AERODYNAMIC OPTIMIZATION OF MISSILE EXTERNAL

CONFIGURATIONS

A THESIS SUBMITTED TO
THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES
OF
MIDDLE EAST TECHNICAL UNIVERSITY

BY

KIVAN ARSLAN

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR
THE DEGREE OF MASTER OF SCIENCE
IN
AEROSPACE ENGINEERING

SEPTEMBER 2014

Approval of the thesis

AERODYNAMIC OPTIMIZATION OF MISSILE EXTERNAL
CONFIGURATIONS
Submitted by KIVAN ARSLAN in partial fulfillment of the requirements for the
degree of Master of Science in Aerospace Engineering Department, Middle East
Technical University by,

Prof. Dr. Canan zgen

Dean, Graduate School of Natural and Applied Sciences

_______________

Prof. Dr. Ozan Tekinalp

Head of Department, Aerospace Engineering

_______________

Prof. Dr. Serkan zgen

Supervisor, Aerospace Engineering Dept., METU

_______________

Examining Committee Members:

Prof. Dr. Hseyin Nafiz Alemdarolu
Aerospace Engineering Dept., METU

_______________

Prof. Dr. Serkan zgen

Aerospace Engineering Dept., METU

_______________

Prof. Dr. Yusuf zyrk

Aerospace Engineering Dept., METU

_______________

Assoc. Prof. Dr. Sinan Eyi

Aerospace Engineering Dept., METU

_______________

Assoc. Prof. Dr. Metin Yavuz

Mechanical Engineering Dept., METU

_______________

Date: 05/09/2014

I hereby declare that all the information in this document has been obtained
and presented in accordance with academic rules and ethical conduct. I also
declare that, as required by these rules and conduct, I have fully cited and
referenced all material and results that are not original to this work.
Name, Last name

: Kvan ARSLAN

Signature

iv

ABSTRACT

AERODYNAMIC OPTIMIZATION OF MISSILE EXTERNAL

CONFIGURATIONS

Arslan, Kvan
M.S., Department of Aerospace Engineering
Supervisor: Prof. Dr. Serkan zgen
September 2014, 85 Pages

In this thesis, design optimization methods capable of optimizing aerodynamic

performances of missiles and rockets are developed. Sequential Quadratic
Programming (SQP) and Random Search (RS) methods are used for optimization,
whereas Missile DATCOM, which is a semi-empirical aerodynamic analysis tool,
is used to calculate aerodynamic coefficients of missile configurations. As the first
part of the work, capabilities and limitations of SQP and RS optimization methods
are investigated on a complex test function. In addition to that, a validation for
aerodynamic analysis tool is done. Then, using reverse engineering approach,
aerodynamic performance parameters of NASA Tandem Control Missile
(TCM) configuration are defined as objectives to the developed design
optimization method with the aim of reaching TCM configuration at the end of
aerodynamic performance optimization process. Geometric properties and
aerodynamic performance parameters of the optimum configurations obtained
using the developed design method are compared to the ones of the TCM
configuration. Moreover, an optimization case study for a generic air-to-ground
missile is carried out. It is concluded that the developed design optimization
v

method is able to determine missile external configurations for pre-defined

aerodynamic performance parameters.

Keywords: Missile Aerodynamics, External Geometry Optimization, Sequential

Quadratic Programming, Random Search Optimization

vi

FZE DI GEOMETRLERNN AERODNAMK AIDAN EN

YLETRLMES

Arslan, Kvan
Yksek Lisans, Havaclk ve Uzay Mhendislii Blm
Tez Yneticisi: Prof. Dr. Serkan zgen
Eyll 2014, 85 Sayfa

Bu tezde, fze ve roketlerin kavramsal tasarm srecinde kullanlabilecek

aerodinamik performans en iyiletirmesi yapan bir tasarm metodu gelitirilmitir.
Bu tasarm metodunda fze konfigrasyonlarnn aerodinamik katsaylarn
hesaplanmasnda yar ampirik bir aerodinamik analiz yazlm olan Missile
DATCOM, tasarm en iyiletirmesinde ise ardk ikinci derece programlama ve
rastgele tarama yntemleri kullanlmtr. alma kapsamnda ilk olarak,
kullanlan en iyileme yntemleri optimum deerleri bilinen bir test fonksiyonu ile
dorulanmtr. Bunun yan sra, kullanlan aerodinamik analiz arac iin bir
dorulama almas yaplmtr. Ardndan literatrde bulunan NASA Tandem
Control Missile (TCM) konfigrasyonuna

ait

aerodinamik

performans

parametreleri tersine mhendislik anlay ile, gelitirilen tasarm aracna hedef

olarak tanmlanm ve tasarm aracnn TCM konfigrasyonuna ulamas
amalanmtr. Tasarm aracndan elde edilen konfigrasyonun geometrik
zellikleri ve aerodinamik performans parametreleri TCM konfigrasyonu ile
karlatrlmtr. Bu almaya ek olarak jenerik bir havadan karaya fze
konfigrasyonu iin aerodinamik tasarm optimizasyonu uygulamas yaplmtr.
vii

Sonu olarak, gelitirilen tasarm metodunun nceden tanmlanan aerodinamik

performans parametreleri iin fze d geometrilerini belirleyebildii grlmtr.
Anahtar Kelimeler: Fze Aerodinamii, D Geometri En yiletirmesi, Ardk
kinci Derece Programlama, Rastgele Arama ile En yileme

viii

To My Family

ix

ACKNOWLEDGEMENTS

I would like to express my deepest gratitude to my supervisor Prof. Dr. Serkan

zgen for his support, guidance and advice throughout the thesis.
I would like to thank my department manager Mr. Ali Akgl for his guidance,
advice and criticism during this study. I also would like to thank my colleagues in
Aerodynamic Design and Analysis Department of ROKETSAN for all their help
and support during the thesis.
I am very thankful to my parents Mrs. Semra Arslan, Mr. Kl Arslan, my sister
Ms. Cansu Arslan and my brother Mr. Ahmet Emre Arslan for their help,
motivation and endless love.
I want to express my best wishes to Mr. Hayri Yiit Akargn and Mr. Mehmet
Akgl for their friendship and support during this study.

TABLE OF CONTENTS

ABSTRACT ............................................................................................................v
Z ......................................................................................................................... vii
ACKNOWLEDGEMENTS ...................................................................................x
TABLE OF CONTENTS ..................................................................................... xi
LIST OF FIGURES ........................................................................................... xiii
LIST OF TABLES ............................................................................................. xvi
CHAPTERS
1. INTRODUCTION ..........................................................................................1
1.1 Aerodynamic Forces and Moments Acting on a Missile ......................... 4
1.2 Classification of Missiles ......................................................................... 5
1.3 Missile Aerodynamic Design Parameters ................................................ 8
1.3.1 Lift-to-Drag Ratio (L/D) ................................................................. 8
1.3.2 Stability ........................................................................................... 9
1.3.3 Maneuverability ............................................................................ 10
1.3.4 Control Effectiveness .................................................................... 11
1.4 Literature Survey .................................................................................... 12
1.5 Aim of the Thesis ................................................................................... 13
1.6 Organization of the Thesis ..................................................................... 14
2.

METHODOLOGY .......................................................................................17
2.1 Mathematical Optimization .................................................................... 17
2.2 Design Optimization............................................................................... 18
2.2.1 Random Search Method................................................................ 18
2.2.2 Sequential Quadratic Programming (SQP) Method ..................... 22
2.3 Aerodynamic Analysis Methodology ..................................................... 25
2.4 Aerodynamic Optimization Procedure ................................................... 28

3.

VERIFICATION OF THE DESIGN OPTIMIZATION METHOD ......33

3.1 Goldstein and Price Function ................................................................. 33
3.2 Verification of the Sequential Quadratic Programming (SQP) Method 35
3.3 Verification of the Random Search Method ........................................... 40
3.4 Validation of the Aerodynamic Analysis Method .................................. 43
xi

4.

MISSILE AERODYNAMIC OPTIMIZATION STUDIES .................... 47

4.1 Aerodynamic Shape Optimization for NASA Tandem Control Missile 48
4.1.1 NASA Tandem Control Missile (TCM) Configuration ................ 48
4.1.2 Aerodynamic Analyses of TCM ................................................... 48
4.1.3 Missile Design Optimization Study for TCM ............................... 51
4.1.3.1 Objectives, Constraints and Variables of Optimization ......
Study .............................................................................. 51
4.1.3.2 Optimization using DONLP2 .......................................... 53
4.1.3.3 Optimization using ACRS .............................................. 62
4.2 Aerodynamic Shape Optimization for a Generic Air-to-Ground Missile
(AGM) .................................................................................................... 72
4.2.1 Missile Design Optimization for AGM......................................... 72
4.2.1.1 Objectives, Constraints and Variables of Optimization ......
Study .............................................................................. 72
4.2.1.2 Optimization using DONLP2 .......................................... 75
4.2.1.3 Optimization using ACRS............................................... 77
4.2.1.4 Discussion of Results ...................................................... 80

5. CONCLUSIONS AND FUTURE WORK ................................................ 81

REFERENCES .................................................................................................... 83

xii

LIST OF FIGURES

FIGURES
Figure 1.1 Iterative Missile Design Process [4] ...................................................... 3
Figure 1.2 Forces and Moments Acting on a Missile ............................................. 5
Figure 1.3 Control Alternatives for Missiles........................................................... 7
Figure 1.4 Static Stability........................................................................................ 9
Figure 1.5 Statically Stable Missile Configuration [4] ........................................ 10
Figure 2.1 Conventional Wing-body Missile Configuration and Coordinate
System ................................................................................................................... 25
Figure 2.2 Variation of Interference Factors with r/sw or r/st [5] .......................... 27
Figure 2.3 Flowchart for Design Optimization with Sequential Quadratic
Programming Method ........................................................................................... 29
Figure 2.4 Flowchart for Design Optimization with Random Search Method ..... 31
Figure 3.1 Goldstein and Price Function............................................................... 34
Figure 3.2 Global Minimum Search with SQP Method, Successful .................... 36
Figure 3.3 Global Minimum Search with SQP Method, Converged to Local
Minimum ............................................................................................................... 37
Figure 3.4 Global Minimum Search with SQP Method, Converged to Local
Minimum ............................................................................................................... 38
Figure 3.5 Change of Objective Function Values with Iterations for 15
Optimization Runs ................................................................................................ 39
Figure 3.6 Evaluated Points During Global Minimum Search with Random
Search Algorithm .................................................................................................. 40
Figure 3.7 Evaluation of Iterations During Optimization ..................................... 41
Figure 3.8 Change of Objective Function Values with Iterations for Best and
Worst Solutions ..................................................................................................... 42
Figure 3.9 Lift Coefficient Obtained from MD, CFD and Experiment for TCM
Configuration (Mach = 1.75) ................................................................................ 43
xiii

