The Leviathan [Architecture Masters Thesis] by Yaseen Bhatti

The Leviathan A Computationally derived conceptual Pavilion By Yaseen Bhatti, Alex Kendall & Ben Miller

Thesis Statement Problem The contemporary urban environment is characterised by rapid change. Unless resilient systems are developed, factors such as the Climate Emergency will bring widespread disruption and unrest.

Theory Building on the ideas of Researching For Design we are looking at Resilience theory and Generative Design Theory derive answers which combine understanding of mathematics, engineering and the natural environment to create solutions.

Solution Computational design tools will allow us to generate architectural interventions for tackling issues centred around rapid change in our urban environment.

Contents 01 | Context

Framing the project primary goals and narrative

02 | Generative Systems

Introducing generative design and a systems design logic

03 | What is Tested

p11

Researching Resilience Theory for Generative Exploration

04 | The CPU Brief

p15

Applying resilience theory to pavilion design

05 | Site Analysis

p18

Multi-scale assessment of the site to identify local issues

06 | Design Strategies

p22

Addressing climate change qualitatively and quantitatively

07 | Generative Design Process

p32

Formulating a systems strategy to satisfy design goals generatively

08 | Generating Design Iterations

p43

Producing initial canopy iterations of varying performance

09 | Iteration Optimisation

p50

Refining Canopy iterations for highest performers

10 | Landscaping The Leviathan

p57

Applying field theory for circulation routes

11 | Water Runoff Study Methods

p61

Methods for testing flood mitigation strategies

12 | Water Runoff Rate Experiment Methods for testing flood mitigation strategies

13 | Detailing Finalising pavilion structure

p80

p71

Contents 01 | Context

Framing the project primary goals and narrative

02 | Generative Systems

Introducing generative design and a systems design logic

03 | What is Tested

p11

Researching Resilience Theory for Generative Exploration

04 | The CPU Brief

p15

Applying resilience theory to pavilion design

05 | Site Analysis

p18

Multi-scale assessment of the site to identify local issues

06 | Design Strategies

p22

Addressing climate change qualitatively and quantitatively

07 | Generative Design Process

p32

Formulating a systems strategy to satisfy design goals generatively

08 | Generating Design Iterations

p43

Producing initial canopy iterations of varying performance

09 | Iteration Optimisation

p50

Refining Canopy iterations for highest performers

10 | Landscaping The Leviathan

p57

Applying field theory for circulation routes

11 | Water Runoff Study Methods

p61

Methods for testing flood mitigation strategies

12 | Water Runoff Rate Experiment Methods for testing flood mitigation strategies

13 | Detailing Finalising pavilion structure

p80

p71

01 | Context Framing The Primary Project Drivers and Narrative

Problem

Traditional Pavilion Test-bed Taught Knowledge

Sylvia Lavin

Design System #2 Design System #3

K n ow l e d g e

Design System #1

K n ow l e d g e

Project Area of Research

Reference Materials

A pavilion should be â&#x20AC;&#x153;not predicated on the difference between art and the world, but rather facing their profound imbricationâ&#x20AC;?

Untested Architectural Concept

Pavilion Testbed

Solution

Research For Design Focusing less on the Solution as an isolated element instead as an accumulation and representation of the method used to reach this particular outcome. Therefore there is a shift in focus from the outcome itself and more looking and representing the process as a whole.

Pavilion Design Approach

Proven idea

Failed Test

The Contemporary Pavilion Traditionally the pavilion was designed to be used as a testbed for new ideas and experimentation, but the idea of a pavilion has become diluted. Currently, the pavilion is more about art than architecture. Pavilions are no longer being used to test ideas and feedback into the profession, instead area repeating what has already been shown.

Problem

Traditional Pavilion Test-bed Taught Knowledge

Sylvia Lavin

Design System #2 Design System #3

K n ow l e d g e

Design System #1

K n ow l e d g e

Project Area of Research

Reference Materials

A pavilion should be â&#x20AC;&#x153;not predicated on the difference between art and the world, but rather facing their profound imbricationâ&#x20AC;?

Untested Architectural Concept

Pavilion Testbed

Solution

Pavilion Design Approach

Proven idea

Failed Test

3.00 2.75 2.50 2.25 2.00 1.50 1.25 1.00

365 Majority

0.75 0.50 -0.25 -0.00

UK Climate Debates

-0.25 -0.50 1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

The Climate Crisis Global Temperature Rise The Climate Crisis poses hardly imaginable existential repercussions for our planet and species if sufficient measures are not taken to resolve our exploitative and destructive relationship with our natural environment. Meanwhile Far-Right Nationalism is being elected across the world from Brazil to India, from the UK to the US. These leaders are defying any counter climate change action and protecting the big business that frequently finances them and who are also causing the vast majority of the damage.

Designing Resilient Architecture

Engaging People with the Problem

Rising global temperatures are already causing extreme weather events and disruption throughout the globe including:

Developing Resilient architecture has no impact if there is no will to use the findings.

Rising Seas Rising Temperatures Increased severe weather Widespread Disease Crop damage and food shortages

ÎŠ

Measures must be taken to cope with this, as well as to control our relationship with resources to create more resilient systems. Resilience Theory may contain answers to these issues, and this is what we want to explore.

Potential

1950

Îą

Recent events, such as the UK General Election, have demonstrated that there is a lack of prioritisation from the public on this issue. With others taking the forefront, the stance elected is that of silence. However Climate Action, such as protests and strikes have made substantial headway in bringing this to peoples attention. A pavilion is a public space and an opportunity to explore topical issues. It is also an emotive experiential environment.

ÎŠ Connectedness

Through this Climate Action can be taken to engage visitors with the issues that drive our design.

