1 Numerical Simulations
For the study of blade sweep and its impact on the aerodynamic and aeroacoustic
performance of the axial flow fans, a series of numerical simulations were carried
out for the radial, forward and backward sweep fans using GAMBIT as pre-
processor and FLUENT as solver and post-processor. As Direct Numerical
Simulation (DNS) and Large Eddy Simulation (LES) are still computationally
expensive, Reynolds Averaged Navier-Stokes (RANS) equations were employed
for the simulations in FLUENT. Fan blades were designed using NACA 0065
airfoils and have a radius of 0.2 meters. SolidWorks was used for 3D solid
modeling of three fans. A total of fifteen (15) simulations were performed for
three (forward, radial and backwards) fans at five differnt flow rates (0.675,
0.81,0.9, 0.99 and 1.125 m3 /s) corresponding to the inlet velocities of 5.37,
6.45, 7.16, 7.88 and 8.95 m/s respectively. The working fluid for all simulations
is air at STP (Standard Temperature and Pressure). The highest flow Reynolds
number (Re) corresponding to highest inlet velocity and fan diameter is 2.4E05.
1.1 Fan Test Bench Modeling
Computational domain for the simulations has been modelled similarly to the
experimental test bench of the axial flow fans in LEMFI (ENSAM Paris). We
have modelled the 1/6th part of the test bench for our simulations (like the fan
itself). The diagonal of the test bench is used as the diameter for the circular test
bench in the simulations because the circular test bench easily makes possible
the use of conditions of periodicity. In the middle of the box, there is a wall
with holes for homogenisation of the flow. Model and schematic diagram of the
test bench are shown in the figures 1 and 2.
1.2 Dimensions of the Computational Domain
Dimensions of the computational domain for the forward and backward-swept
fans are given as under,
• Entry tube length 1760 mm with a radius of 725 mm
• Exit tube length 600 mm with a radius of 700 mm
• Rotor tube length 54 mm with a radius of 222.5 mm
1.3 Symmetry and Periodicity
As the six blades of the fan are evenly and regularly arranged around the hub of
the propeller, we can use the symmetry and thus one blade is modelled instead
of all the six blades. This is shown in the figure 3. This shall save the time of
calculation and economy of resources. The side walls of the domain are periodic
and thus are included in the boundary conditions of the computational domain.
The meshes of the periodic side walls are “linked” in the GAMBIT so that they
fulfil the conditions of periodicity.
1
Figure 1: Test bench model for axial flow fans
Figure 2: Schematic of the test bench for axial flow fans
2
Figure 3: 1/6th of fan used in simulations
Figure 4: Interfaces at Entry-Rotor and Rotor-Exit
1.4 Interfaces
For all the three fans, Interfaces have been at the entry and at the exit of the
rotor domains. Interface is a surface that belongs to the two zones and therefore
is called as interface. Interfaces used in the simulation are shown in the figure
4.
1.5 Mesh details
The total mesh for forward-swept fan domain consists of total triangular cells
equal to 1590673 and more than 50 % of these cells belong to the rotor volume
because of its fine mesh. The mesh or grid quality was checked with Eqian-
gular skew, which is the default measure for checking grid quality in Gam-
bit/FLUENT. The worst element quality value was 0.8159 which is less than
the criteria fixed for these cells (which is 0.85).
3
Figure 5: Moving-walls boundary conditions for fan
1.6 FLUENT Solver
FLUENT solver selected was “Steady, Viscous, Segregated, Implicit and First
Order”.
1.7 Turbulence Model and Specification
Both k- (with Enhanced Wall Treatment) and RSM models of turbulence were
used in simulations. Turbulence specification method adopted in FLUENT sim-
ulations was Turbulent Intensity and Hydraulic Diameter. Details are given as
below,
Turbulent Intensity = 0.16(Reh )−1/8 = 5%, where Reh is Reynolds number
based on Hydraulic Diameter
Hydraulic Diameter Dh = 4Sm /Pm , where Sw is the wet surface and Pw is the
wet perimeter
Hydraulic Diameter Dh Entry = 4πR2 /2π R = 500 mm
Hydraulic Diameter Dh Exit = 1400 mm
Rotational Speed of Rotor = 314 rad/s (3000 rpm)
1.8 Common Boundary Conditions
For all the three fans, we have fixed the side walls as periodic, the upper bound-
aries as fixed walls, inlets as velocity inlets, outlets as pressure outlets, speed
of rotation for the rotor fluid is fixed as 314 rad/s (3000 rpm). The blade, hub
and support are modelled as moving walls as shown in the figure 5 while the
rest have been modelled as stationary walls.
