AA222 Lecture1

Course for Winter 2011: AE 714-1 Special Topics
Multidisciplinary
Optimization
AA222: IntroductionDesign
to Multidisciplinary
Design Optimization
Joaquim R. R. A. Martins
http://mdolab.engin.umich.edu/index.php/martins
Tuesdays and Thursdays at 1:30-3:00pm in 1012FXB

(0)
(0)
x0 , y t,(0)
xi
0
12
x0
System
Optimization
x0 , y t
x0 , y t
x0
x0 , yjt =i
xi
xi
1,8
7,11(1,8)
8
11
f 0 , c0 , cC
f 0 , c0 , cC
10
9
7,10
Metamodel i
yi
6
1
6
xi
xi
xi
Optimization i
2
5(2)
2
5
f i , ci
f i , ci
4
3
4
yi
yi
yi
Analysis i
AA222
Lecture 1
April 2, 2012
This course is an introduction to numerical optimization and its application to the design of
multidisciplinary aerospace systems. The course will cover:
Mathematical formulation of multidisciplinary design problems: selection of objective functions,

design variables and constraints, examples of aerospace design optimization problems
algorithms
Overview of gradient-based and gradient-free
1
Optimality conditions: unconstrained and constrained, Pareto optimality, aircraft design
AA222: examples
Introduction to MDO
Course Administrative Information

Instructors: Dr. Jason E. Hicken and Prof. Juan J. Alonso
Course Assistant: TBA
Office Hours: J. Hicken: Fridays 2:00-3:30pm (Prof. Alonso: TBA)
Schedule: Mon/Wed 2:153:30pm, McCullough 126
Course Web Page: http://adl.stanford.edu/aa222
Course Mailing List: aa222-class. Subscribe by going to
https://itservices.stanford.edu/service/mailinglists/tools
and entering aa222-class in the Subscribe and unsubscribe to a list box.
Pre-requisites: multivariable calculus, basic linear algebra, familiarity with a
programming language (C, C++, F90/95, Python, MATLAB).
AA222 satisfies the MS and PhD Math Requirement. It is therefore going to
deal with the mathematics of optimization and MDO.
AA222: Introduction to MDO
Homeworks & Projects: You can discuss ideas in groups but you MUST carry
out and write up solutions on your own. Late policy is 10% of grade per day
/ fraction of day. Assignments will start from more mathematical and will
shift towards more applied.
Grading: 75% Homework, 25% Final Project.
Textbooks
No book is required. Detailed course notes will be handed out for every lecture.
However, if you are interested in more details, the following books have been
placed on reserve in the library and we would recommend that you get the first
one (if you are likely to continue to work in optimization).
Optimization Concepts and Applications in Engineering. Belegundu, A. and
Tirupathi, R., Prentice Hall, 1999.
Introduction to Engineering Design Optimization. Onwubiko, C., Prentice
Hall, 2000.
Applied Optimization with MATLAB programming.
Interscience, 2001.
Venkataraman, P.,
Acknowledgements & Additional Courses

The first version of these notes was jointly put together by Prof. Joaquim R.
R. A. Martins, from the University of Michigan, and Prof. Juan J. Alonso,
from Stanford University. This is the fifth+ iteration of these notes which
contains a tailoring for the subject matter covered in AA222, additional lectures
in approximation theory, hierarchical decomposition, and setup of the MDO
problem. In addition, a number of additional interactive optimization problems
have been added, together with new homework problems and graphics for the
explanation of concepts. Prof. Ilan Kroo and Dr. Dev Rajnarayan have
contributed greatly to the current status of these notes.
AA222 is not meant to be a replacement for optimization courses offered
elsewhere at Stanford. If you are interested in optimization you are strongly
encouraged to get a depth of knowledge from some of the following courses:
MS&E 111, 112, 120, 211, 212, 310, 311, 312, 318, Math 151 (or more
advanced), EE 364A/B/S, and STATS 310 A,B,C, among others.
Course Outline
Introduction:
1.
2.
3.
4.
5.
What is MDO?
Terminology and Problem Statement
Classification of Optimization Problems
Methods of Solution
Practical Applications
Single Variable Optimization:

1. Optimality Conditions
2. Line Search Methods
Gradient-Based Optimization:
1. Optimality Conditions
2. Steepest Descent and Conjugate Gradient Methods
3. Quasi-Newton Methods
Sensitivity Analysis:
1.
2.
3.
4.
Finite Differences
Complex-Step Derivative Approximation
Algorithmic Differentiation
Semi-Analytic Methods
Handling Constraints:
1.
2.
3.
4.
5.
KarushKuhnTucker (KKT) Conditions

