You are on page 1of 18

See discussions, stats, and author profiles for this publication at: https://www.researchgate.


Performance Evaluation of Divided Intake Ducts: Effect of Area Ratio and Inlet
Reynolds Number

Article  in  International Journal of Fluid Mechanics Research · January 2003

DOI: 10.1615/InterJFluidMechRes.v30.i5.60


7 106

4 authors, including:

S. Bharani Namita Singh

Thomas & Betts, A Member of ABB Group Guru Jambheshwar University of Science & Technology


V. Seshadri
Indian Institute of Technology Delhi


Some of the authors of this publication are also working on these related projects:

Bioconversion of lignocellulosic biomass into value-added products View project

Transportation characteristics of High concentration fly ash slurries through pipelines View project

All content following this page was uploaded by S. Bharani on 06 February 2016.

The user has requested enhancement of the downloaded file.

International Journal of Fluid Mechanics Research, Vol. 30, No. 5, 2003

Performance Evaluation of Divided Intake Ducts:

Effect of Area Ratio and Inlet Reynolds Number†

Sanjeev Bharani
Department of Mechanical, Materials and Aerospace Engineering
University of Central Florida, Orlando, USA

S. N. Singh,
Indian Institute of Technology Delhi
Hauz Khas, New Delhi-110016, India

V. Seshadri and R. Chandramouli

Department of Applied Mechanics, Indian Institute of Technology Delhi
Hauz Khas, New Delhi-110016, India

Investigations carried out on divided intake ducts with 22.5◦ /22.5◦ angle-of-
turn to establish the effect of Reynolds number (Re = 6.27 · 104 to 1.88 · 105 )
and area ratios (2.0 to 4.0) using Computational Fluid Dynamics (CFD) code
are reported. The core flow remains close to the convex wall at the duct
outlet with magnitudes lower in the central region of the cross-section. It is
found that static pressure recovery is around 65% and is independent of in-
let Reynolds number for area ratio 2 duct (AR = 2). With increase in
area ratio the static pressure recovery increases to 74% for AR = 2.5, 78%
for AR = 3 and 80 % for AR = 4. The cross-velocity magnitudes at
the exit for all the cases are found to be less than 5 % of the mean flow.

* * *


AR area ratio;
A aspect ratio;
Cpr static pressure recovery coefficient, 2(Pex − Pin )/ρu2 ;
C1ε , C2ε , Cµ , σk , σε constants of the turbulence model;
Gk generation term (kinetic energy);
k turbulent kinetic energy;
M number of dependent variables;

† Received 01.11.2003

ISSN 1064-2277
2003 Begell House, Inc.
Ls duct length;
L0 distance between two inlet ducts;
P static pressure;
Pdyn(in) inlet dynamic pressure;
Pwall wall pressure;
Ri sum residuals for a dependent variable;
Sm mass added to continuous phase;
SN φ normalizing factor;
Uavi mass averaged inlet velocity;
u mean velocity;
u0 velocity perturbation;
ub bulk velocity;
V cell volume;
Vf mass flux (velocity) through the faces;
x longitudinal coordinate;
α under relaxation factor;
ε turbulence dissipation rate;
θ angle-of-turn;
ρ density of fluid;
µ dynamic viscosity;
µt eddy/turbulence viscosity;
ν kinematic viscosity.


i, j indices of tensorial notation;

ex diffuser exit;
in diffuser inlet.


The intake of a fighter aircraft must meet the engine mass flow demand in a steady and symmet-
ric manner over a wide range of aircraft speed and altitudes [1] with higher static pressure recovery
and low distortion. Divided intake ducts are widely used for ingestion of atmospheric air to the
single-engined fighter aircraft. These facilitate the diffusion of the incoming air over a short duct
length with minimal pressure loss. The intakes are normally side-mounted and the two limbs of the
duct merge inside the fuselage into one and are expected to feed air at the compressor inlet at Mach
number in the range of 0.3 to 0.5 with minimum turbulence. Due to offset positioning of the in-
takes with respect to the engine it poses a difficult design problem to tailor the S-shaped ducts while
satisfying constraints imposed by other aspects of the aircraft design. Martin and Holzhauser [2]
observed that as the intake mean flow ratio is reduced by some form of exit control, a critical point
is reached below which unequal flows develop in the two limbs of Y intake. With incompressible
flow analysis, they have shown that static pressure recovery characteristics at the junction of the two
ducts govern the flow instability and the flow reversal. They have also indicated that in a steady state
of equal flow in the two limbs even a small disturbance grows to a larger magnitude at the junction.
In [3] a simple model is used to explain the phenomenon that causes transition from symmetric to
asymmetric operation of these ducts in supersonic flights. Investigations on other aspects of such

