multi-objective optimisation of building designs · building design optimisation different design...
TRANSCRIPT
![Page 1: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/1.jpg)
Multi-objective optimisation of
building designs
Sandy Brownlee
Senior Research Assistant
![Page 2: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/2.jpg)
Outline
Evolutionary multi-objective optimisation
Building design optimisation
Example problems
Decision making
Algorithm improvements
Summary, questions etc.
2
![Page 3: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/3.jpg)
Recap and definitions:
evolutionary multi-objective optimisation
3
![Page 4: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/4.jpg)
EMO
Single objective GA
Moving to multi-objective
Constraints
NSGA-II
4
![Page 5: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/5.jpg)
Single objective GA
1. Generate random population
2. Assign a fitness to members of the
population
3. Choose the best ones and recombine them
to produce offspring
4. Mutate the offspring
5. Repeat 1-4 until we’re done
5
![Page 6: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/6.jpg)
SO GA Example
6
0 1 1 1 0 1
1 0 1 1 0 0
0 0 1 1 0 1
0 1 1 0 1 0
0 0 0 0 1 1
1 0 0 0 0 0
4
3
3
3
2
1
0 1 1 1
0 1 1 0 1 1
0 0
1 0 1 1 1 0
0 1 1 0 0 1
1 0 1 0 0 0
0 0 0 0 1 0
4
3
2
1
0 1 1 1 0 1
1 0 1 1 0 0
0 1 1 1
1 1 1 0 1 1
0 0 0 1 1 1 0 0
1 0 1 1 1 1
3
5
![Page 7: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/7.jpg)
Multi-objective
Multi-objective optimisation…
In reality, most problems are multi-objective,
often with conflicts – e.g. cost vs performance
How do we define fitness for more than one
objective?
Could just add them together, but how do we
weight them?
Better to find the trade-off an make an
informed decision
7
![Page 8: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/8.jpg)
Definition: Dominance
This time there are two
“fitnesses” (objective
values) for each solution
One solution dominates
another if it is “better” in
both objectives
Can plot the objectives of
population in 2D >>>
Set of non-dominated
solutions is the Pareto front
8
0 1 1 1 0 1
1 0 1 1 0 0
0 0 1 1 0 1
0 1 1 0 1 0
0 0 0 0 1 1
1 0 0 0 0 0
1
2
4
3
2
4
3
4
6
7
2
1
![Page 9: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/9.jpg)
Constraints
Some solutions “good” or “bad”
Building with no ventilation is cheap and low-energy,
but not very comfortable!
E.g.: max hours over 28oC, min lighting, compliance
with law
9
How to handle?
Whole research area
Can be included in the
concept of dominance
Constraints can be hard
to satisfy
![Page 10: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/10.jpg)
NSGA-II
A popular GA for MO
optimisation
Selection biases search
towards:
Feasible solutions
Non-dominated
solutions (low rank)
Non-crowded solutions
Basis for the
experiments here
10
Non-dominated sorting / ranking
![Page 11: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/11.jpg)
Building design optimisation
11
![Page 12: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/12.jpg)
Building Designs
Broad concepts
3 example building problems
Variables, objectives, constraints
Variable sensitivity – decision making
12
![Page 13: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/13.jpg)
Building designs
Why optimise?
Climate change!
Over 50% of UK
carbon emissions
are related to
energy consumed
buildings
Cost, comfort
No mass production
13
![Page 14: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/14.jpg)
Building design optimisation
Buildings are complex!
Many variables
Dimensions, materials, layout, systems (heat /
light etc), control configuration
Many objectives / constraints
Energy use, Construction cost, Comfort
Compliance
Highly suitable for EA
14
![Page 15: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/15.jpg)
Building design optimisation
Different design stages
Conceptual – overall shape
Detailed – materials, equipment
Change at concept stage can be big
But also dependent on getting things
right later
Project blurring lines between stages;
optimise across stages (e.g.
