Beruflich Dokumente
Kultur Dokumente
CSDL-T-1062 TRAJECTORY OPTIMIZATION FOR AN ASYMMETRIC LAUNCH VEHICLE by Jeanne Marie Sullivan
June 1990
i:_R 125
Aq p
ASV_'_
T_,_tC
LAUh_CH
V_ICL_]
M._.
The Charles
Stark Draper
02139
Laboratory,
Inc.
TRAJECTORY ASYMMETRIC
AN
Jeanne
B.S. Physics, Carnegie
Marie
Mellon
Sullivan
University, (1988)
Submitted to in Partial
MASSACHUSETTS
OF TECHNOLOGY
Jeanne
Marie
Sullivan,
to M.I.T. and the C.S. Draper distribute copies of this thesis or in part.
Signature
of
Certified
Approved
Accepted
Table
Chapter
of Contents
Page
1:
3 3 4
Background. Method
..............................................................
Ovcrvicw
of A.L.S.
Framcs
2.7 3:
Constraints Design,
J_
Chapter 3.3 In-Flight 3.3.1 3.3.2 3.4 Sensing 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.5 Pre-Launch 3.5.1 3.5.2 3.5.3 3.5.4 3.5.5 3.6
4"
Page Guidance Phase Phase and Sensed Angular Angle Dynamic Acceleration Trajectory Introduction Phase Phase Phase Phase of 1 and and Control .................................. and Control Control ................... 31 32 35 38 38 39 40 43 43 45 45 46 46 47 50 50 54 54 54 58 ....................... 62 73
2 Guidance and
..........................
.......................................... ...........................................
Direction Design
.............................................. Rise ................................... Maneuver of Attack Explicit Design ............................. Profile ....................... ...................
Guidance
Automation Simulation
...............................
................................................... ......................................................
Order Control
Model
............................................
Simulation
......................................................
Chapter
Page
5:
................................................. ................................................. of Numerical Gradient Along Approximation Method Search Optimization Methods ........
...................................
6:
Trajectory Trajectory
.....................
..............................
7:
Recommendations
...................................
Conclusions Recommendations
............................................... ........................................
iii
List
Figure
of Figures
Page
A.L.S. Thrust
Configuration Model
...................................................
8 10
Relationship Frames
................................................................... and Rates Orientation Free Body Body Frame Relationship Frame With Pitch Plane .............
2.4 2.5 2.6 2.7 2.8 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 4.1 4.2 4.3
In Body
Profile
.....................................
Representation Representation of Angle Profile for Design Coordinate Parameters Flow for
Trajectory Process
.......................
Trajectory Simulation
................................ .............................
Frames
Orientation Simulation
iV
Figure
Page
4.4
Relationship Predictive
Between Simulation
Inertial
Reference
Frames
of Full
and 64
4.5
Angle
66
66
Comparison Boost
Between
Entire
67
Nozzle
Deflection for
67
70
71
Comparison Boost
Between
Partial
....................................................... Comparison Partial Path Path Boost for for Between Full and Predictive
71
Nozzle
Simulations 5.1 5.2 5.3 Steepest Steepest Polak-Ribiere Minimization 5.4 Bracketing by Triplet
......................................... Function Function Algorithm Contours Contours for Function .............. ............
72 79 80
Conjugate
84
of Abscissas
..............................................
86
Figure
Page
5.5
Bracketing by Triplet
Interval
for
Function
Minimum
Bracketed 86 Interval Interval Extrapolation 92 98 ..... 100 103 of 104 Solutions of 104 ....................... Solutions of 107 107 ...... ....... 88 88
- Minimum - Minimum of Function of Slopes Decision a Headwind for Alpha Old Alpha Plots
Location
by for
Plot
Profile:
6.5
Old
100% 6.6 6.7 On-Orbit New 60% 6.8 New 100% 6.9 A. 1 A. 2 In-Flight
Van69 Mass
Profile:
Alpha
....................................................... Update and #70 for Wind #69 and Stronger Profiles #70 Wind Winds In Flight .....
Trajectory #69
Vandenberg Linearized
......................... Profiles
Vandenberg
List
Table
of Tables
Page
Engine
Characteristics (Datum
9 11
Properties Error
State
Comparison Entire
and
After Error
69
Between Boost
After Profile
72
102
102
New Alpha
Trajectory Profile
105
Optimization
Headwinds Profile
106
Alpha
Van69
Headwinds
..................................................
106
vii
ABSTRACT
A numerical process (A.L.S.). vehicle for optimization technique configuration is used of the to fully automate the trajectory Launch design system of the
an asymmetric
proposed process
Advanced
The objective
trajectory orbit.
design
is the maximization
The trajectories
shape model
that could
be described reduce
by a small set of parameters. the computation A predictive vehicle state, time required simulation
can significantly
wind control
This simulation
an idealized
by increasing
The conjugate The method simulation, parameters. Prelaunch the trajectory initial modified guess. The trajectory head winds. predictive deviations guess
of on-orbit using
requires and
the predictive
the gradient
of the on-orbit
function
with respect
to the trajectory
with finite differencing. carried out using the optimization to be highly procedure. sensitive shape choice For to the was of the
originally
proved
of the optimal
resulted
in a procedure
simulation produced
is used
in flight
to redesign conditions
to account than
for
expected of
trajectory portion
parameter
is modified
-- the value
aerodynamic
of the trajectory.
-1-
ACKNOWLEDGEMENTS
I received like to thank assistance. I would study Diane, Kamala, buddy. Camille, Kelly, much help and support Richard Goss, Frederick me much while Boelitz, working on my masters thesis. I would first and
and Gilbert Stubbs for all of their advice in my two years at Draper. for being Mike (since such a great friend
and
Chavela, Duncan,
Carde,
Kris,
and Cathy. my family for all of their encouragement to be when and support. play left
I would always
be anything
I wanted
for the 'Skins. was prepared the National Contract Johnson at the Charles Space Stark Draper Laboratory, Inc. Langley and under Task
Research
Aeronautics
of this report does not constitute agency of the findings of ideas. of this thesis or conclusions
approval contained
Laboratory
It is published
my copyright
to the Charles
Stark
Draper
Laboratory,
Cambridge,
Massachusetts.
Jeanne
M. Sullivan
Chapter One
INTRODUCTION
1.1 Background
Both planning manned to design system
and Problem
and unmanned trajectories launch vehicles currently require a large amount of
flown.
This leads to high costs and causes or orbital changes. wind averages from those In addition, rather than expected for orbit
to be inflexible vehicles
payload
for these
are designed
If winds orbit
the vehicle
with insufficient
maneuvers. A new launch vehicle the Advanced associated wind Launch is being developed System (A.L.S.), by NASA to alleviate vehicle these problems. decrease Called the costs
the proposed
should
preparation vehicle
variations.
measurements
launch.
trajectory once
of the vehicle
This mass
constraint design
on trajectory limits.
loads on the
Several asymmetrical
models
of the A.L.S.
have
on an of a seven
version
by General attached
Dynamics
to a single booster.
The resulting
asymmetry
in thrust
leads to trajectories
large pitch
-3-
The guidance
concepts
he
restricted
vehicle
to the pitch plane by nulling both yaw and roll torques. so that comparisons in ascent guidance between
This thesis
will be based on
a different
stage-to-orbit by a
(SSTO)
a simple trajectory
1.2 Method
costly
prelaunch should
preparation be fully
time
system
flexibility, vehicle
the mode, be
design
process
automated. before
a specific
acquired
shortly
that determines
mass, That
parameters
memory
to command
the trajectory
to adapt to winds that differ significantly algorithm utilizes has been developed simulation
a predictive
F.W.,
MGuidance,
Steering,
Load Relief
and Conffol
of an Asymmetric
Launch Vehicle H.
Massachusetts M.A.,
Institute of Technology
"Ascent Guidance
for a Winged
Technology
Master of Science
-4-
anda numerical
function. The predictive simulation system
optimization
scheme
mass to define
ifs objective
simulation
is a simplified mass.
simulation
on-orbit
simulation.
greatly
mass
because
equations
a much
integration stage,
time
The inputs
to the
include
the current
the prelaunch
wind measurement,
shape parameters. scheme uses the negative is equivalent to evaluate of the on-orbit to maximizing its objective mass on-orbit function as the mass.
optimization
This is required
the optimization
at a given
function
function, parameters
derivatives approximated
of the on-orbit
with
respect
to the
1.3 Overview
model
is discussed
in Chapter characteristics
2.
The
configuration, The is
pressure,
density,
and dynamic
equations Earth.
the vehicle's
load constraint
Chapter Boelitz
3 describes
the trajectory
design,
guidance, used
and control
concepts
developed
by
and implemented
-5-
chapter,the
proposed
trajectory
optimization
procedure
is presented
are described. simulation along is discussed in Chapter 4. The simplified approximation. different time kinematics and
control
for several
was very good. 5 justifies of on-orbit the choice of the conjugate theory gradient method for the numerical
mass. are
and procedure 6.
simulation worked
defined
to improve
the procedure's
robustness
to arbitrary
conclusions
-6-
Chapter
Two
VEHICLE
DESCRIPTION
AND MODELING
2.1 Physical
Description
of A.L.S.
Vehicle
model Launch
guidance proposed
of The boost
Dynamics
to provide
an unmanned
components booster
of a core
and
a single
in an asymmetric
configuration. Both the core and booster mixture thrust of liquid hydrogen stages (LH) have identical, and liquid lbs. oxygen non-throtfleable (LOX). Each engines engine fueled by a 6:1
level of approximately
612,000
has seven
core has only three engines. are gimballed the booster When
in both the pitch and yaw directions has more engines, its fuel is separated tanks
to provide
control. of the
will be depleted
this occurs,
the booster
from the core. the payload The lower The inertial bay. section The diameter of this section is
portion
the same
as the booster
measurement
at the base of the core. The Module from core. booster's (BRM). engines, This occurs servos, and fuel lines are contained in a Booster Separation Recovery of the BRM from the
separates
and recovery
is made
at sea.