3.00 2.75 2.50 2.25 2.00 1.50 1.25 1.00

365 Majority

0.75 0.50 -0.25 -0.00

UK Climate Debates

-0.25 -0.50 1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

Designing Resilient Architecture

Engaging People with the Problem

Rising global temperatures are already causing extreme weather events and disruption throughout the globe including:

Developing Resilient architecture has no impact if there is no will to use the findings.

Rising Seas Rising Temperatures Increased severe weather Widespread Disease Crop damage and food shortages

ÎŠ

Potential

1950

Îą

ÎŠ Connectedness

Through this Climate Action can be taken to engage visitors with the issues that drive our design.

02 | Generative Systems Introducing Generative Design & a Systems Design Logic

NATURAL SELECTION Natural processes develop unique organisms which are each adapted to the functional requirements of the environment they occupy. This is attained through a generational process of Selection, Breeding and Mutation. Genotype Phenotype Environment Genotype Phenotype Genotype Phenotype Environment Environment

Generative Design Learning From Nature The natural environment offers a vast body of information on how problems arising have been resolved. We can study how design occurs in nature, and the process that natural evolution takes in arriving at its final design solutions.

X Y Z

Exploitation // Exploration One of the most successful elements of natural design is its unique ability to negotiate the trade-off between exploitation and exploration.

MIN

X Y Z

MIN

MAX MAXMIN X Y Z

MAX

In Nature

GENOTYPE

PHENOTYPE

ENVIRONMENT

The encoded information or DNA of an organism.

The physical expression of an organism.

The context the forms the organism.

Unknown systems must be explored to gain understanding. However when a system is partially known, deciphering whether to optimise knowledge or seek more can be hard to determine.

Exploitation Selection

Breeding

Mutation

Using existing information to find the optimum.

Exploration Selection

Breeding

Random development knowledge.

In Design

Mutation

new

Genotype Phenotype Environment Genotype Phenotype Environment GENOTYPE PHENOTYPE Environment ENVIRONMENT Genotype Phenotype MIN MAX MIN MAX X MIN Y MAX X X Y Z Y Z Z

The parameters of a design.

The physical representation of these parameters.

The context of the design.

X Y Z

Exploitation // Exploration One of the most successful elements of natural design is its unique ability to negotiate the trade-off between exploitation and exploration.

MIN

X Y Z

MIN

MAX MAXMIN X Y Z

MAX

In Nature

GENOTYPE

PHENOTYPE

ENVIRONMENT

The encoded information or DNA of an organism.

The physical expression of an organism.

The context the forms the organism.

Unknown systems must be explored to gain understanding. However when a system is partially known, deciphering whether to optimise knowledge or seek more can be hard to determine.

Exploitation Selection

Breeding

Mutation

Using existing information to find the optimum.

Exploration Selection

Breeding

Random development knowledge.

In Design

Mutation

new

Genotype Phenotype Environment Genotype Phenotype Environment GENOTYPE PHENOTYPE Environment ENVIRONMENT Genotype Phenotype MIN MAX MIN MAX X MIN Y MAX X X Y Z Y Z Z

The parameters of a design.

The physical representation of these parameters.

The context of the design.

Neural Networks Neural Networks are compositions of a huge number of simple calculations which have the ability to create highly complex non-linear functions. They are structures of input parameters. These have been successful at modelling many real-world systems.

Rule of thumb

Unexpected yet high performing

Hidden Layers

Input Layer

Expressionistic

Output Layer

Population No

Generate random options within the Design Space

Fitness defines the favourability of a system in achieving its defined purpose. The principals of the Selection Breeding and Mutation process can be applied to systems in order to find favourable solutions.

Novelty Vs Performance As designers we are striving for novel design solutions. However, we are also bound by general rules of thumb in order to achieve a guaranteed performance. Our aspiration should be to achieve a hybrid solution.

Choose options with favourable qualities.

Breed favourable pairs.

Generate mutations based on favourable qualities.

Features such as the process of breeding are all also parametric within the system. A designer must decide on the level of inclusion versus elitism in order to avoid missing out on novel outputs or pursue extreme optimums within the Design Space.

Breeding Chance

Fitness

Mutation

Breeding Chance

“Fitness” Complex systems tend to have a high number of fitness criteria. Often parameters therefore have non-linear or complex relationships with Fitness values. Sometimes these parameters collaborate, conflict or do not interact at all. Ultimately the overall best fit is decided by the priorities of the designer.

Parameter X

Compramise

Fitness

Crossover

Breeding Approaches

Fitness

Selection

Output

Fitness

Processes for finding favourable solutions

Yes

Fitness

Pursuing “Fitness”

Fit

Parameter Y

Rule of thumb

Unexpected yet high performing

Hidden Layers

Input Layer

Expressionistic

Output Layer

Population No

Generate random options within the Design Space

Choose options with favourable qualities.

Breed favourable pairs.

Generate mutations based on favourable qualities.

Breeding Chance

Fitness

Mutation

Breeding Chance

Parameter X

Compramise

Fitness

Crossover

Breeding Approaches

Fitness

Selection

Output

Fitness

Processes for finding favourable solutions

Yes

Fitness

Pursuing “Fitness”

Fit

Parameter Y

Design Optimisation

Input Parameters

Methods of Data Harvesting

Objective Functions

Available data sets are categorised into 3 types

identified

and

Continuous values MIN MAX MAXIMISE

TARGET

Goals are determined for the system, as to what ideal value is wanted for each data set.

While some simple optimization problems have direct solutions, most require the problem to be solved incrementally using ‘optimization algorithms’.

Discrete values B C

Constraint functions

Extents of the data are determined to proscribe the limits to a feasible design space.

Genetic Algorithm START

Roulette selection of parents

Generate inital population

Crossover to produce children

Calculate fitness of individuals

Satisfy stop criterion

END

Deterministic Methods Input Layer

Hidden Layer

This process achieves a solution through the direct application of a series of defined steps.