intakes are not available in open literature, whereas several studies pertaining to single S-shaped
diffusing ducts are available [4–17].
Investigations have been carried out for S-shaped diffusing ducts to establish the effect of cur-
vature (turn angle), area ratio, shape of cross section of the duct and center line shape for laminar
and turbulent flow at the inlet. The following important features are observed.

(i) The development of secondary motion is similar to the case of constant area curved ducts [18,
19] with the magnitude of cross flow velocity reducing towards the exit [4]. The velocity
distribution is skewed towards the convex surface [14, 15].

(ii) The flow uniformity at the outlet reduces with increase in curvature where as it improves with
increase in aspect ratio [13].

(iii) A pair of contra rotating vortices exists through out the diffuser. The size of the vortices is
dependent on the area ratio and angle of turn [17].

(iv) The pressure recovery in these diffusers is low. It reduces with increase in curvature [4, 12,
13, 16].

(v) The pressure recovery coefficient and total pressure loss coefficient are relatively independent
of Reynolds number and the overall performance is comparatively much lower than straight
diffuser of similar specifications [14].

Over the last decade, computational fluid dynamics is being used extensively as a tool for analyzing
the flow in complex geometries and has also been used in diffusers [13, 14, 16, 17, 20, 21].
Studies on S-shaped ducts are not directly applicable to the divided intake ducts because of
basic difference in geometry and complexities of the flows, but have provided a good understanding
of the flow mechanism in these ducts. In the present study, an attempt has been made to establish the
effects of inlet Reynolds number and area ratio (2 to 4) on the performance of circular cross-section
22.5◦ /22.5◦ divided intake ducts.

Mathematical Model

The governing equations in the reduced form for steady and incompressible turbulent mean
flows are:

(ρui ) = Sm , (1)
· µ ¶¸
∂ui ∂P ∂ ∂ui ∂uj 2 ∂ul ∂ ³ ´
ρuj =− + µ + − δij + −ρu0i u0j , (2)
∂xj ∂xi ∂xj ∂xj ∂xi 3 ∂xl ∂xj
where i, j = 1, 2, 3.
Additional terms −ρu0i u0j , in the momentum equations are modeled using the Renormalization
Group Theory (RNG) k-ε turbulence model for the closed form solution. The model improves the
predictions for wall bounded flows having curved geometries [22]. The Boussinesq hypothesis is
used to relate the Reynolds stresses to the mean velocity gradient as given below:
µ ¶ µ ¶
0 0 ∂ui ∂uj 2 ∂ui
−ρui uj = µt + − ρk + µt δij , (3)
∂xj ∂xi 3 ∂xi

where µt is the eddy viscosity and k is the turbulence kinetic energy and δij is the Kronecker delta.

Two additional transport equations, one for turbulent kinetic energy (k) and the other for the
turbulence dissipation rate (ε) are solved to evaluate µt , which is computed as,

µt = ρCµ k 2 /ε, (4)

where Cµ is a constant.
The additional equations for k and ε for steady incompressible flow in simplified form in the
absence of the thermal gradients are:
· ¸
∂k ∂ ∂k
ρui = αk µeff + Gk − ρε, (5)
∂xi ∂xi ∂xi

· ¸
∂ε ∂ ∂ε ε ρε2
ρui = αε µeff + C1ε (Gk ) − C2ε − R, (6)
∂xi ∂xi ∂xi k k
where, Gk is the generation of turbulent kinetic energy due to the mean velocity gradient, calculated
as Gk = µt S 2 . Here, S is the modulus of the mean rate of shear stress tensor, defined as,
S = (2Sij Sij ) , (7)

Sij is given by,

µ ¶
1 ∂ui ∂uj
Sij = + , (8)
2 ∂xj ∂xi
C1ε and C2ε are constants. αk and αε are the inverse effective turbulent Prandtl numbers for k
and ε, respectively. The effective viscosity is modelled in the RNG theory using scale elimination
procedure resulting in a differential equation for turbulent viscosity as

ρ2 k ν̂dν̂
d √ = 1.72 √ 3 , (9)
εµ ν̂ − 1 + cν̂

where ν̂ = µeff /µ and cν̂ ≈ 100.

