Original Article

Proc IMechE Part C:

J Mechanical Engineering Science
A pointing error analysis model 0(0) 1–13
! IMechE 2018

for large reflector antennas Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav

under wind disturbance DOI: 10.1177/0954406218779611

journals.sagepub.com/home/pic

Jie Zhang , Jin Huang, Pengbing Zhao , Wei Liang,

Youran Zhang and Congsi Wang

Abstract
Large reflector antennas have been widely used in extensive researches, including satellite communications, detection of
gravitational waves, origin of galaxies, and so on. With increasing in the diameter and the working frequency of these
antennas, the requirement for pointing accuracy becomes more demanding simultaneously. However, wind disturbance
has serious impacts on the pointing accuracy of antennas, which makes it be a severe challenge to the development of
large antennas. To reduce the influence of wind disturbances on pointing accuracy, it is important to construct an
accurate model for estimating the pointing error caused by wind. The traditional modeling method focuses on treating
the wind disturbance as a torque on the motor output, whilst ignoring the oscillations of flexible structures caused by the
wind pressure, so the pointing error caused by structural deformation cannot be estimated which lead to a limitation of
compensation effect. In this article, based on an entire working condition dynamic model (EWCDM), a pointing error
analytical model is proposed to estimate the pointing error caused by deformation of the antenna structure. A series of
simulations and tests are carried out on an antenna with a diameter of 7.3 m. The results prove the feasibility of the
proposed model.

Keywords
Reflector antennas, pointing accuracy, wind disturbance, dynamic model, analytical model

Date received: 19 December 2017; accepted: 2 May 2018

torques at antenna drives, wind disturbance on the

Introduction
servo system, manufacturing imperfections and so
In engineering projects, the influence of wind disturb- on. Meanwhile, some approximate solutions to these
ance is determined with respect to the pointing per- problems were stated in this study. A wind disturb-
formance of large reflector antennas, e.g. Green Bank ance model was developed which was mainly essential
Telescope (GBT) with a diameter of 100  110 m, for estimating disturbance torque. The wind disturb-
which is the largest radio telescope in the world. ance was assessed with respect to antenna pointing
Under the wind speed of about 7 m/s, the pointing performance,4 and three types of disturbance in simu-
error of this telescope is about 14 arcs, which is lation analysis were modeled as follows: i) an equiva-
much higher than the observational requirement of lent wind force was applied to the reflector surface, ii)
1.5 arcs.1 For QiTai radio Telescope (QTT), a reflec- motor shaft torque was applied to the antenna driver,
tor antenna of 110 m aperture was designed in China. and iii) speed disturbance was applied to the input of
A study has found that the equivalent torque reaches
1.7  107 Nm when the lateral wind blows at a speed
Key Laboratory of Electronic Equipment Structure Design, Ministry of
of 10 m/s; meanwhile, the wind pressure on the struc-
Education, Xidian University, Xi’an, China
ture surface was about 1.2  105 N. Without effective
control compensation, the maximum pointing error Corresponding authors:
caused by wind disturbance was up to 80 arcs.2 Jin Huang, Key laboratory of Electronic Equipment Structure Design,
The influence of wind disturbances has been exten- Ministry of Education, Xidian University, Xi’an 710071, Shaanxi, China.
Email: jhuang@mail.xidian.edu.cn
sively investigated. According to a study conducted by
Congsi Wang, Key laboratory of Electronic Equipment, Structure
Wodek Gawronski, many difficulties were encoun- Design, Ministry of Education, Xidian University, Xi’an 710071, Shaanxi,
tered in maintaining high-precision pointing control China.
of large antennas,3 including backlash and friction Email: congsiwang@163.com
2 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