Output Layer

Mutation of children

Direct Analysis

Gradient Descent

Exhaustive search

Calculate fitness of children

Heuristics (Rule of thumb)

Application: Finite Element Analysis A vertices based analysis of force distribution through a solid. These can be used to assess the functionality of structural systems and optimise material distribution.

Calculate generation by “Elitism”

Direct Analysis

Genetic algorithms use a top-down external system that tests a given model to try to learn how it works from the outside in, to iterate and ‘evolve’ solutions.

Stochastic Methods Gradient Descent

Exhaustive search

Heuristics (Rule of thumb)

Monte Carlo (Random testing)

Stochastic Gradient Descent

Metaheuristics

Monte Carlo (Random testing)

Stochastic Gradient Descent

Metaheuristics

This process works by introduce some level of randomness while ‘searching’ for a solution.

Application: Fluid Dynamics An agent based system that simulates dynamics of nodes affected by geometry fields and other factors to see how they interact. These can be used for crowd flow or water flow simulations.

Permutation Sequence MINIMISE >=

Design Optimisation

Input Parameters

Methods of Data Harvesting

Objective Functions

Available data sets are categorised into 3 types

identified

and

Continuous values MIN MAX MAXIMISE

TARGET

Goals are determined for the system, as to what ideal value is wanted for each data set.

While some simple optimization problems have direct solutions, most require the problem to be solved incrementally using ‘optimization algorithms’.

Discrete values B C

Constraint functions

Extents of the data are determined to proscribe the limits to a feasible design space.

Genetic Algorithm START

Roulette selection of parents

Generate inital population

Crossover to produce children

Calculate fitness of individuals

Satisfy stop criterion

END

Deterministic Methods Input Layer

Hidden Layer

This process achieves a solution through the direct application of a series of defined steps.

Output Layer

Mutation of children

Direct Analysis

Gradient Descent

Exhaustive search

Calculate fitness of children

Heuristics (Rule of thumb)

Calculate generation by “Elitism”

Direct Analysis

Genetic algorithms use a top-down external system that tests a given model to try to learn how it works from the outside in, to iterate and ‘evolve’ solutions.

Stochastic Methods Gradient Descent

Exhaustive search

Heuristics (Rule of thumb)

Monte Carlo (Random testing)

Stochastic Gradient Descent

Metaheuristics

Monte Carlo (Random testing)

Stochastic Gradient Descent

Metaheuristics

This process works by introduce some level of randomness while ‘searching’ for a solution.

Permutation Sequence MINIMISE >=

03 | What is Tested Researching Resilience Theory for Generative Exploration

CONTEXT

ACUTE SHOCK

Continuous stress CONTINUOUS POSITIVE STRESS

SHOCK TO THE SYSTEM

NEW STATE OF EQUILIBRIUM

PAVILION

RESILIENCE CONTINUOUS NEGATIVE STRESS

OLD STATE OF EQUILIBRIUM

Systems PROCESS

Adaptation Scales

Resilience Theory Resilience Theory explores how systems limit, cope with and facilitate change. Biology and ecologies can provide answers on how to achieve resilient systems. ‘In nature, the biota and the physical environment interact such that not only does the environment shape the biota but the biota transform the environment.’ These systems adapt to change at all scales in different, non-linear ways. They are characterised by cyclical processes of Exploitation, Conservation, Release and Reorganisation.

Ecologies Potential

Systems of change

Ω Connectedness

CONTINUOUS VARIANT STRESS

CONTEXT

ACUTE SHOCK

Continuous stress CONTINUOUS POSITIVE STRESS

SHOCK TO THE SYSTEM

NEW STATE OF EQUILIBRIUM

PAVILION

RESILIENCE CONTINUOUS NEGATIVE STRESS

OLD STATE OF EQUILIBRIUM

Systems PROCESS

Adaptation Scales

Ecologies Potential

Systems of change

Ω Connectedness

CONTINUOUS VARIANT STRESS

System fails when external instances exceed the parameters of resilience

Calculated Constraints

Undefined Boundary

Catagorised

Diversity

Equilibrium Point

Maximum Resilient Range

New Factor Imposed

Time

Temporary Instances

Maximised Efficiency

Equilibrium Point

Resilience Theory Engineered Resilience In the creation of a resistant system, there are 2 approaches the first is engineered resilience, this focuses on understanding and preserving a current state. For instance building to withstand weather conditions within a specific expected range. This method maintains the efficiency of a function while under load and always aims to return a system to the initial point of equilibrium. The quality of this Engineered Resilience can be measured by 2 quantitative measures; a systemâ&#x20AC;&#x2122;s resistance to disturbance and, speed of return to the equilibrium.

Continuity Productivity Efficiency

Cons: Requires Control Vulnerability

Equlibrium Point

Redundency

Point of Equilibrium Begins To Shift New Equilibrium Point

Diversity Opportunity Persistance Cons: Unpredictability No Redundency

Equlibrium Point

Equilibrium Line Begins To Deform

Resilience Theory Ecological Resilience An Ecological Resilient system is not constrained within an expected range of conditions but is instead an interconnection of adaptable systems. Ecology recognises the natural development and variation of a state with multiple equilibrium points and as part of a wider system. This maintains the existence of a system but not necessarily the function of the system itself. Ecological Resistance can be measured in terms of the magnitude of disturbance that can be absorbed before the system changes structure.

