Computation of Fluctuating Wind Pressure

and Wind Loads on Phased-Array Antennas

Qiang Du and Pingan Du

University of Electronic Science and Technology of China

School of Mechatronics Engineering
No.2006, Xiyuan Street, Hi-tech Zone, Chengdu, China, 611731
Tel: 86-28-61831056; E-mail: duqiang2007@hotmail.com; dupingan@uestc.edu.cn

Abstract

A necessary step in the design of a high-accuracy antenna system is to establish the signal-error budget due to structural,
pointing, and environmental parameters. Wind disturbances are the main source of vibration, and, consequently, of
pointing errors of the antenna in the open air. For these purposes, a method for computing the time history of wind
pressure and wind loads on antennas within the atmospheric boundary layer is presented. The method is based on
simulation of the wind eld and a computational uid dynamics (CFD) technique. Taking a phased-array antenna as an
example, the uctuating wind pressure and wind loads on the antenna were computed and compared with measured
data. The results showed that the uctuating wind pressure and wind loads computed by the presented method were
in good agreement with the data from a wind-tunnel test. The methods presented can be used to estimate the pointing
error induced by wind, and for design of an antenna controller for different purposes.

Keywords: Pressure control; load management; shape control; numerical analysis; wind; phased arrays; antenna
mechanical factors

1. Introduction the dimensionless wind torques, and compared them with the

T
field-measured torques [5]. Based on the experimental data,
he pointing accuracy of antennas has an important influ- Gawronski obtained the wind-velocity field, calculated the
ence on the accuracy of radar systems. Wang investigated wind forces on the antenna dish, and simulated the antenna
the performance of planar phased-array antennas with pointing errors [6]. Furthermore, Gawronski et al. simulated the
mechanical errors, and revealed that the effects on antenna wind-induced pointing errors, and the simulated results showed
performance are the loss of peak response in the scan direction that the pointing error due to fluctuating wind pressure was of
and the broadening of the main lobe, while the far-out sidelobe the same order as the error due to steady pressure [7]. Finally,
structure remains relatively intact [1]. Gawronski discussed three control algorithms for antenna-
pointing errors, and addressed their basic properties, tracking
Because the wind load is one of the most important loads precision, and limitations as applied to antenna tracking [8].
acting on antennas, reliable estimates of pointing errors due to
wind disturbances are required for large-size antennas in the In summary, it is necessary to counteract the effects of
open air. Antenna wind loads can be separated into two com- deformation that the wind pressure and wind loads on anten-
ponents: mean loads and dynamic loads. Dynamic wind loads nas produce. However, the expenses of tests are large. Although
are used to estimate the antennas pointing accuracy, and mean computational fluid dynamics (CFD) technology has been
wind loads are used to determine the sizes of antenna drives and quickly developed in wind engineering, there is no application
motors. The steady-state error due to mean wind pressure was to the computation of wind loads of antennas.
investigated by Katow [2]; the dynamic error of the DSS-14
antenna was analyzed by Massoudi [3]. In this paper, by combining the numerical computation of
the mean wind pressure and the simulation of the fluctuating
The primary source of wind load data is from wind-tun- wind velocity, a method for computing the wind pressure
nel tests. Lombardi carried out a wind-tunnel test on a rotating and wind loads on antennas in the time domain is proposed.
antenna, and compared the experimental results with those First, the mean-wind-pressure coefficients of the points on the
obtained from tests on a static model [4]. Gawronski et al. also antenna are numerically computed. Then, based on Gaussian
conducted wind tunnel tests on the DSS-13 antenna, calculated stationary random process theory, and by adopting the power

66 ISSN 1045-9243/2012/$26 2012 IEEE IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012
spectral density functions of longitudinal wind velocity fluc- variables, the Reynolds stresses, uiu j , appear. Equations (3)
tuation proposed by Simiu, and weighted-amplitude wave-
and (4) thus have ten unknown variables including the Reynolds
superposition technology (WAWS), the wind velocity of
stresses, velocities, and pressure in the flow field. These cant
the antenna is simulated. Finally, based on a quasi-steady
be directly solved for, and a turbulence model to compute the
assumption of the wind load, the wind pressure on the antenna
Reynolds stresses is needed. In this paper, the RNG k -
is computed, and the wind-load time history of the antenna can
turbulence model is adopted [12].
be obtained.

