Design Optimization

Footer Text 2016-10-24 1

 Today’s Lecture
• Introduction to Design Optimization
o What is it?
o How does it work?
What kind of methods are available?
o How can it be used for System Design?
o Some examples of Design Optimization applied to engineering systems

Footer Text 2016-10-24 2

 Introduction
What is System Design?

Examples of Engineering Systems

Footer Text 2016-10-24 3

 Introduction (2)
• Aircraft System Design

Footer Text 2016-10-24 4

 Introduction (3)
Cost Group Propulsion Group

Pilots Choice

Fuselage Group
Optimum Design
Empennage Group

Wing Group
Weight Group

Undercarriage Group
Weapon Group

Footer Text 2016-10-24 5

 Traditional Design Hierarchy
•General arrangement and performance
Mission •Representative Configurations
Requirements •General internal layout

Conceptual Conceptual
Design Baselines
•Optimization •Systems Specifications
•Parametric •Detailed Subsystems
Preliminary Selected
•1st level analysis Design Baseline •Internal Arrangements
•Process Design
•Sophisticated analysis
•Problem Decomposition Detailed Production
•Disciplinary Optimization Design Baseline

Production
 May lead to Sub-optimal designs and Support
Source: AIAA MDO White Paper, 1991

Footer Text 2016-10-24 6

 Traditional Approach To Product Development
100% CONCEPTUAL 100% PRELIMINARY 100% DETAILED 100%

KNOWLEDGE ABOUT DESIGN

• AERODYNAMICS
• PROPULSION
• STRUCTURES
• CONTROLS
• MANUFACTURING
• SUPPORTABILITY
• COST

DESIGN FREEDOM

TIME INTO DESIGN PROCESS

• Short Conception phase + unequal distribution of disciplines
o limited scope for optimization & assessing impact of inter-disciplinary
o couplings.
• Correction of later problems
o Costly/ Lost Time/ Futile
o Solutions limited to specific discipline

Source: AIAA MDO White Paper, 1991

 The Design Process Paradigm Shift
100% CONCEPTUAL 100% PRELIMINARY 100% DETAILED 100%

KNOWLEDGE ABOUT DESIGN

GOAL

• AERODYNAMICS
• PROPULSION
• STRUCTURES
• CONTROLS
• MANUFACTURING
• SUPPORTABILITY
GOAL
• COST

DESIGN FREEDOM

TIME INTO DESIGN PROCESS

Design process reorganized to gain information earlier and to retain design freedom longer
• More up-front design
• More evenly distributed efforts of disciplines in early design
• Alleviate paradox
• Design decisions/trade-off reordered

Source: AIAA MDO White Paper, 1991

 Modernizing Design
• How do we achieve this
improvement?
• One way is to implement
numerical optimization
o Integrate all of the different
analysis methods that drive the
design
Aircraft Design Example
Aerodynamics
•
o Aerodynamics
o Structures Structures
o Performance
• Implement numerical
optimization algorithms Performance

Footer Text 2016-10-24 9

 Design Optimization
• Design Optimization definition:
o Design Optimization consists of mathematical processes for
determining the maximum or minimum of a function
o OR: Make something the best it possibly could be

• Components of a Design Optimization Problem

o Design Variables
o Objective Functions
o Constraint Functions
o System Equations

Footer Text 2016-10-24 10

 Objective Functions
• In order to select between design alternatives, we
need a measure of ‘goodness’ that allows us to
compare different designs quantitatively
• We need to be able to estimate this numerically
• Objective functions need to be chosen carefully
o Should reflect the major design goal of the system
o If we choose the wrong objective function, the results will be useless
• The objective function will lead the design to
being the best possible in that category
• Example: If we’re designing an airplane, we could
use the following
o Weight, speed, range, or any other measurable quantity
• Objective functions are represented as

Footer Text 2016-10-24 11

 Objective Functions (2)
• Is there one aircraft which is best in every way?
The lightest, fastest, most efficient, cheapest?

