Professional Documents
Culture Documents
Optimal Conceptual Design of Two-Stage Reusable Rocket Vehicles Including Trajectory Optimization
Optimal Conceptual Design of Two-Stage Reusable Rocket Vehicles Including Trajectory Optimization
Many candidate concepts of reusable space transportation vehicles have been proposed around the world. This
paper applies an optimization method for conceptual designs of winged fully reusable rocket two-stage-to-orbit
(TSTO) vehicles in both horizontal takeoff and vertical launch styles. We first describe our methods for analyzing
vehicle design. Then we integrate those methods into the optimization problem, the solution of which yields the
minimized parameter, defined as the total gross weight of the first-stage booster and the second-stage orbiter. This
information allows us to determine the optimal vehicle configuration and flight trajectory for a highly feasible
Downloaded by Stanford University on October 7, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.1082
TSTO vehicle. The optimal solutions show the necessity of lightening and miniaturizing components. Vertical
launch vehicles are lighter in total gross weight than horizontal takeoff vehicles. In addition to optimizing vehicle
configuration, this study also optimizes ascent and return trajectories. These optimizations enable the booster to
glide back to the launch site without propellant, despite the long downrange path from the staging point of the
ascent trajectory.
are applied to each vehicle. These methods use the surface panels weight to compute the gear weight of the horizontal takeoff booster
from the vehicle definition. The methods, however, cannot estimate and, in the case of the vertical launch booster and the orbiter, it is
parasite drag. Especially, base drag accounts for a large part of the the maximum landing weight.
total drag because a vehicle with rocket engines has a cutoff aft 2) The orbiter has a reaction control system and orbital maneu-
fuselage. According to Ref. 16, we estimate the base drag of each vering engines (RCS/OME) and their propellants. The system has
vehicle from its base area and the Mach number. In addition, we thrusters with a specific impulse of 310 s and a velocity increment
compute the skin-friction drag by the sum of the frictions from the capability of 340 m/s in orbit. The propellant is supposed to weigh
flow velocities on the panels. Transonic aerodynamic forces result 70% of the total RCS/OME weight.
from subsonic and supersonic forces, as explained in Ref. 16. 3) This study estimates the weight of the thermal protection sys-
The simple methods in this study are applicable to each vehicle tem (TPS) more precisely than the HASA method. We presume that
separately and not to the coupled vehicle. We presume that the cou- the nose edge and wing leading edges of the orbiter are provided
pled vehicle’s lift force is equal to that of the booster and that its drag with C/C materials (area density 40 kg/m2 ), and, on the other sur-
force is the sum of the total drag of the booster and the parasite drag faces of the orbiter and the booster, thermal protection materials
of the orbiter. Thus, this study does not consider the interference equivalent to 30-mm-thick ceramic tiles are bonded.
drag caused by the coupling of the two vehicles. 4) The propellant tanks are made of aluminum alloy, and each is
The methods used in this study are simpler and lower fidelity than composed of a cylinder whose ends are dome shaped. The working
the latest sophisticated computation methods, but they are useful in pressure of the tank is given by the sum of the ullage pressure
that the aerodynamics analysis, by which the optimization process (3 kgf/cm2 ) plus the head pressure. We compute the tank weight
computes the aerodynamic characteristics repeatedly, needs only from the tank wall thickness, which is obtained considering the
a short computation time. These methods are applied widely to working pressure with the membrane theory. For manufacturability,
aerodynamic preliminary analyses of candidate space transportation however, the thinnest wall is determined to be 3 mm thick in this
vehicles.6,17 paper. In addition, the weight of the isogrid structure that is generally
patterned on the inner surface of the tank, plus the weight of spot
Propulsion Analysis welds, sensor attachments, and so on, require the preceding tank
Given a flight altitude by the trajectory analysis, the propul- weight to be multiplied by a constant number 1.4 in order to avoid
sion analysis calculates a thrust force and specific impulse in the underestimation.
following manner. The rocket engine uses the cryogenic propel- 5) An equation to obtain rocket engine weights is quoted from
lant (LOX/LH2). This study considers a conventional rocket engine Ref. 19. The following equation holds for the thrust force Tinf (N)
given by historical perspective and not a specific engine. in vacuum and the engine weight Weng (kg):
A thrust force Tinf (N) per engine in vacuum is one of the design
variables. According to existing or developed rocket engines, we Weng = 0.0119Tinf
0.856
(6)
presume that a cross-section area of the rocket nozzle exit Ae (m2 )
of the booster and orbiter is proportional to Tinf , as expressed here: A thrust-to-weight ratio of a rocket engine with a maximum thrust
force of 1.5 MN amounts to approximately 65 using the preceding
Ae = 2.771 × 10−6 Tinf (1) equation.