-7-
293 ft.
331,000
Ibs.
Liquid Oxygen
Tank
Inter-Tank Adapter
l_\,,...._j./_\,..,__._/'_-'_
IMU
7 LH/LOX
Engines
***_**"*_
3 LH/LO
Engines
-8-
The A.L.S. vehicle uses a total of 10 gas generator engines. Several important features
of the engines are presented in Table 2.1. The engines are non-throttleable, meaning that
the thrust level cannot be changed during flight. both assumed to be constant.
The mass flow rate and vacuum thrust are of each engine within the atmosphere
SPECIFICATION
I I I
Gas Generator LOX/LH Fixed 1,427 Lbs/sec 612 KLbs 6,744 Lbs 88.0 in 150 in
Each engine is
angle of the engine nozzles of each engine is + 9". with a 5" cant angle in capability
can be changed during flight. The limit on the gimballing Because of the asymmetry the pitch plane.
capability
This design feature allows the vehicle to have a large gimballing angle.
-9-
For this study, it was assumed be represented be controlling thrust vectors by a resultant the vehicle. are limited
that the thrust for the set of engines such that two thrust vectors, is described by Figure
2.2.
to a deflection
cant angle
Booster,
Core,
nes
o/i i
_ Thrust ..-'''--._:." .o I_ i _I deflection cant Installed
.:"
..
"_----_ i 9
/__
Gimballing
capability
J
NOTES 1) All 10 engines are installed
/
with a 5 cant. gimballing capability of
2) All 10 engines have the same :t: 9 from installed cant. 3) Resultant 4) Resultant
thrust vector of core, Tc, acts through point A. thrust vector of booster, Tb, acts through point B.
Figure
2.2:
Thrust
Model
angle,
8.
control
while steering
to the trajectory
by the guidance
- lO-
2.2 Mass
Properties
the linear
acceleration properties:
during
flight, and
knowledge moment of
of the time-varying
(m), cg position,
as fuel is expended.
change.
is constrained of cg location
axis (lyy) is required. of the dry mass fuel mass properties mass properties properties is computed before launch and is combined calculation flight since before mass
can then
continuously
to be constant. used is the same as that developed by Boelitz, geometric based solids upon and
recommendations shells.
from
NASA.
The vehicle
is separated
All components
were assumed
in both the core and the booster as a solid cylinder modeled as solid
modelled length.
shells
modules
bay
are presented
2.2 below:
Xcg
?.cg
lyy
fit)
138.0 63.5 112.2
(ft) 0 0 -11.1
(slu_
ft 2)
Table
2.2:
Dry Mass
Properties
(Datum
at base of core)
-11-
2.3 Aerodynamic
Characteristics
to compute to know
the aerodynamic
forces
plane,
it is These
force.
of both Mach
NASA
with updated
aerodynamic
for 0<Mach<8
between
Linear interpolation
and cubic
interpolation
2.4 Environmental
Conditions
coefficients density,
above,
knowledge
is required
of to
of sound)
pressure,
Atmosphere.
These feet.
parameters
of altitude
range of 0 to 282,000
Above
be zero and the speed of sound It is assumed during flight. radar-tracked in this thesis
is assumed
winds
both before launch and during flight. provided a wind profile from design Vandenberg before AFB. A certain
launch,
and a different
simulation.
- 12-
the
guidance
system
for robustness
in the
presence
of unexpected
wind
this study
is limited
plane,
the winds
were
assumed
to the local Earth-relative line approximations and its linearized above 66,000
horizontal. between
straight
number shown
of data
The profile
by this study
approximation ft.
in Appendix
A. In this prof'de,
the winds
dissipated
2.5 Coordinate
Frames
and
Kinematics
Three
reference
frames
are used
to simulate
the motion
of the vehicle
about a spherical
Earth.
(1) Inertial
Earth-Centered
Reference
Frame:
(X, Y, Z)
of motion
are referred
frame. Pole.
The The
of the earth.
zero longitude
the right-handed
(2) Local
Geographic
Frame:
(ON,
liE,
liG)
is at the center
of gravity
of the vehicle.
ut; axis
points
the center
are calculated
this reference
-13-
Frame:
(xB, YB, ZB) is fixed to the vehicle's center of gravity As stated plane. and assumes earlier, this
of this frame
the vehicle
to the centerline
and points
product
of u6 and xB. The zn axis completes on the vehicle are computed in this
and torques
XB roll
UN
Headin
uE
Earth
z
Relative
Horizontal
yaw
B pitch
UG
Figure
2.3:
Relationship
Between
Body
and Local
Geographic
Frames
-14-
t.OEarth
Z
North Pole
Longitude = 0 at time = O)
Figure
2.4:
Inertial
Relationship
With Pitch
Plane
-15-
the body
reference
frames plane
is shown
in the
dynamics, angle
is set to
pitch attitude
between
the inertial
The angle of attack with respect _, are also depicted in this figure.
to the Earth-relative
of the inertial
frame into the body frame was described azimuth, elevation, matrix
a sequential in Etkin.
is described
The rotation
obtained
cos O cos
9" 9"
cos O sin 9" sin sin @ sin 9" + cos Ocos 9" 9"
[C]=
sin sin @ cos - cos sin 9" cos sin O cos + sin sin 9"
matrix
of a vector B is known
known
and vice-versa.
in the inertial
such that:
B =XlUXI
+yluyI
+ZlUTl'
=XB
UXB
+yBuyB
+ZBUZB
(2.2)
but
the
components
known,
then
these
components
can
be
determined
{x,}{x,}
YB
zB
=[C]
Yl
Zl
(2.3)
-16-
The rotation
known
matrix,
so that [C] q
= [C] T.
Therefore,
a vector
which
is
in the body-fixed
can be expressed
in the inertial
=[c] r yB
Zl ZB
(2.4)
2.6 Dynamics
and
Rigid
Body
Equations
Figure where:
2.5 shows
the angular
notation
frame
XB
\
L, fo r
YB M, fop
N, COy z B
Figure
2.5:
Angular
Rates
in Body Frame
- 17-
Since experiences
to the pitch-plane
motion,
it was assumed
L -N - 0
The rolland yaw rates are therefore also zero:
(2.5)
o_r = toy = 0
(2.6)
Parameters
The following
parameters, plane:
pictured
in Figure
The
air-relative The
velocity angle
between
velocity
and the
wind velocity.
with respect
to the air-relative
in the positive
YB direction.
of attack
with respect
to the air-
relative
of the aerodynamic
coefficients.
-18-
+X B
V A
cg
Earth
Relative
Horizontal
Figure
2.6:
Flight
Orientation
Parameters
2.6.2 Forces
and Torques
The forces
during
endoatmospheric
flight
forces A free
of gravity
into the
vehicle
forces
in the body-fixed
frame
FN =- S Q CN uzs
(2.7)
FA =- S Q CA UXB
(2.8)
-19-
where:
S = reference Q = dynamic p = air density "CA = magnitude CN --"coefficient CA = coefficient uzB = unit vector uxB = unit vector of air-relative of aerodynamic of aerodynamic in z-direction in x-direction velocity normal force C/v (a, Mach) Mach) area = constant pressure -- p VA2/2
+XB/.
F.
datum
Figure
2.7:
Vehicle
- 20 -
Theaerodynamic
pitching
moment
is expressed
in a similar form:
(2.9)
of center
of pressure
of center of pressure
(2.10)
(2.11)
where:
Tb = booster
thrust
to the pitching
moment
exerted
on the vehicle
is:
MrtRUST
= Tt, Xcg sin _ + Tc Xcg sin t_ - TI, (D + zcg) cos 6 - Tc zcg cos
t5
(2.12)
where:
x-axis z-axis
cg position cg position
(measured (measured
from datum
from centerline
between
centerlines
= constant
-21
The expressed
force
of gravity
points as:
toward
the
center
of the
Earth
along
UG and
can be
in the inertial
frame
Fg = - m g (R/R)
(2.13)
where:
R = position R = magnitude
expressed vector
in inertial
coordinates
position
m = re(t) = vehicle
mass at time t
of Motion
acting
can be summed
in the body-fixed
frame
to give
the
force in body-f'Lxed
FNET = FN + F A + T b + Tc + Fg
(2.14)
This
force
can
be resolved
into
the
inertial
frame
by the
use
of equation
2.4.
The
Iranslational
equations dR= dt V
of motion
are then:
(2.15)
dV=(1)F dt
(2.16)
The principle:
rotational
equations
of motion
are
calculated
using
the
angular
momentum
M -
d([/]o dt
acrll _t
)
relative to
body frame
x [/]0,)
(2.17)
- 22 -
This setof
equations
can be expanded
+[i
_9c0
(2.18)
where
as:
and is expressed
in the body-fixed
flame
M --
(2.19)
velocity
in the body-fixed
frame
as:
(2.20)
inertia
at time t
equation
2.18,
First,
to form Second,
a principal
the inertia
rate
was
contribution equation
moment
neglected.
Using
assumptions,
in scalar
d_ L = I_-_-
_paXj
(lyy
lzz)
(2.21)
M = 6y aco.
(2.22)
N = Iz_ _-
_a b {l_x- lyy)
(2.23)
- 23 -
Theseequationsare known as
that: the rolling moment, yawing
Euler's moment,
equations
of motion.