Taking a phased array antenna as an example, the com- 2.2 Simulation of Fluctuating Wind Velocity
standard model with the same shape [9-11], and the reliability Wind velocity can be considered as a spatial vector, and
and validity of the computed results were checked. The results the instantaneous wind velocity can be decomposed as
showed that the method can accurately compute the mean wind
loads and dynamic wind loads on the antenna, and can provide Vx= u + u ( t ) ,
input for wind vibration analysis of the antenna in the time
domain.
Vy = v ( t ) , (5)

2. Computational Method Vz = w ( t ) ,

2.1 Numerical Computation of where u ( t ) , v ( t ) , and w ( t ) are the fluctuating wind velocities
Mean Wind Pressure in the along-wind, cross-wind, and vertical directions,
respectively. u is the mean wind velocity in the along-wind
For an incompressible wind field with low velocity, the direction.
continuity equations, the law of mass conservation and the
momentum equations followed by the law of momentum con- Because there is no correlation between u ( t ) and v ( t ) ,
servation, are expressed as the three-dimensional fluctuating wind field can be decom-
posed into independent one-dimensional fluctuating wind
div ( U ) = 0 , (1) fields. In computational wind engineering, the power spectrum
of the fluctuating wind velocity in the along-wind direction,
ui p ji proposed by Simiu, is [13]
+ div ( ui U ) =
+ , (2)
t xi x j
u*2 200
( i, j = 1, 2,3 ), S u ( z , ) = (6)

where U = u1i + u2 j + u3k ; u1 , u2 , u3 are the instantaneous

(1 + 50 )
53

velocity components of a spatial point in the flow field in the z

direction i; p is the instantaneous wind pressure; and ji is the where in non-dimensional coordinates, = ;
2 u z
component of viscosity stress in the airflow. u* = Ku z ln ( z z0 ) ; is the angular frequency of the wind
Although all parameters in Equations (2) can be obtained velocity; the Karman constant is K = 0.4 ; the mean wind-
from unsteady numerical computation for the wind field, steady velocity profile of the atmospheric boundary layer takes the

numerical computation for the mean value is readily adopted in exponential law form u z = ur ( z zr ) ; ur is the mean wind
engineering for its reduced computational complexity. If all the velocity at the reference height; is the exponent of the veloc-
terms of Equations (1) and (2) are averaged in the time domain, ity profile; and z0 is the roughness height of the ground.
the equations become
As for the stationary Gaussian fluctuating wind field with
ui n spatial points, the matrix of the spectral density for the wind
=0, (3)
xi velocity is

ui u 2ui uiu j S11 ( ) S12 ( ) S1n ( )

1 p
+uj i = + . (4) S ( ) S22 ( ) S 2 n ( )
t x j xi x j x j x j S ( ) = 21 , (7)


Through computation using Equations (3) and (4), the mean S
n1 ( ) S n 2 ( ) Snn ( )
wind pressure and mean wind velocity of a point in the wind
field can be derived. In Equations (4), a set of new unknown where Sii ( ) is the function of the self-power spectral density,

IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012 67
 which is real, and Sij ( ) , ( i j ) , is the complex function of where N is number of frequency intervals; H jm (ml ) and
the cross-power spectral density. It can be calculated from the jm (ml ) are the mode and argument of the corresponding
equation
elements in H ( ) , respectively; u is the cutoff angular
frequency of the fluctuating wind-velocity spectrum; t is the
Sij ( ) = Sii ( ) S jj ( ) Cohij ( ) , (8)
sampled time; and ml is the random phase angle between 0
and 2.
where Cohij ( ) is the spatial coherence function. The spatial
coherence function can be calculated by
2.3 Computation for Fluctuating Wind
Cohij ( )
Pressure and Dynamic Wind Loads