The estimation of the component weights needs the gross weight,
This means that each rocket engine has a fixed nozzle. The atmo- which is the sum total of the component weights. For example, the
spheric pressure p (Pa) changes the thrust force to TRE (N) in the landing gear weight, as part of the gross weight, is expressed as
air: an equation with a variable of the gross weight. Thus, the method
analyzes the weights using the gross weight in the design variables to
obtain the gross weight. We define an equality constraint condition
TRE = K Tinf (2)
that both the weights correspond exactly.
K = 1 − Ae ( p/Tinf ) (3) Knowledge of the layout of the components in the vehicle makes
it possible to compute the center of gravity. The flight situations
define the constraint conditions by which the c.g. locations remain
Simultaneously, a specific impulse ISP (s) in the atmosphere is pro- within the limits that the aerodynamics analysis determined. There
vided by the specific impulse ISPinf (s) in vacuum (fixed to be 445 s is also an equality constraint condition that the c.g. location of the
in this paper) and the preceding parameter K : booster does not move when the orbiter is released from the booster.
This implies that the only c.g. location determines the attachment
ISP = K ISPinf (4) location of the orbiter in this paper.
TSUCHIYA AND MORI 773
Trajectory Analysis station to monitor rockets launched from Japan. We also have a plan
This study implements a three-degree-of-freedom (3DOF) tra- to develop infrastructures for a future spaceport.
jectory analysis.20 State variables are altitude, longitude, latitude, This study considers two takeoff styles: the horizontal takeoff and
velocity, flight-path angle, heading angle, and mass. Control vari- the vertical launch. In the case of vertical launch, an initial condition
ables are angle of attack, bank angle, and thrust throttle. Generally, is that the flight-path angle is 90 deg. After the takeoff, having no
the trajectory analysis computes state variable profiles by integrating bank angle, the coupled vehicle rises and accelerates east by the
equations of motion with given control variable profiles and exam- power of the booster’s rocket engines; this paper does not employ
ines satisfaction ratings of constraint conditions during the flight. parallel burn by the orbiter or cross-feeding of the propellants.
One of the characteristics in this study is to optimize the flight tra- The horizontal takeoff vehicle has constraint conditions involving
jectory with the other design variables, as already described. The the angle of attack, throttle, dynamic pressure, thrust, body normal
trajectory analysis, however, must deal with dynamic variables de- acceleration, and body axial acceleration. The values for these are
pending on time, while the other design variables are static ones limited to the maximum values in the design variables, except for the
independent of time. Therefore, we discretize the state and control angle of attack and the throttle, which are below 20 deg and above
variables to transform them into static variables like the other de- 60%, respectively. The vertical launch vehicle has the same set of
sign variables.21 In the same manner we also discretize constraint constraint conditions, plus the constraint condition that the flight-
conditions and the motion equations expressed as differential equa- path angle must decrease steadily from the initial 90 deg. This is
tions, which are functions of the state and control variables. The because a free flight-path angle results in the optimal solution, in
number of discrete elements is fixed at 200 in total. Nevertheless, which the vehicle descends with the flight-path angle negative im-
if we were to discretize all 10 state and control variables with the mediately after liftoff to enhance acceleration just above the ground
200 elements, the more than 2000 resultant design variables would and then gains altitude by the lift force from its wings. This is an
make for a very heavy computational burden in the optimization. unrealistic flight path with respect to safety, though it has the advan-
Downloaded by Stanford University on October 7, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.1082
Therefore, we classify the state variables into those of position and tage that the vehicle can accelerate even if powered by small rocket
lateral motion (longitude, latitude, and heading angle) and others engines. From this we constrain the flight-path angle of the vertical
(altitude, velocity, flight-path angle, and mass). The structures of launch vehicle to decrease monotonously.
the motion equations exhibit strong interaction among the latter and
a weak influence of variation of the former on the latter. It follows Orbiter Ascent Phase
that we discretize only the latter four of the seven state variables and In this phase the released orbiter accelerates by its rocket engines
the three control variables. When the values of the former three state and climbs to an altitude of 120 km. At that altitude the orbiter
variables are required, they are computed by numerical integration cuts off the rocket engines temporarily and zooms up with no thrust
of their motion equations. in an elliptical orbit whose apogee altitude is 200 km. Finally, an
The flight styles of the vehicles change the equations of motion. apogee kick puts it into a 200-km circular low Earth orbit (LEO).