System fails when external instances exceed the parameters of resilience

Calculated Constraints

Undefined Boundary

Catagorised

Diversity

Equilibrium Point

Maximum Resilient Range

New Factor Imposed

Time

Temporary Instances

Maximised Efficiency

Equilibrium Point

Continuity Productivity Efficiency

Cons: Requires Control Vulnerability

Equlibrium Point

Redundency

Point of Equilibrium Begins To Shift New Equilibrium Point

Diversity Opportunity Persistance Cons: Unpredictability No Redundency

Equlibrium Point

Equilibrium Line Begins To Deform

SYSTEM DRIVERS SYSTEM FLOWS SYSTEM ACTIVE DESPITE DRIVER FAILURE

DRIVER BREAKS RESULTING IN A SYSTEM FAILURE

NEW SYSTEM STATE

FUNCTIONAL SYSTEM ACTORS

DEFUNCT SYSTEM ACTORS

Resilience Theory Defining system properties ‘Human foresight and innovation can reverse trends [of exploitation and destruction and develop paths that sustain natural diversity and create opportunity.’

SELF HEALING

SELF ORGANISING

MULTI FUNCTIONAL

MULTI APPLICABLE

SYSTEM DRIVERS SYSTEM FLOWS SYSTEM ACTIVE DESPITE DRIVER FAILURE

DRIVER BREAKS RESULTING IN A SYSTEM FAILURE

NEW SYSTEM STATE

FUNCTIONAL SYSTEM ACTORS

DEFUNCT SYSTEM ACTORS

SELF HEALING

SELF ORGANISING

MULTI FUNCTIONAL

MULTI APPLICABLE

04 | The CPU Brief Applying Resilience Theory to Pavilion Design

Existing Knowledge: existing material knowlage

Architects Statement What is the intent of the pavilion? Every year the designer of the serpentine pavilion as a change to explain their hopes for their pavilion, explaining the reasoning, context and meaning behind what they aimed to create.

The Climate Crisis is the current largest existential threat to humanity. Increasingly unpredictable weather events are already having severe effect as a direct result. In the UK, whole months worth of rain are falling in the space of days. We want to use our pavilion proposal as a testbed to explore how architecture can intervene to create systems resilient to these dramatic changes. In selecting this approach for developing our pavilion, we will return it to its original function; testing an idea to push the field of architecture forward. This will steer it away from the repetitive, commodified and superficial designs that currently saturate the typology.

Untested Architectural Concept: can a building be optimised using genetic algorithms to advance resilient building techniques?

This idea will also inform how we create the public space within our pavilion. Through an exploration of methods for developing emotive atmospheres, we create a spatial realisation that communicate to visitors the theme that have driven our design. Pavilion Testbed

Proven idea

Failed Test

Existing Knowledge: existing material knowlage

Untested Architectural Concept: can a building be optimised using genetic algorithms to advance resilient building techniques?

Proven idea

Failed Test

Decentralised Systems The pavilion is designed as a fragment of a wider system. As such, failure would have limited impact on the wider issue.

Resilience Applied Steps Taken to Ensure a Resilient Design We made decisions throughout the project with a priority towards ensuring Resilience informed our decision making.

Redundency The pavilion is developed for extreme scenarios, meaning it has redundency instilled in its fabrication. The varying uses of the pavilion and the ability to recycle rainwater ensure this is still a valuable design decision.

Structurally Optimised

Replaceable Elements

Using Karamba we have engineered the structure to ensure it is definitely capable of coping with the most extreme water loading it could receive as a result of extreme weather.

The pavilion is constructed of smaller, easily refabricatable components that can be replaced if damaged.

MultiFunctionality

Ease of Fabrication

The pavilion serves a range of purposes, including a sheltered cafe, a piece of rainwater harvesting infrastructure, a piece of sculpture, a think-piece.

In line with our UN SDGs we have designed the structure to be easily fabricatable, but still requiring laymen work in order to avoid eliminating jobs as the fabrication industry transitions into being more automated.

Decentralised Systems The pavilion is designed as a fragment of a wider system. As such, failure would have limited impact on the wider issue.

Resilience Applied Steps Taken to Ensure a Resilient Design We made decisions throughout the project with a priority towards ensuring Resilience informed our decision making.

Structurally Optimised

Replaceable Elements

Using Karamba we have engineered the structure to ensure it is definitely capable of coping with the most extreme water loading it could receive as a result of extreme weather.

The pavilion is constructed of smaller, easily refabricatable components that can be replaced if damaged.

MultiFunctionality

Ease of Fabrication

The pavilion serves a range of purposes, including a sheltered cafe, a piece of rainwater harvesting infrastructure, a piece of sculpture, a think-piece.

05 | Site Analysis Multi-Scale Assessment of Site to Indentify Local Issues

Site Analysis: Meso Key Nodes & Circulation The site is the serpentine pavilion, indicated in the site plan below. The diagram illustrates the parks access points and circulation to and from these points, key nodes and the site. The circulation allows us to assess the frequency and density of pedestrian flow through the park towards the site. This will eventually influence out form to develop a footprint that is more considered towards access etc.

Park Access Points

Local Underground Links

Key Nodes in Park

Circulation line Key :

Circulation in Park

Circulation to Park

Serpentine Pavilion Site

Park Access Points

Local Underground Links

Key Nodes in Park

Circulation line Key :

Circulation in Park

Circulation to Park

Serpentine Pavilion Site

Sunlight Hours in June

Sunlight Hours in August

Sunlight Hours in October

June is the first month the serpentine is open, and is the brightest months of the year.

August is the midpoint for the serpentine pavilions opening time of the year.

October is the last month the serpentine is open. It is the darkest month, yet most of the site remains well lit.

Serpentine Gallery Highest density of pedestrian flow (based on the meso analysis)

Sunlight Hours Key : 0H

15H

Park Access Points

Site Analysis: Micro Access & Environmental Analysis The serpentine pavilion site sits within the bounds of the serpentine gallery site. The circulation is based on the access points and the meso analysis to find which access points would receive a higher proportion of visitors flowing through. It is simple, but can be manipulated with a complex pavilion intervention. The solar analysis also feeds directly in to form, allowing us to then consider our intervention with all information required to consider.

