Transient results with periodic boundary conditions?
 

Someone applies periodic boundary conditions to a turbulent channel flow to mimic an infinite domain size in spanwise and streamwise direction.

Does it make sense to analyse the flow during the transient stage, i.e. before a statistically stationary flow is reached?

A practical example: I see in literature simulations of particle-laden channel flows. It is analyzed how an initially uniformly distributed particulate phase obtains preferential locations close to the wall. I understand that the statistically stationary flow field has a physical meaning. But before that, during the transient stage, is that physical? Note that the particles in the simulation have all the same residence time. In reality that would mean that they are not at the same location...

My opinion is that you should distinguish between the transient of the velocity field and the transient of the particle laden problem.
Starting from the arbitrary initial condition, the velocity field develops only a numerical transient with no physical meaning. After that a statistically equilibrium is reached, you can set an initial uniform distribution of particles that develop a certain transient over a physically meaning unsteady velocity field. The question could be now in the fact if a uniform initial distribution of particles can or not be a good physical assumption.

Thank you for that answer!

Can you please tell me why you think the transient of the velocity field (e.g. from an initially laminar profile to turbulent) has no physical meaning whereas the transient of the particles (e.g. from initially uniform distribution in a turbulent field to non-uniform) is physical?

because using periodic BC.s you are not simulating a spatial evolving boundary layer over two flat plates... This flow problem is just a model for studying the wall turbulence, therefore we let the flow to forget the initial condition and study the turbulence only once it is fully developed...

the Lagrangian tracking of the particle depends only on the velocity field, the initial conditions for the particles never corresponding to a real experiment.

However, if you want, you can try starting from an exact Poiseuille velocity field without any superimposed perturbation and compute simultaneously the tracking of particles starting from an initial uniform distribution.
What is the physical meaning of such simulation is not clear...

great, I got it, thanks!