Penalty and Barrier Methods
Reduced Gradient and Gradient Projection Methods
Sequential Quadratic Programming (SQP)
Constraint Agglomeration
Gradient-Free Optimization:
1.
2.
3.
4.
NelderMead Simplex
DIRECT Method
Genetic Algorithms and Pareto Optimality
Particle Swarm Algorithms
Function Fitting and Regression:

1.
2.
3.
4.
5.
Polynomial approximations
Design of Experiments
Gaussian processes / Kriging
Multi-fidelity approximations
Other topics
MDO Architectures:
1.
2.
3.
4.
Collaborative Optimization (CO)

Concurrent Subspace Optimization (CSSO)
Bi-Level Integrated System Synthesis (BLISS)
Coupled-Sensitivity Analysis
1
1.1
Introduction
What is MDO?
Engineering Design Optimization

The DO in MDO.
In industry, problems routinely arise that require making the best possible design
decision.
However, optimization is still underused in industry. . .
Aerospace is one of the leading applications of engineering design optimization.
Why?
Conventional Versus Optimum Design Process

Specifications
Baseline
design
Specifications
Analyze or
experiment
Evaluate
performance
Change
design
No
Is the design
good?
Yes
Final design
Baseline
design
Analyze
Evaluate
objective and
constraints
Change
design
No
Is the design
optimal?
Yes
Final design
10
Multidisciplinary Design Optimization (MDO)

Most modern engineering systems are multidisciplinary and their analysis is often
very complex, involving hundreds computer programs, many people in different
locations. This makes it difficult for companies to manage the design process.
In the early days, design teams tended to be small and were managed by a single
chief designer who knew most about the design details and could make all the
important decisions.
Modern design projects are more complex and the problem has to be decomposed
and each part tackled by a different team. The way these teams should interact
is still being debated by managers, engineers and researchers [2, 1, 12].
11
1.2
Terminology and Problem Statement
Objective Function
What do we mean by best?
Objective function is a measure of goodness that enables us to compare two
designs quantitatively. Need to be able to estimate this measure numerically...
If we select the wrong goal, it doesnt matter how good the analysis is, or how
efficient the optimization method is. Therefore, its really important to select a
good objective function. Underrated.
Objective function may be linear or nonlinear and may or not be given explicitly.
We will represent it by the scalar f .
12
Is there one aircraft which is the fastest, most efficient, quietest, most
inexpensive?
You can only make one thing best at a time.
13
Example: Trade-off Between Aerodynamics and Strutures

Aerodynamic
Optimization
Sequential optimization does not lead to the

true optimum.
Structural
Optimization
1.6
Aero-structural optimization requires coupled

sensitivities.
1.4
1.2
Lift
1
0.8
Aerodynamic optimum
(elliptical distribution)
0.6
Add structural element sizes to the design

variables.
0.4
Aerostructural optimum
(maximum range)
0.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Spanwise coordinate, y/b
0.8
0.9
Optimizer
Aerodynamic
Analysis
Structural
Analysis
Student Version of MATLAB
Including structures in the high-fidelity wing

optimization will allow larger changes in the
design.
14
Design Variables
Design variables are also known as design parameters and will be represented by
the vector x. They are the variables in the problem that we allow to vary in the
design process.
Optimization is the process of choosing the design variables that yield an optimum
design.
Design variables should be as independent of each other as possible.
Design variables can be continuous or discrete. Discrete variables are sometimes
integer variables.
15
Constraints
Few practical engineering optimizations problems are unconstrained.
Constraints on the design variables are called bounds and are easy to enforce.
Like the objective function, constraints can be linear or nonlinear and may or not
be given in an explicitly form. They may be equality or inequality constraints.
At a given design point, constraints may be active of inactive. This distinction
is particularly important at the optimum.
16
Optimization Problem Statement

minimize
f (x)
by varying
x Rn
subject to hp(x) = 0,
gm(x) 0,
p = 1, 2, . . . , Nh
m = 1, 2, . . . , Ng
f : objective function, output (e.g. structural weight).

x : vector of design variables, inputs (e.g. aerodynamic shape); bounds can be
set on these variables.
h : vector of equality constraints (e.g. lift); in general these are nonlinear
functions of the design variables.
g : vector of inequality constraints (e.g.
nonlinear and implicit.
structural stresses), may also be
17
Classification of Optimization Problems

Smooth
Discontinuous
Linear
Continuity
Linearity
Nonlinear
Static
Continuous
Dynamic
Optimization
Problem
Classification
Time
Quantitative
Discrete
Design
Variables
Qualitative
Deterministic
Data
Stochastic
Constraints
Convexity
Unconstrained
Constrained
Convex
NonConvex
18
1.3
Optimization Methods
Intuition: decreases with increasing dimensionality.
Grid or random search: the cost of searching the design space
increases rapidly with the number of design variables.
Genetic algorithms: good for discrete design variables and very
robust; but infeasible when using a large number of design variables.
Multi-objective optimization.
NelderMead algorithm: simple and robust but inefficient for more
than a few design variables.