speed loop. The first equivalent method was based on vibration of the reflector,12 and it cannot estimate the
the finite element model of antenna, which could not be pointing error caused by the deformation of the reflec-
directly used in the controller design; the latter two tor. The finite model was used to determine the deform-
methods are mainly used in antenna design stage, and ation of the structure under wind load, but it cannot be
they are useful in analyzing wind disturbance effects of used in designing the controller directly.13 To analyzing
the servo system. For reflector antennas with diameters the pointing error caused by wind accurately, the first
of 34 m and 70 m, the influence of pointing perform- step is to establish a dynamic model for assessing the
ance from the gain of the controller has been discussed, deformation of the antenna structure. After obtaining
and the results indicate that increasing the gain would few modal transition functions, the EWCDM was
reduce the rise time, shorten the settling time, increase derived based on the modal superposition, and the
the overshoot, improve the closed-loop bandwidth, deformation of the antenna structure could be
and reduce the pointing error caused by wind.5 A described under different operating conditions.
50 m antenna combined different control methods
comprising proportional–integral–derivative (PID) The dynamic model based on the modal
and linear-quadratic-Gaussian (LQG) controllers in
superposition method
current, velocity, and position loops were compared.
The results show that LQG controller significantly out- As shown in Figure 1, the antenna structure was
performed PID controllers in reducing pointing errors deformed under wind load, and the nodes of the
caused by wind gusts.6,7 The flexible deformations in reflector would produce some displacements. After
the antenna structure were estimated by performing obtaining the distribution of wind load on the reflec-
finite element analysis, and the antenna performance tor, the dynamic model could be expressed as follow
was analyzed on the basis of electromechanical cou-
pling model.8,9 A uniform load of wind disturbance X
s

was applied on reflector surface. The antenna pointing Mq þ Dq_ þ Kq ¼ ðBodi fdi þ Boli fli þ Bosi fsi Þ ð1Þ
i¼1
error was determined by analyzing the dynamic model,
and an active compensation mechanism was designed y ¼ Coq q ð2Þ
to effectively reduce pointing error.10,11
Without considering the deformation of the _ and q€ are the nodal displacement, velocity,
where q, q,
antenna structure, the servo error caused by wind and acceleration vectors, respectively; fdi , fli , and fsi
was effectively reduced by using optimal control are the inputs of the equivalent wind pressure in dif-
methods. Although a few methods consider the ferent directions of different parts of the reflector; s is
deformation of the antenna structure, these methods the number of reflector partitions; Bodi , Boli , and Bosi
cannot be used directly in controllers nor are they are the corresponding input matrixes that provide the
accurate enough to estimate the pointing error position of the force action; y is the output vector; M,
under different operating conditions. Meanwhile, the D, and K are the mass, damping, and stiffness matri-
requirement of pointing accuracy is more demanding ces, respectively; Coq is the output matrix that pro-
and the vibration of the antenna structure is more vides the position of the out point.
obvious, especially for those antennas with low nat- A variable qm was introduced to express the struc-
ural frequency and constructed without a radome. ture in modal coordinates; this variable is referred as
Hence, it is an urgent need to propose a method the modal displacement, and it satisfies the following
for analyzing the pointing error caused by wind equation
accurately.
A new model was proposed to estimate the point- q ¼ qm ð3Þ
ing error, and a detailed study of this method is pre-
sented in this paper. An entire working condition where  is the matrix of modal shape defined as
dynamic model (EWCDM) of the antenna structure follows
is derived in the Entire working condition dynamic 2 3
u11 u21 . . . un1
model section 2. An approximate model based on
6u 7
the geometric algorithm was used to analyze the 6 12 u22 . . . un2 7
6 7
pointing error caused by reflector deformation in 6 ... ... ... ... 7
6
U ¼ ½ u1 u2 . . . un  ¼ 6 7
the Pointing error analytical model section. Based 7
6 u1k u2k . . . unk 7
on the simulation of an analytical example explained 6 7
4 ... ... ... ... 5
in the Model application section, conclusions are pre-
u1nd u2nd . . . unnd
sented in the last section.
ð4Þ
where un is the nth order of modal shape, and unnd
Entire working condition dynamic model represents the component with the node number nd .
The rigid model was usually used in antenna pointing Here, n is the order of modes and nd represents the
control system. This model cannot describe flexible number of nodes.
Zhang et al. 3

Figure 1. The structure of large reflector antenna.