In the high-Reynolds-number limit Eq. (9) gives

µt = ρcµ , (10)
with Cµ = 0.0845, the effective viscosity is calculated using Eq. (10).
The additional source term (R) in the ε-equation is,
" µ ¶2 #µ ¶
3 η ±¡ 3
¢ ε2
R = Cµ ρη 1 − o 1 + βη k , (11)
η k

where, η = Sk/ε; η o = 4.38; β = 0.012. The value of constants in the turbulence model used in the
present study are the standard values reported in literature (C2ε = 1.42, C1ε = 1.68, Cµ = 0.0845,
σk = σε = 0.72). The RNG k-ε model has been used as the additional term “R” in ε equation
improves the accuracy of the solution significantly besides being more responsive to the effects of
rapid strain and streamline curvature.

100 1 - 0o section
2 2 - 30o section
50 3 - 90o section

30o 4 - 90o/60o section
INLET 5 - 90o/90o section
60o 100


50 382 50

30 296.6 30


Rc 573.0

O 52.5
O 72.5

Rc 573.0




Fig. 1. Geometry of S-diffusers: a) 90◦ /90◦ rectangular-square, b) 15◦ /15◦ circular-circular.

Validation of the Code

The geometry for validation of the FLUENT code [23] is the same as that used by [12] and [16]
as shown in Fig. 1a and b, respectively. The experimental results reported are compared with the
predictions using air as working fluid. The 90◦ /90◦ rectangular-square diffuser geometry of [12]
(Fig. 1a) was generated using the CFD code with boundary conditions, grid and meshing scheme
as given in the Table. The boundary conditions at inlet are specified by assigning known values of
the variables. At the outlet, the continuity has to be satisfied and hence the boundary condition is
specified in a manner so that velocity is directed outwards. Wall boundary conditions are specified

Test cases for validation of CFD code.

Majumdar et al. [12] Singh et al. [16]

Inlet velocity 40 m/sec 26.5 m/sec
Cross-section rectangular-square circular-circular
Inlet Reynolds number 8.21 · 105 0.95 · 105
Working fluid air air
Element shape hexahedral tetrahedral
Number of volume cells 23100 67136

at the walls as per the direction given in [23]. The mean velocity at the exit plane at three stations
is compared with the experimental data as shown in Fig. 2a. The wall static pressure distribution is
depicted in Fig. 2b. It is seen that predicted mean velocity compares reasonably well (±0.1u/Uavi ).
Similarly the pattern of static pressure along the walls is well predicted with minor differences at
few locations.
Validation of the code was further established by comparing (Fig. 3) the predicted static pres-
sure recovery and total pressure loss for the 15◦ /15◦ circular diffuser investigated by Singh et
al. [16]. The details of the number of cell volumes and type of mesh for the circular diffuser are
given in the Table. Fig. 3 shows that the experimental and predicted pressure recovery coefficient
closely matches up to ≈ 12◦ turn in the first bend after that the experimental pressure recovery
obtained is lower than the predicted values. At the exit of the diffuser, the predicted pressure re-
covery obtained is ≈ 46 % whereas the experimental value reported is ≈ 38 %. The predicted total
pressure loss also closely matches the experimental values in the initial portion but subsequently
the predicted values are lower than the experimental values. The higher values of pressure recovery
and lower values of total pressure loss by prediction can be attributed to the wall roughness present
in the experimental investigations. Incorporation of wall roughness factor (ε/D) of the order of
0.01 a provision which is provided in the CFD code FLUENT [23] into the predictions improved
the matching considerably. Comparison of velocity contours for 22.5◦ /22.5◦ for S-shaped diffuser
with work of Rojas et al. [4] has been shown elsewhere [17], depicts reasonably good matching.
On the basis of a reasonable matching of the predicted and the reported results on the similar type
of geometrical configurations, the CFD code FLUENT [23] can be considered to be suitable for
analyzing flow characteristics in divided intake ducts.

