Master of Aerospace Engineering

BE : Control of Launcher H∞ Control Practice

Report

Authors : M. Oscar L’H OPITAULT

M. Alexandre D’O RGEVAL
M. Gerald I MHOF

Due date of report: 14/04/2021

Actual submission date: 14/04/2021
Contents
1 Introduction 2

2 Rigid launcher models 2

2.1 Time domain simulation model (Rigid Launcher) . . . . . . . . . . . . . . . . . . . 2
2.2 Synthesis model (rigid launcher) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Augmented system 4
3.1 Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Augmented model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4 First H∞ controller synthesis and assessment 7

4.1 H∞ controller synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Frequency domain assessment (rigid launcher) . . . . . . . . . . . . . . . . . . . . . 9
4.3 Time domain simulation (rigid launcher) . . . . . . . . . . . . . . . . . . . . . . . . 10

5 Controller Settings 12
5.1 H∞ rigid motion controller - sensitivity study . . . . . . . . . . . . . . . . . . . . . 12
5.2 H∞ rigid motion controller settings . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1
1 Introduction
This workshop aims to synthesize and assess a H∞ for a space launcher. The H∞ design is applied on
the controller of a heavy launcher of an ARIANE 5 class, during its atmospheric ascent trajectory.

2 Rigid launcher models

2.1 Time domain simulation model (Rigid Launcher)
In the statement, ∆α is given as :
∆Ż − ∆W
∆α = ∆θ +
VR
Which can also be written as :
Y = CX + DU
With :  
∆θ̇  
 ∆θ  ∆βc
 
Y = ∆α  ∆Ż 
X = U =  ∆W 

∆β˙R  ∆βFZ
∆βR
And matrices C and D filled as :
 
0  
1 0
1
C=  
 VR  D = − V1R 
0 0
0

As a reminder, A and B are given below :

   
0 A6 A3 0 K1 0 B1 K1
1 0 0 0 0  0 0 0
   
A=  0 A 1 A 2 0 K2   02 B2 K2 
B=  

0 0 0 −2ξ0 ω0 −ω 2  ω 0 0
0 0
0 0 0 1 0 0 0 0

To be filled in Ẋ as :
Ẋ = AX + BU
The rigid launcher is modeled as described in the figure below with A, B, C and D given above.

2
 Figure 1: Rigid launcher Open loop

According to the following plot, the uncontrolled launcher is indeed naturally unstable :

Figure 2: Rigid launcher Open loop output

2.2 Synthesis model (rigid launcher)

The SISO model is the rigid launcher open loop without wind nor deflection bias.Commanded deflec-
tion is the only input. Numerical data of table 1 from the statement were used for this diagram.

3
 Figure 3: Rigid launcher Open loop (SISO model)

Figure 4: Rigid launcher Open loop output (SISO model)

According to the previous plot of the angle of attack, the behaviour of the launcher is naturally unsta-
ble.
Eigenvalues are to be modified to control the launcher behavior.

3 Augmented system
3.1 Weights
The figure below represents the augmented system that we use for H∞ synthesis.

4
 Figure 5: Augmented system for H∞ synthesis

The output errors Z1 and Z2 are related to the disturbances W1 and W2 through:
    
Z1 T11 T12 W1
=
Z2 T21 T22 W2

With :    1 G 
T11 T12 W1 1+KG −W1 1+KG W3
TZW = = K KG
T21 T22 W2 1+KG −W2 1+KG W3
W1 , W2 and W3 associated transfer functions are given in the statement.
1 1 1
The following figure gives the Bode diagrams of the inverse transfer functions W1 , W2 and W3 .

5
 1 1
(a) Bode diagram of W1 (b) Bode diagram of W2

1 1
(a) Bode diagram of W1W3 (b) Bode diagram of W2W3

For the bode diagram of W11 , the magnitude increases from a step of -22dB to a step of 10 dB.
The system attenuates for a frequency range [10−3 ; 0] rad.s−1 and amplifies for a frequency range
[0; 102 ] rad.s−1 . W2 and W3 being constants, bode diagram of W12 describes a constant magnitude and
phase as bode diagram of W21W3 should, W21W3 seems wrong. W11W3 shall have the same behaviour than
1
W1 since W3 is a constant and it is correct.

3.2 Augmented model

The figure below shows the full augmented model including disturbances.