By substituting equation (3) into equations (1) and Ignoring the damping, Mm and Km are satisfied
(2) and left multiplying by T , the dynamic model with the following equations under free vibration
could be expressed as follow
X2 ¼ M1
m Km ð11Þ
Mm q€ m þ Dm q_ m þ Km qm
2 3
X s
ð5Þ !1 0 . . . 0
¼ ðUT Bodi fdi þ UT Boli fli þ UT Bosi fsi Þ 6 0 ! ... 0 7
6 2 7
i¼1 X¼6 7 ð12Þ
4... ... ... ...5
y ¼ Coq Uqm ð6Þ 0 0 ... !n

In this expression, mass, damping, and stiffness where X is the natural frequencies matrix, and !i is
matrices are transformed into modal mass matrix the natural frequency of the ith order.
(Mm ), modal damping matrix (Dm ), and modal stiff-
ness matrix (Km ), respectively.
Modal transition function
Mm ¼ T M ð7Þ From the aforementioned modeling process, a
dynamic model was developed by the modal mass,
damping, and stiffness matrices. In these matrices,
Km ¼ T K ð8Þ the corresponding parameters always changed accord-
ing to the elevation angle under different working con-
ditions, so it was necessary to find a way to express
Dm ¼ T D ð9Þ these changes. The modal transition function was
proposed.
The damping matrix could be obtained by the fol- Despite changes in the elevation angle, the top-
lowing equation ology structure shape and the connection do not
change; therefore, it is only a movement of the natural
D ¼ 1 K þ 2 M ð10Þ frequency values and the vibration amplitude; the
low-level modes still vibrate in the same modal
where 1 and 2 are the Rayleigh damping scale shape in different elevation angles. Moreover, the
coefficients. characteristics of wind spectrum are mainly associated
4 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

with low frequency, which leads to vibration and y ¼ Coq qm ð18Þ

pointing error in low frequency. Usually, the fre-
quency is restricted within 20 Hz.14 where Melm , Delm , and Kelm are the new modal mass,
Specifically, by discretizing elevation angle and ana- damping, and stiffness matrices that change with the
lyzing the structure model at a periodic interval angle elevation angle, respectively; Uel is the new matrix of
(every h degree), the elevation angles were fixed at ph modal shape.
p ¼ 1, 2, 3 . . .; the corresponding natural frequencies
!ph1 , !ph2 , . . . , , !phm were obtained by modal analysis.
Natural frequencies with the same mode shape in dif-
Pointing error analytical model
ferent elevation angles were selected to compose a new For cassegrain antenna, the following factors have an
sample vector ! a . With this mode shape, the frequency impact on the antenna pointing precision as a conse-
in any elevation angle, !ela , is expressed as follow quence of the deformation of the reflector: deform-
ation of the main-reflector, displacement of the feed,
!ela ¼ f! ðel Þ ð13Þ and displacement and rotation of the sub-reflector.
To estimate the influence of the deformed main
where f! ðel Þ is the modal transition function of nat- reflector, a best fitting parabolic is used. Nodal dis-
ural frequency (MTFNF), which could be obtained by placements of each point of the main reflector were
a method using data modeling fitting, such as least calculated using equation (18). Under the criterion of
square methods or support vector machines. ! a is minimum axial error, the parameters for the best fit-
used as the sample vector. ting parabolic were obtained from the following
Similarly, the modal mass in any elevation angle equation
mela is obtained
 ¼ A1 H ð19Þ
mela ¼ fm ðel Þ ð14Þ
Here, A is a weighting parameter matrix con-
where m a is the modal mass vector with the same structed with nodal coordinates, and H is a weighting
modal shape. Moreover, fm ðel Þ is the modal transition parameter matrix constructed with nodal displace-
function of modal mass (MTFMM), and m a is used as ments.15 The parameters of  consist of displacements
the sample vector. of the vertex of the fitted parabolic reflector, which is
f’ ðel Þ is the modal transition function of modal defined as du, dv, dw; the rotation angles of the focal
shape (MTFMS); the modal shape vector in any axis of the fitted parabolic reflector are defined as
elevation angle, ’ela , could be derived as follow u , v , and the change in focal length is defined as
df in UVW coordinates.
’ela ¼ f’ ðel Þ’0a ð15Þ When analyzing the pointing error in the azimuthal
direction, the four parameters, dv, dw, u , and df, do
where ’0a is the first element of ’ a , which is the modal not affect the pointing performance in the UOW
shape vector with the same modal shape. Because there plane. Combined with other factors as shown in
are so many freedom degrees of antenna structure, it is Figure 2, when the vertex moves upward by do and
difficult to find some information on a single point to the focal axis rotation is v , the feed moves by db ; the
express the change in vibration amplitude. Unlike the displacement and rotation of the sub-reflector are dn
sample vectors of MTFNF and MTFMM, the sample and s , respectively. The distance between B0 and the
data, i , of MTFMS is defined as follow focal axis of the sub-reflector is ds1 , which produces a
lateral displacement, ds2 , of the focus F.
sffiffiffiffiffiffiffiffiffiffiffiffiffi
eeli k0m
i ¼ ð16Þ jdb  dn  tan s L1 j
keli e0m ds1 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ cos s jdb þ dn þ tan s L1 j
1 þ ðtan s Þ2

