Professional Documents
Culture Documents
Subir Patra
Roll No.09301025
Examiner
~....
. ...........................
\~................
Guide
~.~.
...
""""" ........
Chairman
~~..~.~..1r.'" fu'
Date:
..J:~
I
.
Place:...~.!...
1:?:..~..~.~.~
~
:5.:F..!...0
Declaration
I declare that this written submission represents my ideas in my own words and where others' ideas or
words have been included, I have adequately cited and referenced the original sources. I also declare that
I have adhered to all principles of academic honesty and integrity and have not misrepresented,
fabricated or falsified any idea/data/fact/source in my submission. I understand that any violation of the
above will be a cause for disciplinary action by the Institute and can also evoke penal action ITomthe
sources which have thus not been properly cited or ITomwhom proper permission has not been taken
when needed.
cfMoC?\
SubirPatra
(09301025)
,
iii
Certificate
This is certified that this M.Tech Project Report titled Pitch Control of Flexible Launch Vehicle by
Subir Patra is approved by me for submission. Certified further that, to the best of my knowledge the
report represents work carried out by the student.
Prof.Hari B.Hablani
Guide
iv
Abstract
Due to the use of lightweight composites, launch Vehicles of recent times are more flexible, with their
modal frequencies lower and hence closer to the control bandwidth than earlier. This causes a
destabilizing control-structure interaction in the launch vehicle control loops. The structural modes,
therefore, as in the past, need to be considered in the design of control systems for launch vehicles. The
scope of this project is limited to the pitch control of a flexible launch vehicle in its first stage, tracking
an optimum trajectory to a desired point in space with desired velocity. First, an optimum pitch profile
of a launch vehicle is presented. Second, a pitch controller for a rigid launch vehicle is designed using
the classical control theory. Actuators, deflecting the engine nozzle, modelled as first- and second-order
dynamics are considered. A launch vehicle is modelled as a slender beam, and its modal frequencies and
shapes are determined using Ansys. The first bending mode in the pitch plane is considered in the design
of the pitch controller, and a detailed study of its interaction with the controller is undertaken. In order to
gain stabilize the mode, an unsymmetrical notch filter, tuned with the first bending mode of the launch
vehicle in the control loop, is used. Stability analysis is carried out by means of the root locus, Bode, and
Nyquist plots. Stability margins are determined over entire flight duration at an interval of 20s. Based on
the specifications of gain margin, phase margin, and stability margin, a zone of exclusion that satisfies
these specifications is drawn in a Nyquist plot to show clearly the stability of the designed controller.
Step responses are examined at each 20s interval of the flight time to verify that the time-domain
specifications (percentage overshoot, rise time, settling time) are met.
Table of contents
List of figures
ix
List of tables
xiii
Nomenclature
xiv
1. Introduction
1.1 Motivation
10
13
13
14
4. Pitch Control of Launch Vehicle Rigid Body Dynamics with FirstOrder Actuator
15
15
4.2 Details of parameter variation staring from launch to completion of first stage
18
20
21
22
23
23
27
27
28
29
30
31
32
33
vi
33
34
34
36
36
38
38
42
42
43
44
45
45
45
47
47
48
49
49
51
52
52
52
53
6.5 Slope
54
55
56
56
60
60
63
vii
63
66
68
70
71
72
73
73
74
75
75
75
76
76
77
77
78
79
79
80
81
7.1 Conclusion
81
81
viii
List of figures
Page
Figure No
Title
No
Fig1.1
Fig.1.2
Fig.2.1
Fig.2.2
Fig.2.3(a-b-c-d)
Fig.2.3(e)
Pitch profile
Fig.2.3 (f)
Fig.2.4
11
Fig.3.1
14
Fig.4.1
15
Fig.4.2
17
Fig.4.3
18
Fig4.4
18
Fig.4.5
19
Fig.4.6
20
Fig.4.7
20
Fig. 4.8
21
Fig.4.9
22
Fig.4.10
22
Fig. 4.11
Zone of exclusion
24
Fig. 4.12
25
Fig. 4.13
25
Fig. 4.14
26
Fig.4.15
26
Fig.4.16
27
Fig.4.17
28
Fig.4.18
29
Fig.4.19
Ramp response ( = 0)
30
ix
Fig.4.20
31
Fig.4.21
32
Fig.4.22
Fig.4.23
33
Fig. 5.1
34
Fig.5.2
36
Fig.5.3
37
Fig.5.4
37
Fig. 5.5
Step response of rigid body system to a pitch step command (at t=20sec)
39
Fig.5.6
39
Fig. 5.7
40
Bode plot of rigid body system for updated gain (at t=20 sec)
40
Fig.5.9
41
Fig.5.10
41
Fig.5.11
42
Fig.5.12
43
Fig.5.13
44
Fig.5.14
44
Fig5.15
45
Fig.5.16
46
46
47
47
48
Fig.6.1
51
Fig.6.2
52
Fig.6.3
53
Fig.6.4
54
Fig.6.5
55
Fig.6.6
57
Fig.6.7
57
Fig .6.8
58
Fig.6.9
59
Fig.6.10
59
Fig.6.11
Nyquist plot of the controller without the notch filter, and zone of exclusion
60
Fig.6.12
61
Fig. 6.13
62
Fig.6.14
63
Fig.6.15
Step response for different Place of Notch filter in the control loop
64
Fig.6.15
65
Fig.6.16
65
Fig.6.17
66
Fig.6.18
Nyquist plot of the controller with the notch filter, and zone of exclusion
66
Fig.6.19
67
Fig6.20
68
Fig6.21
68
Fig6.22
69
Fig.6.23
69
Fig.6.24
70
Fig.6.25
71
Fig.6.26
71
Fig.6.27
72
xi
72
Fig.6.29
73
73
74
74
75
Fig.6.34
76
Fig6.35
76
Fig6.36
Nozzle deflection
77
Fig.6.37
77
Fig.6.38
78
Fig.6.39
78
Fig.6.40
tracking error
79
Fig6.41
Nozzle deflection
79
xii
List of Tables
Table
No
Page
Title
Number
Table1
Final conditions
13
Table2
13
Table3
13
Table4
28
Table5
30
Table6
42
Table7
43
Table8
52
Table.9
Tracking error rate (from Matlab) (at flight time t=20s and = 0)
70
xiii
Nomenclature
Aerodynamic force
Control location
Actuator deflection
Angle of attack
Control thrust
Mass of engine
Moment of inertia-pitch
Le
Actuator damping
Amplifier gain
Integrator gain
Distance from centre of pressure in pitch plane to origin of body axis system.
