You are on page 1of 17

Available online at www.sciencedirect.

com

ScienceDirect
Advances in Space Research 62 (2018) 3281–3297
www.elsevier.com/locate/asr

Response surface modeling-based analysis on launch vehicle capability


Sungjun Ann a, Namhoon Cho a, Youdan Kim a,⇑, Sangjoon Shin a, Jaemyung Ahn b
a
Department of Mechanical and Aerospace Engineering and Institute of Advanced Aerospace Technology, Seoul National University, Seoul 08826,
Republic of Korea
b
Department of Aerospace Engineering, Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea

Received 28 May 2017; received in revised form 28 September 2017; accepted 5 March 2018
Available online 13 March 2018

Abstract

This paper proposes a framework to analyze the capacity of a launch vehicle systematically. Important mission requirements and
flight parameters that influence the payload capacity of a given launch vehicle configuration are identified. The response surface models
(RSM; the ‘‘analyzer”) for the launch capacity, coasting/burning arcs, and the velocity increment elements are developed. A series of
trajectory optimization results, which are obtained with the ‘‘optimizer”, are used to determine the coefficients of the models. The case
study to analyze the capacity of a low Earth orbit (LEO) launch vehicle configuration is conducted to demonstrate the validity of the
proposed framework.
Ó 2018 COSPAR. Published by Elsevier Ltd. All rights reserved.

Keywords: Trajectory optimization; Multi-stage launch vehicle; Launch vehicle capability; Small satellite; DV analysis

1. Introduction The adjustable part in the commercial launch vehicle is


related to its trajectory. Many studies have been performed
Recently, the applicability of small satellites (e.g. Cube- to design the trajectory of multi-stage launch vehicle. Most
Sat) for space missions has been rapidly increasing based of the studies have focused on optimizing the trajectory of
on its cost-effectiveness. There are many missions utilizing the launch vehicle to deliver maximum payload to a single
small satellites, and the target orbits for a single launch target orbit considering various point and path constraints.
configuration can be various as well (McIntyre et al., The optimal control histories and design variables can be
2016). Considering complexity and economic feasibility, it obtained through these analyses. The techniques used for
is not suitable to design a new launch vehicle for each mis- trajectory optimization of launch vehicles can be catego-
sion to deliver a satellite to a target orbit. Therefore, a rized as the indirect method and the direct method (Betts,
launch vehicle system with a specific configuration should 1998).
be used for different missions involving various target The indirect method can check the optimality of the
orbits. The analysis on the capability of the launch vehicle solution using the Hamiltonian. If the system dynamics is
is useful to select a proper launch vehicle for delivering a highly nonlinear or the constraints are complicated, the ini-
small satellite to a target orbit. tial guess of co-state variable profiles is very difficult. This
results in the usually very small region of convergence,
which is the disadvantage of the method. There are many
⇑ Corresponding author.
studies on launch vehicle trajectory optimization using
E-mail addresses: sungjun9037@snu.ac.kr (S. Ann), nhcho91@snu.ac.
the indirect method considering various constraints such
kr (N. Cho), ydkim@snu.ac.kr (Y. Kim), ssjoon@snu.ac.kr (S. Shin), as normal force (Gath and Calise, 2001), free coasting
jaemyung.ahn@kaist.ac.kr (J. Ahn). (Lu et al., 2008), and bending moment (Lu and Pan,

https://doi.org/10.1016/j.asr.2018.03.007
0273-1177/Ó 2018 COSPAR. Published by Elsevier Ltd. All rights reserved.
3282 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

2010). Recently, Pontani and Teofilatto (2014) and Pontani Mukhopadhyay, 2010). It is easy to practically use the
(2014) proposed a particle swarm optimization (PSO) RSM. For example, it can provide the sensitivity of the
based procedure to overcome the difficulty in the initial response with respect to the change in input parameters
guess of co-states and the small region of convergence by simple differentiation. The RSM has been adopted in
inherent in the indirect method. On the other hand, the various fields such as analytic chemistry (Bezerra et al.,
direct method can handle problem with complicated 2008; Gupta et al., 2017), biochemistry (Basß and Boyacı,
dynamics/constraints relatively easily. One of potential 2007), and combustion/fluid dynamics (Shirvan et al.,
issues in the direct method is its large number of design 2017). There are also many studies using the RSM in var-
variables, which are usually found using the nonlinear pro- ious disciplines of aerospace engineering such as system
gramming. Weigel and Well (2000) optimized the ascent design (Nguyen et al., 2013; Sobieski and Kroo, 2000;
trajectory for a dual payload launch considering the splash- Knill et al., 1999), experimental aerodynamics (Landman
down constraint using a direct multiple-shooting method. et al., 2007; DeLoach and Erickson, 2003), and guidance
Maddock et al. (2017) used the direct multi-shooting tran- and control (Ahn and Seo, 2013).
scription method to optimize the trajectory of a two-stage The purpose of this study is to develop a framework to
launch vehicle with complicated inequality constraints analyze the payload capacity of a launch vehicle for various
including the bounds on states of each stage. Roh (2001), target orbits. The proposed framework consists of two
Roh and Kim (2002) compared the performance of indirect modules  the optimizer and the analyzer as shown in
and direct methods for trajectory optimization of a multi- Fig. 1. The optimizer module conducts trajectory optimiza-
stage launch vehicle. In addition, they proposed a tech- tion that maximizes the payload mass considering various
nique to obtain the initial guess of co-state trajectory using constraints such as target orbit parameters and launch con-
the direct method, which enhances the convergence charac- ditions, which are provided as inputs to the module along
teristics of the indirect method. with launch vehicle configuration. The analyzer module
While information on the capacity of launch vehicles is constructs the response surface models for the launch
available from the launch service providers (ULA, 2013; capacity and velocity increment/loss elements using the tra-
Orbital ATK, 2015; Orbital, 1999; Charania et al., 2016), jectories obtained in the optimizer module.
the maximum payload that can be delivered for a specific The contribution of this paper can be summarized as
mission should be obtained by optimizing the vehicle’s follows. First, a framework to model/analyze the capacity
ascending trajectory for the mission. This paper introduces of a given launch vehicle configuration for various target
a response surface method (RSM) based framework to orbits using the response surface modeling (RSM) is devel-
model the maximum payload capacity of a launch vehicle oped. The framework can be practically used by launch ser-
with its mission elements (e.g. orbital altitude and inclina- vice providers and their users during the conceptual design
tion) as input parameters. phase of the launcher system or space mission. Secondly,
The RSM has been widely used since early 1950s to find realistic case studies demonstrating the effectiveness of
appropriate functional relationship between input parame- the proposed framework have been conducted. It is shown
ters and a response of interest using regression (Khuri and that the framework can provide the velocity increment/loss

Fig. 1. Response surface modeling based launch capability analysis: schematic diagram.
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3283

models and the sensitivity of the capacity of a launch vehi-


cle with respect to the changes in mission parameters,
which can be used to establish the launch vehicle’s strategy
to reach the desired orbit.
This paper is organized as follows. In Section 2, equa-
tions of motion are provided considering the characteristic
of the launch vehicle. In Section 3, the framework is pro-
posed. The optimizer and the analyzer are described in Sec-
tions 3.1 and 3.2, respectively. The specifications of the
launch vehicle are defined, and case studies are performed
in Section 4. Finally, conclusions are made in Section 5.