Here eeli and keli are the kinetic energy and modal ð20Þ
stiffness with the same modal shape in different eleva-
tion angles, respectively; e0m , k0m , and ’0a are the ds1
ds2 ¼ ð21Þ
corresponding values of the benchmark angle. Ma
After obtaining the modal mass, damping, and
L1
stiffness matrices in different elevation angles, Ma ¼ ð22Þ
EWCDM could be expressed as L3

Melm q€ m þ Delm q_m þ Kelm qm where L1 is the distance between points B and N; Ma
X s
ð17Þ is the multiplying factor; and L3 is the distance
¼ ðUTel Bodi fdi þ UTel Boli fli þ UTel Bosi fsi Þ between points N0 and F. By omitting the vertex dis-
i¼1 placement of the paraboloid along the W axis, the
Zhang et al. 5

distance du is obtained.Kw is the beam deviation

Model application
factor, which is related to the ratio of the paraboloid As shown in Figure 3, a 7.3-m Ka-band antenna is
focal distance to its open diameter and aperture-field taken as an example; real, finite element, and compu-
distribution. The actual pointing error caused by tational fluid dynamics (CFD) models are depicted in
structural deformation can be expressed as n this figure.
 
jtan v ðL1 þ L3 cos s  ds2 sin s Þ
ðdn þ L3 sin s þ ds2 cos s Þ þ tan v L2  do j
du ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ ðtan v Þ2
ð23Þ
u ¼ dus Kw =f ð24Þ
n ¼  v þ u ð25Þ

Figure 4. Reflector model shares.

Figure 5. Wind pressure coefficient on the front and back in

different directions. (a) Wind direction ¼ 0 ; elevation
angle ¼ 60 . (b) Wind direction ¼ 90 ; elevation angle ¼ 60 . (c)
Figure 2. Pointing error caused by deformation of dual Wind direction ¼ 0 ; elevation angle ¼ 40 . (d) Wind direc-
reflector antenna. tion ¼ 90 ; elevation angle ¼ 40 .

Figure 3. Real, finite element, and CFD models.

6 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

Figure 6. Wind direction wd and elevation angle el .

Zhang et al. 7

By numerical calculation, the CFD model is used compliant with a real antenna, and some analytical
to analyze wind pressure distribution on the reflector; results indicate that the proposed model is effective.
it consists of the main reflector, feed and sub-reflector; Next, it is explained in detail.
the finite element model is used to derive mode infor-
mation of different vibration shape, which lays a
Numerical calculation
foundation for the modal transition function. What
is more, some tests are used to prove that the finite The wind pressure on the pillar-mount and the central
element model of the antenna (7.3 m diameter) was hub is not considered because they have large stiffness

Figure 7. Main vibration shape of the structure. (a) First order; (b) second order; (c) third order; (d) fourth order; (e) fifth order.
8 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

Table 1. Parameters of modal analysis.

