Professional Documents
Culture Documents
2-Propagation and Transformation of Mean Elements at Geostationary Orbits
2-Propagation and Transformation of Mean Elements at Geostationary Orbits
Engineering Notes
Propagation and Transformation of theory of Kaula [12] using an iterative procedure, thus allowing
precise computation of the mean elements. These works were mainly
Mean Elements at Geostationary Orbits focused on low Earth orbits (LEO), and so the perturbation from the
aspherical gravity of the Earth received the most attention because it
dominates all other periodic effects on LEO satellites. Ely [13] used a
Hong-Xin Shen,∗ Zheng-Zhong Kuai,† and Heng-Nian Li‡ numerical method to compute the Fourier series coefficients; this
Xi’an Satellite Control Center, 710043 Xi’an, method has the advantage of finding the short-period terms due to any
People’s Republic of China perturbing acceleration, but probably does not provide physical
insight into the satellite motion. The work by Li [14] using a singular
DOI: 10.2514/1.G003654
value decomposition method is similar to the work by Ely.
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
Section IV separates short-period terms from the long-period and parameter θk is the angle between the two position vectors rk and r, as
secular terms, thus providing the equations for the mean elements. given by krk − rk2 r2k r2 − 2rk r cos θk.
Section V describes the suggested averaging algorithm and derives A simplified expression of the solar radiation acceleration
the analytic expressions for the short-period amplitude, which is used commonly used for numerical simulations is adopted as given by
to switch between osculating and mean elements. The mean elements Eq. (7) [22]:
propagation and transformation models are tested in Sec. VI.
Rsp −CR A∕mP0 xs x ys y zs z (7)
II. Preliminaries where CR means the radiation pressure coefficient, A∕m is the
A GEO has a near-zero inclination i and eccentricity e, and so the satellite’s area-to-mass ratio, and coefficient P0 4.56e-6 N∕m2 is
right ascension of the ascending node Ω and the argument of the the solar radiation pressure in the neighborhood of the Earth. Note
perigee ω are not conveniently defined. Here only semimajor axis, that the solar radiation pressure is eliminated when Earth shadow
mean longitude, and eccentricity vector are concerned. A set of occurs. The time of Earth shadow is easily predetermined.
nonsingular orbital elements [21] is used here to avoid singularity
issues. The variables ey e sinω Ω and ex e cosω Ω,
which relate to the change of the eccentricity magnitude and rotation
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
of the line of apsides, are introduced. Based on the fact that GEO III. Propagation of Osculating Elements
satellites have near-zero inclination and eccentricity, derivatives of In this section, the semi-analytical astrodynamical models used for
the semimajor axis a, mean longitude l, and eccentricities ex and ey the propagation of the osculating elements will be discussed,
can be reduced to the form given by Eqs. (1–4): including Earth’s aspherical gravity, third-body gravity, as well as
solar pressure.
da 2 ∂R
(1)
dt na ∂l A. Semimajor Axis
According to Eqs. (1) and (5) the rate of change of the semimajor
dl 2 ∂R axis caused by Earth’s gravity is given by Eq. (8):
− (2)
dt na ∂a
da p
2 μ−6J22 r2e a−5∕2 sin2l − l22
dey 1 ∂R dt E
2 (3)
dt na ∂ex 3∕2J31 r3e a−7∕2 sinl − l31 − 45J33 r3e a−7∕2 sin3l − l33
15J 42 r4e a−9∕2 sin2l − l42 − 420J44 r4e a−9∕2 sin4l − l44
dex 1 ∂R
− 2 (4) (8)
dt na ∂ey
p
where R is disturbing function, n μ∕a3 is the mean motion, and The third-body gravity disturbing function given by Eq. (6)
μ is the Earth’s gravitational constant. Also, l Ω ω M − Θ contributes to the variation in the semimajor axis according to
λ − Θ is mean longitude, where Θ is the Greenwich sidereal angle. Eq. (9):
The disturbing function R is the sum of three main disturbing
functions: RE , Rls , and Rsp , resulting from the perturbations due to s
da a3 μk ∂cos θk
the nonspherical Earth gravity, the lunisolar attraction, and the solar 2 1 a∕rk 2
radiation pressure, respectively. Without loss of generality we dt ls μ r2k ∂l
consider the gravity field up to degree and order 4 as shown by Eq. (5) − 2a∕rk cos θk −3∕2 − 1 (9)
[14]:
μ 1 We here neglect the influences of small eccentricity and inclination
RE − J2 re ∕r2 3J22 re ∕r2 cos2l − l22 on cos θk , thus ∂cos θk ∕∂l is given by ∂cos θk ∕∂l
r 2
−xk sin λ yk cos λ, where λ is defined by λ l Θ.
