Vous êtes sur la page 1sur 14

中国科技论文在线 http://www.paper.edu.

cn
Can. Aeronaut. Space J., Vol. 56, No. 1, pp. 17–30, 2010

Conceptual design of wrap-around-fin rockets


using the MDO methodology
Liangyu Zhao, Gyung-Jin Park, Kwan Soo Lee, and Shuxing Yang

Abstract. The conceptual design of rockets is a system engineering that involves configuration, aerodynamics, engine,
dynamic stability, and coupling effects. To obtain benefits from the synergistic effects, the latest research achievements, and
the most powerful analysis tools efficiently, this paper focuses on how to apply the multidisciplinary design optimization
(MDO) technology to the conceptual design of the long-range, large length to diameter ratio wrap-around-fin (WAF)
rockets. The software iSIGHT was employed as the MDO framework, the multidisciplinary feasible (MDF) method as the
MDO architecture, and the multi-island genetic algorithm combined with sequential quadratic programming as the search
algorithm. By maximizing the payload ratio, the structural analysis, aerodynamics, engine, trajectory, and dynamic
stability were considered comprehensively. A case study demonstrated that MDO could improve the rocket
performance effectively.

Résumé. Le modèle conceptuel des fusées est associé au concept d’ingénierie des systèmes complexes qui comprend la
configuration, l’aérodynamique, le moteur, la stabilité dynamique ainsi que leurs effets de couplage. Afin de tirer profit
dans un souci de plus grande efficacité des effets synergétiques, des derniers développements en recherche et des outils
d’analyse les plus puissants, on met l’accent dans cette étude sur la façon d’appliquer la technologie d’optimisation
multidisciplinaire (OMD) au modèle conceptuel des fusées à ailettes enveloppantes (WAF) longue portée à rapport
longueur/diamètre élevé. Le logiciel d’optimisation iSIGHT a été utilisé comme cadre OMD, la méthode MDF
(« multidisciplinary feasible ») comme architecture OMD et l’algorithme génétique utilisant le modèle à ı̂lots multiples
(« multi-island genetic ») a été combiné à la programmation quadratique séquentielle à titre d’algorithme de recherche.
Dans une optique de maximisation du taux de charge utile, divers éléments ont été examinés en détail dont l’analyse
structurale, l’aérodynamique, le moteur, la trajectoire ainsi que la stabilité dynamique. Une étude de cas a démontré que la
technologie OMD pouvait s’avérer efficace pour l’amélioration de la performance des fusées.
[Traduit par la Rédaction]

Nomenclature Lcy length of cylindrical segment of control section


LE length of engine
A equatorial moment of inertia Lf distance between WAF trailing edge and rocket
AT21 burn time of first stage tail
AT22 burn time of second stage Ln length of nozzle
c absolute thickness of WAF Lr length of WAF root
c relative thickness of WAF LW wingspan of WAF
D rocket diameter p value of thrust
DT delay time PR payload ratio
Isp mass-specific impulse Tp period of pitch oscillation
I% ratio of total impulse of first stage to total Tr period of roll oscillation
impulse of entire engine THn absolute thickness of nozzle
L length of rocket V* volume
LA length of arc segment d installation angle of WAF
Lb length of warhead lA length to diameter ratio of arc segment
LC length of control section lb ratio of warhead length to rocket length

Received 2 October 2009. Published on the Web at http://pubservices.nrc-cnrc.ca/casj on 30 September 2010.


L. Zhang, and K.S. Lee. Department of Mechanical Engineering, Hanyang University, Seoul 133-791, Republic of Korea.
G.-J. Park.1 Department of Mechanical Engineering, Hanyang University, Gyeonggi-do 426-791, Republic of Korea.
S. Yang. School of Aerospace Engineering, Beijing Institute of Technology, Beijing, 100081, China.
1
Corresponding author (e-mail: gjpark@hanyang.ac.kr).