Response surfaces: good for noisy functions, still requires a large

number of evaluations to create fit.
Gradient-based: the most efficient for a large number of design
variables; assumes the objective and constraints are smooth
functions.
19
1.4
Practical Applications
Airfoil Design (D. Zingg, UTIAS / M.Nemec, NASA) [10]

Lift-to-drag ratio maximization
64
62
-4
Initial Design
Final Design
-1
-2
58
-2
-1
54
0
0
0.2
0.4
0.6
0.8
1
X/C
-3
56
Cl/Cd
Gradient
52
-4
log(||Gradient||)
60
Cl/Cd
Cp
-3
-5
50
-6
10
20
30
40
50
Flow Solves and Gradient Evaluations
M = 0.25, Re = 2.88 106

20
Drag minimization with fixed lift

0.024
RAE 2822
Final Design
-1
0.022
0.02
0
Cd
Cp
-0.5
0.018
0.5
RAE 2822
Final Design
1
0
0.25
0.5
0.75
1
X/C
0.016
0.014
0.65
0.7
Mach Number
0.75
Single flight condition, M = 0.74.

Baseline: RAE2822, = 2.9o. Final: = 1.9o. Drag reduced by 36.4%.
21
Drag minimization for fixed lift with multiple flight conditions

0.024
-1
RAE 2822
Final Design
0.022
0.02
Cd
Cp
-0.5
0.018
0.5
RAE 2822
Final Design
1
0
0.25
0.5
0.75
1
X/C
0.016
0.014
0.65
0.7
Mach Number
0.75
Four flight conditions: M1 = 0.68, M2 = 0.71, M3 = 0.74, M4 = 0.76.

For M = 0.74, = 1.65o, drag was reduced by 33.8%.
22
Trade-off between Lift and Drag Coefficients
0.012
0.0115
L=0.99
CL=0.549
CD=0.0121
L=0.99
L=0.7
L=0.5
L=0.3
L=0.2
0.011
CD
L=0.1
L=0.05
0.0105
L=0.01
CL=0.132
CD=0.0104
0.01
0.0095
-0.5
-0.4
-CL
-0.3
L=0.01
-0.2
23
Induced-drag Minimization (J. Hicken and D. Zingg, UTIAS) [4, 5]

Induced drag represents approximately 40% of the total drag on a conventional
aircraft in cruise flight. In this example, the induced drag is minimized while the
lift is held fixed. The initial wing has a span of b = 3.
Variables: Wing sections at six spanwise stations control the shape of the wing.
These sections are allowed to twist and scale about the quarter chord, and
the sections can translate vertically. The wing span is also allowed to change
as is the angle of attack.
Constraints: CL and surface area are constrained to their initial values. The
entire geometry is constrained to remain inside the box defined by |y| 2.5
and |z| 0.25
What do you think will happen?
24
Structural Topology Optimization (K. James and J. Martins, UTIAS)

This example is from recent work by James and Martins [6]. The objective of
structural topology optimization is to find the shape and topology of a structure
that has the minimum compliance (maximum stiffness) for a given loading
condition.
25
26
Aircraft Design with Minimum Environmental Impact

The environmental impact of aircraft has been a popular topic in the last few
years. In this example, Antoine and Kroo [3] show the trade-offs between cost,
noise, and greenhouse gas emissions, by solving a number of multiobjective
problems.
787
Blended-wing-body (BWB)
27
28
Pareto fronts of fuel carried, emissions and noise vs. operating cost
29
Pareto surface of emissions vs. fuel carried vs. noise
30
Aerodynamic Design of a Natural Laminar Flow Business Jet

(P.Sturdza and I.Kroo, Stanford) [7]
ASSET Corporation configuration. M = 1.5; Cruise Altitude: 50,000 feet;

