See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261210309

A generalized staging optimization program for space launch vehicles

Conference Paper · June 2013

DOI: 10.1109/RAST.2013.6581333

CITATIONS READS

8 2,810

2 authors, including:

M. Kemal Özgören
Middle East Technical University
96 PUBLICATIONS 739 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ezgi Civek on 11 April 2018.

The user has requested enhancement of the downloaded file.

 A Generalized Staging Optimization Program For
Space Launch Vehicles

Ezgi Civek-Co�kun, MSc Prof. Dr. Kemal Ozg6ren

System Design Department Mechanical Engineering Department
Roketsan Missile Industries Inc. Middle East Technical University
Ankara, Turkey Ankara, Turkey
ecivek@roketsan.com.tr ozgoren@metu.edu.tr

Abstract-This paper addresses the staging optimization Another advantage of staging is that launch vehicle
problem for multistage rockets which carry payloads from the configuration can be optimized for the requirements of a
Earth's surface into the Earth orbits. In the early design phases, particular mission by adjusting the amount of propellant and
requirements are not so strict, there are many unknowns and engine thrust, and using different types of engines, propellants
problem arises as to what is the optimum staging to achieve the and structural materials for various stages. Stages can also be
given mission. Therefore, designers need simplified tools
designed for best performance considering their operating
providing a quick insight on the vehicle performance with
conditions.
minimum basic vehicle data. For this purpose, a Matlab® based
computer program has been written to determine staging The optimum stage mass distribution between stages for a
parameters (number of stages, mass distribution between stages, multistage launch vehicle and the propellant and structural
and the propellant and structural masses for each individual masses for each individual stage can be determined by staging
stage) which minimize the gross lift-off mass of the launch vehicle optimization for a given set of technology options. By optimal
for a specific mission. staging, launch vehicle can achieve its specified mission with
minimum gross lift-off mass which can be considered as a key
In this study, staging optimization problem has been
driver of both performance and cost.
formulated based on Delta-V equations and solved by method of
Lagrange Multipliers. The problem has been stated in a general Till today, many efforts have been made to optimize launch
form to handle launch vehicles having arbitrary number of stages vehicle staging for minimum gross lift-off mass [1-9]. In the
and with various configurations involving serial, parallel and early papers, problem has been solved with many simplifying
clustered stages; and with different structural ratios and assumptions. Malina and Summerfield [1] were the first to
propellant exhaust velocities in each stage. optimize staging; however, the solution was limited to stages
having equal propellant exhaust velocities and structural ratios.
Staging optimization program developed in this study has
Vertregt [2] extended the solution to the case when all stages
been verified for different missions using available data of
have different exhaust velocities. Goldsmith [3] offered a
existing launch vehicles. Thus, a quick and effective tool to find
solution for two stage rockets when the structural masses are
optimal vehicle configurations in the conceptual design phase of a
proportional to propellant masses. Weisbord [4], Subotowicz
generic multistage launch vehicle has been achieved.
[5], Hall and Zambelli [6] have all presented general solutions
Keywords-multistage rockets; space launch vehicles, staging
for minimum gross lift-off mass with non-homogenous stages
optimization; minimizing gross lift-off mass; conceptual design and the solution holds for arbitrary number of stages.
All of these authors used the method of Lagrange
I. INTRODUCTION Multipliers, which was proven many times, and they ignore the
Exploration and utilization of space for the benefit of effects of gravity, drag and steering in order to obtain the
mankind require space launch vehicles, which carry payloads derivatives analytically. Later on, Srivastava [7], Tawakley [8]
from the Earth's surface into the Earth orbit and beyond. Space and Adkins [9] have examined the isolated effects of gravity,
launch vehicles are multistage rockets composed of two or steering and drag, respectively.
more stages, each of which contains its own propellant and This paper presents a staging optimization method well
structure. The idea behind staging is to improve performance suited to be used in the conceptual design phase of a generic
by reducing the vehicle's mass on the way to orbit. Once the multistage launch vehicle having different structural ratios and
propellant of a stage is consumed, the empty stage which is no propellant exhaust velocities in each stage. This method allows
longer useful and only adds weight to the vehicle is discarded considering the gravitational, aerodynamic and propulsive loss
and the next stage is ignited. This stage then accelerates the rest factors with proper margins under the simplest possible
of the vehicle much faster. Thus, less propellant is required to conditions without determining the flight trajectory.
achieve the desired orbit.

