Czech Technical University in Prague

Faculty of Electrical Engineering

Department of Cybernetics

Bachelor's Thesis

Modeling of Launch Vehicle during the Lift-o Phase in

Atmosphere

Matyá² Lustig

Supervisor: doc. Ing. Martin Hrom£ík, Ph.D.

Study Programme: Cybernetics and Robotics

Field of Study: Robotics

Bachelor's degree

May 18, 2017

ii
 Czech Technical University in Prague
Faculty of Electrical Engineering

Department of Cybernetics

BACHELOR PROJECT ASSIGNMENT

Student: Matyáš L u s t i g

Study programme: Cybernetics and Robotics

Specialisation: Robotics

Title of Bachelor Project: Modeling of Launch Vehicle during the Lift-off Phase
in Atmosphere

Guidelines:
The objective of this thesis is to provide a platform for model-based simulation and control
laws validation of launch vehicles. The emphasis is put on the lift-off phase of the rocket flight
which is analysed using a mathematical flight dynamics model. An expected result of the
Bachelor thesis are also simulation and practical experiment illustrating advanced physical
phenomena of a real rocket flight.

1. Get familiar with the principals of rocket engineering and rocket science.
2. Develop a flight mechanics model of a launch vehicle during the lift-off phase of a flight
through the atmosphere.
3. Create a dedicated simulation scheme in the MATLAB/Simulink environment.
4. Design, execute and document selected practical experiments, related either to flight
identification and validation of the simulation model, or to flight testing of selected control
algorithms.

The water-rocket developed within a FEL SGS project, with its on-board unit, may be used as
the experimental platform.

Bibliography/Sources:
[1] Arthur E. Bryson, Jr. - Control of Spacecraft and Aircraft - 1994, Princeton, New Jersey,
USA
[2] John H. Blakelock - Automatic Control of Aircraft and Missiles, 2nd edition - 1991, USA

Bachelor Project Supervisor: doc. Ing. Martin Hromčík, Ph.D.

Valid until: the end of the summer semester of academic year 2017/2018

L.S.

prof. Dr. Ing. Jan Kybic prof. Ing. Pavel Ripka, CSc.
Head of Department Dean

Prague, December 2, 2016

iv
 v

Acknowledgements
I would like to express my gratitude to my supervisor, doc. Ing. Martin Hrom£ík, Ph.D. for
all the suggestions and advice he has oered me during the work on this thesis. I am grateful
to my parents for their continuous encouragement and to Eva who read my rst thesis.
vi
 vii

Declaration
I declare that the presented work was developed independently and that I have listed all
sources of information used within it in accordance with the methodical instructions for
observing the ethical principles in the preparation of university theses.

In Prague on May 18, 2017 .............................................................

viii
Abstract
The objective of this thesis is to provide a platform for model based simulation and control
laws validation of launch vehicles. The lift-o phase of the rocket ight is analyzed using
a mathematical nonlinear dynamics model. Full set of six degrees of freedom equations of
motion is developed with combination of auxiliary computations and necessary reference
frames and transformations in between them. The mathematical model is employed by
comprehensive MATLAB and Simulink simulation which can be easily adjusted for any type
of launch vehicle or rocket. The simulation model has been validated and linearized for a
further analysis. Water rocket practical experiment demonstrates the simulation platform
capabilities.

Abstrakt
Práce pojednává o vývoji platformy pro po£íta£ovou simulaci letu raketových nosi£·, jeº
m·ºe být pouºita i pro návrh a validaci automatického °ízení. Dynamický nelineární model
sestávající z pohybových rovnic uvaºujících ²est stup¬· volnosti rakety, podp·rných výpo£t·
a transformací mezi sou°adnými systémy popisuje vzletovou fázi letu rakety v atmosfé°e. Na
základ¥ matematického modelu je v prost°edí MATLAB a Simulink navrºen model sim-
ula£ní. Výstupy simula£ní platformy jsou ov¥°eny a porovnány s o£ekávaným chováním
raketového nosi£e. Nelineární model je dále zlinearizován pro hlub²í analýzu. Moºnosti
simula£ního modelu demonstruje praktický experiment s vodní raketou, kdy jsou výsledky
simulace srovnány s daty z reálného letu.

ix
x
Contents
1 Introduction 1
2 Goals 2
3 Mathematical Model 3
3.1 Reference Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.1.1 Earth Centered Inertial Reference Frame . . . . . . . . . . . . . . . . . 3
3.1.2 Earth Centered Fixed Reference Frame . . . . . . . . . . . . . . . . . . 3
3.1.3 Earth Geographic Reference Frame . . . . . . . . . . . . . . . . . . . . 4
3.1.4 Body Reference Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Rotational Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3.1 Translational Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3.2 Rotational Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Subsystem Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4.1 Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4.2 Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4.3 Fin Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4.4 Thrust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4 Simulation 24
4.1 Simulink Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Input Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 Linear Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.4.1 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

5 Water Rocket Experiment 38

5.1 Input data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

6 Results 43
7 Conclusion 45

xi
A Contents of the attached CD 46

xii
List of Figures
3.1 Earth centered inertial and Earth centered xed reference frames . . . . . . . 4
3.2 Earth centered xed and Earth geographical reference frames . . . . . . . . . 5
3.3 Body reference frame with yaw, pitch and roll angles of rotation . . . . . . . . 6
3.4 Air pressure as a function of altitude up to 80 km . . . . . . . . . . . . . . . . 14
3.5 Air pressure as a function of altitude up to 1000 km . . . . . . . . . . . . . . 14
3.6 Air temperature as a function of altitude up to 80 km . . . . . . . . . . . . . 16
3.7 Air temperature as a function of altitude up to 1000 km . . . . . . . . . . . . 17
3.8 Air mass density as a function of altitude up to 80 km . . . . . . . . . . . . . 18
3.9 Air mass density as a function of altitude up to 1000 km . . . . . . . . . . . . 18
3.10 Relation between lift & drag and normal & axial forces acting on a body . . . 19
3.11 Water rocket model ns arrangement [10] . . . . . . . . . . . . . . . . . . . . 21
3.12 Thrust vectoring schema with yaw and pitch control angles . . . . . . . . . . 23

4.1 Simulink model - main scheme . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.2 Simulink model - transformations . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Total thrust of all nine rocket engines over time (tburn = 162 s) . . . . . . . . 27
4.4 Flight trajectory of launch vehicle . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.5 Thrust vectoring over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.6 Vehicle's Euler angles over time . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.7 Aerodynamic moments acting on the launch vehicle over time . . . . . . . . . 30
4.8 Nonlinear and linear model pitch angle θ responses comparison . . . . . . . . 32
4.9 Nonlinear and linear model angular velocity q responses comparison . . . . . 32
4.10 Poles and zeros of the system in relation to velocity . . . . . . . . . . . . . . . 33
4.11 Poles and zeros of the system in relation to moment of inertia . . . . . . . . . 34
4.12 Poles and zeros of the system in relation to pitch moment coecient for α . . 35
4.13 Launch Vehicle Simulation MATLAB GUI - rst tab . . . . . . . . . . . . . . 36
4.14 Launch Vehicle Simulation MATLAB GUI - second tab . . . . . . . . . . . . . 37
4.15 Launch Vehicle Simulation MATLAB GUI - third tab . . . . . . . . . . . . . 37

5.1 Water rocket experimental and simulation altitude results comparison . . . . 40

5.2 Water rocket experimental and simulation acceleration results comparison . . 41
5.3 Water rocket experimental angular velocity results . . . . . . . . . . . . . . . 42

xiii
List of Tables
3.1 WGS84 model dening constants . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Constants dening atmospheric layers . . . . . . . . . . . . . . . . . . . . . . 15

4.1 Simulation model input data . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5.1 Water rocket model input data . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.2 Estimated water rocket model input data . . . . . . . . . . . . . . . . . . . . 39

xiv
Chapter 1

Introduction
The ultimate task of a launch vehicle is to safely transport any kind of payload to an orbit
around the Earth. Most of the missions are covering highly valuable satellites or deep space
probes and some even serve for manned mission - to the Moon, to the ISS and maybe one
day in not so far future to other planets. Thus, a great attention is being paid to carrier
rockets reliability and development.

Aerospace along with military industry have always been the driving power of progress
and innovation in many dierent elds in modern history. The fastest advancement took
place during the Cold War renowned space race between the USA and the USSR resulting in
landing a man on the Moon. However, we have never really stopped exploring our surround-
ings. Recently, several initiatives became rife trying to push boundaries of humanity further
again. Elon Musk has revealed fearless plans for manned missions to Mars with SpaceX com-
pany which already successfully operates its highly innovative Falcon 9 rocket. Je Bezos
with his company Blue Origin has set himself the task of opening up the space to broader
public. And even the Czech Technical University is close to establishing a partnership with
a company which specializes in small class launch vehicles manufacturing. Apparently, the
aerospace industry is once again coming alive.

