LES of Plane Channel flow using Gerris

An LES (pretty coarse, though) of Plane channel flow using Gerris Flow Solver, based on the ref[1]:
The bottom and top boundaries are walls.
The left and right boundaries are periodic.
The back and front boundaries are periodic
Channel length:width:height = 2:1:1
The kinetic energy dissipation at the subgrid scale is modeled with the Smagorinsky-Lilly model with Cs = 0.1
The simulation was done for a Reynolds number (based on friction velocity and channel half width) of 395.
Since the GfsMetric feature for 3D, in Gerris, had a bug, i had to use the octree mesh.
As a result,
-the total number of cells was huge, around 844,000 as against 180,000 in the ref[1]
-yet the wall y+ is around 2.4 (calculated based on the obtained u_tau) which is twice coarser
-and hence incorrect u_tau (25% smaller)
Expecting an incorrect u_tau ahead, the simulation was based on the Reynolds number based on bulk velocity and channel half width, calculated to be 6808 using the formula Re_bulk = 7.33*Re_tau^1.143 [Dean(1978) cited in ref[2]]. Aiming this Re_bulk, the source term in the stream-wise momentum equation was adjusted to yield a constant value everywhere, to drive the flow. [The final source term is ~1.12e-4 L/T^2 and U_bulk=0.1368 L/T, given a kinematic viscosity of 2e-5 L^2/T]
No synthetic turbulence was given as initial condition. White noise was given for both the stream wise momentum source term and velocity at the begining for one flow through time; consequently turbulence sets up only after ~20 flow through times.
The simulation ran for ~140 flow through times and took ~8 days on a 8 cores of 3.6GHz machine!
The first and second order statistics are, however, not time averaged. From the simulation of last time step, ensemble averaging was done of the values at every plane in the y-direction.
In the charts for comparing the results of first and second moments, scaling with respect u_tau was done based on the correct u_tau (~0.008) instead of the obtained one (~0.006)
References:
1. Eugene de Villiers, 2006, "The Potential of Large Eddy Simulation for the Modeling of Wall Bounded Flows", PhD Thesis
2. Johan Larsson, 2006, "Towards Large Eddy Simulation of Boundary Layer Flows at High Reynolds Number: Statistical Modeling of the Inner Layer" PhD Thesis

[@jeltirb]

Hello, this is a nice result, I am also trying to run a simulation of this test case, did you implement the LES Model to Gerris or is there a free version with this possibility ?
Thank you for your answer

[@balapesar]

Dear Eugene
Greetings,
Very nice video!
I want to install Gerris on my computer using the instructions given in its official website. However, there are several different files and I don't know exactly that which one I should install.
I have Ubuntu on my computer, would you please let me know that what other files I should install on Ubuntu in order to be able to run Gerris on my pc and show its graphical results as well?
Thanks very much,
Sayed

[@molinos94]

Try to install Gerris using darcs ;)