( ) ( ) ( )
2 2 2 2
2 C x xi x j + C y2 yi y j + C z2 zi z j
= exp According to the relationship between airflow velocity
( ui + u j ) and airflow pressure, the airflow pressure, i , which is gener-
ated by the airflow velocity, Vx , at point i can be decomposed
(9)
into the mean airflow pressure and fluctuating airflow pressure:
Using the Cholesky decomposition method, the matrix of the
1 1 1 2
spectral density for the wind velocity can be decomposed into i = Vix2 = ui2 + ui ui ( t ) + ui ( t ) , (13)
2 2 2
S ( ) = H ( ) H*T ( ) , (10)
1
where the mean airflow pressure is i = ui2 , and the fluctu-
2
where H ( ) is a lower triangular matrix, and H*T ( ) is the 1 2
ating airflow pressure is = i ( t ) ui ui ( t ) + ui ( t ) . Due to
conjugated and transposed matrix of H ( ) . Because S ( ) is 2
a complex matrix, the elements of H ( ) can be calculated by 1 2
ui ui ( t ) ui ( t ) , i ( t ) ui ui ( t ) .
2
decomposing the columns of S ( ) one by one.
Because the airflow pressure is influenced by antennas in
H11 = S11 , the wind field, the mean wind pressure acting on point i on
antenna pi is
H i1 = Si1 H11 ,
(11) 1
=pi =
si i si ui2 , (14)
j 1 2
H jj
= S jj H jk H jk ,
k =1 where si is the shape coefficient of point i on the antenna.

j 1 In fact, pi has an another expression in numerical simula-

H
= ij Sij H jk H ik H jj , tion compared with theoretical analysis, and it can be denoted

k =1 as

(1 < j < i n) . 1
pi = C pi ur2 , (15)
2
On the basis of the theory presented by Deodatis [14], the
fluctuating wind velocity of a simulated point can be expressed where C pi is the non-dimensional coefficient of mean wind
as
pressure, and can be obtained from Equation (15); pi is solved
j N from Equations (4). Therefore, by combining Equation (14) and
u j (t )
= H jm (ml ) 2 cos ml t + jm (ml ) + ml Equation (15), the relationship between the shape coefficient
m= 1 =l 1 and the coefficient of mean wind pressure can be found:

=u , (12) 2
N zr
si = C pi , (16)
m zi
ml = ( l 1) + ,
n
where zr is the height of the reference point, and zi is the
( j = 1, 2, , n ) height of the computed point. This is based on the quasi-steady