However,
we stated
earlier
L =N
=o_
=o_
=0
(2.24)
Therefore,
equations
and equation
2.22 simplifies
to:
dt
I;_M
(2.25)
where
moment,
M, acting
from
the thrust
= MAERO
+ MTHRUST
(2.26)
angle rates
rate transformation 1
[W] =
0 0
Using describes
matrix,
set which
of the vehicle
with respect
o =[w]
L_yJ (2.28)
where:
(assumption
made
analysis)
= dO/dt
- 24 -
In summary, the equationsof motion implemented in the six degreeof freedom simulationusedfor this thesisare: Translational:
dR= dt V (2.29)
av=(_)F dt
Rotational:
_Er
(2.30)
(2.31)
o =[w]
L_yJ (2.32)
2.7 Constraints
moment
on the vehicle.as
moves
bending
that
to constrain equation
aerodynamic
force
can
be
FN
= S Q c_v
(2.33)
- 25 -
where:
S = reference Q = dynamic p = air density Va = magnitude CN = coefficient of air-relative of normal velocity force = C_/(og Mach) area = constant pressure - p VA2/2
aerodynamic
of angle
of attack,
the coefficient
of normal Given
aerodynamic
function as:
of angle
of attack.
this approximation,
CN = CNa Ot
(2.34)
where:
(2.35)
The normal
aerodynamic
by:
FN
= S Q Qua ot
(2.36)
Since
and
CNais
approximately
equal
to
the
normal
is roughly Therefore,
proportional to control
of dynamic aerodynamic
Q, and the is
of attack, controlled
the vehicle
of Q and a. limit
by the
This flight.
is the
on the vehicle's
endoatmospheric
- 26 -
air density,
of
air-relative
velocity
with to a
is that
within
the atmosphere
has left is
A typical
pressure
behavior
illustrated There
methods
which
have
to control
the Qu product
so that II
on normal
aerodynamic capability
Boelitz,
study,
within could
the control
system.
Corvin
the specified
800
500
400 300
Time
(sec)
Figure
2.8:
Typical
Dynamic
Pressure
Profile
- 27 -
Chapter Three
TRAJECTORY
AND CONTROL
3.1 Introduction
The
new
trajectory
design by Boelitz.
and
guidance Trajectory
concepts design
developed
in this
developed objectives
and constraints,
Boelitz's
objective
of the guidance
commands
while mass"
it has
elliptical
on-orbit Also,
is that
it allows
to be available
minimizing payloads
for endoatmospheric
boost
will allow
the A.L.S.
to carry larger
A disadvantage loop manner dispersions the trajectory an attempt predictive for on-orbit those
of the Boelitz
study was that the guidance was not updated occur during
such that the trajectory from the pre-launch designed is made simulation mass. before
occurs,
to update
to maximize as
and estimation
done
by Boelitz
and introduces
in trajectory
and
Section phases.
is divided
- 28 -
of
different
guidance
and control
schemes.
Section
3.3 describes
which
signals
can be that
signals
the estimators
for angular
dynamic
concepts trajectory
the pre-launch
phase. design
3.6 introduces
the method
for automating
the pre-launch
and updating
the trajectory
in flight.
3.2 Mission
and
Flight
Phases
must The
payloads
through
the
atmosphere
and planning
The vehicle
must also have the capability and still have orbit. of the A.L.S. and can be divided thus different 3.1. enough
dispersions ascent
into the desired The ascent involves schematic Phase tower. Attitude Phase
phases.
Each schemes.
phase A
different
constraints is shown
and control
in Figure
rise from
the launch
can clear
attitude
of the vehicle
during
be held
constant
is thus used to meet this objective. by a rapid pitchover designed to orient the vehicle to the initial from the
2 is characterized by Phase
achieved while
1. Therefore,
remains is
below
in this phase
on pitch attitude
rate calculated
as an analytical
- 29 -
Phase
Four
Exoatmospheric Flight Phase Predictive-adaptive Powered Explicit Guidance (PEG) T = 120 sec.
Maximum Q a
Phase
Three
Constrained endoatmospheric flight phase Acceleration-direction steering, guidance, and control, subject to a Q (x limit.
Relatively unconstrained rapid launch maneuver Attitude control unimpaired by Q a constraint. T _=8 sec.
m
Phase
One
Figure
3.1"
A.L.S.
Flight
Phases
- 30-
Phase 3 is the endoatmospheric by the Qtt limit. above the specified profile The Qa product
not constrained.
command
is used that will follow specified limit. Phase that 4 begins when pressure flight
is below is above
if the Qa product
to a sufficiently from
small value
so to
the dynamic
transition
occurs
A modification
Shuttle Phase 4
Guidance
(PEG)
is inserted
The specified
miles miles
(3) Horizontal
_rpeJ
where
I.t = gravitational
constant
3.3 In-Flight
presents
and involves
control
concepts
developed
by
in the previous
to improve
on the control
concepts.
-31
3.3.1Phase
I and 2 Guidance
and Control
Figure 3.2 shows a generalblock diagram describing the guidance and controlfor both Phase I (vertical rise) and Phase 2 (launch maneuver). The guidance commands of both
Phase I and Phase 2 are based on pitchattitude rate, COo For Phase 1 the vehicle must risevertically to clearthetower and so:
_(0= 0
8cmrr= 90" (3.1)
2, the vehicle
must
conditions
for Phase
1 and
then orient
conditions
3. Corvin
and Boelitz
form is given
by the following
(3.2)
of Phase of Phase
TKick = duration
The
shape
is shown design
in Figure process
3.3.
Both
discussed in Figure
in the next
of tOc shown
analytically
the following
for commanded
pitch attitude:
<33,
- 32 -
Figure
3.2:
Phase
1 and 2 Guidance
and Control
Block
Diagram
33
2t_{deg 0 ,_,
Ih,
Referring open-loop,
employed
loop during Phase 1 and 2 because it was decided that the altitude of the vehicle during both of these phases was sufficiently The attitude commands low so that wind dispersions by the open-loop were not a serious problem. system are fed directly attitude and the control. The
generated
guidance
into the control system which tries to null the error between the commanded sensed attitude, e0. The attitude error is compensated resulting signal is then summed with the fed-forward by proportional-integral commanded
in a pitch rate error signal, e_. This error signal is multiplied nozzle angle command by a proportional nozzle gain, KNKv, Boelitz to provide linearized an engine the vehicle
to the engine
servos.
dynamics,
of a planar trajectory,
_, and pitch attitude, O. On the basis of this linearization, the gain transfer function is approximately Kv where:
Kv = T Xc...._._g
lyy
(3.4)
- 34 -
which
monotonically
increases
with time.
Therefore,
was perfectly
achieved:
8=
(3.5)
and Control
launch
vehicles
direction"
steering
direction"
acceleration
of the vehicle
by gravity. or a pitch
loop
is commanded
In addition signals
acceleration a parallel
in the a
is added. component
employs vehicle's
proportional axis.
to the
measured
normal
is approximately
proportional
is to rotate
the vehicle
and thereby
load relief
used
which
overcomes mode,
of the
disadvantages
It is a dual
acceleration by Bushnell.
developed
A general control
3 acceleration system
guidance
and mode,
is shown
in Figure
The control
"sensed"
to follow
a stored
The sensed
to vehicle
acceleration mode
and aerodynamic
The secondary
- 35-
mode)
loads.
are constantly
an angle pressure,
of attack _):
limit defined
by the division
(3.6)
(3.7)
where:
_t = estimated
eA = accelerafion-directon
angle
of attack
of attack limit,
then
and thus null out any difference acceleration-direction and the vehicle trajectory autopilot
between
following
by an add-on
load relief
function
when needed. or open-loop guidance is shown on the block profile diagram can be of in
closed-loop design
guidance, in flight
a new
acceleration
This closed-loop
will be discussed
- 36-
_E
t--
"i
_T
,_ 8,0-
I_loI
8 J g _ 8^
(T+
_T
I
_J_-
_j
Figure 3.4:
t_oj
Phase 3 Guidance and Control Block Diagram
- 37 -
system
shown
switch either
in Figure
in
1 and 2. of attack
be modified tac"
by proportional-integral is compared
producing pitch
rate,
This signal
by a time-varying
proportional nozzle
nozzle
command,
are idealized
deflection
is perfectly
dynamic sets
characteristics, are
of gains
designated 3.,*.
by the Boelitz,
subscripts approximations
in the gains
in Figure
involved error
command
be continuous
3.4 Sensing
and
Estimation
3.4.1
Sensed
Signals
that
Measurement
Unit
(IMU)
contains
an
attitude
measured
attitude
an integrating of sensed
accelerometer acceleration-
This velocity
produced
by thrust
gravity.
are processed
and control:
1) Vehicle 2) Inertial
relative
to a local earth AV
horizontal:
increments:
- 38 -
3) Inertialvelocity: V
- this signal where 4) Inertial gravity R of inertial velocity is the sum of the sensed is specified by a gravity inertial model velocity and the integral of gravity,
position:
- the integral
In addition nozzle
to these
four processed
signals,
the nozzle
deflection,
8, is determined
from
actuator measurements. are several signals directly needed during for the guidance These velocity excluding and the control of the vehicle that
flight.
rate (w), the angle of pressure (Q), signals and the must be
acceleration estimated
flight.