3
− J31 re ∕r2 cosl − l31 15J33 re ∕r3 cos3l − l33 The solar pressure disturbing function given by Eq. (7) contributes
2 to the rate of change of the semimajor axis as shown by Eq. (10):
3 15
J 4 re ∕r4 − J42 re ∕r4 cos2l − l42
8 2 s
105J44 re ∕r4 cos4l − l44 (5) da a3
−2 C A∕mP0 −xk sin λ yk cos λ (10)
dt sp μ R
where r is the distance to the satellite in Earth-centered J2000 frame,
re is the Earth’s equatorial radius, and l22 , l31 , l33 , l42 , and l44 With the assumption of small eccentricity and inclination, the
represent the coefficients for the gravitational model that can be position vector becomes r a cos λ a sin λ 0 T .
found in Ref. [14]. Therefore, the semimajor axis variation is given by Eq. (11):
The disturbing function resulting from the third-body
perturbations can be given in Earth-centered coordinates as shown
in Eq. (6) [14]: da da da da
(11)
dt dt E dt ls dt sp
Rls μk 1∕krk − rk − r∕r2k cos θk (6)
∂r
dl p −a sin λ (21)
−2 μ3∕2J2 r2e a−7∕2 − 9J22 r2e a−7∕2 cos2l − l22 ∂ey
dt E
6J31 r3e a−9∕2 cosl − l31 −60J33 r3e a−9∕2 cos3l − l33
∂r cos θk
−15∕8J4 r4e a−11∕2 75∕2J42 r4e a−11∕2 cos2l − l42 axk −3 cos2λ yk sin2λ∕2 (22)
∂ex
−525J44 r4e a−11∕2 cos4l − l44
(12) ∂r cos θk
axk sin2λ − yk 3 cos2λ∕2 (23)
The third-body gravity disturbing function contributes to the rate ∂ey
of change of mean longitude as given by Eq. (13):
It is worth noting that the small values of eccentricity cannot be
r ignored in cos θk when considering eccentricity variation, and so the
dl a
−2 μ r2 a2 − 2rk a cos θk −3∕2 rk cos θk − a expression for cos θk is different from that in Eq. (9). In contrast,
dt ls μ k k when only phasing variation is considered, the terms involving
− cos θk ∕r2k (13) eccentricity are too small and thus can be neglected. Using Eq. (19),
we have ∂cos θk ∕∂l xk − sin λ − ex sin2λ ey cos2λ
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
The solar radiation pressure does not influence the rate of change yk cos λ ex cos2λ ey sin2λ, which reduces to
of the mean longitude because ∂RP ∕∂a 0, which therefore ∂cos θk ∕∂l −xk sin λ yk cos λ due to the small values of
gives dl∕dtsp 0. eccentricity. This expression is in agreement with the earlier
Finally, because the satellite’s mean rotating rate may be different derivations.