Order of modal
Modal shape Elevation angle ( ) Natural frequency (Hz) Kinetic energy (J) shape

Main reflector vibrating in 0 7.19650589 761885.71 1

azimuth direction 10 7.13977302 738918.35 1
20 7.08526336 697067.14 1
30 7.05185883 967339.92 1
40 7.08929227 664470.24 2
50 6.94865482 633221.68 2
60 6.86799539 618374.76 2
70 6.79127448 610167.57 2
80 6.89516746 621175.55 2
90 7.00813535 536715.61 2
Main reflector vibrating in 0 8.19642377 906854.42 2
elevation direction 10 7.67670966 1066287.22 2
20 7.69912160 1627628.01 2
30 7.08806870 1293005.54 2
40 6.53314481 1301506.36 1
50 6.27198185 756497.22 1
60 5.66799485 537849.11 1
70 5.54455238 392719.60 1
80 4.91384040 292117.19 1
90 4.59042072 216904.18 1
Main reflector vibrating 0 8.44440333 1348024.68 3
around its focal axis 10 8.25681593 1444801.34 3
20 8.26789357 1351324.45 3
30 8.13973036 1432604.58 3
40 8.30425407 1437300.29 3
50 8.59675493 1866852.15 3
60 8.83009913 2019369.55 3
70 8.96740499 1983635.64 3
80 10.1864468 1693424.18 3
90 11.9747541 1296898.53 3
Sub-reflector vibrating 0 15.7365 76668.54 4
around its focal axis 10 16.03614 78250.35 4
20 16.32107 72851.39 4
30 16.33524 74668.55 4
40 16.35293 77505.43 4
50 16.24926 80261.33 4
60 16.11182 83955.61 4
70 16.20136 85780.54 4
80 16.12006 88400.35 4
90 16.16183 87329.89 4
Sub-reflector vibrating in 0 20.65868 358980.92 5
azimuth direction 10 22.53585 204348.75 5
20 21.44242 151559.81 5
30 20.9374 141697.26 5
40 20.57687 120182.73 5
50 20.22041 108946.87 5
60 19.91318 113868.74 5
70 19.73323 139187.62 5
80 19.91863 109627.95 5
90 19.50963 97093.11 5
Zhang et al. 9

and small windward area. As shown in Figure 4, the As shown in Figure 5, fluid dynamics was analyzed
reflector model was divided into 38 parts; it comprised using CFD software (ANSYS Fluent); wind pressure
of four parts from the feed, four parts from the sub- coefficient in a different direction of different share
reflector, and 30 parts from the main reflector. was obtained. Then, by giving a wind speed, the
wind force in different direction of each part,
fdi , fli , fsi , could be estimated. Figure 6 illustrates the
definition of wind direction wd and the elevation
angle el .
The antenna structural model contains the pillar-
mount, the central hub, and the reflector, most of
whose materials are steel; however, the reflector is
made from aluminum for the light weight. The finite
model of the antenna is built by using ANSYS 17.2.
It consists of 18,414 elements and 19,483 nodes, and
applying fixed constraints at the bottom of the struc-
ture, the simulating is the nearest the actual condition.
Because the frequency of the dynamic vibration
caused by wind is usually within 20 Hz, the frequency
range of the analysis is also set to 20 Hz. After per-
forming mode analysis, it determines the natural fre-
Figure 8. The Modal Assurance Criterion value between two quency, modal mass, kinetic energy, and the modal
model orders. shape of different order in different elevation angles