Serpentine Site Sun Path The sun path illustrated to the right represents the sun positions at all times of the day between June and October. Coupled with the shadow analysis this allows us to consider spacial qualities induced by our form.

Sunlight Hours in June

Sunlight Hours in August

Sunlight Hours in October

June is the first month the serpentine is open, and is the brightest months of the year.

August is the midpoint for the serpentine pavilions opening time of the year.

October is the last month the serpentine is open. It is the darkest month, yet most of the site remains well lit.

Serpentine Gallery Highest density of pedestrian flow (based on the meso analysis)

Sunlight Hours Key : 0H

15H

Park Access Points

City Water Management

Drainage & Flooding

How a city manages its water can have a dramatic effect on the way the city functions efficiently.

Urban foundation construction decreases the permeability of the earth for run off of water.

Surface or Ground Water Pavilion

City

London’s Water Crisis Applying Resilience theory

Industrial Wastewater Treatment

Water Recycling

Domestic Water Treatment

Industrial Water Treatment Industrial Trade

Waste Water Treatment Plant

Resource recovery Biosolids and Biogas

Centralised vs De-centralised

Increase in population and loss of green space

The Victorian system of water control in London, built for 2.5 million people, is stretched to failure. As a centralised system, this affects far more people than otherwise. In contrast when a de-centralised fails minimal areas are affected.

‘Loss of green space, growing population and ageing water pipes are putting the capital at an increased risk of flooding and drought.’ - Leonie Cooper

Population

As a coastal city on an island nation its assumed that London has water in abundance. However, in fact, due to un-resilient infrastructure and drainage, the city faces crisis on multiple fronts.

Possible Responses Water systems should be more decentralised to avoid major disruption. Rainfall and flood water needs restrategising to offset existing issues within the city. As conditions escalate, efforts must be made to predict and pre empt major disruption.

Green Space

Time

Water wasted

Water available

Centralised

De-centralised

City Water Management

Drainage & Flooding

How a city manages its water can have a dramatic effect on the way the city functions efficiently.

Urban foundation construction decreases the permeability of the earth for run off of water.

Surface or Ground Water Pavilion

City

London’s Water Crisis Applying Resilience theory

Industrial Wastewater Treatment

Water Recycling

Domestic Water Treatment

Industrial Water Treatment Industrial Trade

Waste Water Treatment Plant

Resource recovery Biosolids and Biogas

Centralised vs De-centralised

Increase in population and loss of green space

‘Loss of green space, growing population and ageing water pipes are putting the capital at an increased risk of flooding and drought.’ - Leonie Cooper

Population

As a coastal city on an island nation its assumed that London has water in abundance. However, in fact, due to un-resilient infrastructure and drainage, the city faces crisis on multiple fronts.

Green Space

Time

Water wasted

Water available

Centralised

De-centralised

06 | Design Strategies Addressing Climate Issues Qualitatively and Quantativley

Material Tectonics Hardness Texture

interior / Exterior Tension Connectivity with wider world Perception of transition between inside and outside Detachment from environment or element

Finish

Atmospheric Design

OPENNESS

PATH FINDING AND MOVEMENT

Defining a Spatial Intent Anxiety Whilst elements of our design are highly function orientated, the spatial quality of our pavilion is also a key contributor to our design. Following reading Silvia Lavinâ&#x20AC;&#x2122;s text on Pavilions we agreed that investigating the interplay of Art and Architecture would yield a more relevant proposal. We used Peter Zumpthorâ&#x20AC;&#x2122;s text on atmospheres to explore how he develops architectural environments.

Calmness

CONTEXTUAL CONNECTION

Material Tectonics Hardness Texture

interior / Exterior Tension Connectivity with wider world Perception of transition between inside and outside Detachment from environment or element

Finish

Atmospheric Design

OPENNESS

PATH FINDING AND MOVEMENT

Calmness

CONTEXTUAL CONNECTION

MOCA PAVILION Tom Wiscombe

Sharp, jutting and intersecting geometry

Changes of material and texture

Variations in route width

Pleat rises to figure Pleat

Advanced water-based polymer with Kevlar textile reinforcement

Interface

2D // 3D Tension

Changes in level

Reduced sense of relative scale and repetition

Playing with interior exterior tension

Presecriptive circulation, modulating from high shelter to low

Focus on contrast from high volume and views spaces to low

Atmospheres Applied Pavilion Scale Strategies With our research from our atmospheres exploration complete and our form optimised to create emergent conditions, strategies were employed to further orchestrate the experience of our pavilion in the design of the details and landscaping.

Distribute pools to give space back to site and generate a cooled environment within the pavilion

MOCA PAVILION Tom Wiscombe

Sharp, jutting and intersecting geometry

Changes of material and texture

Variations in route width

Pleat rises to figure Pleat

Advanced water-based polymer with Kevlar textile reinforcement

Interface

2D // 3D Tension

Changes in level

Reduced sense of relative scale and repetition

Playing with interior exterior tension

Presecriptive circulation, modulating from high shelter to low

Focus on contrast from high volume and views spaces to low

Distribute pools to give space back to site and generate a cooled environment within the pavilion

Volume of rain

Traditional Design

Resilient Design

Flooding

No flooding

Planned Flooding Slow Dispersion

ROTTERDAM: AN EXAMPLE

As a cities are large areas covered in primarily hard none porous surfaces it isnâ&#x20AC;&#x2122;t uncommon for heavy rain to cause short term localised flooding.

In the city of Rotterdam a lot of land has been reclaimed from the sea and for this reason most of the city is susceptible to flooding with sudden heavy rain, it is therefore important to slow the rate of water reaching the sewer system.