Design and manufacturing of rockets is a highly demanding and protracted process.

Before you can even start, you have to penetrate the physical processes occurring during
an atmospheric and space ight. This thesis is aiming to develop a mathematical model of
launch vehicle's ight focusing on the lift-o phase in the Earth's atmosphere. We begin
with dening various reference frames embracing the whole launch vehicle's environment.
Kinematic and dynamic models of rocket ight are developed resulting in a set of equations
of motion. Likewise, individual physical sub-models are utilized and described.

All together, the mathematical model serves as a basis for computer simulation develop-
ment. Mighty software MATLAB is utilized with combination of Simulink to create an exact
and interactive simulation of launch vehicle's ight. Basic ight laws may be veried and
certain situations or maneuvers demonstrated. A practical experiment with water rocket
model has been conducted at the CTU in the past and we will pick up its threads by
validating our simulation model with ight data obtained during that occasion. Let's start
the countdown!

1
Chapter 2

Goals
Based on the thesis' guidelines enclosed, several goals have been set. The following tasks
have been laid out and the work will hopefully resolve all of them.

1. Get acquainted with the state of the art of rocket engineering and rocket science.

2. Develop a ight mechanics model of a launch vehicle during the lift-o phase of a ight
through the Earth's atmosphere.

3. Create a simulation scheme in the MATLAB or Simulink environment and validate the
simulation's results.

4. Linearize the nonlinear model and perform a basic sensitivity analysis.

5. Design, execute and document a practical experiment with the water rocket model.
Calibrate the simulation model with respect to the obtained ight data.

2
Chapter 3

Mathematical Model
3.1 Reference Frames
Firstly, several frames of reference are introduced in order to describe the position and motion
of the launch vehicle in space in accordance with Zipfel [1]. Reference frames are understood
as right handed Cartesian coordinate systems specied by their orientation in space. Certain
trade-o has to be made between the number of frames used by the model and the clarity
of problem description. Following reference frames have been chosen because together they
are capable of complete characterization of the launch vehicle's motion.

The use of more reference frames implies necessity to coordinate transformations between
them. Appropriate transformation rotational matrices are presented for each two subsequent
frames. Any more complex or inverse transformations can be built upon the individual ones.

3.1.1 Earth Centered Inertial Reference Frame

For any space travel in our solar system, the heliocentric frame represents the inertial frame
of reference. However, for ights close to the surface or in the atmosphere, the Earth centered
frame is sucient. Origin of such coordinate system lies at the center of mass of the Earth.
We use the J2000 system with basis vectors î, ĵ, k̂ given as follows. The vector î lies in the
equatorial plane pointing towards the vernal equinox (intersection of ecliptic with the Earth's
equator in spring) in J2000 epoch (1st January 2000, 12:00 Terrestrial time). The k̂ vector
is normal to the equatorial plane, coincides with the Earth's axis of rotation and points to
the North pole. The remaining basis vector ĵ lies in the equatorial plane and completes the
right-handed Cartesian system. The frame may be referred to as the inertial one or by the
abbreviation ECI.

3.1.2 Earth Centered Fixed Reference Frame

The origin of the Earth centered xed frame also lies in the center of Earth. Let's describe
it with another set of basis vectors îe , ĵe and k̂e . First vector îe is dened as pointing from
the center of the Earth to the intersection of the Earth's prime Greenwich meridian with
the equator. Vector k̂e is aligned along the Earth's spin axis and thus coincides with the

3
4 CHAPTER 3. MATHEMATICAL MODEL

ECI's basis vector k̂ . Vector ĵe completes the right-handed Cartesian coordinate system.
The frame may be referred to as an Earth or ECF one.

k̂, k̂e

ĵe

ĵ
î Ω

îe

Figure 3.1: Earth centered inertial and Earth centered xed reference frames

The relation between ECF and ECI visible in gure 3.1 is determined by the Earth's
rotation about the k̂ vector for the angle between vernal equinox in J2000 and Greenwich
meridian (i.e. from î to îe ). The size of the angle can be expressed as Ω = ωE (t − tJ2000 ),
where ωE = 7.2921 × 10
−5 rad/s is the angular velocity of the Earth [2], t indicates the

current time and tJ2000 stands for the J2000 epoch. One complete Earth's rotation with
respect to the stars takes approximately 23h 56min, a period of time also called one sidereal
day. The transformation matrix from ECI to ECF frame can be expressed as
 
cos(Ω) sin(Ω) 0
ECF
TECI = − sin(Ω) cos(Ω) 0 (3.1)
0 0 1

3.1.3 Earth Geographic Reference Frame

This is the frame in which any position on the Earth's surface can be described. Longitude
is understood to be positive from the Greenwich meridian to the east, while latitude from
the equator to the north.
3.1. REFERENCE FRAMES 5

The frame's origin lies at the point of interest on the Earth's surface and orientation is
described by a set of basis vectors îg , ĵg and k̂g . In order to simplify the model, axes of
geographic coordinate system are aligned with body frame axes (see 3.1.4) at the launch
pad before takeo. The system's origin thus lies at the launch pad longitude and latitude
coordinates. The vector îg is normal to the Earth's surface pointing upward, vector k̂g points
towards the north and ĵg to the east completing the triad. The frame may be referred to as
a geographic or EG one.

k̂e

k̂g

ĵg
îg

ĵe

îe

Figure 3.2: Earth centered xed and Earth geographical reference frames

Transformation from ECF to EG frame consists of two parts and can be further examined
in gure 3.2 above. Firstly a rotation about k̂e (z -axis) for longitude angle λ and secondly
rotation about new y -axis for negative latitude angle Φ. The latitude angle is negative in
order to achieve the desired orientation of the geographic frame as described above. The
total transformation matrix from ECF to EG is formulated as

EG
TECF = R(y, −Φ)R(z, λ) (3.2)
 
cos(λ) cos(Φ) cos(Φ) sin(λ) sin(Φ)
EG
TECF = − sin(λ) cos(λ) 0  (3.3)
− cos(λ) sin(Φ) − sin(λ) sin(Φ) cos(Φ)
6 CHAPTER 3. MATHEMATICAL MODEL

3.1.4 Body Reference Frame

The last of the four frames is used to describe relative attitude of the launch vehicle with
respect to its initial position at the launch pad and serves as a basis for reference points
inside the vehicle.

Traditionally in space industry, the body frame is dened by basis vectors îb , ĵb and k̂b
set as follows. The frame's origin is merged with the spacecraft's center of gravity ( c.g.),
otherwise called the center of mass. Vector îb is aligned with the body main symmetry
axis pointing towards the rocket's nose. Coordinate system is further xed by vector ĵb
aligned with right n of the rocket and vector k̂b completes the right-handed triad pointing
downwards from the rocket's belly. The body reference frame may be simply called as the
body one or by abbreviation B.

îb

c.g.
ĵb
θ
k̂b
ψ
Figure 3.3: Body reference frame with yaw, pitch and roll angles of rotation

Any relative attitude of a body in space can be described using three consecutive rotations
about its three main body axis, in our case îb , ĵb and k̂b . In aerospace industry these axis and
angles ψ, θ, φ are typically referred to as yaw, pitch and roll successively. Overall orientation
of the body frame can be examined in gure 3.3. The total rotation from geographical to
body frame can be expressed by transformation matrix

B
TEG = R(x, φ)R(y, θ)R(z, ψ) (3.4)
 
cos(θ) cos(ψ) cos(θ) sin(ψ) − sin(θ)
B
TEG = cos(ψ) sin(θ) sin(φ) − cos(φ) sin(ψ) cos(φ) cos(ψ) + sin(θ) sin(φ) sin(ψ) cos(θ) sin(φ) 
sin(φ) sin(ψ) + cos(φ) cos(ψ) sin(θ) cos(φ) sin(θ) sin(ψ) − cos(ψ) sin(φ) cos(θ) cos(φ)]
(3.5)

Although the inertial frame used throughout this work is the ECI one, there seems to be
no point in determining the absolute launch vehicle's attitude with respect to it. Instead, the
3.2. ROTATIONAL KINEMATICS 7

yaw, pitch and roll angles ( YPR ) will describe rocket's attitude with respect to the initial
position at the launch pad, to the EG frame.

3.2 Rotational Kinematics

In the previous chapter various reference frames and transformation relations in between
them have been discussed. The yaw, pitch and roll angles hold great signicance for describ-
ing spacecraft attitude and behavior. In order to obtain their values throughout the ight,
dierential equations are necessary. An angular velocity is introduced as a derivative of the
YPR angles. The total angular velocity of launch vehicle's body frame with respect to the
geographic reference frame is expressed in body frame coordinates and can be decomposed
into three axes components
ω
~ = pîb + q ĵb + rk̂b (3.6)

As we are dealing with rotational angles used in reference frames transformation, the
angular velocity can also be represented as shown by Wie [3].

 
p
~ = q 
ω (3.7)
r
     
φ̇ 0 0
=  0  + R(x, φ) θ̇ + R(x, φ)R(y, θ)  0  (3.8)
0 0 ψ̇
 
φ̇ − ψ̇ sin θ
= θ̇ cos φ + ψ̇ cos θ sin φ
 (3.9)

ψ̇ cos θ cos φ − θ̇ sin φ

  
1 0 − sin θ φ̇
= 0 cos φ sin φ cos θ   θ̇  (3.10)
0 − sin φ cos φ cos θ ψ̇

We have expressed angular velocity using derivatives of YPR angles. However, the inverse
situation is more common in real life, and thus from equation 3.10 the kinematic dierential
equation can be obtained.