Geometries Investigated and Boundary Conditions

Flow predictions have been made for 22.5◦ /22.5◦ divided intake ducts having circular cross-
section to understand the effects of inlet Reynolds number and area ratio on their perfomance. The
schematic layout of these ducts created in x-y plane along with major dimensions is shown in Fig. 4.
The figure also depicts the planes selected for presentation of the results. The iso-velocity contours,
static pressure and total pressure distribution are analyzed at planes P1, P2, P3 (flow mixing plane),
P4 and P5. These planes are selected at the intersection of curvature of centerline and the angle-
of-turn. The cross-velocity plots are presented at S1, S2, S3, S4, S5 and exit plane, which are
perpendicular to the flow along the length of one limb of the divided entrance intake duct. Cross-
velocity plots only at the outlet planes are presented in this paper for the sake of brevity. The
diameter at the inlet for the limbs was selected as 48 mm. A constant area duct of 50 mm length
was provided at the upstream. The exit diameter was dependent on the area ratio selected. The

ST-9 ST-6 ST-3
a) 1 1 1

0.8 0.8 0.8

0.6 0.6 0.6



0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
0 0.5 1 0 0.5 1 0 0.5 1
U / Uavi U / Uavi U / Uavi

b) 0.2

Pwall / Pdyn(in)


CVCC wall - experimental
CCCV wall - experimental
bottom wall - experimental
-0.6 CVCC wall - computional
CCCV wall - computional
bottom wall - computional

0 30 60 90 90/30 90/60 90/90
Angle of turn (degrees)

Fig. 2. Comparison of experimental [12] and predicted values in 90◦ /90◦ rectangular S-diffusers
with AR = 2: a) longitudinal velocity magnitudes (solid – computed, markers – expeimental),
b) wall static pressure distribution (CVCC – Convex Concave Wall, CCCV – Concave Convex Wall).

diameters for the corresponding area ratios are given in Fig. 4. A constant area duct of 150 mm
length was provided downstream for uniform, smooth and continuous flow conditions [4, 24]. Both
the intake limbs were symmetrically designed about the axis of the downstream outlet duct with
radius of curvature for each as 382.0 mm with angle-of-turn (θ) = 22.5◦ /22.5◦ .
Three-dimensional diffuser geometries were developed using the GAMBIT package of the CFD
code and were meshed with 53192 tetrahedral cell volumes for AR = 2.0, 53487 for AR = 2.5,
58450 for AR = 3.0 and 64567 for AR = 4.0 ducts. These cell volumes were arrived by doing a

pressure recovery - computational
total pressure loss - computational
40 pressure recovery - experimental

Coefficient (%)
total pressure loss - experimental




0 7.5 15 15/7.5 15/15
Angle of turn (degrees)

Fig. 3. Comparison of experimental (Singh et al., [16]) and predicted pressure recovery coefficient
static pressure recovery along the length of 15◦ /15◦ S-diffuser with AR = 1.9 (smooth wall).

50 Ls 150

P1 P2 P3
Od S1 S2


S4 S5

Od Rc
/2 P4 P5

Fig. 4. Geometry of divided intake duct.

grid independence check with respect to the pressure recovery coefficient prediction. Fig. 5 shows
the grid independence check for the AR = 2.5, as a representative case. Air as the working fluid
is fed at the inlet planes with uniform velocity of 40 m/sec. It is seen that increasing the number
of cells from 53487 to 82663 does not change the predicted pressure recovery trend and the final


Coefficient (%)


20 pressure recovery for mesh with 82663 cells

pressure recovery for mesh with 53487 cells

0 11.25 22.5 22.5/11.25 22.5/22.5
Angle of turn (degrees)

Fig. 5. Mesh independence check for divided intake duct with AR = 2.5.

Flow analysis is carried out for AR = 2 duct at Re = 6.27 · 104 (20 m/sec), Re = 1.25 · 105
(40 m/sec) and Re = 1.88 · 105 (60 m/sec) to establish the effect of Reynolds number. There after
further investigations are carried out with Re = 1.25 · 105 (40 m/sec) for ducts with area ratios of
2.5, 3.0 and 4.0. Longitudinal and cross-velocity contours along with the static pressure recovery
and total pressure loss coefficients are presented for each case.