3.4.2 Angular
Rate
rate estimate,
_, is used
as the inner
variable
during
1, 2, and 3. It is also used for the estimation of tangential and and centripetal intrinsic noise acceleration in the IMU
The
will be magnified
in the derived
rate, especially
if the sampling
is small. on a first order digital over a sampling representation through In the complementary interval) is shown f'tlter
used an angular
rate estimator
based
in pitch attitude
a low-pass
estimate
implementation an estimate
frequency acceleration
is used
in which
of angular as
is passed
This procedure
gives
- 39-
filtering The
a high angular
frequency acceleration
rate
signal
based
upon
integration
estimate
is based
upon
the engine
deflections, forces.
and corrections
and aerodynamic
low-pass
filter
o_lowfr_u=%-y
'cs + 1
+
_-_h_h _quency
xs + 1 high-pass filter
Figure
3.5: Continuous
Signal Representation
of Angular
Rate Estimator
angle
of attack
estimate,
in Phase feedback
3.
It is used
in the
mode of
variable
in the control
system
order
digital
filter
continuous-time input,
representation
in Figure determined
procedure:
at the angular
center
of gravity
from
the IMU
sensed
aerodynamic
the estimated
- 40 -
4) Determinethenormal
aerodynamic pressure. 5) Use the normal the aerodynamic force
by dividing
the estimated
normal dynamic
force coefficient
and Math
number
to search
through
to find a corresponding
low-pass
2_0)nS
filter
2 + COn
fraqu_ey
2 S2 + 2_0)nS + COn
-I-
-t-
O_high frequency
S2 + 2_0)nS
+ 0) 2
high-pass
filter
Figure
3.6:
Continuous
Signal
Representation
of Angle
of Attack
Estimator
high back
frequency to Figure
is
input
to the complementary
filter
is based
upon
pitch
attitude. input,
frequency,
equivalent
into a lower
filter.
filter is shown
The Chapter
problem
lies in estimating
a discrete-time as:
shown
in
a = 0- y+ C_w
(3.8)
-41
low-pass
filter 2
2_cons + co n
O_w Ir_u_mey 2 S2 + 2_C0nS + O)n
ahlgh lr=qu_=y
S2 + 2_(OnS + (02
high-pass
filter
Figure
3.7:
Alternate
Representation
of Angle
of Attack
Estimator
Taking
the derivative
_=o-_'+_w
(3.9)
This equation
can be expressed
in the discrete-time
domain
as:
Aa = A0 - AT+
Aew
(3.10)
The
incremental Because
change
by variations be measured,
normal
vector.
is neglected
of attack estimation.
A7 is also neglected
the relationship:
Ao_= za0
(3.1])
The varying
omission
change
in flight order
path
angle
can cause
a small
time-
A second
complementary
angle of attack
- 42 -
3.4.4 Dynamic
Pressure
Phase 3 utilizes in the Qa limiting estimator. The dynamic the air-relative during no wind
estimated mode.
dynamic
pressure, pressure
swRching
logic
and
The dynamic
of attack
density density
of
is assumed because
atmospheric
However,
measurement of horizontal
angle
to accomplish
3.4.5 Acceleration
Direction
acceleration of Phase
direction,
3. The increments
velocity cycle,
the inertial
velocity
increments
are expressed
of the acceleration
vector
is then computed
in terms
of the pitch
and yaw
(3.12) (3.13)
- 43 -
The AV
and yaw
signals
because
Therefore,
the pitch
are each
sent through
continuous-time
fl(s)
zlfl + 1
(3.14)
where
U ^, representing
the estimated
filtered
acceleration
axes is calculated
UA = Unit value
of
The estimated
acceleration
direction
(3.16)
control the
system,
signal
between is
acceleration
estimated
direction
between
unit vector)
C = UA x U_
(3.17)
The angle,
I_A, between
is:
flA = sin11C
(3.18)
The vector
composed
in roll, pitch,
UE=
fla[ unit(C)]
(3.19)
-44-
The pitch
component
of this vector
is the error
signal,
CA, shown
in Figure
3.4.
3.5 Pre-Launch
3.5.1 Introduction
Trajectory
Design
system
which
is implemented prior
in an open-loop
manner
to launch
by a trajectory
design
for phases
1 through
3 was open-loop.
closed-loop current
Explicit
program
program
of aerodynamic
has reached
the upper
atmosphere. Suitable constraints. orbit m/_,ua. needed. heavier trajectories must meet the desired the primary The objectives objective vehicle while satisfying the specified is to reach mass, be for is
vehicle,
of trajectory have
design
should
on-orbit
for any
post-boost
Minimizing payloads
the fuel
to be carried experienced
must be within
the designer
limit at
design
process
is greatly shapes
simplified
if a trajectory
shape
or form by:
is
The trajectory
are described
rise.
attitude
of an angle of attack
shapes
and the
trajectory entire
pre-launch
automated
- 45 -
numericaloptimization technique. The automationof the pre-launchtrajectory design processis discussedin Section3.6. Section3.7 discusses how this sametechniqueis appliedto in-flight trajectorydesign. 3.5.2Phase
1: Vertical Rise
is required of
of the vertical
possibility
it was decided
at 400 feet.
Maneuver
to the overall
trajectory
this quickly
the vehicle
well below
Q_x limit. As discussed a sinusoidal earlier in this chapter, the trajectory shape used for this phase is based on
function
of pitch angular
rate:
TKick
(3.20)
where:
commanded 1 2
pitch rate
of Phase of Phase
Integration
of this equadon
yields:
oi
ZKick
(3.21)
- 46 -
where:
be noted be 90".
phase 1 involves
a vertical
rise from
the shape
of the pitch
reached
Boelitz
of this phase
He chose
Of and
The procedure
order/idealized Using
control
simulation
to predict
state at the end of the launch motion, the desired tolerance. he iterated
maneuver.
simulation
of the
on the value
of TKick until
than a small
in one dimension
Profile
In Phase
design
process
As the trajectory the acceleration loop during The angle phase. attitude value, achieved dividing
program
simulates
of the vehicle
is calculated
and stored
flight. of this angle guidance of attack profile system is shown in Figure 3.8. Figure 3.9 shows the
shape
of attack
design system
simulation
of this
in angle
in phases value
be equal limit
of attack Q. When
- 47 -
vehicleis commanded
the angle of attack "bucket" shape
to fly along
Because
the specified
to the dynamic
pressure.
constant
of attack, al,
three limit.
which
define
is constrained
1) The
Qa
limit
used
must
be less loads
specified
by a finite structural
2 must need
be able to meet
transition
0_
Qoc limiting
OC1 I I I I I I I ! I I ! ! I I
Qtz limiting
ends
a 2
! I !
Figure
3.8:
Phase
3 Angle
of Attack
Profile
for Trajectory
Design
-48
.=_ca
Figure
3.9:
Phase
3 Control
System
For Trajectory
Desgin
- 49 -
3.5.5Phase
Phase
4: Powered
Explicit
Guidance
4 is initiated
when For
the vehicle
is in the upper
atmosphere
a version
Explicit
trajectory important
design
prior
of the overall
is to produce
updates
acceleration
direction
angle
in pitch
every
six
using a "linear
tan
OA = Ko + (t -to) K1
(3.22)
where
angle
in the pitch
plane
to minimize mass.
the propellant
required
into orbit,
thus maximizing
This program and the model only variable and acceleration of the vehicle.
as a point mass
dynamics degrees
six degrees
of freedom
of freedom.
inputs
inertial
position,
velocity,
acceleration mass
of gravity
the on-orbit
3.6 Automation
of Trajectory
Design
phases
with each
1 through
3, can be completely
described
parameters:
- 50 -
1) Of
2 for Phase 3
4) Qa limit = specified
for mission
specified
Qtz limit
This trajectory.
is a very
simple
only four
parameters four
are needed
to describe
the
these
parameters
each
of these
parameters
using reached
the full six degree an maximum. has been reached. determine is based function
of This
until
it seemed
that on-orbit
the a
maximize
on-orbit
upon it seeks
optimization
scheme
which
uses on-orbit
mass
A predictive
simulation trajectory
mass. 3.10.
A simplified
is shown
is initialized
state
The only
information
available
optimization
algorithm
supplies
simulation PEG
the current
is used
optimization which
algorithm optimize
multivariable maximum
on-orbit
-51
&
ffl
Figure 3.10:
Automated
Trajectory
Design
Process
- 52-
This unexpected
scheme
is used
before
for it
changes
in payload
this scheme
the trajectory
of wind
dispersions
the complete
sensors
so the disturbance
information dispersions,
of an updated
describe
the components
in detail. simulation
A reduced
the integration
A conjugate
gradient
because
is highly
nonlinear.
gradient
is approximated
finite
differencing.
- 53-
Chapter Four
PREDICTIVE
SIMULATION
4.1 Introduction
for selecting
trajectory
parameters
to maximize this process, is the on-orbit the vehicle's end the the of a
In order
optimization
scheme of on-orbit
is employed
conditions which
A simplified of computation
simulation
written mass.
to calculate
the on-orbit
This chapter
characteristics freedom
with the full six degree in Chapter used for 3). the predictive an idealized 4.3. Section Remarks
simulation
reduced-order
to increase
the integration
time step of the simulation, is described for various in Section wind conditions.
system
4.2 Reduced-Order
Model
to only three
degrees plane.
of freedom
the vehicle
pitch plane,
By choosing
is needed are of
to specify needed
to describe frame.
the location
of the vehicle
the inertial
simulation,
it is still necessary
to calculate
mass
- 54 -
aerodynamic simulation,
and thrust
are greatly
simplified frames
simulation
in Figure 4. I and
Reference
Frame: of motion
(x,y, z) used is fixed in the predictive to the surface of the Earth set. simulation of the fiat Earth are referred to this site. The
equations frame.