978-1-4673-6396-9/13/$3l.00 ©2013 IEEE 857

 II. MATHEMATICAL MODEL

A. Tsiolkovsky's Rocket Equation

The famous fundamental rocket equation derived from
Newton's second law of motion governs the relationship
between the propellant consumption and the velocities that can
be attained by the rocket vehicle and is given by
mO,2
L'1 Vvehicle = C In A
. (1)

where:
L'1 Vvehicle : maximum change of speed
C : exhaust velocity
A : mass ratio mO,1
Mass ratio (A) is just the ratio of the initial mass to the final
mass
Fig, I, Mass definitions for serial staging,
_ !!!.::... _ mo m,+mp +mpl
A (2) In line with the single stage rocket, mass of the kth stage
mj mo-mp m,+mpl and the initial and fmal masses before and after the operation of
the kth stage are
where:
mo : initial mass (8)
ml : final mass
ms : structural mass (9)
mp : propellant mass
m I,k =ms,k + mO,k+1 (10)
mp/ : payload mass
Exhaust velocity (C) is defined by
1) Serial staging: Several stages stacked on top of each
(3) other and one stage burns alone until its propellant is
exhausted. It is then jettisoned and the next stage is ignited
where: (Fig. 1).
lsp : specific impulse
2
go : gravity at seal level, go = 9.81 mls
Mass ratio of the kth stage (Ak) is
B. Multistage Rocket Parameters
Ideal velocity increment for an N-stage rocket is the sum of (11)
the velocity increments of the individual stages.
N Structural ratio of the kth stage (ck) is
L'1v"ehlcle = L Ck ·In A k (4)
k=1 __ m_...:.. ,,k
__ ms,k
E:k = = (12)
In the analysis of N-stage rocket, payload of any particular
stage (k) can be considered as the mass of subsequent stages
(k+1, ... , N) as illustrated in Fig. l. Payload ratio of the kth stage (Ak) is

m pi,k =mO,k+l (5)

(13)
And the payload of the last (Nth) stage is the actual payload
of the launch vehicle.
A, C and A are not independent. If we express ms and mp/ in
m pl,N =m pI (6) terms of mp from (12) and (13), respectively and insert those in
(11), then we get the mass ratio for the kth stage (Ak) as
Total payload ratio (A,) is defmed as
k 1+ A
m mp
Ak = --- (14)
A = --.!!!... = __1 E:k + Ak
r (7)
2) Parallel staging: Several stages are mounted in parallel
where: and burn in simultaneously. Some launch vehicles use both
mo,! : gross lift-off mass (GLOM) of the launch vehicle serial and parallel staging (Fig. 2).

858
 /\
(21)
mO,N
m c -m p.1O
-�

C Orbital Velocity Equation

Velocity of the satellite at any point in an orbit can be
mO,1 mp,11 m l= mc - mp,IO calculated using the following vis viva equation derived from
orbital energy conservation equation,

Vorbit. = �GM(�-�) (22)

mp,IO mb mO= mb+ mp,IO r a

mo,o where:
VOrbll : orbital velocity at radial distance r
GM : Earth's gravitational parameter, GM 398600 Ian /S2
= 3

Fig, 2, Mass definitions for parallel staging, r : radial distance from the Earth's center to the satellite
a : semimajor axis
The analysis of a parallel staged rocket is quite similar to
the one presented above, the main difference is the stage D, Delta- V Calculations
numbering and the need for calculation of average exhaust Velocity change of an N-stage launch vehicle (L1Vvelticle) can
velocity, be calculated from rocket equation (4) knowing the fact that
When the parallel boosters and the core fIrst stage are this is the ideal velocity change, and gravitational and
burning simultaneously, they are taken together and called the aerodynamic forces, flight maneuvers and all other velocity
zeroth stage, while the propellant remaining in the core's fIrst losses/gains are neglected, Tewari [12] proposed to add a total
stage after discarding the parallel boosters is called the fust of 1,5 km/s margin for velocity losses/gains for a launch to the
stage of the rocket low earth orbit and 2 km/s for a launch to the geosynchronous
orbit
According to Fig, 2, mass, structural and payload ratios of
the zeroth stage equivalent to a serial rocket are given by Velocity losses/gains can also be included separately in the
calculation of the total velocity increment needed to get into
m o, o orbit (L1Vmission) '
A0 =
(15)
ms,b + mO,l