2. Dynamic modeling of launch vehicle

A launch vehicle delivering a satellite from the Earth’s


surface to the orbit experiences rapid changes in mass, Fig. 2. Coordinate systems.
position and velocity. Therefore, to design a trajectory of
the launch vehicle, the rotation of the Earth, the aerody-
namic forces, and gravitational force should be considered. Table 1
In this study, the launch vehicle is considered as a point Definition of the coordinate systems.
mass, and the Earth is modeled as a rotating sphere. The Coordinate Systems Index Definition
mass, position, and velocity of the launch vehicle are Earth-centered inertial (ECI) I zI : the rotation axis of the
defined as state variables, and the angles determining the Earth
direction of thrust are selected as control variables. The Earth-centered, Earth-fixed E T EI ¼ Rotz ðxe DtÞ
(ECEF)
state variables related to the position of the launch vehicle Local horizontal L T LE ¼ Roty ðkÞRotzðKÞ
are distance from the center of the Earth r, longitude K, Velocity-carried axis V T VL ¼ Roty  p2 þ c Rotx ðvÞ
and latitude k. The relative speed of the launch vehicle with Body-fixed B T VB ¼ Rotz ðbÞRoty ðaÞ
respect to the rotating Earth V, flight path angle c, and
flight azimuth angle v are state variables related to the dm T vac
velocity of the launch vehicle. To consider the orientation ¼ ð1Þ
dt g0 I sp
of the launch vehicle, it is assumed that the direction of
thrust is equal to the axial direction of the launch vehicle. dr
¼ V sin c ð2Þ
The angle of attack a and the angle of sideslip b are control dt
variables. In Fig. 2, the coordinate systems considered in dK V cos c sin v
this study are shown, where O represents the center of ¼ ð3Þ
dt r cos k
the Earth, and the center of mass of the launch vehicle is
dk V cos c cos v
marked with pcm . ¼ ð4Þ
The definitions of the coordinate systems are summa- dt r
rized in Table 1 (Markl, 2001), where T BA represents a coor- dV T cos a cos b  D le sin c
¼ 
dinate transformation from coordinate system A to dt m r2
coordinate system B, and Roti represents a rotation matrix þ rxe cos kðsin c cos k  cos c sin k cos vÞ
2
ð5Þ
with respect to axis i. The basis vectors of the coordinate  
dc T sin a þ L V l
system A are xA ; y A , and zA . In the ECI (Earth-centered ¼ þ  2 e cos c
inertial) and ECEF (Earth-centered, Earth-fixed) coordi- dt mV r rV
nate systems, xI and xE are both pointing toward the rx cos k
2
þ e ðcos c cos k þ sin c sin k cos vÞ
Greenwich in equatorial plane. Note that xI points toward V
the Greenwich at the initial time t0 , and is fixed in space. þ 2xe cos k sin v ð6Þ
On the other hand, xE is pointing toward the Greenwich
only at the specific time ðt0 þ DtÞ. In the local horizontal dv T cos a sin b þ Y V
¼ þ cos c tan k sin v
coordinate, the direction of xL is defined as radially out- dt mV cos c r
2
ward direction from the Earth, and zL is pointing toward rxe
þ sin k cos k sin v
the North Pole. xV and xB are aligned with the velocity V cos c
and the axial direction of the launch vehicle, respectively. þ 2xe ðsin k  tan c cos k cos vÞ ð7Þ
Using the relation of state and control variables, the
equations of the motion can be expressed as follows where T vac is a thrust at vacuum, g0 is a gravitational accel-
(Roh, 2001). eration of the Earth at sea level, and I sp is a specific impulse
3284 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

at vacuum. The terms containing angular velocity of the the target orbit. For launch conditions, initial heading
Earth, xe , represent the effects of rotation of the Earth, angle of the launch vehicle is considered, which can be
and the terms containing the standard gravitational param- imposed as a launch direction constraint.
eter of the Earth, le , represent the effects of the gravita- Trajectory optimization is performed to obtain a refer-
tional force. Thrust T, and aerodynamic forces, drag D, ence trajectory of the launch vehicle, which delivers the
lift L, and sideforce Y, can be represented as follows. satellite into the desired orbit. The histories of control vari-
T ¼ T vac  Ae p ð8Þ ables (aðtÞ and bðtÞ) and the values of design parameters x
¼ [DT c1 ; DT c2 , . . ., DT cn ; DT f ] are obtained through the
D ¼ F N sin a þ F A cos a ð9Þ optimization, where DT ci is duration of the ith coast arc,
L ¼ F N cos a  F A sin a ð10Þ and DT f is duration of final burn. It is assumed that the
1 coast arc occurs between two powered stage outside the
Y ¼ qV 2 SC N ;b ð11Þ Earth’s atmosphere. For the burn arc of the final stage,
2
maximum capacity of propellant at the final stage is consid-
with ered. The objective of the optimization problem is to max-
1 1 imize the payload mass, which can be expressed as
F A ¼ qV 2 SC A ; F N ¼ qV 2 SC N ;a ð12Þ  
2 2 J  ¼ max m tf ð13Þ
aðtÞ;bðtÞ;x
where Ae is a cross-sectional area of nozzle exhaust exit, S
is a reference surface area, C A is an axial force coefficient, Note that the duration of the burn is closely related to the
and C N is a normal force coefficient. The pressure and den- amount of propellant at the corresponding stage. Consider-
sity of the air are modeled as pðhÞ ¼ p0 eh=h0 , and ing the maximum capacity of the propellant in the final
qðhÞ ¼ q0 eh=h0 , respectively, where h is an altitude. stage, the inequality constraint on the burn time of the final
stage is imposed as follows
3. Capability analysis
05DT f 5DT f ;max ð14Þ
In this section, a framework to analyze the capability of To reduce the effects of the aerodynamic loads on the struc-
a launch vehicle is proposed. The analysis procedure con- ture of the launch vehicle, gravity turn is usually adopted
sists of two parts - the optimizer (explained in Section 3.1) during the first stage. Therefore, input constraints are con-
and the analyzer (explained in Section 3.2). sidered as follows
The optimizer generates a family of optimal trajectories
for a single launch vehicle configuration with various mis- aðtÞ ¼ bðtÞ ¼ 0; t 2 during the first stage
 