from Earth’s rotating rate, one has Eq. (14): One also gets the variations in eccentricity that result from the solar
pressure perturbation given by Eqs. (24) and (25):
dl dl dl r
n − ωe (14) dey a A 3 1 1
dt dt E dt ls − CR P0 − xs xs cos2λ ys sin2λ
dt sp μ m 2 2 2
where ωe is the Earth’s mean rotating angular velocity. (24)
C. Eccentricity r
dex a A 3 1 1
As Earth’s aspherical perturbation does not provide significant CR P0 − ys xs sin2λ − ys cos2λ
dt sp μ m 2 2 2
influence on the variation in eccentricity, only J2 perturbation is
considered given by Eqs. (15) and (16): (25)
Therefore, these result in Eqs. (26) and (27):
dey 3 p
− J 2 μr2e a−7∕2 cos λ (15)
dt E 2
dey dey dey dey
(26)
dt dt E dt ls dt sp
dex 3 p
J2 μr2e a−7∕2 sin λ (16) dex dex dex dex
dt E 2 (27)
dt dt E dt ls dt sp
The third-body perturbation contributes to the variation in
eccentricity according to Eqs. (17) and (18):
s IV. Propagation of Mean Elements
dey 1
μk r2k r2 − 2rk r cos θk −3∕2 To construct a model for the propagation of mean elements, the
dt ls μa secular perturbation and long-term perturbation should be separated
∂r cos θk ∂r 1 ∂r cos θk from the short-term perturbation. Note that these mean elements are
× rk −r − 2 (17) single-averaged values resulting from eliminating short-period
∂ex ∂ex rk ∂ex
variation. They therefore have no influence on the secular behavior,
and hence it is not necessary to separate the secular terms and long-
s period terms in the current work. The mean values are used to
dex 1 establish the osculating elements as given by Eq. (28)
− μ r2 r2 − 2rk r cos θk −3∕2
dt ls μa k k
(18)
∂r cos θk ∂r 1 ∂r cos θk ds ds ds
× rk −r − 2 (28)
∂ey ∂ey rk ∂ey dt dt dt short
r a1 − ex cos λ − ey sin λ, and where s is the averaged variation of an orbit element s, which is
influenced by a disturbing function. ds∕dt
represents the variation of
mean elements, which will be described in this section, and
ds∕dtshort would be useful for the switch between mean and
cos θk xk cos λ − 3ex ∕2 ex cos2λ∕2 ey sin2λ∕2
osculating elements in the following section.
yk sin λ − 3ey ∕2 ex sin2λ∕2 − ey cos2λ∕2 (19) We assume that the differential equations for mean elements can be
written as Eqs. (29–32):
∂r da da da da
−a cos λ (20) (29)
∂ex dt dt E dt ls dt sp
J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES 2135
dl dl dl dl dl x2k y2k x2k y2k
(30) cos2 θk mean , cos2 θk short cos2 θk −
dt dt E dt ls dt sp dt K 2 2
(36)
de y de y dey de y Thus, lunar-solar gravity results in both a long-term effect and
(31) short-term effect as given by Eqs. (37) and (38):
dt dt E dt ls dt sp
r
dl a μk 3
−2 −1 x2k y2k a∕rk
de x de x dex de x dt ls μ r2k 2
(32)
dt dt E dt ls dt sp 3 35
1 − 5x2k y2k x2k y2k 2 a∕rk 3 (37)
2 8
where dl∕dt
K is an additional term resulting from semimajor axis
variation rather than perturbations. r
dl a μk
−2 3cos2 θk short a∕rk
A. Semimajor Axis dt ls,short μ r2k
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
It can be observed that there are no periodic terms in the right hand −9 cos θk 15cos3 θk ∕2a∕rk 2 (38)
of Eq. (8), for instance, the radius r and mean longitude l of
geostationary orbits change quite slightly, and so the variation in the According to the analysis in Sec. III, the mean and short-period
semimajor axis resulting from Earth’s gravity has a linear trend variations resulting from the solar pressure are null. Moreover, the
without short-term variations and thus can be written as short-term oscillation of the semimajor axis Δa results in a short-term
da∕dt
E da∕dtE , da∕dtE,short 0. This should be expected, oscillation of the mean longitude with the same frequency. Therefore,
because the longitude acceleration can be considered to be constant the mean and short-term perturbations for the mean longitude are
depending on the longitude location [4], which means that the drift obtained from Eq. (39):
rate is almost a constant. A constant drift rate results in a linear trend
in the semimajor axis. r
dl dl 3 μ
To approximate 1 a∕rk 2 − 2a∕rk cos θk −3∕2 − 1 precisely, n − ωe , − Δa (39)
in particular for the Moon’s perturbation (i.e., k m), Taylor dt K dt K,short 2 a5
expansions up to order three are taken into account in terms of a∕rk as
shown in Eq. (33):
C. Eccentricity
1 a∕rk 2 − 2a∕rk cos θk −3∕2 − 1 3a∕rk cos θk According to Eqs. (15) and (16) the Earth’s aspherical perturbation
p
is quite small because 3∕2J 2 μR2e a−7∕2 ≈ 2.7 × 10−7 , and so the
3 5
−1 5cos2 θk a∕rk 2 −3 cos θk 7cos3 θk a∕rk 3 variation in eccentricity resulting from Earth’s gravity is
2 2 neglected here.