ms,b where:
=
[;
0 (16)
mb + mp,iO L1Vg : velocity loss due to gravity
L1V d : velocity loss due to aerodynamic drag
mO.l L1Vp : propulsive losses due to steering and pressure change
A0 =
(17) L1Vgain: velocity gain due to Earth's rotation or initial altitude
mb + mp,iO
or initial velocity
where: Gravity losses (llVg) and drag losses (llV d) are the most
c : subscript for core fust stage signifIcant loss terms, and they are primarily dependent upon
b : subscript for boosters the initial thrust-to-weight ratio (T/W), TIW needs to be greater
mp,IO : propellant mass burned in parallel with the boosters in than unity for the vehicle to leave the launch pad, and typical
zeroth operation lift-off T/W values are in the range 1,3 to 2 [10],
Average exhaust velocity of the zeroth stage is Approximated values obtained from real data samples can
be used for rough estimations of gravity and drag losses, In this
!sp,b ,mp,b + !sp,c ,mp,1O
Co =
study, IIVg and L1V d variations versus T/W presented by [11]
go (18)
have been used (Fig, 3), which is valid for vertical take-off
mp,b + mp,1O
vehicles, For horizontal take-off, thrust losses will be higher,
Similarly, the equivalent ratios for the fIrst stage are but gravity loss is much lower, For horizontal take-off highest
value of T/W can be used to approximate this behavior,
(19) Launch vehicles also experience propulsive losses due to
the maneuvering and static pressure difference at the nozzle
==
m exit during their flight These losses are smaller compared to
&' __ ,-,-s,c,--
IIVg and L1V d especially for vertical take-off vehicles and it is
1 _
(20)
m -m 1
c p, O diffIcult to estimate the magnitude without having the flight
trajectory,

859
 1000
1-- gravity loss
III. PROBLEM FORMULATION
900
The objective of staging optimization problem is to fmd the
"'-
drag loss

800 optimal mass ratios of stages (Ak) which will minimize the
700 "'- gross lift-off mass of the launch vehicle (GLOM, mo) for

�'" 600 '" specified payload mass (mpl).

'"'"
OJ 500 '" mo = mO, =
N

] I m, + mp1 (27)
.2
> 400 ""- k=!
<] .......
300 ....... Dividing (27) by mph one can write the objective function
�r-. (minimize mo) with respect to mass ratios of stages (Ak)' Curtis
200
-- ---
--
[lO] showed that the following relation is obtained by using the
100 ---- defmitions of Ak and Ck given in (11) and (12), respectively.
o � (1- c,) A,
IT
1 1.5 2 2.5 3 3.5 4
TNV !!i = (28)
mp1 k=] 1- ckA,
Fig. 3. �V losses vs T/W.

It is more convenient to take the natural logarithm of both

On the other side, rotation of the Earth fortunately assists in sides to obtain the following easily differentiable function.
launching. Launch vehicle has a latitude-dependent initial
velocity at the time of launch owing to the eastward rotational
velocity of the Earth. Component of velocity due to Earth's (29)
rotation can be calculated as

v " = Q. ro • cos rAo (24) Launch vehicle must provide the required energy to insert
'0,1"0
the satellite into desired orbit. Thus, �Vvehicle must be equal to
where: �Vmission'
v,'''o : Earth's speed at radius ro and latitude <Po (eastward) N

Q : Earth's angular velocity, Q 7.292 X 10.5 radls =

�Vveluell'
. =�
L..J Ck ·In Ak = �Vmission (30)
ro : radial distance from Earth's center, r o R + ho =
k=l
<Po : geographic latitude of the launch site Finally, the optimization problem can be formulated as
Considering the launch azimuth (A) and Earth's rotation, Minimize
one can find the velocity required to obtain target orbit
velocity.
(objective function,f) (31)
North