This divides the flight trajectory into four phases, as described next. To estimate propellant consumption, we actually compute the flight
In addition, if each vehicle flies with multiple rocket engines the trajectory until the temporary engine cutoff and approximate the
engines are cut off one after another because the body axial ac- apogee kick by an impulse thrust.
celeration increases as the propellants are consumed. Because the At the beginning of this phase, the angle of attack is allowed to
equations of motion change discontinuously at the shutdown points, have values that are discontinuous with the terminal angle of attack
the phases must be subdivided into intervals, the number of which in the preceding phase. The bank angle remains 0 deg during the
corresponds to the number of engines. Initial and terminal times flight for the orbiter to fly east. The constraint conditions in this
in the intervals are contained in the optimized variables. We de- phase are the angle of attack, throttle, dynamic pressure, thrust,
fine equality constraint conditions to make the time and variables normal acceleration, and body axial acceleration. Their values are
equal between a series of intervals. The phases are explained as limited to the orbiter’s flight performance parameters in the design
follows. variables, except that the angle of attack is kept below 20 deg and
the throttle is kept in the 60–100% range.
Booster Ascent Phase
This phase indicates the flight from liftoff until the orbiter sep- Booster Fly-Back Phase
arates from the booster. We suppose that the launch site is Aeon After releasing the orbiter, the booster turns to fly back to the
airfield (longitude, 157.36◦ west and latitude, 1.78◦ north) on Christ- launch site, Aeon airfield. As the booster in this paper has no separate
mas Island of the Republic of Kiribati (Fig. 3). On the island, airbreathing engines for the return flight, the only rocket engines can
National Space Development Agency of Japan has a downrange propel the booster in this phase. In this paper they are not compared
with the vehicles with the airbreathing engines on the boosters. The
bank angle, which until now has been fixed at 0 deg, is allowed to be
free. The initial angle of attack must be equal to the terminal angle
in the booster ascent phase. As in the ascent phase, the angle of
attack, throttle, dynamic pressure, thrust, normal acceleration, and
body axial acceleration are constrained during the return flight. The
terminal constraint conditions are that the latitude and longitude of
the booster are identical with those of the launch site and the altitude
is zero.
of the wing, whereas the normal acceleration and dynamic pressure the vertical launch booster uses the gear for landing and not for both
limits are given by the design variables. In addition, the following takeoff and landing. For either booster the takeoff weight is about
constraint conditions are prepared: the lift-to-drag ratio of the orbiter five times the landing weight. This makes the landing gear of the
at subsonic speed must be greater than five, which is equivalent to horizontal takeoff booster six times heavier than that of the vertical
that of HOPE-X,23 and the orbiter can pull itself up at the maximum because of Eq. (5). The difference affects the total gross weight and
angle of attack during landing without violating the body normal the dry weight. This result shows the necessity of more accurate
acceleration limit. estimation of the landing gear weight. The difference in total gross
weight and dry weight between the vertical and horizontal takeoff
Optimization Problem vehicles demonstrates that the former seems more feasible than the
In addition to the design variables and constraint conditions just latter, as the definition of the minimized parameter of this study
described, the total gross weight of the booster and orbiter with a makes it clear that smaller vehicles are more likely to be realized.
payload weight is defined as a minimized parameter. We originally In the future, however, when the field of candidate space transporta-
minimized a dry weight of the vehicles, and the vehicles obtained tion systems is narrowed down, weight will not be the only criterion.
were very large. To restrict the vehicle size, we redefined the mini- Other criteria include cost, operability, reliability, and so on.