Results and Discussion

The results are presented and analyzed to clearly bring out the effect of parameters mentioned
in the previous section.
Static Pressure Recovery and Pressure Loss in AR = 2 Intake Ducts. The static pres-
sure recovery trends at Reynolds number, Re = 6.27 · 104 , Re = 1.25 · 105 and Re = 1.88 · 105
corresponding to the flow velocities of 20 m/sec, 40 m/sec and 60 m/sec showed that the effect of
Reynolds number is negligible. Hence, only single curve is given for pressure recovery (Fig. 6) as
the maximum deviation was only around 2 % in the range of Reynolds numbers covered.
At all the Reynolds numbers it is seen that the pressure recovery increases linearly almost up to
the end of first turn, with values close to 64 %. Thereafter it reduces by around 4 %, in the second
bend up to 11.25◦ and recovers back to a value of approximately 65 ± 1 % at the outlet plane. The
static pressure recovery obtained with these ducts is approximately 10 % less than the ideal pressure
recovery. In the initial length of the duct, increase in pressure recovery is due to the increase in the
flow passage area. The growth of boundary layer in this length of passage is negligible because of
which rapid pressure recovery takes place. The fall in pressure recovery in the second bend could be
due to the growth of boundary layer as a result of separation and mixing losses at the inflexion plane
(plane P3). Further increase in area of cross-section does not lead to diffusion as the effective area



Coefficient (%)





0 11.25 22.5 22.5/11.25 22.5/22.5
Angle of turn (degrees)

Fig. 6. Pressure recovery (solid) and total pressure loss coefficient (dashed)
at all inlet Reynolds numbers in the velocity range 20 – 60 m/sec.

of flow does not increase due to the growth of boundary layer. The total pressure loss registered
for the three inlet flows increases almost linearly along the length of duct at a much slower rate
compared to pressure recovery and is nearly identical for all the three Reynolds numbers and hence
only single curve is given for total pressure loss (Fig. 6). The values for each case remained almost
same at around 8 ± 1 %.
Velocity Distribution in AR = 2 Intake Ducts. Longitudinal velocity distribution at differ-
ent cross-sectional planes of the 22.5◦ /22.5◦ intake duct for the three Reynolds numbers are shown
in Fig. 7 as iso-velocity contours.
For Re = 6.27 · 104 , the longitudinal velocity distribution is uniform at the inlet plane. As
expected, at plane P1 the core velocity shifts towards the convex wall having a higher velocity
(20.24 m/sec). At plane P2, close to the inflexion plane, core flow shifts back to the center as a result
of the centrifugal force. The velocity magnitude also reduces which shows that diffusion is taking
place. Similar pattern of flow has also been reported by Dey et al. [17] in S-shaped circular cross-
section diffusers. The reduction in the velocity magnitude is due to the diffusion of the flow with
increase in the area of cross-section of the duct. At the mixing region (plane P3), the flow from both
the limbs mix but retain their identity. The core flow shifts towards the convex walls of the second
bend as a result of change in the direction of centrifugal force. Similar velocity distribution remained
till plane P5 with size of the core flow covering a major area of cross-section (11.45 to 11.04 m/sec).
The velocity distribution from plane P3 to P5 also shows that the longitudinal velocity magnitude
at the center of the duct cross-section remains lower. This may be due to the streamlines coming
from the two limbs forming concave streamline surfaces parallel to the concave walls. With the
increase in the Reynolds number values, the pattern of velocity distribution does not change. It
is similar to the previous case with higher velocities at each plane. The core velocity magnitudes
at plane P4 increases to 25.33 m/sec and 37.99 m/sec for Re = 1.25 · 105 and Re = 1.88 · 105 ,

Fig. 7. Longitudinal velocity contours for 22.5◦ /22.5◦ duct with AR = 2 at Re = 6.27 · 104 .

Fig. 8. Cross-velocity plots for 22.5◦ /22.5◦ duct with AR = 2 at Re = 6.27 · 104 .

Fig. 9. Longitudinal velocity contours for 22.5◦ /22.5◦ duct for AR = 2.5 at Re = 1.25 · 105 .

