Search

Although turbulence is known for its ability to efficiently disperse matter, inertial particles carried by a turbulent flow do not explore all its regions homogeneously. In this thesis, two experimental studies have thus addressed the problem of preferential exploration of particles denser than the working fluid.In a first experimental device, the spatial distribution of polyamide or polyacetal beads whose sizes are of the order of the integral length scale of the flow, suspended in a turbulent swirling flow at high Reynolds number, was studied. As these particles are denser than the carrier fluid, their dynamics is driven by the competition between turbulent agitation and gravitational effect. When the rotation frequency of the disc that generates the flow is low, the settling of the beads dominates their resuspension by the turbulence so that they remain confined at the bottom of the tank. Above a certain rotation frequency, turbulent fluctuations overcome the effect of gravity and the particles start to explore the entire vessel. However, two extremely different behaviours emerge depending on the size of the beads. While the smaller ones are almost homogeneously distributed, the larger ones are preferentially found in the upper part, as if gravity was reversed. This trapping is related to the topology of the mean flow and the fact that large particles cannot reach the downward regions of the flow located in the corners due to their size and therefore only rarely fall back down.A second experimental device was constructed in which a turbulent flow is generated by an array of randomly actuated, firing upwards, jets. This kind of forcing shows large velocity fluctuations while having a weak mean secondary flow, but a spatial decay of turbulence is observed with increasing distance from the jets. This flow allows us to study the modification of the settling velocity of glass particles slightly larger than the dissipative scale by turbulence as compared to the case in a quiescent fluid. By simultaneously measuring the flow properties and the trajectories of the inertial particles as they fall, it was possible to directly show that the beads preferentially sample the streaming zones with downward fluid velocity. As a result, their settling velocity is increased and this effect is all the more pronounced the greater the turbulence intensity and the smaller the particle size.; Bien que la turbulence soit connue pour sa capacité à disperser efficacement la matière, des particules inertielles transportées par un écoulement turbulent n’explorent pas toutes ses régions de manière homogène. Au cours de cette thèse, deux études expérimentales ont ainsi permis d’aborder le problème de l’exploration préférentielle de particules plus denses que le fluide porteur.Dans un premier dispositif expérimental, la distribution spatiale de billes en polyamide ou en polyacétal de tailles comparables à l’échelle intégrale de la turbulence, transportées par un écoulement tourbillonnaire fermé à grands nombres de Reynolds, a été étudiée. Ces particules étant plus denses que le fluide porteur, leur dynamique est pilotée par la compétition entre l’agitation turbulente et l’effet de la gravité. Lorsque la fréquence de rotation du disque qui engendre l’écoulement est faible, la sédimentation des billes domine leur resuspension par la turbulence de telle sorte qu’elles restent confinées en bas de la cuve. À partir d’une certaine fréquence de rotation, les fluctuations turbulentes surpassent l’effet de la gravité et les particules explorent l’entièreté de la boîte. Cependant, deux comportements extrêmement différents apparaissent en fonction de la taille des billes. Alors que les plus petites sont distribuées de façon presque homogène, les plus grosses se trouvent préférentiellement dans la partie supérieure, comme si la gravité avait été inversée. Ce piégeage trouve son origine dans le fait que ces dernières sont trop grosses pour atteindre les zones descendantes de l’écoulement situées dans les coins et ne retombent donc que très rarement vers le bas.Un second dispositif expérimental a été construit dans lequel un écoulement turbulent est généré dans une tour par un réseau de jets orientés vers le haut, s’allumant et s’éteignant de façon aléatoire. Un tel forçage permet de minimiser l’existence d’un écoulement moyen tout en garantissant d’importantes fluctuations de vitesse, mais une décroissance spatiale de la turbulence est observée en s’éloignant des jets. Cet écoulement a permis l’étude de l’altération de la vitesse de sédimentation de particules en verre de tailles légèrement supérieures à l’échelle dissipative par la turbulence. En s’appuyant sur une mesure simultanée des propriétés de l’écoulement et des trajectoires des particules inertielles au cours de leur chute, il a été possible de mettre directement en évidence que les billes explorent préférentiellement les régions descendantes de la turbulence. Il en résulte que leur vitesse de sédimentation est accrue d’autant plus fortement que l’intensité de la turbulence est grande et que la taille de la particule est petite.

International audience; The current article presents a regularization procedure of the Lagrangian point-force approach commonly used to account for the perturbation of a fluid phase by a dispersed particle phase. The regularization procedure is based on a nonlinear diffusion equation to naturally ensure parallel efficiency when the regularization length scale extends over several grid cells. The diffusion coefficient thus becomes a function of the particle source term gradient and expressions allowing to approximately adjust the regularization length scale according to the local particle to mesh size ratio are proposed, so that mesh refinement or polydisperse sprays may be handled. Elementary numerical test cases confirm the convergence of the present procedure under mesh refinement and its ability to locally adapt the regularization length scale. Furthermore, the chosen regularization length scale allows to match the leading order term of the perturbation flow field set by the particle beyond approximately two particle diameters in the Stokes regime. When applying the presented source term regularization procedure, the terminal velocity of a particle settling under gravity in the Stokes regime becomes relatively insensitive to mesh refinement. However, errors with respect to the theoretical settling velocity remain substantial and removal of the particle's self induced velocity appears necessary to recover the undisturbed fluid velocity at the particle location and correctly evaluate the drag force. As the current regularization procedure yields source terms that are close to c 2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Gaussian, an analytic expression from the literature is used to estimate the particle's self induced velocity. When combining source term regularization and removal of the particle's self induced velocity, good results are obtained for the terminal settling speed in the Stokes regime. Results obtained for horizontally separated particle pairs settling under gravity in the Stokes regime show equally good agreement with theoretical results. Because analytic expressions for the particle's self-induced velocity are no longer available at finite particle Reynolds numbers, correlations recently proposed in the literature are used to obtain correct settling velocities beyond the Stokes regime.