Figure 3.10 Drag Coefficient Obtained from MD, CFD and Experiment for TCM
Configuration (Mach = 1.75) ................................................................................. 44
Figure 3.11 Pitching Moment Coefficient Obtained from MD, CFD and
Experiment for TCM Configuration (Mach = 1.75) .............................................. 44
Figure 4.1 TCM B1T4C4 Configuration (Dimensions in cm) .............................. 48
Figure 4.2 Normal Force Coefficient versus Angle of Attack .............................. 49
Figure 4.3 Axial Force Coefficient versus Angle of Attack .................................. 50
Figure 4.4 Missile Center of Pressure versus Angle of Attack ............................. 50
Figure 4.5 Canard Placement Area (red) and Random Possible Canard Geometries
(dashed lines) ......................................................................................................... 52
Figure 4.6 Optimum Configuration Geometry (top), TCM B1T4C4 (bottom) ..... 55
Figure 4.7 Change of Leading Edge Position throughout Optimization ............... 56
Figure 4.8 Change of Root Chord throughout Optimization ................................. 57
Figure 4.9 Change of Taper Ratio throughout Optimization................................. 57
Figure 4.10 Change of Semi-span throughout Optimization................................. 58
Figure 4.11 Change of Normal Force Coefficient (M=1.75, =16) throughout
Optimization .......................................................................................................... 58
Figure 4.12 Change of Center of Pressure Location (M=1.75, =4) throughout
Optimization .......................................................................................................... 59
Figure 4.13 Change of Center of Pressure Location (M=1.75, =16) throughout
Optimization .......................................................................................................... 59
Figure 4.14 Change of Axial Force Coefficient (M=1.75, =0) throughout
Optimization .......................................................................................................... 60
Figure 4.15 Change of Objective Function throughout Optimization ................... 60
Figure 4.16 Change of Objective Function Values with Iterations ....................... 61
Figure 4.17 Optimum Configuration Geometry (top), TCM B1T4C4 (bottom) .. 64
Figure 4.18 Change of Leading Edge Position throughout Optimization ............. 65
Figure 4.19 Change of Root Chord throughout Optimization ............................... 66
Figure 4.20 Change of Taper Ratio throughout Optimization .............................. 66
Figure 4.21 Change of Semi-span throughout Optimization................................. 67

xiv

Figure 4.22 Change of Normal Force Coefficient (M=1.75, =16) throughout

Optimization.......................................................................................................... 67
Figure 4.23 Change of Center of Pressure Location (M=1.75, =4) throughout
Optimization.......................................................................................................... 68
Figure 4.24 Change of Center of Pressure Location (M=1.75, =16) throughout
Optimization.......................................................................................................... 68
Figure 4.25 Change of Axial Force Coefficient (M=1.75, =0) throughout
Optimization.......................................................................................................... 69
Figure 4.26 Change of Objective Function throughout Optimization .................. 69
Figure 4.27 Change of Objective Function Values with Iterations ....................... 70
Figure 4.28 Change of Objective Function Values with Iterations (y axis is
narrowed) .............................................................................................................. 71

xv

LIST OF TABLES

TABLES
Table 1.1 Classification of Missiles Based on Platform and Target Location [4] .. 6
Table 1.2 Advantages and Disadvantages of Missile Control Types [4] ................ 7
Table 2.1 Indices in Equations (2.33) - (2.35) ....................................................... 26
Table 3.1 Variable and Objective Function Values for Best and Worst Solutions
............................................................................................................................... 39
Table 3.2 Variable and Objective Function Values for Best and Worst Solutions 42
Table 4.1 Flight Conditions ................................................................................... 49
Table 4.2 Objective Aerodynamic Performance Parameters ................................. 51
Table 4.3 Missile Geometry Variables and Constraints ........................................ 51
Table 4.4 Change of Configuration Geometry along Optimization ..................... 54
Table 4.5 Optimum and TCM B1T4C4 Configuration (Objective) Aerodynamic
Performance Parameters ........................................................................................ 55
Table 4.6 Optimum and TCM B1T4C4 Missile Geometric Proportions .............. 56
Table 4.7 Aerodynamic Performance Parameters of Best and Worst
Configurations ....................................................................................................... 61
Table 4.8 Change of Configuration Geometry along Optimization Run .............. 63
Table 4.9 Optimum and TCM B1T4C4 Configuration (Objective) Aerodynamic
Performance Parameters ........................................................................................ 64
Table 4.10 Optimum and TCM B1T4C4 Missile Geometric Proportions ............ 65
Table 4.11 Aerodynamic Performance Parameters of Best and Worst
Configurations ....................................................................................................... 70
Table 4.12 Missile Geometry Variables and Limits .............................................. 72
Table 4.13 Aerodynamic Performance Parameters for Sample AGM .................. 73
Table 4.14 Flight Regime for AGM Optimization Case Study ............................. 73
Table 4.15 Optimum AGM and Objective Aerodynamic Performance Parameters
............................................................................................................................... 76
xvi

Table 4.16 Optimum AGM Geometric Proportions ............................................. 76

Table 4.17 Change of AGM Configuration Geometry along Optimization Run . 77
Table 4.18 Optimum AGM and Objective Aerodynamic Performance Parameters
............................................................................................................................... 78
Table 4.19 Optimum AGM Geometric Proportions ............................................. 78
Table 4.20 Change of AGM Configuration Geometry along Optimization Run . 79

xvii

xviii

CHAPTER 1

1. INTRODUCTION

An engineering design optimization problem can be defined as obtaining the best

possible engineering system under given system constraints, performance
requirements and time limits. Designers generally need a tool to overcome the
design optimization problem, where an analysis/simulator is coupled with an
optimization method [1]. The outcomes of the optimization problem are defined
as objective functions which may represent structural, aerodynamic or chemical
performance parameters or properties. Objectives of the problem are functions of
design variables that shape the properties of the engineering system [1].

In missile aerodynamic design, it is necessary to investigate a large design space.

Complexity of a design space is dependent on the number of input variables.
Missile diameter, nose, body and aft body length, number of fin sets, number of
fins for each fin set, size and shape of each fin and fin cross-section can be given
as some examples to the design variables. Geometric limitations are imposed to
the design problem when launch platform compatibility, cost issues and other
subsystem integrations are taken into account. Aerodynamic performance
parameters such as maneuverability, control effectiveness, static and dynamic
stability, lift-to-drag ratio are objectives for the design problem. Accordingly, in
missile aerodynamic design optimization problems, there exist numerous
parameters, variables and constraints that make the problem difficult to solve in
terms of time and computational effort. In addition to that, performance
parameters obtained as outcomes of the design process generally conflict with
each other. Consequently, it is difficult to obtain the best possible solution unless
the designer has enough computational resource for searching the whole design
1

space or he/she is using mathematical optimization methods together with

engineering analysis/simulation tools. In this thesis, it is aimed to develop a
design optimization method that can be used to overcome difficulties of missile
aerodynamic design problem and reduce time spent in conceptual design phase of
missiles.

Gradient-based and stochastic optimization methods are widely used in

engineering design optimization problems. Gradient-based methods determine the
maxima or minima of the objective functions of the design problem by using
gradient information of the function. Newtons method, steepest descent method
and Sequential Quadratic Programming (SQP) are some examples to gradientbased optimization methods and are widely used in the literature for engineering
problems such as aerodynamic or structural design optimization. A detailed
description of SQP method is given in Chapter 2. These optimization methods are
generally regarded as fast and demand low computational resources. Nevertheless,
they are sometimes called local optimizers since they tend to converge to the local
minimum of the design space and may produce biased results as they require an
initial point to start the optimization process. In addition to that, gradient-based
methods may end up with infeasible solutions since engineering optimization
problems are generally highly nonlinear and discontinuous in nature. Accordingly,
stochastic optimization methods that do not require gradient information of the
problem are employed in design optimization problems to overcome negative
effects of gradient-based methods. Random Search (RS) methods, Genetic
Algorithms (GA) and Simulated Annealing (SA) are examples of stochastic
optimization methods [2]. A detailed description of an RS method can be found in
Chapter 2. Stochastic methods are more successful for determining global
minimum, whereas they generally require more computational resources in
contrast to the gradient-based methods [3].

In conceptual design of missiles, an iterative process shown in Figure 1.1 is

followed. A baseline configuration that has similar mission requirements and
propulsion system is selected as a starting point. Design variables that affect
2

aerodynamics, propulsion, weight and flight trajectory of the baseline

configuration are changed iteratively to meet the mission requirements established
for the design process [4]. Here, as aerodynamics has priority over other design
characteristics, defining a configuration that is optimized in terms of aerodynamic
performance reduces the overall number of resizing iterations and decreases the
time spent on the design process.

Figure 1.1 Iterative Missile Design Process [4]

Aerodynamic design of missile/rocket configurations are based on the

investigation of alternative external shapes [4]. Missile aerodynamic performance
parameters are generally similar to parameters that are used for aircraft design.
3

Since missiles/rockets are uninhabited and expendable, elevated levels of speed,

altitude and maneuvering accelerations can be considered in the design process.
However, these elevated levels come together with some aerodynamic problems
such as nonlinearity of aerodynamic coefficients at high angles of attack and very
high wing loadings in these elevated flight envelopes [5]. As missiles are
uninhabited, it is possible to use extensive roll motion in flight that results in
additional dynamic stability considerations. Guidance systems that compensate
pilot input are necessary for missiles and some problems of stability and control
that are not encountered in aircraft design are introduced with these systems.
Missiles/rockets are generally slender, therefore slender-body theory is sufficient
for understanding aerodynamic characteristics of missile configurations [5].

1.1

Aerodynamic Forces and Moments Acting on a Missile

Missile aerodynamics is considered under three aerodynamic forces and three

moments for a six degree-of-freedom analysis and body fixed coordinate system.
In Figure 1.2, directions of these forces and moments in coefficient form are given
[6]. Note that, forces divided by free stream dynamic pressure ( ) and reference
area (

gives force coefficients (e.g.

) whereas moment coefficients (e.g.

) are calculated by dividing the moments with free stream dynamic pressure,
reference area and reference length (

. Here, reference length is generally

taken as missile diameter and reference area is cross-sectional area of the missile
body.

(1.1)

(1.2)

Figure 1.2 Forces and Moments Acting on a Missile

Force coefficients for a missile are shown in Figure 1.2 as CA, CN, and CY. Axial
force coefficient, CA, is positive in negative x direction whereas side force
coefficient, CY, is positive in the y direction and normal force coefficient, CN, is
positive in the negative z direction. In Figure 1.2, moment coefficients are given
as Cl, Cm, and Cn. Rolling moment, Cl, are about x-axis, pitching moment
coefficient, Cm, is about y-axis and yawing moment coefficient, Cn, is about zaxis, shown in their positive senses in the figure.

1.2

Classification of Missiles

Missiles are classified depending on their launch platform and target location,
mission and target type, propulsion system, design drivers, type of control etc.
[4,5]. An example classification based on launch platform and target locations for
state-of-the-art missiles is given in Table 1.1 [4].