The
origin
The z axis points y axis completes (2) Body-Fixed The origin roll motion. the nose cone.
toward
the center
(xs, YB, za) is fixed to the vehicle's to the centerline is restricted center of gravity and assumes no
and points
towards
Because
to pitch
as the inertial
y axis.
of the body-fixed
frame
with respect
frame
to
x-z plane
about
y axis.
R, locates
with respect
origin.
The
frame matrix
is described from
by a single inertial
0, the pitch is
a vector
to body
coordinates
[C]prexlictiv
sim
cos 0 sin 0
0 0
-sin 0 cos 0
(4.1)
- 55 -
R Flat +z B +x
Launch
Site
+z
NOTES: Launch site = origin of inertial frame Vehicle cg = origin of body-fixed frame
Simulation
Coordinate
Frames
The aerodynamic
and thrust
forces
are defined
the same
FN =- S Q C,v UZB
(4.2)
F A = - S Q CA .XB
(4.3)
(4.4)
(4.5)
-56-
subtends
around as being
the Earth during boost is very small and so it flat during boost. Using inertial this approximation, z axis:
to approximate vector
will always
in the direction
of the positive
Fg -- m g uz
(4.6)
forces
are resolved
from matrix
frame 4.2.
into
the
in equation
These
in the inertial
frame
on the vehicle
in inertial
FNET = FN + FA 4- Tb 4- T 4- Fg
(4.7)
Since
it is assumed
that
there
are
acting
on
the
vehicle,
the
y-
component
(4.8)
The translational
equations
of motion
simulation:
dR=
dt
(4.9)
dV=(1 dt
) F
(4.10)
where
V = inertial
velocity
= VE
Because always
forces,
the y-component
of both position
and velocity
is
R = Rx ux + (0) uv + Rz uz
(4.11)
(4.12)
- 57 -
from:
dt
Gy -1 M
(4.14)
attitude,
needed
to specify
the attitude
of the vehicle
and is
d_a dt= mp
(4.15)
Boelitz
simulation When
trajectory simulation
design
and tuned
the
pitch attitude
was utilized
controller
and angle
3) were digitally
in the 6DOF
simulation
with a sampling
time of 0.1
integration
simulation
To achieve
this goal,
simulation The
both
response that
benefits systems
The
replacement
of the control
large
integration the
the computations
to determine
- 58 -
actuator
commands
for the
actual system.
control
systems since
by much system
Finally,
control
either pitch attitude or angle to either equation be specified of motion simulation used
control of attack.
allows Thus,
or calculated
twice to solve for pitch attitude. design purposes so it must mimic design. the
the trajectory
for phases
of attack control
systems
is made
control
will be perfectly
o--o:-- {I,Ph_ 3:
LTKick
{,.
+o:,,,,.,
(4.17)
(4.18)
the idealized
control
one
parameter must
has be
specified
attitude
of attack
at all times
attitude frame
is needed
to specify
of the
body frame
with respect
to the inertial
to determine
coefficients
orientation
shown attitude
in Figure
4.2,
- 59 -
+x B
VW V A
cg
Earth
Relative
Horizontal
Figure
4.2:
Flight
Orientation
Parameters
The following
equation
can be derived
from Figure
4.2:
0=0_
+ )' - OrW
(4.19)
where:
pitch attitude with respect contribution with respect to the air-relative from winds velocity = aas velocity (VA)
to the Earth-relative
- 60 -
Both the flight path angle, be determined Only time, one flight either from
contribution
aw, can
the assumed
of motion. at a be
orientation
parameter or angle
parameter
determined
4.20 or equation
Phase
1 & 2:0
= Oc therefore
a = 0 - 7 + aW
(4.20)
Phase3:
a=acthereforeO=a
+y-aW
(4.21)
control
systems to rotate
used
to determine
attitude
these control
systems
idealized, given
the question
how to deflect
nozzles
in equations
deflection using
This equation
can be rewritten
approximation
where
= t_ and cos(b') = 1:
for nozzle
deflection:
The pitch moment, in flight from knowledge analytically assumption: for the fast
I0, is the only term in the above of the current and second vehicle
equation
that cannot
phases
Phase
1:I0
= IOc = 0
(4.24)
-61
,.
Phase2:
lO=lOc=
.0
__211_ (Trick}Sin[
(t2rtLZKick Zvert) ]
(4.25)
vehicle's
pitch
moment
during
the third
phase
has
a very maneuver.
small
average
varying
deflection, 3:
to approximate
of the vehicle
as zero during
phase
Phase
3:I0
= 0
(4.26)
4.4 Predictive
Simulation
of the
predictive
is shown state.
4.3. can
The
predictive on the
wind
information prior
given to launch.
be the
in the 6DOF
simulation. and
include winds,
properties,
atmospheric transformations.
density
procedure
commands
attitude routine
during
phases
of
control or angle
then assumes
perfect
the two flight parameter. and is used acceleration integrated guidance continues on-orbit comparison
parameters,
of attack, (zero
and computes
the unknown
The
pitch moment
is specified deflection
for phases
to calculate
the nozzle
to achieve frame
control.
of motion environmental,
and control
Explicit
Guidance
routine the
At this time,
aerodynamic
forces
to the thrust
- 62 -
Phases
Guidance 1 & 2: 9c
no
Phase 3: (Xc
e = ec(t)
Phase 2: o_= 9- 3'+ aW
.
0 = f(_,
TKick, t)
dt
(x = _c(t) Phase 3: 0=
,,
dR=v
, dt
o_+'y-otw
0=0
.=
'
t
Compute Acceleration:
= _(e, I F AERO I, I
F_HRUSTI,
Simulation
Flow Chart
- 63 -
between
it was
decided
to use kinematic
an
trajectory
simulation. simulation's
position, required
trajectory between
would reference
for comparison
simulation
in Figure
inertial,
transformation vectors
between
is constant.
are expressed
coordinates
simulation
06 D
(3.
Figure
4.4:
Relationship
Between Predictive
Inertial (3DOF)
Reference
Frames
of Full (6DOF)
and
Simulations
- 64 -
simulation
is to give
prediction
of the of
computation
to a minimum. states
simulation,
the following
were compared
by the full
simulation: (1) angle (2) flight (3) height (4) nozzle of attack path angle (H) deflection angle were (6) run. 3DOF The first set corresponded to starting both the (o0 (_)
Two 6DOF
sets
of comparisons
simulation
simulation
set because
The second
by initializing
simulation a "partial"
trajectory. simulation
improvement
in accuracy
of the predictive
it is initialized
state at a later time in flight. a value of integration was greatly time step for the predictive enhanced by using a small simulation. time step, dt, of the vehicle For phase 3,
1 and 2. This is due to the fact that the dynamics are much faster than during the rest of the flight.
the launch
four different
dt = {0.1,
0.3, 0.5,
1.0} seconds
(4.27)
The
first
to "entire" plot,
boost
is shown
in Figures the
For each
computed
by the 3DOF
simulation
time steps
- 65 -
12 _ _
.......................... t ! i i i
LEGEND
"
I...........
,o I.............. \ ..........................
,,1=
<
_.--'_ ............
:
! ...............
:
':.........
; _
_. -: _-.-_.;_ ........
4 ............... 20
i ............... 20
! 80
_'"_'
:
i
"" 120
40 Time
100
Figure 4.5:
Angle
of Attack
Comparison
Simulations
For Entire
i 60 ....................................
50
_ 40
30 20 10 0
..............
..............................
_dt=
i................
'- ..............
: - ..............
2. , -
_"
-% ......
_...............
Increasing (:It i
"." - -.
20
40 Time
80
100
120
Figure 4.6:
Simulations
- 66 -
12 x104
10
................. :
:
Solid Une:
_: ....... 1
0 0 20 40 Time Figure 4.7: Height Comparison 60 (sec) Full and Predictive Boost Simulations 80 100 120
Between
For Entire
5 ................ Dashed Lines: Pred. Sim ............ Increasing dt ........ (dt = 0.1, 0.3, 0.5, 1.0 sec :. i i_l_ ! z _
:,
Z 3 2.5 0 20 40 Time Figure 4.8: Nozzle Deflection Comparison For Entire 60 (sec) Between Boost Full and Predictive Simulations 80 100 120
- 67 -
for
The phases
correspondence 1 and 2 was very control control because dynamic for angle
simulation
utilizes control.
of attack
utilizes
simulations,
occur
during
limit is computed
of the Qa
Q, is a function
is calculated
in air-relative of attack
of the angle
predictive
simulation
steer
angle of attack
limit, but this limit may not be the same as for the 6DOF velocity. 4.5 that the angle predicted of attack given
simulation
by the 6DOF
simulation
the angle
of attack
predicted
by the 3DOF
This is a result
sources sources
produces error.
time steps produces tend to be offsetting. in flight path angle generated simulation
a positive
and height.
The
height
for
simulation between
the height
curves
simulations
steadily
of integration
in the equations
- 68 -
deflection.
The nozzle
deflection
shown
cant of 5".
deflection
of the 6DOF
simulation
deflections
in slope.
for these
discontinuities than
is thus larger
approximation
used for phase 3 of the idealized The average error of the nozzle
deflection
grows with time yet is stiff very small. the absolute The absolute errors in the final error in predicted states of the predictive mass simulations along at with
on-orbit
is presented is calculated
This quantity
as follows:
(4.28)
for 6DOF
of core
3DOF dt
O__ll'or
_elTor
H_T0f
t_tmror
mferrof
(sec)
0.1 0.3 0.5 1.0
Table 4.1:
Between After
and Predictive
(3DOF)
Entire
- 69 -
The second
Figures predictive accuracy have where height. are 4.9
runs above,
boost
is shown
in the The
to 4.12.
by initializing
simulation
simulation
because
errors 4.11 of
less time
in accuracy cannot
even be observed
shown
errors that
boost runs.
runs Also,
significantly percentage
were
produced is improved
in "entire"
_" <
...............