ð15Þ
sion parameters. To analyze the capacity of a launch vehi- 30 5aðtÞ530 ; 30 5bðtÞ530 ; 8t
cle, multiple scenarios are properly selected. The scenario,
For the circular orbit, the shape of the orbit can be deter-
which is an input to the optimizer, is composed of orbital
mined by the inertial speed V i and altitude h, i.e.,
parameters and launch conditions. The optimizer determi-
nes the optimal trajectory/control and associated environ- h ¼ r  Re ð16Þ
mental variables. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
The outputs of the optimizer, a family of optimized tra- V i ¼ V 2 þ 2Vrxe cos c sin v cos k þ ðrxe cos kÞ ð17Þ
jectory variables associated with input scenarios, are used
by the analyzer for modeling the payload capacity and per- where Re is the radius of the Earth. The orientation of the
formance parameters as response surface models using orbit only depends on the inclination i, which can be calcu-
curve fitting (in Section 3.2.1). The analysis on the created lated as
models is conducted and its results are presented in Sec- cos kðV cos c sin v þ rxe cos kÞ
tion 3.2.2. During the ascending phase, a launch vehicle cos i ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
is affected by various factors. The effects of these factors ðV cos cÞ þ 2Vrxe cos c sin v cos k þ ðrxe cos kÞ
on the launch vehicle performance are studied using DV ð18Þ
analysis. Finally, the strategy to maximize payload of
launch vehicle with respect to orbit conditions are Also, considering the launch site of the vehicle, the initial
explained using the analysis method. conditions are chosen as follows.
mðt0 Þ ¼ free; hðt0 Þ ¼ 800 m;
3.1. Optimizer
Kðt0 Þ ¼ 127:32 ; kðt0 Þ ¼ 34:25
ð19Þ
For orbital condition, the LEO is selected as a target V ðt0 Þ ¼ 143:25 m=s; cðt0 Þ ¼ free;
orbit in this study. Circular orbits with inclination between vðt0 Þ ¼ free; 180
0 and 140 (Vallado and McClain, 2013) are considered as
target orbit, which are widely used for the LEO satellites. Note that the initial position is fixed, and the initial velocity
Various altitudes are also considered as the parameter of is partially given. The initial heading angle of the launch
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3285

vehicle is set free for the case of vðt0 Þ ¼ free. The final con- 3.2. Analyzer
ditions are chosen as follows.
3.2.1. Response surface models
mðtf Þ ¼ free; Kðtf Þ ¼ free; cðtf Þ ¼ 0
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The optimizer provides the optimized output parame-
l
hðtf Þ ¼ hLEO ; V i ðtf Þ ¼  e ; iðtf Þ ¼ iLEO ters, i.e., mass of the payload mpayload , duration of the coast
h t f þ Re arcs DT c , and duration of the final burn DT f for given
ð20Þ launch vehicle configuration and orbit parameters (altitude
h and inclination i). In this study, the response surface
The altitudes of the orbits considered are from 160 km
models representing these output parameters are created
to 2,000 km with 20 km intervals, i.e., hLEO ¼
based on the optimization results using the polynomial fit-
½160 180    1980 2000km. The orbit insertion speed is
ting method. The models can provide approximate optimal
determined as the circular orbit speed for the given
values for a case not considered in the optimization, with-
altitude.
out actually performing the trajectory optimization.
The latitude of launch site affects the possible range of
The models are assumed to be polynomial functions of
target orbit’s inclination as follows (Larson et al., 1995).
input parameters (h and i) expressed as follows:
iLEO;T ¼ fi 2 Rjkðt0 Þ 5 i 5 180  kðt0 Þg ð21Þ X
nh X
ni

where iLEO;T is set of possible inclination, and iLEO  iLEO;T . ^y ¼ f ðh; iÞ ¼ prs hr is ð22Þ
r¼0 s¼0
In this study, the inclinations of the orbits considered are
from 40 to 140 with 10 intervals, i.e., iLEO ¼ where f is a curve fitting function of each data, ^y is the
½40 50    130 140 . Fig. 3 shows particular trajectories value of the curve fitting function, nh is degree of polyno-
for the three cases. Despite the inefficiency associated with mial with respect to h, and ni is degree of polynomial with
the satellite’s rotational direction, orbits with high inclina- respect to i. The coefficients of polynomial psr can be
tion (i > 100 ) are used for military purposes sometimes. expressed as a vector form p ¼ [p00 ; p10 ; p01 ; . . . ; pnh 0 ; . . . ;
The constraint on the initial heading angle is associated pnh ni ]. The parameter vector p can be determined using
with the flight safety; it should be imposed considering the the least squares regression. In this study, a nonlinear least
flight-restricted zone around the launch site. While the con- squares method is adopted, which minimizes the sum of
straint is given as the range of initial heading angle, it can squares of error (SSE) as
be treated as a fixed initial value if the allowable range is X X 2
extremely limited. To analyze the influence of initial head- SSE ¼ ½y data ðh; iÞ  ^y  ð23Þ
h2hLEO i2iLEO
ing condition on launch vehicle’s capability, a fixed initial

heading case, vðt0 Þ ¼ 180 , is considered in this study. J ¼ minSSE ð24Þ
p
The trajectory optimization problems associated with
target altitude and inclination values specified in the input To check the goodness-of-fit, R-square (R2 ) and root mean
scenarios are solved using the collocation method with squared error (RMSE) are examined.
sequential quadratic programming (SQP) (Hargraves and SSE
Paris, 1987). R2 ¼ 1  ð25Þ
SST
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SSE
RMSE ¼ ð26Þ
jhLEO jjiLEO j  jpj

where
X X 2
SST ¼ ½y data ðh; iÞ  y  ð27Þ
h2hLEO i2iLEO

1 X X
y ¼ y data ðh; iÞ ð28Þ
jhLEO jjiLEO j h2hLEO i2i
LEO

and jj is the cardinality of the set. Using the values of


SSE; R2 , and RMSE, minimum values of nh and ni are
selected as degree of each polynomial. For the goodness-
of-fit, it can be stated that curve fitting is well performed
when the value of R2 is close to 1 and the value of RMSE
is close to 0, respectively.

3.2.2. DV analysis
Fig. 3. Target orbits with three inclinations and trajectories of the launch The main factors affecting the launch vehicle’s trajectory
vehicle. are thrust, gravity, drag, steering, and rotation of the
3286 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