Table 1.1 Classification of Missiles Based on Platform and Target Location [4]

Platform and Target

Location

Air to Air (ATA)

Air to Surface (ATS)

Surface to Surface (STS)

Surface to Air (STA)

Range / Mission

Example

Short Range ATA

AIM-9

Medium Range ATA

AIM-120

Long Range ATA

AIM-54

Short Range ATS

AGM-65

Antiradar ATS

AGM-88

Medium Range ATS

Apache

Antitank ATS

AGM-114

Long Range ATS

AGM-86

Long Range STS

BGM-109

Long Range Anti-armor STS

MGM-140

Man-portable STS

Javelin

Short Range STA

FIM-92

Medium Range STA

MIM-23

Long Range STA

MIM-104

Missile Defense

PAC-3

Another classification can be made based on type of control. Apart from

unconventional control systems such as thrust vector or reaction jet control,
missiles are controlled by deflecting aerodynamic lifting surfaces [4].
Accordingly, aerodynamic characteristics and the overall performance of a missile
are hugely affected by control surface location. Mainly, missiles are classified as
tail controlled, canard controlled and wing controlled (Figure 1.3) [4].

Figure 1.3 Control Alternatives for Missiles

Each type of control alternative has advantages and disadvantages that need to be
considered in missile design. These advantages and disadvantages are summarized
in Table 1.2 [4].

Table 1.2 Advantages and Disadvantages of Missile Control Types [4]

Control
Type

Advantages

Tail
Control

Canard
Control

Efficient
packaging
Low actuator
torque
Low induced
rolling moment
Efficient at high
angles of attack
Efficient
packaging
Simplified
manufacturing
Increased lift at
low angles of
attack

Disadvantages

Maverick AGM-65 [4]

Wing
Control

Fast response
Maneuverability
at low angles of
attack
Small trim angle

Examples

Decreased lift
at low angles
of attack

Stall at high
angles of
attack
High induced
roll

Poor
packaging
High hinge
moments
Large wing
size
Large
induced roll

AMRAAM AIM-120
[4]

AIM-9L [4]

Gopher SA-13 [4]

Sparrow AIM-7 [4]

HARM AGM-88 [4]

1.3

Missile Aerodynamic Design Parameters

Lift-to-drag ratio, stability, maneuverability and control effectiveness are the main
design parameters for missile aerodynamic design problems [7].

1.3.1 Lift-to-Drag Ratio (L/D)

Lift-to-drag ratio, also known as aerodynamic efficiency, is one of most important
parameters in missile design. It has a significant impact on range and also affects
maneuverability [4]. It is defined as the ratio of the total lift to the total drag of a
missile. It can also be written in coefficient form. In coefficient form, relation of
lift and drag to the axial and normal forces can be written as;

(1.3)

In these equations,

represents the missile angle of attack. Range of a missile is a

function of lift-to-drag ratio according to the Breguet range equation [4];

) (

(1.4)

where cruise velocity is denoted as , specific impulse denoted as

and weight

of the missile before launch and weight of fuel are denoted as

and

respectively.

1.3.2

Stability

Stability of a missile is considered in terms of static stability and dynamic

stability. A statically stable missile can produce some amount of pitching moment
with increasing angle of attack in opposing manner [4]. In other words, a statically
stable missile weathercocks to flow direction when flight conditions are changed.
In order to hold this aerodynamic characteristic, sign of the

(pitch stiffness

derivative) must be negative as shown in Figure 1.4 [8].

(1.5)

Figure 1.4 Static Stability

Figure 1.5 Statically Stable Missile Configuration [4]

Static stability is generally quantified with the static margin, distance between
center of pressure (
diameter (

and center of gravity (

divided by the missile

[4];

(1.6)

Negative, zero and positive values of static margin represents stable, marginally
stable and unstable missile configurations, respectively. A smaller negative static
margin means a stable missile that can be trimmed (

at high angles of

attack, accordingly is able to produce high normal force and has increased
maneuverability.

1.3.3 Maneuverability
Capability of a missile to change speed and direction in a given time is termed as
maneuverability [9]. Load factor is a measure of this aerodynamic performance
parameter and defined as the ratio of the normal acceleration due to aerodynamic
force to the gravitational acceleration [8].

where

10

(1.7)

(1.8)

At a given speed, a missile with a higher load factor is able to achieve higher
normal force; accordingly, it is able to maneuver tighter (low turn radius) and
faster (high turn rate) [10].

1.3.4

Control Effectiveness

Control effectiveness of a missile is defined as the ratio of the control surface

deflection at trim condition to the angle of attack [8].

(1.9)

In missile design, generally a control effectiveness value that is less than 1 is

desired [7]. A missile with smaller control effectiveness can be trimmed at a
higher angle of attack, accordingly has better control power and maneuverability
[8].

When missile aerodynamic design parameters are considered in the pitch plane,
aerodynamic coefficients such as axial force coefficient (
coefficient (

), pitching moment coefficient (

), normal force

) and flight conditions such as

velocity (Mach number), angle of attack ( ), control surface deflection ( ) are the
main factors that affect the aerodynamics of a missile.

11

1.4

Literature Survey

In literature, some studies on missile shape optimization are available. These

studies are conducted for different design objectives, using both stochastic and
gradient-based methods for optimization.

Genetic Algorithms (GA) and

Simulated Annealing (SA) methods are preferred in most of the studies. Using
pareto GAs, Anderson, Burkhalter and Jenkins [11] investigated missile
aerodynamic shape optimization. An aerodynamic prediction code is coupled with
a GA optimizer. A number of design variables concerning the missile shape and
aerodynamic performance objectives are included in his study. More recently,
Tanl [12] developed an external configuration design tool that uses GA as
optimization method. The method proposed is able to size missiles depending on
the flight performance objectives. Karako [13] worked on missile external
configuration optimization using GA. Main focus for his work is maximizing
flight performance and minimizing radar cross-section of missiles. Yong et al.
[14] studied aerodynamic shape optimization for canard controlled missiles for
maximizing flight range. A GA optimizer is used together with a 3 degree-offreedom flight simulation code and an aerodynamic analysis tool. Tekinalp and
Bingl [15] developed a missile trajectory optimization method. SA optimization
method is employed as the optimizer. A 2 degree-of-freedom flight simulation
model and an aerodynamic solver are used together in the proposed method.
ztrk [2] worked on multiobjective design optimization of rockets and missiles.
Objectives on range, flight time, hit angle and velocity, aerodynamic coefficients
are considered together. A continuous SA optimization algorithm called Hideand-Seek is used. In addition to these stochastic design optimization methods,
there exists the work of Tanrkulu and Ercan [16] that uses gradient-based
optimization. They proposed a method to be used for external configuration
optimization of unguided missiles at the conceptual design phase. A gradientbased optimization algorithm is used to determine optimal configurations.
Objectives of the optimization are defined considering flight performance of the
configurations.

12

As discussed above, most of the work on missile shape optimization is based on

genetic algorithms and simulated annealing method. Popularity of these methods
can be recognized in different fields of science and technology. As an alternative
to these stochastic methods, Random Search (RS) approaches can be considered
in optimization. Especially for functions that are discontinuous and non-smooth,
RS method is easy to implement. There exist numerous applications that show RS
approaches are quite robust even though convergence to the global optimum is not
guaranteed [17]. Computational resource requirements for RS are also seen as
acceptable [17]. Silveira Jr. et al. [18] suggested that Adaptively Random Search
(ARS), which is a type of RS, is superior to the other stochastic methods such as
GAs. Tsoulos and Ragaris [19] compared some RS algorithms with SA on
different test functions and showed that RS yields faster solutions in most of the
cases. Recent applications of RS can be found in the literature [17,18,20].
Apart from stochastic approaches, gradient-based methods are employed in design
optimization problems. One of the most recently developed methods is Sequential
Quadratic Programming (SQP) and it is considered as one of the most suitable for
gradient-based optimization [21]. SQP methods are successfully employed for
aerodynamic shape optimization in the literature [22,23].

1.5

Aim of the Thesis

The aim of this thesis is to develop a design optimization method that can be used
for the conceptual design phase of missile aerodynamic design. The methodology
is based on shaping external geometry of the missile under given constraints in an
automated manner. It aims to determine the best possible configuration in terms of
pre-defined aerodynamic design objectives inside the design space. Most
importantly, the proposed design optimization method reduces the time spent for
missile external configuration design in conceptual design process when
compared to the conventional design approach.

13

In this work, a gradient-based and a random search optimization algorithm

coupled with a semi-empiric missile aerodynamics solver are used to determine
the optimum aerodynamic external configurations. In the literature, most of the
work related to the missile design optimization use genetic algorithms as
optimization method. In global optimization problems, random search methods
show sufficiently successful results with acceptable computational effort.
Accordingly, a random search optimization method is employed in this thesis to
investigate its capabilities in missile external geometry optimization. In addition
to that, a gradient-based method, Sequential Quadratic Programming (SQP) is also
coupled with the same aerodynamic solver. SQP is known as one of the best
performing and state-of-the-art gradient-based optimization methods. Two
different design optimization procedures are developed using two different
optimization approaches and compared to each other in order to understand their
applicability and performances in missile aerodynamic design. FORTRAN
programming language is used for the implementation of these methods.

In order to analyze each configuration during optimization, an aerodynamic solver

code is used. Since obtaining aerodynamically the best performing external
geometry in minimum time is a driving factor to develop this design optimization
methodology, aerodynamic coefficient fast prediction tool Missile DATCOM is
used for aerodynamic analyses.

1.6

Organization of the Thesis

This thesis consists of 5 chapters. Background information on missile conceptual

design and missile aerodynamics are given in Chapter 1. Some methods that can
be applied for missile design optimization together with existing works in
literature are also discussed. Aim of the thesis is also given in Chapter 1.

Chapter 2 gives detailed information on methods that are used in missile

aerodynamic optimization procedure developed in this work. SQP and RS based
14

optimization algorithms employed in this work are discussed together with

aerodynamic analysis methodology.

Verification of the developed optimization methodology is investigated in Chapter

3. A global optimization test function is used for this purpose. Results are
discussed and compared.

In Chapter 4, two applications of missile aerodynamic optimization are presented.

For the first application, a missile configuration for wind tunnel test cases, NASA
Tandem Controlled Missile (TCM), is used as baseline. With reverse-engineering
approach, canard geometry of TCM is optimized. In the second application,
optimization of a generic Air-to-Ground Missile (AGM) is discussed. For these
missile aerodynamic optimization applications, results are given separately for
SQP and RS optimization methods employed in the aerodynamic optimization
procedure.

Chapter 5 finalizes this thesis with concluding remarks and recommendations for
future studies.

15

16

CHAPTER 2

METHODOLOGY

Methodology for the development of the design optimization method is described

in this chapter. Governing equations of employed optimization algorithms are
given first. After the description of governing equations, aerodynamic analysis
method used for performance evaluation of missile configurations is explained.
Finally, integration of optimization methods with aerodynamic analysis tool to
form the design optimization method is discussed.

2.1

Mathematical Optimization

In general terms, mathematical optimization is defined as [24];

(2.1)

subject to the constraints:

(
(2.2)

where ( ,

and

are scalar functions of the real column vector x.

17

In

represents the (design) variables, whereas (

is

called as the objective function. Inequality constraints and equality constraints are
represented by

2.2

and

( , respectively [24].