68 IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012
assumption that the shape coefficients of points in antennas 3. Computational Example
subjected to fluctuating wind are same as those subjected to
mean wind, that is, si ( t ) = si . The fluctuating wind pressure,
3.1 Numerical Computation of
pi ( t ) , can then be obtained from its shape coefficient, si , and Mean Wind Pressure
its fluctuating wind velocity as
In this paper, a phased-array planar antenna with a rectan-
1 gular cylinder is used as a computational example. For
pi ( t ) =
= si i ( t ) C pi ( t ) ur .
2
(17)
2 comparison, the shape of the antenna was modified to be the
same as the model tested in [9]. Its array surface was a rectan-
The wind pressure on point i, including the mean wind pressure gular plane with sides (B H) of 4.5 m 18 m and depth D = 3
and the fluctuating wind pressure, can therefore be known. The m.
wind pressure coefficient of point i can be calculated from
The boundary conditions at approaching flow were defined
pi ( t ) + pi as follows. The antennas height, H, was chosen to be the
C=
pi (t ) = C pi ( t ) + C pi , (18) reference height in the computation, and the wind velocity of
1 2
ur the reference height was ur = 12.7 m/s. The wind velocity in
2
the y and z directions were both equal to zero. Besides these, the
where C pi ( t ) is the fluctuating wind pressure coefficient, turbulent kinetic energy, k z , and its dissipation rate, , at the
which is time dependent and can be calculated from Equa- inlet were calculated per the following equations:
tion (17). Furthermore, the transient wind loads on antennas can
3
be derived from the sum of the products of the fluctuating wind kz = ( u z I )2 ,
2
(22)
Through the steady numerical simulation, the mean drag 1
= C 3 4 k z3 2 ,
force and lift force on the antenna can be obtained. The mean l
drag coefficient, Cd , and the lift coefficient, Cl , of the antenna
can then be found as where l and I are the turbulence-integral length scale and the
turbulence intensity, respectively. These were calculated from
Fd
Cd = ,
1 2 z
v A l = 1200 ,
2 10
(19) (23)
Fl 1.7
Cl = . z
1 2 I = 1.5 .
v A 10
2
As for the boundary condition in the outflow, we supposed that
According to [10], the drag coefficient of the antenna is defined
the turbulence flow in the outflow was fully developed, and that
as
the normal gradients of all physical variables in the outflow of
the flow field were all zeros. The wall conditions defined as the
Fd ( t )
Cd ( t ) = . (20) top, left, and right surfaces of the computational domain were
1 H
sliding walls, and the bottom of the computational domain and
B ( z ) u z2 dz
2 0 the surfaces of the antenna were all non-sliding walls. For the
computational domain near to the non-sliding walls, a standard
The root mean square (rms) coefficients of the drag force on the wall function was adopted to simulate the flow field. The
antenna, which reflect the fluctuations of the wind loads, can be computational domain, coordinate definitions, and boundary
calculated from conditions are shown in Figure 1. After that, the computational
domain mesh was generated using an unstructured grid. The
Fx mesh near the surfaces of the antenna had to be refined. The
C F = , (21) smallest size of the grid close to the antenna was 100 mm.
x 1 2
ur A
2 Finally, approximately 9.2 105 unstructured grids were
obtained. Because the blockage ratio of this computational
where A is the projected area of the antenna in the along-wind model was 2.94%, it was not necessary to correct the computed
direction. results due to wall interference.

IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012 69
 The wind pressures of the leeward surface on the antenna were
all negative. C pmax on this surface was in the lower part, and
the wind-pressure coefficients on the leeward surface changed
slowly.

Because the antenna used in the computations had the

same shape and wind conditions as the standard model tested,
C p was in the same place on the antenna and the standard
model results should have been identical. In this paper, the
mean wind pressure was numerically simulated, and the mean
pressure coefficients of the antenna at 2 3 H , compared with
the test data measured at Tongji University [9, 11], are shown in
Figure 1. The computational domain and boundary condi- Figure 3. From Figure 3, the mean wind pressure acting on the
tions. elements of the phased-array antenna were different. On the
windward surface, the pressure coefficients on the front surface
given by the numerical computation were over-predicted. To
the contrary, the pressure coefficients on the leeward surface
obtained from the numerical computation were under-predicted.
The reason was that the steady numerical computation could
not show all of the eddies in the wind field.

On the side surface, the distribution patterns predicted by

the RNG k -
the experiment, which showed that the steady numerical com-
putation cannot be used to accurately obtain the pressure on the
side surface.

3.2 Simulation for Fluctuating Wind Velocity

on the Antenna

Because the leeward surface was vertical to the airflow

direction, the power spectrum of the leeward surface was sup-
posed to be same as that of the windward surface. For simula-
tion of the fluctuating wind velocity, the windward and leeward
surfaces of the antenna were partitioned into 2 60 regions,
and numbered according to a transverse interval of 1.5 m
and a vertical interval of 0.9 m. The regions on the windward
surface of the antenna are shown in Table 1, and those regions

Figure 2. Contours of the mean pressure coefcients on the

antennas surface: (l) front, (r) back.