(33) By using Taylor expansions, one obtains Eqs. (40) and (41):
s
Note that one-order Taylor expansion could be precise enough for de y 15 a5 μk
the computation of the Sun’s perturbation due to the very small value x 4 − 5x2k − 5y2k
of a∕rs . dt ls 16 μ r4k k
(40)
Because of the influence of the term ∂cos θk ∕∂l −xk sin λ dey dey dey
yk cos λ in Eq. (9), only short-period perturbation arises from the −
dt ls,short dt ls dt ls
lunar-solar attraction. Taking first-order Taylor expansions for Eq. (9)
as an example, one obtains Eq. (34):
s
r de x 15 a5 μk
da a a2 2 − y 4 − 5x2k − 5y2k
3 μ y − x2k sin2λ 2xk yk cos2λ (34) 16 μ r4k k
dt ls μ k r3k k dt ls
dex dex dex
It can be seen that lunar-solar gravity results in a semidaily term − (41)
dt ls,short dt ls dt ls
perturbation given by da∕dt
ls 0 and da∕dtls,short da∕dtls ,
and this fact also can be observed when using third-order Taylor
series approximation without needing the detailed expressions. from which it can be seen that both the short-period and long-period
Also, using Eq. (10) we get da∕dt perturbations result from the lunar and solar attractions. Also, the
sp 0 and da∕dtsp,short
da∕dtsp . solar pressure contributes to the variation in eccentricity in both long-
period and short-period forms as given by Eqs. (42) and (43):
B. Mean Longitude r
de y a A
3
Similar to that for the semimajor axis, Earth’s gravity results only C P x
dt sp μ Rm 0 s
2
in a secular perturbation on the mean longitude as given by r
dl∕dt
E dl∕dtE and dl∕dtE,short 0. dey 1 a A
− C P x cos2λ ys sin2λ (42)
Through second-order Taylor expansions of Eq. (13) we get dt sp,short 2 μ Rm 0 s
Eq. (35):
r
dl a μk
−2 −1 3cos2 θk a∕rk r
dt ls μ r2k dex 3 a A
− C P y
−9 cos θk 15cos3 θk ∕2a∕rk 2 (35) dt sp 2 μ Rm 0 s
r
dex 1 a A
It can be observed that the odd powers of cos θk only provide a C P x sin2λ − ys cos2λ (43)
short-term effect, and we have Eq. (36): dt sp,short 2 μ Rm 0 s
2136 J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES
To have a full view of the characteristics of the perturbations, the Equation (46) is rewritten as follows by using dθk ∕dt n − nk :
perturbation influences on elements of semimajor axis, mean s Z
longitude, and eccentricity are summarized in Table 1. It is worth a3 μk
noting that some secular or long-period perturbations were ignored in Δals −2 sin θk 1 a∕rk 2 − 2rk a cos θk −3∕2
recent literature [14]. For instance, the lunar-solar attraction for mean μ r2k
longitude and eccentricity was not considered, whereas these flaws − 1∕n − nk dθk − C (49)
are fixed in this work.
where nk xk y_k − x_ k yk ∕x2k y2k means the Moon’s or the Sun’s
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
rotating angular velocity relative to the Earth. Note that the Moon’s
V. Transforming Mean and Osculating Elements rotation must be considered in order to maintain good accuracy,
whereas the Sun’s rotation rate relative to the Earth is so small that it
An analytical mean elements calculation scheme isolates the short-
can be neglected.
period terms Δs as defined by Eq. (44):
One obtains the ampl of semimajor axis given by Eq. (50) by using
the approximation given by Eq. (33):
s s − Δs (44)
s
a3 μk 3
In the current work the amplitude is obtained using an indefinite Δals −2 − a∕rk cos2 θk short
integral of the short-period terms associated with the removal of the μ r2k n − nk 2
nonperiodic part, that is, as given by Eq. (45): 3 cos θk − 5cos3 θk ∕2a∕rk 2
Z 5
Δs
ds
dt − C (45) 6cos2 θk short − 7cos4 θk short a∕rk 3 (50)
dt short 8
s
where Δals and Δasp are given by Eqs. (50) and (52), respectively.
a3 μk
One obtains the expression of ΔlK as follows: Δey ls sin λ − x2k sin3 λ 6xk yk cos λ
μ n − nk r3k
r
3 μ 1 − 2xk yk cos3 λ−3x2k sin λ y2k sin3 λ
ΔlK − A1 xs sin λ − ys cos λ
2 a5 n − ns 3
a∕rk xk cos2λ − yk sin2λ (58)
1 4
A x2 − y2k sin2λ∕4 − xk yk sin2λ∕2 (56)
n − nk 2 k
p p
where A1 −2 a3 ∕μCR A∕mP0 ∕ωe , A2 3 a5 ∕μ1∕ s
rk μk ∕n − nk .
