The interaction of solid particles with turbulence has for long been a topic of interest
for predicting the behaviour of industrially relevant flows. The flow interacts with the
solid particles, the particles in turn modify the turbulent flow field, and particles collide
with one another. The multi-level of interactions between the two-phases challenges our
physical comprehension of these dynamical systems, as well as the numerical modelling
strategies. Large-scale predictions on two-phase flows encompass the numerical solving of
the largest turbulent structures and the tracking of large numbers of individual particles.
Large Eddy Simulation methods are widely used for their low computational cost and
accuracy in resolving most of the turbulent scales. In one-way coupled simulations however,
the smallest scales can be relevant when considering second order dispersions of solid
particles. A subgrid scale model is presented which enforces a correlation of the subgrid
scale velocity of neighbouring particles, hence ensuring particle-pair dispersion statistics.
Depending on their size and inertia, particles tend to accumulate in preferential regions
forming clusters. A new numerical investigation on the turbulence modulation due to
these large scale clusters is presented. A novel method for computing a characteristic
length scale of clusters is introduced. Two-way coupled Direct Numerical Simulations
of homogeneous isotropic turbulence are performed for a large range of particle Stokes
number and mass loadings. A correlation is found between one the turbulent modulation
length scales and the characteristic length scale of clusters.
The Lagrangian framework is appreciated for its accuracy in tracking particle trajectories.
However, in case of four-way coupling approach, the deterministic detection of colliding
particle pairs becomes computationally very expensive. A new correlated stochastic model
for particle collisions in the Lagrangian framework is introduced. The model is first validated
for a homogeneous isotropic turbulence case, and then in a vertical channel flow.