The RNG k - turbulence model was used to compute the

mean wind pressure on the antenna. The contours of C p on the
windward and leeward surfaces of the antenna are shown in
Figure 2. From Figure 2, it was seen that the windward surface
was dominated by positive pressure. The maximum of the mean
wind pressure coefficients, C pmax , was in the upper part of the
antenna. The gradient of the wind pressure at the superior edge Figure 3. The mean pressure coefcients of the antenna at
of the antenna was intense, and C p at the edges was a minimum. 2 3H .

70 IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012
 Table 1. The regions on the windward and height, so the power spectrum of the wind velocity in the along-
leeward surfaces of the antenna. wind direction could be considered to be invariant along the
horizontal direction [13]. Meanwhile, according to the distance
1(61) 21(81) 41(101) between two points at different heights, there existed different
2(62) 22(82) 42(102) correlations for fluctuating wind velocity at the central points in
different regions.
3(63) 23(83) 43(103)
4(64) 24(84) 44(104)
5(65) 25(85) 45(105)
3.3 Computation of
6(66) 26(86) 46(106) Fluctuating Wind Pressure
7(67) 27(87) 47(107)
8(68) 28(88) 48(108) In [8], the fluctuating wind-pressure distributions on the
standard model were compared. These were measured in vari-
9(69) 29(89) 49(109)
10(70) 30(90) 50(110) Bristol. It was obvious that the root mean square of the pressure
11(71) 31(91) 51(111) coefficient on this antenna, computed in this paper, was larger
than that of the standard model measured in the wind tunnel
12(72) 32(92) 52(112)
tests. Because the height of the antenna was lower than the
13(73) 33(93) 53(113) standard model, the turbulence intensity near the ground was
14(74) 34(94) 54(114) greater than at higher places. With the same shape and wind
15(75) 35(95) 55(115)
16(76) 36(96) 56(116)
17(77) 37(97) 57(117)
18(78) 38(98) 58(118)
19(79) 39(99) 59(119)
20(80) 40(100) 60(120)

Table 2. The parameters for simulation of the

uctuating wind velocity.

Parameter Value
0.3
u [Hz] 4
0.0122
N 512
n 120
t [s] 0.25
t [s] 600
ur [m/s] 12.7

zr [m] 18

z0 [m] 1

on the leeward surface are represented by the bracketed

number. During simulation, the fluctuating wind velocities in a
region were thought to be completely relevant. The important
parameters in simulation of the fluctuating wind velocity are
shown in Table 2. The time history of the fluctuating wind
velocity at the central points in the 21st, 30th, 81st, and 90th
regions are shown in Figure 4.

From Figure 4, there was a strong correlation for the fluc- Figure 4. The uctuating wind velocities at the central
tuation of the wind velocity between two points at the same points in regions.

IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012 71
 conditions, the root mean square of the fluctuating pressure computation were in good agreement with the values from the
coefficients and dynamic wind loads of different structures Selvam model at the windward and leeward surfaces, but the
have to do with the proportion to their height. The root mean results computed from the present computation were less than
square of fluctuating pressure coefficient, C prms s , and the root those for the Selvam model at the side surfaces. Due to the
mean square of the dynamic wind loads, C F , of the standard complexity of the flow on the side surfaces, the prediction
xs models based on the RANS k- turbulence model could not be
test model, can be changed and compared with the computed used to obtain the fluctuating pressure on the side surfaces to
results: agree with the wind-tunnel tests. However, the results computed
for the present example are basically similar to the test data: the
1.7
C prms a C Fx a za fluctuating wind pressure on the windward and leeward surfaces
= = , (24) can be used for the computation of the wind-induced vibration
C prms s C F zs
xs of the antennas surfaces. Comparing with the Selvam model,
the root mean square of the drag coefficient computed by the
where za is the height of the antenna, and zs is the height of presented method should be very close to the Selvam model.
the standard model. The test data of the standard model can The presented method has the advantage over the Selvam
therefore be used to compare with the results computed by the approach of supplying the time history of wind pressure to the
computation of wind-induced vibration of antennas in the time
domain.
In the numerical simulation based on the RANS turbu-
lence model, the root mean square of the pressure coefficient Furthermore, the changes of C p in Figure 3 and C prms in
cannot be directly calculated. The computational model for Figure 5 were alike in the along-wind direction. The corre-
computing the root mean square of the fluctuating wind-pres- sponding amplitudes of the elements distributed in the central
sure coefficient was proposed by Paterson and Holmes from the part of the windward surface induced by the fluctuating wind
Bernoulli equation [15]: pressure will be larger than those distributed on the edge part.