3
a3 μk
Therefore the amplitude of the mean longitude is Δex ls − − cos λ − x2k cos3 λ − 6xk yk sin λ
μ n − nk r3k
104
4.217 5
Semimajor axis,km
,m
4.2165
-1
-3
4.216 -5
0 7 14 21 28
Time,day
a) Semimajor axis a
10-3
146.4 5
Mean longitude,deg
,deg
146.2 0
146 -5
0 7 14 21 28
Time,day
b) Mean longitude l
104 106
0 5
-0.5 MEPP OSC OSC-HPOP 4
-1 3
ey
ey
-1.5 2
-2 1
-2.5 0
-3 -1
0 7 14 21 28
Time,day
c) Eccentricity ey
104 104
0 2
MEPP OSC OSC-HPOP
-1 0
ex
ex
-2 -2
-3 -4
0 7 14 21 28
Time,day
d) Eccentricity ex
Fig. 1 Comparison between MEPP and HPOP (case 1). a) Semimajor axis a. b) Mean longitude l. c) Eccentricity ey . d) Eccentricity ex .
2138 J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES
r
1 a A 1
Δex sp − C P x cos2λ ys sin2λ (61) facilitate the comparison with the results we use the high precision
4 μ Rm 0n s orbit propagation (HPOP) method of STK that uses the same models.
In the following discussion, the resulting mean elements will
To summarize, the total amplitude of the eccentricity vector is be shown.
given by Eqs. (62) and (63):
A. Propagation of Mean Elements
Δey Δey ls Δey sp (62)
In Figs. 1 and 2 we can observe the osculating orbital elements over
28 days and the mean elements precise propagation (MEPP) method
Δex Δex ls Δex sp (63) presented in this Note. In the plots the legend “OSC” represents the
osculating elements generated using the method described in Sec. III.
The osculating semimajor axis has a total variation up to about 2 km.
The MEPP results agree very well with the averaged evolution of the
VI. Validation osculating results. Because the mean longitude does not vary
Tests of the above expressions start with osculating trajectories that significantly, as shown in Figs. 1b and 2b, this fact results in an almost
include perturbations from Earth’s nonspherical gravity, Moon and linear change in the average semimajor axis over the current time
Sun’s gravity, and solar radiation pressure. For the central body span. Results given by the other y axis stand for the difference
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
gravity all harmonics up to degree and order 4 are considered. The between the OSC method and the HPOP method, denoted by the
osculating trajectories are generated with the initial conditions given legend “OSC-HPOP” in the plots. The comparison results show that
in the J2000 frame. Two illustrative scenarios are considered: a the osculating-elements model has good agreement with the HPOP of
42164.8 km and M 280.0 deg for case 1, whereas a STK. It is worth noting that the osculating-elements model and
42.166.8 km and M 180.0 deg for case 2. The other orbital HPOP cannot be exactly the same, as the osculating-elements model
elements and epoch are the same for both scenarios: e 0.0003, uses a new system based on the geostationary assumptions.
i 0.09 deg, Ω 54.0 deg, ω 162.0 deg, and Epoch = March Moreover, the change trend of mean longitude is visible in the
12 2018 12:00:00 UTC. For both cases, the area-to-mass ratio is evolution of mean semimajor axis. This is because the mean value of
A∕m 0.02. The radiation pressure coefficient is CR 1.5. To the geostationary semimajor axis is around 42,165.7 km [4,14],
10 4
4.217 5
Semimajor axis,km
,m
4.2165
-1
-3
4.216 -5
0 7 14 21 28
Time,day
a) Semimajor axis a
10 -3
46.2 2
Mean longitude,deg
46 0
45.8 -2
0 7 14 21 28
Time,day
b) Mean longitude l
10 -6
10 -4
0.5 6
0 MEPP OSC OSC-HPOP 5
-0.5 4
-1 3
ey
ey
-1.5 2
-2 1
-2.5 0
-3 -1
0 7 14 21 28
Time,day
c) Eccentricity ey
10 -6
10 -4
0 1
MEPP OSC OSC-HPOP
-1 0
ex
ex
-2 -1
-3 -2
-4 -3
0 7 14 21 28
Time,day
d) Eccentricity ex
Fig. 2 Comparison between MEPP and HPOP (case 2). a) Semimajor axis a. b) Mean longitude l. c) Eccentricity ey . d) Eccentricity ex .