Laplace operator
Time
xiv
xv
Chapter1
Introduction
1.1Motivation
Launch vehicles have very complicated dynamic characteristics due to structural vibrations of
their slender body, fuel sloshing, aerodynamic effects and engine gimbal dynamics. However,
launch vehicles follow simple trajectories determined in advance. The guidance and control
systems of a launch vehicle act together for the vehicle to fly a path, taking it to some desired
terminal conditions. The vehicle is designed to maximize the payload for a given takeoff
weight, and the inert weight of the structure is reduced to a minimum. The trajectory is
designed to cause minimum aerodynamic load in atmospheric phase. The trajectories have
dispersion due to imperfections such as variation of thrust-time curve, aerodynamic
coefficients, autopilot errors, and so forth. These all need not be to be corrected by the
guidance system in the atmospheric phase, because once the vehicle is outside the atmosphere
there is sufficient time for correcting the trajectory. Hence the speed of response is not high
in the atmospheric phase of the flight.
Flexibility must be considered by the control designer if the lowest frequency of the launch
vehicle vibration is less than about six times the desired control bandwidth. Otherwise there
is a possibility that this mode will be destabilized by the control system (the control effort
spills over outside the control bandwidth and destabilizes the vibration mode). Launch
vehicles have bending modes that can be excited by control motion. In one experiment, this
interaction manifested itself as a servo-elastic instability on a test bench when a feedback
loop was closed from a gyro in the nose to the control surface at the tail. The inertial force
associated with the controls excited the first bending mode of the launch vehicle, and a sensor
feedback designed for the rigid launch vehicle had a wrong sign. This is a classic case of the
sensor separated from the actuator by compliance.
The flexibility effect is incorporated in the dynamic model for the control system design due
to following reasons
i) In a flexible launch vehicle, inertial navigation system measures the attitude and angular
velocities of the deflection as well as the rigid body motion, and feed these signals back to the
control loop. If the bending vibration frequencies are near the control system frequency,
feeding back of the deflection degrades the control system stability and, in the worst case,
makes it unstable. Moreover, the interaction of control forces with elastic deformations could
cause undesired excitations, leading to resonance.
deformation angle(F ).
The angles R , F and S versus time are shown in Fig.1.2, and = +
The feedback signal will thus be based not only on the rigid vehicle state but also on
contribution from the vehicle flexibility. Thus structural interaction is caused by the control
loop through the actuator and the sensors. This control-structure interaction could cause
divergent oscillations leading to a structural failure of the launch vehicle. So the control
system design need to ensure that control- structure interaction is stable and does not lead to
diverging oscillations.
ii) When the launch vehicle deforms, the local angle of attack along the length of the vehicle
changes, which introduces additional aerodynamic load on the launch vehicle, increasing the
bending moment along the launch vehicle. Thus aeroelasticity is introduced in the control
system.
Chapter 2
Gravity Turn: Modelling and Simulation
2.1 Dynamics of gravity turn [17]
Since the performance of a launch vehicle depends on the amount of fuel it carries, the only
mass that can be reduced is its structural mass. Further, because the launch vehicle must be
able to withstand high launch loads, which are mostly axial as shown in Fig.2.1, its transverse
strength is sometimes sacrificed to gain longitudinal strength and to decrease structural mass.
Velocity
v
Drag, D
Thrust, T
mg
Fig.2.1 Force acting on satellite booster (T=Thrust, D =Drag) [Wiesel, W .E, McGrawHill]
Launch vehicles are thus weak in the transverse direction; hence their trajectory is designed
to pass through the atmosphere at zero angle of attack. It follows that thrust vector must be
aligned with the velocity vector of the vehicle at all times during the flight, as shown in
Fig.2.1. Also, during the flight to a designated position vector in space orbit, the vehicle
must obviously be rotated from its vertical position of launch to horizontal position at burn
out, with a desired final velocity. This transpires through dynamics automatically by what is
known as gravity turn. To analyze this motion, we need equations of motion in tangentnormal coordinates.