mized parameter to be the total gross weight including the propellant Figure 4 and Tables 2 and 3 show that, even if the orbiter has no
weight related to a vehicle volume. As a result, however, the sizes payload capability, the total gross weights of the orbiters and boost-
and weights of the optimal vehicles in both the case of the min- ers taking off vertically and horizontally are 126 and 146 Mg, re-
imized parameters differed by not more than 10%. Furthermore, spectively. An increase in the payload weight by 1 Mg increases the
qualitative trends of the solutions were same. For rocket vehicles total gross weight by 40–50 Mg. If fully reusable rocket TSTO sys-
with hydrogen–oxygen propellants, the dry weight and the gross tems are made to replace Japan’s expendable H-IIA rocket, which
weight are closely linked. This study obtains the optimal solutions has a payload capability of more than 10 Mg to LEO, the total
Downloaded by Stanford University on October 7, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.1082
for both the vertical and horizontal takeoff vehicles, in which three gross weights of the vertical launch vehicles and the horizontal
sets of payload weights and volumes are delivered to/from orbit: takeoff vehicles will probably reach 550 and 600 Mg, respectively,
0 Mg and 0 m3 , 4 Mg and 40 m3 , and 8 Mg and 80 m3 . whereas the H-IIA rocket weighs about 290 Mg. Compared with
A sequential-quadratic-programming method is employed as an
optimization method. There are approximately 1450 optimized de-
sign variables. The reason for the large number is that we have not
only the geometric and flight performance parameters, but also the
seven state and control variables discretized to the 200 elements, as
explained in the preceding subsections.
Optimal Solutions
Optimal Configurations
Figure 4 shows the variation in weight configurations of the hor-
izontal and vertical takeoff vehicles with the payload weights and
volumes. Tables 2 and 3 detail the component weights as well as
the size and performance data for the vehicles. The results demon-
strate that the vertical launch vehicles are 10–15% lighter than the
horizontal takeoff vehicles in total gross weight. The vertical launch
vehicles are about 20% lighter in the total dry weight as well. The
difference is attributed to the landing gear weights of the respec-
tive boosters. As described in the weight analysis, the landing gear
weight of the vertical launch booster is estimated on the basis of the
maximum landing weight, rather than the takeoff weight, because Fig. 4 Component weight comparison for the two takeoff styles.
work again, the downrange distances reach 400–500 km. The boost- Systems, AIAA, Reston, VA, 2001.
7 Sobieszczanski-Sobieski, J., and Haftka, R. T., “Multidisciplinary
ers start a turn with the bank angle that exceeds 70 deg. At a high
bank angle the boosters gain a short turning radius but lose altitude Aerospace Design Optimization: Survey of Recent Developments,” AIAA
and velocity. The optimization provides the best bank angle and Paper 96-0711, Jan. 1996.
8 Wolf, D. M., “TRANSYS-Space Transportation System Preliminary De-
angle of attack between the losses and gains. The boosters subtly sign Software,” Journal of Spacecraft and Rockets, Vol. 31, No. 6, 1994,
adjust the two angles in order to prevent the dynamic pressure and pp. 1067–1071.
body normal acceleration from exceeding the optimized limits. As 9 Rahn, M., Schottle, U. M., and Messerschmid, E., “Multidisciplinary
the ground projection of the flight path points to the launch site, Design Tool for System and Mission Optimization of Launch Vehicles,”
the boosters gradually lessen the bank angle to 0 deg, and the angle AIAA 96-4130, Sept. 1996.
778 TSUCHIYA AND MORI
10 Koelle, D. E., “Economics of Small Fully Reusable Launch Systems 18 Harloff, G. J., and Berkowitz, B. M., “HASA—Hypersonic Aerospace
(SSTO vs. TSTO),” Acta Astronautica, Vol. 40, No. 2–8, 1997, pp. 535–544. Sizing Analysis for Preliminary Design of Aerospace Vehicles,” NASA CR-
11 Miele, A., and Mancuso, S., “Design Feasibility via Ascent Optimality 182226, Nov. 1988.
for Next-Generation Spacecraft,” Acta Astronautica, Vol. 45, No. 12, 1999, 19 Koelle, D. E., TRANSCOST Statistical-Analytical Model for Cost Esti-
pp. 705–715. mation and Economic Optimization of Space Transportation Systems, TCS-
12 Mehta, U. B., and Bowles, J. V., “Two-Stage-to-Orbit Spaceplane Con- Trans Cost Systems, Ottobrunn, Germany, 1991.
cept with Growth Potential,” Journal of Propulsion and Power, Vol. 17, 20 Zipfel, P. H., Modeling and Simulation of Aerospace Vehicle Dynamics,
No. 6, 2001, pp. 1149–1161. AIAA, Reston, VA, 2000.