orientation, envelope, controls) but
more to be done
15
![Page 16: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/16.jpg)
Building design optimisation
16
Evolutionary
algorithm
Simulation
(energy, cost
modelling, comfort
prediction…)
fitness
building
Optimal building(s)
![Page 17: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/17.jpg)
Example 1: Cellular Windows
Optimise glazing for an atrium in a building
Variables:
Switch on glazing and shades in 120 cells
240 bits encoding
Objectives - minimise:
Construction cost
Energy for lighting, heating and cooling
Constraints:
number or aspect ratio of “windows” (mutually
neighbouring cells)
17
![Page 18: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/18.jpg)
Example 1: Cellular Windows
18
![Page 19: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/19.jpg)
Single Objective
With “number” constraint,
glazing falls in central area
Where the light sensors are
located
With aspect ratio constraint,
glazing tends to be spread
out, still usually 3 windows
Better coverage of facade
19
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 2 3 2 2 2 20 0 0 0 0 0 0 0 0 1 0 0 0 0 0
2 2 2 2 2 2 2 2 2 3 2 2 2 3 20 0 0 0 0 0 0 0 1 0 0 0 0 0 0
2 2 2 2 2 2 2 2 3 2 2 2 2 2 20 0 0 0 0 0 0 0 1 0 0 0 0 0 0
2 2 2 2 2 2 2 2 3 2 2 2 2 2 20 0 1 0 0 0 0 1 1 1 0 0 0 0 0
2 2 3 3 2 2 2 3 3 3 2 2 2 2 20 0 0 0 0 0 1 0 0 0 0 0 0 0 0
2 2 2 2 2 2 3 2 3 2 2 2 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 2 2 2 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
2 3 2 2 2 2 2 2 2 2 2 3 2 2 20 0 1 1 0 0 0 1 0 0 1 0 1 0 0
2 2 3 3 2 2 2 3 2 2 3 2 3 2 20 0 0 0 0 0 0 0 0 0 0 0 0 1 0
2 2 2 2 3 2 2 3 2 2 2 2 2 3 20 0 0 0 0 0 0 1 0 0 0 0 0 0 0
2 2 2 3 2 2 2 3 2 2 2 2 2 3 20 0 0 0 0 0 0 1 0 0 0 0 0 0 0
2 2 3 2 2 2 2 3 2 2 2 2 2 3 20 1 0 0 0 0 0 0 1 0 0 0 1 0 0
2 3 2 2 2 2 2 2 3 3 2 2 3 2 20 0 1 0 0 0 0 0 0 0 0 0 0 0 0
2 2 3 2 2 2 2 2 2 2 2 2 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
![Page 20: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/20.jpg)
Multi-objective
Trade-off for energy vs cost
Simple linear cost per glazed cells & shades
Larger window still tends to centre
Hard to meet constraints
Seeding the population helps
20
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 3 2 2 2 2 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 1 0
2 2 3 2 2 2 2 2 3 2 2 2 2 3 20 1 0 0 0 0 0 0 1 0 0 0 0 0 0
2 3 2 2 2 2 2 2 3 2 2 2 3 2 31 0 0 0 0 1 1 0 0 0 0 0 0 0 0
3 2 2 2 2 3 3 2 3 2 2 3 2 2 20 0 0 0 0 0 0 0 0 0 0 1 0 0 0
2 3 2 2 2 2 2 3 2 2 2 3 2 2 20 1 0 0 0 0 0 0 1 0 0 0 0 0 0
2 3 2 2 2 2 2 2 3 2 2 3 2 2 20 0 0 0 0 0 0 0 0 0 0 1 0 0 0
2 2 2 2 2 2 2 2 2 3 2 3 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 3 2 2 2 2 2 2
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
2 2 2 3 2 3 3 2 2 2 2 2 2 3 20 1 0 0 0 0 0 0 0 0 0 0 0 0 0
2 3 3 2 2 3 2 2 2 2 2 2 2 2 30 1 1 0 0 0 0 1 1 0 0 1 1 1 1
2 3 3 2 2 3 2 3 3 3 2 3 3 3 30 0 0 0 0 1 1 1 1 0 0 0 0 0 0
3 2 2 2 2 3 3 3 3 2 2 3 2 2 20 1 1 1 0 0 0 0 0 0 0 1 0 0 0
2 3 3 3 2 2 2 3 2 3 2 3 2 2 20 0 0 0 0 0 0 0 1 0 0 0 0 0 0
2 3 2 2 2 2 2 2 3 2 2 3 2 2 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 3 2 2 2 2 20 0 0 0 0 0 0 0 1 0 0 0 0 0 0
2 2 2 2 2 2 2 2 3 2 2 2 2 2 2
![Page 21: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/21.jpg)
Example 2: Office block
Small 5 zone office; a single floor of a larger
building
Variables:
Orientation, glazing area, type, wall/floor
types, HVAC set points and times
Objectives:
Energy use, cap cost
Constraints:
Thermal comfort, air quality (CO2 levels)
21
![Page 22: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/22.jpg)
Results
Example building with glazing altered
22
![Page 23: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/23.jpg)
Results
23
![Page 24: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/24.jpg)
Example 3 : Risk of mould growth
Variables: heating, ventilation, aircon system
setup and operation
Objectives: Energy, Mould Risk (related to
long, warm, damp periods)
Hospital ward,
Kuala Lumpur
24
![Page 25: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/25.jpg)
Variable sensitivity – decision making
Decision making
Why is a given solution optimal?