E 2010 CASI 17 转载
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

lB length to diameter ratio of rocket applied to the design of a winged missile and a hypersonic
lW aspect ratio of WAF air-breathing missile (Jun et al., 2003; Starkey et al., 2008),
new-concept civilian airplanes such as the blended-wing-
r* equivalent density
body (BWB) aircraft and supersonic transport (Choi et al.,
h0 fire angle 2008; Herrmann, 2008; Ko et al., 2003), a reusable launch
vc critical rotation rate vehicle (Brown and Olds, 2006), the unmanned air vehicle
ve balance rotation rate (UAV) (Lee et al., 2009; Rajagopal et al., 2007), and so on.
x0 sweepback angle of WAF leading edge Researchers have benefitted greatly from the MDO
methodology.
x1 forward-swept angle of WAF tailing edge Some MDO strategies, also referred to as MDO
j opening angle of WAF architectures, have already been proposed. In general,
MDO strategies can be classified into two categories based
on the number of optimizers in the entire architecture. One is
Introduction single-level architecture, which has one optimizer and
includes the multidisciplinary feasible (MDF) method and
The long-range multiple launch rocket is a type of the individual discipline feasible (IDF) method (Cramer et
traditional and popular ground-to-ground munition. It has al., 1994; Renaud, 2002). The other is multilevel architecture,
received a lot of attention, and significant support has been which has two or more optimizers and includes the
provided for its unique advantages in many countries. A collaborative optimization (CO) method, the concurrent
large length to diameter ratio with a wrap-around fin subspace optimization (CSSO) method, and the bilevel
(WAF) is the most popular aerodynamic configuration. integrated system synthesis (BLISS) method (Braun, 1996;
Many long-range rockets in current use are designed using Hajela et al., 1992; Kodiyalam and Sobieszczanski-Sobieski,
these criteria, including the multiple launch rocket system 1999; Sobieszczanski-Sobieski et al., 1998). A comparison of
(MLRS) and guided multiple launch rocket system these MDO architectures with mathematical examples can
(GMLRS) developed by the US, the 9M55K series be found in Yi et al. (2008). In addition, analytical target
projectiles of the BM-30 ‘‘Smerch’’ system developed by cascading (ATC) and isoperformance are also considered by
the former Soviet Union, and the A-100 system developed some designers as two kinds of MDO strategies and have
by China. With developments in science and technology, been receiving more and more attention (Kim et al., 2003;
these kinds of rockets will be used not only in the military Weck and Jones, 2006). Each is briefly explained here. MDF,
domain but also in civil engineering, such as in the field of which is the foundation of other MDO strategies, is easy to
artificial rainfall. In the foreseeable future, rocket use, but a complicated system analysis has to be involved in
developments will be made in terms of longer range, every step of the iteration. IDF was proposed to eliminate
smaller miss error, greater interchangeability of parts, and the system analysis by solving every subdiscipline
higher intelligence level. With increasing pressure from independently. Multilevel MDO strategies were proposed
competition in the world weapons market, the cost of to handle large-scale problems. CO, a two-level MDO
rockets should be minimized at the same time as the strategy, does not have to perform system analysis by
performance is maximized. This is the prime motivation of decomposing subdisciplines. The system level manages the
the multidisciplinary design optimization (MDO) method in overall process with the lower level, and all of the
the conceptual design of weapon systems just like it has been subdisciplines are performed concurrently. High-fidelity
used extensively in the aircraft design process (Kroo, 1997). models of subdisciplines are quite important in CO. CSSO,
Conceptual design is the first stage of the three major also a two-level MDO strategy, has a system optimizer and
phases in the design of a flying vehicle; the other two multiple subsystem optimizers. The system and the
stages are preliminary design and detailed design. subsystems optimizers have the same optimization
Designers have the greatest freedom at the conceptual objectives and the same or different design variables.
design stage, and decisions at this stage can directly affect Approximation techniques are fairly essential for CSSO.
the final performance of the flying vehicle (AIAA Technical BLISS was proposed based on the decomposition
Committee for Multidisciplinary Design Optimization, techniques. The key advantage of BLISS is that it allows
1991). The MDO, initiated in the 1980s, can be described designers to stop the optimization process at any time.
as a methodology for the design of systems where the ATC is based on the principles of the preliminary design of
interaction between several disciplines must be considered products, and isoperformance is based on the theories of
and where the designer is free to significantly affect the inverse design. However, we prefer to consider ATC and
system performance in more than one discipline (AIAA isoperformance as product design methods, such as MDO.
Technical Committee for Multidisciplinary Design Unfortunately, very little literature on the MDO can be
Optimization, 1991). The MDO is responsible for many found in the field of WAF rocket design. The aim of this
achievements in the fields of aeronautics and astronautics research was to apply the MDO methodology to the
(Weck et al., 2007). Currently, this methodology has been conceptual design of WAF rockets. This paper is organized

18 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

in three major sections. First, the MDO problem and the characteristics can be calculated, such as mass, center of
interactions of concerned disciplines are described, and gravity, moments of inertia, aerodynamic coefficients (e.g.,
some of the geometry variables, geometry constraints, and coefficients of lift, drag, and pitch moment), and change of
optimization objectives are described based on the thrust against time. Substituting these characteristics into the
simplification of a benchmark rocket. Second, the mathematical model of the ballistics, the range and change of
subdisciplines including component configuration or angle of attack with time and the change of angle of sideslip
parametric modeling, aerodynamics, engine analysis, with time are obtained. Lastly, the stability can be analyzed
trajectory simulation, and stability analysis are described. based on the flight information. This is one iteration of the
Third, a case is used to demonstrate that MDO techniques entire design process. If the performance is not good enough,
can truly improve the rocket performance effectively. UG NX the iteration should be performed again and again. The
(Siemens Co., available from www.plm.automation.siemens. relationships and interactions of the required disciplines
com/en_us/products/nx/index.shtml), a commercial computer- are illustrated in Figure 1, and a more detailed description
aided-design (CAD) package, was used for the parametric is given in the following sections.
modeling. A surrogate model technique was involved in the Basically, an MDO problem is an optimization problem
aerodynamics discipline; the MDF method was adopted as the with or without constraints. Design variables, optimization
MDO architecture, and the multi-island genetic algorithm objectives, and design constraints are the three key factors.
(MIGA) was combined with sequential quadratic programming Before defining these factors, a simplified benchmark rocket
as the search algorithm. is defined. It is based on the sophisticated A-100 rocket
system (the weapon system and rocket are shown in
Figures 2 and 3). The body of the simplified rocket, which
Definition of the problem is illustrated in Figure 4, includes five main components,
Multidisciplinary design optimization of WAF rockets, namely the control section, the warhead, the engine, the
similar to the traditional design method (Zhou and Ju, nozzle, and the stabilizer (the tail fins – wings). The
2005), requires a number of iterations to achieve a balance control section consists of an arc segment that starts from
between performance and design parameters. First, the the rocket nose and a cylindrical segment just after the arc
aerodynamic configuration is determined based on former segment. The former is generated through rotation of a
experience or new concepts. The geometry parameters and specified generatrix along the longitudinal axis of the
type of engine propellant (fuel) are then determined. After rocket, and the latter is assumed to be a solid cylinder. The
the burn time of the engine is chosen, the rocket warhead and engine are both simplified similar to the

Figure 1. Interaction of the concerned disciplines.