respectively. The core velocities at plane P5 are 22.87 m/sec and 34.31 m/sec, respectively. The
velocity magnitudes in the center are about 8 – 9 % lower compared to the core velocities. It was
also seen that with the increase in the Reynolds number the flow pattern at the exit does not change.
There is only a corresponding increase in the outlet velocity magnitudes. The cross-flow velocities
observed at planes S1 to S5 for Re = 6.27 · 104 showed the presence of pair of contra-rotating
vortices up to plane S3, which is also observed in S-diffusers. At plane S4, which is the mixing
plane for the flow from the two limbs, formation of two pairs of contra-rotating vortices is observed,
which continues to exist even at plane S5, with reduced magnitude. It was also observed that the
flow pattern for cross-flow velocities does not change with Reynolds number, the change being only
in relative magnitudes of cross-flow velocities. For sake of brevity, cross-flow velocities in form of
vector plots are shown for planes S4 and S5 only for the Reynolds number = 6.27 · 104 in Fig. 8.
Effects of Area Ratio on Velocity Distribution. The studies on divided intake duct of AR =
2 at different Reynolds numbers has shown that the performance is not significantly affected by
Reynolds number in the range over which investigations are carried out. Hence studies on higher
area ratio ducts is performed only at one Reynolds number corresponding to an inlet velocity of
40 m/sec. The longitudinal velocity pattern through out the duct for different area ratio ducts is
similar to the pattern observed for the AR = 2 (Fig. 7). Similar to the previous case (AR = 2)
lower velocity magnitudes are present in the central region at plane P5 compared to the core flow
velocity magnitudes towards the top and bottom side. The velocity magnitudes at the central region
are lower by approximately 9 % for AR = 2, 12.8 % for AR = 2.5, 21 % for AR = 3, 32.5 % for
AR = 4. This shows that flow uniformity at the exit plane reduces with increase in area ratio. For
sake of brevity only longitudinal velocity distribution obtained at different cross-sectional planes
for AR = 2.5 divided intake duct is depicted in Fig. 9.

a) b)

c) d)

Fig. 10. Longitudinal and cross-velocity contours at the exit plane for Re = 1.25 · 105 :
a) AR = 2, b) AR = 2.5, c) AR = 3, d) AR = 4.

Coefficient (%)
pres. recovery (AR=2.0)
pres. recovery (AR=2.5)
40 pres. recovery (AR=3.0)
pres. recovery (AR=4.0)
total pres. loss (AR=2.0)
total pres. loss (AR=2.5)
20 total pres. loss (AR=3.0)
total pres. loss (AR=4.0)

0 11.25 22.5 22.5/11.25 22.5/22.5
Angle of turn (degrees)

Fig. 11. Pressure recovery coefficient and total pressure loss coefficient
for 22.5◦ /22.5◦ duct function of area ratio (Re = 1.2 · 105 ).

For better understanding of the flow in these ducts, the longitudinal and cross-velocity plots at
the exit plane are shown in Fig. 10. It is seen that the core flow is always located close to the top
and bottom walls of the duct with lower values in the central region as observed for the planes P3
to P5. Compared to plane P5, it is seen that the fall in central region flow velocity has reduced from
9 % to 3.8 % for AR = 2; 12.8 % to 7 % for AR = 2.5, 21 % to 13.1 % for AR = 3 and 32.5 %
to 30 % for AR = 4. This implies that the flow continues to develop beyond plane P5 and achieves
higher flow uniformity up to the exit plane, the development being more pronounced up to AR = 3
The cross-velocity vectors at the exit plane indicate four vortices symmetrically placed in each
quadrant. These vortices are formed due to secondary flow induced by pressure gradients resulting
from the streamline curvature. The pressure gradient and the deflection of transverse vorticity com-
ponent of the boundary layer causes secondary flows in the form of two pairs of counter-rotating
vortices [26]. For AR = 2, the vortices have magnitudes approximately 0.07 m/sec at the center
and 0.28 m/sec in the outer region. The velocity magnitudes in the wall region are higher (1.3
to 0.72 m/sec). With the increase in the area ratio, the variation between the core velocity of the
vortices to the outer region is reduced and is in the range of 0.14 to 0.29 m/sec for AR = 2.5
duct and 0.13 to 0.27 m/sec for AR = 3 duct. For AR = 4 duct, the cross-flow velocity distri-
bution shows a tendency of reducing to one pair of contra-rotating vortices from two pairs. This
reformation may be attributed to the increase in the flow diffusion with increase in the area ratio. A
relook at the velocity magnitudes shows that the percentage of cross-velocity magnitudes relative
to the longitudinal velocity magnitudes are 1.3 % for AR = 2.0 duct 1.7 % for AR = 2.5 duct,
1.75 % for AR = 3 duct and 1.72 % for AR = 4 duct. The secondary flow velocity magnitudes are
always less than 5 % of the longitudinal velocity, which suggests that the flow is moving into the
compressor without much distortion [25].

