Gas-solid flows are abundant both in nature and in industrial applications, therefore the ability to accurately predict their behaviour is of crucial importance. The main goal of the research project presented in this thesis was to develop a methodology for eff cient true direct numerical simulations (DNS) of turbulent
flows with solid particles. True DNS (TDNS) in this case implies that not only all the spatial and temporal scales of the f ow f eld are directly computed, but also that the appropriate boundary conditions
are imposed on surfaces of the particles allowing for the boundary layer development.
The designed approach is strongly based on the ideas of the Immersed Boundary Method (IBM). In this technique, two separate computational grids are used: a fixed fluid grid and a moving triangulated one, representing the body surface. The flow equations are modif ed in the regions were the grids overlap. Various implementations of the IBM are discussed, along with the most common diff culties encountered while using this approach. These challenges include accurate imposition of the boundary conditions, evaluation of the fluid-particle momentum transfer and spurious pressure oscillations observed in the case of moving bodies. A number of improvements, designed for addressing the main IBM challenges,
are proposed and evaluated on a set of test cases. Additionally, a parallel triangulation library, MFTL, designed in the course of the research project is presented.
The IBM technique is subsequently adopted for the investigation of flows past non-spherical particles at a range of Reynolds numbers and orientations. The results of this study lead to the development of
shape-specif c correlations evaluating the drag, lift and torques on non-spherical particles as functions of Reynolds number and the angle of incidence. Also, an approach for describing the motion of such
particles is presented as well.