Earth. By analyzing these factors, the launch vehicle’s 4. Case study


strategy to maximize its payload can be obtained.
If the orbit insertion condition changes, the effects of This section presents the procedure and results of a case
these factors vary as well, which results in the change in study that can demonstrate the effectiveness of the pro-
the maximum payload obtained by the optimization. posed framework. The specifications of the launch vehicle
Therefore, proper concept for the analysis is needed. It is are described in Section 4.1. The results of capability anal-
difficult to analyze the force itself, because it is a three- ysis are presented according to the launch condition in Sec-
dimensional vector. Alternatively, the energy, which is a tions 4.2 and 4.3. In both sections, results of curve fitting
scalar variable containing the effect of the force, can be are presented, and the analysis based on DV is performed.
considered. However, it is hard to compare the results For the curve fitting, the degree of polynomial is limited to
using energy itself because the energy is directly affected 15nh ; ni 53. As criteria of the goodness-of-fit, R2 =0:99 and
by the mass. Therefore, in this study, DV analysis is RMSE510 (for mpayload : kg; DT : second) are selected. In the
adopted to investigate the effects of the factors (Larson curve fitting function, unit of h and i are km and degree ( ),
et al., 1995). respectively.
It starts with the equations of motion related to the rel-
ative speed V in Eq. (5). In the physical point of view, Eq. 4.1. Launch vehicle specifications
(5) is related to the acceleration. Therefore, the forces and
rotation of the Earth affecting the launch vehicle are The launch configuration of Taurus (Orbital, 1999), a
reflected in Eq. (5), where the effect of the mass is normal- four-stage launch vehicle for LEO missions, is modified
ized. The orbit conditions can be also considered, because and used for case study. Important launch events and asso-
the speed depends on the altitude of the target orbit. ciated event times are illustrated/summarized in Fig. 4 and
Let us integrate Eq. (5) to check the change of the Table 2. It is assumed that the coast arcs occur at the ends
launch vehicle’s speed. of the burn, and the number of the coast arcs are specified
Z tf Z tf as two in this study. The coast arcs begin after the burn
T cos a cos b D le sin c
Vdt ¼   dt arcs of the stage 2 and the stage 3. The separation of fairing
t0 t0 m m r2 occurs at t4 , the middle of the burn arc of the stage 3.
Z tf
Values characterizing the propulsion system of the launch
þ rx2e cos kðsin c cos k  cos c sin k cos vÞdt
t0 vehicle used in the case study are presented in Table 3.
ð29Þ Dry mass, Dmdry , of each stage is also presented, which is
separated from the launch vehicle at the ends of the each
The change is decomposed into five components, i.e., total burn. Note that Dt1 is a burn time of the stage 1, Dt2 is a
propulsion impulse DV T , gravity loss DV G , drag loss DV D , burn time of the stage 2, Dt4 is a burn time of the stage 3
steering loss DV S , and the Earth rotation inertial gain DV R . before the separation of the fairing, and Dt5 is a burn time
Z tf of the stage 3 after the separation of the fairing, where
T
DV T ¼ dt ð30Þ Dti ¼ ti  ti1 with i ¼ 1; . . . ; 7.
t0 m The duration of coast arc DT c1 after the burn arc of the
Z tf
le sin c stage 2, the duration of coast arc DT c2 after the burn arc of
DV G ¼ dt ð31Þ the stage 3, and the burn time DT f of the final stage are
t0 r2
Z tf considered as the optimization parameters. These arc dura-
D tions can be expressed using Dti values as follows:
DV D ¼ dt ð32Þ
t0 m
Z tf DT c1 ¼ Dt3 ; DT c2 ¼ Dt6 ; DT f ¼ Dt7 ð36Þ
T
DV S ¼ ð1  cos a cos bÞdt ð33Þ For the limit on propellant of the final stage, DT f ;max is con-
t0 m
    sidered as 68.4s.
DV R ¼ V i tf  V tf þ V ðt0 Þ The aerodynamic coefficients in Eq. (12) are modeled as
Z tf follows.
 rx2e cos kðsin c cos k  cos c sin k cos vÞdt ð34Þ
t0 C A ¼ ftnðMachÞ; C N ;a ¼ ftnðaÞ; C N ;b ¼ ftnðbÞ ð37Þ

Then, Eq. (29) can be expressed as follows. where Mach is the Mach number. At each stage, the value
of S is given as constant. For the atmospheric data,
V Orbit ¼ DV T  DV S  DV G  DV D þ DV R ð35Þ
h0 ¼ 7600 m; p0 ¼ 101; 325 Pa, and q0 ¼ 1:225 kg=m3 are
where the inertial speed needed for the orbit is used.
V Orbit ¼ V i ðtf Þ. Note that positive values of DV T and DV R Total 92 nodes are used for the collocation method
are related to the increase of the speed. On the other hand, which is one of direct optimization methods, and the num-
positive values of DV G ; DV D , and DV S are related to the ber of nodes for each time interval are summarized in
decrease of the speed. Table 4.
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3287

Fig. 4. Time line of the launch vehicle.

Table 2 4.2.1. Response surface


Launch event sequences. The output parameter models for free initial heading
Time Events subcase are presented in Fig. 5. The performance index,
t0 End of the kick-turn, Start of the stage 1 burn, Start of the gravity- the mass of maximum achievable payload, is shown in
turn Fig. 5(a). The maximum achievable payload decreases as
t1 End of the stage 1 burn, Start of the stage 2 burn, Separation of the altitude of orbit h increases. The lower the inclination
the stage 1 of desired orbit i is, the more maximum payload can be
t2 End of the stage 2 burn, Start of the coast arc 1
achieved. DT c1 is zero in all cases as shown in Fig. 5(b).
t3 End of the coast arc 1, Start of the stage 3 burn, Separation of the
stage 2 On the other hand, DT c2 increases as h increases and i
t4 Separation of the fairing decreases as shown in Fig. 5(c). As shown in Fig. 5(d),
t5 End of the stage 3 burn, Start of the coast arc 2 DT f maintains the maximum value for h < hc , where hc is
t6 End of the coast arc 2, Start of the stage 4 burn, Separation of the a critical altitude of the orbit. However, DT f decreases as
stage 3
h increases and i decreases for h=hc . The gradient of max-
t7 End of the stage 4 burn, Start of orbital motion, Separation of the
stage 4 imum achievable payload and DT c2 are changed for h=hc
as well. The characteristics of the optimal results are signif-
icantly changed near hc . Therefore, the curve fitting of the
optimal results is separately performed considering the
4.2. Free initial heading angle condition between h and hc .
To obtain hc , maximum h satisfying DT f ¼ DT f ;max is
In Section 4.2.1, response surfaces for free initial head- selected as a data set for the given i. Using the data set con-
ing angle are obtained, and the accuracy of the response sisting of h and i; hc can be approximated as a function of i
surfaces is validated. The analysis of the optimization via polynomial regression. The results are plotted in Fig. 6,
results is performed in Section 4.2.2. For free initial head- and the function is given as follows.
ing angle, the relation between orbit parameters and max-
imum achievable payload are investigated, and the strategy hc ¼ f c ðiÞ ¼ 1296 þ 2:691i ð38Þ
of the launch vehicle is explained. Furthermore, the critical Considering the condition of h and hc , the maximum
altitude which exist on free initial heading angle case is ana- achievable payload mpayload ; DT c1 ; DT c2 , and DT f are
lyzed in Section 4.2.3. approximated as polynomial functions of h and i.

8 5 2
>
< 1778  0:6896h  5:294i þ 9:772  10 h þ 0:001235hi h5hc
mpayload ¼ ð39Þ
>
:
1458  0:2879h  5:79i þ 0:0005625hi þ 0:0007352i 2
h > hc
DT c1 ¼ 0 8h ð40Þ
8 7 3 7 2
>
< 0:8148 þ 1:361h  0:2725i  0:0007145h  0:0009831hi þ 1:839  10 h þ 1:546  10 h i
2
h5hc
DT c2 ¼
>
:
 1651 þ 3:018h þ 2:014i  0:00166h2 þ 0:01219hi  0:1255i2 þ 5:922  107 h3  8:395  106 h2 i þ 2:617  105 hi2 þ 0:000428i3 h > hc
ð41Þ
8
< DT f ;max
> h5hc
DT f ¼ ð42Þ
>
: 5 2
379:9  0:3473h  0:6701i þ 8:186  10 h  0:0002177hi h > hc
3288 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

Table 3
Specifications of the launch vehicle.
Stage 1 Stage 2 Stage 3 Fairing Stage 4
T vac (kN) 1686 485 118 – 32
I sp (s) 284.439 294.4 292.3 – 288.9
Ae ðm2 Þ 1.77 1.6 1.266 – 1.266
Burn time (s) 69 72.4 73.3 – DT f
Dmdry (kg) 4850 1860 345 – 350
mfairing (kg) – – – 500 –

Table 4
The number of nodes for each time interval.
Dt1 Dt2 Dt3 Dt4 Dt5 Dt6 Dt7 Fig. 6. Plot of critical altitude versus inclination (free initial heading
Number of nodes 16 16 6 6 16 16 16 angle).

