Hi,

I am solving Turbulent channel flow by LES. But, i am facing problem as follows.---

I am not using any software. i am using my own code in fortran language.

For the channel----

i am taking periodic b.c in streamwise and spanwise direction.

dimension--4pi*2*2pi and grid 50*56*66

Re=205, corresponding friction velocity and half channel height.

So, i am using mean driving pressure gradient=1.0

for initial condition: u(x,y,z)=(1-y**8)+0.005*(rand(iseed)-0.5). /////Rand gives value , between 0,1, y is wall normal coordinate.

also, v(x,y,z)=0.0

w(x,y,z)=0.0

using---2nd order cd for convective and diffusive discretization. and adam-bashforth for time .

time step--0.0002

The problem is , i saw the mean velocity increase and reaches a very high value during the simulation.

also, along the channel centerline, i check the instantaneous u, v, w velocity----it does not show any turbulence.

what is the remedy....can anybody share his experience , during channel flow simulation.

thanks in advance.

How long have you run it for? Are you sure its mean velocity that is increasing? How is instantaneous velocity behaving, increasing, decreasing or staying constant?

Regards

1) I am assuming you are solving invompressible NS equations. Are you making sure your initial condition satisfies continuity? Check Kim, Moin and Moser's 1987 JFM paper for good initial condition.

2) Is your grid resolution enough to resolve the wall shear stress region? I had solved LES (Re=180) and had finer mesh than yours.

3) Make sure your LES model is working fine, unless you have already tested this in some other simulation.

Maybe you could start with a lower Re (180) as in the JFM paper and try to figure out what is going on (just a suggestion)

-agg

The resolution shouldn't be a problem for this Reynolds number. Probably the flow become laminar which implies that you have to change the initial conditions. What are using? Just some random small-scale disturbances? Take some initial disturbances that have a larger length scale and make sure that the disturbances are large enough. In my DNS I usually just take some random sine disturbances. Works fine.

Hi,

First of all , thanks to agg and tom. My, Re is 205. And , my resolution near the wall is 0.68 wall unit. Still, according to agg, i will try, with Re=180.

How, could i understand , my intial condition is divergence free or not?

As, i am using a profile like below for i.c----

u(x,y,z)=(1-y^8)+10*(rand(iseed)-0.5) v(x,y,z)=0.0 w(x,y,z)=0.0

rand(iseed) will give value between 0,1.

So, my disturbence is -5,+5 at random values.

Could, you please, tell me , whethere, this field is satisfying continuity or not? ------------------------------------------------------

Tom, is my random disturbence fullfill the necessary large scale and amplitude.

tom, could you please, tell me what is the exact expression for i.c condition you , used.

Do, you give v=0.0, w=0.0 or anything else.

--------------------------------------------------------

Thanks pankaj

The disturbances that you impose have a very short wave length and viscosity or the subgrid model will almost immediately dissipate/damp most of these these disturbances. Therefore, I suggested to use disturbances with a longer wave length because these will no be damped so easily. Take just a couple sine/cosine disturbances, the fine details don't matter so much, just try something. It doesn't have to be divergence free (at least with my code).

Thanks. Tom.---

Could you please, send me the expression you used.

one more thing---how you specify the mean pressure gradient?

I have used dp/dx=1.0, beccause, my code is nondimesionalized by half channel height and friction velocity. am i correct?

or what else you are using?

Hi,

First of all , thanks to agg and tom. My, Re is 205. And , my resolution near the wall is 0.68 wall unit. Still, according to agg, i will try, with Re=180.

How, could i understand , my intial condition is divergence free or not?

As, i am using a profile like below for i.c----

u(x,y,z)=(1-y^8)+10*(rand(iseed)-0.5) v(x,y,z)=0.0 w(x,y,z)=0.0

rand(iseed) will give value between 0,1.

So, my disturbence is -5,+5 at random values.

Could, you please, tell me , whethere, this field is satisfying continuity or not? ------------------------------------------------------

Tom, is my random disturbence fullfill the necessary large scale and amplitude.

tom, could you please, tell me what is the exact expression for i.c condition you , used.

Do, you give v=0.0, w=0.0 or anything else.

-------------------------------------------------

Could you please, send me the expression you used.

one more thing---how you specify the mean pressure gradient?

I have used dp/dx=1.0, beccause, my code is nondimesionalized by half channel height and friction velocity. am i correct?

or what else you are using?

[rowsanalam]

hi pankaj myself rowsan --plz help me with your code i am also facing with same problem as you--plz send me the code--in my id --rowsanalam@gmail.com

Originally Posted by pankaj saha
;55894

Hi,

First of all , thanks to agg and tom. My, Re is 205. And , my resolution near the wall is 0.68 wall unit. Still, according to agg, i will try, with Re=180.

How, could i understand , my intial condition is divergence free or not?

As, i am using a profile like below for i.c----

u(x,y,z)=(1-y^8)+10*(rand(iseed)-0.5) v(x,y,z)=0.0 w(x,y,z)=0.0

rand(iseed) will give value between 0,1.

So, my disturbence is -5,+5 at random values.

Could, you please, tell me , whethere, this field is satisfying continuity or not? ------------------------------------------------------

Tom, is my random disturbence fullfill the necessary large scale and amplitude.

tom, could you please, tell me what is the exact expression for i.c condition you , used.

Do, you give v=0.0, w=0.0 or anything else.

--------------------------------------------------------

Thanks pankaj