5
Velocity, v
Centre of mass
Local horizon
Drag, D
Thrust, T
mg
h
C
(Trajectorys centre of curvature)
Fig.2.2 Launch vehicle boost trajectory, is the flight path angle [Curtis, H., Orbital
Mechanics for Engineering Student, Elsevier, 2007, p.552]
Satellite launch vehicle forces during powered ascent is illustrated in Fig.2.2, where
T= thrust produced by the nozzle at the base acting along the vehicles longitudinal axis
aligned with or v
1
(2.1)
CD = coeff. of drag,
Tangential acceleration: at = dt
(2.2)
(2.3)
v2
E +h
cos
(2.4)
= m m gsin
(2.5)
Normal acceleration: =
(2.6)
Tangential:
dt
v2
cos
RE + h
RE
E +h
(2.7)
vcos
( 2.8)
h = vsin
(2.8)
1+
(2.9)
Numerical methods are used to solve Eqs. (2.5-9).While doing so, one must account for the
variation of thrust, mass, atmospheric pressure, gravity and . A numerical example of this
gravity turn is shown below. However, in this example, all variations just stated are not
considered, nor is the atmospheric drag considered. In this simulation, the initial flight path
angle at t=0 is taken to be 850 and the desired velocity is 7.71 km/sec corresponding to a
satellite in a circular orbit at an altitude of ~327 km, explained more fully later. From
Fig.2.3b and Fig.2.3f, however, we see that the launch vehicle has achieved a higher
altitude ~500km . This is because here we consider initial perturbation angle as = 850
arbitrarily to initiate a gravity turn trajectory. To get a desired altitude with a desired velocity
we have to choose 0 in such a way that the desired altitude is achieved. This is in fact a
two -point boundary value problem.
As stated earlier, to keep the transverse lifting load close to zero, the angle of attack is
controlled to be close to zero by keeping the launch vehicle longitudinal axis aligned with the
instantaneous velocity vector. At lift off, the launch vehicle is vertical and = 900 . After
clearing the tower and gaining speed, the vernier thrusters or gimballing of the main engines
produce a small, programmed pitchover, establishing an initial flight path angle 0
slightly<900 and some rate 0 .
Thereafter, continues to decrease according to Eq. (2.7) shown above. For instance, if, after
initial pitch over,
Eq.(2.7). But this initial condition must be such that the desired final velocity of the payload,
at the desired altitude, with = 0 and = 0 is achieved. As stated earlier, this is a two-point
boundary value problem but that is beyond the scope of this project.
2.1.2 Analytical solution of tangential velocity versus flight path angle [7]
Assuming that there is no drag and the thrust acceleration is tangential, the velocity equation
Eq. (2.5) simplifies to
v = g 0 sin
(2.10)
(2.11)
We eliminate time t by dividing v equation (2.10) with equation (2.11); the integration of
the subsequent equation of
yields
1 + tan 2
logv = logcos 0
+ constant
1 tan
2
10
(2.12)
Because
1+sin
1sin
1tan
2
1+tan
(2.13)
o
2
= constant
(2.14)
Note that at t = t f (the final time), = 0 because the velocity is horizontal then; therefore,
Eq.(2.14) yields: constant = v(t f )= vf = The desired velocity.
v
The normalized velocity ( v ) is plotted in Fig.2.4 for o =2, starting with the initial value at
f
influence of gravity as the launch vehicle is accelerated. Clearly, the launch vehicle must also
be pitched at the same rate as [Fig.2.3d], Eq. (2.11), and the resulting velocity profile must
yield the desired altitude and the desired horizontal velocity at timet = t f . For a circular orbit
of a satellite injected by a launch vehicle, the final altitude and the final velocity are related
as:
vf =
where = 398600.4415
km 3
s2
(2.15)
R E +h f
km
s
altitude, downrange, flight path angle and the rate profiles are illustrated in Fig.2.3,obtained
by integration of motion equations (2.7)-(2.8). We observe in Fig.2.3a that the desired
velocity is achieved; however, as noted earlier, the final altitude, 500 km, in fig.2.3b and
Fig.2.3f is higher then the desired altitude 327.3298 km. This is a consequence of selecting
the initial = 850 arbitrarily. To determine the pitch profile that achieves desired velocity at
desired altitude is a two-point boundary value problem, related to maximizing the payload
(satellite) mass for a specified horizontal velocity at fixed terminal time subjected to the
terminal constraint of a desired altitude. One simple version of an optimal pitch program
under some simplifying condition is given by [7]
= arctan tano ct
(2.16)
where, the constants 0 and c are chosen to satisfy the terminal conditions. This optimal
solution is referred to as a linear pitch steering law. For details see Ref.7.
12
Chapter -3
Optimal Pitch Profile
3.1 Launch vehicle trajectory optimization
The main objective of the launch vehicle is to place a satellite in an orbit satisfing the
requirements of a particular mission, a geosynchronous transfer orbit (GTO) for instance. For
one particular GTO, a set of, final conditions are:
Final conditions
Apogee
42161 km radius
Perigee
300 km altitude
Latitude
90 deg
Table1.Final conditions [7]
The rocket Nozzle area for the three stages estimated as 2.96, 0.78 and 0.06 sq.m respectively
[7].
Stage
Structure(tonnes)
Propellant(tonnes)
VaccumThrust(kN)
Burn time(sec)
17.5
157
4*748
138
4.325
34
760
130
1.2
10.7
62
735
307.23deg.
Latitude
5.43 deg.
Radius
6378.14 km.
Air velocity
0.0 km.
90 deg.