4
2 RANS and Turbulence Models
2.1 Reynolds-Averaged Navier-Stokes (RANS)
Reynolds-Averaged Navier-Stokes (RANS) have been derived from Navier-Stokes
equations by decomposing instantaneous quantities into an average and fluctuat-
ing quantities. The three-dimensional viscous incompressible (constant density)
Navier-Stokes equations (for both laminar and turbulent flows) without gravity
terms are given as,
The x-momentum equation,
∂u ∂u ∂u ∂u 1 ∂p
+u +v +w =− + ν∇2 u (1)
∂t ∂x ∂y ∂z ρ ∂x
The y-momentum equation,
∂v ∂v ∂v ∂v 1 ∂p
+u +v +w =− + ν∇2 v (2)
∂t ∂x ∂y ∂z ρ ∂y
The z-momentum equation,
∂w ∂w ∂w ∂w 1 ∂p
+u +v +w =− + ν∇2 w (3)
∂t ∂x ∂y ∂z ρ ∂z
whereas the continuity equation is given as,
∂u ∂v ∂w
+ + =0 (4)
∂x ∂y ∂z
Here u,v and w represent instantaneous velocities in x-, y- and z-directions,
t represents time, p is the static pressure, ν is the kinematic viscosity and
∂2 ∂2 ∂2
∇2 = ∂x 2 + ∂y 2 + ∂z 2 . In the indical notation, the same equations can be
written compactly as,
1
∂t ui + uj ∂j ui = − ∂i p + ν∇2 ui (5)
ρ
with the continuity equation given as,
∂i ui = 0 (6)
The RANS equations for a three-dimensional viscous incompressible (con-
stant density) turbulent flow without gravity terms can be written as,
The x-momentum equation,
∂u ∂u ∂u ∂u 1 ∂p ∂u02 ∂u0 v 0 ∂u0 w0
+u +v +w =− + ν∇2 u − ( + + ) (7)
∂t ∂x ∂y ∂z ρ ∂x ∂x ∂y ∂z
The y-momentum equation,
5
∂v ∂v ∂v ∂v 1 ∂p ∂u0 v 0 ∂v 02 ∂v 0 w0
+u +v +w =− + ν∇2 v − ( + + ) (8)
∂t ∂x ∂y ∂z ρ ∂y ∂x ∂y ∂z
The z-momentum equation,
∂w ∂w ∂w ∂w 1 ∂p ∂u0 w0 ∂v 0 w0 ∂w02
+u +v +w =− + ν∇2 w − ( + + ) (9)
∂t ∂x ∂y ∂z ρ ∂z ∂x ∂y ∂z
whereas the continuity equation can be written as,
∂u ∂v ∂w
+ + =0 (10)
∂x ∂y ∂z
Here the overbar () denote mean quantity and a prime ()0 denotes fluctuat-
ing quantity. In the indical notation, the RANS momentum equations can be
compactly written as,
1
∂t ui + uj ∂j ui = − ∂i p + ν∇2 ui − ∂j u0j u0i (11)
ρ
whereas the continuity equation can be written as,
∂i ui = 0 (12)
2.2 Turbulence Models
The use of Reynolds averaging procedure introduces additional unknowns to
the equations. Further approximations are required to represent the fluctuating
quantities known as Reynolds stresses in order to reduce the number of un-
knowns to equal the number of equations. The Reynolds stresses -ρu0i u0j need to
be modeled through a turbulence model. Two models, the standard k- (with
Enhanced Wall Treatment) and Reynolds Stress Model (RSM) were used in our
simulations.
2.3 k- model
The k- is the most widely used general-purpose turbulence model. It is based
on the turbulent viscosity νt concept with eddy viscosity given by
k2
νt = Cµ (13)
where Cµ is a constant (0.09) and k and are obtained from differential
equations which represent the transport of turbulent kinetic energy k and rate
of dissipation . These equations are,
∂k ∂k ∂ νt ∂k ∂u
u +v = ( ) + νt ( )2 − (14)
∂x ∂y ∂y αk ∂y ∂y
6
∂ ∂ ∂ νt ∂ ∂u 2
u +v = ( ) + C1 νt ( )2 − C2 (15)
∂x ∂y ∂y α ∂y k ∂y k
Here the three terms of the right-hand side of equation 14 represent turbu-
lent diffusion, turbulent energy production and viscous dissipation respectively
where the parameters Cµ = 0.09, C1 = 1.44, C2 = 1.92, αk = 1 and α = 1.3.
These equations apply only to free shear flows. Wall damping functions were
later added to the original model of Jones and Launder (1972) to enable this
model to be used for wall-bounded shear flows too.
2.4 Reynolds Stress Model
RSM closes the Reynolds-Averaged Navier-Stokes equations by solving addi-
tional transport equation for six independent Reynolds stresses. Transport equa-
tions are derived by Reynolds averaging the product of momentum equations
with a fluctuating quantity. Closure also requires one equation for turbulent
dissipation. Additionally, the isotropic eddy viscosity assumption (used in ear-
lier models) is avoided. RSM is physically the most complete model. The exact
equation for the transport of the Reynolds Stress Rij is given as:
DRij
= Pij + Dij − ij + Πij + Ωij (16)
Dt
The equation can be read as:
-The rate of change of Rij = u0i u0j plus the transport of Rij by convection, equals
-rate of production Pij , plus
-transport by diffusion Dij , minus
-rate of dissipation ij , plus
-tranport due to turbulent pressure-strain interactions Πij , plus
-transport due to rotation Ωij
This equation describes six partial differential equations, one for the trans-
port of each of the six independent Reynolds stresses.