Figure 9. Test vibration shapes of the structure. (a) First order; (b) second order; (c) third order; (d) fourth order; (e) fifth order.
10 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

from 0 to 90 at a periodic interval angle (every 10 ). same. However, the value of which order is smaller
Figure 7 illustrates that there are five main vibration than 0.4 could be considered that the two model
shapes of this structure: they are the main reflector orders are with different model shape.
vibrating in azimuth direction as shown in
Figure 7(a), the main reflector vibrating in elevation
Experimental testing
direction as shown in Figure 7(b), the main reflector
vibrating around its focal axis as shown in Figure 7(c), These tests were conducted by loading impulse loads
the sub-reflector vibrating around its focal axis as onto the actual antenna, and 58 accelerometer sensors
shown in Figure 7(d), and the sub-reflector vibrating were placed on the structure. Using modal analyzer,
in azimuth direction as shown in Figure 7(e). The par- natural frequencies of different vibration shapes were
ameters are listed in Table 1. calculated, and the results were compared with those
Using Modal Assurance Criterion which relates of the finite element model with the same vibration
two sets of eigenvectors, the results could be seen as shape. As shown in Figure 9, test vibration shapes of
in Figure 8. The values represent the relevance the structure were depicted using modal test system;
between two model orders; the higher the value, the the shapes are same as those obtained by analyzing
stronger the relevance. The specific values are listed in finite element model. Table 2 presents the comparison
the following table. When the diagonal value is equal of natural frequencies between the test and analysis;
to 1, the order shape of the two models would be the the maximum relative error is about 27.3%.

Table 2. Comparison of natural frequencies.

1st vibration shape (Hz) 2nd vibration shape (Hz) 3rd vibration shape (Hz) 4th vibration shape (Hz) 5th vibration shape (Hz)

ELA TR AR RE TR AR RE TR AR RE TR AR RE TR AR RE

0 7.04 7.20 2.3% 7.21 8.20 13.7% 7.33 8.44 15.1% 15.61 15.74 0.8% 18.00 20.66 14.8%
10 7.45 7.14 4.2% 7.21 7.68 6.5% 7.34 8.26 12.5% 15.63 16.04 2.6% 18.77 22.54 20.1%
20 7.71 7.09 8.0% 7.13 7.70 8.0% 7.37 8.27 12.2% 15.67 16.32 4.1% 18.81 21.44 14.0%
30 7.70 7.05 8.4% 6.76 7.09 4.9% 7.38 8.14 10.3% 15.55 16.34 5.1% 18.43 20.94 13.6%
40 7.66 7.09 7.4% 6.71 6.53 2.7% 7.39 8.30 12.3% 15.47 16.35 5.7% 17.74 20.57 16.0%
50 7.55 6.95 7.9% 6.58 6.27 4.7% 7.41 8.60 16.1% 15.27 16.25 6.4% 17.03 20.22 18.7%
60 7.43 6.87 7.5% 6.50 5.67 12.8% 7.42 8.83 19.0% 14.81 16.11 8.8% 16.47 19.91 20.9%
70 7.18 6.79 5.4% 6.37 5.54 13.0% 7.42 8.97 20.9% 14.02 16.20 15.5% 16.21 19.73 21.7%
80 6.84 6.90 0.9% 6.32 4.91 22.3% 8.13 10.19 25.3% 15.64 16.12 3.1% 16.04 19.92 24.2%
90 6.64 7.01 5.6% 6.35 4.59 27.7% 9.38 11.97 27.6% 15.62 16.16 3.5% 16.00 19.51 21.9%
MRE 8.4% 27.7% 27.6% 13.5% 24.2%

Figure 10. Loading position and direction.

Zhang et al. 11

Figure 12. Sampling values and fitting curves.

Figure 11. Experimental verification.

Table 3. Comparison of deformation.

Test Simulation Calculation

Loading results results results
direction EL angle (mm) (mm) (mm)

Lateral 70 0.80 0.63 0.65

direction 50 0.66 0.55 0.52
30 0.41 0.50 0.45 Figure 13. Tests values and fitting curves.
Vertical 70 1.30 1.23 1.22
direction 50 1.45 1.44 1.40
30 1.63 1.52 1.61