Table3. Initial conditions (Kourou Launch site for European space Agency)
13
14
Chapter 4
4. Pitch Control of Launch Vehicle Rigid Body Dynamics with FirstOrder
Actuator
4.1 Rigid body Model of Launch Vehicle
Here only the pitch-plane dynamics is studied. Several assumptions are made that allow for
simplification of the equations. The equations of motion of a launch vehicle are complicated
by the fact that the vehicle has time-varying mass and inertia. There can also be relative
motion between various masses within the vehicle, such as fuel sloshing, engine gimbal
rotation, and vehicle flexibility. The derivation of equations of motion neglects nozzle inertia
and sloshing effect.
Dynamic Pressure ( ): =2 2
Aerodynamic Forces: = , =
15
The system dynamics, neglecting the nozzle inertia and sloshing effect, can be written in the
form [2]
d
Cmq
q d
l
Tc lc
2u
q=
+
+
Iyy
Iyy
Iyy
(4.1)
In terms of notations used in Ref.2, Eq. (4.1) can be written more compactly as
(using =
W
V
q = mw w + m + mq q
where mq =
W
V
(4.2)
C mq
d
q
2u
, mw =
I yy
l
I yy
, m =
T c lc
I yy
= mw V + m + mq
= + mq + c
(4.3)
The aerodynamic damping provided by the mq term is usually very small and can be
neglected in the first cut design [2]. The vehicle transfer function is then given by
c
= P = s 2
(4.4)
where c =
T c lc
I yy
L l
I yy
Here we consider a case where the control is provided by secondary injection thrust vector
control where effect of inertia of control effectors can be neglected and the actuator is
considered as a first order actuator, i.e.,
+ K c = K c c
(4.5)
Taking Laplace transform with zero initial states, we obtain the transfer function relating
and as follows
Kc
=G=
c
s + Kc
(4.6)
16
The close-loop transfer function of the signal flow diagram shown in Fig.4.2 is given by
C=
=
K s GP
(4.7)
1+K s GP (1+K r s)
c K s K c
s3 + K c s2 + c K s K r K c s + K c (c K s )
=
s+P
(s 2
K
+ 2sc c + 2c )
(4.8)
(4.9)
2c (K c 2c c )
+
Kc
(4.10)
2c + 2c c K c 2c c +
c K s K c
(4.11)
The gain schedule for the entire flight duration is obtained by taking appropriate values of
and c .
17
Thrust(N)
1550000
1500000
1450000
1400000
1350000
20
40
60
80
100
120
140
160
Flight Time(sec)
Xcg(m)
20
15
10
5
0
0
50
100
150
200
Flight Time(sec)
18
250
300
The aerodynamic load per unit of angle of attack for a launch vehicle
Lalpha(N/rad)
800000
700000
600000
500000
400000
300000
200000
100000
0
0
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.4.5 The aerodynamic load per unit of angle of attack [11]
Aerodynamic load data at three instants of time (0.5 sec, 72 sec and 152 sec) are taken from
Ref.11. At any other time in this time interval the load is determined by interpolation
[Fig.4.5]. Total mass of launch vehicle and its cg position changes during the flight time. This
changes moment of inertia of the launch vehicle during the flight time. Here we have
determined the varying moment of inertia during the first stage of a launch vehicle [Fig.4.6].
19
Moment of Inertia(Kgm2)
1200000
1000000
800000
600000
400000
200000
0
0
20
40
60
80
100
120
140
160
140
160
Flight Time(sec)
Fig.4.6 Moment of Inertia ( ) vs Flight time
T c lc
L l
I yy
I yy
9
8
c(1/sec2)
7
6
5
4
3
2
1
0
0
20
40
60
80
100
Flight Time(sec)
Fig.4.7 Control moment coefficient
20
120
alpha(1/sec2)
3
2.5
2
1.5
1
0.5
0
20
40
60
80
100
120
140
160
Flight Time(sec)
21
Design bandwidth ( )of 6 r/s and damping coefficient (c ) of 0.65 has been chosen.
ii)
Figs. (4.9-10) shows the variation of Gains (Forward gain & Feedback gain) of a typical
launch vehicle.
>6dB
<-6dB.
3. Phase margin
>300
4. Stability margin R= 1 + GH
min
> 0.5
=6dB
= -6dB
Phase margin
=300
The boundary of zone of exclusion for the above specification is given by [Ref.2]:
x2
y2
+
=1
0.752 0.5852
23
6.5
0.265
-0.2
4.3
The close-loop poles of the system with gain K s = 6.5 are shown with cross marks on the
root locus. We see the close-loop poles are in the left half plane of the s-plane.
24
Fig. 4.12 Root Locus for Simplified autopilot (at time t=20 sec)
The gain and phase magnitude (Bode) plot of a rigid launch vehicle and the step response to a
pitch command are shown in Fig.4.13 and Fig.4.14 respectively. While phase margin of the
controller is 50 degrees, the gain margin is infinite.
26
From Fig.4.15, the steady state error can be found as 7E-3. Now we will determine the
steady- state error analytically for comparison.
Considering the autopilot schematic of Fig.4.2 the steady sate response of the system to a
step command c is given by [11]
=
(4.17)
27
The ramp response to a pitch ramp command is shown in Fig.4.16. Here the tracking error
grows continuously. From the ramp response plot, we determined the following tracking
error rate:
Tracking error(deg)
at t=3ses
0.264
at t=8sec
0.30
(0.3-0.264)/5=7.2E-3 deg/sec.
28
10
s
( ) = 7.1047957E 3 (o/sec) which agrees with the simulation results above. Because this
rate error is small, the grown in the tracking error is significant. However, as flight
continues, becomes larger and the error growth also become significant, as will be shown
in Sec.4.7.3.
