You are on page 1of 9

Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067

www.springerlink.com/content/1738-494x(Print)/1976-3824(Online)
DOI 10.1007/s12206-016-0412-2

Optimization of a subsonic wind tunnel nozzle with low contraction ratio via
ball-spine inverse design method†
Hadi Hoghooghi1, Mahdi Nili Ahmadabadi1 and Mojtaba Dehghan Manshadi2,*
1
Department of Mechanical Engineering, Isfahan University of Technology, Isfahan, 84156-83111, Iran
2
Department of Mechanical and Aerospace Engineering, Malek-Ashtar University of Technology, Isfahan, 83145-115, Iran

(Manuscript Received September 25, 2015; Revised November 19, 2015; Accepted December 25, 2015)

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract

The goal of wind tunnel design is to generate a uniform air flow with minimum turbulence intensity and low flow angle. The nozzle is
the main component of wind tunnels to create a uniform flow with minimal turbulence. Pressure distribution along nozzle walls directly
affects the boundary layer thickness, pressure losses and non-uniformity of flow velocity through the test section. Although reduction of
flow turbulences and non-uniformity through the test section can be carried out by nozzles with high contraction ratio, it increases the
construction cost of the wind tunnel. For decreasing the construction cost of nozzle with constant test section size and mass flow rate, the
contraction ratio and length of nozzle should be decreased; that causes the non-uniformity of outlet velocity to increase. In this study, first,
three types of nozzle are numerically investigated to compare their performance. Then, Sargison nozzle with contraction ratio of 12.25
and length of 7 m is scaled down to decrease its weight and construction cost. Having scaled and changed to a nozzle with contraction
ratio of 9 and length of 5 m, its numerical solution reveals that the non-uniformity of outlet velocity increases by 21%. By using the Ball-
spine inverse design method, the pressure distribution of the original Sargison nozzle is first scaled and set as the target pressure of the
scaled down nozzle and geometry correction is done. Having reached the target nozzle, numerical solution of flow inside the optimized
nozzle shows that the non-uniformity just increases by 5% in comparison with the original Sargison nozzle.
Keywords: Subsonic wind tunnel; Nozzle; Contraction ratio; Adverse pressure gradient; Flow non-uniformity; Inverse design method
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

rated, flow quality to decrease and flow losses to increase.


1. Introduction
Thus, the pressure gradient in these areas should be examined
The nozzle is the main component of wind tunnels for creat- after design. If the non-uniformity of velocity at any cross-
ing a uniform flow with minimal turbulence. Nozzle design section of the test section is less than 0.2 percent of its average
should be based on certain principles under special attention. value in the same cross-section, the air flow is favorable. In
The main task of a nozzle is accelerating air flow, slowing these circumstances, the dynamic pressure non-uniformity in
down to become comfortable in the settling chamber and in- the cross-section should be less than 0.4 to 0.6 percent. Also,
crease the required velocity through the test section. an air deviation angle of 0.1 degree is acceptable [2]. One of
Because the nozzle cross-section is gradually reduced, the the important aspects in the design of a wind tunnel is the
outlet flow of the nozzle has less non-uniformity and turbu- opposite problem between performance and cost of wind tun-
lence. Shape and dimensions of a nozzle determine not only nel construction. Quality of test section during test determines
the velocity, but also the non-uniformity of air flow at outlet the performance of the wind tunnel. Energy losses through the
[1]. The nozzle shape and size directly affects its energy losses. wind tunnel have also an important role in determining the
The most of these energy losses refer to wall friction. The power consumption and efficiency of the wind tunnel.
nozzle profile is designed to reach a uniform flow distribution Reducing flow turbulences in a test section can be obtained
at the nozzle outlet. Air velocity along the nozzle can be by nozzles with high contraction ratio. Although a great set-
changed in a range through which separation does not happen tling chamber causes the high contraction ratio and reduces
[1]. The adverse pressure gradient near the inlet and outlet the flow non-uniformity, it increases the construction cost of
zones of nozzle walls causes the boundary layer to be sepa- the wind tunnel [3].
Bradshaw [3] showed that the pressure gradient and type of
*
Corresponding author. Tel.: +98 3145225044, Fax.: +98 3145225044 convergence and divergence duct affected the growth rate of
E-mail address: mdmanshadi@mut_es.ac.ir