Subject to

] cos i
( ) N

A = sin- I c, . In A, = �v,,,;,,;o,, (constraint function, g) (32)

cosrPo k=]
i: orbit inclination
IV. METHOD OF SOLUTION
�:---+'----------- East If the Method of Lagrange Multipliers is applied, optimal
mass ratios (Ak) can be found by specifying the values for
Fig. 4. Velocity gain due to Earth's rotation. �Vmission, ck, Ck and N. Introducing the Lagrange multiplier (P),
and combining (31) and (32), one can have / f+ pg =

According to Fig. 4, speed gained due to Earth's rotation

can be calculated as
(33)

Expanding the logarithm

Launching from a carrier aircraft with an initial velocity N

(Vi) also helps decreasing the total required �V. Therefore, j' = I [ ln (l-ck) + In A, -In (l-ckAk)]
�Vgam. in (23) is k=l
( :t
N
(34)
�vgam =�v1'01 + VI (26) +p C, ·In A, - �v,,,;,,;o,,
)

860
 Differentiating with respect to Ak and equating it to zero Knowing the following facts;
• Ak> 1
(35) • 0 < Ck < 1
• CkAk < 1 in order to satisfy (39)
• Ck>O
One can fmd Ak as

p , Ck
And from (36)
1+
Ak =

p,Ck 'ck
(36)
(40)
For / to be minimum at the mass ratios Ak given by (36),
a2/ > 0 must hold for all values of Ak• Taking the second From (40)
differential of f
(41)
(37)
Rearranging
(32) can be rewritten as
(42)
(38)
(42) must hold for every k. Hence, the upper bound is
Burghes [13] showed that optimum number of stages is
p< ----
-1
--
min [Ck (1 - ck )]
between 2 and 4 for most satellite launching operations. Based (43)
on this fact, the minimum number of stages that is practical
should be chosen and solved fust, and then significant
differences should be compared after evaluating different After the calculation of Ak, Ak can be obtained from (14) as
values of N.
C
Structural ratios (ck) and exhaust velocities ( k) of stages (44)
strongly depend on operational aspects and the state of the art
of propellant and materials technology. Typical values of care With the known values of Ak each stage mass can be
in the range 0.1 < c < 0.2. Practical values of Isp are in the range calculated from (13) by the following recursive equation
250 sec < lsp < 475 sec, and the corresponding values of C are
in the range 2500 mls < C < 4750 mls. Therefore, the values
beginning with Nth stage down till the fust stage.

for N, Ck and Ck are initially specified by the designer mpl

considering available technology options, and �Vmission is mN = -
A-
calculated using the method given in Section II-D. N
(38) is a transcendental equation with single unknown (P),
and can be solved by iterative methods. Having evaluated for p (45)
a given set of Ck and Ck, optimal mass ratios of stages (Ak) can
be found by substituting p into (36).
In this study, Newton-Raphson method, which is a widely
mpl +mN +mN_1 +"'+m2
m1
used method for solving transcendental equations, has been =

used to solve (38). The Newton-Raphson method in one

variable is implemented starting with an initial guess for the
root, then the function is approximated by its tangent line and Finally, GLOM (minimized value) is
the X-intercept of this tangent line is computed. This process is
repeated as until a sufficiently accurate value is reached. N
Initial guess plays an important role in the solution to
mo Lmk
k=l +mpl
= (46)

overcome the shortcomings such as infinite iteration cycles and

non-convergence. Determining the bounds for the root will Structural mass of each stage is found from (12)
help to define the initial guess. It is clear that natural logarithm
(In) function in (31) is defined only for positive values. (47)

(l-ck)Ak > 0 (39)

And propellant mass of each stage is

(48)

861
 So far, staging has been optimized based on equations of 300,000 ,------

serial staging. But, many launch vehicles have two to six

250,000 +----------F---
number of boosters strapped on to the first stage to provide
�
more thrust at lift-off. For launch vehicles with parallel staging, -;;: 200,000
an equivalent launch vehicle with serial staging can be defmed. 1')
E
After staging has been optimized based on serial staging, � 150,000 +-------/-'------::;£=---
- ___ V = 9 km/s
optimal staging data can be converted to a parallel ,;
� 100,000 +-------7----:;£'--
:; -�It--
-+-V = 10 km/'