E 2010 CASI 19
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

Optimization objective
The selected optimization objective is to maximize the
payload ratio (PR), which is defined as the ratio of the
warhead mass (Mwarhead) to the rocket takeoff mass
(Mtakeoff). The objective is selected based on the designer’s
intention, and PR is defined as follows:
M warhead
PR ~
M takeoff
ð1Þ
V w rw
~
V ac rac z V cc rcc z V w rw z V e rei z V n rn z V s rs

where V is the volume; r is the equivalent density; and the


subscripts ac, cc, e, ei, n, s, and w denote arc segment,
cylindrical segment, engine, initial engine, nozzle, WAF,
and warhead, respectively. Normally, the volume of the
Figure 2. A-100 weapon system. specified model can be calculated automatically using
CAD software. However, equivalent densities must be
cylindrical segment of the control section and consist of solid defined by the designer. The values and units of the
cylinders. The nozzle is a hollow cylinder. The profile of the concerned equivalent densities are shown in Table 3, and
simplified stabilizer, which is shown in Figure 5, consists of all were deduced based on the benchmark rocket.
six panels of WAF.
Design constraints
Design variables
The constraints of the geometry can be regarded as the
Based on the basic theory of rocket design (Zhou and Ju, design space. For example, the rocket diameter belongs in
2005) and the aforementioned simplifications, the the range of [100.0, 400.0] (in mm), and the rocket length to
independent parameters to define the rocket geometry are diameter ratio belongs in the range of [10.0, 30.0]. The design
shown in Table 1. They are also the design variables for space of the geometry is defined according to popular
the rocket geometry. rockets worldwide and the current industrial foundation of
Though the parameters in Table 1 are independent and a specified country (China). Some other design constraints in
generally used to describe the rocket geometry, they are the subdisciplines are discussed in detail later in the paper.
not convenient enough to construct the solid rocket model,
especially using CAD software. To fix this problem, the
Multidisciplinary design optimization
relevant design parameters, which can be deduced from the
aforementioned independent design variables and are more The MDF method, which is also referred to as the fully
convenient to construct the geometry model, are shown in integrated optimization (FIO) method or nested analysis and
Table 2. design (NAND) method (Keane and Nair, 2005), was
employed as the MDO architecture. The whole MDO
system includes structural analysis, aerodynamics, engine
analysis, trajectory simulation, and stability analysis, which
are described in detail in the following sections.

Structural analysis
Structural analysis computes the mass, center of gravity,
and moments of inertia of the rocket. Stress and fatigue have
not yet been considered. This subdiscipline can also be called
parametric modeling or components configuration.
A novel geometry representation method, called the ‘‘point-
orientation and vector-direction’’ method, is proposed based
on the simplified rocket and is illustrated in Figure 6. Five
points (solid circles) and one vector (solid line with an
arrowhead) are involved. Orientation points are employed
to describe the starting and end points of each component,
and the direction vector is used to describe the longitudinal
Figure 3. A-100 rocket. axis of the rocket (from nose to tail). Five orientation points

20 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

Figure 4. Profile of the body of the simplified rocket.

Figure 5. Profiles of the stabilizing device of the simplified rocket.

Table 1. Design variables for the rocket geometry. Table 2. Design parameters for the rocket geometry.

Symbol Description Symbol Description


D (mm) Rocket diameter; diameter of warhead, engine, and L (mm) Length of rocket; calculated as L 5 lBD
cylinder segment of control section LA (mm) Length of arc segment; calculated as LA 5 lAD
lB Length to diameter ratio of rocket; defined as ratio Lcy (mm) Length of control section cylindrical segment;
of rocket length to rocket diameter, and can be assumed to be a function of D and L; Lcy 5
described as lB 5 L/D f(D, L) 5 exp[(D 2 300.0)/D] 6 506.0 6 L/7600.0
lA Length to diameter ratio of arc segment; defined as LC (mm) Length of control section; LC 5 LA + Lcy
ratio of arc segment length to rocket diameter Lb (mm) Length of warhead; Lb 5 lb(L 2 LC)
lb Ratio of warhead length to rocket length (excluding Lf (mm) Distance between WAF trailing edge root and rocket
control section); can be used to describe the tail; Lf 5 L 6 121.0/7600.0
warhead power and calculate the payload ratio LW Wingspan of WAF; LW 5 D sin(j)/2, where j ƒ 2p/n
j (u) Opening angle of WAF; shown in Figure 5 and (n is the number of WAFs and is kept constant at 6
defined as angle between the chord and the for the benchmark rocket)
tangent to the surface of the nozzle section Lr (mm) Length of WAF wing root; Lr 5 LW/lW
lW Aspect ratio of WAF; defined as ratio of WAF span Ln (mm) Length of nozzle; assumed to be a function of Lf and
to chord length of WAF wing root Lr: Ln 5 Lf + Lr 6 1.1 + 10.0
c Relative thickness of WAF; defined as ratio of WAF THn (mm) Absolute thickness of nozzle; assumed to be a
thickness to chord length of WAF wing root function of D, THn 5 D/10.0
d (u) Installation angle of WAF; defined as angle between LE (mm) Length of engine; LE 5 L 2 LC 2 Lb 2 Ln
WAF wing root axis and the rocket longitudinal c (mm) Absolute thickness of WAF; c ~ cLr
axis; illustrated in Figure 5 x1 (u) Sweepback angle of trailing edge of WAF; constant
x0 (u) Sweepback angle of WAF leading edge; shown in and equals 20u for the benchmark rocket
Figure 5
Note: All of the mathematical relationships in the table are determined
based on the benchmark rocket. Some of these parameters are shown in
Figures 4 and 5.
combined with the direction vector can describe the nose, the
end plane of the control cabin and the starting plane of the There are two advantages with this geometry
warhead, the end plane of the warhead and the starting plane representation method. First, every component can be
of the engine, the end plane of the engine and the starting constructed in parallel as long as the orientation points,
plane of the nozzle, and the tail of the rocket. direction vector, and relevant parameters are known. The