29
(4-3.75)=0.25
(8-7.75)=0.25
,error
30
4.5
0.28
3.2 sec 2
6.1sec 2
31
Fig.4.21 Commanded pitch rate ( ) and actual pitch rate ( )(at flight time t=100sec)
32
10
, = 3.2 sec 2 , c = 6.1 sec 2 & K s = 4.5 we get tracking error rate ( ) =
33
Chapter-5
5. Pitch Control of Launch Vehicle with Second-Order Actuator dynamics
5.1 Launch vehicle rigid body dynamics
In launch vehicle the control torque is generated by gimballing the engine or by deflecting
the flexible nozzle, so we need to consider actuator as a higher order transfer function
(second and third order) transfer function. In the previous case we have considered first order
actuator.
The dynamic equation considering the influence of engine inertia is written by including the
as follows [2],
Iyy = Tc lc + L l + me Le lc
(5.1)
me Le (s + me Le )
= G = c
Tc
(s 2 )
write, K d = c
(5.2)
m e Le
Tc
c +K d s 2
G=
(5.3)
s 2
2a
=P= 2
c
s + 2a a s + 2a
(5.4)
By using the procedure in [chapter-3], we get a characteristic equation for this system as
follows [2]
s 4 + 2a a + K s K r K d 2a s 3 + 2a + K s K d 2a s2 + K s K r 2a c 2a a s
+ c K s 2a = 0
[Ref. 2]
(5.6)
(5.7)
Here, c and c define the desired pole location of the system and and define the
remaining two poles.
Equating coefficients of like powers of s, we get [Ref.2]
2a a + K s K r K d 2a = 2a c + 2
(5.8)
2a + K s K d 2a = 2 + 2c + 2c c . 2
(5.9)
K s K r 2a c 2a a = 2c . 2 + 2 . 2c c
(5.10)
c K s 2a = 2 2c
(5.11)
Ks =
1K d 2c
2c c K d L + 1 K d 2c P
Kr =
L 1 K d 2c 1 42c P[2c c 2c K d ]
(5.12)
(5.13)
Where,
K d =
Kd
c
(5.14)
M = 2a a 2c c
(5.15)
N = 2a 2c 2c c 2a a 2c c
(5.16)
L = +
2c
N
2a
(5.17)
35
2a a
2c
2c c
P=
+
M
+
N
2a
2a
2a
(5.18)
36
Feedback gain( )
Fig.5.4 Launch vehicle autopilot with 2nd order actuator and integrator [Ref.11]
37
GH= K s (
= K s K r 2a c
w 2a
s 2 +2 a w a s+w 2a
m e Le
s+K i
Tc
Tc
m e Le
s2 +
s 2
c
2
m e L e (s +m e L e )
(1+K r s)
(s 2 )
Tc
1
...(5.20)
s 2 +2 a w a s+w 2a
(s + K )
...(5.21)
the
gain,
at
flight
time
t=20sec
we
set
the
autopilot
gains K s = 3.1 . K r = 0.2531 and K i = 0.4 . We chose K i by trial and error method. K i is
chosen such that control system gives satisfactory performance. For the range K i = 0.4 to 0.9
we get good performance of the control system so we chose K i = 0.4. By using the gain
schedules from Fig.(5.2-3) the settling time, rise time gain and phase margin specifications
are met [Figs(5.5-6)], but the overshoot specification is not met. So we need to increase
feedback gain values to introduce enough damping in the system.
The step response to pitch command and Bode magnitude and phase plots are shown in
Fig5.5 and Fig 5.6 respectively.
38
Fig. 5.5 Step response of rigid body system to a pitch step command (at t=20sec)
Fig. 5.7 Step response of rigid body system to a pitch step command for updated gain
(at t=20sec)
Fig.5.8 Bode plot of rigid body system for updated gain (at t=20 sec)
40
We can see with these gains values all the specification is met [2]
The root locus of the rigid body system is shown in Fig.5.9. The closed loop pole of the
system with gain K s = 3.1 is shown with cross mark on the root locus.
41
-33.5653 +29.6407i
-33.5653 -29.6407i
Aerodynamic pole
-3.3505 + 4.9097i
-3.3505 - 4.9097i
-0.3945
Table6.Close Loop Poles
It is seen from the Fig. (5.9-10), the close loop poles are located on the left half of the splane. So the system is stable.
3.1
0.3
0.2 sec 2
4.3sec 2
From the Fig.5.11 we can see that tracking error grows with time. From the ramp response
plot, we have determined the tracking error rate.
Tracking error(deg)
At time t=3sec.
0.322
0.3249
5.8E-3 deg/sec
43
44
From the Nyquist plot we can see that GH plot does not enter the zone of exclusion. So we
conclude that the stability margin specifications are met [Ref.2].
45
Overshoot(%)
20
18
16
14
12
10
8
6
4
2
0
0
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.5.16 Variation in the overshoot during atmospheric flight of the Launch vehicle for
designed autopilots
3.5
Settling time(sec)
3
2.5
2
1.5
1
0.5
0
0
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.5.17 Variation in the Settling time during atmospheric flight of the Launch vehicle
for designed autopilots
46
0.6
Rise Time(sec)
0.5
0.4
0.3
0.2
0.1
0
0
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.5.18 Variation in the Rise time during atmospheric flight of the Launch vehicle for
designed autopilots
Phase Margine(deg)
40
35
30
25
20
15
10
5
0
0
20
40
60
80
100
120
140
160
Flight time(sec)
Fig.5.19 Variation in the Phase margin during atmospheric flight of the Launch vehicle
for designed autopilots
47
Fig.5.20 Variation in the Gain margin during atmospheric flight of the Launch vehicle
for designed autopilots
Stability margin for a case have good performance during flight trials [2].