The accuracy of the curve fitting results is summarized


in Table 5. The degree of polynomials (nh ; ni ) are selected
such that the number of coefficients are minimized while
Table 5
satisfying the criteria for R2 and RMSE. By checking the Accuracy of curve fitting (free initial heading angle).
values of R2 , it can be stated that the optimal results fit well nh ni SSE R2 RMSE
to each polynomial response surface - except for the case if
mpayload (h5hc ) 2 1 6:8098  10 4
0.9982 9.4411
DT c2 for h > hc , whose RMSE is relatively high. It means mpayload (h > hc ) 1 2 5:3034  103 0.9987 4.6150
that there exist some outliers with large variance, which DT c2 (h5hc ) 3 1 1:8179  104 0.9995 4.8843
should be considered when prediction is required. DT c2 (h > hc ) 3 3 6:9675  104 0.9961 16.8983
To validate the accuracy of the response surface, a num- DT f (h > hc ) 2 1 205.7107 0.9944 0.9089
ber of trajectory optimizations with free initial heading

Fig. 5. Response surfaces for payload mass and coast/burn arc durations (free initial heading angle).
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3289

Table 6 Now, the results of DT c1 ; DT c2 and DT f are analyzed


Validation of the response surface (free initial heading angle). considering physical meaning. Firstly, the launch vehicle
mpayload ðkgÞ DT c1 ðsÞ DT c2 ðsÞ DT f ðsÞ can deliver more payload by utilizing the coast arc DT c2
h ¼ 315 km; i ¼ 60 
1276.2 0 328.7 68.4 rather than the coast arc DT c1 . It is obvious because DT c1
(0.55%) (0%) (2.32%) (0%) is zero for the all cases as shown in Fig. 5(b). During the
h ¼ 455 km; i ¼ 120 916.6 0 405.3 68.4 coast arc, total energy always decreases because of the
(0.86%) (0%) (0.63%) (0%) gravitational force. Therefore, it is reasonable for the
h ¼ 600 km; i ¼ 98 953.2 0 519.2 68.4 launch vehicle to reduce the mass during the coast arc.
(1.50%) (0%) (0.59%) (0%) Secondly, the launch vehicle has different strategy to
h ¼ 1200 km; i ¼ 98 717.6 0 800.8 68.4 deliver more payload according to hc , which can be seen
(1.27%) (0%) (0.10%) (0%) in Fig. 5(a), (c), and (d). For h5hc , the launch vehicle
can deliver more payload by using maximum amount of
h ¼ 1800 km; i ¼ 98 542.2 0 1188.8 47.3
(0.94%) (0%) (0.11%) (1.83%) propellant in the final stage, DT f ;max . On the other hand,
for h > hc , the launch vehicle can deliver more payload
by reducing the amount of the propellant in the final stage,
angle are conducted with inputs at ‘‘confirmation points.” DT f , and using longer coast arc, DT c2 . As mentioned
The optimization results are compared with the output of
above, strategy for h5hc is reasonable because the amount
the response surfaces, whose difference indicates the accu-
of propellant cannot exceed a limit. However, strategy for
racy of the proposed framework. Table 6 presents the per-
h > hc is quite not in accord with intuition. To analyze the
centage difference between the RSM and optimization
results, DV T ; DV G ; DT c2 , and DT f are examined. If the
results defined as ((the output-the optimal value)/(the opti-
amount of propellant decreases, the amount of available
mal value))100 at the confirmation points. The worst-case
thrust and mass of the launch vehicle decrease. For
gap between the results is 2.32%, and it can be stated that
h > hc , the results show that decrease of mass is more effec-
the response surface model fits pretty well with the actual
tive to deliver more payload than increase of duration of
optimization results.
available thrust. Decrease of mass enables the launch vehi-
cle to utilize longer coast arc. As a result, for h > hc , the
4.2.2. Analysis of the results
gradient of DV G with respect to h becomes small, and con-
To analyze the results of the optimization, each value of
sequently, the gradient of DV T with respect to h becomes
DV is obtained as shown in Fig. 7(a)–(e), and
smaller as well.
DV T ; DV S ; DV G , DV D , and DV R are plotted.
The trajectory changes for h > hc can be analyzed by
Most of the speed is generated by the thrust, and most
DV D . Most of loss due to drag occurs during the stage 1.
of the speed is lost by the gravity. The loss of speed by
Therefore, large DV D implies the launch vehicle’s slow
the drag varies with h, and DV R is highly affected by i. In
passing through the atmosphere. Also, the altitude at the
case of free initial heading angle, steering loss is small as
end of the stage 1 is plotted as shown in Fig. 8. For
expected.
h > hc , the launch vehicle prepares to meet the orbit condi-
A couple of observations on the optimized payload mass
tion in advance, because the available thrust at the final
are presented as follows.
stage is short. Therefore, the altitude of the launch vehicle
First of all, if the direction of the rotation of target orbit
is more slowly changed, and as a result DV D increases.
is much different from that of the Earth, then less payload
can be delivered to the target orbit, which can be seen in
Fig. 5(a). As i increases, the maximum achievable payload 4.2.3. Analysis on critical altitude
decreases. To study this issue, DV T and DV R are examined We found close relationship between the existence of
with respect to i. As i increases, DV R decreases and DV T critical altitude, hc , and the maximum burn time (inter-
increases. The speed generated by the rotation of the Earth preted as the propellant storage capacity) of the final stage
decreases, and even the speed is lost for high i. As a result, DT f ;max .
more thrust is utilized to overcome the rotation of the The maximum amount of propellant in the final stage
Earth. can be treated as a constraint on maximum burn time
Secondly, the higher the altitude of the target orbit is, DT f ;max in Eq. (14), because the mass flow rate in Eq. (1)
the less payload can be delivered to the target orbit, which is constant for each stage. To analyze the effect of
can be seen in Fig. 5(a). As h increases, the maximum DT f ;max on hc , additional simulations for the cases
achievable payload decreases. To study this issue, DT f ;max ¼ 40s and DT f ;max ¼ 120s are performed. The
DV T ; DV G , and V Orbit are investigated. As h increases, all response surfaces of DT f for these cases are presented in
of DV T ; DV G , and V Orbit increase, where V orbit =
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi Fig. 9(a) and the associated plots of hc versus inclination
le =ðh þ Re Þ. This implies, as h increases, loss of energy angle are presented in Fig. 9(b). The decrease in hc is
due to the gravitational force increases, and the speed observed as the value of DT f ;max increases. In addition,
needed to maintain orbit increases. Consequently, more the optimal value of DT f increases as altitude h decreases,
thrust is needed to cope with the effects. while DT f not exceeding DT f ;max . The critical altitude, hc ,
3290 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

Fig. 7. Response surfaces for velocity increments/losses (free initial heading angle).

occurs when the optimal value of DT f coincides with the of Section 4.2 are quite different from that of Section 4.3
value of DT f ;max . Furthermore, the critical altitude and inclination which
exist on fixed initial heading angle case are analyzed in
Section 4.3.3.
4.3. Fixed initial heading angle