• We can only make one thing the best at a time

• That thing is the objective function

Footer Text 2016-10-24 12

 Design Variables
• Design variables are parameters we can change
to adjust the design
• Optimization is the process of choosing the design
variables that yield an optimum design with
respect to the objective function
• We need to select the design variables carefully
o They should be independent of each other
o They should affect the objective function in a significant way

• Design variables are represented by

• is a vector

Footer Text 2016-10-24 13

 Design Variables (2)
• Aircraft Wing Design Example
• Objective function: maximize L/D

Footer Text 2016-10-24 14

 Constraint Functions
• Almost all engineering design problems have
constraints
• Some constraints are simply boundaries of design
variables, which are easy to handle
o For example, we might know that the wing span should be between 7
and 10 meters
• Some constraints are implicit
o For example, maximum speed > 200 kts
o This is a quantity that is calculated by some equations or analysis
programs, not simple limits on the design variables
• Some are equality constraints
o For example, Lift=Weight
• Some are inequality constraints
o For example Weight < 600 kg

Footer Text 2016-10-24 15

 Constraint Functions (2)
• Aircraft Design Example
Constraints
Endurance > 8 hours
Rate of climb > 4.5 m/s
Takeoff distance < 300 m
Landing distance < 350 m
Static margin > 10%
Range > 1500 km
Weight < 600 kg

Footer Text 2016-10-24 16

 System Equations
• Equations that calculate the performance of the
system that is being designed
• System equations
o read in a set of design variables as input
o Calculate the objective function value
o Calculate the values of each equality constraint
o Calculate the values of each inequality constraint

• Examples
o Finite element method (can calculate weights, stiffness, and stress)
o CFD aerodynamics (calculates drag, pressure distribution, etc)

Footer Text 2016-10-24 17

17
 System Equations (2)
• VLM Aerodynamics • FEM Structural Solver
Solver

Footer Text 2016-10-24 18

 Problem Statement
• Minimize
• By varying
• Subject to ,
•

• objective function
• equality constraint functions
• inequality constraint functions

Footer Text 2016-10-24 19

 Simple Examples
• Unconstrained
optimization

• The optimizer begins

at an initial point
• Steps are taken that
reduce the value of
until the a minimum
value is reached

Footer Text 2016-10-24 20

 Simple Examples (2)
• Constrained
optimization

• Such that

• Steps are taken that

reduce the value of
until the a minimum
value is reached that
doesn’t violate the
constraint

Footer Text 2016-10-24 21

 Engineering Examples

• 18 Bar Truss
o Load applied along each top node

• What is the best design?

o How thick should each beam be?
o How long should each beam be? (where to we put the connecters)

Footer Text 2016-10-24 22

 Engineering Examples (2)

• We need to define the structure

in terms of a set of design
variables
• We can alter these variables to
try and search for better designs

Footer Text 2016-10-24 23

 Engineering Examples (3)
• We have represented the structure as a set of 12
design variables
• Now we need to set up our optimization problem
• What makes one truss design better than
another?
• We need to select an objective function
o We decide that a lighter design is better
o Minimize mass=sum(each element length x each element thickness x density)

• What are the design constraints?

o The truss shouldn’t break when certain loads are applied to it
o Stress < yield stress in each element
o Stress < buckling stress in each element

Footer Text 2016-10-24 24

Engineering Examples (4)
•Now we can select or develop an optimization
program and run it to find the best design

•The optimization algorithm gives us the x values

corresponding to the position of each node and the
thickness of each beam
Footer Text 2016-10-24 25
 Methods
• How can we determine what the optimum
design is?
• There are many computational methods available
for doing so
o Random search
o Simplex method
o Bisection method
o Newton’s method
o Genetic Algorithms
o Sequential Quadratic Programming
o Many others

Footer Text 2016-10-24 26

 Classification of
Optimizers

Footer Text 2016-10-24 27

 Additional Examples
• Topology
Optimization
• Each square is a
design variable
that is either 1 or
0
• Can result in
highly optimized
structural designs

Footer Text 2016-10-24 28

 Additional Examples (2)
• Regional Jet Conceptual Design
• Objective Function: minimize fuel consumption
• Constraints:

Footer Text 2016-10-24 29

 Additional Examples (3)
• System Equations
o VLM Aerodynamics solver
o Performance Equations
o Weight Equations

o Aircraft configuration

Footer Text 2016-10-24 30

 Additional Examples (3)
• Design Variables

Footer Text 2016-10-24 31

 Additional Examples (4)
• Constraints

Footer Text 2016-10-24 32

Additional Examples (5)

33
Aircraft Design Case Study
Deterministic P=0.80

P=0.99 Boeing 737-800

34