Any method that can slow the rate water is released into the cities system will have a dramatic effect on the chances of flooding within the city.

One of the methods of storage is through sunken park areas that are used normally 90% of the time but when required can store vast amounts of water helping to prevent flash flooding.

Volume of rain

Traditional Design

Resilient Design

Flooding

No flooding

Planned Flooding Slow Dispersion

ROTTERDAM: AN EXAMPLE

As a cities are large areas covered in primarily hard none porous surfaces it isnâ&#x20AC;&#x2122;t uncommon for heavy rain to cause short term localised flooding.

Any method that can slow the rate water is released into the cities system will have a dramatic effect on the chances of flooding within the city.

One of the methods of storage is through sunken park areas that are used normally 90% of the time but when required can store vast amounts of water helping to prevent flash flooding.

Form The form of the architectural intervention can be parametrised to control its effect at points through.

Density

Tree Distance

Variance

Opening

Volume

Shelter

Entrance

Light

Wind

Trees

Rain

Existing Paths

Capacity

Space Volume

No. of Spaces

Entrance size

Route Length

No. of Routes

Building Distance

Supports

Porosity / Openings

Boundary

Points of Interest

Context The attributes of the site context can be measured and manipulated as parameters to control the nature of the engagement of the user.

Levels

Agents The behaviour of agents can be influenced. Capacity can be varied and natural movement and stopping can be developed.

Shelter

Free Space

Input Parameters Isolating the input variables for parametric massing iterations When designing using a algorithm as a base the inputs that will make up part of this algorithm need to be considered carefully.

Project Specific The approach to our project, based on a theoretical, problem centric intention, provides its own unique parameters to control.

Water Capacity

Spans

Support Distance

Area

Tectonics

Form The form of the architectural intervention can be parametrised to control its effect at points through.

Density

Tree Distance

Variance

Opening

Volume

Shelter

Entrance

Light

Wind

Trees

Rain

Existing Paths

Capacity

Space Volume

No. of Spaces

Entrance size

Route Length

No. of Routes

Building Distance

Supports

Porosity / Openings

Boundary

Points of Interest

Context The attributes of the site context can be measured and manipulated as parameters to control the nature of the engagement of the user.

Levels

Agents The behaviour of agents can be influenced. Capacity can be varied and natural movement and stopping can be developed.

Shelter

Free Space

Project Specific The approach to our project, based on a theoretical, problem centric intention, provides its own unique parameters to control.

Water Capacity

Spans

Support Distance

Area

Tectonics

TARGET

Objectives & Constraints Primary Goals For the Design Defining goals relevant to our core design drivers and Methods for dealing with them I should be incapable of drawing a single stroke at the present moment; and yet I feel that I never was a greater artist than now. When, while the lovely

TARGET

Objective Functions

Constraint functions

MAXIMISE

MINIMISE <=

MAXIMISE

MINIMISE <=

OBJECTIVES & CONSTRAINTS

RESILIENCE

Water Collection Water Runoff Rate Water Distribution Total Footprint Structural Performance

ATMOSPHERE

Variance in Enclosure Variance in Internal Height Reduced wayfinding Vast expanses Reverberation Coarse textures Structural Intensity

GENERATIVE DESIGN

COMPUTATIONAL EXPERIMENTATION Water Simulation Structural Simulation

TRADITIONAL DESIGN

TARGET

Objective Functions

Constraint functions

MAXIMISE

MINIMISE <=

MAXIMISE

MINIMISE <=

OBJECTIVES & CONSTRAINTS

RESILIENCE

Water Collection Water Runoff Rate Water Distribution Total Footprint Structural Performance

ATMOSPHERE

Variance in Enclosure Variance in Internal Height Reduced wayfinding Vast expanses Reverberation Coarse textures Structural Intensity

GENERATIVE DESIGN

COMPUTATIONAL EXPERIMENTATION Water Simulation Structural Simulation

TRADITIONAL DESIGN

Catenary Surface Deciding the Base Form A Catenary Surface is an example of Extensive Curvature which has seen extensive use in Pavilion design. It relies on the Parabola, a shape found throughout nature and in structural systems.

Parabolas

Existance in Nature Structural Performance

Working With Gravity

A Catenary is based on a system of parabolas. This efficiently transfers load along the surface and to the ground.

A parabola is a naturally occuring phenomena. The force of gravity causes projectiles to follow this shape as they move.

Long Spanning

Water Run-Off Water falling on the structure will naturally run off the edges of the surface. The form will also support the water load.

Natural Suspensions

Water Collection

This shape features in natural structures such as spiders webs of silk, networks of radial thread cross connected by secondary threads which the spider can hang from. Gravity gives them this shape

Multiple sequacious Catenary forms will generate pools at their internal anchor points.

Minimised Weight

Robust Shells

The form of the structure is naturally materially efficient, distributing material along load paths.

The perfomativity of the form means that they have evolved in nature. For example this tortoise shell exoskeleton robustly protects the tortoiseâ&#x20AC;&#x2122;s organs.

Proportion Collected

The pavilion is a public space which, especially on the rainy days in which the pavilion exhibits its potential, will require a large amount of shelter space.

Working With Water

More extensive structures with more internal anchors, well distributed, will collect a high proportion of the rain fall.

Parabolas

Existance in Nature Structural Performance

Working With Gravity

A Catenary is based on a system of parabolas. This efficiently transfers load along the surface and to the ground.

A parabola is a naturally occuring phenomena. The force of gravity causes projectiles to follow this shape as they move.

Long Spanning

Water Run-Off Water falling on the structure will naturally run off the edges of the surface. The form will also support the water load.

Natural Suspensions

Water Collection

This shape features in natural structures such as spiders webs of silk, networks of radial thread cross connected by secondary threads which the spider can hang from. Gravity gives them this shape

Multiple sequacious Catenary forms will generate pools at their internal anchor points.