(
C prms 2 kr 3 + 0.816 C pi u z k z
= ) ur2 . (25) After the shape coefficients of the antenna have been
computed by numerical simulation of the steady wind field,
Selvam also derived a model for computing the root mean and the time history of the wind velocities at the central points
square of the fluctuating wind-pressure coefficient [16]: in all regions have been obtained from simulation for the
fluctuating wind field, the wind pressures, including the mean
wind pressures and the fluctuating wind pressures on the cen-
= (
C prms 2 C pi 1.414ur kr + kr ) ur2 , (26) tral points of all regions, can be calculated. The time history of
the wind pressure coefficient at the central points in 21st, 30th,
where kr is the kinetic energy at the reference point in the inlet. 81st, and 90th regions are shown in Figure 6.
Figure 5 shows the experimental and computed root mean
square values of the pressure coefficients of the antenna at Taking the shape coefficients of the antenna into account,
2 3H . the time history of the wind pressure at the central points in
the antenna had some changes compared with the time history
From Figure 5, The Paterson model obviously under-pre- of the wind velocity at these points. From Figure 6, as the
dicted the fluctuating wind pressure. C prms a from the present fluctuating wind velocity simulated, the fluctuation of the
wind pressure on the upper part of the windward surface was
smaller than its fluctuation on the middle part. To the contrary,
the fluctuation of the wind pressure on the upper part of the
leeward surface was stronger than its fluctuation on the middle
part. The fluctuating wind pressures on the same surface had
some coherence. Because the wind pressure on the leeward
surface was negative, the wind pressure on the corresponding
regions in the windward and leeward surfaces of the antenna
were characteristic of reversed phase. The curves in Figure 5
also showed that the fluctuation of the wind pressure on the
windward surface was stronger than on the leeward surface,
and the fluctuation of the wind pressure on the leeward surface
was very small.

The fluctuating wind pressure on the phased array

antenna must bring about an element-spacing perturbation. It
tends to produce errors in beam pointing and interferometric
Figure 5. The root mean square of the pressure coefcients angle measurements, thus degrading the radar target-location
of the antenna at 2 3 H . capability.

72 IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012
 streamlines, it was clear from Table 3 that the mean drag coef-
ficient of the antenna computed by this presented method was
slightly lower than the test data, because of its failure to simulate
all eddies in the wind field.

As for the dynamic wind loads on the antenna in the

along-wind and cross-wind directions, the result derived from
the present method was in good agreement with the test data.
Therefore, the method presented in this paper can be applied
to the predictions for wind-induced vibration of antennas, and
the deformation of antennas can be calculated. The time histo-
ries of the drag coefficients of the antenna are shown in Fig-
ure 7. From Figure 7, it could be seen that the amplitude of the
drag coefficients computed by the presented method were big.
The drag coefficients at some time were even negative, which
indicated that the wind loads on the phased-array antenna near
to the ground were more complicated. Furthermore, using the
method to compute the fluctuating wind pressure and dynamic
wind load, increasing the regions on the antennas surface can
further improve the computational accuracy.

In order to enable a radar antenna to acquire good func-

tioning electrically and in mobility, the antennas framework
must meet the requirements of good rigidity and be lightweight.
Wind loads play a decisive role in the strength calculations and
stiffness calculations for antennas working in the open air,
and the time history of wind loads can supply an input to the
structural analysis in the time domain. However, the dynamic
wind loads are the main source of antenna vibrations, and
consequently pointing errors are proportional to the square of
the velocity [17].

