33
optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction , Van Nostrand Reinhold, New York

Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

Embed Size (px)

Citation preview

Page 1: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

optimization

1/33

Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction,

Van Nostrand Reinhold, New York

Page 2: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

optimization techniques

2/33

● which technique depends on● nature of problem● nature of information required

● historically● mathematical problem – single optimal value● not interested in suboptimal values

● design● both decisions and suboptimal solutions important● suboptimal solutions options● may be more acceptable in terms of unstated objectives

Page 3: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

optimization techniques

3/33

● nature of problem / information● design variables discrete &/or discontinuous &/or non-contiguous● steel beams in discrete sizes● no. of lifts discontinuous,● materials non-contiguous● nonlinear relationships heat loss

Page 4: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

types of techniques

4/33

● calculus● continuous and differentiable

● linear programming (LP)● well developed method● linear relationship among variables

● nonlinear programming (NLP)● nonlinear relationships

● dynamic programming (DP)● discrete, nonlinear, handles constraints

● evolutionary computation● population

Page 5: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

general strategies

5/33

● exhaustive enumeration● all possible solutions

● implicit enumeration● e.g. branch & bound, DP

● hill-climbing● moving from existing solution to an improved solution

Page 6: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

linear programming● best developed technique

● most frequently used● guarantees optimum solution

● 3 conditions● variables must be continuous, >= 0● O.F. must be linear, OF=20x1 +12x2

● constraints must be linear, 4x1 + 3x2 > 18

6/33

Page 7: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

linear programming● convex spaces

● feasible solution space

● O.F. moves away from origin● optimum solution at vertex

7/33x1

x2

Page 8: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

dynamic programming● design problems

● not continuous or linear

● definition (Richard Bellman)● an optimal set of decisions has the property that whatever the first decision is, the remaining decisions must be optimal with respect to the outcome which results from the first decision

8/33

if you don’t’ do the best with what you have,you will never do the best with what you

should have had

Page 9: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

dynamic programming● stage-state formulation

● implicit enumeration of all paths

● guarantees global optimum● non-serial DP doesn’t guarantee● but pretty good

9/33

Stage 1 2 3

es

4 5State

Page 10: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation● hill-climbing

● one solution at a time● in direction of steepest slope

● local optima● variables &/or constants● equations – y = mx + c

● EC in parallel● populations● survival of the fittest – probability● random generation 10/33

LO

GO

Page 11: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

11/33

Page 12: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

12/33

Page 13: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

13/33

Page 14: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

14/33

Page 15: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

15/33

Page 16: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

16/33

Page 17: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

17/33

Page 18: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

18/33

Page 19: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

19/33

Page 20: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

20/33

Page 21: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

21/33

Page 22: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

22/33

Page 23: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

23/33

Page 24: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

24/33

Page 25: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

25/33

Page 26: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

26/33

Page 27: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

27/33

Page 28: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

28/33

Page 29: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

29/33

Page 30: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

30/33

Page 31: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

evolutionary computation

31/33

Page 32: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

artificial life

32/33

Page 33: Optimization 1/33 Radford, A D and Gero J S (1988). Design by Optimization in Architecture, Building, and Construction, Van Nostrand Reinhold, New York

multicriteria optimization

33/33

● single objective● e.g. cost

● several conflicting objectives● max light – min heat● best looking, min cost car

● Pareto solutions● best compromise

● tradeoffs

c1

c2