Valentin Borisavljevic 2140173

RANS and LES simulation of a pulsating jet

1. Abstract

CFD simulation of fluids is a highly relevant subject for many current engineering
applications. Typical quantities calculated with the use of CFD include the aerodynamic
characteristics of automobiles and aircrafts, the hydrodynamic drag on boats and the torque
produced by the wind on windmill propellers. The two main CFD methods are RANS and LES.

This report looks into the simple case of a pulsating fluid jet in a two-dimensional, symmetric
domain. The goals are the comparison of two different CFD methods, and the analysis of the
effects of different boundary conditions. The two CFD methods considered are RANS and
LES.

To achieve this, the report shows a comparison of the results of four different CFD
simulations: one uses RANS, the three others use LES but have different domain sizes.

The report states two conclusions. Firstly, a clear difference is visible between the results of
the RANS and LES methods. Also, there is an influence of the boundary conditions on the
flow field.

2. Introduction

A flow field is defined as an area with moving fluid. Many quantities characterise it. The basic
ones are the velocity, the pressure, and the temperature. Velocity is represented by a vector
⃗ = 𝑉𝑥 ⃗⃗⃗
𝑉 𝑒𝑥 + 𝑉𝑦 ⃗⃗⃗⃗
𝑒𝑦 + 𝑉𝑧 ⃗⃗⃗
𝑒𝑧 , with 𝑉𝑖 = component of the velocity in the i-direction (and the total
magnitude = 𝑉 = √𝑉𝑥2 + 𝑉𝑦2 + 𝑉𝑧2 ). Furthermore, each of the flow characteristics
mentioned above eventually varies with the position within the field, and with time in the
case of a transient situation. As a result, the flow field in terms of velocity is a vector field,
because there is a different vector for each position within the field. More specifically, it is
called the velocity field. The report looks at the flow field in terms of the velocity magnitude
V.

Two types of flows can be distinguished: laminar flows and turbulent flows. In the turbulent
̅𝑖 + 𝑉𝑖′ . In other
case, the velocity components 𝑉𝑖 (and thus the total magnitude V) vary as 𝑉

1
Valentin Borisavljevic 2140173

words, they have an extra variating component beside of their normal variations with time
and position.

For any application within the flow field, it is necessary to calculate the velocity. The
calculation of the velocity can either happen by an experimental model, or a mathematical
model. The latter is based on a set of differential equations (named “governing equations”),
more specifically the Navier-Stokes equations, together with a set of boundary conditions.
An analytical solution could not (yet) be found, except for very simple theoretical situations.
This is why numerical solution methods play an important role. The most common numerical
methods are the CFD computer codes. The foundation of the numerical approach is the
discretization of the domain (so not 𝑥 ∈ [𝑥0 ; 𝑥𝑛 ] but x can only be 𝑥0 , 𝑥1 , …). This converts
the continuous domain to a “mesh” or “grid”, which is an ensemble of discrete points
(distinct positions in the domain). The space between the grid points defines how fine the
grid is: the less space between points, the finer the grid.

There are 2 important types of CFD codes: Reynolds-averaged Navier-Stokes simulation

(RANS), and large eddy simulation (LES). RANS finds a time-averaged solution (in other
words, a solution averaged over 𝑡 ∈ [𝑡𝑖 ; 𝑡𝑖+1 ] , then for 𝑡 ∈ [𝑡𝑖+1 ; 𝑡𝑖+2 ] , …). LES finds a
spatially-averaged solution (local averages of grid points). Furthermore, LES requires more
grid points than RANS, in other words a finer mesh (less space between grid points).

The report will look into the difference in solutions from RANS and LES, and into the
influence of certain boundary conditions. Four cases are considered: case 1 is a RANS
simulation in a domain of a certain size, case 2 is a LES simulation in the same domain to see
the difference with RANS, and cases 3 and 4 are LES simulations in domain of different sizes
to see the influence of the side and outflow boundary conditions.

