LES with dynamic model
 

Hi; I am solving a turbulent channel flow with complex geometry by LES. To avoid the averaging in the homogeneous plane, I have used a Lagrangian Dynamic Model of Meneveau (1996). For initial condition, Cs=0.16 is chosen. I recognized that the value of Cs varies between 0 and 1. The high values of Cs causes to remove the turbulence fluctuations (makes the flow laminar). To avoid high values for Cs I clipped the Cs when it exceeds 0.16. I have now a turbulent channel, but the velocities are over predicted. I have some questions: 1. Is it usual in dynamic model to get the Cs higher than ~0.16 in some regions of the domain? 2. Is it a correct way to clip the Cs when it becomes large? Thanks

Re: LES with dynamic model
 

Hi,

I am also working on turbulent channel flow. Could you share some aspect?

What initial condition you are using ? i hope you are using periodic b.c. then what is the mean driving pressure gradient.

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

As , for your quearry, i can tell , what i am using.

First--You do localised volume averaging of cs over the test filter cell. Then you use clip.

everybody use clip to avoid numerical instability.

thanks

Re: LES with dynamic model
 

Thanks pankaj saha for help. I am using periodic b.c. in stream and span-wise directions. For the normal direction, no-slip in the bottom and free sleep at the top. The bottom is not flat. I am solving the flow over a wavy bed. I use a Cartesian grid with ghost cell immersed boundary condition. The initial condition is randomly by using sin and cosine for u, v and w. The flow is driven by pressure gradient to make the bulk velocity unit. As I mentioned, I am using Lagrangian Dynamic model. In Lagrangian Dynamic Model by using ramp function for I_LM, we avoid the negative viscosity. The problem arises only from the high value positive viscosity.

Can you please let me know, which kind of Dynamic Model you use? Plane averaging or Lagrangian?

Re: LES with dynamic model
 

I am using Localized volume average. It is not palne average.

See, From formulae you will get, cs. Then do the ageraging over the neighbouring test volume cell. Like evering time you will get 27 cell average value.

Could you please post, the expresion you are using for the 3 componeent of velocity with random function....menas the for u, v, w .

Thanks.--

If you have any problem, i will send you a paper, which told how to do vol.average

Re: LES with dynamic model
 

Thanks pankaj for reply. I initialise as follow:

do k = 0,kmax+1

do j = 0,jmax+1

do i = 0,imax+1

u(i,j,k) = 0.01 * sin(i+j*k+0.) + 1.0 ! 1.0 is the bulk velocity

v(i,j,k) = 0.01 * sin(i+j*k+1.)

w(i,j,k) = 0.01 * sin(i+j*k+2.)

p(i,j,k) = 0.0

enddo enddo enddo

Can you please send me the paper for volume averaging of cs. Thanks

Re: LES with dynamic model
 

Physics of fluids, A5(12), pp-3186-3196.

Zang,Street, Koseff.

---------------------------------- If, you dont have, any access to p.o.f, then tell me your mail id, i will send you a copy.

here , in appendix, the vol, avg technique is there.

------------------------------------------------------ Doubt: You, use 0, 1, 2 in the three component of velocity. What is that?

Also, you did not use, coordinate distance. Means, only the i,j,k----No, x,y,z is there.?

Could you write in details?

Also, are you supplying driving pressure gradient by itarative technique or simply a body force term??

Re: LES with dynamic model
 

Thank you to send me the paper. I have access to download it. I used i, j, k to make a random field of velocity in the domain. If you use x, y and z, in this case for two points near together, you will have not much differences in their velocities. Because you have no other source to apply the perturbations (because of periodic boundaries) the turbulence will maybe die for relatively low Reynolds numbers. If we apply i,j and k for all three components with the same function (here sin) we will have the same perturbation for all three components. To make a difference between the velocities components, I added arbitrary numbers 1, 2 and 3. After giving a random field, you need to apply the boundary conditions. For pressure gradient, I used a body force. The body force is calculated from the bulk velocity. In any time step, I considered " fx=(Ubulk0-Ubulk)*cof ". Ubulk0 is the desired and Ubulk is the calculated bulk velocities. Because the difference is not very high to balance fx, you need to multiply it by a coefficient. For example I use 100. Don't forget to use Ubulk0 a little higher that the desired value to have a body force in case you reach Ubulk0. Determining value for cof and Ubulk0+eps is a try and error problem. If you have large fluctuations, fx will maybe oscillates, because Ubulk will be not constant in time. In this case you need to calculate the average Ubulk in time. Thanks

Re: LES with dynamic model
 

fx=(Ubulk0-Ubulk)*cof

I hope, you are adding 'fx' in streamwise momentum equation.

see, for any Re, Ubulk0 has a desired value. So, it is fixed before the code is run. for e.g--Re=4000=Ubulko*H/v So, IF you take h=1 and Ubulk0=1 v=1/4000 is the setting condition for code.

Now, for any time you will get Ubulk, from code and it will give fx, thereafter.

So, when Ubulk0=Ubulk then fx=0, so, the code does not get any body force. You, are saying so you are taking an eps. How, you set that one . That means, code is not calculating eps, rather you supply it manually.

i did it, by some formuallae----like below.

p_new=p_old(1+cof((Ubulk0/Ubulk)**2-1)

First, you take a guess value of p_old, Also, Ubulk0, is known from Re. Ubulk is calculated , everytime.

And your body force term, p_new, will be updated at everytime step. when, you reach , Ubulk=Ubulk0, then after, P_new=P_old...so, no need of trial and error.

You just put this, P_new to your x-momentum equation.

try this--this is more scientific.

QUICK Scheme
 

Hi,

I wanted to know if anyone has used QUICK scheme in LES despite it is too dissipative. According to Mittal and Moin (1997) the QUICK is too dissipative for reactive flows and generated noise flows and there central schemes should be used. But for ordinary flows it is nearly sufficient.

I would appreciate very much it if anyone could tell me about any resent journal articles in which QUICK (or B-QUICK) is used.

Sincerely,
Maani