    
φ̇ cos θ sin φ sin θ cos φ sin θ p
 θ̇  = 1  0 cos φ cos θ − sin φ cos θ   q (3.11)
cos θ
ψ̇ 0 sin φ cos φ r

The biggest drawback of using Euler angles is that any rotational sequence has a singu-
larity at some point. In our case we can see the problem right away from the equation 3.11.
π
It does not make any sense for pitch angle θ= 2 +kπ where k∈Z because of cosine function
in denominator of the fraction. Such singularity is often referred to as a gimbal lock, term
originally describing a mechanical phenomenon when two axes of rotation coincide causing
the system to lose one degree of freedom. In other words, it becomes unrecognizable about
which axes of the two the system is rotating [4]. One of the possible solutions is to introduce
8 CHAPTER 3. MATHEMATICAL MODEL

quaternions, and thus fully avoid the danger of a gimbal lock. The use of quaternions in
computer simulation has also a great advantage of reducing the computation time as the
number of demanding trigonometric expressions is cut down.

Rotation quaternion for our rotation sequence 3 − 2 − 1 is described by Diebel [4] as

 
cos φ2 cos 2θ cos ψ2 + sin φ2 sin 2θ sin ψ2
 
q0
q1  sin φ2 cos 2θ cos ψ2 − cos φ2 sin 2θ sin ψ2 

~q =  =
q2  cos φ sin θ cos ψ + sin φ cos θ sin ψ 
 (3.12)
2 2 2 2 2 2
q3 φ θ ψ φ θ ψ
cos 2 cos 2 sin 2 − sin 2 sin 2 cos 2

where q0 is a scalar and q1 , q2 and q3 are forming a vector q̂ . For total angular velocity
ω
~ of the body with respect to EG frame, the rotational kinematics equation 3.11 has its
equivalent in quaternions [1] given by

    
q̇0 0 −p −q −r q0
q̇1  1 p 0 r −q  q1 
 =    (3.13)
q̇2  2 q −r 0 p  q2 
q̇3 r q −p 0 q3

Moreover, any rotational matrix can be built from quaternions with the help of equation

R = (q02 − q̂ T q̂)I3 + 2q̂ q̂ T + 2q0 [q̂x ] (3.14)

where [q̂x ] is the so-called skew-symmetric matrix expressing the cross product operation. If
we apply this equation to the transformation from geographical to body frame, the result is
an equivalent transformation matrix to the equation 3.5.

 2
q0 + q12 − q22 − q32

2q1 q2 − 2q0 q3 2q0 q2 + 2q1 q3
B
TEG =  2q0 q3 + 2q1 q2 q02 − q12 + q22 − q32 2q2 q3 − 2q0 q1  (3.15)
2q1 q3 − 2q0 q2 2q0 q1 + 2q2 q3 q0 − q12 − q22 + q32
2

The Euler angles representing yaw, pitch and roll rotations can be obtained back from
quaternion notation by three equations

2q1 q2 + 2q0 q3
ψ = arctan( ) (3.16)
q02+ q12 − q22 − q32
θ = arcsin(−2q0 q2 + 2q1 q3 ) (3.17)

2q2 q3 + 2q0 q1
φ = arctan( 2 ) (3.18)
q0 − q12 − q22 + q32

3.3 Equations of Motion

In this section, the dynamics of the launch vehicle is described accommodating six degrees of
freedom, in accordance with Wu [5]. Two parts are treated separately, the translational and
rotational movement. General equations are set up with respect to various reference frames.
Individual parts of the equations will be described in the following sections.
3.3. EQUATIONS OF MOTION 9

3.3.1 Translational Dynamics

The rst three degrees of freedom belong to translational movement. We draw on Newton's
second law in the form of
1 ~
~a = F (3.19)
m
As has already been stated in section 3.1, the total position of the launch vehicle is given
with respect to the inertial frame and can be described by vector

~r = xî + y ĵ + z k̂ (3.20)

The velocity and acceleration in inertial frame are then obtained one by one as

~v = ~r˙ = ẋî + ẏ ĵ + ż k̂ (3.21)

~a = ~v˙ = ẍî + ÿ ĵ + z̈ k̂ (3.22)

That is for the left side of the 3.19 equation. For the right side, we need to sum all the
external forces acting on the launch vehicle. Several sources are taken into consideration
- atmosphere, gravity and propulsion. The atmospheric (or aerodynamic) forces as well
as thrust will be most conveniently expressed in body frame. However, the gravity force
varies based on the location on the Earth and, as shown below, can be expressed in both
geographical and inertial frame. In equation 3.23 the inertial one is chosen. With the
appropriate transformations, we can express rst three dierential equations as

1 h ECI ~ i
~v˙ = Fatm + F~thr + F~g


TB (3.23)
m

Next three dierential equations are provided simply by the equation

~r˙ = ~v (3.24)

3.3.2 Rotational Dynamics

Although the 3 DOF simulation treating an object as a point mass is sucient during an
initial mission design, more comprehensive approach is assumed in this work. Three more
degrees of freedom are introduced in order to describe the launch vehicle's attitude. While
its position is calculated with respect to the ECI inertial frame, the attitude for simulation
purposes is measured with respect to the initial position of the spacecraft on the launch pad,
i.e. the EG reference frame. This way the yaw, pitch and roll angles are clearly dened.

The Euler's second law can be written as

M ~˙
~ =L (3.25)

where ~
M is the sum of external moments/torques acting on the body and ~
L represents the
angular momentum, in the simplest case about the center of gravity of the body. It can be
further deduced that
~ = Jˆ · ω
L ~ (3.26)
10 CHAPTER 3. MATHEMATICAL MODEL

where ω
~ represents total angular velocity of the body and Jˆ is the moment of inertia express-
ing how much torque has to be exerted on the body to reach a desired angular acceleration.
It can be obtained by integrating the second moments of mass of all body points with respect
to their distance d from the rotational axis over the whole body mass M.
Z
Jˆ = d2 dm (3.27)
M

When spread all over three main body axes x, y, z , tensor Jˆ consists of three moments
of inertia for each of the axes [3].

  
J11 J12 J 13 îb
Jˆ = îb ĵb
 
k̂b J21 J22 J23
   ĵb  (3.28)
J31 J32 J33 k̂b

Finally, the equation of motion for an object treated as a rigid body can be written in
the form of
M ~˙ + ω
~ = Jˆ · ω ~ × (Jˆ · ω
~) (3.29)

Now we will parse the sum of the external moments acting on the launch vehicle. When
proceeding from the previous part, three main sources of forces and torques are considered
acting on the spacecraft - atmosphere, gravity and propulsion. Only the gravitational force is
acting in the center of gravity of the spacecraft, and thus producing no torque. Moments from
atmospheric (aerodynamic) and propulsion forces have dierent reference points. Correct
reference frames have to be maintained. With respect to the moment of inertia tensor, the
body frame is chosen as the inertial one this time. The equation 3.29 can be rearranged, the
total external moment substituted and with combination of equation 3.13 we obtain another
set of seven equations of motion

h i
~˙ = (J)
ω ˆ −1 M~ atm + M
~ thr − ω
~ × (Jˆ · ω
~) (3.30)

    
q̇0 0 −p −q −r q0
q̇1  1 p 0 r −q  q1 
 =    (3.13 revisited)
q̇2  2 q −r 0 p  q2 
q̇3 r q −p 0 q3
3.4. SUBSYSTEM MODELS 11

3.4 Subsystem Models

Two sets of total thirteen equations of motion were developed throughout the previous sec-
tion. First one deals with forces and launch vehicle's translational movement in space. The
second set considers moments exerted by external forces on spacecraft and their impact on
attitude dynamics. Each of the force and moment sources is described in this section.

3.4.1 Gravity

Planet Earth is often pictured as an ideal sphere. However, the reality lies far from that. Our
planet is attened on the poles and thus resembles more of an ellipsoid. In fact, the proper
term used for the Earth's shape is called geoid. One of the latest geometric estimations of the
Earth's shape is the World Geodetic System 1984 model [6], which is in its simplied form
used throughout this work. Our motivation is to describe the variations in gravitational
acceleration ~g magnitude based on location around the Earth. The WGS84 ellipsoid is
dened by several constants summarized in the table 3.1 below.