48
Chapter-6
Pitch Control of Flexible Launch Vehicle
6.1 Flexible body dynamics [Ref.11]
Flexibility of a launch vehicle is of primary concern to the control engineer because the
sensing instrument (rate gyro) picks up not only rigid body motion but also local elastic
deflection. This introduces an unstable control-structure interaction. This unstable controlstructure interaction could cause divergent oscillations, ending up with structural failure of a
launch vehicle.
Vehicle flexibility influence on the autopilot design is demonstrated by assuming the first
bending mode. Here we assumed that the bending modes are primarily excited by launch
vehicle engine deflection (First-order effect) [11]. The actuator input to the launch vehicle is
only the rigid vehicle state but also contribution due to vehicle flexibility.
Bending equation [Ref.11]
Bending deflection W x, t =
n
i
qi (t)i (x)
...(6.1)
Fi
qi + 2i i q i + 2i qi = M i
...(6.2)
L
0
f(x, t)i dx
...(6.3)
l
0
m(x) i (x) 2 dx
...(6.4)
Considering only the first-order effect, the bending modes is excited by rocket engine
deflection. Considering only the first bending mode, the corresponding generalized force for
a concentrated force is determined as follows. Recall that the Dirac delta function (xa) is
defined by[13],
49
L
0
1 x (x LT ) x
= [Tc + me Le ]1 LT
...(6.5)
The point of application of thrust force is at engine swivel point, x=LT . We normalized
bending modes at engine swivel point by taking 1 LT = 1, therefore Eq(6.5) becomes[11]
1
s2 + 2a w1 s + w12 q1 = M (me Le s 2 + Tc )
...(6.6)
m e Le 2
( +m L )
M1
e e
(s 2 +2a w 1 s+w 21 )
...(6.7)
(6.8)
Where LG denotes the point on the vehicle where a gyro is located and NL1G = L
Rigid body equation [2]
The rigid body pitching motion is coupled to other dynamic modes of the system by
considering the first order effect the rigid body equation becomes
Ir = Tc lc + L l r + me Le lc
Hence,
...(6.9)
T
c
2
m e L e (s +m e L e )
Tc
(s 2 )
...(6.10)
...(6.11)
= = 2 +2
...(6.12)
2
+
50
...(6.13)
...(6.15)
51
s2 + 21 1 s + 12 22
= 2
s + 22 2 s + 22 12
(6.16)
Where 1 is equal to the frequency of the mode that interacts with the controller.
52m
1.13m4
3.79 m2
150 GPa
0.33
The modal frequencies depend upon the mass and length of the launch vehicle. So during the
atmospheric flight the launch vehicle modal frequencies change with time. Here we have
determined the frequencies at three different points of flight time. At any other time in this
time interval these are determined by interpolation. Fig.6.3 shows the lowest frequency of the
launch vehicle. The operation points are 0 sec, 72 sec and 138 sec. [Ref.11]. Fig6.3 shows the
lowest frequency of the launch vehicle.
Fi
(6.17)
Mi
53
W(x,t)=
n
i
q i t i x
(6.18)
L
f(x, t)i
0
dx
(6.19)
l
0
m(l) i (x) 2 dx
(6.20)
The variation of the generalised mass of the lowest frequency mode and the total mass of a
launch vehicle are shown in Fig.6.4 below. At each time instant the generalized mass is
determined by interpolation from the known values at 0, 72sec, and 138 sec.
6.5 Slope (
We determined the slope of the lowest frequency of the mode at the point of gyro location.
The gyro is located at 15 m from the nose of the launch vehicle. We measure the slope at
three different time instant (0 sec, 72 sec, and 138 sec) at gyro location. At any other time the
slope is determined by interpolation [Fig.6.5].
54
.
Fig.6.5 Slope changes in first phase of flight in sensor location[11]
55
s+K i
s
s 2 +2 a w a s+w 2a
Denoting A=c
m e Le
Tc
2 1 a A
A B
and B=
GH = K s K r 2a (A B) s
Where C=
w 2c
c
2
m e L e (s +m e L e )
(s 2 )
Tc
m e Le
M1
m L
T
e e (s 2 + c )
M1
m e Le
(s 2 +2 1 w 1 s+w 21 )
NL1g (1+K r s)
NL1g
Tc
1
) s+
(s 2 +Cs +D)
m e Le
Kr
s 2 +2 a w a s+w 2a s 2 (s 2 +2 1 w 1 s+w 21 )
and D=
... (6.17)
s+K i (s 2 +
... (6.18)
A 21 B
A B
6.7.1 Input to launch vehicle autopilot at the flight time t=20sec [11]
The actuator frequency ( ) =50 rad/s, damping coefficient ( ) =0.7. [14]
Damping coefficient for 1st flexible mode = 0.5% [9]
Mass of engine [me ] = 437.36 kg [11]
Distance between engine CG and gimbal point (Le ) = 0.7680 m. [11]
Control thrust (Tc ) =1454568 N
Generalized mass (1 ) = 33600 Kg
Control moment coefficient ( ) = 4.3sec-2
Aerodynamic moment coefficient ( ) = -0.2 sec-2
Slope per unit length at gyro location (N1L g ) = 0.0896 rad/m.