In Section 4.3.1, response surfaces for fixed initial head- 4.3.1. Response surface
ing angle are obtained, and the accuracy of the response The output parameter models for fixed initial heading
surfaces is validated. The analysis of the optimization subcase are presented in Fig. 10. The performance index,
results is performed in Section 4.3.2. For fixed initial head- the mass of maximum achievable payload, is shown in
ing angle, the relation between orbit parameters and max- Fig. 10(a). The tendency of each optimal result is signifi-
imum achievable payload are investigated, and the strategy cantly different from that of free initial heading angle.
of the launch vehicle is explained. Due to constraint on The maximum achievable payload is highly affected by i.
heading angle of the launch site, the relation and strategy The critical altitudes exist for both of DT c1 and DT f .
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3291

The gradient of DT c2 starts to be changed rapidly at hc1 and


hc2 .
To obtain hc1 , minimum h satisfying DT c1 ¼ 0 is selected
as a data set for the given i. Using the data set consisting of
h and i; hc1 can be approximated as a function of i via poly-
nomial regression. The results are plotted in Fig. 11(a), and
the function is given as follows.
hc1 ¼ f c1 ðiÞ ¼ 8647  263:5i þ 2:239i2  0:005293i3 ð43Þ

Likewise, to obtain hc2 , maximum h satisfying


DT f ¼ DT f ;max is selected as a data set for the given i. Using
the data set consisting of h and i; hc2 can be approximated
as a function of i via polynomial regression. The results are
plotted in Fig. 11(b), and the function is given as follows.
Fig. 8. Response surface for Stage 1 burn-out altitude (free initial heading
angle). hc2 ¼ f c2 ðiÞ ¼ 7184  134:8i þ 0:8i2 ð44Þ

DT c1 has non-zero value for h < hc1 as shown in Fig. 10(b), Considering the condition of h with hc1 and hc2 , the max-
where hc1 is a critical altitude of DT c1 . For DT f , the shape imum achievable payload mpayload ; DT c1 ; DT c2 , and DT f are
of a critical altitude hc2 is different as shown in Fig. 10(d). approximated as polynomial functions of h and i.

8
>
> 3299 þ 3:4h þ 125i  0:001366h2  0:0753hi  1:01i2 þ 4:538  107 h3 þ 1:819  106 h2 i þ 0:0004025hi2 þ 0:002099i3 h < hc1
>
>
>
>
<
mpayload ¼ 1472 þ 0:1333h þ 74:99i þ 0:0001272h2  0:01701hi  0:5857i2  7:965  107 h2 i þ 0:0001104hi2 þ 0:0009926i3 hc1 5h5hc2
>
>
>
>
>
>
:
382:6  0:2317h þ 34:2i  5:229  105 hi  0:21i2 h > hc2
ð45Þ
8 7 3 6 2
>
< 1770 þ 1:604h  51:29i  0:0004903h  0:0514hi þ 0:4472i þ 2:531  10 h  1:581  10 h i þ 0:0002979hi  0:001129i h < hc1
2 2 2 3

DT c1 ¼
>
:
0 h=hc1
ð46Þ
8 6 3 6 2
>
> 1684  10:54h  33:3i þ 0:004501h þ 0:255hi þ 0:0312i  1:77  10 h  1:264  10 h i  0:001407hi þ 0:001014i
2 2 2 3
h < hc1
>
>
>
>
<
DT c2 ¼ 233 þ 1:068h þ 5:975i  0:000498h2 þ 0:001044hi  0:03603h2 þ 1:04  107 h3 þ 1:153  107 h2 i  7:754  106 hi2 hc1 5h5hc2
>
>
>
>
>
>
:
2:461  104  17:96h  559:7i  0:000141h2 þ 0:4412hi þ 2:825i2 þ 9:229  107 h3  4:984  105 h2 i  0:001576hi2  0:00333i3 h > hc2
ð47Þ

Fig. 9. Characteristics of hc for various values of DT f ;max .


3292 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

Fig. 10. Response surfaces for payload mass and coast/burn arc durations (fixed initial heading angle).

Fig. 11. Plot of critical altitude versus inclination (fixed initial heading angle).

8
>
< DT f ;max h5hc2 heading angle. Especially, DT c1 (h < hc1 ), DT c2 (h < hc1 ),
DT f ¼ and DT c2 (h > hc2 ) are not fitted well. Thus, when predic-
>
:
701:7  0:09127h  12:11i þ 0:0001754hi þ 0:07044i2 h > hc2 tion is required, this effect should be considered.
ð48Þ Trajectory optimizations at confirmation points for
fixed initial heading angle are also conducted to validate
The accuracy of the curve fitting results is summarized the accuracy of the proposed framework. Table 8 compares
in Table 7. The degree of polynomials (nh ; ni ) are selected the results from the response surface model and the opti-
to minimize the number of coefficients of polynomials mization at the confirmation points. The worst-case gap
while satisfying the criteria for R2 and RMSE. By checking between the results is 2.84%, and it can be stated that the
the values of R2 and RMSE, it can be observed that the response surface model fits pretty well with the actual opti-
overall goodness-of-fit is lower than that of free initial mization results.
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3293

Table 7 of fixed initial heading angle, steering loss is significant and


Accuracy of curve fitting (fixed initial heading angle). highly affected by i. DV R is highly affected by i, similar to
nh ni SSE R2 RMSE that of free initial heading angle. Especially, the relation
mpayload ðh < hc1 Þ 3 3 7:9658  10 3
0.9975 7.6252 of DV T and i is significantly different from that of free ini-
mpayload ðhc1 5h5hc2 Þ 2 3 9:1685  104 0.9983 10.7935 tial heading angle case. The loss of speed by the drag is also
mpayload ðh > hc2 Þ 1 2 348.2624 0.9979 2.1549 affected by i.
DT c1 ðh < hc1 Þ 3 3 5:9983  104 0.9627 20.8244
DT c2 ðh < hc1 Þ 3 3 3:0113  105 0.9625 46.8828
We can make a couple of observations in the result of
DT c2 ðhc1 5h5hc2 Þ 3 2 4:0380  104 0.9990 7.1630 fixed initial heading subcase.
DT c2 ðh > hc2 Þ 3 3 2:1599  104 0.9898 17.5658 First of all, if the initial heading of the launch vehicle is
DT f ðh > hc2 Þ 1 2 83.0350 0.9902 1.0522 similar to the direction of the rotation of the orbit, then
more payload can be delivered to the target orbit, which
can be seen in Fig. 10(a). Let us define ic as a critical incli-
@m 
Table 8 nation of the orbit, which satisfies payload @i  ¼ 0. As i
Validation of the response surface (fixed initial heading angle). i¼ic