E 2010 CASI 21
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

Table 3. Equivalent densities (kg/m3). modeling is developed based on the secondary


development of the commercial CAD package called UG
Value NX (Siemens Co., available from www.plm.automation.siemens.
Symbol Description (6103) com/en_us/products/nx/index.shtml) using C++ programming
rac Equivalent density of arc segment 1.3840 language (Kirch-Prinz and Prinz, 2001). The three-
rcc Equivalent density of cylindrical segment 1.1399 dimensional (3D) solid model of one WAF rocket with a
rw Equivalent density of warhead 1.7856 Karman nose and six WAF panels is shown in Figure 7, and
rei Initial equivalent density of engine 1.7856 the six 3D WAF panels are shown in Figure 8. The opening
ref Final equivalent density of engine 0.5603
angle of the WAFs is shown in Figure 9, which is the right tail
rn Equivalent density of nozzle 0.5603
rs Equivalent density of WAF 7.8000 view of the rocket. The characteristics of the rocket, such as
mass, center of gravity, and moments of inertia, can be
obtained using the nested function of UG NX. The
rocket model is obtained just by assembling these generation of invalid designs is avoided during the process
components. Second, this method is compatible with of parameterization. This can increase the robustness of the
various fidelity models, which is very important and very entire optimization process. A more detailed description of
useful for MDO. parametric modeling can be found in Zhao (2007).
Based on the theory of Samareh (2001), parametric This section provides rocket geometry parameters for the
modeling techniques can be classified into three categories, aerodynamics subdiscipline, mass, center of gravity and
namely the discrete approach, free-form deformation moments of inertia for the ballistic simulation, and
methods, and CAD-based approaches. The last category is relevant geometry parameters for the engine analysis. In
the most widely used because there are so many benefits addition, it provides two-dimensional (2D) figures and 3D
from the current CAD software. The code of parametric models for the design documentations.

Figure 6. Orientation points (solid circles) and direction vector (arrow).

Figure 7. Three-dimensional (3D) solid model of rocket with a Karman nose.

22 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

fluid dynamics (CFD) is a good substitute for wind tunnels


(Anderson, 1995), but there are still some problems integrating
CFD techniques into the MDO system. For instance, CFD
tools are not automatic and robust enough, especially on mesh
generation. In addition, the CFD technique is time-
consuming. As a compromise in terms of computation
accuracy and time, the surrogate model technique was
employed in our MDO system. This technique is a key
element of MDO methodology and has been used in almost
every MDO application (Simpson et al., 2004; 2008).
The orthogonal array is employed as the design of
experiment (DOE) technology and radial bases function
(RBF) neural network as the approximation method (Hick
and Turner, 1999; Keane and Nair, 2005). High-fidelity data
are obtained using the commercial CFD package FLUENT
(Tu et al., 2007). A structured mesh of a representative WAF
rocket is shown in Figure 10, and the mesh of the WAF and
its vicinity are shown in Figure 11. Figure 12 shows a typical
velocity contour near the nose of the WAF rocket from a
high-fidelity model. In this case, the Mach number is 3.0 and
the angle of attack is 0.0. Steady flow fields were simulated to
obtain the static aerodynamic coefficients, such as the
Figure 8. Six panels of the WAF. coefficients of lift, drag, and pitch moment. Unsteady flow
fields with a dynamic mesh were simulated to obtain the
Aerodynamics dynamic derivatives, like the coefficient of the derivative of
the roll damping. The governing equations here were the
The MDO methodology emphasizes that the high-fidelity Reynolds averaged Navier–Stokes (RANS) equations with
computation models should be employed. In terms of the Spalart–Allmaras 1 equation turbulence model (Tu et al.,
aerodynamics, experimental data obtained using wind 2007). More information about how to calculate relevant
tunnels are undoubtedly closest to reality but are too aerodynamic coefficients and validation of this calculation
expensive and time-consuming to be practical and method can be found in Mao et al. (2007a). A more detailed
necessary at the conceptual design stage. Computational description of how to construct surrogate models and
comparisons of several surrogate model techniques, such
as the polynomial response surface, the Kriging method,
and the RBF neural network, are given in Zhao (2007) and
Zhao et al. (2007). The RBF neural network has been
identified as having the highest approximation accuracy
(relative error less than 5.0%) for our problem (Zhao,
2007). It is also known that the RBF neural network will
provide higher approximation accuracy if it is constructed
for every coefficient of the aerodynamic force–moment
separately (Zhao et al., 2007). The neural network toolbox
of MATLAB was employed to construct an RBF neural
network (The MathWorks, Inc., 2006), and the surrogate
model was constructed off-line.
This section describes the key element of the MDO system.
It provides a fast and effective method to integrate an
aerodynamics high-fidelity model into the MDO system.
For example, an evaluation using FLUENT needs about
3 h, and an evaluation using the well-established surrogate
model needs only 8 ms.

Engine
Generally, the WAF rocket uses a solid rocket engine as a
Figure 9. Tail view of the rocket. thrust generator. It is unnecessary to design the engine in

E 2010 CASI 23
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

Figure 10. Side view of the mesh for CFD.

Figure 12. Typical flow field of the rocket nose.

Figure 11. Mesh of the wing.

detail because it is enough for conceptual design to know the


changes in thrust and rocket mass with the burn time. In
response to this characteristic, two types of simplified
propulsion were introduced, one with a single combustion
chamber with a single thrust value (type 1), and one with a
two combustion chambers with two thrust values (type 2)
(Dong and Shang 1996), as sketched in Figure 13. It is
assumed that the mass-specific impulse (Isp) of the solid
engine is constant and is obtained using the following
equation based on the benchmark rocket: Figure 13. Thrust changes versus operating time.