First modal frequency (at t=20 sec) =17.1 rad/sec.
Fig.6.6 and fig.6.7 show the pitch controller of the flexible launch vehicle with notch filter. In
Fig.6.6, the notch filter filters the control torque command before the actuator, whereas in
Fig.6.7 the notch filter filters the sensor output. These are called location-1 and location-2.
Intuitively, the notch filter after the sensor causes lag in the measurement, whereas the notch
56
filter before the actuator filters out the modal frequency component from the control torque
command.
subsequently.
57
The root locus of pitch controller, shown in Fig.6.1 without notch filter, is shown in Fig.6.8.
The close-loop poles of the system for gain = 3.1 are shown with cross-marks on the root
locus. We see bending poles on the right side of the s-plane implying unstable controlstructure interaction. This instability arises because of noncolocation of the sensor and the
actuator shown in Fig.6.2.
58
The step response to a pitch command is shown in Fig.6.9. Flexible and rigid body responses
are shown separately. The flexible response shows diverging oscillations.
Fig.6.9 Flexible response and rigid body response (at t=20 sec)
The total response to a pitch command is shown below, compared with the rigid body
response.
Fig.6.11 Nyquist plot of the controller without notch filter, and zone of exclusion (at
t=20sec)
From the Nyquist plot we see that GH plot enters the zone of exclusion. So we conclude that
the stability margin specifications are not met [2].
60
17.1 rad/sec and the amplitudes of the peak is well above the zero dB line.
61
Fnotch
s2 + 21 1 s + 12 22
= 2
s + 22 2 s + 22 12
The Notch filter pole-zero patterns is shown in Fig.6.13 and the frequency response is shown
in Fig. 6.14.
62
s+K i
s
) s 2 +2
w 2c
2
a w a s+w a
c
2
m e L e (s +m e L e )
Tc
(s 2 )
m L
T
e e (s 2 + c )
1
(s 2M+2
m e Le
2
1 w 1 s+w 1 )
s 2 +2 s+2 2
(6.19)
Taking A=c
GH =
m e Le
Tc
K s K r 2a (A
where C=
2 1 a A
A B
and B=
B) s
and D=
m e Le
M1
NL1g
Tc
1
) s+
(s 2 +Cs +D)
s 2 +21 1 s+21
m e Le
Kr
2
2
2
2
2
s +2 a w a s+w a s (s +2 1 w 1 s+w 1 ) s 2 +22 2 s+22
s+K i (s 2 +
22
21
(6.20)
A 21 B
AB
component that would otherwise cause excitation of the elastic mode. In location-2, the filter
filters attitude and rate sensor measurements.
Fig.6.15 Step response for different Place of Notch filter in the control loop
In the case of location-2, the step response in steady-state is oscillatory about one, whereas in
location-1 the step response settles to unity without oscillations. This is because, as
experienced earlier, in the case of location-1, the filter filters out the modal frequency
component of the control torque command. So we use the notch filer locaton-1 in the
subsequent autopilot design.
Here we assume that there is 5% error in computation generalized mass and modal frequency
determination. At flight time t=20 sec the computed first bending frequency is 17.1 rad/sec.
With 5% error, modal frequency varies from 17.1 rad/sec to 17.95 rad/sec. Now we will
check whether this Notch filter [2] is suitable for this variation of frequency. The frequency
response of notch filter is shown in Fig.6.14.
The Bode plot of the gain stabilized system with notch filter for bending frequency 17.1
rad/sec is shown in Fig.6.16.
64
65
Now we use the same Notch filter. Bode plot with notch filter is shown below in Fig.6.17.
We see from the controller is now stable and that it satisfies the performance specifications
[2].
Fig.6.18 Nyquist plot of the controller with notch filter, and Zone of exclusion (at t=20
sec)
66
From the Nyquist Plot [Fig.6.18] we see that GH plot does not enter in the zone of
exclusion. So we conclude that control system performance satisfies the stability margin
specifications [Chapter4].
The root locus of the controller with bending filter is shown in Fig.6.19. The close-loop poles
of the system with the gain K s = 3.1 are shown with the cross-marks on the root-locus. We
see that the flexible poles move to the left side of the imaginary axis with a significant
damping coefficient. The change of damping ratio of the flexible poles is from -0.0564 to
0.0113.
67
68
0.3247
0.3624
7.54E-3 deg/sec.
70
71
Overshoot(%)
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.6.27 Variation in the overshoot during atmospheric flight of the Launch vehicle for
designed autopilots
RiseTime(Sec)
0.5
0.4
0.3
0.2
0.1
0
0
20
40
60
80
100
120
140
160
Flight Time(sec)
Fig.6.28 Variation in the Rise Time during atmospheric flight of the Launch vehicle for
designed autopilots
72
Settling Time(Sec)
3.5
3
2.5
2
1.5
1
0.5
0
0
20
40
60
80
100
120
140
160
Flight Time(Sec)
Fig.6.29 Variation in the Settling time during atmospheric flight of the Launch vehicle
for designed autopilots
Gain Margine(dB)
16
14
12
10
8
6
4
2
0
0
20
40
60
80
100
120
140
160
Flight time(sec)
Fig.6.30 Variation in the Gain Margin during atmospheric flight of the Launch vehicle
for designed autopilots
73
Phase Margine(dB)
20
40
60
80
100
120
140
160
Flight time(sec)
Fig.6.31 Variation in the Phase Margin during atmospheric flight of the Launch vehicle
for designed autopilots
Fig.6.32 variation of notch filter frequency during atmospheric flight of the Launch
vehicle for designed autopilots
74
75
77
78
79
80
Chapter7
Conclusion and Future Work
7.1 Conclusion
In this project, we present a pitch control design methodology for a flexible launch vehicle
using classical control theory. We begin with the control system design of a simplified rigid
launch vehicle. We use a first-order actuator. A rigorous study is carried out to determine the
gain schedules for the first stage of the launch vehicle. The transfer function of this simplified
model is developed.