To verify the effectiveness of EWCDM, a load-

deformation test is presented. As shown in
Figure 10 (the red point represent targets and the
arrows denote load direction), impulse loads were
applied in two different directions (lateral and verti-
cal). Targets were set on the maximum deformation in
two different oscillation stages, and displacements
were measured in different oscillation directions
using a laser tracker (API). As shown in Figure 11,
‘‘1’’ represents API, ‘‘2’’ represents target, and ‘‘3’’ rep-
resents the 100 kg load. The maximum displacements
were compared with the calculated results using Figure 14. Pointing error estimation under different operat-
EWCDM, and the simulation results calculated using ing conditions.
the software. Three different elevation angles, 30 , 50 ,
and 70 , were calculated in the tests. Table 3 presents
the results of maximum displacements. by performing mode analysis from 5 to 85 at a peri-
By comparing maximum deformations, the model- odic interval angle (every 10 ). The results indicate
ing method based on EWCDM was verified. that the modal transition function complies with the
fitting date of different elevation angles.
Figure 14 illustrates the pointing error results when
Analytical results
the wind speed is 12 m/s and the wind direction is 70 .
As shown in Figure 12, the technology of modal tran- The maximum pointing error caused by structure
sition function was used to calculate the natural fre- deformation was 0.00487 , 0.00644 , and 0.00837
quency in any elevation angle. The sampling values of when the elevation angle was 10 , 45 , and 80 ,
the five vibration shapes are FOM1–FOM5, and the respectively.
fitting curves are FF1–FF5. Figure 13 illustrates some According to the analysis model of the pointing
test data for verifying fitting curves; they are obtained error, a controller based on PID is designed as
12 Proc IMechE Part C: J Mechanical Engineering Science 0(0)

Figure 15. Antenna servo system.

Figure 16. The system architecture.

Table 4. System parameters.

Symbol Quantity Value Figure 17. The pointing performance with PID controller.
Ra () Motor resistance 0.55
La (H) Motor inductance 0.011 Under a wind with the speed of 12 m/s, when the
kb (Vs/rad) Armature constant 0.35 elevation angle is 45 and wind direction is 70 , the
km (Nm/A) Motor torque constant 1.15 pointing errors in azimuth direction are compared.
kc (V/A) Current feedback coefficient 0.55
Figure 17 shows the different pointing performances
of the antenna servo system with PID controller.
kt (Vs/rad) Velocity feedback coefficient 0.12
Without considering the structure deformation (the
Jm (Nm/s2) The inertia of the motor 0.03
feedback signal of the position loop is ), the max-
and the brake
imum pointing error (DPE) would be up to 0.012 and
kg (Nm/rad) The stiffness of the reducer 2.5  108
the RMS error was 0.0057 . But the maximum point-
N() The reducer gear ratio 354 ing error (FPE) was about 0.008 by introducing n
du
dt () Differentiator – into controller of the position loop, and the RMS
Ja (Nm/s2) The inertia of the 12,000 error was only 0.0034 . The RMS pointing error is
antenna structure reduced by 40.3%. Therefore, it was necessary to
build an accurate model to estimate the pointing
error caused by wind under different operating
shown in Figure 15 (PID controller is the most conditions.
common technique in engineering projects). After
obtaining the wind velocity and direction, the wind
Conclusion
pressure could be obtained and be used to estimate
the pointing error d , and the operation condition Structural deformation caused by wind load has been
(antenna rotation angle ) should be introduced into the main source of pointing errors, especially for those
EWCDM in real time. The architecture and param- antennas with low natural frequency. Thereby, it will
eters of the antenna drivers and rigid model are shown result in significantly decrease of electrical perform-
in Figure 16 and Table 4. ance of antenna. The traditional method analyzed
Zhang et al. 13