Design Optimization

Design optimization is defined as a process in which design variables are updated

automatically in order to achieve a better design output [25]. As discussed in the
previous chapter, there exist numerous optimization methods that can be used for
design optimization purposes. In this study, Random Search (RS) and Sequential
Quadratic Programming (SQP) methods are employed for the development of the
design optimization method for missile external configurations.

2.2.1 Random Search Method

Random Search Method is a stochastic optimization method. Design variables
throughout the optimization process are selected randomly; therefore, this method
does not require gradient information of the objective function [26]. Since it is not
possible to define an exact objective function for the aerodynamic design problem
solved in this work, using random search algorithm is advantageous.

In this study, a random search algorithm named as Derivative-Free Adaptively

Controlled Random Search (ACRS) is employed in the developed design
optimization method. Well known Prices algorithm is the base of ACRS;
however, it improves the Prices algorithm by gathering as much as available
information from the objective function evaluations during the optimization
process [26]. In order the determine minimum of the objective function, ACRS
follows the steps given below [26]:

18

Given a positive integer

such that

, where n is number of

variables;

Step 0: Initial set determination;

Set

Points

(2.3)

are chosen randomly over the optimization domain and

objective function f is evaluated at each point

Step 1: Determine the maximum and minimum objective function values

(

) and corresponding points (

) for the current set;

(2.4)

(2.5)

Stop when convergence criterion is satisfied.

Step 2: Random
the n points

points

are chosen over

Centroid

is determined from;

where

19

(2.6)

of

(2.7)

(2.8)

where

(2.9)

is a sufficiently large positive constant (e.g.

).

Step 3: A weighted reflection is performed to determine the trial point ; let,

(2.10)

then take

with

(2.11)

given as;
(

(
(

{
Compute (
Step 4: If (

if

; if not go to Step 2.
then take;

20

)
(2.12)

(2.13)

Set

and go to Step 2.
(

Step 5: If

then take;

Set
Step 6: If

and go to Step 1.
(

then;

and select the subset

values of

(2.14)

(2.15)

points in corresponding to the smallest

of

To use a quadratic model of the objective function, diagonal matrix

, the vector

and the scalar

are determined such that

(2.16)

Step 7: If the diagonal entries of

are not all positive, then take

set
Step 8: If

(2.17)

and go to Step 1.
is positive definite, let

(2.18)

21

If

or ( )

then take

(2.19)

else take

where

{ }

(2.20)

determined from
(

Set

(2.21)

and go to Step 1.

2.2.2 Sequential Quadratic Programming (SQP) Method

Sequential Quadratic Programming is a gradient-based optimization method.
Along the optimization process, design variables are determined according to the
gradient information of the objective function. Nevertheless, due to nature of the
aerodynamic design problem discussed in this work, gradient information of the
objective function is non-existent. Therefore, gradients need to be calculated
numerically by using objective function directly.

In addition to ACRS method discussed previously, aerodynamic shape

optimization is also done by using SQP method. The SQP algorithm employed in
the developed design optimization method, DONLP2, is capable of locating a
local minimum inside the design space quickly [27]. However, this quickness
comes with an expense; locating the global minimum is very much dependent on
the initial design point, especially in a complex design space containing numerous
local minima. As an SQP algorithm, DONLP2 requires gradients of the objective
function. Accordingly, gradients are calculated numerically using a sixth order
22

approximation computing a Richardson extrapolation of three symmetric

differences [27]. In order the determine minimum of the objective function,
DONLP2 follows the methodology given below [21,27]:

For a quadratic problem that is defined as;

Find

that minimizes the quadratic objective function ;

(2.22)

subject to the linear equality constraints defined as;

(2.23)

where
(2.24)

or
The Lagrange function, , for the problem given in (2.22) is defined as

where

(2.25)

is the Lagrange multiplier associated with the th equality constraint.

The Kuhn-Tucker necessary conditions can be stated as;

(2.26)

and
23

(2.27)

Equation (2.22) can be identified as a standard form of the original optimization

equation without inequality constraints;

(2.28)

subject to:
(

(2.29)

Accordingly, equation (2.22) can be solved iteratively for the solution of the
problem given in (2.28). Addition of inequality constraints to the general
optimization problem defined in (2.28), following can be written:

Find

which minimizes
(2.30)

subject to
(2.31)

with Lagrange function given by

24

(2.32)

2.3

Aerodynamic Analysis Methodology

Aerodynamic coefficients and related aerodynamic performance parameters of a

design configuration need to be determined during the design optimization
process. Accordingly, Missile DATCOM aerodynamic coefficient estimation
software is employed in the developed optimization method. Missile DATCOM is
a semi-empirical tool that predicts the aerodynamic coefficients for a wide variety
of missile configurations. It has adequate accuracy for preliminary design phase
and outputs the results very quickly. Therefore, it is convenient to use this
software for design optimization since numerous configurations need to be
evaluated in a short time in order to determine the most suitable one, considering
the objective performance parameters and geometric constraints.

Missile DATCOM is developed based on Component Build-Up (CBU) methods.

Aerodynamic force and moment coefficients of each component of the missile
together with their interactions are calculated using CBU methods. Total forces
and moments acting on a conventional missile configuration shown in Figure 2.1
is the summation of coefficients of each component that form the missile [28].

Figure 2.1 Conventional Wing-body Missile Configuration and Coordinate System

Normal force, pitching moment and axial force coefficients for a conventional
missile configuration can be described as follows [28]:
25

(2.33)

(2.34)

(2.35)

Indices in equations (2.33) - (2.35) are given in Table 2.1.

Table 2.1 Indices in Equations (2.33) - (2.35)

Body
(

For

Wing in presence of body

Body in presence of wing

Tail in presence of body

Body in presence of tail

Tail in presence of wing

= 0, equation (2.33) can be expanded as [28]:

(2.36)

(2.37)

(2.38)

(2.39)

26

(2.40)

K terms given in above equations are carry-over interference factors between

components. These interference factors are obtained from slender-body theory
that is developed for wing-body configurations [5]. Relation of interference
factors with body radius (r) and wing/tail semi-span (sw, st) is given in Figure 2.2.
2,00
1,80

Interference Factors

1,60
1,40
1,20
1,00

0,80
0,60

KW(B) or KT(B)

0,40

kW(B) or kT(B)

0,20

kB(W) or kB(T)
KB(W) or KB(T)

0,00
0,0

0,2

0,4

0,6

0,8

1,0

r/sw or r/sT
Figure 2.2 Variation of Interference Factors with r/sw or r/st [5]

Note that relation between interference factors given in Figure 2.2 is;

Aerodynamic derivatives for wing/tail fins such as

(2.41)

and

are determined

using theoretical or semi-empirical methods. Empirical methods are generally

employed to predict derivatives of body aerodynamic coefficients. Equivalent
27

angle of attack approach and vortex algorithms are also incorporated when
combining body and fin aerodynamic predictions for configuration synthesis [29].

2.4

Aerodynamic Optimization Procedure

Aerodynamic optimization process consists of many sub-problems such as

designation of a configuration, determining aerodynamic performance parameters
of the configuration, evaluation of the configuration according to design
objectives and finding a better configuration in terms of aerodynamic
performance. In this study, a design optimization method that is capable of
tackling these problems automatically is developed. As two different optimization
algorithms are used, flowcharts and descriptions for iterative steps for both are
given in the following discussion.

Design optimization method employing DONLP2 algorithm coupled with

aerodynamic analysis tool follows the procedure given in Figure 2.3.

28

Figure 2.3 Flowchart for Design Optimization with Sequential Quadratic Programming
Method

Steps of aerodynamic design optimization with DONLP2 algorithm can be

described as follows:

In order to start a design optimization, input files for baseline design

configuration,

geometric

constraints

and

objective

aerodynamic

performance parameters need to be supplied to the design tool. Since

29

DONLP2 uses a SQP method, it is also necessary to initialize the design

variables.

As the first step of design optimization, DONLP2 algorithm solves the

quadratic programming sub-problem formed by the objective function and
design constraints. During the solution process, it is necessary to supply
derivative information of the objective function. As this information is not
available analytically, it is computed numerically. Aerodynamic analysis
tool is called for evaluation of the objective function and its derivatives
using a sixth order approximation.

After the solution of the quadratic programming sub-problem, direction of

descent is determined.

Then, step size is calculated to determine new design variables along

minimization.

In the fourth step, a new set of design variables is determined using the
descent direction and step size information. Objective function value is
calculated with the aerodynamic analysis tool and stored together with the
configuration information.

Hessian matrix that consists of second-order partial derivatives of the

quadratic problem is updated.

Convergence criterion is checked. If it is not satisfied, quadratic problem is

solved again using the new design variables determined in step 4. This
procedure is repeated until convergence is achieved. External proportions
and aerodynamic performance parameters of the optimum design
configuration are provided as output.

Design optimization method employing ACRS algorithm coupled with

aerodynamic analysis tool follows the procedure given in Figure 2.4.

30

Figure 2.4 Flowchart for Design Optimization with Random Search Method

Steps of aerodynamic design optimization with ACRS algorithm can be described

as follows:

31

In order to start a design optimization, input files for baseline design

configuration,

geometric

constraints

and

objective

aerodynamic

performance parameters need to be supplied.

As the first step of design optimization, ACRS algorithm generates a set of

random configurations that lie inside the design space defined by
geometric constraints. Size of the set is proportional to the number of
design variables.

Aerodynamic coefficients for all of the configurations in the starting set

are calculated by using aerodynamic analysis tool. Accordingly, objective
function that represents the aerodynamic performance is calculated for all
of the configurations. Since each configuration has a unique objective
function value, they are sorted and configuration with maximum objective
value (fmax) is found.

After generation of starting set, ACRS generates a new random

configuration based on information obtained about the design space while
generating the starting set. Design parameters of this new configuration are
sent to the aerodynamic analysis tool and aerodynamic coefficients of the
configuration are determined. Using aerodynamic coefficients, objective
function value for this configuration (fp) is calculated.

Since fmax and fp is known, ACRS compares them in order to determine

whether this new configuration is better than the configuration with fmax or
not. If fp is less than fmax, new configuration is added to solution set
whereas configuration that has fmax is discarded from the set.

ACRS continues to generate random configurations, calculate fp and

update the configurations in the solution set upon convergence criteria is
satisfied. When generating these random configurations, information
gathered from previous iterations are taken into account to direct solutions
to the global minimum. This is achieved by use of a weighted centroid, a
weighted reflection and a quadratic model of the objective function.

Optimization process is stopped when the convergence criteria is met.

Accordingly

external

proportions

and

aerodynamic

parameters of the optimum design configuration are provided.

32

performance

CHAPTER 3

VERIFICATION OF THE DESIGN OPTIMIZATION

METHOD

A verification study on an optimization test function is done for the methods used
in this work. Goldstein and Price function is chosen as the test function since it is
accepted as a global optimization test function for optimization algorithms in the
literature [30]. Details about this function are given in the following section.
Moreover, aerodynamic analysis tool employed in developed optimization method
is also validated in this chapter.

