CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Turbulent Jet (https://www.cfd-online.com/Forums/main/193293-turbulent-jet.html)

selig5576 September 21, 2017 11:44

Turbulent Jet
 
1 Attachment(s)
I am interested in simulating turbulent jets with a 3D solver I've been working on. I was curious if anyone has suggestions on what I should use as an inlet condition. So far I'm using Gaussian noise where my standard deviation is equal to the maximum velocity. I am also setting u, v, w, p = 0 for my initial condition. I am using no-slip everywhere except the inlet and outlet. Does anyone have suggestions on some an inlet condition?

FMDenaro September 21, 2017 12:04

I suppose you are working with an incompressible formulation, therefore you cannot simulate the pysical transient from the rest.
At the first time step the flow immediately accomodate for the balance between inflow and outflow and the divergence-free constraint in the interior. That means you have to wait that the solution forgets the initial condition to get a physically correlated flow condition. Therefore, don't worry too much about the initial condition.

tiam September 21, 2017 14:42

What type of simulation are you doing, LES or RANS? Is the incoming flow turbulent or laminar?


Sent from my iPhone using CFD Online Forum mobile app

selig5576 September 22, 2017 14:28

Jet parameters
 
So far I have not implemented any turbulence models (RANS or LES). I want to start with a simple jet. Should I expect it to take awhile for the jet to form given my initial condition for my velocities are 0 however my inlet is just noise. Just to confirm, u, v, w = 0 except the outlet, is that the correct set of BCs?

FMDenaro September 22, 2017 14:39

If you have a good fine grid, you can work in a DNS formulation just setting a parabolic inlet profile (laminar inflow) and let the transition of the jet along x (that is for increasing Re_x). You need a sufficient lenght of the domain and have to wait a physical correlation of the velocity.

selig5576 September 24, 2017 12:41

Inlet condition
 
I have also reached the conclusion that a parabolic inlet profile is good. Is the following parabolic profile sufficient? (X = left, right, Y = front, back, Z = top, bottom)

Code:

for (int i=0; i < nx; i++)
{
  for (int k=0; k < nz; k++)
  {
    u[i][0][k] = y[i][0][k]*(2.0 - y[i][0][k])
  }
}

In other words

u_{i,0,k} = y_{i,0,k} \left(2 - y_{i,0,k}\right)

FMDenaro September 24, 2017 13:08

You need the inlet solution coming from a pipe flow, not a 2D channel

selig5576 September 24, 2017 14:56

Inlet BC (correction)
 
Much apologies about that, what I meant to say was

Code:

for (int i=60; i < 80; i++)
{
  for (int k=60; k < 80; k++)
  {
    u[i][0][k] = y[i][0][k]*(2.0 - y[i][0][k])
  }
}

Essentially I define my inlet (in terms of the grid) to be a square shaped pipe. Is a square pipe that is 20 x 20 points sufficient in side, for now? I am running nx = 140, ny = 280, nz = 140 for testing purposes.

FMDenaro September 24, 2017 15:09

Quote:

Originally Posted by selig5576 (Post 665472)
Much apologies about that, what I meant to say was

Code:

for (int i=60; i < 80; i++)
{
  for (int k=60; k < 80; k++)
  {
    u[i][0][k] = y[i][0][k]*(2.0 - y[i][0][k])
  }
}

Essentially I define my inlet (in terms of the grid) to be a square shaped pipe. Is a square pipe that is 20 x 20 points sufficient in side, for now? I am running nx = 140, ny = 280, nz = 140 for testing purposes.


if the inlet is a duct with squared section, the solution is not that of circular pipe or plane channel... you have to solve the equation

Lap u = Re*dp/dx
u= 0 on the boundaries of the duct

selig5576 September 24, 2017 15:17

Inlet BC
 
You are 100% correct. If my paper work is right I should have

Code:

r = sqrt(pow((x[i][0][k] - 50),2.0) + pow((z[i][0][k] - 50,2.0))

if ( r < radius)
{
u[i][0][k] = y[i][0][k]*(2.0 - y[i][0][k]);
}

This should ensure a circular pipe like profile.