3 60
70
80 Time
90 (sec)
1O0
110
120
Figure
4.9:
Angle
of Attack
SimuIations
For
- 70 -
55
i..............
.i ..............
i
i ............
i
I SolidUne:
FuUSire
Between Boost
Simulations
12
10
..............
.............
: ............... ! _'_
: .... __"
..............
xO
2 6O
70
80
90 Time (sec)
100
110
120
Figure
4.11"
Height
Simulations
-71 -
LEGEND
o5 ................
4.5 ............... 4
"'_ .......
!...............
! .......
1
,..,
3.5
6(
70
80 Time
90 (sec)
100
110
120
Figure 4.12:
Nozzle
Deflection
Between Boost
Simulations
tit
Table
4.2:
Comparison Simulations
Between
and Predictive
(3DOF)
After Partial
- 72 -
4.5 Conclusions
(3DOF)
simulation complete
developed simulation.
eliminates control
computation and
integration
simplifications simulation.
assumptions
predictive good
of the predictive
simulation
to be very
in comparison
simulation.
- 73 -
Chapter Five
NUMERICAL
OPTIMIZATION
5.1 Introduction
The primary objectiveof trajectory design for the A.L.S. is the maximization of onorbit mass. An optimization procedure isdescribedin this chapterwhich willmaximize onorbitmass. This mass is determined by the trajectory thatis flown and is,therefore, a functionof theparticular parameters chosen to specifythetrajectory shape. For a given set of trajectory parameters,theon-orbit mass isdctcrrnincd by using thepredictive simulation described in the previous chapter. The simulation approach is requiredbccausc itisnot practical to develop a closed-form solution for the on-orbitmass. Consequently, the onorbitmass functionisnot analytical. Any optimizationprocedure which seeks to maximize on-orbitmass must, therefore, bc numerical in form. For the approach used in thisthesis, the on-orbitmass function ismultidimensional because as many used to definethetrajectory shape. This chapter firstcompares in Section 5.2 several multi-dimensional numerical as threeparameters are
optimizationalgorithms thatarcdescribedin the currentliterature. The particular method chosen was a version of the conjugate gradientmethod. The overall procedure for
implementing the conjugate method and some of the underlying theory is described in Section 5.3. Sections5.4 and 5.5 describeseparatesubroutinesthathad to bc pcfforrncd in conjunction with the algorithm. Section 5.4 describeshow theon-orbitmass function was optimized along a specific searchdirection. Section5.5 describeshow the gradientof theon-orbit mass functionwas approximated using finite differencing.
- 74 -
5.2 Comparison
of Numerical
Optimization
Methods
numerical
optimization most
algorithms suitable
To determine
of this thesis,
Two primary
et. al.4. optimization schemes these rely on function information and sometimes them, it
information. to provide
methods
between
information
of a multi-dimensional which column minimize maximization the minimization a function vector the
function
must
variables were
function. mass.
design
is the to as:
of on-orbit
the maximization
of on-orbit function
is equivalent
of the negative
of on-orbit
will be defined
F(x) =- mf
(5.1)
where
mf
is the There
on-orbit are
mass three
of the
A.L.S. that
and define
from of the
the
predictive in the
simulation. predictive
parameters
trajectory
simulation
x = [Of o[ 1 a2] r
(5.2)
3 Scales, L.E., Introductign LTD., pp. 1-106. 4 Press, W.H., Flannery, Cambrdige:
to Non-Linear O_timization.
1985.
London:
MacMillan
Education
B.P., Teukolsky,
S.A., Vetterling,
W.T., Numerical
Recipes.
1986.
-75
the n first
partial
g = VF(x)
(5.3)
information mass,
on the shape
of the function.
function
based
this gradient
be approximated
composed
of the
derivatives
G(x) = V2F(x)
(5.4)
by the following
tensor
notation:
_xi_xj
(5.5)
This matrix
is also not analytical this matrix and would function matrix about with
if on-orbit finite
mass
function. number of
differencing
require
is used can
also
because
the shape
of the objective
schemes
can be divided
into three
main groups
they utilize. are commonly called "direct about search the methods". gradient These vector or
evaluations. Without
is utilized.
of the
function,
these
-76-
to the direct
is provided
by a second
group
of
first derivative
in addition to function
evaluations.
of the gradient vector, g. Included in this category are the conjugate gradient methods and variable metric
information,
are able to change the vector x function. The first derivative because these Direct
makes these methods more efficient than direct search methods any points that would increase
function value.
are useful for highly discontinuous for the trajectory For continuous
However, function.
problem in this thesis, the on-orbit mass is not a functions, with continuous f'u:st derivatives, the
gradient methods
than direct search methods. about the Hessian matrix of the objective function
The third group requires information in addition to the gradient vector. is utilized. This group includes
and variations
in computer
of approximating
inaccuracies impractical
of on-orbit mass. to
A member of the second group has been chosen for this study because it is possible approximate the gradient group, the method of the on-orbit mass function using finite differencing. descent was eliminated from consideration method. section.
Of this its
of steepest slower
because
is usually
gradient
is presented
in the following
method and the variable metric methods metric methods information the conjugate construct a rough
approximation number
collected gradient
over a successive
metric methods
its simplicity.
- 77 -
5.3
Conjugate
Gradient
Method
Gradient
methods,
to within
tolerance.
The iteration
on x is achieved
relationship:
xk+l = xk + 0kPk
(5.6)
where xk is the previous and xk+l is the updated a positive F(x), along scalar
will minimize
Pk is defined to give
weighting
the greatest
vector. described
by a separate
optimization
In the method
of steepest
descent,
the search
is defined
as:
Pk = - gk
(5.7)
where where
vector. always
The
steepest
descent
method
is useful
points
of the minimum.
5.1 where
the function
descent
the minimum
in one not
the minimum.
In general,
this does
Using found
descent F(x)
a new
point,
when
which
VF(xk+l),
vector,
leads in a slowly
-78
Figure
5.1 Steepest
Descent
Contours
gradient
method
makes
an improvement
on the definition
faster
convergence
of steepest to examine
To understand function
the properties
of this method,
the case
the objective
to be minimized
is a quadratic
function
of n independent
F(x)
=IxTAx
+ bTx +C
(5.8)
where
A is a constant
symmetric is:
n by n matrix,
b is a constant
n-vector
and c is a scalar.
The gradient
of this function
g(x) = Ax + b
(5.9)
The Hessian
matrix
function
is simply:
G(x) = A = a constant
matrix
(5.1o)
- 79 -
x0
Figure
5.2
Steepest
Descent
Function
Contours
If A, and thus G, is positive global assumed matrix. A unique point xl: property minimum
definite,
function
will be convex
and have
For the rest of the discussion, 5.8 does have a positive definite
it is A
defined
in equation
of quadratic
functions
can be developed
by taking
the gradient
at a
g(xl)
= Axl
+ b
(5.11)
g(x2) = Ax2 + b
(5.12)
- 80-
(5.13)
results in the following property which
Substituting is unique
equation
equation
to quadratic
g(xl)
(5.14)
it now possible to describe the conjugacy gradient property method or
Having
defined
gradient a quadratic
in n iterations
used in equation
pi T G Pk = 0 for j _ k
(5.15)
condition
can be changed
physically
meaningful
form by
by the scalar
(5.16)
Using
equation
to:
(xj+l - xj)ZGpk
= 0
(5.17)
Substituting
the transpose
of equation
5.13 results
in:
(gj+l - gj)r Pt = 0
(5.18)
The above
equation
means
search
direction, that
to
in gradients,
= 0(1)k-l,
-81
This propertyassures that the minimum of a quadraticfunction with a positive definite Hessianmatrix will befoundin at mostn
iterations. A proof of this is given in Scales.
Scales
also derives
the following
definition
of a search vector
which
satisfies
the conjugacy
condition:
Pk -- - gk + fit Pt-I
(5.19)
vector
is:
Po " - go
(5.21)
There
versions The
gradient chosen
method
differ
only
in the
definition method
particular
for this
"Polak-Ribiere"
flk is given
by the expression:
,Ok =(gk-
gk-l)Tgk
gk_lZgk_ 1
(5.20)
algorithm
was recommended
and Press
in finding be noted
the minimum
of a non-quadratic gradient
method's
of the search
vector
any more
evaluations vectors
gradient
method
search
vectors
a positive
optimization
procedures,
mass
functions could
positive
Hessian
matrix.
occur.
the gradient
algorithm
will generally
go through
- 82 -
more than n
function, The guess
iterations
for convergence.
If local
minima
are present
in the objective
then computes
chosen
gradient
information
the main
algorithm vector
the scalar,
VF(xk.j)
algorithm
direction,
is based
conjugate
If this
is smaller
a specified
estimate
of the minimum.
5.4 Minimization
Along
Search
Direction
function Recalling
out
xk+l = xk + r/k Pk
(5.22)
r/k must
be chosen Pk.
search
direction
minimization value
algorithm
computes
a value
and substitutes
it into equation
direction, evaluated
a new estimate
to the one-dimensional
minimization
algorithm.