13 Tsuchiya, T., and Suzuki, S., “A Simultaneous Optimization Technique 21 Tsuchiya, T., and Suzuki, S., “Spaceplane Trajectory Optimization with
for Spaceplane Shape and Trajectory,” AIAA Paper 2001-1847, April 2001. Vehicle Size Analysis,” 14th IFAC Symposium on Automatic Control in
14 Tsujimoto, T., Kouchiyama, J., and Shirouzu, M., “Current Status of the Aerospace, International Federation of Automatic Control, Seoul, Repub-
H-II Orbiting Plane Experimental (HOPE-X) Development,” Proceedings lic of Korea, 1998, pp. 444–449.
of 22nd International Symposium on Space Technology and Science, Japan 22 Harpold, J. C., and Graves, C. A., Jr., “Shuttle Entry Guidance,” Journal
Society for Aeronautical and Space Science, Motioka, Japan, 2000, Paper of the Astronautical Sciences, Vol. 27, No. 3, 1979, pp. 239–268.
00-g-06. 23 NAL/NASDA HOPE Joint Research Team, “Aerodynamic Data
15 Morino, L., Chen, L., and Suciu, E., “Steady and Oscillatory Subsonic Book—HOPE-X FY09,” National Space Development Agency, NASDA
and Supersonic Aerodynamics around Complex Configurations,” AIAA Jour- TMR-000001, Tokyo, March 2000 (in Japanese).
nal, Vol. 13, No. 3, 1975, pp. 368–374.
16 Raymer, D. P., Aircraft Design: A Conceptual Approach, AIAA, Reston,
VA, 1999.
17 Sova, G., and Divan, P., “Aerodynamic Preliminary Analysis System II. J. Martin
Part II: User’s Manual,” NASA CR-182077, April 1991. Associate Editor
Downloaded by Stanford University on October 7, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.1082
This article has been cited by:
6. Mathieu Balesdent, Nicolas Bérend, Philippe Dépincé, Abdelhamid Chriette. 2010. Multidisciplinary Design
Optimization of Multi-Stage Launch Vehicle using Flight Phases Decomposition. International Journal for Simulation
and Multidisciplinary Design Optimization 4:3-4, 117-125. [CrossRef]
7. Yunjun Xu. 2009. Trajectory Analysis for Vertical Takeoff and Vertical Landing Reusable Launch Vehicle’s Upper
Stage. Journal of Aerospace Engineering 22:1, 58-66. [CrossRef]
8. Mateen-Ud-Din QaziLinshu HePermoon Mateen. 2007. Hammersley Sampling and Support-Vector-Regression-
Driven Launch Vehicle Design. Journal of Spacecraft and Rockets 44:5, 1094-1106. [Citation] [PDF] [PDF Plus]
9. Xiao Huang, ; Bernd Chudoba, Overview of a HTHL Hands-On SAV Design Synthesis Methodology . [Citation]
[PDF] [PDF Plus]
10. Vernon T. Bechel, Mathew Negilski, Joshua James. 2006. Limiting the permeability of composites for cryogenic
applications. Composites Science and Technology 66:13, 2284-2295. [CrossRef]
11. Mateen-ud-Din Qazi, He Linshu. 2006. Nearly-orthogonal sampling and neural network metamodel driven conceptual
design of multistage space launch vehicle. Computer-Aided Design 38:6, 595-607. [CrossRef]
12. Takeshi TSUCHIYA. 2006. Near-Optimal Guidance Method for Maximizing the Reachable Domain of Gliding
Aircraft. TRANSACTIONS OF THE JAPAN SOCIETY FOR AERONAUTICAL AND SPACE SCIENCES 49:165,
137-145. [CrossRef]
13. Takeshi TsuchiyaTakashige Mori. 2005. Optimal Design of Two-Stage-To-Orbit Space Planes with Airbreathing
Engines. Journal of Spacecraft and Rockets 42:1, 90-97. [Citation] [PDF] [PDF Plus]