Dear Bart,
thank you for your reply and detailed explanation! If I’m right, then for a simple 2D poiseuille flow I’d like to run I will have to do the following…
physical extension
x = 0.100m
y = 0.010m (direction of flow)
u_p = 1m/s
nu_p = 135.2m^2/s (air)
g_p = 9.81 m/s^2
with 50 cells resolution in x direction, and 5 cells in y...
dx_p = 2e-3
dt_p = dx_p^2 = 4e-6
in lattice units
u_lb = u_p * (dt_p / dx_p) = 1 * (4e-6 / 2e-3) = 2e-3
nu_lb = nu_p * (dt_p / dx_p^2) = 135.2e-7 * (4e-6 / 2e-3^2) = 135.2e-7
g_lb = g_p * (dt_p^2 / dx_p) = 9.81 * (4e-6^2 / 2e-3) = 78.48e-9
So far, I understood well. Referring to J.Latt’s documents, I thought I could use the reynolds number to get viscosity by…
nu_lb = ( u_lb * resolution ) / Re
thus specifying the characteristic velocity for the case, u_lb, implicitly!?
However, I’ve set up a test for that simple 2D poiseuille flow, using the above conversion. I use periodic boundaries in flow direction, thus I do not specify a velocity, only dx, dt, viscosity and acceleration (all as converted above). In my results I receive a reasonable velocity profile, but since I do not specify a velocity (only periodic boundaries and an acceleration), how to interpret the results?
First I thought of using the following…
same extend in x and y, thus having...
dx_p = 2e-3
with e.g. Re=10, and u_lb = 0.01
nu_lb = (u_lb * resolution) / Re = (0.01 * 50) / 10 = 0.05
dt_p = dx_p * u_lb = 20e-6
for gravity, the same as above...
g_lb = g_p * (dt_p^2 / dx_p) = 9.81 * (20e-6^2 / 2e-3) = 1.962e-6
What, if using this approach? Is it wrong? Results looking good from qualitative point of view, just the maximum magnitude of the velocity is about 0.00122, which is far from u_lb=0.01… I’m really confused!
So, finally, I think I understand what you explained concerning conversion, but I think I do not understand how to interpret the settings according to the results!?
Finally, in my code I’ve just used something like…
// u_x already computed as sum of pdf's with respect of direction vectors,
// just add gravity...
u_x = u_x + tau*g_x
Maybe, I totally misunderstood the principal of how to implement it?
Please, I would appreciate continue dicussion on that topic
Thanks in advance, Francois