3.5 Application of the Wind Loads

Figure 6. The coefcients of wind pressure at the central
points in the regions.
In this paper, based on the computational fluid dynamics tech-
nique and simulation of the wind velocity, the steady wind
forces and dynamic wind loads acting on the antenna have been
3.4 Computation of Dynamic Wind Loads
computed. What is next is to calculate the antennas structural
deformation, and to define the power of the drive located in
Similarly to the mean pressure coefficient, the mean drag the antennas pedestal, in the design phase. Additionally, the
coefficient of the antenna should be identical to that of the determination of the antenna velocity loop model is part of the
standard model. According to Equation (19) and Equation (21), controller, which significantly improves the tracking precision
the mean drag coefficient, Cd , and the root mean square in wind gusts, and the accuracy of the model is the factor
coefficient of drag force, C F , of the antenna in the along- that influences the controllers performance. The accurate
x

wind direction were computed, respectively. Although the lift computation of wind loads on antennas is helpful in establishing
force acting on the antenna in the cross-wind direction is subject the velocity loop model of antennas. An antenna control system
to more influences than the drag force, the time characteristics is shown in Figure 8.
of the wind force do not depend on its direction [17]. Therefore,
the root mean square coefficient of lift force, C F , should be
y

equal to the root mean square coefficient of drag force. For

Table 3. A comparison of results from different methods.
comparison of dynamic wind load, the root mean square
coefficient of force on the standard model was expressed as that
Method Cd C F Cl C F
of the antenna, and this is shown in Table 3. x y

Wind test [10] 1.79 0.899 0 0.97

Because the RNG k - turbulence model used in the
simulation for the mean wind pressure can better deal with the Present
1.61 0.951 0.009 0.951
kinds of flow with high strain rates and bigger bending computation

IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012 73
 loads on the phased-array antenna in the along-wind and cross-
wind directions were in good agreement with the test results.

3. Because the validity of the presented method has nothing to

do with the size or shape of the studied antenna, the computa-
tional method for wind pressures and wind loads can be
extended to compute the wind loads on antennas with other
shapes, including reflector antennas.

5. References

1. H. S. C. Wang, Performance of Phased-array Antennas with

Mechanical Errors, IEEE Transactions on Aerospace and
Figure 7. The drag coefcients of the antenna. Electronic Systems, 28, 2, April 1992, pp. 535-545.

2. M. S. Katow and H. D. McGinness, Wind Load Predic-

tions for the 64-Meter Diameter Antenna, Technical Report
32-1526, Jet Propulsion Laboratory Pasadena, California, 15,
June 15, 1973.

3. M. Massoudi, Tracking Error of 100-m Antenna Due to Wind

Gust, TDA Progress Report 42-48, September and October,
Jet Propulsion Laboratory, Pasadena, California, December 15,
1978, pp. 94-101.

4. G. Lombardi, Wind-Tunnel Tests on a Model Antenna

Figure 8. An antenna control system. Rotating in a Cross Flow, Engineering Structures, 13, 4,
October 1991, pp. 345-350.