The autopilot design for controlling the pitch incorporating the nozzle inertia is presented
next. Detailed gain scheduling is carried out for the first stage of the launch vehicle. Pitch
step response is examined to check if the time domain specifications (like percentage
overshoot, rise time, settling time) are met. This is done at a 20 second interval of first-stage
flight duration. Frequency-domain analysis is conducted to determine the gain margin and the
phase margin of the controller for the entire first stage of the flight.
We then present the design of an autopilot for controlling the pitch of a flexible launch
vehicle. Here we examine the gain stabilization method. An unsymmetrical notch filter is
added in the loop before actuator. The filter is centred at the first flexible mode. 5% error in
determination of the modal frequency shape and generalized mass is considered. The notch
filter is designed to meet the performance requirement of the control system for this
uncertainty. The gains are selected to meet the rigid body control requirements. If any of the
specifications are not met, the gains are adjusted to meet the specifications. We also
determine the step response at 20 second interval of the flight time of the first stage of the
launch vehicle. Bode plots, Nyquist plots and zone of exclusion are drawn to check if the gain
margin and the phase margin requirements are met for the entire first stage of the flight.
ii) In this project we have included only first bending mode in the control system design. We
suggest to incorporate at least first three modes in the control system design.
iii) Here we have carried out what is known as a short period analysis we need to study the
vehicle performance using a detailed simulation model where all nonlinearities and
disturbances including atmospheric disturbance (wind, gust), longitudinal and lateral
acceleration and yaw control should be carried out. Such a study would be a combination of
six-degree-of freedom simulation and long period analysis and it would includes the effect
of autopilot performance parameters on the overall trajectory.
82
References
[1]Jenkiss, K. and Roy, J., Pitch Control of a Flexible Launch Vehicle, IEEE Transactions
on Automatic Control, ISSN: 0018-9286
[2] Kadam, N.V., Flight control system for Launch vehicle and Missiles, Allied Publishers
private Limited
[3] A.E Bryson, Jr., Control of Spacecraft and Aircraft, Princeton university Press,
Princeton, New Jersey
[4] Plaisted, C.E., and Leonessa, A., Expendable Launch Vehicle Adaptive Autopilot
Design, AIAA Guidance, Navigation and Control Conference and Exhibit18 - 21 August
2008, Honolulu, Hawaii
[5] Matlab control system ToolBox Users guide ,The Math works Inc, 1996
[6]Geissler,E.D.,
Wind
Effect
on
Launch
Vehicle,
Technivision
Services
Slough,England,1970
[7] Noton,M. Spacecraft Navigation and guidance, Springer-verlag London Limited, 1998.
[8]Franklin,G.F, Powell,J.D,and Emami-Naeini,A., Feedback Control of Dynamic
Systems, 5th Edition, Pearson Education, Delhi,2006
[9] Jiann-Woei, J., Abram, A., Robert, H., Nazareth, B., Charles, H., Stephen, R., and Mark,
J., Ares I Flight Control System Design
[10] Kisabo, A.B., Agboola, F., Adebimpe, O.A. and Lanre, A.M., Autopilot Design for a
Generic based Expendable Launch Vehicle Using Linear Quadratic Gaussian (LQG) Control
Approach, European Journal of Scientific Research, Vol.50, No.4 (2011), pp.597-611
[11] Greensite, A. L., Analysis and design of space vehicle flight control systems, Spartan
Books, New York, 1970
83
[12] Ashok joshi., Nabajit Barman., Control system design for flexible missile, M Tech
thesis, IIT Bombay, 2009
[13] http://en.citizendium.org/wiki/Dirac_delta_function
[14] VJTI students internship project report under Prof Hari B. Hablani, IIT Bombay
[15] Bryson, A. E., Control of Spacecraft and Aircraft, Princeton University Press, 1994
[16] Jang, J.-W., Bedrossian, N., Hall, R., Norris, H., Hall, C., and Jackson, M., Initial AresI Bending Filter Design,AAS07-059, February 2007
[17] Hablani , H.B., Research Notes: Space Flight Dynamics and Control, IIT Bombay, 2011
[18] Byung , C.S., Yong-Kyu .P., Woong-Rae .R., and Gwang-Rae. C., Attitude Controller
Design and Test of Korea Space Launch Vehicle-I Upper Stage Intl J. of Aeronautical &
Space Sci. 11(4), 303312 (2010)
[19]Steve.R. Vibration Challenges in the Design of NASAs Ares Launch Vehicles,
NASA/MSFC, September 1, 2009
84
Acknowledgment
I take this opportunity to convey my sincere thanks to Prof Hari B.Hablani for his invaluable
guidance which enabled me to complete this project work successfully. His innovative ideas
and suggestions inspired and guided me to explore new material for effective completion of
the report.
Date:
Subir Patra
Place:
(09301025)
85