Minimised Weight

Robust Shells

The form of the structure is naturally materially efficient, distributing material along load paths.

The perfomativity of the form means that they have evolved in nature. For example this tortoise shell exoskeleton robustly protects the tortoiseâ&#x20AC;&#x2122;s organs.

Proportion Collected

The pavilion is a public space which, especially on the rainy days in which the pavilion exhibits its potential, will require a large amount of shelter space.

Working With Water

More extensive structures with more internal anchors, well distributed, will collect a high proportion of the rain fall.

Boundary Anchoring

Full Edge

Catenary Surface

PROS

Converting Our Footprint to a Mesh Can we develop something to functionally complete a structural simulation. Can we achieve complex footprints Can we explore different grid systems. Kangaroo takes load, node and surface data from a script in order to output a structurally rational surface.

Parameters

Load Force

Edge Length

Mesh Density

Anchors

Limited Anchor Points PROS

Enclosure Form Flexibility

Views out Flexible Access

CONS

No Openings Irrational Mesh

High Exposure Form Limitation

Nature of Mesh Polygons

Boundary Anchoring

Full Edge

Catenary Surface

PROS

Parameters

Load Force

Edge Length

Mesh Density

Anchors

Limited Anchor Points PROS

Enclosure Form Flexibility

Views out Flexible Access

CONS

No Openings Irrational Mesh

High Exposure Form Limitation

Nature of Mesh Polygons

Process Explore ways of generating meshes.

Anchors

Mesh Experimentation Explore square and triangle based mesh constructions.

Generate boundaries from straight edged and curved edges shapes. Randomise anchor locations.

Goals

Can we achieve complex footprints Can we explore different grid systems. What density of mesh division is required.

Can we develop something to functionally produce a catenary surface.

Conclusion 7

Any polygonal geometry can be used by subdividing larger boundaries into polygons that do not self intersect.

10 13

Process Explore ways of generating meshes.

Anchors

Mesh Experimentation Explore square and triangle based mesh constructions.

Generate boundaries from straight edged and curved edges shapes. Randomise anchor locations.

Goals

Can we achieve complex footprints Can we explore different grid systems. What density of mesh division is required.

Can we develop something to functionally produce a catenary surface.

Conclusion 7

Any polygonal geometry can be used by subdividing larger boundaries into polygons that do not self intersect.

10 13

Complex Meshes High Polygon and Line Based Anchors Experimentation was completed with increasingly complex base polygons, connecting sequences of varied scales, proportions and edge counts.

Process Apply mesh logic to larger footprints and more complex shapes. Generate anchors based on whole lines and curves.

Goals To achieve random anchor distribution across multiple polygons. To combine square and triangle based meshes. To work out how and to what extent lines can be used for anchors.

Conclusion Any polygonal geometry can be used by subdividing larger boundaries into polygons that do not self intersect. Line based anchors are created using adjacent anchor points from a mesh surface. This limits the nature of anchorage. However limited line based anchor processes can be completed.

Complex Footprints

Line base anchors were also explored, as apposed to point based ones. These rely on a sequence of adjacent points from the mesh.

Line Based Anchors

Complex Meshes High Polygon and Line Based Anchors Experimentation was completed with increasingly complex base polygons, connecting sequences of varied scales, proportions and edge counts.

Process Apply mesh logic to larger footprints and more complex shapes. Generate anchors based on whole lines and curves.

Goals To achieve random anchor distribution across multiple polygons. To combine square and triangle based meshes. To work out how and to what extent lines can be used for anchors.

Complex Footprints

Line base anchors were also explored, as apposed to point based ones. These rely on a sequence of adjacent points from the mesh.

Line Based Anchors

07 | Generative Design Process Formulating a Systems Strategy to Satisfy Design Goals Generativley

Generative Design Process Optimising Based on Project Criteria A catenary surface can be parametrically generated and analysed to achieve project specific properties using a Generative Design solver. We generated a script in Grasshopper that incorporates our selected most important features and parameters of the design.

PARAMETERS

FITNESS CRITERIA Phenotype

Number of Edge Points

View Area Variance

Random Distribution Variant

Canopy Height Variance

Number of Internal Points

Maximum Height

Random Distribution Variant

Maximum Water Retention

Find max height

OPTIMISE As Close to Chosen Max as Possible

12.

4. Divide footprint into grid

DeďŹ ne Building Footprint

Convert into uninverting polygons

9. (A - MEAN)2 + (B - MEAN)2 + Etc Number of Data Points

Select Random Edge Points

Cut Mesh Horizontally at Eye Height: Adults & Children

Select Random Inside Points Input Chosen Points as Anchors DeďŹ ne Canopy Extension

4. Sample a random number of edge and internal mesh points

OPTIMISE

Maximise Variance

Measure water runoff distribution 5. From the original footprint, divide into an equally spaced grid of points

6. Generate the catenary surface from the mesh and random anchor points

7. Calculate the proportion of rainwater retained on the surface

Calculate Variance

Subdivide into panels

3. Subdivide the polygons into a triangulated mesh of panels

Run Views Analysis for each grid point

11.

Structural Simulation

Convert polygons to mesh

2. Subdivide this into polygons of similar area and proportion

Calculate Variance

1. Define a polygon footprint based on requirements for the site

10.

Measure height of mesh at each point

8. Measure the height of the canopy at each grid point

9. Run a views analysis at each eye height, using the sections and context geometry

Calculate Proportion That Stays on Surface

10. Calculate the variance of canopy height and views at each grid point

11. Run an optimisation to balance variance of height, views out and rain retained

12. Optimise the catenary surface height for the most high performing genome

PARAMETERS

FITNESS CRITERIA Phenotype

Number of Edge Points

View Area Variance

Random Distribution Variant

Canopy Height Variance

Number of Internal Points

Maximum Height

Random Distribution Variant

Maximum Water Retention

Find max height

OPTIMISE As Close to Chosen Max as Possible

12.