3.1

Goldstein and Price Function

Goldstein and Price function is a two variable function that has multiple local
minima and defined as [26]:
(

( 9

( 8

3
(3.1)

3
48

36

subject to
(

33

(3.2)

Goldstein and Price function takes the global minimum value of ( )

( )

3 for

In Figure 3.1, Goldstein-Price function is shown. Since function values change

very rapidly depending on the variables, contours are drawn logarithmically. As
seen from Figure 3.1, local and global minima are located at an area where change
of ( ) is very limited, imposing a challenge to the determination of the global
minimum.

1E+06

800000

f(x,y)

f(x,y)

600000

400000

200000

1.0E+06
4.0E+05
1.6E+05
6.6E+04
2.6E+04
1.1E+04
4.3E+03
1.7E+03
7.0E+02
2.8E+02
1.1E+02
8.0E+01
4.6E+01
1.8E+01
7.4E+00
3.0E+00

2
1

0
-1

-1

-2 -2

Figure 3.1 Goldstein and Price Function

In order to determine the capabilities and limitations of the design optimization

methods used in this study, global minimum of the Goldstein-Price function is
searched. Two different approaches under consideration in this work are tested;
34

stochastic, random search algorithm ACRS and gradient-based Sequential

Quadratic Programming (SQP) algorithm DONLP2.

3.2

Verification of the Sequential Quadratic Programming (SQP) Method

Using the DONLP2 algorithm, global minimum point of the Goldstein and Price
function is searched. Since DONLP2 is based on Sequential Quadratic
Programming, a gradient-based optimization method, it is necessary to supply a
starting point in the design space to begin the optimization process. In the
verification study, a random starting point is selected as ( )

9 .

Starting from the initial point, optimization algorithm is able to find the global
minimum of the function in 15 iterations with 255 function evaluations as
( )

. In Figure 3.2, advance of the iterations as search marches

towards the global minimum is shown.

35

2
1.5

f(x,y)
1.0E+06
4.0E+05
1.6E+05
6.6E+04
2.6E+04
1.1E+04
4.3E+03
1.7E+03
7.0E+02
2.8E+02
1.1E+02
8.0E+01
4.6E+01
1.8E+01
7.4E+00
3.0E+00

0.5
0

-0.5
-1
-1.5
-2
-2

-1.5

-1

-0.5

0.5

1.5

X
Figure 3.2 Global Minimum Search with SQP Method, Successful

Another random starting point is selected as ( )

85

35 . Starting

from the initial point, optimization run converges to a local minimum of the
function in 16 iterations with 272 function evaluations as ( )
where objective function has a value of ( )

3 . In Figure 3.3, advance of

the iterations as search marches towards the local minimum is shown.

36

2
1.5

f(x,y)
1.0E+06
4.0E+05
1.6E+05
6.6E+04
2.6E+04
1.1E+04
4.3E+03
1.7E+03
7.0E+02
2.8E+02
1.1E+02
8.0E+01
4.6E+01
1.8E+01
7.4E+00
3.0E+00

0.5
0

-0.5
-1
-1.5
-2
-2

-1.5

-1

-0.5

0.5

1.5

X
Figure 3.3 Global Minimum Search with SQP Method, Converged to Local Minimum

One more random starting point is selected as ( )

( 94

6 . Starting

from there, optimization run converges to a local minimum of the function in 12

iterations with 197 function evaluations as ( )
function has a value of ( )

( 8

, where objective

84. In Figure 3.4, advance of the iterations as

search marches towards the local minimum is shown.

37

2
1.5

f(x,y)
1.0E+06
4.0E+05
1.6E+05
6.6E+04
2.6E+04
1.1E+04
4.3E+03
1.7E+03
7.0E+02
2.8E+02
1.1E+02
8.0E+01
4.6E+01
1.8E+01
7.4E+00
3.0E+00

0.5
0

-0.5
-1
-1.5
-2
-2

-1.5

-1

-0.5

0.5

1.5

X
Figure 3.4 Global Minimum Search with SQP Method, Converged to Local Minimum

It is known that, during a global minimum search, accuracy of SQP methods are
very dependent on the initial point and design space topology. Accordingly,
Goldstein-Price function is solved 15 times to investigate repeatability and
consistency of the results obtained using SQP algorithm namely DONLP2.
Random initial points are generated for each solution to see the effect of the
starting point on the optimization process. Change of objective function values
with iterations for best, worst and remaining runs are shown in Figure 3.5. Best
and worst solutions to the problem are given in Table 3.1.

38

10

10

BEST
ITER
WORST

F(x,y)

103

102

101

10

10

15

20

Iterations
Figure 3.5 Change of Objective Function Values with Iterations for 15 Optimization Runs

Table 3.1 Variable and Objective Function Values for Best and Worst Solutions

f(x,y) 3.0

-2.657E-09

-1.0

-7.994E-15

Worst

1.2

0.8

837.0

OBJECTIVE

0.0

-1.0

0.0

Best

Previous discussion and Table 3.1 prove that there are solutions to the GoldsteinPrice test problem that converge to global or local minima when DONLP2
algorithm is used. Therefore, this gradient-based SQP method should be used
carefully when the design space is complex.

39

3.3

Verification of the Random Search Method

Using the derivative-free Adaptively Controlled Random Search (ACRS)

algorithm, global minimum of the Goldstein-Price function is investigated. ACRS
algorithm is able to find the global minima of the function in 514 iterations as
( )

. In Figure 3.6, all of the points used by the optimization method

during global minimum search are shown. Change of these points with iterations
as search marches towards the global minimum is illustrated in Figure 3.7.

2
1.5

f(x,y)
1.0E+06
4.0E+05
1.6E+05
6.6E+04
2.6E+04
1.1E+04
4.3E+03
1.7E+03
7.0E+02
2.8E+02
1.1E+02
8.0E+01
4.6E+01
1.8E+01
7.4E+00
3.0E+00

0.5
0

-0.5
-1
-1.5
-2
-2

-1.5

-1

-0.5

0.5

1.5

X
Figure 3.6 Evaluated Points During Global Minimum Search with Random Search
Algorithm

40

Iteration 1

Iteration
1
1. Iterasyon
Iteration
Iteration
109109
109. Iterasyon
Iteration
514. Iterasyon
Iteration
514514

2
2
1.5
1.5
1
1

YY

0.5
0.5
0
0

-0.5
-0.5
-1
-1
-1.5
-1.5
-2
-2-2
-2

-1.5
-1.5

-1
-1

-0.5
-0.5

X0

0.5
0.5

1
1

1.5
1.5

2
2

Figure 3.7 Evaluation of Iterations During Optimization

Generally, stochastic methods such as random search method are fairly good at
locating global minimum in optimization problems. However, they generally
require more computational time and effort, compared to gradient-based methods.
Using random search optimization algorithm ACRS, Goldstein-Price function is
solved 15 times to investigate repeatability and consistency of results. Note that it
is not necessary to supply a starting point for optimization when using ACRS.
Change of objective function values with iterations for best and worst resulting
runs is shown in Figure 3.8. It can be seen that both best and worst run converge
to global minimum meaning that none of the 15 runs ended up in local minima.
Table 3.2 gives solutions for best and worst runs to the problem.

41

10

WORST
BEST

F(x,y)

10

10

10

10

100

200

300

400

500

Iterations
Figure 3.8 Change of Objective Function Values with Iterations for Best and Worst
Solutions

Table 3.2 Variable and Objective Function Values for Best and Worst Solutions

f(x,y)-3.0

Best

-9.982E-07

-1.0

8.350E-10

Worst

-4.130E-06

-1.0

2.173E-08

0.0

-1.0

0.0

OBJECTIVE

Previous discussion and Table 3.2 show that all of the 15 solutions to the
Goldstein-Price optimization test problem ended up in the global minimum of the
function, whereas SQP method discussed previously may get trapped in local
minima of the function as seen in Figure 3.3, Figure 3.4 and Figure 3.5.
Accordingly, it can be concluded that ACRS is more robust in locating global

42

minimum and avoiding local minima inside a complex design space, compared to
DONLP2.

3.4

Validation of the Aerodynamic Analysis Method

In this study, Missile DATCOM software is used to predict the aerodynamic

coefficients of missile configurations. Results obtained from Missile DATCOM
(MD) for NASA Tandem Control Missile (TCM) configuration are compared to
experimental and Computational Fluid Dynamics (CFD) results found in the
literature [31]. It is aimed to investigate the applicability and accuracy of the
aerodynamic analysis method employed for missile shape optimization. More
detail on TCM configuration can be found in Section 4.1.1.
Figure 3.9 - Figure 3.11 present lift, drag and pitching moment coefficients
calculated for the TCM configuration at Mach=1.75, respectively.

14
Experiment

12

CFD

10

MD

CL

6
4
2
0
-2
-4
-5

15

25

a []
Figure 3.9 Lift Coefficient Obtained from MD, CFD and Experiment for TCM
Configuration (Mach = 1.75)

43

35

8
Experiment

CFD

MD

CD

4
3
2
1
0
-5

15

a []

25

35

Figure 3.10 Drag Coefficient Obtained from MD, CFD and Experiment for TCM
Configuration (Mach = 1.75)

Cm

0
-2
-4
Experiment

-6

CFD
MD

-8
-5

15

25

35

a []
Figure 3.11 Pitching Moment Coefficient Obtained from MD, CFD and Experiment for
TCM Configuration (Mach = 1.75)

44

From Figure 3.9, Figure 3.10 and Figure 3.11, it can be said that MD is
sufficiently successful for aerodynamic coefficient prediction of conventional
missile configurations. Especially for angles of attack up to 10, results are very
good and almost coincide with experiment data. Predictions start to disagree with
CFD and experiment when the angle of attack is over 10. Even then, MD has an
acceptable error for conceptual and preliminary design optimization purposes.

45

46

CHAPTER 4

MISSILE AERODYNAMIC OPTIMIZATION STUDIES

In this part of the thesis, aerodynamic optimization of missile external

configurations is carried out to investigate the capabilities of the developed design
optimization methods. Two different optimization applications are demonstrated.
In the first section, NASA Tandem Control Missile (TCM) configuration is
considered as baseline. As a starting point, aerodynamic performance parameters
of the TCM configuration are determined using aerodynamic coefficient
prediction tool Missile DATCOM. Then, using the reverse engineering approach,
aerodynamic performance parameters of TCM configuration are given as
objectives to the design tool and it is aimed to obtain a configuration similar to the
original TCM configuration. In the second section, an optimization application for
a generic air-to-ground missile (AGM) configuration is demonstrated.
Aerodynamic performance parameters for the AGM are defined considering
typical mission requirements of such missiles. Each optimization application is
done using both optimization algorithms, ACRS and DONLP2. Optimum
configurations are discussed separately and compared to each other in order to
investigate the performances of the algorithms.