Recommended by Associate Editor Kyu Hong Kim
boundary layer thickness. This is a phenomenon of curvature
© KSME & Springer 2016 effects. For more study of this phenomenon, Saddoughi [4]
2060 H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067

defined the convergence parameters. ing the wall frequently. They developed this method for in-
Ardekani [5] changed the parameters of a vertical wind tun- verse design of airfoils in subsonic and transonic external flow
nel and tried to reduce the length of the nozzle and cost of regimes [20]. Overall, the BSA is a physical based inverse
construction with little change in the quality of the outlet flow. design algorithm that changes a fluid dynamic problem to a
One of the optimal shape design methods for nozzles is the fluid solid interaction one. Thus, its convergence rate is very
surface shape design that involves finding a shape associated suitable. Moreover, in comparison with the other methods, it
with a prescribed distribution of surface pressure or velocity. is a simple algorithm that can be easily incorporated into any
There are basically two different algorithms for solving sur- open source or commercial flow solver as a block box.
face shape design problems: decoupled (iterative) and coupled In this study, first, different methods for design of wind
(direct or non-iterative) techniques. In coupled solution ap- tunnel nozzle are presented and discussed. By applying the
proach an alternative formulation of the problem is used in presented design methods, some nozzles are designed and
which the surface coordinates appear (explicitly or implicitly) then, numerically studied. Then, Ball-Spine inverse design
as dependent variables. method is used for optimization of a nozzle with low contrac-
The iterative shape design approach relies on repeated tion ratio to prevent increasing the non-uniformity of outlet
shape modifications such that each iteration consists of flow velocity. In this approach, the pressure distribution of Sargison
solution followed by a geometry updating scheme. In other nozzle with contraction ratio of 12.25 and low non-uniformity
words, a series of sequential problems are solved in which the is scaled down and set as the target pressure of a nozzle with
surface shape is altered between iterations so that the desired contraction ratio of 9. Then, geometry correction is done to
pressure distribution is finally achieved. reach a small nozzle with acceptable flow uniformity at its
One of the iterative shape design methods is based on the outlet. Finally, the performance of the designed nozzle is stud-
residual-correction approach. In this method, the key problem ied by numerical solution.
is to relate the calculated differences between the actual pres-
sure distribution on the current estimate of the geometry and
2. Principles and methods of nozzle design
the target pressure distribution (the residual) to required
changes in the geometry. The nozzle is the most important component of a wind tun-
Garbedian-McFadden [6, 7] presented an iterative inverse nel, directly affecting its outlet flow quality. Some considera-
design method based on a mathematical approach called flexi- tion, as follows, should be satisfied in nozzle design.
ble membrane method (GM design). This method was later (1) The nozzle should be designed such that a uniform flow
modified by Malone et al. [8-10], presently known as MGM velocity is created at its outlet. Thus, velocity or pressure gra-
(modified Garbedian-McFadden or Malone-Garbedian-McFad- dients should not cause the growth of boundary layer thick-
den) technique. In this mathematical approach the surface of ness or flow separation [1].
an aerodynamic body is modeled as a membrane that deforms (2) In nozzles, it is preferred the area ratio of inlet to outlet
under aerodynamic loads. Dulikravich [11] presented an in- is high so that outlet flow quality increases and turbulence
verse design method based on an analytical Fourier series intensity decreases. For a wind tunnel with low turbulence
solution for MGM equation. The method was successfully intensity, the contraction ratio (c) should be more than 12, but
tested at subsonic and transonic flow regimes for both airfoils nowadays for decreasing the cost of construction, this ratio
and wings. usually changes from 4 to 12 [2].
Barger et al. [12] presented the streamline curvature method The major loss in nozzles is due to friction. According to
that attributes the changes in surface curvature to the change the fundamentals of fluid mechanics, the friction loss can be
in flow velocity. Campbell et al. [13] used this method for full calculated from the following equation [2]:
potential equation, Bell et al. [14] for Euler equation and
Malone et al. [15] for Navier-Stokes equations. L
Nili et al. presented a physical algorithm for internal flows K n = 0.32 f a v , (1)
Dsc
in which the duct wall is considered as a flexible string that
deforms under the difference between target and current pres-
sure distribution. They developed this method for non-viscous where fav is the average friction coefficient, L is the length of
compressible [16, 17] and viscous incompressible internal nozzle and Dsc is the hydraulic diameter of settling chamber
flow regimes [18]. [2].
Recently, Nili et al. [19] developed a novel inverse design In a subsonic wind tunnel, there are two regions with ad-
method called Ball-spine algorithm (BSA) for quasi-3D de- verse pressure gradient that are located near the beginning and
sign of centrifugal compressor meridional plane. In BSA, the end of the nozzle. Indeed, the primary region has a lower ad-
passage walls are composed of a set of virtual balls that freely verse pressure gradient and is hence less effective in separa-
move along specified directions called spines. The difference tion and uniformity of the flow in test section [21].
between target and current pressure distribution at each modi- Pressure coefficient for wind tunnel walls can be defined as
fication step is applied to each ball as an actual force deform- the following equation:
H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067 2061

