The stochastic model

A "more physical" model is available under Fluent. It is called the stochastic model. It is based on a gaussian distribution of velocity field such as:
U= <U> + u'. Fluent gives <U> average field and  urms =<u'u'>^0.5  fluctuation field. A number S is randomly chosen such as instantaneous u'= S urms where urms is the local fluctuation velocity.  Then the trajectories depend not only on the average velocity field but also on the fluctuation field.
However k-epsilon model considered that the fluctuations are isotropic, what is not true in the reality.

Now some particles are taken in the recirculations. The residence times increase considerably.

Several tries can be done with the same particle to obtain several random trajectories. A lot of tries with the same particle can give a statistical average trajectory.

Thanks to this model, some conclusions about the flow can be done easilier. It seems to appear that particles injected in the center of the surface can be easily rejected on the contrary to other particles that can be trapped by recirculations. It can be remarked that a "red particle" is capted after the first intern and stays in the recirculation.

Comparisons between particles of different mass are possible, but conclusions are difficult to obtain because of the little number of tries.

Same idea for different diameters and same conclusions. Here big and small particles are trapped in the recirculation. A lot of  tries could give interesting conclusions.