configuration using the defmitions given in Section II-B-2. � ......V-llkm/s

50,000 +-__."'-:.A....-""='------
.
V. DISCUSSION AND CONCLUSION
A Matlab® script has been written to determine the optimal a 1000 2000 3000 4000 5000 6000

staging of an N-stage rocket for a set of given technology pavload mass (kg)
options. Thus, a quick and effective tool to find optimal vehicle
configurations in the conceptual design phase of a generic Fig. 7. Variation of GLOM with payload mass.
multistage launch vehicle has been achieved.
It is necessary to point out that the method proposed in this
Staging optimization program developed in this study has study for estimating gravitational, aerodynamic and propulsive
been verified for different missions using available data of losses during flight is a poor approximation, but useful for
existing launch vehicles. In order to illustrate the method of preliminary evaluation during the conceptual design phase. It is
application, variation of GLOM with N, Ll Vmission and mp/ have always necessary to carry out trajectory simulations by solving
been shown graphically in Fig. 5 to Fig. 7. equations of motion for detailed performance analysis in order
Fig. 5 illustrates the variation of GLOM with the number of to check the validity of the assumptions.
stages. The decrease in GLOM between N 2 and N 3 is = =

quite appreciable, but for higher values of N the decrease REFERENCES

becomes very small. [I] F.J. Malina and M. Summerfield. 'The problem of escape from the earth
by rocket." Journal of the Aeronautical Sciences 14 (1947): 471.
350,000 ,-----
[2] M. Vertregt. "A method of calculating the mass ratio of step rockets."
300,000 Journal of the British interplanetary Society 15 (1956): 95.
mp'= 500 kg [3] M. Goldsmith. "On the optimization of two stage rockets." Jet
� 250,000 Propulsion 27 (1957): 415-416.
�

� 200,000 +----\---------- [4] 1. Weisbord. "A generalized optimization procedure for n-staged
;:: ___ v = 9 km/s
� +----+---
150,000
-+-v = 10km/, [5]
missiles." Jet Propulsion 28 (1958): 164-167.
M. Subotowicz. 'The optimization of the n-step rocket with different
[100.000 +--....-\--------- ...... V - 11 km/s construction parameters and propellant specific impulses in each stage."
Jet Propulsion 28 (1958): 460-463.

[6] H.H. Hall and E.D. Zambelli. "On the optimization of multistage
rockets." Jet Propulsion 28 (1958): 463-465.
3 6 7 [7] T.N. Srivastava. "Optimum staging with varying thrust attitude angle."
number of stages Defence Science Journal 16 (1966): 153-164.

[8] V.B. Tawakley. "On the calculation of optimum mass distribution of a

Fig. 5. Variation of GLOM with number of stages. multi-stage rocket vehicle." Defence Science Journal 17 (1967): 53-66.
[9] C.N. Adkins. "Optimization of multistage rockets including drag."
Fig. 6 and Fig. 7 illustrates the variation of GLOM with Journal of Spacecraft (1970): 751-755.

Ll Vmission and payload mass (mp/), respectively. Figures show [10] H.D. Curtis. Orbital mechanics for engineering students. 1st.
that GLOM increases exponentially with increasing Ll Vmission Burlington, MA: Elsevier Butterworth-Heinemann, 2005.

and increases linearly with the increasing mpl. [11] J.P. Loftus and C. Teixeira. "Chapter 18: Launch systems." Space
mission analysis and design. Ed. W.J. Larson and J.R. Wertz. 3rd. EI
60,000 ..,------ Segundo, CA & New York, NY: Microcosm Press & Springer, 1999.
719-744.
50,000 +----------,1''---- [12] A. Tewari. Atmospheric and space flight dynamics: Modeling and
� simulation with Matlab and Simulink. Boston: Birkhaeuser, 2007.
-;;: 40,000
1') [13] D.n. Burghes. "Optimum staging of multistage rockets." international
E Journal of Mathematical Education in Science and Technology 5.1
� 30,000 +------�=------:---- ___ 100 kg
(1974): 3-10.
,;
�
e
20,000 +----11=--------:;£----­
::;
-+-SOOkS

...... lOOOkg
aD

10,000 1-----..=----

8 10 11 12

mission d.ltaV (km/s)

Fig. 6. Variation of GLOM with mission Delta-V.

862

View publication stats