Rouse: L=7m, Ds=3.5m, i/L=0.6


P - P¥ ( P - P0 ) - ( P¥ - P0 ) P + q¥
Cp = = Þ C p = gage , (2) 1.8
Bell: L=7m, Ds=3.5m
Sargison: L=7m, Ds=3.5m, i/L=0.6
q¥ q¥ q¥
1.7
1.6
where p is wall pressure, P0 is total pressure, p∞ is static pres- 1.5
1.4
sure and q∞ is the air dynamic pressure at the test section of
1.3
wind tunnel. 1.2

Y (m)
Nozzles can be usually designed by two curves, internal 1.1
curve and external curve, with a turning point at xi. The equa- 1
0.9
tion of these two curves for the first time was presented by
0.8
Rouse [1] as follows: 0.7
0.6
0.5
R D D ( x / L )3 0.4
0 £ x £ xi = s - [ s - 1] 0 1 2 3 4
L (m)
5 6 7
D/2 D D ( xi / L) 2
(3)
R D (1 - x / L)3 Fig. 1. Comparison of nozzle obtained by different methods.
xi £ x £ L = 1 + [ s - 1]
D/2 D (1 - xi / L) 2
é 30i 4 20i 3 12i 2 6i ù é 0 ù éaù
ê 6 3 ú êh - H ú ê ú
where L is the length of nozzle, D is the hydraulic diameter of L L5 L4 L ú ú , w = êb ú .
A=ê 5 , B=ê
nozzle outlet, R is vertical distance from walls to the central ê 6L 5L 4
4 L3 3L 2ú
ê 0 ú êc ú
axis of nozzle and Ds is the hydraulic diameter of nozzle inlet. ê 4
ú ê ú ê ú
ëê30 L 20 L 12 L2
3
6 L ûú ë 0 û ëd û
Note that Mikhail concluded the optimal length of nozzle is 2
(8)
to 2.5 times the length of the internal curve [22].
Bell and Metha [23] provided a five-degree polynomial for
For nozzle with 3.5 m×3.5 m inlet section, 1 m×1 m test
nozzle profile of wind tunnel as follows.
section, maximum velocity of 80 m/s, the nozzle length of 7 m
and contraction ratio of 12.25, the nozzle profile can be drawn
x for three mentioned methods as shown in Fig. 1.
z = ,
L (4)
h = [-10 z 3 + 15z 4 - 6z 5 ]( H i - H o ) + H i
3. Numerical study of the three original nozzles
where Hi and Ho is the height of nozzle inlet and outlet from Reynolds–averaged continuity and Navier-Stokes equations
the center axis and L is the length of nozzle. are used as the governing equations for the two-dimensional
Sargison et al. [24] provided a six-degree polynomial for turbulent incompressible flow inside the three nozzles [5].
nozzle profile of wind tunnel as follows: Finite volume method (FVM) is used for discretizing the gov-
erning equations with second-order accuracy. Shear stress
transport (SST) model is applied for modeling Reynolds
y = ax 6 + bx 5 + cx 4 + dx 3 + ex 2 + fx + g . (5)
stresses. The steady state equations of mass and momentum
can be written as follows in a stationary frame [25]:
The coefficients of this equation are determined according
to the constraints of the nozzle profile as follows: ¶u ¶v
+ =0 (9)
¶x ¶y
y ( x = 0) = H , y '( x = 0) = 0 , y ''( x = 0) = 0 ¶u ¶u ¶P ¶ ¶
u +v =- + μÑ 2u + (- r u¢2 ) + (- r u¢v¢) (10)
y ( x = L) = h , y '( x = L) = 0 , y "( x = L) = 0 (6) ¶x ¶y ¶x ¶x ¶y
y "( x = i ) = 0 ¶v ¶v ¶P ¶ ¶
u +v =- + μÑ 2v + ( - r v¢2 ) + ( - r u¢v¢) . (11)
¶x ¶y ¶y ¶x ¶y
where i the horizontal location of turning point, L is the length
of nozzle, H is the height of the inlet and h is the height of the To do this, first, the structural grids are generated for the
outlet. In that study, the location of optimum turning point was three nozzles and very fine grids are used for regions near the
obtained at i/L = 0.6. To solve the unknown coefficients, the walls so that y+ reaches lower than 10. Fig. 2 shows the gener-
following matrix equation should be solved [24]: ated grid through the nozzle. Mass flow rate for the inlet and
pressure outlet for the outlet are used as boundary conditions.
Aw = B , (7) Outlet pressure is considered such that the total pressure at the
inlet of each nozzle is equal to atmospheric pressure. In this
where case, for all nozzles mass flow and inlet pressure are equal that
2062 H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067