24 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

Taking the burn time of the second stage (AT22) as an


independent design variable, the thrust value of the second
stage (p2) is calculated as follows:

p2 ~ I ð1 { I% Þ = AT22 ð8Þ

The total operating time of the engine (TAT2) is


calculated as

TAT2 ~ AT21 z DT z AT22 ð9Þ

This section provides changes of the thrust versus flight


time and changes of the engine mass center against the flight
time to the trajectory simulation sub-discipline.

Trajectory simulation
The six-degree-of-freedom (6-DOF) trajectory equations
of a spinning missile were employed to simulate the rocket’s
Figure 14. Mass changes versus operating time. flight ballistics (Qian et al., 2000). MATLAB/Simulink was
employed as the solver of the ballistics ordinary differential
equations (ODEs). This discipline contributes a design
I sp ~ 839:31=333:55 ~ 2:5157kN:s:kg{1 ð2Þ
variable (h0) called the fire angle, which is defined as the
The changes in mass with burn time for the two types of angle between the ground floor and the longitudinal axis
engines are sketched in Figure 14. This discipline provides of the rocket when launching. In addition, this discipline
five more design variables, namely the propulsion type (PS), offers a constraint, the range, which is one of the major
the ratio of the total impulse of the first stage to the total criteria to evaluate the rocket performance.
impulse of the entire engine (I%, equals 1.0 for type 1), the
delay time (DT) between the first stage burning and the Stability analysis
second stage burning, the burn time of the first stage The stability analysis consists of the static stability
(AT21), and the burn time of the second stage (AT22). analysis and the dynamic stability analysis. Dynamic
PS is equal to zero (type 1) or one (type 2). The mass stability analysis includes the coning motion analysis, the
difference of the engine before and after burning (Dm) can following stability analysis, and the resonance effects
be calculated as follows: analysis. This discipline will introduce four more
constraints into the MDO system as follows: the static
Dm~Ve (rei {ref ) ð3Þ
stability margin must be larger than 10.0%, the ratio of the
The total impulse of the solid engine (I) can be then be pitch oscillation period to the roll oscillation period must be
determined using the following equation: larger than one safe factor (1.2 for this kind of rocket) when
the rocket is flying, the dynamic equilibrium angle must be
I ~ Isp Dm ð4Þ smaller than 3.0u at the highest flight point, and the
difference between the balanced rotation rate (ve) and the
For type 1, I% equals 1.0, only the total burn time of the critical rotation rate (vc) must be smaller than 0.0.
first stage (TAT1) is the independent design variable, and p The static stability margin (ssm) is calculated as follows:
can be calculated as follows:
ssm ~ ðxF { xG Þ = L ð10Þ
p ~ I = TAT1 ð5Þ
where xF is the distance between the nose and the center of
For type 2, I% is an independent design variable, so the the pressure from the air, xG is the distance between the nose
total impulse of the first stage (I1) can be calculated as and the center of gravity, and L is the length of the rocket. If
follows: ssm . 0, the rocket has static stability; ssm 5 0 means a
transient static stability; and ssm , 0 means a static
I1 ~I|I% ð6Þ
instability. More information on how to calculate static
Taking the burn time of the first stage (AT21) as another stability can be found in Qian et al. (2000).
independent design variable, the thrust value of the first The following stability is caused by the gravity of the
stage (p1) is calculated as follows: rocket (Xu, 2004). To avoid this phenomenon, the
dynamic equilibrium angle (Da) must be smaller than 3.0u
p1 ~ I1 = AT21 ð7Þ at the highest flight point. This equilibrium angle, also

E 2010 CASI 25
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

considered as the total angle of attack at the highest point, is MDO framework and search algorithm
defined as the angle between the longitudinal axis and the
The commercial software iSIGHT developed by
vector of the flying velocity of the rocket and can be
Engineous Software, Inc. was employed as the MDO
calculated using the following equation (Xu, 2004):
framework. Because it is a highly nonlinear problem, the
kzd kzd MIGA combined with sequential quadratic programming
Da~ 2 h_ 2 2 g cos h ð11Þ nested in iSIGHT was employed as the search algorithm
k V k V
(Engineous Software, Inc., 2004).
where A is the equatorial moment of inertia, k 5 (|kz|)1/2, kz 5 Similar to other widely used genetic algorithms, the
rSLmz/2A, kzd 5 rSL2mzd/2A, mz is the coefficient of the MIGA employs a so-called elitism technique which
pitch moment, mzd is the derivative of the coefficient of the guarantees that the best genetic material is carried over to
damping of pitch moment, r is the air density, L is the the child generation and allows the best individuals from the
reference length, S is the reference area, and h is the angle previous generation to be preserved without alteration. In
of the obliquity of the trajectory. addition, the selection operation in MIGA employs the
The period of the pitch oscillation Tp is just the period of so-called ‘‘tournament selection’’ scheme. The main feature
the angle of attack and can be calculated as of MIGA that distinguishes it from traditional genetic
algorithms is the fact that each population of individuals is
Mz z Mzvz ~ Aa
€ ð12Þ
divided into several subpopulations called ‘‘islands.’’ Some
where Mz is the pitch moment, Mzvz is the damping of the individuals are then selected from each island and migrated
pitch moment, and a is the angle of attack, and with the to different islands periodically. This operation is called
initial condition described as follows: ‘‘migration.’’ Two parameters control the migration
 process, namely the migration interval, which is the
at~0 ~a0 number of generations between each migration, and the
ð13Þ
a_ t~0 ~0 migration rate, which is the percentage of individuals
migrated from each island at the time of migration
where t is the time. (Engineous Software, Inc., 2004).
The period of the roll oscillation Tr is calculated by solving
Flow chart and realization
Tr ~ 2p = vx ð14Þ
Based on the previous descriptions, the flow chart of an
Thus, the following equation must be satisfied to avoid the MDO system of the conceptual design of WAF rockets is
resonance effects: illustrated in Figure 15. First, all of the independent design
Tp variables are organized into four groups based on the special
§1:2 ð15Þ requirement of each subdiscipline. The organizations are
Tr
shown in Table 4. Each group is then transformed to the
Equation (15) can also be written as corresponding solver. After that, the mass, center of
gravity, moments of inertia, and PR are obtained through
vx § 1:2 | 2p = Tp ð16Þ structural analysis; coefficients of aerodynamic force and
aerodynamic moments, both of which depend on the flight
More detailed information on how to deal with the following
velocity and the angle of attack or the angle of sideslip, are
stability and resonance effects can be found in Xu (2004).
calculated by the surrogate model in the aerodynamics
The balanced rotation rate (ve) and the critical rotation
discipline; the change in thrust with burn time, the change
rate (vc) can be calculated as follows:
in engine mass with burn time, and the change in engine mass
c1 V center with burn time are deduced in engine analysis; and the
ve ~{ ð17Þ launching system module in the flow chart is to set the fire
c3 L
angle and submit it to the trajectory simulation. All of the
c7 c9 qSL2 outputs of the aforementioned four disciplines are provided
vc ~ ð18Þ to the trajectory simulation. Then, the range, PR, thrust
(C{A)c5 V
value of the first stage, ratio of the first thrust to the
where V is the velocity of the rocket, q is the dynamic second thrust, static stability margin, ve, vc, ratio of the
pressure, C is the pole moment of inertia, and c1, c2, c5, c7, pitch period to the yaw period, and equilibrium angle are
and c9 are obtained based on the aerodynamic coefficients obtained. The latter four outputs are used to evaluate the
by the weighted least squares method. stability of the rocket in the stability analysis module, and
When ve , vc for every point of the ballistics of the the other outputs are used to evaluate the rocket
rocket, the coning motion would result in asymptotic performance. The entire process is under the control of the
stability. More detailed discussions on coning motion can MDO framework, and the realization on the iSIGHT
be found in Mao et al. (2006; 2007b) and Zhao (2007). platform is shown in Figure 16.