Quantity Symbol Value

semi-major axis a 6,378,137.0 m
mean radius Rm 6,371,008.8 m
attening f 1/298.257223563
eccentricity e 0.08181919084
gravitational constant µ 3,986,004.418 m3 s−2

Table 3.1: WGS84 model dening constants

The geographic reference frame introduced in section 3.1.3 works with geocentric latitude
Φc . In order to determine gravitational acceleration, we need to introduce geodetic latitude
Φg which is the value commonly seen on maps. The geocentric latitude represents the angle
between equatorial plane and the position on spherical Earth surface. On the other hand,
geodetic latitude is dened as the angle between equatorial plane and plane tangent to the
ellipsoid Earth surface. Relation between geocentric and geodetic latitudes can be quite
easily gured out.
tan Φc
tan Φg = (3.31)
(1 − f )2
12 CHAPTER 3. MATHEMATICAL MODEL

The gravitational acceleration depending on geodetic latitude and altitude above the
surface can be then expressed in EG by employing colatitude angle Υ = 90 − Φc as follows
[7].
"
µ 3  R0  2  R 3
0
gr = − 2 1 − J2 (3 cos2 Υ − 1) − 2J3 cos Υ(5 cos2 Υ − 3)
r 2 r r
# (3.32)
5  R0 4 4 2
− J4 (35 cos Υ − 30 cos Υ + 3)
8 r
" #
µ  R0  2 1  R0  2
gΥ = − 3 2 sin Υ cos Υ J2 + J3 sec Υ(5 cos Υ − 3)
r r 2 r
" # (3.33)
5  R 2
0 2
+ J4 (7 cos Υ − 3)
6 r

Ji are so called oblateness terms estimating the geoid shape: J2 = 1.0826 × 10−3 ,
where
J3 = −2.54 × 10−6 and J4 = −1.61 × 10−6 and reference ellipsoid radius is approximated by
equation " #
e 5
R0 = a 1 − (1 − cos 2Φg ) + e2 (1 − cos 4Φg ) (3.34)
2 16
Moreover, gravitational acceleration can be also expressed in ECI inertial frame using
Cartesian coordinates, which may be useful during latter simulation. Terms higher than J2
are neglected here [7]. If the total gravitational acceleration is represented by vector

~g = gx î + gy ĵ + gz k̂ (3.35)

then its individual components can be calculated as

" " ##
µ 3  R0 2  z 2 x
gx = − 2 1 + J2 1−5 (3.36)
r 2 r r r
" " ##
µ 3  R0 2  z 2 y
gy = − 2 1 + J2 1−5 (3.37)
r 2 r r r
" " ##
µ 3  R0 2  z 2 z
gz = − 2 1 + J2 3−5 (3.38)
r 2 r r r

The total gravitational acceleration is for lower altitudes almost directly proportional to
the altitude.

Finally, gravitational force aecting the spacecraft of total mass m is established.

   
Fgx gx
F~g = Fgy  = m gy  (3.39)
Fgz gz

Gravitational force has its point of application in the center of gravity of the launch
vehicle. Thus, it produces no torque on the launch vehicle about the c.g. which is the origin
of our body frame.
3.4. SUBSYSTEM MODELS 13

3.4.2 Atmosphere

This work is focused on a lift-o phase of ight in the Earth's atmosphere which cannot be
compared to a free space ight. Atmosphere plays a signicant role in the real ight and
thus must be taken into consideration in the mathematical model.

As the launch vehicle ascends upward, it ies through several distinct layers of atmo-
sphere. Each of them has unique properties and the rate of change in pressure, temperature
or air density varies with altitude. Primarily, air density characteristics is crucial for de-
scribing forces acting on the spacecraft, as we will see later. Simplied calculations based on
the U.S. Standard Atmosphere 1976 model [8] are introduced here. The model itself uses a
very broad set of equations and tables to precisely describe atmospheric variations over an
altitude and its implementation is beyond our topic. Therefore, some simplications have
been made. However, the work can always be built on and expanded with more precise
models.

The air pressure is approximated by equation

  gM
Lh RL
p = p0 1 − (3.40)
T0

where p0 = 101325 P a is the sea level pressure, L = 0.0065 Km−1 is the molecular scale
temperature gradient, h [m] is the the altitude above ground, T0 = 288.15 K is the sea
level temperature, g [ms
−2 ] is the gravitational acceleration magnitude at current position

obtained from gravity model, M = 0.028964 kgmol

−1 is the molar mass of air and R =

8.31447 JK −1 mol−1 is the universal ideal gas constant. Note that these variables are not
constant in real life and vary slightly with altitude. Nevertheless, for the purpose of our air
pressure model, they are considered to be constant.

The dependence of the air pressure on altitude above the Earth's surface with general
g = 9.80665 ms−2 is plotted in gures 3.4 and 3.5 on the next page.
14 CHAPTER 3. MATHEMATICAL MODEL

80

70

60

50
h [km]

40

30

20

10

0
0 20 40 60 80 100 120
p [kPa]

Figure 3.4: Air pressure as a function of altitude up to 80 km

1000

900

800

700

600
h [km]

500

400

300

200

100

0
10 -6 10 -4 10 -2 10 0 10 2
p [kPa]

Figure 3.5: Air pressure as a function of altitude up to 1000 km

3.4. SUBSYSTEM MODELS 15

The temperature varies abruptly over several distinct atmospheric layers up to the alti-
tude of 120 km and then follows exponential curve. The equations for each of the layers are
fully described below [8].

ˆ For altitude from sea level up to h = 86 km, the temperature can be approximated by
the equation
T = T0b − Lb (h − hb ) (3.41)

where L is picked from the table 3.2 below based on the current altitude for each of the
atmospheric layers and T0 is simply the temperature at the borders between obtained
from previous calculations; for b = 0 is T0 = 288.15 K .

b hb [km] L [K]
0 0 −6.5
1 11 0.0
2 20 +1.0
3 32 +2.8
4 47 0.0
5 51 −2.8
6 71 −2.0
7 86 −

Table 3.2: Constants dening atmospheric layers

ˆ For altitude from 86 km to 91 km, the temperature remains constant at value

T = 186.8673 K (3.42)

ˆ For altitude from 91 km up to 110 km, the temperature is approximated as an elliptic

curve expressed by
" #1/2
 h − 91 2
T = 263.1905 − 76.3232 1 − (3.43)
−19.9429

where all the values used are dened constants.

ˆ For altitude from 110 km to 120 km the temperature is governed by the equation

T = 240 + 12(h − 110) (3.44)

ˆ Finally, for altitude from 120 km up to 1000 km which is more than sucient for our
purposes, the temperature has an exponential form of

T = Tinf − (Tinf − 360) exp(−0.01875ζ) (3.45)

16 CHAPTER 3. MATHEMATICAL MODEL

where Tinf = 1000 K and ζ is dened with the help of r0 = 6.356766 × 106 km as

 r + 120 
0
ζ = (h − 120) (3.46)
r0 + h

We have fully described the temperature model as stated in U.S. Standard Atmosphere
1976 model [8]. The dependence of temperature on altitude above the Earth's surface with
general g = 9.80665 m/s2 is plotted in gures 3.6 and 3.7 below.

80

70

60

50
h [km]

40

30

20

10

0
180 200 220 240 260 280 300
T [K]

Figure 3.6: Air temperature as a function of altitude up to 80 km

3.4. SUBSYSTEM MODELS 17

1000

900

800

700

600
h [km]

500

400

300

200

100

0
200 300 400 500 600 700 800 900 1000 1100
T [K]

Figure 3.7: Air temperature as a function of altitude up to 1000 km

The air mass density can now be calculated using pressure and temperature values at
certain altitude as
pM
ρ= (3.47)
RT
The dependence of air mass density on altitude above the Earth's surface with general
g = 9.80665 m/s2 is plotted in gures 3.8 and 3.9 on the next page.
18 CHAPTER 3. MATHEMATICAL MODEL

80

70

60

50
h [km]

40

30

20

10

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
ρ [kgm-3 ]

Figure 3.8: Air mass density as a function of altitude up to 80 km

1000

900

800

700

600
h [km]

500

400

300

200

100

0
10 -8 10 -6 10 -4 10 -2 10 0 10 2
ρ [kg/m3]

Figure 3.9: Air mass density as a function of altitude up to 1000 km

3.4. SUBSYSTEM MODELS 19

With air density determined at any altitude, the forces acting on a spacecraft in the
atmosphere can be presented. When we speak about aerodynamic forces, we usually mean
lift and drag. Lift L is a force acting upward, normal to the direction of ight. Drag
D is pointing backward, in opposite to the velocity. However, when dealing with bodies
symmetrical along the main axis of rotation such as rockets or missiles, dierent set of forces
may be easier to employ. Those forces are called normal and axial and they are dened
in the body frame, which brings us some benets. The normal force N is acting upward,
normal to the îb body axis. Axial force A is pointing backward, along the main body axis îb .
A third force perpendicular to both of the previous forces is called side force S and remains
unchanged.