3. Problem setup (situation)

The situation of the flow field is a pulsating jet of air in a 2-dimensional domain. A first
aspect of the situation is the geometry of the domain. It is 2-dimensional, and also
symmetric, because the jet inlet is located in the middle of the horizontal length (x) of the
domain. The inlet stretches over a horizontal length of D=0.1m. The domain size is defined
by the values Lx and Ly. As mentioned at the end of the previous paragraph, three domain
sizes are considered: Lx=2m and Ly=4m , Lx=2m and Ly=2m , Lx=4m and Ly=4m .

2
Valentin Borisavljevic 2140173

Lx [m] Ly [m]
case 1 2 4
case 2 2 4
case 3 2 2
case 4 4 4

There is a horizontal solid wall on both sides of the inlet, which delimitates the bottom of the
domain. The other domain boundaries (left, right and top) are free.

A second aspect of the situation is the mesh which discretizes the domain. As mentioned in
the last paragraph, the foundation of a numerical method is the definition of a mesh. The
ICEM extension of the ANSYS computer software is used for this. The grid spacing is defined
by the number of grid points versus the domain lengths: Nx / Lx is the number of grid points
per unit of horizontal length, Ny / Ly is the number of grid points per unit of vertical length.
Firstly, it is advisable to implement a finer mesh at horizontal positions of the inlet. So, the
grid spacing is small at those positions and then gradually increases towards the horizontal
boundaries. The vertical grid spacing is chosen to be uniform. Secondly, as mentioned at the
end of the last paragraph, LES requires a finer mesh than RANS. As a result of those two
considerations, following grid spacing are chosen:

Nx / Lx , inlet [points/m] Nx / Lx , walls Ny / Ly

case 1 200 200  20 30
case 2 270 270  40 40
case 3 270 270  40 40
case 4 270 270  40 40

A third aspect of the situation is the definition of the boundary conditions. The first
boundary condition is the fluid velocity at the inlet. For case 1, it is assumed to be vertical (y-
direction), with a constant value of V= 10 m/s. For cases 2, 3 and 4, it is also vertical but the

3
Valentin Borisavljevic 2140173

value is a sinusoidal function of time (V = 10*(1+sin(2*π*20*t)) m/s ). The second boundary

condition is the velocity at the walls. The velocity there is 0 (known as the “no slip
condition”).

Another aspect of the situation is the time-dependence. For RANS, it is assumed that it is not
the case, so the situation is steady. For LES, this assumption is not acceptable, so the
situation is transient. For the LES cases, a time interval of 2s is considered.

Lastly, there is no heat flux within the domain, because the temperature is assumed to be
constant.

4. Results

Following paragraph discusses the results of the 4 different cases. The simulations are
realized with the FLUENT extension of ANSYS.

The comparison of cases 1 and 2 shows the difference between RANS and LES. The 1st figure
shows the RANS result, and the figure on the right shows the LES result at t=1s (middle of
the time interval). RANS gives a laminar velocity field, while LES gives a turbulent result with
recognizable vortices. This is because LES resolves more grid points (finer grid) than RANS,
which allows a more detailed result for the flow field. (Remark: the turbulence may also be
caused by the sinusoidal boundary condition at the inlet for case 2 ?)

4
Valentin Borisavljevic 2140173

The maximum velocity is 10 m/s (= inlet velocity) in case 1, while it is 14 m/s in case 2 (> max
inlet velocity of 13 m/s). The latter clearly shows the turbulent behaviour of the flow,
because a rise of the velocity above the maximum inlet velocity would not be possible with
the fluctuations that apply in a turbulent flow.

The comparison of cases 2 and 3 (case 2 is the 2nd figure) shows the effect of the outflow
boundary condition, because the position of the outflow boundary (top of the domain)
changes (different Ly).

5
Valentin Borisavljevic 2140173

The comparison of cases 2 and 4 (case 2 is the 2nd picture) shows the effect of the side
boundary condition, because the position of the side boundaries (left and right boundaries
of the domain) changes (different Lx).

6
Valentin Borisavljevic 2140173

5. Conclusions

A clear difference is visible between the RANS and LES method. Also, the influence of the
side and outflow boundary conditions can be seen.

6. References

Xia, J., Lecture notes for Advanced thermofluids. Brunel University, London (United
Kingdom).