26 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

Figure 15. Flow chart of MDO system.

Table 4. Organization of the design variables into groups based on the requirements of each subdiscipline.

D lB lA lb Dj lW c d x0 PS I% AT21 DT AT22 h0
Structural analysis ! ! ! ! ! ! ! ! ! ! ! ! ! ! 6
Aerodynamics ! ! ! ! ! ! ! ! ! 6 6 6 6 6 6
Engine ! ! ! ! 6 6 6 6 6 ! ! ! ! ! 6
Trajectory simulation 6 6 6 6 6 6 6 6 6 6 6 6 6 6 !

E 2010 CASI 27
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

Figure 16. Realization on the iSIGHT platform.

Case study the equality constraint is a little difficult to reach. Dj in


Table 5 is calculated as follows:
One case is studied here. The design variables and design
space are described in Table 5, and the objective and Dj ~ j0 { j ð19Þ
constraints are described in Table 6. To evaluate the
optimization results more conveniently, the propulsion where j0 is the opening angle of the benchmark rocket and is
type was not considered as an optimization variable and equal to 60.0u.
kept the same as that of the benchmark rocket. Based on In this case, the relevant parameters of MIGA are
the current industrial capability, the thrust value of the first described as follows: the size of subpopulation is 20, the
stage was limited to 250.0 kN. The range was supposed to be number of islands is 10, the number of generations is 100,
70.0 km but was set in the range of [69.95, 70.05] km because the rate of crossover is 1.0, the rate of mutation is 0.01, the

Table 5. Description of design variables and design space.

Design Optimization Lower Benchmark Upper Final


variable variable limit value limit result
D (mm) ! 100.00 300.00 400.00 395.76
lB ! 10.00 25.30 30.00 21.33
lA ! 2.00 2.30 2.50 2.49
lb (%) ! 10.00 35.00 50.00 44.45
Dj (u) ! 0.00 0.00 20.00 7.27
lW ! 0.400 0.600 0.800 0.628
c ! 0.010 0.015 0.020 0.012
d (9) ! 30.0 40.0 60.0 48.0
x0 (u) ! 15.00 20.00 30.00 20.67
PS 6 0 1 1 1
I% (%) ! 30.0 70.0 90.0 58.2
AT21 (s) ! 0.00 5.00 20.00 5.00
DT (s) ! 0.00 2.00 50.00 9.04
AT22 (s) ! 0.00 10.00 20.00 12.65
h0 (u) ! 40.00 45.00 60.00 58.00

28 E 2010 CASI
中国科技论文在线 http://www.paper.edu.cn
Canadian Aeronautics and Space Journal / Journal aéronautique et spatial du Canada

Table 6. Objective and constraints.

Lower limit Benchmark value Upper limit Final result


Objective
Payload ratio (q) 0.0000 0.3446 — 0.4248
Constraint
Range (km) 69.950 47.770 70.050 70.026
Difference between ve and vc (s–1) — 246.96 0.00 243.02
Ratio of pitch period to roll period 1.20 2.30 — 2.20
Equilibrium angle at highest point (u) — 0.87 3.00 0.95
Static stability margin (%) 10.0 23.6 — 22.9
Thrust value of first stage (kN) — 108.37 250.00 134.96
Ratio of first thrust to second thrust 2.00 4.70 — 3.51

