25/11/2014

ONERA-ISAE / Airbus
Presented by: Yann DENIEUL
PhD Student

Integrated Design and

Control of a Flying Wing
Using Nonsmooth
Optimization Techniques
PhD Thesis
Thesis directors: Daniel Alazard & Joël
Bordeneuve (ISAE-ONERA)
Supervisors: Clément Toussaint (ONERA)
Gilles Taquin (Airbus-EIXOG)
 25/11/2014

Contents

1. Problem setup

2. Integrated Design and Control

3. Way forward

Page 2

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Contents

1. Problem setup

2. Integrated Design and Control

3. Way forward

Page 3

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Context and problem

Previous PhD thesis: “Handling Qualities resolution for Blended-Wing

Body” (Saucez, 2013) at the Future Projects Office in Airbus, Toulouse and
ISAE-Supaero
Flying Wing configuration very promising
Handling qualities were a major challenge and not deeply studied yet

Initial configuration:

Page 4

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Context and problem

Previous PhD thesis: “Handling Qualities resolution for Blended-Wing

Body” (Saucez, 2013) at the Future Projects Office in Airbus, Toulouse and
ISAE-Supaero
Flying Wing configuration very promising
Handling qualities were a major challenge and not deeply studied yet

Initial configuration:
PhD conclusions:
No major showstopper
concerning Handling qualities
Vertical surfaces needed
Multicontrol surfaces necessary
Active control mandatory

Page 5

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 April 2nd 2014

Context and problem

• Focus on the need for active stabilisation

Natural Aircraft

Page 6

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 April 2nd 2014

Context and problem

• Focus on the need for active stabilisation

Control
laws

Natural Aircraft Augmented Aircraft

• Consequences of the active stabilization:

High-rate control surfaces
create large secondary power demand
• Cost of instability on A/C design
Page 7

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 Use Tab 'Insert - Header & Footer' for Presentation Title - Siglum - Reference Month 20XX

How unstable is the flying wing? Longitudinal modes

• Max instability: 1,25 rad/s @ Mlight & low Mach

• Impact on actuators Bandwidth? Optimal actuators sizing?

Page 8

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Contents

1. Problem setup

2. Integrated Design and Control

3. Way forward

Page 9

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Control Problem Setup

• Aircraft Model:
𝛿𝑉
𝛿𝛼
• Longitudinal Model, 4 states 𝑋 =
𝛿𝑞
𝛿𝜃
• 11 Controls

𝛿𝛼
• Measures for control: Y = 𝛿𝑞
𝛿𝜃

Page 10

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Standard form for H2/H∞ control

• Standard form:

Page 11

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Standard form for H2/H∞ control

• Standard form:

• H2/H∞ control problem:

Page 12

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Standard form for H2/H∞ control

• Standard form:

• H2/H∞ control problem:

• Weighting function on pitch acceleration:

Page 13

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Controller structure

• Static 11x3 State-feedback

• Linear Control Allocation is performed by the compensator

• Allocation strategy is given by the optimisation
• How to mix Nonlinear Control Allocation (ie including saturations) with structured
controller is an open question for us

Page 14

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design approach

• Co-design:
• Meaningful physical parameters are considered as controller parameters

• Physical parameters are optimised in the controller synthesis

• Example: in (Alazard et al., 2013), a delay accounting for sensor quality is

optimised conjointly with an attitude controller

Page 15

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design approach

• Co-design:
• Meaningful physical parameters are considered as controller parameters

• Physical parameters are optimised in the controller synthesis

• Example: in (Alazard et al., 2013), a delay accounting for sensor quality is

optimised conjointly with an attitude controller

• In litterature:
• Integrated design and control (also known as plant-controller optimization) was
performed using LMI framework (Niewhoener et al., 1995)
• Full-order controllers
• Handling Qualities requirements hardly translated into H∞ constraints

• Adress this problem using nonsmooth optimization tools for structured controllers
Page 16

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design approach

• Parametrized first-order bandwidth:

Page 17

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design approach

• Parametrized first-order bandwidth:

• Closed-loop model for synthesis:

Page 18

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design approach

• Parametrized first-order bandwidth:

• Closed-loop model for synthesis:

• New optimization problem:

Page 19

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Systune

• Why using Systune?

• Allows for mixed H2/H∞ synthesis and multiobjective optimization.

Page 20

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Systune

• Why using Systune?

• Allows for mixed H2/H∞ synthesis and multiobjective optimization.

• Allows for structured parameters for the controller and physical parameters;
bounds on these variables are easily applicable

Page 21

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Systune

• Why using Systune?

• Allows for mixed H2/H∞ synthesis and multiobjective optimization.

• Allows for structured parameters for the controller and physical parameters;
bounds on these variables are easily applicable

• Directly specifying closed-loop structure and tunable blocks. Single Simulink

model for linear synthesis and nonlinear simulation.

• Variety of constraints: H∞ but also pole placement constraints: more applicable

for Handling Qualities purpose

Page 22

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

First results

• Frequency-domain response of
𝑇𝑤→𝑧𝑖𝑛𝑓 (blue) and 𝑊 −1 (yellow)

Page 23

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

First results

• Frequency-domain response of
𝑇𝑤→𝑧𝑖𝑛𝑓 (blue) and 𝑊 −1 (yellow)

• Comparison: 𝑇𝑤→𝑢 = 0.6 for LQ 2

minimal energy control
• 𝑇𝑤→𝑢 2 = 1.26 for mixed H2/H∞
control with infinite bandwidth
Page 24

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

First results

• Frequency-domain response of Controller:

−1
𝑇𝑤→𝑧𝑖𝑛𝑓 (blue) and 𝑊 (yellow)

• Comparison: 𝑇𝑤→𝑢 = 0.6 for LQ 2

minimal energy control
• 𝑇𝑤→𝑢 2 = 1.26 for mixed H2/H∞
control with infinite bandwidth
Page 25

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Co-design first results

Page 26

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Contents

1. Problem setup

2. Integrated Design and Control

3. Way forward

Page 27

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Way forward
• Control surfaces size parametrization and aero model calculation

2D Sketcher
Page 28
AVL Mesh

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Way forward

• LFT form of the parametrized aerodynamic model

• Co-design on the LFT form for control surfaces sizing

Page 29

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Way forward

• LFT form of the parametrized aerodynamic model

• Co-design on the LFT form for control surfaces sizing

• Pole placement constraints

Page 30

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Way forward

• LFT form of the parametrized aerodynamic model

• Co-design on the LFT form for control surfaces sizing

• Pole placement constraints

• More physical criterion: mass / energy minimization through actuators mass

models

Page 31

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Thank you for your attention

Questions?
Page 32

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Annexes

Page 33

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Simulink

Page 34

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Context and problem

• Focus on the need for multicontrol

 Internship on multicontrol allocation

 Developement of a control allocation module

Page 35

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Control allocation: Attainable Moments Subset

Page 36

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

Control allocation for multicontrol surfaces - EIXOG - Ref. PR1316811 - Issue 1Use Tab 'Insert - Header & Footer' for Presentation Title - Siglum - Reference

Direct control allocation

• Calculation of the intersection facet

Page 37

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.

 25/11/2014

Page 38

© AIRBUS Americas. All rights reserved. Confidential and proprietary document.