Relation between the two sets of forces is guided by an angle of attack α, the angle
between the relative air mass velocity and the main body axis îb , as can be seen in gure
3.10 below. The second angle determining the spacecraft's orientation towards the air mass
velocity is side slip angle β . The angle of attack is dened as
v 
airz
α = tan−1 (3.48)
vairx
and the side slip angle as
v 
airy
β = tan−1 (3.49)
vairx
Assuming that the air mass surrounding the vehicle is still in relation to the ground and
any disturbances such as wind are neglected, the relative velocity is equal to the vehicle's
velocity
~vair = ~v (3.50)

~
N ~
L

îb

~
D c.p.
α
~
A
~vair

Figure 3.10: Relation between lift & drag and normal & axial forces acting on a body

We can transform from the pair of drag and lift to normal and axial forces using the
equations

N = L cos(α) + D sin(α) (3.51)

A = −L sin(α) + D cos(α) (3.52)

20 CHAPTER 3. MATHEMATICAL MODEL

and vice versa

L = N cos(α) − A sin(α) (3.53)

D = N sin(α) + A cos(α) (3.54)

Aerodynamic forces act on the spacecraft at its center of pressure ( c.p.) and can be all
expressed in a similar form as

1 2
A = CA ρSvair (3.55)
2
1 2
S = CS ρSvair (3.56)
2
1 2
N = CN ρSvair (3.57)
2
where CN is the normal coecient, CA is the axial coecient and CS is the side coecient,
ρ [kgm−3 ] is the air mass pressure at current altitude, S [m2 ] is the reference area and ~vair
[ms−1 ] is the velocity of the spacecraft relative to the surrounding air mass.
The aerodynamic coecients are in reality not constant as we assume them to be. Their
dependency on Mach number, angle of attack and side slip angle needs to be determined
experimentally or by using complex Computational Fluid Dynamics simulations. For our
purposes, we can dene them as

CA = CA0 + αCAα + βCAβ (3.58)

CS = CS0 + βCSβ (3.59)

CN = CN0 + αCNα (3.60)

However, constant values CS0 and CN0 are neglected in further work.

All necessary parts were developed and we can now express the atmospheric aerodynamic
forces in body coordinates as

   
Fatmx −A
F~atm = Fatmy  =  S  (3.61)
Fatmz −N

In order to dene the moments exerted by aerodynamic forces on the launch vehicle, the
location of the center of pressure as their point of reference needs to be claried. The body
frame has its origin located at the center of gravity and to maintain rocket stability the center
of pressure has to be located behind it - negative x coordinate in the body reference frame.
 T
Let's dene position of c.p. in the body coordinates as ~rc.p. = −xc.p. 0 0 . However,
the aerodynamic moments about the c.g. are usually called yaw, pitch and roll moments

   
Matmx Mroll
~ atm
M = Matmy  = ~rc.p. × F~atm = Mpitch  (3.62)
Matmz Myaw
3.4. SUBSYSTEM MODELS 21

and they are expressed in a similar way as aerodynamic forces

1 2
Mroll = CM r ρSvair l (3.63)
2
1 2
Mpitch = CM p ρSvair l (3.64)
2
1 2
Myaw = CM y ρSvair l (3.65)
2
where l [m] is the reference body length and the aerodynamic coecients dependent on
the angle of attack, side slip angle and vehicle's angular velocity are described as

CM r = pCM rp (3.66)

CM p = αCM pα + qCM pq (3.67)

CM y = βCM yβ + rCM yr (3.68)

It has been proven that when the aerodynamic forces are applied at certain location along
the body, the magnitudes of aerodynamic moments remain almost constant, independently
of the angle of attack. Such position is called the aerodynamic center [9]. The position of
c.p. has to be determined experimentally from the relation 3.62.

3.4.3 Fin Control

One of the options how to steer a launch vehicle is implementing a n control. Fins are used
commonly on rockets for stabilization during ight. On our water rocket model described in
greater detail in chapter 5, two sets of ns are used. There are four orthogonal ns at the
bottom for passive stabilization and four orthogonal controlled ns close to the rocket head
for active ight control. The ns arrangement can be explored on the following sketch.

Figure 3.11: Water rocket model ns arrangement [10]

22 CHAPTER 3. MATHEMATICAL MODEL

Position of all four steerable ns A, B, C and D is dened by vectors from rocket's c.g. to
the geometric center of each n ~rf inA , ~rf inB , ~rf inC and ~rf inD . The aerodynamic forces acting
on ns are similar to the three dimensional forces acting on the whole spacecraft. However,
due to the ns arrangement around the body, one of the three forces can be neglected while
one n is always hidden behind the launch vehicle's body. Thus, for each n we obtain a
pair of forces - axial and side or axial and normal. Specically, for ns A and C we take into
consideration axial and normal forces and for ns B and D axial and side ones.

Although the ns' aerodynamic forces depend on the air velocity and pressure which
varies with altitude, the quantities are assumed to be constant for simulation purposes. The
forces vary only with the n turn angle. Apart from that, the equations are equivalent to
those developed above, see equation 3.61.

3.4.4 Thrust

Enough of forces hindering our launch vehicle! What drives it up to the sky is propulsion.
There are plenty of propulsion types used in aerospace industry but so far only the rocket
engines has been suitable for big launch vehicles due to their high thrust level. Simply said,
they work by expanding high pressurized gas through a converging-diverging nozzle and thus
accelerating propellant to hypersonic velocities [11].

The thrust of such rocket motor is basically given by the equation

T = ṁVe + (pe − pa )Ae (3.69)

where T [N ] is the thrust force,ṁ [kgs−1 ] is the propellant mass ow rate, Ve [ms−1 ] is the
exhaust velocity at nozzle exit, pe [P a] is the exhaust pressure at nozzle exit, pa [P a] is the
ambient pressure equal to
2
zero in vacuum and Ae [m ] the nozzle exit area which is usually
xed.

Thrust is by no way constant as it depends on ambient pressure which varies with altitude.
The higher the rocket ies, the higher thrust we get with some limitations. In order to
compare power of dierent rocket engines, equivalent exhaust velocity is introduced

T
Veq = (3.70)
ṁ
as well as specic impulse describing the rate of thrust which is gained from an engine
for unit propellant mass ow rate in sea level gravity g0 . Specic impulse is the primary
measurement of rocket performance and can always be found in the engine's specications.

T Veq
Isp = = (3.71)
ṁg0 g0

The thrust vectoring is considered the main control mechanism in our model. Engine
nozzle is attached to gimbals which can be turned around slightly in all directions, steering
the rocket. Two independent angles are responsible for the nozzle attitude - δψ represents

rotation of nozzle about yaw axis k̂b and δθ represents rotation of nozzle about pitch axis ĵb
in body coordinates as seen in gure 3.12. Thus only pitch and yaw angles can be controlled
by thrust vectoring.
3.4. SUBSYSTEM MODELS 23

îb

c.g.

xgim
ĵb
Fthrx k̂b
δθ δψ
Fthrz
T
Fthry

Figure 3.12: Thrust vectoring schema with yaw and pitch control angles

   
Fthrx T cos δψ cos δθ
F~thr = Fthry  =  T sin δψ  (3.72)
Fthrz T sin δθ
Thrust force has its point of action placed where the gimbaled nozzle is connected to
the launch vehicle. Such location can be expressed in the body reference frame by vector
 T
~rgim = −xgim 0 0 . Total moment exerted on the spacecraft by thrust force is zero
when no thrust vectoring is in action. In other cases the moment is non-zero.

 
Mthrx
~ thr
M = Mthry  = ~rgim × F~thr (3.73)
Mthrz
Chapter 4

Simulation
Based on the four dierential equations fully describing both translational and rotational
movement of the launch vehicle a computer based simulation model can be developed. The
powerful MATLAB has been employed with combination of Simulink graphical modeling
environment. The main nonlinear equations 3.23, 3.24 and 3.30, 3.13 are repeated here for
reference.

1 h ECI ~ i
~v˙ = Fatm + F~thr + F~g


TB (3.23 revisited)
m

~r˙ = ~v (3.24 revisited)

h i
~˙ = (J)
ω ˆ −1 M~ atm + M
~ thr − ω
~ × Jˆ · ω
~ (3.30 revisited)

    
q̇0 0 −p −q −r q0
q̇1  1 p 0 r −q  q1 
 =    (3.13 revisited)
q̇2  2 q −r 0 p  q2 
q̇3 r q −p 0 q3

4.1 Simulink Model

At rst, each of the subsystems has been modeled and, wherever possible, a block from
Simulink library has been used for greater precision, mostly fromAerospace Blockset. For
Zonal Harmonic Gravity Model block
example, gravitational acceleration is obtained from
and atmospheric data from COESA Atmosphere Model block (1976 U.S. Standard Atmo-
sphere). Further mathematical relations are then modeled using standard building blocks.
The launch vehicle's ight complex Simulink model is closely interconnected making any
debugging quite tricky. A couple of screen shots of various parts of the whole design are
presented below in gures 4.1 and 4.2.