To put it more cleanly,

r = \sqrt{(x_{i,0,k} - r_{a})^{2} + (z_{i,0,k} - r_{b})^{2}}

FMDenaro September 24, 2017 15:30

But you do not have a circular pipe...the 2d profile in a squared section is quite different.
Here you can find the solution
https://books.google.it/books?hl=it&...20duct&f=false

selig5576 September 26, 2017 13:04

Strange problem
 
2 Attachment(s)
So, I decided to change my test for the round jet as its more related to what I need to do. As such,

Code:

for (int i=0; i < nx; i++)
{
      for (int k=0; k < nz; k++)
      {
          round_jet = sqrt((X[i][0][k]-0.5)*(X[i][0][k]-0.5) + (Z[i][0][k]-0.5)*(Z[i][0][k]-0.5));

          if (round_jet < 0.08)
          {
              u[i][0][k] = 1;
          }
      }
}

Even though I'm setting the u-velocity to be 1, I should still get a jet, however I get this weird problem. For some reason I get this build up by the outlet and it ultimately "kills" the development of my jet. Even after 10k time steps, the jet starts going to the side (??). I've also tried gaussian noise and I get the same result of the jet not forming.....

FMDenaro September 26, 2017 13:52

I don't understand what are you plotting....
Start debugging the first time step, check if the flow rate in inlet is equal to that at the outlet, check the continuity in the interior.
Plot the 1D sections of the inlet profile.

selig5576 September 26, 2017 15:10

Clarification
 
The plots I attached are that of the u-velocity from a top perspective and a rotated perspective. I will double check the inlet and outlet. On a side note, for visualization, should I be plotting the magnitude of the velocities? Thanks!

FMDenaro September 30, 2017 19:58

Quote:

Originally Posted by selig5576 (Post 665765)
The plots I attached are that of the u-velocity from a top perspective and a rotated perspective. I will double check the inlet and outlet. On a side note, for visualization, should I be plotting the magnitude of the velocities? Thanks!

At the inlet plane the u velocity or the magnitude are the same. Plot the isovalue lines of the velocity over the inlet plane

selig5576 October 4, 2017 15:12

Jet formulation
 
5 Attachment(s)
So after a few days of working on this problem. I have the following findings. With a parabolic inlet profile I have a problem where the jet does not fully form. What I mean is it starts expanding in terms of the radius of the jet, but not the length of the jet. I have been checking my conservation of mass and it is on the order of 1e-7.

FMDenaro October 4, 2017 15:25

And do you get a steady state solution with the jet not entering in the domain?

selig5576 October 5, 2017 12:43

Jet formation
 
I am of the understanding I should change the inlet to u = 0 so the solution is appropriately contained in the domain?

FMDenaro October 5, 2017 12:47

Quote:

Originally Posted by selig5576 (Post 666678)
I am of the understanding I should change the inlet to u = 0 so the solution is appropriately contained in the domain?


No, I wonder what happens just after 1 time step and if you have a flow rate in outlet. Then plot in the plane yz the contour of the u velocity to understand if the inlet profile is correct.

selig5576 October 5, 2017 14:34

Jet
 
1 Attachment(s)
Dr. Denaro, my jet profile at t = 1.0 is attached. My inlet profile is

Code:

for (int i=0; i < nx; i++)
    {
        for (int k=0; k < nz; k++)
        {
            round_jet[i][0][k] = sqrt((X[i][0][k]-1.0)*(X[i][0][k]-1.0) + (Z[i][0][k]-1.0)*(Z[i][0][k]-1.0));

            if (round_jet[i][0][k] < 0.05)
            {
                u[i][0][k] = X[i][0][k]*(5.0 - X[i][0][k]);
            }
        }
    }



All times are GMT -4. The time now is 15:29.