Advanced CFD methods for turbulent flows

This thesis reports further developments of the non-oscillatory forward-in-time multidimensional positive definite advection transport algorithm (NFT-MPDATA) class of solvers in simulating incompressible and variable-density low-speed flows. MPDATA due to its self-regularisation property can be employed to enable implicit large eddy simulation (ILES). In contrast to conventional large eddy simulation (LES), the subgrid-scale (SGS) effects in ILES are implicitly provided through the numerical dissipation rather than explicitly modelled with an SGS model, allowing for a simpler numerical implementation and a cost-effective computation. The capabilities of NFT-MPDATA to perform both LES and ILES are examined through several systematic numerical studies simulating various turbulent flows, namely cavity flows, flow past a sphere, and jets/plumes. The validity of ILES and the implemented SGS models for conventional LES is established by comparing the computed results against other numerical studies or notable experimental data. The solutions to the variable-density low-speed flows are obtained by developing a new type of NFT-MPDATA solver that integrates the low Mach number (LMN) equations. It is capable of capturing variable-density effects driven by large thermal variations. The accuracy of the new method is evaluated using benchmark computations of laminar differentially heated cavity (DHC) flows. The LES and ILES capabilities of the proposed LMN scheme are also demonstrated through simulations of turbulent DHC flows and non-isothermal turbulent free-jets. Additionally, it is illustrated using the turbulent Helium plume as a test case showing that the proposed LMN scheme can be augmented to capture variable-density effects arising from compositional variation in binary species flows assuming an isothermal condition.