Table 1. Non-uniformity of outlet velocity for three nozzles. Rouse: L=7m, Ds=3.5m, i/L=0.6
Bell: L=7m, Ds=3.5m
1 Sargison: L=7m, Ds=3.5m, i/L=0.6
Nozzle Outlet velocity non-uniformity (Percent)
Rouse 0.12 0.8

Bell 0.20
0.6
Sargison 0.14
0.4

CP
0.2

-0.2

0 1 2 3 4 5 6 7
L (m)

Fig. 4. Pressure coefficient of three nozzles.

Rouse: L=7m, Ds=3.5m, i/L=0.6


Bell: L=7m, Ds=3.5m
0.86 Sargison: L=7m, Ds=3.5m, i/L=0.6
Fig. 2. Grid of nozzle.

Rouse: L=7m, Ds=3.5m, i/L=0.6


Bell: L=7m, Ds=3.5m 0.855
82 Sargison: L=7m, Ds=3.5m, i/L=0.6

81.8
CP

0.85
81.6
Outlet Velocity (m/s)

81.4
0.845
81.2

81
0.84
0.2 0.4 0.6 0.8 1 1.2 1.4
L (m)
80.8

80.6
Fig. 5. Inlet zone pressure coefficient of three nozzles.

0.2 0.3 0.4 0.5 0.6 0.7 0.8 Rouse: L=7m, Ds=3.5m, i/L=0.6
Y (m)
Bell: L=7m, Ds=3.5m
-0.16 Sargison: L=7m, Ds=3.5m, i/L=0.6
Fig. 3. Outlet velocity profile of three nozzles.
-0.165

-0.17
it is a real state. The residual of 10-7 is considered as the con-
vergence criterion in this numerical solution. -0.175

To investigate the performance of these nozzles, outlet ve-


CP

-0.18
locity distribution and wall pressure distribution are compared
-0.185
for these three nozzles, as shown in Figs. 3 and 4, to select the
best nozzle for more optimization. -0.19
Fig. 3 indicates that the outlet velocity distributions for
-0.195
Sargison and Rouse nozzles are more uniform than those for
Bell nozzle. Moreover, it can be seen the Sargison nozzle has -0.2
6.4 6.6 6.8 7 7.2
the least boundary layer thickness at the outlet of nozzle. It L (m)

means for Sargison nozzle that uniform velocity area through Fig. 6. Outlet zone pressure coefficient of three nozzles.
the test section is larger than the two others. In Table 1, the
non-uniformities of velocity for the three nozzles are quantita-
tively compared. As seen, Rouse nozzle has the best uniform- zoomed, especially near the inlet and outlet region of the noz-
ity, while Bell nozzle has the worst uniformity among these zle according to Figs. 5 and 6, respectively. As seen, the ad-
three nozzles. verse pressure gradient of Rouse nozzle near both the inlet and
To focus on the wall pressure coefficients of these three outlet is less than that of the other ones. The effect of this phe-
nozzles, the wall pressure coefficients shown in Fig. 4 are nomenon should be checked in the quality of velocity profile
H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067 2063

Table 2. Compression of pressure loss for three nozzles. -0.05


Sargison: L=5.6m, Ds=3.0m, i/L=0.6
Sargison: L=7.0m, Ds=3.5m, i/L=0.6
Nozzle Pressure loss (Pa) Average outlet velocity (m/s)
Rouse 55.2 81.67 -0.1
Bell 55.6 81.67
Sargison 48.2 81.63

CP
-0.15

Table 3. Specifications of original and scaled down nozzle.