5. W. Gawronski, J. A. Mellstrom, and B. Bienkiewicz,

4. Conclusions Antenna Mean Wind Torques: A Comparison of Field and
Wind-Tunnel Data, IEEE Antennas and Propagation Maga
When an antenna working in the atmospheric boundary zine, 47, 5, October 2005, pp. 55-59.
layer is designed, it is necessary to not only analyze its electri-
cal performance, but to also consider the influence of structural 6. W. Gawronski, Modeling Wind-Gust Disturbances for the
deformation induced by wind. Traditionally, the mean wind Analysis of Antenna Pointing Accuracy, IEEE Antennas and
pressure is obtained from a steady-state wind tunnel, and the Propagation Magazine, 46, 1, February 2004, pp. 50-58.
dynamic wind pressure is generated by the power spectrum
of the wind velocity and the wind-tunnel data. In view of the 7. W. Gawronski, B. Bienkiewicz, and R. E. Hill, Pointing-
shortcomings of high costs and long times in wind-tunnel Error Simulations of the DSS-13 Antenna Due to Wind Dis-
testing, a computational method combined with numerical turbances, The Telecommunications and Data Acquisition
simulation for the steady wind field and simulation for the Progress Report 42-108, October-December 1991, Jet Propul-
fluctuating wind field was presented in this paper. The pre- sion Laboratory, Pasadena, California, February 15, 1992, pp.
sented method has the following characteristics: 109-134.

1. Taking the mean wind-velocity profile of the atmospheric 8. W. Gawronski, Antenna Control Systems: From Pi to H ,
boundary layer for the velocity inlet of the computational IEEE Antennas and Propagation Magazine, 43, 1, February
model, and adopting the RNG k - turbulence model to com- 2001, pp. 52-60.
pute the mean wind pressure on the antenna, the computed
results for the mean wind pressures on the windward and lee- 9. P. Huang, M. Gu, and Y. Quan, Wind Tunnel Test Research
ward surfaces of the antenna were close to the experimental on CAARC Standard Tall Building Model, Chinese Quarterly
data. Consequently, wind loads on the antenna were slightly of Mechanics, 29, 4, December 2008, pp. 627-633.
lower than the experimental data.
10. E. D. Obasaju, Measurement of Forces and Base Over-
2. Integrating the advantages of the computational fluid turning Moments on the CAARC Tall Building Model in a
dynamics technique to compute the mean wind pressure and Simulated Atmospheric Boundary Layer, Journal of Wind
to simulate the fluctuating wind velocity, the results computed Engineering and Industrial Aerodynamics, 40, 2, June 1992,
by this method for fluctuating wind pressure and dynamic wind pp. 103-126.

74 IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012
11. W. H. Melbourne, Comparison of Measurements on the
CAARC Standard Tall Building Model in Simulated Model
Wind Flows, Journal of Wind Engineering and Industrial
, 6, 1-2, July 1980, pp. 73-88.

12. S. B. Pope, Turbulent Flow, Cambridge, UK, Cambridge

University Press, 2000.

13. E. Simiu and R. H. Scanlan, Wind Effects on Structures: an

, New York,
Wiley, 1986.

14. G. Deodatis, Simulation of Ergodic Multivariate Stochastic

Processes, Engineering mechanics, 22, 8, 1996, pp. 778-787.

15. D. A. Paterson and J. D. Holmes, Computational of Wind

Flow Around the Texas Tech Buildings, Proceedings of
Workshop on Industrial Fluid Dynamics, Heat Transfer and
Wind Engineering, Highett, CSIRO-DBCE, 1989, pp. 33-34.

16. R. P. Selvam, Computation of Pressure on the Texas

Tech Building, Journal of Wind Engineering and Industrial
, 43, 1-3, 1992, pp. 1619-1627.

17. W. Gawronski, B. Bienkiewicz, and R. E. Hill, Wind-

Induced Dynamics of A Deep Space Network Antenna,
, 178, 1, November 1994, pp.
67-77.

Introducing the Feature Article Authors

Qiang Du was born in Zigong, China, in 1969. He

received the MASc from Xihua University, Sichuan, China,
in 2006. He is currently a PhD student in the Department of
Mechatronics Engineering, University of Electronic Science
and Technology of China. His research interests include the
digital design of phased-array antennas and computational fluid
dynamics applications to antennas.

Pingan Du was born in Chongqing, China, in 1962. He

received the PhD degree from the University of Chongqing,
China, in 1992. He is currently a Professor with the University
of Electronic Science and Technology of China. His research
interests include digital design and simulation of mechatronics
systems.

IEEE Antennas and Propagation Magazine, Vol. 54, No. 1, February 2012 75