J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES 2139
which corresponds to a null drift rate of longitude. Thus, when the B. Converting Osculating to Mean Elements
mean semimajor axis grows above this critical value, an opposite Time series of the osculating and converted mean elements for 28
longitude drift arises, which has a sort of stabilizing effect. days are generated and are presented in Figs. 3 and 4. In the plots the
A significant motivation for the use of the mean-elements method legend “MEAN” implies that short-period perturbations have been
is to increase the allowable propagation step size compared with subtracted from the osculating elements. It can be observed that the
propagating the osculating elements. Propagation of the osculating mean semimajor axis has a variation only about 5 m (see Figs. 3b
elements requires an integration step size of the order of 10 min based and 4b), and the remaining variation of mean longitude almost tends to
on the conventional RKF7(8) integrator. However, if we remove the disappear (see Figs. 3c and 4c). Thus, the mean elements of semimajor
short-period variations, we can propagate the remaining equations of axis and mean longitude in Figs. 3 and 4 look nearly identical to that in
motion for mean elements, and the integration step size can be of the Figs. 1 and 2. It is important noting that the variation of semimajor axis
order of 1 day. This fact demonstrates that the goal of efficient would grow up to tens of meters without considering the Moon’s
propagation of the motion is achieved very well. rotating angular velocity relative to the Earth, as illustrated by Eq. (49).
10 4
4.2168
Semimajor axis,km
MEAN OSC
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
4.2166
4.2164
4.2162
0 7 14 21 28
Time,day
a) Semimajor axis a
10 4
4.21665
Semimajor axis,km
4.2166
4.21655
10 4
4.2165 4.21658
4.21657
4.21645 4.21656
19.5 20.5 21.5 22.5
4.2164
0 7 14 21 28
Time,day
b) Mean semimajor axis a
146.4
Mean longitude,deg
MEAN OSC
146.3
146.2
146.1
0 7 14 21 28
Time,day
c) Mean longitude l
10 -4
0
MEAN OSC
-1
ey
-2
-3
0 7 14 21 28
Time,day
d) Eccentricity ey
10 -4
-0.5
-1 MEAN OSC
-1.5
ex
-2
-2.5
-3
0 7 14 21 28
Time,day
e) Eccentricity ex
Fig. 3 Transforming mean elements from osculating elements (case 1). a) Semimajor axis a. b) Mean Semimajor axis a. c) Mean longitude l.
d) Eccentricity ey . e) Eccentricity ex .
2140 J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES
10 4
4.217
Semimajor axis,km
MEAN OSC
4.2168
4.2166
4.2164
4.2162
0 7 14 21 28
Time,day
a) Semimajor axis a
10 4
4.2168
Semimajor axis,km
4.2167
4.2166
10 4
4.2165
4.2165 4.21649
4.2164 4.21648
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
20.5 21 21.5 22
4.2163
0 7 14 21 28
Time,day
b) Mean semimajor axis a
46.15
Mean longitude,deg
MEAN OSC
46.1
46.05
46
45.95
0 7 14 21 28
Time,day
c) Mean longitude l
10 -4
1
MEAN OSC
0
ey
-1
-2
-3
0 7 14 21 28
Time,day
d) Eccentricity ey
10 -4
-1
MEAN OSC
-2
ex
-3
0 7 14 21 28
Time,day
e) Eccentricity ex
Fig. 4 Transforming mean elements from osculating elements (case 2). a) Semimajor axis a. b) Mean Semimajor axis a. c) Mean longitude l.
d) Eccentricity ey . e) Eccentricity ex .
In Figs. 3d, 3e, 4d, and 4e we can see the osculating and mean osculating results agree very well with the conversion from the
eccentricities. After removing the short-period perturbations averaged evolution of the MEPP to the osculating results. The
resulting from the third-body attraction and solar pressure, the comparison results at the end time are given by Tables 2 and 3,
remaining variation in mean eccentricity is of order 1e-6. Also, the corresponding to case 1 and case 2, respectively.