24
4.1. SIMULINK MODEL 25

Figure 4.1: Simulink model - main scheme

Figure 4.2: Simulink model - transformations

The Simulink model has been designed with the thought of possible future extensions,
and thus should be transparent and easily editable. Setting of thrust vectoring and ns
control commands has to be done in the Simulink model itself (function blocks thrust ...
ctrl and fins ctrl). The simulation run time is set to 200 s by default, while the engines
burn time tburn is 162 s. However, there is no problem of adjusting the simulation run time.
26 CHAPTER 4. SIMULATION

4.2 Input Data

Building a numerical model entails the need for real data. A launch vehicle's ight cannot
be simulated without knowing for example the body mass. In order to ensure some level of
model's reliability, wide research has been conducted resulting in a partial set of numerical
values summarized in the table 4.1 bellow. Most of the data has been taken from the
revolutionary Falcon 9 FT launch vehicle of the SpaceX private company. However, some
values are simply not available to public, and thus had to be estimated. Those are signed
with ∼. Real world values are mostly taken from [12], [13] and estimations are based on
data obtained from [14].

Quantity Symbol Value ∼

LV mass m 570, 000 kg
LV length d 70 m
 T
Position of gimbal ~rgim −21 0 0 ∼
Main moments of inertia J11 , J22 , J33 1.4 × 105 , 3.2 × 107 , 3.2 × 107 ∼
Aerodynamic reference area S 21 m2 ∼
Reference length l 3.24 m ∼
Axial force coecient CA 0.2 ∼
Axial force coecient for α CAα 0.05 ∼
Axial force coecient for β CAβ 0.05 ∼
Side force coecient CS 0 ∼
Side force coecient for β CSβ 0.9 ∼
Normal force coecient CN 0 0 ∼
Normal force coecient forα CNα 0.9 ∼
Roll moment coecient for p CM rp 0.8 ∼
Pitch moment coecient for α CM pα 0.05 ∼
Pitch moment coecient for q CM pq 0.8 ∼
Yaw moment coecient for β CM yβ 0.05 ∼
Yaw moment coecient for r CM yr 0.8 ∼
Total thrust T 9 × 845000 N
Specic impulse Isp 282 s
Burn time tburn 162 s
Exit pressure pe 9 × 105 P a ∼
Nozzle exit area Ae 0.97 m2

Table 4.1: Simulation model input data

Unknown values have been estimated to the best of my knowledge and belief with respect
to meaningful outcomes of the simulation as a whole. It should be mentioned that the results
highly depend on the initial setup of the numerical values, and thus dierent outcomes may
be obtained from dierent setups. The Simulink model has to be tuned extensively in order
to represent the real world.
4.3. OUTCOMES 27

Several assumptions have been made along the simulation process. Only moments of in-
ertia about the main rotational axes are considered in the Simulink model. The aerodynamic
coecients from the table 4.1 above are considered to be constants optimized for velocities
around 600 m/s. The launch vehicle's body is considered to be rigid.

4.3 Outcomes
Apart from exploring the mathematical equations and code les itself, the best way how to
get acquainted with the model is to investigate some results. Flight data from the simulation
are presented here in the form of 2D and 3D plots. In order to visualize more features of the
model, thrust vectoring is utilized.

Most if not all of the quantities from Simulink model are saved to MATALB's workspace
for further analysis. The variables are usually named as sim_(quantity name), for exam-
ple sim_phi_c. Each particular case needs to be checked prior to manipulation with the
variables.

Our mathematical model encapsulates several accurate sub models, for example the atmo-
spheric model or thrust model, the output of which can be seen in gure 4.3 bellow.

6
×10
8.6

8.4

8.2
T [N]

7.8

7.6

7.4
0 20 40 60 80 100 120 140 160 180
time [s]

Figure 4.3: Total thrust of all nine rocket engines over time (tburn = 162 s)

However, it was found that the launch vehicle's behavior varies extensively over the whole
ight. Due to the changes in air density, rocket's thrust and velocity, the ight cannot be
trustworthily simulated with a single set of aerodynamic coecients that highly inuence the
28 CHAPTER 4. SIMULATION

vehicle's dynamics. The coecients are in reality far from being constants as we perceive
them for our purposes. Their values depend heavily on velocity (Mach number). Such
complexity is beyond the capabilities of this work. If we perceive aerodynamic coecients
as constants, they are applicable only for some velocities. For simpler simulation cases, the
thrust was set equal to dragging forces and velocity is maintained at a constant value. The
variations in air density were also neglected. Several signicant plots demonstrating the
launch vehicle's test ight can be seen in gures 4.4 to 4.7 bellow.

Figure 4.4: Flight trajectory of launch vehicle

4.3. OUTCOMES 29

×10 -4

δθ
δ
ψ
4

2
δ [rad]

-2

-4

-6
0 50 100 150 200
time [s]

Figure 4.5: Thrust vectoring over time

4.5
φ - roll
4 θ - pitch
ψ - yaw
3.5

2.5
YPR [deg]

1.5

0.5

-0.5
0 50 100 150 200
time [s]

Figure 4.6: Vehicle's Euler angles over time

30 CHAPTER 4. SIMULATION

×10 4
2

-1

-2
M [Nm]

-3

-4

-5

-6 Mr
Mp
-7 My

-8
0 50 100 150 200
time [s]

Figure 4.7: Aerodynamic moments acting on the launch vehicle over time

4.4 Linear Model

Nonlinear mathematical model has been developed in the previous section and serves as a
good source of equations for the computer simulation. In order to assess the dynamics of
the rigid body launch vehicle, a linear state-space model is introduced here in the form of

ẋ = Ax + Bu (4.1)

y = Cx + Du (4.2)

The state variables, inputs and outputs of our linear system are dened as

 
x = ẋ ẏ ż x y z p q r q0 q1 q2 q3 (4.3)
 
u = δθ δψ (4.4)

y=x (4.5)

The nonlinear model has been linearized with the help of MATLAB's linmod function
for velocity of 600 m/s. The state-space matrices are truly extensive so they are presented
here only in their substituted versions. It can already be stated that matrix C is an identity
matrix and D is a zero matrix. The 0.0 values represent numbers of a small order but still
not equal to zero.
4.4. LINEAR MODEL 31

 
0.0 0 0.0 −0.0 0 0.0 0 0 0 −19.63 0 0 0
0 0 0 0 0.0 0 0 0 0 0 0 0 −19.63
 
0 0 −0.46 0.0 0 −0.0 0 0 0 0 0 19.63 0 
 
1 0 0 0 0 0 0 0 0 0 0 0 0 
 
0 1 0 0 0 0 0 0 0 0 0 0 0 
 
0 0 1 0 0 0 0 0 0 0 0 0 0 
 
A= 0 0 0 0 0 0 −65.32 0 0 0 0 0 0 
 
0 0 −0.01 0 0 0 0 −0.37 0 0 0 0 0 
 
0 0 0 0 0 0 0 0 −0.37 0 0 0 0 
 
0 0 0 0 0 0 0 0 0 0 0 0 0 
 
 
0 0 0 0 0 0 0.5 0 0 0 0 0 0 
 
0 0 0 0 0 0 0 0.5 0 0 0 0 0 
0 0 0 0 0 0 0 0 0.5 0 0 0 0
(4.6)

 
0 0
 10.88 0 
 
 0 10.88
 
 0 0 
 
 0 0 
 
 0 0 
 
B= 0 0  (4.7)
 
 0 6.19 
 
−6.19 0 
 
 0 0 
 
 
 0 0 
 
 0 0 
0 0

If we compare the nonlinear model with the linear one, the results are consistent. With
a short one second thrust vectoring impulse for δθ , the responses are almost identical as
demonstrated in gures 4.8 and 4.9.
32 CHAPTER 4. SIMULATION

1
nonlin model
0 lin model

-1

-2

-3
θ [deg]

-4

-5

-6

-7

-8

-9
0 50 100 150 200
time [s]

Figure 4.8: Nonlinear and linear model pitch angle θ responses comparison

0.02
nonlin model
0.01 lin model

-0.01
q [rad/s]

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07
0 50 100 150 200
time [s]

Figure 4.9: Nonlinear and linear model angular velocity q responses comparison
4.4. LINEAR MODEL 33

The linear model may serve as an initial source for launch vehicle's control laws analysis
and regulators design.

4.4.1 Sensitivity Analysis

As stated above, the vehicle's dynamics highly vary with velocity and other parameters. In
order to qualify such relations, a basic sensitivity analysis has been conducted.