- 83 -
XO
Compute
value of Xo
function at
_,f(xo)
of functiongradient at Xo Compute ,_ Vf(xo) Initialize direction gradient
,.,.*,,,,_,*,o,,4,,,.,.*..,,,,,t
_i-
_)):
....... ",*'
go = vf (Xo)
Po =-go
_,,_Po
x_
of f(x)from direction
_Lf (Xk)
Xk-1 Pk-1
''(
gradient
Xk
k =_k+l
s!...
gk = Vf (Xk) _k = (gkgk-1 )Tgk gk-1 Tgk-1 Pk = "gk + _kPk-1
Xk, Pk, gk
Figure
5.3 Polak-Ribiere
Algorithm
for Function
- 84 -
minimization
algorithm in Press.
of two
the minimum
of abscissas:
A < B < C
(5.23)
and:
F(B)
< F(A)
(5.24)
and:
F(B)
< F(C)
(5.25)
5.24 and 5.25 state that, of the three points value, ff the bracketing conditions within
which
bracket
a function,
B has then a
function
shown
above
illustrate
shown
the
are met,
< F(C)
so the minimum
cannot
satisfy
equation
the interval
uses is based
upon
parabolic If F(B)
curve
fitting.
Two
initial A and
points
> F(A),
then calculates
a guess
A and B compared
to the distance
between
C =B+(2-Gr)(B-A)
(5.26)
- 85-
Minimum
of function
not bracketed:
Figure
5.4
Bracketing
not Bracketed
Minimum
of function
bracketed:
F(C) F(B)
Figure
5.5 Bracketing
Minimum
Bracketed
- 86 -
where:
Gr=3-_
(5.27)
After C is calculated, The routine this point. Examples not bracket 5.6, of this procedure the function fit results solves
the routine
curve
fit between
the three
points. at
and examines
the function
evaluated
are shown
in Figures
points
minimum. in U being
minimum
the curve
minimized minimum
value
in this case.
are met and the minimum 5.7 shows what would is defined
happen
if the parabolic
the function
at U is less than
function
is defined
by F2 then the value of F2 is not bracketed. If the function another the three
at C. The minimum
then
The routine
the initial
points
using mass,
test points.
the optimization
unsuitable.
- 87 -
Minimum
of parabolic
fit, U, outside
[A,B,C]"
bracketed:
[B,C,U]
not bracketed
F(B)
F2(U)
F(C)
s
F_iu/ AB c u
Outside Bracketing Interval
Minimum
of parabolic
bracketed:
[B,U,C]
not bracketed
i
F(B)
F2(U)
F(C)
FI(U)
Interval
- 88 -
As stated find the value current along because being entire estimate
minimization direction,
procedure
is to
it translates
into a test point for r/k that will result estimate, xk. However, estimate.
in the current
the new of
that would
constraints
For example,
cannot
be larger
procedure
component
li<xi
< Ui
for i = l:n
(5.28)
where:
li = lower ui = upper
bound bound
on xi on xi
of r/j, is desired
such
that xk+l,
when
it is used
in equation given
5.22,
none
will violate
the condition
of equation estimate
would
the components
li = xi + 77Pi
for i = 1 :n
(5.29)
- 89 C_JJ
7-
'c
can be defined
which
would
of the
to their upper
ui = xi + rlpi
for i = l:n
(5.30)
Each equation
exist 2n values
of 7/:
= 1: 2n
(5.31)
Of the 2n possible positive upper negative because values or lower values exist. bound of 7/can
values
of 77 that could
be found
from
n the n
the search
direction estimate.
component
of the
Therefore
as the initial
test point
given
to the bracketing
the optimal
r/ used in equation
rldesired e
= I: 2n
(5.32)
The
value
of
17desired
must
equal
the
minimum
of the
above
by the of the C, is
quantity
it by to calculate
by equation
C = B + (2 - Gr )B = (3-
Gr )B
(5.33)
Since
the initial
guess
for B could
be magnified
by (3 - Gr ):
rldesired
(5.34)
of 1/guarantees
triplet
of abscissas
used
the constraints
the constraints
on the trajectory
were chosen
- 90 -
90"
(5.35)
(5.36)
(5.37)
routine
has found
a triplet
values
of 7"/which
satisfy
the bracketing is a
minimization a combination
routine. of the
section
curve
however, a specified
variation to r/k.
on Brent's
method
of the objective
with respect
in Figure triplet,
lie between
The
DF(B) occurs
and DF(C)
arc linearly
extrapolated
is used as the next trial point. of the objective is calculated: function with respect to r/k,
method is defined
the derivative
Xk+l = Xk + r/k Pk
(5.38)
The gradient
at this estimate
is calculated:
gk+l =VF(xk+l)
(5.39)
-91
Based
on slope
information, [B,C]:
next point
be within
F(B) ........................
_,J/'-
DF(B)
A B C
DF(C)
of next
to try
Figure 5.8 Estimating Location of Function Minimum by Extrapolation (or Interpolation) of Slopes to Zero
The derivative
of the objective
function
with respect
to Ok is approximated
as:
OF(FIK)
_-_xl _-_/
gLlPk
(5.40)
- 92 -
5.5 Gradient
Approximation
The difference
gradient
vector
function small
using
the
"finite to
technique". to fast-order
can be used
approximate
of n variables
(5.40)
component
of the gradient
Thej-th
component
of the gradient
as:
gj(x)
is called Another
the "forward
difference
approximation"
and is exact
by defining:
(5.42)
Subtracting
of equation
F(x
+ _ej)2_j
F(x-
_e i) (5.43)
difference" twice
and is exact
requires
function
evaluations
approximation.
To reduce
involved
optimization difference
procedure, approximation.
it was decided
to approximate
the forward-
- 93 -
the
gradient
vector,
the of the
x. Each
component
by first perturbing
the corresponding
x component
by a small
subtracting
if an objective
evaluations
to approximate
the is
than from
of on-orbit
which
simulation
sources, precision
Scales
and
Press, value.
recommend
square
if the procedure
the objective
as the machine.
The perturbation
was chosen
& = 0.001"
(5.44)
5.6 Conclusions
This chapter automate of on-orbit given mass. utilizes conjugate on-orbit matrix. A.L.S. mass.
various design.
numerical
optimization
methods design
that could
be used to
The objective
of trajectory studied
methods
are designed
objective
function
gradient
methods
properties. methods
gradient mass
method
to be more because
practical
optimization would
This matrix
to approximate
differencing.
- 94 -
gradient an initial
method guess
and described
minimize
and iterates
using
located search
to within directions
specified
of n variables, within
are mutually
n iterations. minimization in Press along were value each search direction was also discussed. with a
one-dimensional
routines of points
used.
The
the minimum
is less than
the function
The second
to narrow
until the minimum for making to the current for the search constraints technique
some
located
of the problem. used to approximate was chosen function the gradient over the central vector was
approximation as many
difference
evaluations. requires
of n
variables,
difference
approximation
- 95 -
Chapter
Six
SIMULATION
AND EVALUATION
6.1 Introduction
scheme
described
in the preceding
with
wind velocity
to define
a Qo_ limit to
optimization.
will allow
the vehicle
of large in-flight
wind dispersions
from the
the ability
procedure in Chapter
trajectories
trajectory trajectory
the optimization
sensitive
to the initial
parameters. lead to the definition trajectory of a new trajectory shape which is described in
Pre-launch
results
are presented
use of this trajectory to the initial 6.5 over shape, discusses the flight the only guess the
in an optimization solution.
encountered in flight.
to do that,
trajectory either a
shape option
parameters
using was
for redesigning
in flight
to use
value
optimization.
- 96-
6.2
Decision
Process
for Choice
of Qa
Limit
of this decision
process
is to establish
a Qalimit
a baseline
dispersions on-orbit
wind
measurement
baseline
by mission
planning needed
in orbit to deliver
fuel mass
to accommodate
insertion
The vehicle
by carrying
This decision
that a bound
wind measurement. are shown in Figure 6.1. This mass, figure mr, that
The shows
assumptions
a hypothetical
on-orbit
different
profiles
a strong
The
of on-orbit winds
strong degrade
to strong while
oppose
vehicle's
motion
performance
performance. are shown bound as curves that flatten out with increasing Qa limit
For a large
Qa limit,
mass
the normal
force constraint
will, in effect,
be removed.
- 97 -
mf baseline
\
|
i
= C}o_ lira C}or,lira rid max] Q_ lira [in flight]
i
: Qo_ lim [design spec]
[TD min]
Figure
6.1:
Assumptions
Made
by Decision
Process
The abscissas
axis shown
in Figure
6.1 are:
by vehicle
experienced loads
or normal
vehicle
structural
aalim
[in flight]
Qot limit to be used in flight for Qa limiting This limit certain must be less than the limit
mode
switch
in Phase design
3 by a
set by vehicle
factor
of safety.
Qtz limit to be used for trajectory must be less than the limit
mode
switch
to accommodate performance,
of attack etc.
by the estimator
-98 -
Qat_
design for
on the Qo_ limit that can be used Qo_ limit than this value
A smaller
This value
will increase
from tailwinds
Having decision
explained
the above
concepts,
it is now possible
to outline
the steps
for the
process:
of curves
of on-orbit
mass vs Qatim,
one of these
achieves
pre-launch
are (100-S)%
S% = percentage be cancelled.
of HWraax
that bounds
wind dispersions
2. R%
The reader
to Figure head
6.2. winds
wind
measurement in flight
is is
be experienced
(R+S)%
of the (R+S)%
curve
line. design
(QlxlimrD)
wind
of R% of HWmax.
4.
Qottimr
The
D
value
of the on-orbit
mass
by the R%
wind
profile
at
is mf
TDinit"
headwind
wind
disturbance,
S% of
fPi
= mf
TDinit
"mf
baseline
(6.1)
- 99 -
mf
l
/ j...4__ ue_ Pad:I i
!
!
!_ow
_(R-S)% HWmax
!
i
"
mf
TD init
_._
_.
h___.:,,v,
_
""i
i
r (R+S)% HWmax
HWmax
mf
baseline
"
,,, !