Effect of Area Ratio on Static Pressure Recovery. Static pressure recovery plot along the
length of the divided intake duct as a function of area ratio is plotted in Fig. 11 for a Reynolds
number 1.2 · 105 . The figure also depicts the total loss coefficient in these ducts. It is observed that
pressure recovery increases almost linearly up to 22.5◦ turn with magnitudes being relatively higher
for higher area ratio. The rate of increase in pressure recovery reduces in the second bend due to the
development of boundary layer and mixing losses. The effective area of diffusion reduces due to
growth of boundary layer, reduction being maximum for AR = 2 duct and minimum for AR = 4
duct. One even observes a fall in pressure recovery for AR = 2 and 2.5 duct indicating reduction
in flow area. For AR = 2 intake duct the pressure recovery coefficient is always the lowest with
maximum recovery of 65 %. The maximum values for other area ratios are 74 % for AR = 2.5 duct
78 % for AR = 3 duct and 80 % for AR = 4 duct. On comparing with the ideal pressure recovery,
the predicted values are 10 % lower for AR = 2 duct and 2.5 duct which increases to 13.75 % for
AR = 4 duct. The total pressure loss coefficient varies almost linearly along the length of intake
ducts. Highest total pressure loss (about 10 %) is obtained for AR = 4 duct, whereas for other ducts
it is observed to be lower.


Based on the investigations carried out on divided intake ducts at three Reynolds numbers and
four area ratios, the following conclusions are drawn.

1. The pressure recovery in intake duct of AR = 2 is observed to be independent of the inlet

Reynolds number. The total pressure loss coefficient is also observed to be independent of
Reynolds number.

2. For all Reynolds numbers, pressure recovery increases with increase in area ratio for divided
intake geometries. For increase in the area ratio from 2 to 4, the static pressure recovery
increases from 65 % to 80 %. The pressure loss coefficient also increases with increase in
area ratio but the increase is very marginal.

3. The core flow remains close to the convex wall in each limb of the intake duct. Even after
merger of the flow from the two limbs, the individual core flows remain distinct and positioned
towards the top and the bottom wall of the duct till the exit plane.

4. The cross-velocity plots show two pairs of contra-rotating vortices within the limbs with
higher values of velocity along the duct walls. From the plane of merging of two limbs,
three vortices are seen up to the exit plane. The central region in the duct has the lowest
cross-velocity magnitudes.

5. The magnitudes of the cross-velocity are higher for the AR = 4 intake duct, whereas for
the other area ratio ducts the values for cross-velocity are almost comparable. These ducts
showed less than 5 % distortion at the exit plane.


1. Whitford, R., Design for Air Combat, Jane’s Publishing Co., London, 1987.
2. Martin, N. J. and Holzhauser, C. A., Analysis of Factors Influencing the Stability Characteristics
of Symmetrical Intake Air-Induction Systems NACA Tech. Rept, 1950.