rate of migration is 0.5, and the interval of migration is 5. The experiences of subdiscipline experts are absolutely
Based on these parameters, there would be more than 5000 ignored. The collaborative optimization (CO) method is
iterations. We assumed that it was enough to find a good considered to be the most practical because its
result. iSIGHT decided the final result based on the configuration is consistent with the divisions in the current
feasibilities of all of the valid design points. The final industry organization. We also plan to apply the CO method
results are shown in Tables 5 and 6. to the conceptual design of the WAF rocket in the future.
The final results showed that the payload ratio was
improved from 0.3446 to 0.4248. Additionally, the range
increased significantly. To achieve these goals, the diameter
Acknowledgments
should be increased, and the rocket length to diameter ratio This research was supported by the Second Brain Korea
was decreased. This results in the rocket length increasing 21 Project in the year 2009 and by the World Class Univer-
from 7.95 m to 8.44 m. With the increase in lA, the arc sity (WCU) program through the Korea Science and Engin-
segment length of the control section increased from 0.69 m eering Foundation funded by the Ministry of Education,
to almost 1.0 m. The cylindrical segment length of the control Science and Technology (No. R32-2008-000-10022-0). The
section increased from 0.55 m to 0.72 m, which means that authors are thankful to Mrs. MiSun Park for her correction
there is much more space for the control system. The warhead of the manuscript.
length increased from 2.24 m to 3.00 m. In addition, the ratio
of the total impulse in the first stage to the total impulse for
the engine decreased with the increase in delay time. We think References
this is useful to control the flight velocity and improve the AIAA Technical Committee for Multidisciplinary Design Optimization. 1991.
dynamic stability. The decrease in the opening angle could White paper on current state of the art in multidisciplinary design
also improve the stability of the coning motion. optimization (MDO). American Institute of Aeronautics and
Astronautics, Inc. (AIAA), Reston, Va.

Anderson, J.R. 1995. Computational fluid mechanics: the basics with


Conclusions and discussion applications. 1st ed. McGraw-Hill, Inc., New York.
Research has been conducted on the improvement of the Braun, R. 1996. Collaborative optimization: an architecture optimization for large-
WAF rocket performance by MDO at the conceptual design scale distributed design. Ph.D. thesis, Stanford University, Stanford, Calif.
stage. A test case based on the MDF method obtained
Brown, F., and Olds, R. 2006. Evaluation of multidisciplinary optimization
satisfactory results and improved the payload ratio by up techniques applied to a reusable launch vehicle. Journal of Spacecraft and
to 23% with all the constraints being met. The results Rockets, Vol. 43, No. 6, pp. 1289–1300. doi:10.2514/1.16577.
achieved in this paper can be a reference when designing or
Choi, S., Alonso, J.J., Kroo, M.I., and Wintzer, M. 2008. Multifidelity design
modifying this kind of rocket.
optimization of low-boom supersonic jets. Journal of Aircraft, Vol. 45,
Circular error probability (CEP) is a comprehensive and No. 1, pp. 106–118. doi:10.2514/1.28948.
common criterion for evaluating the performance of the
multiple launch rocket system. Integrating the control and Cramer, E.J., Dennis, J.E., Frank, P.D., Lewis, R.M., and Shubin, G.R.
1994. Problem formulation for multidisciplinary optimization. SIAM
guidance system into the MDO system is one of the future Journal on Optimization, Vol. 4, No. 4, pp. 754–776. doi:10.1137/0804044.
research directions. This can motivate the application of
MDO to the preliminary design stage and even to the Dong, S., and Zhang, Z. 1996. Theory of solid rocket engines. Beijing Institute
detailed design stage. of Technology Press, Beijing, China. [In Chinese.]
Though MDF has the best convergence property of all of Engineous Software, Inc. 2004. iSIGHT_9.0’s reference guide. Engineous
the MDO architectures, it is a single-level MDO strategy. Software, Inc., Cary, N.C.

E 2010 CASI 29
中国科技论文在线 http://www.paper.edu.cn
Vol. 56, No. 1, April/avril 2010

Hajela, P., Bloebaum, C., and Sobieszczanski-Sobieski, J. 1992. Application ASC Structures, Structural Dynamics and Materials Conference, 23–
of global sensitivity equations in multidisciplinary aircraft synthesis. 26 April 2007, Honolulu, Hawaii. American Institute of Aeronautics
Journal of Aircraft, Vol. 27, No. 12, pp. 1002–1010. doi:10.2514/3.45974. and Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 2007-1885.

Herrmann, U. 2008. Multiple discipline optimization and aerodynamic off- Renaud, G. 2002. An application of multidisciplinary feasible formulation for
design analysis of supersonic transport aircraft. Journal of Aircraft, aircraft wing design. NRC Institute for Aerospace Research, Ottawa, Ont.
Vol. 45, No. 5, pp. 1474–1480. doi:10.2514/1.32673.
Samareh, J. 2001. Survey of shape parameterization techniques for high-
Hick, C., and Turner, K. 1999. Fundamental concepts in the design of fidelity multidisciplinary shape optimization. AIAA Journal, Vol. 39,
experiments. 5th ed. Oxford University Press, Eynsham, UK. No. 5, pp. 877–884. doi:10.2514/2.1391.