47

4.1

Aerodynamic Shape Optimization for NASA Tandem Control Missile

4.1.1 NASA Tandem Control Missile (TCM) Configuration

TCM is a supersonic wind-tunnel test case configuration. In this study, a version
of TCM that is known as TCM B1T4C4 in the literature is investigated. Geometry
of TCM B1T4C4 is given in Figure 4.1. This configuration has the same
geometric proportions for each 4 canard and tail fins and these fins are configured
in X shape. It has a cylindrical body with tangent ogive nose. Wind tunnel test
results and analyses done with fast-prediction aerodynamics methods for TCM are
available in the literature [32,33].

Moment Center

Figure 4.1 TCM B1T4C4 Configuration (Dimensions in cm)

4.1.2 Aerodynamic Analyses of TCM

Design optimization procedure employs Missile DATCOM to evaluate
aerodynamic coefficients of each configuration in optimization process.
Accordingly, aerodynamic analyses are done in flight conditions given in Table
4.1 to determine the aerodynamic performance parameters of the TCM
configuration. From these parameters, objectives for optimization study are
defined.

48

Table 4.1 Flight Conditions

1.75

Mach

-4 24

Angle of Attack

= 0

Sideslip Angle

Sea Level

Altitude

Change of aerodynamic coefficients versus angle of attack is given in Figure 4.2,

Figure 4.3 and Figure 4.4. Center of pressure is calculated from the moment
center of missile and non-dimensionalized with missile diameter. Positive values
reflect a center of pressure location forward of moment center whereas negative
values indicate aft of it.

Figure 4.2 Normal Force Coefficient versus Angle of Attack

49

Figure 4.3 Axial Force Coefficient versus Angle of Attack

Figure 4.4 Missile Center of Pressure versus Angle of Attack

50

4.1.3

Missile Design Optimization Study for TCM

4.1.3.1 Objectives, Constraints and Variables of Optimization Study

In missile external configuration optimization study done for TCM, aerodynamic
analysis results of TCM B1T4C4 are taken as objectives. Normal force coefficient
(CN) at 16 angle of attack, center of pressure (XCP/D) at 4 and 16 angles of
attack in addition to axial force coefficient (CA) at 0 angle of attack are chosen as
design objectives (Table 4.2). Note that these coefficients are calculated for Mach
number of 1.75 with zero sideslip angle and at sea level. Results obtained using
two different optimization algorithms are discussed in the following sections,
separately.

Table 4.2 Objective Aerodynamic Performance Parameters

CN (=16)
6.532

XCP/D (=4)
0.370

XCP/D (=16)
0.132

CA (=0)
0.517

Design variables in external configuration optimization are defined for canard

geometry. These variables are canard leading edge location with respect to the
nose tip, semi-span, root chord and taper ratio of canard. Upper and lower limits
for these variables and actual values for TCM B1T4C4 configuration are given in
Table 4.3.

Table 4.3 Missile Geometry Variables and Constraints

Variable
XLE [cm]
SSPAN [cm]
RCHORD [cm]
TR

Description
Canard leading edge
position
Canard semi-span
Canard root chord
Canard taper ratio
51

Lower
Limit
19.0

Upper
Limit
70.0

TCM
B1T4C4
20.32

3.0
5.0
0.3

15.0
20.0
1.0

5.94
9.14
0.625

It can be seen from Table 4.3 that upper and lower limits are defined in a wide
range to investigate capabilities of the developed design optimization methods. A
wide range of design variables leads to increased number of possible
configurations in the design space; thus, making the optimization problem more
difficult. Available area for geometry and placement for canards is shown with
pink in Figure 4.5. Dashed lines represent random possible canard geometries that
may be encountered during optimization runs.

Figure 4.5 Canard Placement Area (red) and Random Possible Canard Geometries
(dashed lines)

Objectives for aerodynamic optimization of TCM configuration can be given as;

(4.1)

([ (

[(

(
(

52

])

(4.2)

(4.3)

where indices

and

represent values regarding TCM configuration and

and 3 are penalty factors for each

optimum configuration, respectively.

aerodynamic coefficient and defined relatively high compared to objective values.

In summary, objective function to be minimized for aerodynamic design

optimization is defined as;

(
(4.4)

4.1.3.2 Optimization using DONLP2

In this part of the study, aerodynamic design optimization is done using DONLP2,
an algorithm based on Sequential Quadratic Programming optimization method,
coupled with aerodynamic analysis tool. Objective aerodynamic parameters and
design variable constraints for external geometry optimization case are given in
Table 4.2 and Table 4.3, respectively. After 73 iterations in 2926 seconds with
2018 different configuration evaluations using aerodynamic analysis software,
design tool determined a configuration that satisfies the objectives of optimization
within design constraints. Some configuration geometries along the optimization
run are given in Table 4.4. Optimum configuration and TCM B1T4C4 canard
geometries are shown in Figure 4.6.

53

Table 4.4 Change of Configuration Geometry along Optimization

Iteration Step

Configuration Geometry

15

30

55

73
(Optimum Geometry)

54

Figure 4.6 Optimum Configuration Geometry (top), TCM B1T4C4 (bottom)

Aerodynamic performance parameters and geometric proportions of optimum

configuration and TCM B1T4C4 configuration are given in Table 4.5 and Table
4.6, respectively.
Table 4.5 Optimum and TCM B1T4C4 Configuration (Objective) Aerodynamic
Performance Parameters

Optimum Conf.

TCM-B1T4C4

CN (=16)

6.531

6.532

XCP/D (=4)

0.370

0.370

XCP/D (=16)

0.132

0.132

CA (=0)

0.517

0.517

55

Table 4.6 Optimum and TCM B1T4C4 Missile Geometric Proportions

Optimum Conf.

TCM B1T4C4

XLE [cm]

20.32

20.32

SSPAN [cm]

5.95

5.94

RCHORD [cm]

9.12

9.14

TR

0.62

0.63

Changes of design variables and objective aerodynamic performance parameters

along the optimization are shown in below figures (Figure 4.7 - Figure 4.15).

30

Leading Edge Position [cm]

28

26

24

22

TCM B1T4C4

20

18

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.7 Change of Leading Edge Position throughout Optimization

56

13
12.5

Root Chord [cm]

12
11.5
11
10.5
10
9.5
TCM B1T4C4

9
8.5

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.8 Change of Root Chord throughout Optimization

0.85

0.8

Taper Ratio

0.75

0.7

0.65
TCM B1T4C4

0.6

0.55

10

20

30

40

50

60

70

80

Iterations

Figure 4.9 Change of Taper Ratio throughout Optimization

57

90

6.2

6
TCM B1T4C4

Semi-Span [cm]

5.8

5.6

5.4

5.2

4.8

4.6

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.10 Change of Semi-span throughout Optimization

7

6.8

CN

6.6
TCM B1T4C4

6.4

6.2

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.11 Change of Normal Force Coefficient (M=1.75, =16) throughout

Optimization

58

0.5
0.4
TCM B1T4C4

0.3

XCP /D

0.2
0.1
0

-0.1
-0.2
-0.3
-0.4

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.12 Change of Center of Pressure Location (M=1.75, =4) throughout

Optimization
0.2
0.15
TCM B1T4C4

0.1
0.05

XCP /D

0
-0.05
-0.1
-0.15
-0.2
-0.25
-0.3

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.13 Change of Center of Pressure Location (M=1.75, =16) throughout

Optimization

59

0.6

0.58

CA

0.56

0.54

0.52

TCM B1T4C4

0.5

0.48

10

20

30

40

50

60

70

80

90

Iterations

Figure 4.14 Change of Axial Force Coefficient (M=1.75, =0) throughout Optimization
14

12

Objective Function

10

-2

TCM B1T4C4

10

20

30

40

50

60

70

80

Iterations

Figure 4.15 Change of Objective Function throughout Optimization

60

90

Optimization case study for TCM configuration is solved 15 times to evaluate

repeatability of the results. Different initial design points are supplied for these
runs. Change of objective function values with iterations for these runs is shown
in Figure 4.16. Aerodynamic performance parameters of best and worst
configurations according to objective function value are given Table 4.7.

Function
Objective
F(x,y)

10

BEST
WORST
ITER

100

10-1

10

-2

20

40

60

80

Iterations
Figure 4.16 Change of Objective Function Values with Iterations

Table 4.7 Aerodynamic Performance Parameters of Best and Worst Configurations

CN

XCP/D

XCP/D

CA

(=16)

(=4)

(=16)

(=0)

Best

6.532

0.370

0.132

0.517

2.999E-04

Worst

6.706

0.001

0.132

0.600

5.873E+00

OBJECTIVE

6.532

0.370

0.132

0.517

0.0

61

f(x)

From Figure 4.16, it can be said that there is no guarantee for reaching to
objectives of the problem when gradient-based DONLP2 algorithm is used for
design optimization. In addition to that, effect of initial points on results is
significant. Moreover, there is a considerable difference between best and worst
configurations in terms of aerodynamic performance parameters, as seen in Table
4.7. It is necessary to consider these weaknesses of the DONLP2 algorithm for
missile aerodynamic optimization problems.

4.1.3.3 Optimization using ACRS

Aerodynamic design optimization study for TCM configuration using Random
Search optimization algorithm ACRS is discussed in this section. Objective
aerodynamic parameters and design variable constraints for external geometry
optimization case are given in Table 4.2 and Table 4.3, respectively. It takes 5539
seconds and 3819 iterations with 3819 different configuration evaluations using
Missile DATCOM to determine a configuration that satisfies objectives within
design constraints. Some configuration geometries along the optimization run are
given in Table 4.8. Optimum configuration and TCM B1T4C4 canard geometries
are shown in Figure 4.17.

62

Table 4.8 Change of Configuration Geometry along Optimization Run

Iteration Step

Configuration Geometry

50

200

500

1000

2000

3819
(Optimum Geometry)

63

Figure 4.17 Optimum Configuration Geometry (top), TCM B1T4C4 (bottom)

Aerodynamic performance parameters and geometric proportions of optimum

configuration and TCM B1T4C4 configuration are given in Table 4.9 and Table
4.10, respectively.
Table 4.9 Optimum and TCM B1T4C4 Configuration (Objective) Aerodynamic
Performance Parameters

Optimum Conf.

TCM-B1T4C4

CN (=16)

6.532

6.532

XCP/D (=4)

0.370

0.370

XCP/D (=16)

0.132

0.132

CA (=0)

0.517

0.517

64

Table 4.10 Optimum and TCM B1T4C4 Missile Geometric Proportions

Optimum Conf.

TCM B1T4C4

XLE [cm]

20.33

20.32

SSPAN [cm]

5.95

5.94

RCHORD [cm]

9.13

9.14

TR

0.62

0.63

Changes of design variables and objective aerodynamic performance parameters

along the optimization are shown in below figures (Figure 4.18 - Figure 4.26).

