1 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Multidisciplinary System
Design Optimization (MSDO)
Multiobjective Optimization (II)
Lecture 17
April 5, 2004
by
Prof. Olivier de Weck
2 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
MOO 2 Lecture Outline
Lecture 2 (today)
? Alternatives to Weighted Sum (WS) Approach
Multiobjective Heuristic Programming
Utility Function Optimization
Physical Programming (Prof. Messac)
Application to Space System Optimization
Lab Preview (Friday 4-9-2003 – Section 1)
3 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Weighted Sum (WS) Approach
1
z
i
MOi
i
i
w
JJ
sf
=
=
|
utopia
Max(J
1
)
Min(J
2
)
miss this
concave region
Pareto
front
convert back to SOP
LP in J-space
easy to implement
scaling important !
weighting determines
which point along PF is
found
misses concave PF
w
2
>w
1
w
1
>w
2
J-hyperplane
J*
i
J*
i+1
4 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Weighted Square Sum Approach
22
11 2 2
J wJ wJ=+
Obj. Fun. Line
J1
J2
Ref: Messac
5 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Compromise Programming (CP)
Obj. Fun. Line
11 2 2
nn
JwJ wJ=+
55
This allows
“access” to the
non-convex part of the
Pareto front
6 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Multiobjective Heuristics
Pareto ranking scheme
Allows ranking of population
without assigning preferences
or weights to individual
objectives
Successive ranking and
removal scheme
Deciding on fitness of
dominated solutions is more
difficult.
Pareto ranking for
a minimization problem.
Pareto Fitness - Ranking
Recall: Multiobjective GA
This number comes
from the D-matrix
7 ? Massachusetts Institute of Technology - Prof. de Weck and Prof. Willcox
Engineering Systems Division and Dept. of Aeronautics and Astronautics
Example Multiobjective GA
()
2
11
1
1
,..., 1 exp
n
ni
i
fx x x
n
=
a o
§