Jun, S., Tischler, A.V., and Venkayya, B.V. 2003. Multidisciplinary design Simpson, T.W., Booker, A.J., Ghosh, D., Giunta, A.A., Koch, P.N., and Yang,
optimization of a built-up wing structure with tip missile. Journal of R. 2004. Approximation methods in multidisciplinary analysis and
Aircraft, Vol. 40, No. 6, pp. 1093–1097. doi:10.2514/2.7219. optimization: a panel discussion. Structural and Multidisciplinary
Optimization, Vol. 27, No. 5, pp. 302–313. doi:10.1007/s00158-004-0389-9.
Keane, J.A., and Nair, B.N. 2005 Computational approaches for aerospace
design: the pursuit of excellence. John Wiley & Sons, Ltd., New York. Simpson, T.W., Toropov, V., Balabanov, V., and Viana, F. 2008. Design and
analysis of computer experiments in multidisciplinary design optimization:
Kim, H.M., Michelena, N.F., Papalambros, P.Y., and Jiang, T. 2003. Target A review of how far we have come — or not. In Proceedings of the 12th
cascading in optimal system design. Journal of Mechanical Design AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, 10–
(Transactions of the ASME), Vol. 125, No. 3, pp. 474–480. doi:10. 12 September 2008, Victoria, B.C. American Institute of Aeronautics and
1115/1.1582501. Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 2008-5802.
Kirch-Prinz, U., and Prinz, P. 2001. A complete guide to programming in C++. Sobieszczanski-Sobieski, J., Agte, J., and Sandusky, R. 1998. Bi-level
Jones and Bartlett Publishers, Sudbury, Mass. integrated system synthesis. In Proceedings of the 7th AIAA/NASA/
ISSMO Symposium on Multidisciplinary Analysis and Optimization, 2–
Ko, A., Leifsson, T., Mason, H.W., Haftka, T.R., Schetz, J., and Grossman, 4 September 1998, St. Louis, Mo. American Institute of Aeronautics and
B. 2003. MDO of a blended-wing-body transport aircraft with distributed Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 98-4916.
propulsion. In Proceedings of the 3rd Annual Aviation Technology,
Integration, and Operations Technical Forum, 17–19 November 2003, Starkey, P.R., Liu, D., Baldelli, H.D., Chang, R., and Che, P.C. 2008. Rapid
Denver, Colo. American Institute of Aeronautics and Astronautics, conceptual design and analysis of a hypersonic air-breathing missile. In
Inc. (AIAA), Reston, Va. AIAA Paper 2003-6732. Proceedings of the 15th AIAA International Space Planes and Hypersonic
Systems and Technologies Conference, 28 April – 1 May 2008, Dayton,
Kodiyalam, S., and Sobieszczanski-Sobieski, J. 1999. Bi-level integrated Ohio. American Institute of Aeronautics and Astronautics, Inc. (AIAA),
system synthesis with response surfaces. In Proceedings of the 40th Reston, Va. AIAA Paper 2008-2590.
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Materials Conference, 12–15 April 1999, St. Louis, Mo. American The MathWorks, Inc. 2006. MATLAB user’s guide. The MathWorks, Inc.,
Institute of Aeronautics and Astronautics, Inc. (AIAA), Reston, Va. Natick, Mass.
AIAA Paper 99-1306.
Tu, J., Yeoh, G., and Liu, C. 2007. Computational fluid dynamics: A practical
Kroo, I. 1997. Multidisciplinary optimization applications in preliminary approach. Elsevier Inc., New York.
design — Status and directions. In Proceedings of the 38th AIAA/ASME/
ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Weck, O., and Jones, M. 2006. Isoperformance: analysis and design of
7–10 April 1997, Kissimmee, Fla. American Institute of Aeronautics complex systems with desired outcomes. Systems Engineering, Vol. 9,
and Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 1997-1408. No. 1, pp. 45–61.

Lee, S., Gonzalez, F., Periaux, J., and Srinivas, K. 2009. Robust multi- Weck, O., Agte, J., Sobieski, J., Arendsen, P., Morris, A., and Spieck, M.
objective aerostructural optimization using advanced evolutionary 2007. State-of-the-Art and future trends in multidisciplinary design
algorithms. In Proceedings of the 47th AIAA Aerospace Sciences Meeting, optimization. In Proceedings of the 48th AIAA/ASME/ASCE/AHS/
5–8 January 2009, Orlando, Fla. American Institute of Aeronautics and ASC Structures, Structural Dynamics and Materials Conference, 23–
Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 2009-970. 26 April 1999, Honolulu, Hawaii. American Institute of Aeronautics
and Astronautics, Inc. (AIAA), Reston, Va. AIAA Paper 2007-1905.
Mao, X., Yang, S., and Xu, Y. 2006. Research on the coning motion of wrap-
around-fin projectiles. Canadian Aeronautics and Space Journal, Vol. 52, Xu, M. 2004. Rocket ballistics. Haerbin Institute of Technology Press,
No. 3, pp. 119–125. Harbin, China. [In Chinese.]

Mao, X., Yang, S., and Zhao, L. 2007a. Numerical simulation on Yi, I.S., Shin, K.J., and Park, G.J. 2008. Comparison of MDO methods with
aerodynamic characteristics of rockets with wrap around fins. Journal mathematical examples. Structural and Multidisciplinary Optimization,
of Solid Rocket Technology, Vol. 30, No. 3, pp. 181–184. [In Chinese.] Vol. 35, No. 5, pp. 391–402. doi:10.1007/s00158-007-0150-2.

Mao, X., Yang, S., and Xu, Y. 2007b. Coning motion stability of wrap Zhao, L. 2007. Research on multidisciplinary design optimization system of
around fin rockets. Science in China Series E: Technological Sciences, conceptual design of wrap-around-fin rockets. Ph.D. thesis, Beijing
Vol. 50, No. 3, pp. 343–350. Institute of Technology, Beijing, China. [In Chinese.]

Qian, X., Lin, R., and Zhao, Y. 2000. Missile flight mechanics. Beijing Zhao, L., Yang, S., and She, H. 2007. Research on constructing surrogate
Institute of Technology Press, Beijing, China. [In Chinese.] model of rocket aerodynamic discipline. Journal of Solid Rocket
Technology, Vol. 30, No. 1, pp. 1–4. [In Chinese.]
Rajagopal, S., Ganguli, R., Pillai, A., and Lurdharaj, A. 2007. Conceptual
design of medium altitude long endurance UAV using multi objective Zhou, C., and Ju, Y. 2005. Theory of design of rocket artilleries. Press of the
genetic algorithm. In Proceedings of the 48th AIAA/ASME/ASCE/AHS/ Beijing Institute of Technology, Beijing, China. [In Chinese.]

30 E 2010 CASI