3. Sudhakar, K. and Ananthkishanan, N., Jump Phenomena in Y-Shaped Intake Ducts, J. Aircraft,
1995, 33, No. 2, pp. 438–439.
4. Rojas, J., Whitelaw, J. H., and Yianneskis, M., Flow in Sigmoid Diffusers of Moderate Cur-
vature, Rept No. FS83/21, Mechanical Engineering Department, Imperial College of Science,
Technology and Medicine, London, UK, 1983.
5. Guo,R. W. and Seddon, J., Swirl Characteristics of an S-Shaped Air Intake with both Horizontal
and Vertical Offsets, Aeronaut. Quart., 1983, 34, pp. 130–146.
6. Guo, R. W. and Seddon, J., The Swirl S-Duct of Typical Air Intake Proportions, Aeronaut.
Quart., 1983, 34, pp. 99–129.
7. Seddon, J., Understanding and Countering the Swirl in S-Ducts: Tests on the Sensitivity of
Swirl to Fences, Aeronaut. J., 1984, 88, pp. 117–127.
8. Shimizu, Y., Nagafusa, M., Sugino, K., and Kubota, T., Studies on Performance and Internal
Flow of U-Shaped and Snake-Shaped Bend Diffusers: 2-nd Report, Trans. ASME, J. Fluids
Engng, 1986, 108, pp. 297–303.
9. Shimizu, Y., Nagafusa, M., Sugino, K., and Nakamura, F., Studies on Performance and Internal
Flow of Twisted S-Shaped Bend Diffuser.– The So-Called Coiled Bend Diffuser: 1-st Report,
Trans. ASME, J. Fluids Engng, 1986, 108, pp. 289–296.
10. Whitelaw, J. H. and Yu, S. C. M., Turbulent Flow Characteristics of S-Shaped Diffusing Ducts,
Flow Measurement and Instrumentation, 1993, 4, No. 3, pp. 171–179.
11. Lien, F. S. and Leschziner, M. A., Computational Modeling of 3-D Turbulent Flow in S-Diffuser
and Transition Ducts, In: Engineering Turbulence Modeling and Experiments, Rodi, W. and
Martelli, F., Eds., Elsevier Science Publishers, Oxford, 1993.
12. Majumdar, B., Singh, S. N., and Agrawal, D. P., Flow Characteristics in S-Shaped Diffusing
Duct, Int. J. Turbo and Jet-Engines, 1997, 14, pp. 45–57.
13. Vinit Gupta, Rajneesh Devpura, Singh, S. N., and Seshadri, V., Effect of Aspect Ratio and Cur-
vature on Characteristics of S-Shaped Diffusers, Indian J. Eng. Mater. Sci., 2001, 8, pp. 141–
14. Anand, R. B., Lajpat Rai, and Singh, S. N., Flow and Performance Characteristics of a
22.5◦ /22.5◦ S-Shaped Circular Diffuser, In: Proc. 28-th Nat. Conf. Fluid Mechanics and Fluid
Power, Chandigarh, India, December, 13–15, 2001, pp. 364–371.
15. Anand, R. B., Lajpat Rai, Singh, S. N., and Sharma, O. P., Flow Characteristics of a Low Aspect
Ratio 90◦ /90◦ S-Shaped Diffuser, J. Aero. Soc. India, 2000, 53 No. 4, pp. 239–252.
16. Singh, S. N., Seshadri, V., and Anand, R. B., Flow Characteristics of S-Shaped Diffusing Ducts,
Closure Rept, AERO/RD-134/100/98-99/1035, Dept Applied Mechanics, Indian Institute of
Technology Delhi, March 2002.
17. Dey, R., Bharani, S., Singh, S. N., and Seshadri, V., Flow Analysis in S-Shaped Diffusers with
Circular Cross-Section, Arab. J. Sci. Eng., December 2002, 27, No. 2c, pp. 197–206.
18. Taylor, A. M. P. K., Whitelaw, J. H., and Yianneskis, M. Developing Flows in S-Shaped Ducts.
Part 1 – Square Cross-Section, Rept No. FS/81/22, Mechanical Engineering Department, Im-
perial College of Science, Technology and Medicine, London, UK, 1981.
19. Taylor, A. M. P. K., Whitelaw, J. H., and Yianneslis, M. Developing Flows in S-Shaped Ducts.
Part 2 – Circular Cross-Section NASA Contract Rept 3759, USA, 1984.
20. Mayer, D. W. and Kneeling, W. D. Evaluation of Two Flow Analyses for Subsonic Diffuser
Design, In: 30-th Aerospace Sciences Meeting and Exhibit. (AIAA), Pap. No. AIAA 92 0273,
Reno, NV, 6–9 January 1992.
21. Bruns, J. E. and Smith, C. F. Full Navier – Stokes Calculations on the Installed F/A-
18 Inlet at a High Angle of Attack, 28-th Joint Propulsion Conference and Exhibit.
(AIAA/SAE/ASME/ASEE), Pap. No. AIAA 92 3175, July 6–9, 1992.

View publication stats

22. Pope, S. B., Turbulent Flows, Cambridge University Press, Cambridge, UK, 2000.
23. FLUENT-5, Fluent Incorporation, USA, 1998.
24. McMillan, O. J. Mean-Flow Measurements of the Flow Field Diffusing Bend, NASA Contract
Rept 3634, 1982.
25. Yung, C. N., Keith, T. G., and De Witt, K. J., Computational Three-Dimensional Turbulent Flow
in S-Shaped Ducts, In: Proc. Sixth Int. Conf. Numerical Methods in Laminar and Turbulent
Flows, Swansea, 11–15 July, 1989, pp. 1377–1387.
26. Harloff, G. J., Reichert, B. A., Sirbaugh, J. R., and Welborn, S. R. Navier – Stokes Analysis
and Experimental Data Comparison of Compressible Flow in Ducts NASA, TM-105796, July