dns of turbulent bubble-laden channel flowsÂ
This thesis deals with numerical simulation methods for multiphase flows where different fluid phases are simultaneously present. In particular, the subject of interest is a system in which the carrier fluid is a liquid that transports dispersed gas bubbles. The simultaneous existence of physical phenomena spanning a wide range of scales of motion is certainly one, if not the most, complex aspect of bubbly multiphase flows. In this context, numerical simulation is a useful and powerful tool for a better understanding of the physics of such systems. The method applied in this thesis is direct numerical simulation (DNS), where all the details of the flow, up to smallest scales, are resolved by the computational grid and time steps. The aim of this thesis is to develop an accurate and computationally efficient tool for DNS of turbulent bubbly channel flow, starting from the volume-of-fluid formulation.
While single phase flow has been studied for a considerable time, the presence of a second phase in the flow drastically changes the structure of the turbulence making the problem significantly more complex and accessible to detailed simulations only much more recently. An important current limitation is that a mathematical fluid is often used. In particular, the mass density ratio between the fluid and gas phase is often set to approximately 10 since high mass density ratios are notoriously challenging from a numerical standpoint. The method developed in the present work attempts to take a step forward in the direction of DNS of turbulent channel flows loaded with thousands of bubbles and mass density ratios closer to a real air-water system at atmospheric pressure. In order to achieve this, efficiency, parallel scalability and accuracy are essential.
The developed method has been validated in a number of steps: first, the transport of the indicator function used to track the second phase has been tested independently from the Navier-Stokes equations by prescribing a velocity field for which the analytical solution is known. Subsequently, a validation with a well-established benchmark was carried out for the simulation of a rising bubble in a viscous liquid under different parameters proving the method to be able to accurately deal with large interface deformation. Finally, a set of simulations of bubbly turbulent channel flow has successfully been carried out and analysed. Statistical quantities have been measured and compared with reference data available from literature. An overall good agreement was found. This indicates the reliability of the method which was subsequently used to investigate the flow statistics at higher mass density ratios of up to 100 and different bubble sizes.