70
65

Leading Edge Position [cm]

60
55
50
45
40
35
30
25
TCM B1T4C4

20
0

1000

2000

3000

4000

5000

Iterations

Figure 4.18 Change of Leading Edge Position throughout Optimization

65

Root Chord [cm]

20

15

10
TCM B1T4C4

5
0

1000

2000

3000

4000

5000

Iterations

Figure 4.19 Change of Root Chord throughout Optimization

1
0.95
0.9
0.85
0.8

Taper Ratio

0.75
0.7
0.65
TCM B1T4C4

0.6
0.55
0.5
0.45
0.4
0.35
0.3
0.25

1000

2000

3000

4000

Iterations

Figure 4.20 Change of Taper Ratio throughout Optimization

66

5000

Semi-Span [cm]

15

10

TCM B1T4C4

1000

2000

3000

4000

5000

Iterations

Figure 4.21 Change of Semi-span throughout Optimization

18

16

14

CN

12

10

8
TCM B1T4C4

1000

2000

3000

4000

5000

Iterations

Figure 4.22 Change of Normal Force Coefficient (M=1.75, =16) throughout

Optimization

67

4
3
2

XCP /D

1
TCM B1T4C4

0
-1
-2
-3
-4
-5

1000

2000

3000

4000

5000

Iterations

Figure 4.23 Change of Center of Pressure Location (M=1.75, =4) throughout

Optimization
4

XCP /D

TCM B1T4C4

-2

-4

1000

2000

3000

4000

5000

Iterations

Figure 4.24 Change of Center of Pressure Location (M=1.75, =16) throughout

Optimization

68

1.4

1.2

CA

0.8

0.6
TCM B1T4C4

0.4

0.2

1000

2000

3000

4000

5000

Iterations

Figure 4.25 Change of Axial Force Coefficient (M=1.75, =0) throughout Optimization
100

Objective Function

80

60

40

20

TCM B1T4C4

0
0

1000

2000

3000

4000

5000

Iterations

Figure 4.26 Change of Objective Function throughout Optimization

69

This missile optimization case study considered for TCM configuration is solved
15 times to evaluate repeatability of the results. Change of objective function
values with iterations for these runs is shown in Figure 4.27 and Figure 4.28.
Aerodynamic performance parameters of best and worst configurations according
to objective function value are given Table 4.11.

Table 4.11 Aerodynamic Performance Parameters of Best and Worst Configurations

CN

XCP/D

XCP/D

CA

(=16)

(=4)

(=16)

(=0)

Best

6.532

0.370

0.132

0.517

1.215E-06

Worst

6.540

0.360

0.149

0.517

2.934E-01

OBJECTIVE

6.532

0.370

0.132

0.517

0.0

f(x)

120
BEST
WORST

Function
Objective
F(x,y)

100

80

60

40

20

0
1000

2000

3000

Iterations
Figure 4.27 Change of Objective Function Values with Iterations

70

20
BEST
WORST

F(x,y)
Function
Objective

15

10

0
1000

2000

3000

Iterations
Figure 4.28 Change of Objective Function Values with Iterations (y axis is narrowed)

From Figure 4.27 and Figure 4.28, it can be seen that all of the 15 runs reach
aerodynamic design objectives in the end. Accordingly, there is a slight difference
between best and worst configurations in terms of aerodynamic performance
parameters, as seen in Table 4.11. Moreover, ACRS algorithm does not require
user defined initial points for optimization runs; therefore, there is no effect of
starting point on results. Note that, DONLP2 reaches the best solution in 2018
configuration evaluations whereas ACRS algorithm evaluate 3819 configurations
for best solution case, meaning that ACRS is almost twice slower than DONLP2.
In the end, all of these results suggest that ACRS algorithm is more robust but
slower in achieving design objectives, when compared to DONLP2.

71

4.2

Aerodynamic Shape Optimization for a Generic Air-to-Ground Missile

(AGM)

4.2.1 Missile Design Optimization for AGM

An aerodynamic optimization case study is carried out for a generic air-to-ground
missile (AGM) using developed optimization method. Both optimization
algorithms, DONLP2 and ACRS, are used for this purpose. Results are presented
separately.

4.2.1.1 Objectives, Constraints and Variables of Optimization Study

It is assumed that body of the AGM under consideration has a fixed geometry due
to sub-system components such as rocket engine, seeker, warhead, control surface
actuators etc. and missile is tail-controlled. It is also assumed that weight is
constant and center of gravity does not change. Wing and tail geometries are taken
as design variables. Geometric constraints defined for these design variables are
given in Table 4.12.

Table 4.12 Missile Geometry Variables and Limits

Variable
XLE1 [cm]
RCHORD1 [cm]
TR1
SSPAN 1[cm]
XLE2[cm]
RCHORD2 [cm]
TR2
SSPAN2[cm]

Description
Wing leading edge position
Wing root chord
Wing taper ratio
Wing semi-span
Tail leading edge position
Tail root chord
Tail taper ratio
Tail semi-span

72

Lower
Limit
30.0
4.0
0.0
3.0
130.0
5.0
0.0
3.0

Upper
Limit
120.0
30.0
1.0
16.0
160.0
30.0
1.0
16.0

In missile external configuration optimization study for a generic AGM

configuration, typical aerodynamic performance requirements and the flight
regime for this type of mission are taken into account. Aerodynamic performance
parameters such as load factor and lift-to-drag ratio at trim condition, control
effectiveness, static margin and hinge moments are defined as objectives and
constraints. These parameters are defined considering maneuverability, range,
stability and weight requirements of a typical AGM and summarized in Table
4.13.

Table 4.13 Aerodynamic Performance Parameters for Sample AGM

Aerodynamic Parameter

Constraint

Objective

Trim Load Factor (ntrim)

ntrim > 3.5

Maximize ntrim

Trim Lift-to-Drag Ratio (CL/CD trim)

Maximize

CL/CD trim > 2.5

CL/CD trim

Static Margin (SM)

-1.0 > SM > -1.5

Control Effectiveness (CE)

-0.5 > CE > -0.8

|HM| < 2 Nm

Minimize HM

Hinge Moment (HM)

Flight regime defined for AGM external optimization case study is given in Table
4.14.

Table 4.14 Flight Regime for AGM Optimization Case Study

0.1 < Mach < 0.8

Mach

-15 15

Angle of Attack

= 0

Sideslip Angle

Sea Level

Altitude

In summary, objectives and constraints given in Table 4.13 for AGM aerodynamic
optimization study is defined mathematically as;

73

35
(

5
(

where

and

(4.5)

(4.6)

(4.7)

are relatively high numbers used to increase weight of the

objectives. Constraints given in Table 4.13 are defined with penalty parameters,

35

[ 5

(4.10)

(
(4.11)

(4.9)

(4.8)

)]

35

5
(4.12)

74

where

is a relatively high number used to increase weight of penalty functions.

Maximum and minimum values for aerodynamic design parameters are

determined using aerodynamic analysis tool for the flight regime given in Table
4.14. In this way, it is guaranteed that a search is done for an optimum
configuration that satisfies aerodynamic objectives inside the given flight
envelope and constraints are taken into account for the entire flight regime.

Finally, objective function to be minimized for aerodynamic design optimization

is written as;

(4.13)

Using above objective function, aerodynamic shape optimization for sample

AGM is studied. Results obtained using two different optimization algorithms are
discussed separately in the following sections.

4.2.1.2 Optimization using DONLP2

Aerodynamic design optimization case study using Random Search optimization
algorithm DONLP2 coupled with aerodynamic analysis tool is discussed in this
section. Design variable constraints and objective aerodynamic parameters for
external geometry optimization case are given in Table 4.12 and Table 4.13,
respectively. After 29 iterations in 718 seconds with 1709 aerodynamic analyses
for candidate configurations, an optimum configuration that satisfies the
objectives of optimization within design constraints is determined. Aerodynamic
75

performance parameters and geometric proportions of the optimum configuration

are given in Table 4.15 and Table 4.16, respectively. Change of AGM geometry
along optimization process is presented in Table 4.17.

Table 4.15 Optimum AGM and Objective Aerodynamic Performance Parameters

Aerodynamic
Parameter
ntrim

CL/CD trim

Constraint

Optimum AGM
Configuration

Objective
Maximize

ntrim > 3.5

(ntrim)max = 3.69 g

ntrim
Maximize

CL/CD trim > 2.5

CL/CD trim

(CL/CD trim)max = 3.16

SM

-1.0 > SM > -1.5

-1.10 > SM > -1.48

CE

-0.5 > CE > -0.8

-0.50 > CE > -0.64

HM

|HM| < 2 Nm

Minimize HM

|HM| < 0.35 Nm

Table 4.16 Optimum AGM Geometric Proportions

Variable
XLE1 [cm]
RCHORD1
[cm]
TR1
SSPAN 1[cm]
XLE2[cm]
RCHORD2
[cm]
TR2
SSPAN2[cm]

Description

Lower
Limit

Upper
Limit

Optimum
AGM
Configuration

30.0

120.0

63.55

4.0

30.0

16.0

0.0
3.0

1.0
16.0

0.30
9.1

130.0

160.0

134.7

5.0

30.0

12.2

0.0
3.0

1.0
16.0

0.84
16.0

Wing leading edge

position
Wing root chord
Wing taper ratio
Wing semi-span
Tail leading edge
position
Tail root chord
Tail taper ratio
Tail semi-span

76

Table 4.17 Change of AGM Configuration Geometry along Optimization Run

Iteration

Configuration Geometry

10

20

29
(Optimum
Configuration)

4.2.1.3 Optimization using ACRS

Aerodynamic design optimization case study using Random Search optimization
algorithm ACRS coupled with aerodynamic analysis tool is discussed in this
section. Design variable constraints and objective aerodynamic parameters for
77

external geometry optimization case are given in Table 4.12 and Table 4.13,
respectively. After 5678 iterations in 2328 seconds, an optimum configuration that
satisfies objectives of optimization within design constraints is determined.
Aerodynamic performance parameters and geometric proportions of optimum
configuration are given in Table 4.18 and Table 4.19, respectively. Change of
AGM geometry along optimization process is presented in Table 4.20.

Table 4.18 Optimum AGM and Objective Aerodynamic Performance Parameters

Aerodynamic
Parameter
ntrim

CL/CD trim

Constraint

Optimum AGM
Configuration

Objective
Maximize

ntrim > 3.5

(ntrim)max = 4.52 g

ntrim
Maximize

CL/CD trim > 2.5

CL/CD trim

(CL/CD trim)max = 3.2

SM

-1.0 > SM > -1.5

-1.07 > SM > -1.46

CE

-0.5 > CE > -0.8

-0.50 > CE > -0.68

HM

|HM| < 2 Nm

Minimize HM

|HM| < 0.64 Nm

Table 4.19 Optimum AGM Geometric Proportions

Variable
XLE1 [cm]
RCHORD1
[cm]
TR1
SSPAN 1[cm]
XLE2[cm]
RCHORD2
[cm]
TR2
SSPAN2[cm]

Description

Lower
Limit

Upper
Limit

Optimum
AGM
Configuration

30.0

120.0

77.14

4.0

30.0

17.0

0.0
3.0

1.0
16.0

0.30
16.0

130.0

160.0

137.5

5.0

30.0

16.3

0.0
3.0

1.0
16.0

0.88
15.9

Wing leading edge

position
Wing root chord
Wing taper ratio
Wing semi-span
Tail leading edge
position
Tail root chord
Tail taper ratio
Tail semi-span

78

Table 4.20 Change of AGM Configuration Geometry along Optimization Run

Iteration

Configuration Geometry

50

200

500

1000

3000

5678
(Optimum
Geometry)

79

4.2.1.4 Discussion of Results

In preceding sections, results for a sample AGM optimization case study are
presented. Evolution of the AGM configuration to the optimum aerodynamic
missile configuration can be seen in Table 4.17 for optimization with DONLP2
and Table 4.20 for optimization with ACRS. Table 4.15 and Table 4.18 show that
both of the algorithms reach objectives and satisfy constraints. Nevertheless, a
slightly better optimum configuration is obtained using ACRS in terms of final
objective function value. 5678 different configurations are evaluated when ACRS
algorithm is used for optimization, whereas evaluation of 1709 configurations is
enough for DONLP2 algorithm to reach the optimum configuration. This is
consistent with the results obtained for TCM study, discussed in the previous
sections. Note that, repeatability of the results especially for the ones obtained
using DONLP2 algorithm should be questioned as done previously.