Scaled down Original Sargison -0.2


Specifications
nozzle nozzle
Inlet dimensions (meter) 3.0×3.0 3.5×3.5
Outlet dimensions (meter) 1.0×1.0 1.0×1.0 -0.25
86 88 90 92 94 96 98 100
Length (meter) 5.6 7.0 X/L (%)

Contraction ratio 9 12
Fig. 7. Wall pressure distributions of original Sargison and scaled
Length to inlet hydraulic ratio 1.85 2.0 down nozzle at the outlet zone.
Approximation of nozzle weight (Kg) 3500 5200
Sargison: L=5.6m, Ds=3.0m, i/L=0.6
Sargison: L=7.0m, Ds=3.5m, i/L=0.6
81.8

at the outlet of each nozzle.


81.6
Note that the adverse pressure gradient at the outlet zone is
more effective than that at the inlet zone on the flow quality of 81.4
the test section [19]. Therefore, for nozzle optimization, it is Outlet Velocity (m/s)
more important to decrease the adverse pressure gradient near 81.2

the outlet.
81
In Table 2, the pressure losses of the three nozzles are com-
pared to each other. According to this table, it can be con-
80.8
cluded that for equal outlet velocity, Sargison nozzle has less
pressure loss than the others. The results reveal that Sargison 80.6
nozzle is preferred as the case of optimization because of uni- 0.2 0.3 0.4 0.5 0.6 0.7 0.8
formity of outlet velocity and minimum pressure loss. Y (m)

Fig. 8. Outlet velocity profile of original Sargison and scaled down


4. Numerical study of length and contraction ratio nozzle.
effects on the performance of Sargison nozzle
If a nozzle is to be designed for a specified test section with nozzle is more uniform than that of the other one by 21%.
minimum construction cost, it is necessary to decrease the In fact, because of scaling down the Sargison nozzle, the
contraction ratio and length of the nozzle. Changes in nozzle wall pressure distribution changes and it causes the non-
dimensions can decrease its performance. Here, the dimen- uniformity of outlet velocity to increase. In other words, wall
sions of the Sargison nozzle are changed to reach a new noz- shape features directly affect its wall pressure distribution,
zle with smaller dimensions. To obtain the effects of length which in turn changes the outlet velocity profile. Thus, the
and contraction ratio on the performance of the nozzle, a outlet velocity profile of the nozzle as an important parameter
Sargison nozzle with contraction ratio of 9 and length to inlet of wind tunnel performance depends on wall pressure distribu-
hydraulic diameter ratio of 1.85 is considered for numerical tion or wall shape. Therefore, Fig. 8 indicates that the scaled
simulation to compute wall pressure distribution and uniform- down nozzle has worse performance and needs to be im-
ity of outlet velocity. Table 3 shows the geometric specifica- proved via inverse design method. For inverse design process,
tions of the original Sargison nozzle and the new scaled one. because the original Sargison nozzle has led to a more uni-
These geometrical changes decrease the construction cost of form outlet velocity profile, its wall pressure distribution is
the nozzle. It is worth noting that for both cases the turning considered as the target function to reach the desired nozzle
point is located at i/L = 0.6. shape.
In Fig. 7, wall pressure distributions of these two nozzles
are compared near the outlet. According to the figure, the re- 5. Inverse design
duction of contraction ratio and length of nozzle increases the
5.1 Fundamentals of bsa inverse design method
adverse pressure gradient near the outlet zone.
Fig. 8 shows the outlet velocity profile of the two nozzles. In inverse design of a duct, the goal is to obtain the geome-
According to the figure, the velocity profile of the original try that corresponds to a target pressure distribution along its
2064 H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067

DP × A × cosq
F = DP × A × cosq = ma Þ a = . (12)
m

If the force along the y direction is applied to the balls


through a period of time (t), the displacement is obtained as
follows:
Fig. 9. Schematic of a duct with balls and spines.
1 2
Dy = a ( Dt ) . (13)
2

If ( r ) is the density of the wall surface, according to previ-


ous equation, the displacement is equal to:

2
1 DP × A × cosq 2 ( Dt ) DP cosq .
Dy = ( Dt ) = (14)
2 rA 2r

In this equation, ( Dt ) is the period of time for ball move-


2
Fig. 10. The force of a ball from the wall. ment in a deformation step. Parameter ( ( Dt ) r ) is a setting
parameter to adjust the convergence rate of inverse design
algorithm. The smaller this value, the slower the convergence
2
walls. Here, the ball-spine inverse design method is used for rate will be. Indeed, if the amount of ( ( Dt ) r ) exceeds from
the design of the nozzle profile. In ball-spine method, the un- a limit, the algorithm will diverge.
known wall of the duct is composed of a certain number of Considering that the surface density is uniform through the
balls with certain mass as shown in Fig. 9. The duct walls are whole wall and the time between two consecutive modifica-
composed of a number of reckoning balls that can move freely tion steps, it can be concluded that relationship between the
across the spines. Flow in the duct causes a pressure distribu- balls displacements and the pressure difference is linear as
tion (present) to be applied to the duct walls. If a target pres- follows.
sure distribution is applied to the outside wall of the duct, it is
logical that the flexible wall deforms so that the computed Dy = C DP , (15)
pressure distribution on the wall reaches the target pressure
distribution. In other words, the force caused by the difference where
between the computed and target pressure distribution at each
point of the wall applies to the corresponding imaginary ball Dt 2
C= cosq . (16)
causing the ball to move. Therefore, the wall geometry is 2r
modified by the difference between target and computed pres-
sure distribution. At each geometry modification step, numeri- (ΔP) is calculated as the difference between the pressure
cal analysis of flow inside the updated nozzle is carried out via distribution in the position of the balls and the target pressure
a viscous flow solver to compute the current pressure distribu- distribution at the same location. In the inverse design algo-
tion along the wall. When target wall shape is obtained, this rithm, it is necessary that the inlet or outlet section be fixed.
pressure difference vanishes. In inverse design problems, for For this, the function of ΔP should be zero at the starting or
the unique solution, characteristic length of the duct should be end point of the duct wall. Here, because the outlet section
fixed. Here, the spines are considered as vertical lines so that should be fixed as the inlet of test section, the target and cur-
the two balls with the same horizontal position on two walls rent pressure of each point is gauged relative to the pressure of
are connected to each other, as shown in Fig. 9. In other words, the end point as the following equation:
the horizontal length of the duct in process of geometry modi-
fication is constant. The details of the Ball-spine algorithm, DPi = éë( PT arg et )i - ( PT arg et ) N ùû - ( Pi - PN ) . (17)
governing equations and its validation method are presented in
Ref. [25]. Then, the new location of each point is calculated as fol-
To obtain the equations of wall deformation, a ball as lows:
shown in Fig. 10 is assumed. Balls can just move along the
vertical direction due to the difference of current and target
xi(
t +Dt )
= xi( )
t
pressure and horizontal location of each ball is constant. The (18)
force causing balls to move is just due to pressure difference ( t +Dt ) (t )
yi = yi + C é( PT arg et )i - ( PT arg et )ref ù - ( Pi - P
t t
ref ). (19)
ë û
( DP ) that is obtained as follows:
H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067 2065

Fig. 11. Implementation of the ball spine inverse design algorithm.

Fig. 12. Comparison of the scaled down Sargison and optimized nozzle.
The residual in this algorithm is defined as follows:

N Sargison: L=5.6m, Ds=3.0m, i/L=0.6


å éë P - ( P
i =1
i T arg et i) ùû 82 Optimized Sargison: L=5.6m, Ds=3.0m, i/L=0.6
Sargison (Target): L=7m, Ds=3.5m, i/L=0.6
residual = N
. (20) 81.8
å éë( P
i =1
) ùû
T arg et i 81.6

81.4

Outlet Velocity (m/s)