How optimal is a given solution?
What design decisions actually impact on the
objectives?
Observe which variables impact the most
Can we ignore some of them to simplify the
search?
What do we learn about the underlying problem?
Can this aid decision making?
25
![Page 26: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/26.jpg)
Variable sensitivity
26
![Page 27: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/27.jpg)
Variable Sensitivity
27
A – HVAC heating set point
B – HVAC cooling set point
C – t’hold temp for nat. vent.
D – glazed area, north upper
E – glazed area, south upper
F – mechanical ventilation rate
G – external wall material
H – ceiling and floor material
I – shading overhang present
![Page 28: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/28.jpg)
Algorithm Improvements
28
![Page 29: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/29.jpg)
Improvements
Two problems!
Constraint handling
Long runtimes
Fitness inheritance
Surrogate model
Experiments / results
29
![Page 30: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/30.jpg)
Problem 1: Constraints
Constraints can be hard to satisfy, and can
limit the extent of the trade-off found
Relaxation – ignore constraints to start with
Normalise / weighting
Constraints weighted equally, or with a bias to
meeting harder constraints first
Include infeasible solutions in population
Allow some infeasible solutions in population
Either keep “least infeasible” or “fittest”
infeasibles
30
![Page 31: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/31.jpg)
Problem 2: Long run times
Typical EA needs thousands of simulations
Building energy simulation takes 1-2 minutes
for example problems
Larger building or more detailed sim takes
longer; also larger search space
31
![Page 32: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/32.jpg)
Long run times: Possible solutions
Simplify the problem
Reduce model complexity
Reduce weather data extent
Parallel execution / caching solutions
“Guess” some of the solutions
Fitness inheritance
Surrogate
32
![Page 33: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/33.jpg)
Fitness Inheritance
Based on idea that “similar” solutions have similar
fitness
After crossover, offspring’s fitness assumed to be
between that of parents
Only inherit ~half the time
Can weight towards a parent
33
![Page 34: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/34.jpg)
Fitness Inheritance
34
0 1 1 1
0 1 1 0 1 1
0 0
Individual Energy Use
kWh
Cost £ Overheating
hours
(max 30)
Max CO2
conc.