Payload: 6-8 Passengers; Range: 5,000 nautical miles; Weight: 100,000 lbs; L/D
= 910; Wings designed with low sweep for natural laminar flow.
31
A CFD Euler code was combined with a boundary-layer solver to compute the
flow on a wing-body. The fuselage spoils the laminar flow that can normally be
maintained on a thin, low sweep wing in supersonic flow. The goal is to reshape
the fuselage at the wing-body junction to maximize the extent of laminar flow
on the wing.
Three design variables were used initially, with quadratic response surfaces and
a trust region update algorithm.
The boundary-layer solution appears superimposed on the inviscid Euler pressures

on the surface grid.
32
Baseline design: a SearsHaack body with wing results in early transition (the
white areas in the boundary-layer solution). N is the measure of laminar
instability, with 1.0 (white) being the prediction of transition. The flow is then
turbulent from the first occurrence of N = 1 to the trailing edge irrespective of
further values of N .
33
From the nose at left, to the tail at right, this is the radius of the original (blue)
and re-designed (red) fuselage after two iterations.
34
With only 3 design variables (the crosses on the fuselage outline that sit on
the wing) and two iterations (not even near a converged optimization) the
improvement is dramatic.
35
With five design variables, and a few more trust-region update cycles, a better
solution is found.
36
The boundary layer is much farther from transition to turbulent flow as can be
seen by comparing the green and yellow colors on this wing with the red and
violet colors two figures ago. Also notice how subtle the reshaping of the fuselage
is.
37
Low-Boom Supersonic Aircraft Design

(F. Palacios, J. Alonso, M. Colonno, J. Hicken and T. Lukaczyk) [11]
The goal in low-boom design is to reduce the strength of the sonic boom
sufficiently to permit supersonic flight overland. The idea is to make subtle
changes to the shape of the aircraft that lead to changes in the sonic boom at
ground level.
NASA N+2 project: Lockheed-Martin configuration

38
Objective: reduce sonic boom strength on the ground by changing shape and
weight of the aircraft.
39
Pressure field (lower half) and equivalent-area shape sensitivity fields.
40
Aero-Structural Design of a Supersonic Business Jet

(J.Martins, UTIAS / J.Alonso, Stanford / J.Reuther, NASA) [8, 9]
Baseline configuration and design variables
WYX"Z![\]X5^`_acb"d7e+f#ghji"[klf4[m\ d7e
n oqpYr4s5t+uvwpyx"zz{x}|~u
LNMPO5Q5RSTU"V
ACBED5F5GHI"J"K
&')(+*,.-0/21",435*5-67(8(:9;=<?>@(
"7". .
!"#"$%

41
Optimized design
LNM
V W!XZY\[]
O
O
P QPPRSUTT

^_^U`aR bdc#e
!#"%$'&)( +* ,$-"./

02143 5 -6 7.897:7<;>=?8@7A7CBEDCF38!GIH8!=KJ

42
Design of Efficient, Quite Open Rotors

(T. Economon and J. Alonso, Stanford)
Open-rotor engines offer significantly improved fuel efficiency relative to ducted
turbofan engines (perhaps as high as 25%). However, these engines also produce
unacceptable noise. In this work, high-fidelity CFD is coupled to an aeroacoustic
model to analyze the performance and noise of a rotor. This analysis will then
be placed inside an optimization framework to shape the geometry of the rotor.
43
Design of Low-Cost Launch Vehicles (M. Colonno, J.Alonso, Stanford,

SpaceX)
Cost: Current U.S. LVs

rewards.
$ 40,000 /kg to LEO. Small improvements yield big
Performance: Payload mass to orbit depends exponentially on many vehicle

parameters.
Coupled environments:
loading tightly coupled.
Wide-ranging aerodynamic, thermal, and structural
44
High-fidelity Aerostructural Optimization of a Subsonic Transport (G.

Kenway, G. Kennedy, and J. Martins)
Problem: Maximize the Breguet range of the Common Research Model
configuration (approx.
dimensions of B777) subject to aerostructural
constraints.

V L
W1
max
ln
cD
W2
Variables: free-form shape variables (154), twist variables (6), tail rotation angle
(2), angle of attack (2), and structural thickness (160).
Constraints: Lcruise MTOW, Lmaneuver 2.5MTOW, Cm,cruise = 0,
Cm,maneuver = 0, KSmaneuver 1, and 27 wing thickness constraints.
45
References
[1] N. Alexandrov and M. Y. Hussaini, editors.
Optimization: State-of-the-Art. SIAM, 1997.
Multidisciplinary Design
[2] N. M. Alexandrov and R. M. Lewis. Comparative properties of collaborative