The distribution of poles resembles an airplane's dynamics with one real stable poles, one
oscillatory mode and one integrator [15]. Three separate quantities have been analyzed. At
rst, the launch vehicle system has been linearized with respect to changing velocity. Note
the pair of complex conjugate poles causing oscillations during the launch vehicle's ight.
They are caused by aerodynamic forces and moments returning the spacecraft into vertical
position after any intended or random disturbance. A slight tendency of the poles to become
more stable as the vehicle speeds up can be observed in gure 4.10. Moreover, from a certain
boundary velocity, the oscillations of complex conjugate pair of poles is slowing down. Note
that the aerodynamic coecients were set to the same constant values for all velocities. Real
stable poles at −75, −20, −5 and −1 are omitted in the gure 4.10.

Pole-Zero Map
0.25 640
0.965 0.935 0.88 0.78 0.62 0.35
0.2

0.15 0.986 320

Imaginary Axis (seconds )

0.1
-1

0.997
0.05 160
Velocity [m/s]
0
0.8 0.6 0.4 0.2

-0.05 80
0.997
-0.1

-0.15 0.986 40

-0.2
0.965 0.935 0.88 0.78 0.62 0.35
-0.25 20
-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1
-1
Real Axis (seconds )

Figure 4.10: Poles and zeros of the system in relation to velocity

34 CHAPTER 4. SIMULATION

The gure 4.11 represents the linearization with respect to the main moments of inertia
with constant vehicle's velocity set to 400 m/s. Although only the J11 moment is displayed
in the legend, another two main moments J22 and J33 have been varied proportionally. It
is observed that as the value of the moments rises, the complex poles' oscillation velocity
slackens and the poles become more stable which corresponds to common sense. The higher
the moment of inertia is, the harder to accelerate the object is. The moment's value used
throughout the simulations is in the middle of the legend range. Real stable poles at −430,
−220, −110, −50 and −25 are omitted in the gure 4.11.

Pole-Zero Map
2.6e5
0.955 0.91 0.84 0.74 0.56 0.3

0.3
0.982
1.3e5
0.2
Imaginary Axis (seconds )
-1

0.995
0.1
6.4e4

J 11 [kgm2]
1.2 1 0.8 0.6 0.4 0.2
0

3.2e4
-0.1
0.995

-0.2
1.6e4
0.982
-0.3

0.955 0.91 0.84 0.74 0.56 0.3

8e3
-1.2 -1 -0.8 -0.6 -0.4 -0.2 0
-1
Real Axis (seconds )

Figure 4.11: Poles and zeros of the system in relation to moment of inertia
4.4. LINEAR MODEL 35

The plot 4.12 shows the relation between launch vehicle's poles and the pitch moment
aerodynamic coecient for angle of attack with constant vehicle's velocity set to 400 m/s.
As the coecient's value rises the rate of system's oscillations increases and its poles move
to the edge of stability. That has been validated by the simulation model. The value of
CM pα used in the simulations is usually set to 0.05. Real stable poles at −14 is omitted in
the gure 4.12.

Pole-Zero Map
0.09
0.96 0.92 0.85 0.74 0.56 0.3
0.15
0.982
0.075
0.1
Imaginary Axis (seconds -1)

0.996
0.05
0.06

α
CMp
0
0.7 0.6 0.5 0.4 0.3 0.2 0.1

0.045
-0.05
0.996

-0.1
0.030
0.982
-0.15

0.96 0.92 0.85 0.74 0.56 0.3

0.015
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Real Axis (seconds -1)

Figure 4.12: Poles and zeros of the system in relation to pitch moment coecient for α
36 CHAPTER 4. SIMULATION

4.5 GUI
One of the ideas behind building a functional launch vehicle ight simulation model has been
the intention to present a simple visualizations to students interested in aerospace eld. Thus
a MATLAB GUI has been designed as a part of the thesis.

The GUI is decomposed into three tabs. The rst tab in gure 4.13 contains launch ve-
hicle's trajectory visualization with ground trace of the trajectory. Next, a simple 3D rocket
model is used to demonstrate the YPR angles. The 3D graphics has been downloaded from
www.thingiverse.com under the name "Classic Rocket" created by user called "Botmaster"
[16]. Visualizations are accompanied by launch vehicle's current position, velocity and values
of its YPR angles.

Figure 4.13: Launch Vehicle Simulation MATLAB GUI - rst tab

In the second tab in gure 4.14 a set of life updating atmospheric data is presented. Plots
of altitude over time and ambient pressure, temperature or air density over altitude can be
found there.

The third tab visible in gure 4.15 bellow represents the thrust over time updating plot.
4.5. GUI 37

Figure 4.14: Launch Vehicle Simulation MATLAB GUI - second tab

Figure 4.15: Launch Vehicle Simulation MATLAB GUI - third tab

Chapter 5

Water Rocket Experiment

The aim of a practical experiment is to demonstrate how well mathematical model reects
reality. We will join an already established team from the Faculty of Electrical Engineering
at the Czech Technical University under the guidance of Ing. Petr Ko£árník, Ph.D. Their
objectives are very extraordinary indeed. With a couple of colleagues they have developed
a launch vehicle reactive water model ELEKTRON I.

As stated in their informational leaet [17], ELEKTRON I is designed as plastic rocket

with propulsion provided by water and pressurized air due to safety reasons. The head of
the model is equipped with parachute and ejection mechanism for safe landing. Nose cone
encapsulates ight sensors (accelerometer and magnetometer) and control unit.

Although our simulation model has been designed for real world big launch vehicles, any
experiments with those are beyond our capabilities. The accessible water rocket model has
to be sucient. The original idea was to simulate a ight of ELEKTRON I in the Simulink
environment with set tilt sequence of control ns. Likewise, the real ELEKTRON I would
be launched with the identical preprogrammed control ns tilt sequence and ight data
including trajectory, velocity and acceleration would be collected. When performed several
times, enough data for the accuracy assessment of our simulation model and its possible
calibration should be obtained.

Unfortunately, due to uninuencable circumstances, we were unable to carry out the

described experiment in time. Nevertheless, data from former model's uncontrolled ight
has been acquired. Those data are used to calibrate the simulation model for the water
rocket model, to estimate unknown quantities and prepare it for further employment.

38
5.1. INPUT DATA 39

5.1 Input data

All available quantities dening the water rocket model are summarized in the table 5.1
below.

Quantity Symbol Value

Mass without water mempty 1134 g
Length l 2080 mm
Diameter d 110 mm
Tank volume Vtank 12.6 l
Initial water volume Vwateri 2, 7 l
Initial air pressure pairi 10 bar
Aerodynamic reference area S 8171 mm2
Drag coecient CD 0.4
MAX achieved altitude hmax 170 m
MAX achieved velocity vmax 45 ms−1
MAX achieved acceleration amax 38 ms−1
Main moments of inertia J11 , J22 , J33 109 , 106 , 109 gmm2

Table 5.1: Water rocket model input data

Other data necessary for the simulation had to be estimated. The water rocket's thrust
has been modeled with a simplied version of equation 3.69 where only the dierence between
exit and ambient pressures is considered. The only known aerodynamic coecient is the drag
coecient and the estimation of others cannot be veried due to lack of experimental data.
The estimated constants are summarized in the table 5.2 below.

Quantity Symbol Value

Exit pressure pe 1 × 106 P a
Nozzle exit area Ae 3.3 × 10−5 m
"Burn" time tburn 3s
Axial force coecient CA 0.4
Axial force coecient for α CAα 0.1
Axial force coecient for β CAβ 0.1
Side force coecient for β CSβ 0.2
Normal force coecient forα CNα 0.2
Roll moment coecient for p CM rp 0.3
Pitch moment coecient for α CM pα 0.01
Pitch moment coecient for q CM pq 0.3
Yaw moment coecient for β CM yβ 0.01
Yaw moment coecient for r CM yr 0.3

Table 5.2: Estimated water rocket model input data

40 CHAPTER 5. WATER ROCKET EXPERIMENT

5.2 Outcomes
The real ight data are compared to the calibrated Simulink model outcomes in the plots
below. A software issue has arisen during the test ight causing the accelerometer on board
of the water rocket to start measuring only after the rst second of ight which can be seen
in all the following gures from 5.1 to 5.3.

With slight tuning of the initializing engine dening constants, the developed simulation
model projects the rocket's altitude very accurately as can be seen in gure 5.1 below. While
the thesis is focused only on the lift-o phase of the ight, model's descent is not covered.

180
exp
160 sim

140

120
altitude [m]

100

80

60

40

20

0
0 1 2 3 4 5 6 7 8 9
time [s]

Figure 5.1: Water rocket experimental and simulation altitude results comparison
5.2. OUTCOMES 41

The biggest source of disturbances for the light water rocket model are denitely wind
gusts. While the simulation model does not include any wind submodel, the rocket's de-
ections cannot be satisfactorily modeled. However, at least a rough match has been found
between the experimental and simulation results for the vertical acceleration visible in gure
5.2 on the next page.