mpayload ðkgÞ DT c1 ðsÞ DT c2 ðsÞ DT f ðsÞ approaches to ic , the maximum achievable payload
increases. To analyze this issue, DV T and DV S are exam-
h ¼ 315 km; i ¼ 60 986.9 0 297.7 68.4
(0.22%) (0%) (2.84%) (0%) ined. As i moves away from ic , steering loss increases,
and consequently, greater thrust is needed to cope with
h ¼ 455 km; i ¼ 120 669.7 0 366.9 68.4
(0.64%) (0%) (2.37%) (0%)
the effects. The effects of the condition on initial heading
can be seen in Fig. 12, which is the trajectories of the
h ¼ 600 km; i ¼ 98 919.8 0 511.3 68.4
launch vehicle for h ¼ 340 km with vðt0 Þ ¼ 180 . For
(0.51%) (0%) (0.54%) (0%)
h ¼ 340 km; ic is calculated as 80:75 from Eq. (45). As
h ¼ 1200 km; i ¼ 98 668.9 0 800.4 68.4 shown in Fig. 12, if i of target orbit is closed to ic , then
(0.41%) (0%) (0.06%) (0%)
the optimal solution does not tend to change the heading
h ¼ 1800 km; i ¼ 98 525.9 0 1111.9 58.1 angle.
(0.57%) (0%) (2.12%) (0.15%)
Secondly, due to the constraint on heading angle condi-
tion of launch site, the launch vehicle has different strategy
The validation in this paper has been performed with to deliver more payload according to hc1 and hc2 , which can
confirmation points in reasonable operational range of be seen in Fig. 10(a)–(d). Especially, if vðt0 Þ is different
the launch vehicle. We observed that the response surface from the direction of the rotation of orbit and h < hc1 ,
models for the coasting time at some points that do not the launch vehicle can deliver more payload by utilizing
comply with normal launch operation have relatively large the coast arc, DT c1 . It is interesting to note that the use
error, which demands for future work on defining the of coast arc, DT c1 , for the launch vehicle is an effective
bound in which the response surfaces are valid. means to maximize the payload as shown in Fig. 10(b).
To analyze this phenomenon, DV T and DV S are examined.
4.3.2. Analysis of the results As i moves away from ic ; DV T and DV S have large values.
To analyze the results of the optimization for fixed ini- In other words, it is hard to change the heading angle of
tial heading angle, each value of DV is obtained as shown the launch vehicle. Moreover, the altitude of the launch
in Fig. 13(a)–(e), and DV T ; DV S ; DV G , DV D , and DV R are vehicle begins to exceed the altitude of the target orbit as
plotted, respectively. shown in Fig. 14(a). Meanwhile, the maximum duration
Again, most of the speed is generated by the thrust, and of the thrust is utilized. The results show deficiency of the
most of the speed is lost by the gravity. Moreover, in case adjustable variables to reach the target orbit when initial

Fig. 12. Launch vehicle trajectories for h ¼ 340 km [vðt0 Þ ¼ 180 ].


3294 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

Fig. 13. Response surfaces for velocity increments/losses (fixed initial heading angle).

heading is fixed. Therefore, although nonzero DT c1 is not are performed, and the results are shown in Fig. 15. Both
suitable to maximize the payload, DT c1 is utilized to reach of hc and hc2 decrease as the value of DT f ;max increases.
the target orbit. The Fig. 14(b) shows the time histories of The difference of hc and hc2 is the shape of the critical alti-
the altitude with and without overshoot of the altitude. tude with respect to inclination.
The existence of the critical inclination ic and altitude hc1
4.3.3. Analysis on critical altitudes and inclination is affected by the constraint on launch azimuth, which is
For fixed initial heading case, the critical altitudes and considered as fixed value of vðt0 Þ. To analyze the effects
inclination are closely related to the launch vehicle’s speci- of vðt0 Þ on the critical values, additional simulations for
fication and condition of the launch site. the cases vðt0 Þ ¼ 150 and vðt0 Þ ¼ 210 are performed.
Similar to hc in Section 4.2.3, the existence of critical The results of the simulations are shown in Fig. 16, and
altitude, hc2 , is closely related to the maximum capacity summarized in Table 9.
of the propellant in the final stage DT f ;max . The additional The relation between ic and vðt0 Þ can be analyzed by the
simulations for the cases DT f ;max ¼ 40 s and DT f ;max ¼ 120 s simulation results. For h ¼ 600 km, the maximum achiev-
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3295

Fig. 14. (a) Response surface for maximum overshoot altitude (fixed initial heading angle), (b) altitude history for various scenario.

Fig. 15. Characteristics of hc2 for various values of DT f ;max .

Fig. 16. Critical inclination and critical altitude with respect to fixed vðt0 Þ.

able payload versus vðt0 Þ is plotted in Fig. 16(a), and each


Table 9
ic for h ¼ 600 km and i (hc1 > 160 km) with respect to fixed vðt0 Þ. corresponding ic is summarized in Table 9. Critical inclina-
tion ic increases as vðt0 Þ increases, and furthermore ic is
ic for h ¼ 600 km i (hc1 > 160 km)
 
approximately equal to vðt0 Þ þ 90 . Let us consider the tar-
vðt0 Þ ¼ 150 58:65 100 5i5140
get orbit that directly passes through area above the launch
vðt0 Þ ¼ 180 79:79 40 5i550 ; 120 5i5140
vðt0 Þ ¼ 210 109:65 40 5i570 ; 140 5i site. As shown in Fig. 17, the projection of the orbit’s tra-
jectory is represented as the direction of the rotation of the
3296 S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297