optimization and other approaches to MDO. In Proceedings of the First
ASMO UK / ISSMO Conference on Engineering Design Optimization, 1999.
[3] N. E. Antoine and I. M. Kroo. Aircraft optimization for minimal
environmental impact. Journal of Aircraft, 41(4):790797, 2004.
[4] J. E. Hicken and D. W. Zingg. Aerodynamic optimization algorithm with
integrated geometry parameterization and mesh movement. AIAA Journal,
48(2):400413, Feb. 2010.
[5] J. E. Hicken and D. W. Zingg. Induced-drag minimization of nonplanar
geometries based on the Euler equations. AIAA Journal, 48(11):25642575,
Nov. 2010.
46
[6] K. James, J. S. Hansen, and J. R. R. A. Martins. Structural topology

optimization for multiple load cases using a dynamic aggregation technique.
Engineering Optimization, 41(12):11031118, December 2009.
[7] I. M. Kroo, R. Tracy, J. Chase, and P. Sturdza. Natural laminar flow for
quiet and efficient supersonic aircraft. AIAA Paper 2002-0146, 2002.
[8] J. R. R. A. Martins. A Coupled-Adjoint Method for High-Fidelity AeroStructural Optimization. PhD thesis, Stanford University, Stanford, CA,
2002.
[9] J. R. R. A. Martins, J. J. Alonso, and J. J. Reuther. Complete configuration
aero-structural optimization using a coupled sensitivity analysis method.
In Proceedings of the 9th AIAA/ISSMO Symposium on Multidisciplinary
Analysis and Optimization, Atlanta, GA, September 2002. AIAA 2002-5402.
[10] M. Nemec, D. W. Zingg, and T. H. Pulliam. Multi-point and multi-objective
aerodynamic shape optimization. AIAA Paper 2002-5548, 2002.
[11] F. Palacios, J. Alonso, M. Colonno, J. Hicken, and T. Lukaczyk.
Adjoint-based method for supersonic aircraft design using equivalent area
47
distributions. In 50th AIAA Aerospace Sciences Meeting, number AIAA2012-0269, Nashville, Tennessee, United States, 2012.
[12] J. Sobieszczanski-Sobieski and R. T. Haftka. Multidisciplinary aerospace
design optimization:
survey of recent developments.
Structural
Optimization, 14(1):123, 1997.
48

AA222 Lecture1

Uploaded by

Copyright:

Available Formats

AA222 Lecture1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AA222 Lecture1

Uploaded by

Copyright:

Available Formats

Course for Winter 2011: AE 714-1 Special Topics

Tuesdays and Thursdays at 1:30-3:00pm in 1012FXB

Mathematical formulation of multidisciplinary design problems: selection of objective functions,

Course Administrative Information

AA222: Introduction to MDO

AA222: Introduction to MDO

Acknowledgements & Additional Courses

AA222: Introduction to MDO

Single Variable Optimization:

AA222: Introduction to MDO

KarushKuhnTucker (KKT) Conditions

AA222: Introduction to MDO

Function Fitting and Regression:

Collaborative Optimization (CO)

AA222: Introduction to MDO

Engineering Design Optimization

AA222: Introduction to MDO

Conventional Versus Optimum Design Process

Multidisciplinary Design Optimization (MDO)

AA222: Introduction to MDO

Terminology and Problem Statement

AA222: Introduction to MDO

You can only make one thing best at a time.

AA222: Introduction to MDO

Example: Trade-off Between Aerodynamics and Strutures

Sequential optimization does not lead to the

Aero-structural optimization requires coupled

Add structural element sizes to the design

Spanwise coordinate, y/b

AA222: Introduction to MDO

Including structures in the high-fidelity wing

AA222: Introduction to MDO

AA222: Introduction to MDO

Optimization Problem Statement

f : objective function, output (e.g. structural weight).

AA222: Introduction to MDO

structural stresses), may also be

Classification of Optimization Problems

AA222: Introduction to MDO

Response surfaces: good for noisy functions, still requires a large

AA222: Introduction to MDO

Airfoil Design (D. Zingg, UTIAS / M.Nemec, NASA) [10]

M = 0.25, Re = 2.88 106

Drag minimization with fixed lift

Single flight condition, M = 0.74.

AA222: Introduction to MDO

Drag minimization for fixed lift with multiple flight conditions

Four flight conditions: M1 = 0.68, M2 = 0.71, M3 = 0.74, M4 = 0.76.

AA222: Introduction to MDO

Trade-off between Lift and Drag Coefficients

AA222: Introduction to MDO

Induced-drag Minimization (J. Hicken and D. Zingg, UTIAS) [4, 5]

AA222: Introduction to MDO

Structural Topology Optimization (K. James and J. Martins, UTIAS)

AA222: Introduction to MDO

AA222: Introduction to MDO

Aircraft Design with Minimum Environmental Impact

AA222: Introduction to MDO

!"#"$%