6
 Figure 8: Full augmented model

4 First H∞ controller synthesis and assessment

4.1 H∞ controller synthesis
At the end of the γ - iteration process, the algorithm gives γ = 63.45
Eigenvalues computed on Matlab are given below :
(−0.0220 + 0.0000i).10−3
(−9.2224 + 0.0000i).10−3
(−0.0148 + 0.0187i).10−3
(−0.0148 − 0.0187i).10−3
(−0.0051 + 0.0000i).10−3
(−0.0001 + 0.0000i).10−3
(−0.0001 + 0.0000i).10−3
After computation of eigenvalues, the following plot diagram is obtained for the transfer function
K(s).

Figure 9: Bode diagram of K(s)

7
This bode diagram has a particular behavior. It is a passing filter for the range [10−2 ; 0] rad.s−1 , then
the system attenuates until a frequency of 10 rad.s−1 and finally, the filter behaves as a low-pass filter
for the range [10; 106 ] rad.s−1 .

1 γ G G
(a) Bode diagram of 1+KG and W1 (b) Bode diagram of 1+KG and W1W3

K γ KG γ
(a) Bode diagram of 1+KG and W2 (b) Bode diagram of 1+KG and W2W3

8
4.2 Frequency domain assessment (rigid launcher)

Figure 12: Bode diagram of the Open Loop of the rigid launcher

Figure 13: Nichols diagram of the Open Loop of the rigid launcher

9
Gain, phase and delay margin given by Matlab are the following ones.

1
For 1+KG :

GainMargin: [0.4178 ∞] dB
PhaseMargin: 171.4633◦
DelayMargin: 3.1688 dB

K
For 1+KG :

GainMargin: [0.6615 6.7309] dB

PhaseMargin: 24.6242◦
DelayMargin: 0.2896 dB

4.3 Time domain simulation (rigid launcher)

Figure 14: Time domain behavior of the controlled rigid launcher : Zero guidance input, no wind,
constant thrust deflection offset = 0.1◦

The launcher is now stable. Deflection angle and angle of attack stabilize to the wanted value after
reaching it.

10
Figure 15: Time domain behavior of the controlled rigid launcher : Zero guidance input, no thrust
deflection offset, wind gradient + gust profile.

With wind and gust profile added, the launcher deflects but the error tends to be corrected and the
angle of attack tends to stabilize to the wanted value.

11
5 Controller Settings
5.1 H∞ rigid motion controller - sensitivity study

Nominal k2= 0.02 k2=0.08 a2<b2 a2>b2 k3=0.35 k3=1.4

y 0.9157 0.6759 1.0946 0.7303 1.0006 0.7944 0.8969
Stability margins
Controller gain (dB) 30.8 33.4 29.1 32.7 29.9 28.3 36
LF gain margin (dB) -7.53 -8.84 -6.91 -8.45 -7.32 -8.06 -7.89
HF gain margin (dB) 9.82 11.9 12.6 10.9 14.5 13.1 12
phase margin (rad) 32.4 39 31.3 36.5 34.1 36.7 33
Time domain response
Case 1 (deflection offset)
Max Attitude angle (°) -0.058 -0.051 -0.041 -0.058 -0.042 -0.037 -0.034
Maximum AoA (°) 0.053 0.046 0.038 0.053 0.039 0.035 0.033
Max Thrust deflect. (°) -0.523 -0.663 -0.857 -0.531 -0.076 -0.667 -0.835
Deflection integral 153.4 184.4 327.4 243.7 243.3 244.6 267.6
Drift (m/s) 0.078 0.063 0.046 0.078 0.051 0.035 0.029
Case 2 (Wind gust)
Max Attitude angle (°) -0.805 -0.768 -0.584 -0.892 -0.631 -0.809 -0.549
Maximum AoA (°) 5.042 5.022 4.914 5.091 4.943 5.033 4.909
Max Thrust deflect. (°) -4.042 -3.985 -3.975 -4.041 -3.91 -3.98 -3.67
Deflection integral (°) 214.8 186.7 328.7 155.8 245.3 213.1 246.3
Drift (m/s) 35.12 35.08 34.99 35.16 35.01 35.21 34.8

Table 1: Synthesis of parameters analysis in two case studies

5.2 H∞ rigid motion controller settings

Regarding the frequency assessment, we can see in table 1, all the margins are respected.

Concerning the behaviour with a wind gust, we can see that with the parameters a2 < b2 , we obtain a
maximum AoA of 5.091° and a deflection integral of 155.8°, which are the closest to the requirement.

Moreover, keeping k2 at 0.04 is the better choice to minimise the effects of thrust deflection offset.

12