\i
,,,
'
i
i
! !
Q(/" li
Qot lim
['FD min] ] l-D ] (TO max] [in flight] (design spec]
Pre-Launch
Measurement
linearly with
time to be zero when an altitude is reached where the wind velocity is zero: _ __L__ t/winas}
fp(t) = fpi(a
(6.2)
time when the wind velocity has decreased fuel pad defined,
to zero.
With design
m/ _,,eti,_+ fp(t)
(6.3)
- 100 -
6.3 Pre-Launch
Trajectory
Optimization
Four sets of pre-launch conjugate second trajectory angle runs gradient integration design method time
trajectory
optimization
They utilized
the
profile described
in Chapter
in the course
second
referred
to as the "new alpha profile" runs. used 60% of the Vandenburg Tables of the 69 (Van69) the
The first set of runs for the old alpha profile headwind optimization iterations" were "avg. profile and the second of these set used
results column
runs
for different
Q0_ limit.
required
1D iterations"
refers
to the average
number of iterations
dimensional
optimization
routine on-orbit
parameters.
Figure
maximum on-orbit
on each run vs. the Qot limit. Qot limit. In addition, wind
the maximum
mass
at each Within
100% Van69
profile
- 101 -
Q a Lin_t
Major
Iterations
Avg.
1D
Maximum
mf
Optimum 01",al, 88.3, 87.0, 85.4, 84.6, 0_2 (deg) 14.5, 13.6, 14.0, 13.8, 8.3 8.6 8.8 10.2
(slu[s)
11061.5 11068.3 11075.8 11082.1
Table
6.1:
Old Alpha
Profile
Pre-Launch
Optimization
Results
Headwinds
||
QaLimit
Major
Iterations
Avg.
1D
Maximum
mf
Optimum
0[, al, 89.1, 88.5, I 88.0, or2 (deg) 15.5, 15.1, 14.9, 9.0 8.9 9.9
(slugs)
11052.9 11059.5 11066.8
Table
6.2:
Old Alpha
Profile
Pre-Launch
Optimization
Results
for 100%
Van69
Headwinds
- 102 -
xlO4 1.109
1.108
:-L-
iiii
J
1.103 2200
2300
2400
2500 Q-Alpha
2600
2700
2800
2900
3000
Figure
6.3:
On-Orbit
Mass
Profile
the plots
of angle
of attack
versus headwind.
solutions
of
could
drive
the solution
in which
in the a, region
zero.
At this point
component
corresponding become
to a, becomes invalid.
identically required
It sometimes before
be considered
a solution
was obtained
in which
was greater
than zero.
- 103 -
.......................................
.......
: ...............
:...............
-"
<
.............. . ...............
i..... I
2'0 40 Time 6.4: Old Alpha Profile: Alpha
Increasing
Qo Limit
i......... ............
l()O 120 of 60% Van69
20
60 (sec)
80
Figure
Solutions
Headwinds 16 14 12 10
t_ d_
Used:
8 6 4 2 0
.<
(alpha
profile)
was developed.
The
2) part of the trajectory at the end of the launch was changed to steering
of attack limit that was defined maneuver 6.3 shows percentage optimization
was simply
the error in the Qot at the of the Qot limit for each alpha mode, the was shape the In of
of spending
to the angle
of attack
parameters
it possible
combination converge
the on-orbit
in the number
of trajectory scheme.
parameters
reduces
the number
% Van69
Winds
Error
in Q_z
Table
6.3:
Error
Simulations
of
Design
- 105 -
profile
for 60%
headwinds
mass obtained
obtained for
The resulting
profiles
are shown
Qo_Limit
Major
Iterations
Avg.
ID
mf
185.3,11.3
I 83.7,9.6
Table
6.4:
New Alpha
Profile
Pre-Launch
Optimization
Results
Headwinds
Qct Limit
Major
Iterations
Avg.
1D
my
Table
6.5:
New
Alpha
Profile
Pre-Launch
Optimization
Results
for 100%
Van69
Headwinds
- 106-
1.109
X104
1.108
__=_1.107
t_
1.106 _, 1.105
8
1.104 .................................................................................................
1.1
O_,oo2300
Figure
2400
2500 Q-Alpha
2600
2700
2800
2900
3000
6.6:
On-Orbit
Mass
Profile
12
..... _0
_. 3000
10
8 .......................
< 6
20
40
60
80
100
120
Time (sec)
Figure 6.7: New Alpha Profile: Alpha Plots for Optimal Solutions of 60% Van69
- 107 -
16
14 12 10 8 < 6 4 2 Q Limits Used: 2500 2750 3000
I.......... i............
!
iiiiiiiiiiiiiiiiiiiiiiiill
o zo 4'0
Time
6'0
(sec)
8'0
_oo
12o
Figure
Profile:
Alpha
Solutions
of 100%
Van69
6.4 In Flight
Trajectory
Design
that any
trajectory
design
in flight
should
occur
after
has low
the launch
maneuver.
During
the launch
maneuver,
the altitude
to cause
the vehicle
to deviate
from its
the prelaunch
presented
section,
the
trajectories Qa
spent most of their time flying in-flight trajectory independent in flight, optimization
of attack
O_1 and
limit in
constant
a: 2 is not as a2
practical
because
the trajectory
is almost later
simulation
is begun
in Chapter
can occur
in flight
shapes proved
as that
on varying
the Qa
limit.
The prelaunch
optimization
- 10S -
theon-orbit
the vehicle As
increase
with increasing
that
the winds
experienced again
measured
before
launch. slugs.
Referring A wind
specified
the baseline
curve
at
this Qt_ limit is 11076 The set of optimal was using then used
defined
trajectory
at a Qcx limit of 2750 lb-deg/ft The 100% full simulation Van69. slugs. was run any
to design
a trajectory The
winds
Without
was 11052
conditions, simulation
the
At 60 seconds, wind
with the
The predictive
returned linearly
of 11063
slugs.
to 8 slugs. a search
The baseline
on-orbit
Therefore,
was made to find the Qo_ limit that would to satisfy the requirements
A Qct limit of 2875 was found using orbit this Qa limit. mass resulting The vehicle from
simulation
is shown
in Figure
An engine
the desired
in flight at a higher
- 109-
16
14
; ...............
100
for Stronger
- 110-
Chapter Seven
CONCLUSIONS
AND RECOMMENDATIONS
7.1 Conclusions
was
to develop process
a computer Advanced
program Launch
that
would System in
design
of the
the possibility
of redesigning
the trajectory
or an engine design
the objective
of the A.L.S.
trajectory
a numerical
to determine on-orbit
maximize calculate
the on-orbit
state,
utilized of
variable, integration
or angle
Because
equation time
of motion
step could
be used. was
over
as one second. gradient method gradient The objective was chosen over
a conjugate
types
of numerical
optimization
schemes.
The conjugate
method
knowledge
function
the predictive
parameters.
is approximated
-111
Theoptimizationprocedure wasappliedto the problemof prelaunchtrajectorydesign. Using the trajectoryshape asoriginally proposed, it wasdiscovered thatoptimal solutions producedtrajectoriesin which the time spentin the constantangleof attackportionsof Phase 3 wasvery small.
dependent guess was As a consequence, for the trajectory solution the solution of attack it was found parameters. that optimal In some solutions cases, when were very the initial could the time In this
to a realistic
solution
In these constant
to a condition
in which zero.
portion
the assumptions
underlying
the optimization
and a correct
solution
was not obtained. this problem, that instead a new trajectory of having shape was defined maneuver was taken neither which was similar to the
except
the launch
of attack
to the angle
of the gradient
procedure
the change
of trajectory reduced
reduced dimensions
Even
in trajectory
shape, as
approximately
the same
from
shapes
a small amount
angle of attack
to re,optimize
to approach redesign
a simple in flight
the launch
simulation from
to design
a new
trajectory
Qo limit. be lower
the predictive
simulation
-112-
redesignedwith a higher Qotlimit. If it wasfoundfrom the predictivesimulationthatthe masswould be higher thanneeded to meetthe fuel padrequirements, then the trajectory couldberedesigned with a lower Qa limit to reducestress on thevehicle.
7.2 Recommendations
study could
should
be made
to determine
if more
effective
trajectory
optimization of Phase 3. In by
be substituted
the constant
efficient
transition
the
optimization. for this thesis uses a specified pitch rate profile maneuver of
is commanded
to follow. improvements
Other in payload
launch
capabilities. that
The impact
in this thesis,
It was
capability in which
for the non-failed nominally thrust For It should this be jet the This case.
should then
increase requirements
mass
the trajectory
and on-orbit
are affected
arm from the longitudinal jets to reduce of attack of the should by profile the average
a cant angle
modification
trajectory
Control the
associated
ability
to transients to modify
Provision thrust
to maintain
model.
- 113-
re-tune
the
in payload,
vehicle of
in terms
This automation
would
in the way
preparation. the steering should be method made employed to compare methods for Phase the 3 was acceleration of the steering. direction trajectory
effectiveness
procedure
with other
steering
- 114-
APPENDIX
NASA Vandenburg altitude. the profile Langley AFB, The profile as linearized Research CA. The Center wind provided this study was with wind given measurements plotted taken versus at
information
as velocity A.I.
in Figure
For simplification,
in Figure
Profile
#70
151
E
v
Profile 10000
_.
__/
"ID .,..,
<
5000
Figure
A. I
Vandenberg
profiles.
- 115-
2xlO 4
_i
Vandenberg #70
...i.........
Figure
A.2
Linearized
Vandenberg
profiles.
- 116-