Table 2 Comparison between mean-to-osculating elements and Table 3 Comparison between mean-to-osculating elements and
osculating elements (case 1) osculating elements (case 2)
Propagation method a, km l, deg ey ex Propagation method a, km l, deg ey ex
MEPP 42,166.2609 146.3463 −0.5651e-4 −2.1147e-4 MEPP 42,164.0064 46.1038 −0.1158e-4 −2.5630e-4
Mean to osculating 42,166.6457 146.3458 −0.3411e-4 −2.5000e-4 Mean to osculating 42,163.5589 46.1047 0.0374e-4 −2.5681e-4
Osculating 42,166.6450 146.3461 −0.3383e-4 −2.4991e-4 Osculating 42,163.5600 46.1051 0.0447e-4 −2.5733e-4
J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES 2141
104
4.217
Semimajor axis,km
4.2165 MEAN OSC
4.216
4.2155
4.215
4.2145
0 50 100 150 200 250 300 350 400
Time,day
Fig. 5 Transforming mean semimajor axis from osculating semimajor axis in 1-year timescale (case 2).
Theoretically, when converting from osculating elements to mean The existing technique of permitted access probably is the SGP4
elements, it is preferred to use mean elements that is unknown as model (http://www.centerforspace.com/downloads/), and so a
input for computing the short-period amplitude so that the iteration comparison is especially relevant. Although there is no certainty
process could be required when using known osculating elements as that the SGP4 theory has no value in GEO, one finds that the SGP4
input; however, because mean and osculating values are close to each performs inferiorly to that in LEO cases. This fact is mainly
Downloaded by National University of Defense Technology on October 16, 2019 | http://arc.aiaa.org | DOI: 10.2514/1.G003654
other, numerical tests show that different input initial conditions do demonstrated in two aspects: 1) the generating osculating semimajor
not make any significant change in the conversion results, and so the axis almost has no short-period oscillation, and this fact may imply
iteration process in fact is not needed in this work. that the model cannot describe the actual position of GEO satellites;
2) the generating osculating eccentricity has no long-period variation,
C. Discussion and due to that the solar radiation pressure is not taken into
This Note introduces a new differential system in order to consideration in this model. Besides, an alternative state-of-the-art
efficiently propagate and transform the orbital elements in the method for osculating/mean elements conversion is developed by Lin
perturbed geostationary regime. The new system contains the Liu from Nanjing University [23]; however, the remaining variation
essential features of the initial system, but it is more efficient to in the mean semimajor axis is normally about 10 m according to the
simulate numerically. As a matter of fact, the new system is effective internal documents. Moreover, the osculating/mean elements
particularly in the geostationary regime. Because of the effect of the conversion method in this Note is fully analytical and of significant
equilibrium points in GEO resulting from the perturbations, the GEO advantage over Liu’s theory in computation efficiency; the average
is somehow stable even for long propagation times, and so the CPU time for single point conversion is 1.562e-5 s vs.0.145 s. This
geostationary assumptions (orbital period close to a sidereal day, computation process is executed on a desktop personal computer
near-zero eccentricity, and inclination) are satisfied with sufficient with an i7 CPU at 3.2 GHz, and Microsoft Visual C++ 2015 is used.
accuracy. An improvement of four orders of magnitude in CPU time (while
A numerical test for a 1-year propagation time is presented keeping the superiority of conversion accuracy) is found.
particularly for scenario 2. Figures 5 and 6, respectively, show the Because the current work is based on a simplified form of the
evolution of semimajor axis and eccentricity. It can be seen that Lagrange planetary equations, which are obtained in case of low
semimajor axis varies around its normal value, and eccentricity does eccentricity and inclination, when the eccentricity and inclination
not diverge from zero. Also, 1 year is almost the period of eccentricity grow beyond some upper limits, the theory developed in this work
evolution due to the dominating role of solar-pressure perturbation, would become inaccurate. Relative simulations have been performed
as shown by Fig. 6. Therefore, the authors have confidence (even to highlight this fact, and then it is found that the increased
though not the certainty) that the method is applicable to a larger eccentricity and inclination, respectively, have the greatest impact on
timescale simulation as long as the orbit is near geostationary. the mean elements of eccentricity and mean longitude. In the
Actually, the mean elements model was developed for the precise concrete, noticeable divergence of the mean eccentricity arises when
maneuver planning of station keeping/acquisition. In this respect, the the initial eccentricity is increased to 0.001, and 2-deg inclination
28-day timescale is rather enough. On the other hand, probably only makes the mean motion of the mean longitude start to diverge from
the secular terms (rather than secular and long-period terms together) the accurate one.
deserve consideration for a larger timescale (years or decades).