(max 1500)
Parent A 54200 370000 40 430
Offspring 57200 365000 25 330
Parent B 60200 360000 10 230
![Page 35: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/35.jpg)
Surrogate Model
Train a model of the fitness function, e.g. ANN
Use the model in place of the FF
35
fitness
building
Optimal building(s)
predicted
fitness
training building
Simulation
(energy, cost modelling,
comfort prediction…)
Evolutionary
algorithm Surrogate
![Page 36: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/36.jpg)
Surrogate Model
1. Generate random population
2. Assign a fitness to members of the
population
3. Choose the best ones and recombine them
to produce offspring
4. Mutate the offspring
5. Repeat 1-4 until we’re done
36
Plain EA
![Page 37: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/37.jpg)
Surrogate Model
1. Generate random population
2. Assign a fitness to members of the
population
3. Choose the best ones and recombine them
to produce too many offspring
4. Mutate the offspring
5. Use surrogate to filter out promising
offspring
6. Repeat 1-5 until we’re done
37
EA with
surrogate
![Page 38: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/38.jpg)
Mining a surrogate model
One part of my research
Examine the surrogate model to gain insight
into the problem
38
0.017 0.017 0.016 0.016 0.015 0.015 0.015 0.015 0.016 0.015 0.015 0.015 0.016 0.016 0.016
0.016 0.016 0.016 0.016 0.016 0.016 0.016 0.016 0.015 0.015 0.016 0.015 0.016 0.016 0.016
0.017 0.017 0.016 0.016 0.015 0.016 0.016 0.016 0.016 0.016 0.015 0.016 0.016 0.016 0.016
0.017 0.016 0.017 0.017 0.016 0.016 0.015 0.015 0.016 0.015 0.016 0.017 0.016 0.017 0.017
0.017 0.017 0.017 0.016 0.016 0.016 0.016 0.016 0.015 0.016 0.016 0.016 0.017 0.017 0.017
0.017 0.018 0.018 0.017 0.018 0.017 0.017 0.016 0.016 0.017 0.016 0.017 0.016 0.017 0.016
0.018 0.017 0.018 0.018 0.018 0.017 0.017 0.017 0.018 0.017 0.017 0.017 0.016 0.017 0.017
0.017 0.018 0.017 0.018 0.018 0.018 0.018 0.019 0.018 0.019 0.017 0.018 0.018 0.017 0.018
![Page 39: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/39.jpg)
In practice…
Both fitness inheritance and surrogate offered
a speed up of around 20-30% in our work
Better constraint handling also helped
More important was correctly framing the
problem
Seeding
Good choice of variables
Parallel execution also made a big difference
39
![Page 40: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/40.jpg)
Summary
Optimisation (particularly with EA) a growing
area in building design community
It really does work in practice
Example problems
Changes what the decision maker does
More information about the problem
Room for improvement
Move to concept stage (form / shape)
Simulation time a big issue
40
![Page 41: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/41.jpg)
Questions
41
![Page 42: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/42.jpg)
A few items for reference if necessary…
42
![Page 43: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/43.jpg)
Spread
We want set to span
the entire length of the
Pareto front
Spread measures how
evenly the solutions
are spaced out
Zero is best
43
![Page 44: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/44.jpg)
Surrogate Model
Limited work done with mixture of continuous
and discrete variables, and with constraints
Approach to constraints same as for FI
i.e. predict value then do cut-off
Using a radial basis function network (RBFN)
Initially tried a single network
Had to retrain whole network if part of it poor
Now one network per objective or constraint
44
![Page 45: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/45.jpg)
RBFN
Feed-forward network
Input layer: problem vars
Hidden layer:
radial basis functions
output similarity to centre
45
Output layer:
linear weighted sum per objective / constraint
Distances
Euclidian (cont), Manhattan (int), Hamming (bits)
![Page 46: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/46.jpg)
Surrogate Model
1. Random init of population
2. Selection of parents
3. Generate too many offspring from parents
3a. Use surrogate to filter out promising offspring
4. Evaluate filtered offspring
5. Combine offspring + parents into Q
6. Non-dom sort Q
7. Replace population with top half of Q
8. If termination criteria not met, back to 2
46
NSGA II with
surrogate
![Page 47: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/47.jpg)
Comparing performance
Hard to compare fronts
What are we measuring?
Closeness to “true” Pareto
front
Spread along the front
Extents of front
Several measures;
hypervolume used here
47
![Page 48: Multi-objective optimisation of building designs · Building design optimisation Different design stages Conceptual – overall shape Detailed – materials, equipment ... Optimal](https://reader034.vdocument.in/reader034/viewer/2022052018/6031b83f5cf3ff15ba382a98/html5/thumbnails/48.jpg)
Hypervolume
The area / volume between
the PF and a nadir point (the
global minimum)
General measure; includes
extent, spread and optimality
of PF
Prefers convex regions of PF
Expensive if many objectives
48