You are on page 1of 5

LARGE EDDY SIMULATION OF A PLANE TURBULENT WALL JET

N. Satish Kumar, AP/Aeronautical Engineering


M. Muthukumar, AP/Aeronautical Engineering
Subhabrata Koley, Lect/Aeronautical Engineering
ABSTRACT:

Large eddy simulation (LES) and Reynolds averaged Navier-Stokes simulation (RANS) of a plane
turbulent wall jet was done using FLUENT software. The basic problem of modelling the eddies in RANS models
is discussed. The eddy structures in the jet region were investigated by using LES model. The alignment of the
standard hairpin vortices is discussed. Turbulent quantities like Reynolds stress and the problem of counter
gradient diffusion were discussed. The results of computational results were compared with standard
experimental results.

Keywords: Large eddy simulation, Turbulent wall jet, RANS.

INTRODUCTION:

Diffusion occurs always in the direction of the gradient. For example heat transfer takes place from hot
to cold areas. There are many occasions in which the counter gradient diffusion takes place. One among those
isa turbulent wall jet.

A turbulent wall jet is formed when a turbulent jet is impinged on the wall at any angle. If the angle of
impingement is perpendicular to the wall then it is known as radial jet and when it is parallel to the wall it is
known as plane jet Fig 1.

Fig 1. Plane wall jet

The main applications of this flow feature is found in the case of vertical take-off aeroplanes in the close vicinity
to ground, flow over multiple flaps etc. So there is a strong reason to study about the flow features. Narasimha
(1989) in his paper said that this counter gradient diffusion might be because of the long arm effect of the jet
region. An experimental result by Narasimha is given in fig. 2 depicts the variation of streamwise velocity and
Reynolds stress (-uvmean).

Fig. 2 Mean velocity and Reynolds stress along wall normal direction
According to RANS models the Reynolds stress is modelled by

Where u and v represents the fluctuating velocity in streamwise and wall normal direction. The over bar
represents the time averaged products of these two components. du/dy represents the mean velocity gradient and
trepresents the turbulent viscosity. The equation tells that if (du/dy>0) then ( <0) and vice versa in order to
make turbulent viscosity as positive. Now from Fig. 2 we can say that within the boundary layer region du/dy is
positive so should be negative within the boundary layer. But there is some region where is positive. It
means that turbulent viscosity is negative to meet that compatibility condition. In turbulence the production term
is defined by - du/dy. If it is positive then it means that the turbulence is produced from mean flow and in
our case it is negative which conveys that from the turbulence the energy is given back to the mean flow.

In LES the eddy which are predominant in the flow are simulated and the small scale eddies are
modelled using RANS approach. In this work we have investigated whether the counter gradient diffusion is
because of large eddies or small eddies. If it is because of small scale eddies, then in LES also we should not
capture the counter gradient diffusion region otherwise we should capture the same.

COMPUTATIONAL DETAILS:

The jet exit has a dimension of 50 x 1 cm and exit domain which is 50 times larger than the jet exit. A
two dimensional domain was analysed in the case of RANS simulation and a three dimensional grid with
0.2x106 elements was used for Large Eddy Simulation. The number of elements was limited because of the
computational facility. The domain was meshed using quadratic and brick elements in the case of 2d and 3d.
The near wall region was finely meshed such that the y+ was in the order of one. This is one of the requirements
for turbulence models in order to capture the near wall effect. The jet exit was specified as velocity inlet to the
computational domain and the value was set to 20 m/s and remaining faces were given as pressure outlet. The
fig 3 shows the velocity contours in the computational domain. K- SST solver was used for RANS
simulations.

RANS RESULTS:

Fig. 3 Velocity contours in a plane wall jet

Fig 4 shows the velocity variation for different streamwise locations. It is obvious from the figure that
the maximum velocity decreases and the jet height increases with streamwise direction. From the figure we can
say that within the boundary layer the du/dy is positive and it is zero at the maximum velocity point. The du/dy
is negative in the jet region. So we expect the opposite in the case of (i.e it is negative within the boundary
layer and positive in the jet region)
Fig 4 Velocity variation for various stream wise locations

Fig 5 Normalised velocity and along wall normal direction

Fig 5 depicts the normalised velocity and along wall normal direction. From the figure we can see
that upto the maximum velocity point is negative and above the maximum velocity region is positive.
But because of the modelling of turbulent viscosity in RANS models we could not get the positive inside
the boundary layer. So RANS models could not capture when the energy is extracted from turbulence and given
to the mean flow.

LES RESULTS:

In the case of large eddy simulation the computational domain was same as RANS simulation but with
a three dimensional geometry. Both the sides of the domain were specified as periodic and at the jet inlet
spectral synthesizer were used to specify the turbulence at the inlet. Fig 6 depicts the variation of streamwise
velocity and Reynolds stress along wall normal direction.
Fig 6. Normalised u velocity and along wall normal direction

In Fig 6 the horizontal line is drawn from the maximum velocity point and a vertical line is drawn from
zero Reynolds stress. Below the horizontal line should be completely negative in order to get a positive
turbulent viscosity. But as observed in the experiments we are predicting a positive region. It means that
this effect is caused because of the large eddies present in the flow. These large eddies are responsible to extract
the energy from turbulence and sending back to the mean flow.

Fig 7. Coherent structures from Top view of wall jet

Fig 8. Coherent structures in three dimensional view

In fig 7 the flow is from right to left. The eddy structures are extracted using TECPLOT using Q
criteria. The eddy structures are flooded with the mean velocity magnitude. Most of the vortices are hairpin
shaped but turned opposite direction to the mean flow. This can be seen clearly in fig 8 in which the three
dimensional shapes are clearly visible.
CONCLUSION:

Large eddy simulation and RANS simulation were done using FLUENT. Counter gradient diffusion
region was captured using LES and the trend is almost comparable with the experimental results. The hairpin
vortices which face backwards to the flow could be responsible for the counter gradient diffusion. A
computation with good mesh accuracy should be done to capture the roots of the eddy. Its significance is that,
whether it is a single eddy which originates from the wall or a detached eddy from the wall.

REFERENCES:

1.Narasimha.R, Turbulence at the cross roads: the utility and drawbacks of traditional
approaches,1989,Mathematical science institute, Cornell university.

2. Turbulence by Davidson.
3. Experimental study of Turbulent plane wall jet. Damotharan, IISc M.E Thesis.
4. Fluent version 13 users Guide.

You might also like