows (natural convection problems). The method consists of adding global stabilization for all scales and then anti-diffusing these effects on the large scales defined by projections into appropriate function spaces. In this way, stabilization acts only on the small scales. We consider two different variations of buoyancy driven flows based on the projection-based stabilization. First, we focus on the steady-state natural convection problem of heat transport through combined solid and fluid media in a classical enclosure. We present the mathematical analysis of the projection-based method and prove existence, uniqueness and convergence of the approximate solutions of the velocity, temperature and pressure. We also present some numerical tests to support theoretical findings. Second, we consider a system of combined heat and mass transfer in a porous medium due to the natural convection. For the semi-discrete problem, a stability analysis of the projectionbased method and a priori error estimate are given for the Darcy-Brinkman equations in double-diffusive convection. Then we provide numerical assessments and a comparison with some benchmark data for the Darcy-Brinkman equations. In the last part of the thesis, we present a fully discrete scheme with the linear extrapolation of convecting velocity terms for the Darcy-Brinkman equations.

Finned heat sink use for electronics cooling via natural convection is numerically investigated. An experimental study from the literature that is for vertical surfaces is taken as the base case and the experimental setup is numerically modeled using commercial CFD software. The flow and temperature fields are resolved. A scale analysis is applied to produce an order-of-magnitude estimate for maximum convection heat transfer corresponding to the optimum fin spacing. By showing a good agreement of the results with the experimental data, the model is verified. Then the model is used for heat transfer from inclined surfaces. After a large number of simulations for various forward and backward angles between 0-90 degrees, the dependence of heat transfer to the angle and Rayleigh number is investigated. It is observed that the contributions of radiation and natural convection changes with the angle considerably. Results are also verified by comparing them with experimental results available in literature.

<p>This thesis is devoted to the investigation of passivesafety and inherent features of subcritical nucleartransmutation systems - accelerator-driven systems. The generalobjective of this research has been to improve the safetyperformance and avoid elevated coolant temperatures inworst-case scenarios like unprotected loss-of-ow accidents,loss-of-heat-sink accidents, and a combination of both theseaccident initiators. The specific topics covered are emergencydecay heat removal by reactor vessel auxiliary cooling systems,beam shut-off by a melt-rupture disc, safety aspects fromlocating heat-exchangers in the riser of a pool-type reactorsystem, and reduction of pressure resistance in the primarycircuit by employing bypass routes.</p><p>The initial part of the research was focused on reactorvessel auxiliary cooling systems. It was shown that an 80 MWthPb/Bi-cooled accelerator-driven system of 8 m height and 6 mdiameter vessel can be well cooled in the case of loss-of-owaccidents in which the accelerator proton beam is not switchedoff. After a loss-of-heat-sink accident the proton beam has tobe interrupted within 40 minutes in order to avoid fast creepof the vessel. If a melt-rupture disc is included in the wallof the beam pipe, which breaks at 150 K above the normal coreoutlet temperature, the grace period until the beam has to beshut off is increased to 6 hours. For the same vessel geometry,but an operating power of 250 MWth the structural materials canstill avoid fast creep in case the proton beam is shut offimmediately. If beam shut-off is delayed, additional coolingmethods are needed to increase the heat removal. Investigationswere made on the filling of the gap between the guard and thereactor vessel with liquid metal coolant and using water spraycooling on the guard vessel surface.</p><p>The second part of the thesis presents examinationsregarding an accelerator-driven system also cooled with Pb/Bibut with heat-exchangers located in the risers of the reactorvessel. For a pool type design, this approach has advantages inthe case of heat-exchanger tube failures, particularly if wateris used as the secondary uid. This is because a leakage ofwater from the secondary circuit into the Pb/Bi-cooled primarycircuit leads to upward sweeping of steam bubbles, which wouldcollect in the gas plenum. In the case of heatexchangers in thedowncomer steam bubbles may be dragged into the ADS core andadd reactivity. Bypass routes are employed to increase the owspeed in loss-of-ow events for this design. It is shown thatthe 200 MWth accelerator-driven system with heat-exchangers inthe riser copes reasonably well with both a loss-of-ow accidentwith the beam on and an unprotected loss-of-heat-sink accident.For a total-loss-of-power (station blackout) and an immediatebeam-stop the core outlet temperature peaks at 680 K. After acombined loss-of-ow and loss-of-heat-sink accident the beamshould be shut off within 4 minutes to avoid exceeding the ASMElevel D of 977 K, and within 8 minutes to avoid fast creep.Assuming the same core inlet temperature, both the reactordesign with heat-exchanger in the risers and the downcomershave similar temperature evolutions after a total-loss-ofpoweraccident.</p><p>A large accelerator-driven system of 800 MWth with a 17 mtall vessel may eventually become a standard size. For thishigher power ADS, the location of the heat-exchangers hasgreater impact on the natural convection capability. This isdue to that larger heatexchangers have more inuence on thedistance between the thermal centers during a lossof- owaccident. The design with heat-exchangers in the downcomers,the long-term vessel temperature peaks at 996 K during aloss-of-ow accident with the beam on. This does not pose athreat of creep rupture for the vessel. However, the locationof the heat-exchangers in the downcomers will probably requiresecondary coolant other than water, like for example oil (fortemperatures not higher than 673 K) or Pb/Bi coolant.</p>