-7
For the design convergence criteria considered 10 and 81.2

when residual reach this amount, geometry correction process 81

stops and the final geometry is obtained. The flowchart of this 80.8

algorithm is shown in Fig. 11. 80.6

80.4
5.2 The results of inverse design 80.2

80
Here, the design problem is to reduce the contraction ratio 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Y (m)
and length of the Sargison nozzle, as an optimum nozzle,
without significant increase in non-uniformity of outlet flow Fig. 13. Comparison of outlet velocity profile of original Sargison,
velocity. Indeed, if the Sargison nozzle is just scaled down scaled down Sargison and optimized nozzle.
without any re-profiling, wall pressure distribution of the new
scaled down nozzle will change so that it increases the non-
uniformity of outlet velocity. Therefore, scaling down the Sargison nozzle. The little difference between the outlet veloc-
Sargison nozzle although reduces its cost and weight, in- ity profile of the optimized and original nozzle is due to the
creases the non-uniformity of outlet velocity. To overcome difference in length and contraction ratio. In other words, be-
this problem, the wall pressure distribution of the original cause wall pressure distribution of the original Sargison be-
Sargison nozzle is considered as the target pressure distribu- comes first dimensionless and is then applied to the inverse
tion for the scaled down Sargison nozzle with lower contrac- design code, the new wall pressure distribution does not ex-
tion ratio and length as initial guess geometry. Thus, in this actly match the original one.
design problem, the target pressure distribution is considered
as constraint and wall profile as design variable. It is worth
6. Conclusions
noting that the pressure coefficient should be first dimen-
sionless and then, it is applied to the inverse design code as the First, three types of nozzles (Rouse, Bell and Sargison)
target pressure. By applying the target pressure distribution to were numerically studied to obtain wall pressure distribution,
the Ball-Spine inverse design code, an improved nozzle with outlet velocity profile and pressure losses. Having compared
lower contraction ratio and length is obtained without signifi- them, the Sargison nozzle was considered for case optimiza-
cant increasing the non-uniformity of the outlet velocity pro- tion. For decreasing the cost construction of the Sargison noz-
file. In Fig. 12, the original Sargison nozzle is compared with zle with constant test section size and mass flow rate, the con-
the optimized nozzle obtained from the inverse design. traction ratio and length of the nozzle were scaled down so
By numerical analysis of flow inside the original and opti- that its outlet section is fixed. But, the flow numerical solution
mized nozzle, their outlet velocity profiles are obtained and of the scaled down nozzle showed that the non-uniformity of
compared to each other as shown in Fig. 13. According to this outlet velocity increased. Thus, to improve the performance of
figure, the non-uniformity of the outlet velocity for the opti- the scaled down nozzle, the nozzle profile was modified using
mized nozzle is only 5 percent higher than that at the original Ball-spine inverse design method.
2066 H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067