Results obtained from different optimization approaches suggest similar outcomes

in missile aerodynamic design point of view. Since defined aerodynamic
performance objectives and constraints conflict with each other, optimization
methodology seeks a balance between them to minimize the objective function.
For example, areas of the wings and control surfaces (tails) increase in the end
since trim load factor objective imposes a requirement as maximization of normal
force of the configuration. However, due to constraints on static margin and
control effectiveness, increase of wing and tail areas are limited. Lift-to-drag ratio
objective has an impact on wing and tail layout, leading to some taper especially
for wings. In addition to these, tail geometry and location is affected by hinge
moment objective. As a result, a low tapered tail geometry that minimizes center
of pressure shift is obtained for optimum configurations.

80

CHAPTER 5

CONCLUSIONS AND FUTURE WORK

In this thesis, an aerodynamic design optimization methodology is developed to

determine optimum external configurations of missiles. Two different
optimization algorithms are employed for the development of an optimization
procedure. One of the optimization approaches used is Sequential Quadratic
Programming (SQP), a state-of-the-art gradient-based optimization method. An
SQP algorithm known as DONLP2 is applied to the design optimization problem.
In addition to that, Adaptively Controlled Random Search (ACRS) algorithm is
also considered for the investigation of optimum aerodynamic missile
configuration. An aerodynamic coefficient fast prediction tool based on
component build-up methods, Missile DATCOM, is used for the aerodynamic
analysis of missile configurations.

A verification study on a global optimization test function known as GoldsteinPrice function is done to investigate the capabilities of the employed optimization
algorithms. Results suggest that SQP method has difficulties on locating global
optimum point in a complex search space. Due to the nature of the method, it may
get stuck in local minima when solutions are repeated for different initial points.
In contrast, Random Search (RS) approach is fairly good at determining global
minimum. For repeated solutions to Goldstein-Price test function, RS method
always converged to the global minimum. Accordingly, it can be concluded that
RS approach is more suitable for design optimization problems where the solution
space is considerably complex. In addition to these verification studies, a
validation for the employed aerodynamic analysis tool, Missile DATCOM, is
carried out. It can be understood that Missile DATCOM is sufficiently accurate
81

for aerodynamic coefficient prediction and has acceptable accuracy to be used in

developed optimization method.

To present the capabilities of the developed methodology on missile aerodynamic

optimization, two different optimization applications are considered. First, a study
with reverse engineering approach is discussed where NASA Tandem Control
Missile (TCM) is taken as baseline configuration. For SQP and RS approaches
this problem is solved separately. In the end, both optimization methods are able
to reach aerodynamic objectives that yield the original geometry of TCM
configuration. However, for repeated solutions, it is determined that RS approach
is superior to SQP in terms of reaching the optimization objectives. The effect of
the initial point on final solution and difficulties on reaching the objectives of the
problem are the identified drawbacks of the SQP method. Even though SQP is
faster to reach the final solution, it is hard to rely on obtaining optimum results by
this method. In expense of some amount of solution time, RS approach work
fairly good for aerodynamic optimization of missile configurations. As a second
aerodynamic optimization application, optimization of a generic Air-to-Ground
Missile (AGM) configuration is discussed. Again, both optimization methods are
used to determine the optimum configuration. Results yield that ACRS algorithm
determined a better aerodynamic missile configuration in terms of aerodynamic
performance objectives. Note that DONLP2 algorithm is almost 4 times faster to
reach the solution in this case.

For future studies, application of developed aerodynamic design optimization

methodology to different types of missiles can be done. For example, an existing
air-to-air missile configuration (e.g. AIM-9X) can be resized for improved
aerodynamic performance with additional design variables. Moreover, different
aerodynamic analysis approaches can be employed. For this purpose, some other
aerodynamic fast prediction tools and CFD methods can be integrated to
aerodynamic optimization methodology developed in this work. A hybrid
optimization algorithm using RS and SQP method can be developed to exploit
robustness of RS and computational efficiency of SQP.
82

REFERENCES

[1] K. M. Rasheed, "GADO: A Genetic Algorithm for Continuous Design

Optimization," The State University of New Jersey, New Brunswick, New
Jersey, 1998.
[2] M. Y. ztrk, "Multiobjective Design Optimization of Missiles," Middle
East Technical University, Ankara, 2009.
[3] X. Wang and M. Damodaran, "Comparison of Deterministic and Stochastic
Optimization Algorithms for Generic Wing Design Problems," Journal of
Aircraft, vol. 37, no. 5, pp. 929-932, 2000.
[4] E.L. Fleeman, Tactical Missile Design. Reston, Virginia: American Institute
of Aeronautics and Astronautics Inc., 2001.
[5] J. N. Nielsen, Missile Aerodynamics. Palo Alto, California: McGraw-Hill,
1960.
[6] W. Blake, L. Auman, J. Doyle, C. Rosema, and M. Underwood, "Missile
DATCOM User's Manual," United States Air Force, Ohio, 2011.
[7] B. Atak, "Enhancement of Aerodynamic Fast Prediction Methods for Slotted
Rockets/Missiles," Istanbul Technical University, Istanbul, 2012.
[8] S. S. Chin, Missile Configuration Design. Orlando, Florida: McGraw-Hill,
1961.
[9] B. Etkin and L. D. Reid, Dynamics of Flight: Stability and Control. New
York: John Wiley and Sons, 1995.
[10] J. D. Anderson, Aircraft Performance and Design. London: McGraw-Hill,
1999.
[11] M. B. Anderson, J. E. Burkhalter, and R. M. Jenkins, "Missile Aerodynamic
Shape Optimization Using Genetic Algorithms," Journal of Spacecraft and
Rockets, vol. 37, no. 5, pp. 663-669, September-October 2000.
[12] . Tanl, "Optimal External Configuration of a Missile," Middle East
Technical University, Ankara, 2009.
83

[13] A. Karako, "Optimization of Aerodynamic Properties of Tactical Missiles,"

Middle East Technical University, Ankara, 2011.
[14] Y. R. Yang, S. K. Jung, T. H. Cho, and R. S. Myong, "Aerodynamic Shape
Optimization System of a Canard-Controlled Missile Using Trajectory
Dependent Aerodynamic Coefficients," Journal of Spacecraft and Rockets,
vol. 49, no. 2, pp. 243-249, March-April 2012.
[15] O. Tekinalp and M. Bingl, "Simulated Annealing for Missile Optimization:
Developing Method and Formulation Techniques," Journal of Guidance,
Control and Dynamics, vol. 27, no. 4, pp. 616-626, 2004.
[16] . Tanrkulu and V. Ercan, "Optimal External Configuration Design of
Unguided Missiles," Journal of Spacecraft and Rockets, vol. 35, no. 3, pp.
312-316, 1998.
[17] J. Jezowski, R. Bochenek, and G. Ziomek, "Random Search Optimization
Approach for Highly Multi-modal Nonlinear Problems," Advances in
Engineering Software, vol. 36, pp. 504-517, 2005.
[18] N. J. Silveira, N. S. M. Cardozo, A. R. Secchi, and K. Wada, "Adaptive
Random Search: A Promising Method for Determining the Stability of
Mixtures," in 10th International Symposium on Process Systems
Enginnering, 2009, pp. 321-326.
[19] I. G. Tsoulos and I. E. Lagaris, "Genetically Controlled Random Search: A
Global Optimization Method for Continuous Multidimensional Functions,"
Computer Physics Communications, no. 174, pp. 152-159, 2006.
[20] R. Simpson, A. Abakarov, and A. Teixeira, "Variable Retort Temperature
Optimization Using Adaptive Random Search Techniques," Food Control,
no. 19, pp. 1023-1032, 2008.
[21] S. S. Rao, Engineering Optimization, Theory and Practice. Hoboken, New
Jersey: John Wiley and Sons, 2009.
[22] P. J. Lin and D. W. Zingg, "Comparison of Optimization Algorithms Applied
to Aerodynamic Design," in 42nd AIAA Aerospace Sciences Meeting and
Exhibit, Reno, Nevada, 2004.
[23] N. B. Kuntawala, J. E. Hicken, and D. W. Zingg, "Preliminary Aerodynamic
Shape Optimization of a Blended-Wing-Body Aircraft Configuration," in
49th AIAA Aerospace Sciences Meeting, Orlando, 2011.

84

[24] J. A. Snyman, Practical Mathematical Optimization: An Introduction to

Basic Optimization Theory and New Gradient Based Algorithms.: Springer,
2005.
[25] Simulia, Application of Metal Processing Simulations.: ASM Handbook
Volume 22B, 2010.
[26] P. Brachetti, M. De Felice Ciccoli, G. Di Pillo, and S. Lucidi, "A New
Version of the Prices Algorithm for Global Optimization," Journal of Global
Optimization, no. 10, pp. 165-184, 1997.
[27] P. Spellucci, "DONLP2 Users Manual," Technical University at Darmstadt,
Darmstadt, 1988.
[28] M. J. Hemsch, "Component Build-Up Method for Engineering Analysis of
Missiles at Low-to-High Angle of Attack," Hampton, Virginia, 1992.
[29] S. R. Vukelich and J. E. Jenkins, "Missile DATCOM: Aerodynamic
Prediction of Conventional Missiles Using Component Build-Up
Techniques," in AIAA 22nd Aerospace Sciences Meeting, Reno, 1984.
[30] L.C.W. Dixon and G.P. Szeg, Towards Global Optimization 2. NorthHolland, Amsterdam, 1978.
[31] A. E. etiner, B. Atak, A. Akgl, and D. C. Oral, "NASA Dual Control
Missile Denek Ta iin Hesaplamal Akkanlar Dinamii Analizleri ve
Dorulanmas," in IV. Ulusal Havaclk ve Uzay Konferans, stanbul, 2012.
[32] M. R., Leavitt, L. D., Leopold, F. Mendenhall, "Experimental Test Cases,"
2005.
[33] D. Lesieutre, J. Love, and M. Dillenius, "Recent Applications and
Improvements to the Engineering Level Aerodynamic Prediction Software
MISL3," in 40th AIAA Aerospace Sciences Meeting & Exhibit, Reno, NV,
2002.

85