Natural convective heat transfer from flat plates and short cylinders inclined at an angle to the vertical in laminar and transition flow regions with isothermal or constant heat flux conditions have been numerically and experimentally studied. When the width of the plate is relatively small compared to its height, i.e., the plate is narrow, the heat transfer rate can be considerably greater than that predicted by these two-dimensional flow results. When the narrow plate is inclined to the vertical, pressure changes normal to the plate surface arise and these pressure changes can alter the nature and the magnitude of the edge effects. When two narrow inclined rectangular flat plates of the same size separated vertically or horizontally, the flow interaction between these heated plates can have a significant effect on the heat transfer. When relatively small square and circular cylinders with exposed top surfaces inclined to the vertical are used, the interaction of the flow over the surfaces that make up the cylinder and inclination angle have, in general, a considerable effect on the magnitude of the mean heat transfer rate and on the nature of the flow over the cylinder surfaces. In the present numerical studies it has been assumed that the fluid properties are constant except for the density change with temperature which gives rise to the buoyancy forces, this having been treated using the Boussinesq approach. The numerical solution was obtained by numerically solving the full three-dimensional form of the governing equations, these equations being written in dimensionless form. The solution was obtained using a commercial CFD code, FLUENT. Results were only obtained for a Prandtl number of 0.7; this being approximately the value of air. In the experimental studies, the average heat transfer rates from cylinders were determined by the transient method, which involves heating the model and then measuring its temperature-time variation while it cools. The average heat transfer rates from the flat plates were determined using a steady state method, which basically involved electrically heating the plate. The tests were carried out inside a large enclosure. / Thesis (Ph.D, Mechanical and Materials Engineering) -- Queen's University, 2011-06-27 19:27:45.724

Among technologies capable to produce electricity locally without contributing to GHG releases, building integrated PV systems (BIPV) could be major contributor. However, when exposed to intense solar radiation, the temperature of PV modules increase significantly, leading to a reduction in efficiency so that only about 14% of the incident radiation is converted into electrical energy. The high temperature also decrease the life of the modules, thereby making passive cooling of the PV components through natural convection a desirable and cost-effective means of overcoming both difficulties. An experimental investigation of heat transfer and fluid flow characteristics of natural convection of air in vertical and inclined open-ended heated channels is therefore undertaken so as provide reliable information for the design of BIPV. Two experimental set ups were developed and used during the present investigations; one located at the CETHIL laboratory in Lyon, the F-device and the other located at the University of New South Wales in Sydney, the R-device. Both channels consisted of two wide parallel plates each of which could be subjected to controlled uniform or non-uniform heat fluxes. The investigation has been conducted by analyzing the mean wall temperatures, measured by thermocouples and mean velocity profiles and turbulent quantity distributions of the flow, measured with a PIV system. Flow patterns close to the heated faces were also investigated. The study is particularly focused on the transition region from laminar to turbulent flow. Three different heating geometric arrangements are examined in the modified Rayleigh number range from 3.86 x 105 to 6.22 x 106. The first is a vertical channel with one wall uniformly heated while the other was unheated, the second was a vertical channel in which both walls were non-uniformly heated and the third is an inclined channel uniformly heated from above. In the vertical configurations the width-to-height channel aspect ratio was fixed at 1:15 and in the inclined ones at 1:16. It is shown that the flow is very sensitivity to disturbances emanating from the ambient conditions. Moreover, the propagation of vortical structures and unsteadiness in the flow channel which are necessary to enhance heat transfer, occurred downstream of the mid-channel section at Ra* = 3.5 x 106 for uniformly and asymmetrically heated channels inclined between 60° and 90° to the horizontal. Indeed, these unsteady flow phenomena appears upstream the location of the inflexion point observed in the temperature excess distribution of the heated wall. In the case of non-uniform heating on both sides of the channel, a stronger 'disruption mechanism' exists, which leads to enhanced mixing and increased Reynolds stresses over most of the width of the channel. Empirical correlations of average Nusselt number as a function of modified Rayleigh number were obtained for each configuration.