Hi,

I am trying to simulate a channel flow using LES turbulence model, and my channel has periodic boundary condition both on the streamwise and spanwise directions.

Although spanwise periodic bc is not much a problem as flow there is literally less, but in the case of streamwise periodic BC, I do have a problem as the flow dies after some time, I was able to find out(from previous yrs postings) that there is necessary to maintain a pressure drop, across the channel.

This actually reflects as a Source term, that should be added, while solving for the streamwise velocity component (u), but I am not sure ? this source term should contain.

If any one can help me in this regard, I would be greately thankful.

Thanks in advance,

Glen

hi Glen, I am also working on LES of channel flow. Now onwards let us share our views and experiences.I too had the same problems. But I overcame those problems. I will tell you what I did? 1. Try to work with Dimensionless equations. So u will non dimensionalize pressure as (P/(rho*Utou**2) 2. Pressure in streamwise direction will have two components: Fixed pressure drop + periodic part

P= P(periodic)+(dP*/dx*) multiplied by (distance in streamwise direction) 3. Once this is done, Then u can fix the pressure gradient in streamwise direction = 1. (dP*/dx*=1)

4. now regarding the problem that flow dies out. It may be because of improper initial conditions. U can get DNS data of instantaneous U,V,W,P from website, which u can use as initial conditions.

In case of any ther help required, do not hesitate to contact me.

Sachin

Hi Glen and Sachin,

I am also working on LES and right now simulating the channel flow case. I too had the same problem. But in my case the CFL number exceeded 1 and the run stopped. I am still not sure if its becasue of this periodic boundary condition issue, although I know that mass will nto be conserved without the source term !!!

so my question is, is this source term very important ?? Can the numerical discretization error act as this source term ???

Please me reply me back at your convinience.

thanx in advance

Shiv,S

Hi, Sachin. I am also into this problem.

1.Is the dimensionalized pressure necessary? I don't see the necessity in your steps.

2. The decomposition of the pressure is actually mean pressure plus the fluctuating part. So you assume the fluctuating pressure is perioic. Is there any theory support for this?

3. According to Pope's book Turbulence, the initial condition for the fully-channel flow in the DNS is unimportant actually. I think the LES also share this aspect.

Thanks.