In Ball-spine method, the nozzle wall is substituted by im- [8] J. Malone, J. Vadyak and L. Sankar, Inverse aerodynamic
aginary balls moving along the specified direction called spine. design method for aircraft components, J. of Aircraft, 24
The difference between current and target pressure distribu- (1987) 8-9.
tion causes the balls to move. At each geometry modification [9] J. Malone, J. Vadyak and L. Sankar, A technique for the
step, a viscous flow solver is used to determine the current inverse aerodynamic design of nacelles and wing configura-
pressure distribution along the wall. The inverse design proc- tions, AAIA paper, AAIA-85-4096 (1985).
ess is continued so that the difference between current and [10] J. Malone, J. Narramore and L. Sankar, An efficient airfoil
target pressure distribution vanishes and the target geometry is design method using the Navier–Stokes equations, AGARD
obtained. (1989) 5.
Finally, to overcome the increased non-uniformity of outlet [11] G. S. Dulikravich and D. P. Baker, Using existing flow-field
velocity, the wall pressure distribution of the original Sargison analysis codes for inverse design of three-dimentional aero-
nozzle was first scaled and then considered as the target pres- dynamic shapes, Recent Development of Aerodynamic De-
sure distribution to be applied to the inverse design code. Hav- sign Methodologies: Springer (1999) 89-112.
ing reached the target nozzle, numerical solution of flow in- [12] R. L. Barger and C. W. Brooks, A streamline curvature
side the optimized scaled down nozzle showed that the non- method for design of supercritical and subcritical airfoils,
uniformity was just increased by 5%. NASA TN D-7770 (1974).
[13] R. L. Campbell and L. A. Smith, A hybrid algorithm for
transonic airfoil and wing design, AIAA Paper 87-2552
Acknowledgment
(1987).
The authors would like to thank Mr. Mohammad Shoomal [14] R. A. Bell and R. D. Cedar, An inverse method for the
for his support of this research. aerodynamic design of three-dimensional aircraft engine
nacelles, Dulikravich (1991) 405-417.
[15] J. B. Malone, J. C. Narramore and L. N. Sankar, An effi-
Nomenclature------------------------------------------------------------------------
cient airfoil design method using the Navier-Stokes equa-
c : Nozzle contraction ratio tions, AGARD (1989).
Kn : Nozzle loss coefficient [16] M. Nili-Ahmadabadi, M. Durali, A. Hajilouy-Benisi and F.
fav : Average friction coefficient Ghadak, Inverse design of 2-D subsonic ducts using flexible
L : Nozzle length string algorithm, J. Inverse Problems in Science and Engi-
Dsc : Settling chamber hydraulic diameter neering, 17 (2009) 1037-1057.
CP : Nozzle wall pressure coefficient [17] M. Nili-Ahmadabadi, A. Hajilouy, M. Durali and F. Gha-
D : Nozzle outlet hydraulic diameter dak, Duct design in subsonic & supersonic flow regimes
Ds : Nozzle inlet hydraulic diameter with & without shock using flexible string algorithm, Pro-
Hi, H : Height of nozzle inlet to the center axis ceedings of ASME Turbo Expo, Florida, USA, GT2009-
Ho, h : Height of nozzle outlet to the center axis 59744 (2009).
i : Place of change the profile or turning point [18] M. Nili-Ahmadabadi, A. Hajilouy-Benisi, F. Ghadak and
M. Durali, A novel 2D incompressible viscous inverse de-
sign method for internal flows using flexible string algorithm,
References
J. of Fluids Engineering (2010) 132.
[1] M. A. Ardekani, Subsonic wind tunnel, K. N. Toosi [19] M. Nili-Ahmadabadi, M. Durali and A. Hajilouy-Benisi, A
University of Technology, Tehran, Iran (2009) (In Persian). novel aerodynamic design method for centrifugal compres-
[2] J. B. Barlow, J. William, H. Rae and A. Pope, Low-speed sor impeller, J. of Applied Fluid Mechanics, 7 (2) (2014)
wind tunnel testing, 3rd Ed., Wiley (1999). 329-344.
[3] P. Bradshaw, Effects of streamline curvature on turbulent [20] M. Nili Ahmadabadi, F. Ghadak and M. Mohammadi, Sub-
flow, AGARD, vol. -AG-169 (1973). sonic and transonic airfoil inverse design via ball-spine algo-
[4] J. P. N. Saddoughi S. G, Lateral straining of turbulent rithm, J. Computers & Fluids (2013).
boundary layers, J. of Fluid Mechanics, 229 (1991) 173-204. [21] J. H. Watmuff, Wind tunnel contraction design, Fluid Me-
[5] M. A. Ardekani, Optimization of a vertical wind tunnel noz- chanics Auckland (1986) (presented).
zle length using numerical and experimental methods, J. of [22] M. N. Mikhail, Optimum design of wind tunnel contrac-
Solid and Fluid Mechanics, 3 (3) (2013) 137-147 (In Per- tions, AIAA, 17 (1979).
sian). [23] J. H. Bell, Robindra and D. Metha, Contraction design for
[6] P. Garabedian and G. McFadden, Computational fluid dy- small low-speed wind tunnels, NAS2-NCC-2-294, vol. CR
namics of airfoils and wings, J. Scientific Computing (1982) 177488 (1988).
1-16. [24] J. E. Sargison, J. G. Walker and R. Rossi, Design and cali-
[7] P. Garabedian and G. McFadden, Design of supercritical bration of a wind tunnel with a two-dimensional contraction,
swept wings, AIAA J., 20 (1982) 289-291. 15th Australasian Fluid Mechanics Conference, The Univer-
H. Hoghooghi et al. / Journal of Mechanical Science and Technology 30 (5) (2016) 2059~2067 2067

sity of Sydney, Sydney, Australia (2004) 13-17. Hadi Hoghooghi, an Engineer at Durali
[25] Ansys, Inc., Ansys CFX solver theory guide, November System Design and Automation Center
(2009). (DSDA), received the M. S. (Mechani-
[26] F. Ghadak M. Nili, M. Dourali and A. Hajilouy-Benisi, A cal Engineering) from Sharif University
new method in inverse design, based on ball-spine for axi- of Technology (2014). His main re-
symmetric ducts with application in gas turbines, Aerospace search areas are aerodynamics, turbo-
Mechanics J., 7 (4) (2010) 65-75 (In Persian). machinery and CFD.

You might also like