and compensated the rotation angle error without subject to wind disturbance. IEEE Antennas Propagat
considering the pointing error caused by structure dis- Magazine 2015; 57: 46–61.
tortion, so it failed to achieve effectiveness against the 3. Gawronski W. Control and pointing challenges of large
wind disturbance. Therefore, based on the entire work- antennas and telescopes. IEEE Transact Control Syst
Technol 2007; 15: 276–289.
ing condition dynamic model which is proposed for the
4. Gawronski W. Modeling wind-gust disturbances for the
first time, the pointing error analytical model is derived
analysis of antenna pointing accuracy. IEEE Antennas
to estimate the pointing error under different operating Propagat Magazine 2004; 46: 50–58.
conditions, and the results are used to understand the 5. Gawronski W. Servo-performance of the NASA deep
impact on the pointing accuracy. From the test results, space network antennas. IEEE Antennas Propagat
it not only provides a verification that the finite model Magazine 2007; 49: 40–46.
is almost close to the actual antenna structure, but also 6. Gawronski W and Souccar K. Control systems of the
gives a verification that the modal transition phenom- large millimeter telescope. IEEE Antennas Propagat
enon exists objectively, which lays a foundation for the Magazine 2005; 47: 41–49.
proposed method. From the analytical results, it shows 7. He G, Li G, Liu M, et al. LQG controller with wind
that the pointing error caused by the structure deform- gust disturbance rejection property for cassegrain
antenna. In: Proceedings of the IEEE international con-
ation should not be ignored, and it is very much neces-
ference on intelligent computing and intelligent systems,
sary to introduce the entire working condition dynamic
Shanghai, China, 20–22 November 2009, pp.344–348.
model into the pointing error analysis, for the pointing 8. Li P, Duan BY, Wang W, et al. Electromechanical cou-
errors at different elevation angle are making a differ- pling analysis of ground reflector antennas under solar
ence. Meanwhile, the proposed method could effect- radiation. IEEE Antennas Propagat Magazine 2012; 54:
ively estimate the pointing error caused by wind, and 40–57.
the final results show that the pointing error could be 9. Haddadi A and Ghorbani A. Distorted reflector anten-
reduced by 40.3%. nas: analysis of radiation pattern and polarization per-
formance. IEEE Transact Antennas Propagat 2016; 64:
Declaration of Conflicting Interests 4159–4167.
10. Zhang J, Huang J, Zhou J, et al. A compensator for
The author(s) declared no potential conflicts of interest with
large antennas based on pointing error estimation under
respect to the research, authorship, and/or publication of
a wind load. IEEE Transact Control Syst Technol 2017;
this article.
5: 1912–1920.
11. Zhang J, Huang J, Wang S, et al. An active pointing
Funding compensator for large beam waveguide antenna under
The author(s) disclosed receipt of the following financial sup- wind disturbance. IEEE/ASME Transact Mechatron
port for the research, authorship, and/or publication of this 2016; 21: 860–871.
article: This work was supported by the National 973 12. Nobuharu U. Wind-induced pointing errors and sur-
Program (grant no. 2015CB857100), the National Natural face deformation of a 10-m submillimeter antenna.
Science Foundation of China (grant nos 51705387, In: Proceedings of SPIE, ground-based airborne tele-
51575419, and 51490660), the National 111 Project (grant scopes, 7733, http://proceedings.spiedigitallibrary.org/
no. B14042), the China Postdoctoral Science Foundation proceeding.aspx?articleid¼750338 (2010 accessed 28
(grant no. 2017M613078), and the Fundamental research July 2017).
Funds for the Central Universities (grant no. XJS17069, 13. Duan B and Wang C. Reflector antenna distortion ana-
and JBX170414). lysis using MEFCM. IEEE Transact Antennas Propagat
2009; 57: 3409–3413.
ORCID iDs 14. Snel RC, Mangum JG and Baars JWM. Study of the
dynamics of large reflector antennas with accelerom-
Jie Zhang http://orcid.org/0000-0002-4024-526X
eters. IEEE Antennas Propagat Magazine 2007; 49:
Pengbing Zhao http://orcid.org/0000-0002-8330-3802
84–101.
15. Wang CS, Duan BY and Qiu YY. Precise algorithm for
References surface errors of reflector antennas and analysis of its
1. Strain D. Optimization of 100-meter Green Bank electrical performance. Chinese J Radio Sci 2006; 21:
Telescope. NASA Sti/recon Technical Report N, 23–28.
pp.94-96, 1994.
2. Zhang J, Huang J, Song R, et al. Analysis of reflector
vibration-induced pointing errors for large antennas