the operational strategy of a launch vehicle to maximize


the payload weight and to understand the performance sen-
sitivity with respect to mission parameters.
The proposed framework can be practically used for
various groups (e.g. developers, service providers, and
users). Firstly, for launch service providers and customers,
the framework provides the capability of a launch vehicle
for various launch condition (e.g. target orbits, launch
operations) without additional trajectory optimization.
The payload capacity can be easily obtained if the launch
condition is given within the range modeled by the
response surface, which helps/accelerates the mission
design procedure. Secondly, the framework can be utilized
by the developers of a launch vehicle during its conceptual
design phase. The response surfaces of different vehicle
configurations associated with various orbital parameters
can be generated, which can be used for screening and/or
Fig. 17. Local horizontal plane with the orbit which directly passes down-selection of the configurations.
through area above the launch site. The discussion on the selection of design points for
response surface modeling for efficient RSM construction
orbit, which is perpendicular to the rotation axis of the and in-depth analysis on the accuracy are left as subjects
orbit. If the initial heading of the launch vehicle and the of future studies.
direction of the rotation of the orbit are aligned
(i ¼ vðt0 Þ þ 90 ), the orbit can be easily reached by the Acknowledgements
launch vehicle. By the definition, the ic represents the incli-
nation of the target orbit, which can be reached with max- This work was supported by Advanced Research Center
imum payload. That is, ic is the inclination of the easiest Program (NRF-2013R1A5A1073861) through the
reachable orbit for given initial heading (vðt0 Þ). Note that National Research Foundation of Korea (NRF) grant
ic is slightly smaller than vðt0 Þ þ 90 due to the effects of funded by the Korea government (MSIP) contracted
the rotation of the Earth. through Advanced Space Propulsion Research Center at
The effect of vðt0 Þ on hc1 can be analyzed by the simula- Seoul National University.
tion results. For the case vðt0 Þ ¼ 150 ; vðt0 Þ ¼ 180 , and
vðt0 Þ ¼ 210 , each hc1 is shown in Fig. 16(b), and the range
References
of i with hc1 > 160 km is summarized in Table 9. As i
moves away from the ic ; hc1 increases. From the relation Ahn, J., Seo, J., 2013. Instantaneous impact point prediction using the
between ic and vðt0 Þ, it is inferred that hc1 is affected by response-surface method. J. Guid. Control Dyn. 36 (4), 958–966.
vðt0 Þ. _
Basß, D., Boyacı, I.H., 2007. Modeling and optimization I: usability of
response surface methodology. J. Food Eng. 78 (3), 836–845.
Betts, J.T., 1998. Survey of numerical methods for trajectory optimization.
5. Conclusions
J. Guid. Control Dyn. 21 (2), 193–207.
Bezerra, M.A., Santelli, R.E., Oliveira, E.P., Villar, L.S., Escaleira, L.A.,
A framework to analyze the capability of a launch vehi- 2008. Response surface methodology (RSM) as a tool for optimization
cle was proposed to deliver satellites with various missions. in analytical chemistry. Talanta 76 (5), 965–977.
The framework consists of the optimizer and the analyzer. Charania, A., Isakowitz, S., Matsumori, B., Pomerantz, W., Vaughn, M.,
Kubiak, H., Caponio, D., 2016. Launcherone: Virgin galactic’s
The optimizer conducts the trajectory optimization to max-
dedicated launch vehicle for small satellites. In: Small Satellite
imize the payload under various constraints on orbital Conference 2016, Logan, UT.
parameters and launch conditions. Altitude and inclination DeLoach, R., Erickson, G., 2003. Low-order response surface modeling of
of the orbit are considered as orbital parameters, and initial wind tunnel data over truncated inference subspaces. In: Proceedings
heading constraints are considered for the launch condi- of the 41st Aerospace Sciences Meeting & Exhibit, Reno, NV.
Gath, P.F., Calise, A.J., 2001. Optimization of launch vehicle ascent
tion. The analyzer constructs response surfaces for payload
trajectories with path constraints and coast arcs. J. Guid. Control Dyn.
capacity and DV elements using the results from the opti- 24 (2), 296–304.
mizer and the method of least squares fitting. To verify Gupta, V.K., Agarwal, S., Asif, M., Fakhri, A., Sadeghi, N., 2017.
the proposed framework, a case study was conducted using Application of response surface methodology to optimize the adsorp-
a four-stage launch vehicle for low Earth orbit missions. tion performance of a magnetic graphene oxide nanocomposite
adsorbent for removal of methadone from the environment. J. Colloid
The maximum achievable payload, duration of coast arcs,
Interface Sci. 497, 193–200.
and burn time of the final stage were fitted as polynomial Hargraves, C.R., Paris, S.W., 1987. Direct trajectory optimization using
functions with respect to altitude and inclination of orbit. nonlinear programming and collocation. J. Guid. Control Dyn. 10 (4),
The outputs of the framework can be used to determine 338–342.
S. Ann et al. / Advances in Space Research 62 (2018) 3281–3297 3297

Khuri, A.I., Mukhopadhyay, S., 2010. Response surface methodology. Orbital, 1999. Taurus user’s guide release 3.0. Orbital, Dulles, VA,
Wiley Interdiscip. Rev.: Comput. Stat. 2 (2), 128–149. Retrieved from <http://www.georing.biz/usefull/Taurus_User_Guide.
Knill, D.L., Giunta, A.A., Baker, C.A., Grossman, B., Mason, W.H., pdf>.
Haftka, R.T., Watson, L.T., 1999. Response surface models combining Orbital ATK, 2015. Pegasus user’s guide Release 8.0. Orbital ATK,
linear and euler aerodynamics for supersonic transport design. J. Dulles, VA, Retrieved from <https://www.orbitalatk.com/flight-sys-
Aircraft 36 (1), 75–86. tems/space-launch-vehicles/pegasus/docs/Pegasus_UsersGuide.pdf>.
Landman, D., Simpson, J., Vicroy, D., Parker, P., 2007. Response surface Pontani, M., 2014. Particle swarm optimization of ascent trajectories of
methods for efficient complex aircraft configuration aerodynamic multistage launch vehicles. Acta Astronaut. 94 (2), 852–864.
characterization. J. Aircraft 44 (4), 1189–1195. Pontani, M., Teofilatto, P., 2014. Simple method for performance
Larson, W.J., Henry, G.N., Humble, R.W., 1995. Space Propulsion evaluation of multistage rockets. Acta Astronaut. 94 (1), 434–445.
Analysis and Design. McGraw-Hill, NewYork, NY, pp. 55–58, 64–69. Roh, W., 2001. A study on the trajectory optimization and inertial
Lu, P., Griffin, B.J., Dukeman, G.A., Chavez, F.R., 2008. Rapid optimal guidance algorithm design for satellite launch vehicles (Ph.D. thesis).
multiburn ascent planning and guidance. J. Guid. Control Dyn. 31 (6), Department of Mechanical and Aerospace Engineering, Seoul
1656–1664. National University. Seoul, Republic of Korea.
Lu, P., Pan, B., 2010. Highly constrained optimal launch ascent guidance. Roh, W., Kim, Y., 2002. Trajectory optimization for a multi-stage launch
J. Guid. Control Dyn. 33 (2), 404–414. vehicle using time finite element and direct collocation methods. Eng.
Maddock, C.A., Toso, F., Ricciardi, L., Mogavero, A., Lo, K.H., Optim. 34 (1), 15–32.
Rengarajan, S., Kontis, K., Milne, A., Merrifield, J., Evans, D., et al., Shirvan, K.M., Mamourian, M., Mirzakhanlari, S., Ellahi, R., 2017.
2017. Vehicle and mission design of a future small payload launcher. Numerical investigation of heat exchanger effectiveness in a double
In: 21st AIAA International Space Planes and Hypersonics Technolo- pipe heat exchanger filled with nanofluid: a sensitivity analysis by
gies Conference, Xiamen, China. response surface methodology. Powder Technol. 313, 99–111.
Markl, A.W., 2001. An initial guess generator for launch and reentry Sobieski, I.P., Kroo, I.M., 2000. Collaborative optimization using
vehicle trajectory optimization (Ph.D. thesis). Institut für Flug- response surface estimation. AIAA J. 38 (10), 1931–1938.
mechanik und Flugregelung der Universität Stuttgart. Stuttgart, ULA, 2013. Delta IV launch services user’s guide. ULA, Centennial, CO,
Germany. Retrieved from <http://www.ulalaunch.com/uploads/docs/Launch_
McIntyre, S., Fawcett, T., Dickinson, T., Maddock, C.A., Mogavero, A., Vehicles/Delta_IV_Users_Guide_June_2013.pdf>.
Ricciardi, L., Toso, F., West, M., Kontis, K., Lo, K.H., et al., 2016. Vallado, D.A., McClain, W.D., 2013. Fundamentals of Astrodynamics
How to launch small payloads? evaluation of current and future small and Applications, fourth ed. Microcosm Press, Hawthorne, CA, pp.
payload launch systems. In: 14th Reinventing Space Conference, 845–852.
London, United Kingdom. Weigel, N., Well, K.H., 2000. Dual payload ascent trajectory optimization
Nguyen, N.V., Choi, S., Kim, W., Lee, J., Kim, S., Neufeld, D., Byun, Y., with a splash-down constraint. J. Guid. Control Dyn. 23 (1), 45–52.
2013. Multidisciplinary unmanned combat air vehicle system design
using multi-fidelity model. Aerosp. Sci. Technol. 26 (1), 200–210.

You might also like