30
a x (exp)
a y (exp)
20
a z (exp)
a x (sim)

10 a y (sim)
acceleration [ms -2]

a z (sim)

-10

-20

-30
0 1 2 3 4 5 6 7 8 9
time [s]

Figure 5.2: Water rocket experimental and simulation acceleration results comparison
42 CHAPTER 5. WATER ROCKET EXPERIMENT

The mass of the water rocket model may not be equally spread causing the rocket to tilt
to one side. That is probably the case here while the angular velocity especially about the
z axis recedes from zero immediately after launch. Another source of disturbances is the
already mentioned wind. Due to these reasons, the experimental results could not have been
matched with the simulation outcomes. So only the experimental ight data are presented
in gures 5.3 bellow.

1.2
p
1 q
r
0.8

0.6
omega [rad/s]

0.4

0.2

-0.2

-0.4

-0.6

-0.8
0 1 2 3 4 5 6 7 8 9
time [s]

Figure 5.3: Water rocket experimental angular velocity results

Data from the onboard magnetometer are not presented here while it holds no signicance
for the developed simulation model.
Chapter 6

Results
Let's recapitulate what have been accomplished by this thesis and summarize the results.

Reference frames. Several reference frames have been partially adopted from literature
in order to maintain conformity with general conventions in aerospace industry. Some, such
as the Earth geographic frame, were on the other hand newly dened with the hope of
better meeting the project's specications. Individual transformation rotational matrices
were described and more complex transformations between frames can be built upon them.

Rotational kinematics. The rotational kinematics serves as a basis for further model-
ing and simulation while it provides us with launch vehicle's attitude in space determination
in terms of yaw, pitch and roll angles. Highly exible quaternions with the advantage of
no dangerous singularities and shorter computation time have been utilized instead of not
so suitable but easily imaginable Euler angles. Necessary relations between both forms are
provided.

Dynamics. Drawing from the Newton's and Euler's laws, translational and rotational
dynamic models have been developed with not fully specied forces and moments. Those are
described in detail in sections about the Earth gravitational eld, atmosphere and aerody-
namics and rocket propulsion and control. Several important assumptions have been made
along the way with the aim to keep the model's description clear but still precise enough.

Equations of motion. When combined together, set of thirteen equations of motion

surfaced. Those equations have been implemented in the MATLAB and Simulink simulation
of launch vehicle's ight. The model has been built with the thought of future expansion
and each subsystem can be easily altered for dierent utilization. Numerical values soon
became necessary for model validation. Some were obtained from published real launch ve-
hicles' specications, others had to be estimated and manually tuned. Probably the greatest
challenge of our simulation model is the partial uncertainty of the numerical quantities. As
we have learned along the way, each value can radically alter the simulation results.

Simulation. After the theoretical equations have been adjusted optimally for our model,
the Simulink model tuning could have begun. The process was demanding due to the absence
of complete set of launch vehicle dening numerical values. Meaningful model results have
been given priority to accurate constants. The model has been veried by multiple test ights.
Each of them with dierent set of thrust vectoring or ns control commands, distinct initial
values, and situations. We focused on comparing the present launch vehicle's behavior to

43
44 CHAPTER 6. RESULTS

theoretical assumptions. For example, it is clearly visible from the gures 4.6 and 4.7 how
the aerodynamic moments are returning rocket to its initial vertical position. The forces and
moments are actually acting as dampers on spring and the coecients can be assimilated to
spring constants which is in agreement with the theory. The modeled ight data are depicted
by the user friendly, life updating GUI for further accessibility.

Stability. The constant aerodynamic coecients also represent probably the biggest
drawback of our approach as they vary a lot with respect to vehicle's velocity in reality.
The aerodynamic forces and moments are likewise aected by the position of the center of
pressure and concept called the stability margin. While the position of c.p. has to be deter-
mined experimentally or using Computational Fluid Dynamics methods, stability margin is
a distance between c.g. and c.p. and tells us how stable the aircraft actually is. While we
see some instability in gure 4.10, the "constantness" of aerodynamic coecients could be
a reason. In other words, the rocket should be stable for all velocities with the right set of
aerodynamic coecients.

Water rocket. Although the initially designed practical experiment could not have been
executed, a set of already obtained ight data has been compared to the simulation model
outcomes. Wherever possible, the estimated constants have been veried with respect to the
real results and may serve as a basis for further water rocket aerodynamics assessment. What
had been seen as the greatest challenge, to readjust the simulation model intended for big
launch vehicles to small rocket models, actually appeared viable as can be stated for example
based on gure 5.1. However, both outer and inner dynamics of a water rocket are dierent
from a launch vehicle's. Should the water rocket be simulated more profoundly, specialized
Simulink subsystem models may be developed and greater set of ight data collected.
Chapter 7

Conclusion
Numerous voluminous books and studies have been written about rocketry and launch vehi-
cles ight laws. It is not possible to cover the problematics in its whole width and depth in
a single thesis. However, satisfactory mathematical model has been developed and platform
for simulating rocket ight described throughout this work.

In sum, all goals initially set in chapter 2 have been fullled. The equations describing
rocket motion have been developed and utilized in the simulation model development. The
Simulink model has been veried and tested in multiple situations, then. A simple yet
informative GUI has been introduced to visualize the simulation results in time. Beyond the
thesis' guidelines, the original nonlinear model has been linearized and a basic sensitivity
analysis has been performed. In the end, a practical experiment has been documented and
our model calibrated with respect to the ight data results.

There is a plenty of space for future research. The Simulink model could be validated
by broader set of experimental water rocket ights with ns control involved. The linearized
model may serve as a basis for control laws development and automatic regulators design.
Further research may look into the rocket's aerodynamics in hypersonic velocities and ex-
perimentally determine the values of aerodynamic coecients. As has been already stated,
the simulation model is highly expandable and versatile. More precise subsystems can be
implemented, multiple stage or exible body rockets introduced. The developed simulation
model can serve as a platform for various cases.

How else to round o a rocketry thesis than by an apt phrase "Per aspera ad astra"!

45
Appendix A

Contents of the attached CD

ˆ rocketsim.slx - Simulink model

ˆ init.m - initialization of simulation

ˆ analyze.m - analysis of simulation results

ˆ gui.m - visualization of simulation results, GUI

ˆ tiltFin.m - ns control used by rocketsim.slx

ˆ rotX.m, rotY.m, rotZ.m - rotation matrices used by rocketsim.slx

ˆ omegaJ.m - part of Euler equations used by rocketsim.slx

ˆ ECF2EG.m - transformation from ECF to EG frame used by rocketsim.slx

ˆ rocket.stl - rocket 3D model used for visualization by gui.m

ˆ README.txt - readme le, instruction manual for the simulation

46
Bibliography
[1] Peter H. Zipfel. Modeling and Simulation of Aerospace Vehicle Dynamics. 2nd. AIAA,
2007. isbn: 9781563478758.
[2] Bulletin B 344. 1.10.2016.
International Earth Rotation and Reference System Service.

[3] Bong Wie. Space Vehicle Dynamics and Control. 2nd. AIAA, 2008. isbn: 9781563479533.

[4] James Diebel.  Representing attitude: Euler angles, unit quaternions, and rotation
vectors. In: Matrix 58 (2006).
[5] Wei Du. Dynamic modeling and ascent ight control of Ares-I Crew Launch Vehicle.
Iowa State University, 2010.

[6] National Imagery and Mapping Agency. World Geodetic System 1984. 3rd. NIMA,
3.1.2000.

[7] Robert F. Stengel. Flight Dynamics. Princeton University Press, 2004. isbn: 0691114072.
[8] National Oceanic Atmospheric Administration, Department of the Air Force, and
NASA. U.S. Standard Atmosphere, 1976. 1976.
[9] The Beginner's Guide to Aeronautics. url: https://www.grc.nasa.gov/WWW/K-
12/airplane/ac.html (visited on 09/04/2017).

[10] Filip Svoboda. Research report: Water rocket project. Czech Technical University, 2015.
[11] Warrick Miller. Lecture notes: Space Vehicle Design (MECH ENG 3104). The Univer-
sity of Adelaide, 2016.

[12] Space Exploration Technologies Corp.Falcon 9 Launch Vehicle Payload User's Guide,
Rev 2. SpaceX, 2015.
[13] Falcon 9 FT. url: http : / / spaceflight101 . com / spacerockets / falcon - 9 - ft/
(visited on 09/04/2017).

[14] Saturn V Launch Simulation. url: http://www.braeunig.us/apollo/saturnV.htm

(visited on 09/04/2017).

[15] Martin Hromcik. Presentation slides: Flight dynamics. Czech Technical University,
2013.

[16] Classic Rocket by Botmaster. url: http : / / www . thingiverse . com / thing : 580192
(visited on 10/04/2017).

[17] Petr Kocarnik. Information leaet: ELEKTRON I. Czech Technical University, 2015.

47