10 -4 VII. Conclusions
4
MEAN OSC
An algorithm for converting between osculating and mean orbit
elements for geostationary satellites has been developed, and the
expressions are provided in this Note. Through perturbation analysis,
2 the short-period terms were separated from long-period and secular
terms, thus obtaining the equations to propagate the osculating and
mean elements. Through an indefinite integration of short-period
0 terms, the amplitude of each element is obtained and used for the
End
conversion. The conversion method does not use any iteration
ey
-6 Acknowledgments
-10 -8 -6 -4 -2
ex -4 This work was supported by the National Natural Science
10
Fig. 6 Transforming mean eccentricity from osculating eccentricity in Foundation of China (No. 11702330). We thank Lorenzo Casalino
1-year timescale (case 2). for his useful suggestions.
2142 J. GUIDANCE, VOL. 42, NO. 9: ENGINEERING NOTES
[6] Kozai, Y., “The Motion of a Close Earth Satellite,” The Astronomical [18] Hoots, F. R., Schumacher, P. W., Jr., and Glover, R. A., “History of
Journal, Vol. 64, Nov. 1959, pp. 367–377.
Analytical Orbit Modeling in the U.S. Space Surveillance System,”
doi:10.1086/107957 Journal of Guidance, Control, and Dynamics, Vol. 27, No. 2, 2004,
[7] Danielson, D. A., Sagovac, C. P., Neta, B., and Early, L. W., pp. 174–185.
“Semianalytic Satellite Theory,” Naval Postgraduate School, NPS MA- doi:10.2514/1.9161
95-002, Monterey, CA, 1995. [19] Gurfil, P., and Seidelmann, P. K., Celestial Mechanics and
[8] Zeis, E., and Cefola, P., “Computerized Algebraic Utilities for the Astrodynamics: Theory and Practice, Springer, Berlin, 2016, Chap. 13.
Construction of Nonsingular Satellite Theories,” Journal of Guidance,
[20] Zhong, W., and Gurfil, P., “Mean Orbital Elements Estimation for
Control, and Dynamics, Vol. 3, No. 1, 1980, pp. 48–54.
Autonomous Satellite Guidance and Orbit Control,” Journal of Guidance,
doi:10.2514/3.55946 Control, and Dynamics, Vol. 36, No. 6, 2013, pp. 1624–1641.
[9] Gao, Y., “Low-Thrust Nonlinear Guidance by Tracking Mean Orbital doi:10.2514/1.60701
Elements,” Journal of Guidance, Control, and Dynamics, Vol. 31, No. 4, [21] Zhao, S. G., Gurfil, P., and Zhang, J. R., “Optimal Servicing of
2008, pp. 1103–1110. Geostationary Satellites Considering Earth’s Triaxiality and Lunisolar
doi:10.2514/1.31256
Effects,” Journal of Guidance, Control, and Dynamics, Vol. 39, No. 10,
[10] Guinn, J. R., “Periodic Gravitational Perturbations for Conversion
2015, pp. 2219–2231.
Between Osculating and Mean Orbit Elements,” AAS/AIAA doi:10.2514/1.G001424
Astrodynamics Specialist Conference, AAS Paper 91-430, Springfield, [22] Montenbruck, O., and Gill, E., Satellite Orbits—Models, Methods, and
VA, 1991. Applications, Springer–Verlag, Berlin, 2000, pp. 77–79.
[11] Spiridonova, S., Kirschner, M., and Hugentobler, U., “Precise Mean [23] Liu, L., and Tang, J., Satellite Orbit Theory and Application (in
Orbital Elements Determination for Leo Monitoring and Maintenance,” Chinese), Publishing House of Electronics Industry, Beijing, 2015,
International Symposium on Space Flight Dynamics, Johns Hopkins
pp. 117–258.
Univ. Applied Physics Lab., Laurel, MD, May 2014.