4. Divide footprint into grid

DeďŹ ne Building Footprint

Convert into uninverting polygons

9. (A - MEAN)2 + (B - MEAN)2 + Etc Number of Data Points

Select Random Edge Points

Cut Mesh Horizontally at Eye Height: Adults & Children

Select Random Inside Points Input Chosen Points as Anchors DeďŹ ne Canopy Extension

4. Sample a random number of edge and internal mesh points

OPTIMISE

Maximise Variance

Measure water runoff distribution 5. From the original footprint, divide into an equally spaced grid of points

6. Generate the catenary surface from the mesh and random anchor points

7. Calculate the proportion of rainwater retained on the surface

Calculate Variance

Subdivide into panels

3. Subdivide the polygons into a triangulated mesh of panels

Run Views Analysis for each grid point

11.

Structural Simulation

Convert polygons to mesh

2. Subdivide this into polygons of similar area and proportion

Calculate Variance

1. Define a polygon footprint based on requirements for the site

10.

Measure height of mesh at each point

8. Measure the height of the canopy at each grid point

9. Run a views analysis at each eye height, using the sections and context geometry

Calculate Proportion That Stays on Surface

10. Calculate the variance of canopy height and views at each grid point

11. Run an optimisation to balance variance of height, views out and rain retained

12. Optimise the catenary surface height for the most high performing genome

Avoidance of Leaf Fall 51%

61%

Positioning the footprint away from the base of trees, especially down wind, will avoid leaves gathering on the canopy.

44%

Site Usage %

Defining Footprint [1]

53%

41%

52%

In order to facilitate itâ&#x20AC;&#x2122;s rainwater harvesting function to the maximum potential the size of the footprint should be maximised.

Site access points. Positioning the polygon to create definite openings relative to key points for accessing the site.

Generating a Basis for Optimisation The base Polygon from which the pavilion catenary surface was generated was based on 6 criteria. Whilst some of these criteria are quantitative, many are qualitative. As such the generative design process is only partially relevant, but still informed our decision process.

The potential views variance. 42%

55%

46%

Fewer definite anchor points as a result of corners in the polygon will allow larger openings between them and therefore more potential variation in views out.

Variation in levels of shelter Shelter is another variable of variance defined by the distance you are from the edge of the canopy at any one point. The variation of this should be maximised.

53%

42%

58%

Optimum Site Locations The site was analysed and regions of optimum light and views were found. These areas have the highest potential variance so should be included in the pavilion.

Avoidance of Leaf Fall 51%

61%

Positioning the footprint away from the base of trees, especially down wind, will avoid leaves gathering on the canopy.

44%

Site Usage %

Defining Footprint [1]

53%

41%

52%

In order to facilitate itâ&#x20AC;&#x2122;s rainwater harvesting function to the maximum potential the size of the footprint should be maximised.

Site access points. Positioning the polygon to create definite openings relative to key points for accessing the site.

The potential views variance. 42%

55%

46%

Fewer definite anchor points as a result of corners in the polygon will allow larger openings between them and therefore more potential variation in views out.

Variation in levels of shelter Shelter is another variable of variance defined by the distance you are from the edge of the canopy at any one point. The variation of this should be maximised.

53%

42%

58%

Optimum Site Locations The site was analysed and regions of optimum light and views were found. These areas have the highest potential variance so should be included in the pavilion.

Grid and Mesh [2-4] Divide the Footprint into a Grid of points and a mesh We divided the footprint into a grid of points. These would be used for our analysis and also as locations to distribute anchors to. 4m was shown to spread the anchors by a minimum amount and in order to provide sufficient atmosphere analysis.

2 - Convert to triangles

We also divided the footprint into a mesh surface. This was subdivided 12 times per triangle. This lead to a satisfactorily detailed canopy without overloading the processing time of the script.

3 - Mesh Subdivision x12

4 - Grid Division 4m

2 - Convert to triangles

We also divided the footprint into a mesh surface. This was subdivided 12 times per triangle. This lead to a satisfactorily detailed canopy without overloading the processing time of the script.

3 - Mesh Subdivision x12

4 - Grid Division 4m

Divide the polygon footprint into a 4m grid of points.

Select Internal Points [5]

Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

Parameter 1 The first parameter selects the number and location of internal anchor points of the form. This is selected from the previously defined regular grid of points, to avoid them being within 4m of one another. These internal anchor points will serve two functions: •

Determine the location of pools of rainwater to collect on the surface.

•

Alter and vary the volumetric quality of the canopy at different points.

# of Anchor Points 1 3 5 7 9 11 13 Random Seed 1 2 3 4 5 6 7 8 9 10

We elected tohave a high range of internal anchor points, from 1 to 13, which would create a high variety of internal spaces. NOTE: A ‘Seed’ is a speciﬁc variation of the randomisation.

Select random points from this grid and then find the closest mesh point to those points.

These selected points inform where the catenary comes down within the boundary, creating more enclosed spaces and channeling water to the inside space.

Divide the polygon footprint into a 4m grid of points.

Select Internal Points [5]

Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

Determine the location of pools of rainwater to collect on the surface.

•

Alter and vary the volumetric quality of the canopy at different points.

# of Anchor Points 1 3 5 7 9 11 13 Random Seed 1 2 3 4 5 6 7 8 9 10

We elected tohave a high range of internal anchor points, from 1 to 13, which would create a high variety of internal spaces. NOTE: A ‘Seed’ is a speciﬁc variation of the randomisation.

Select random points from this grid and then find the closest mesh point to those points.

These selected points inform where the catenary comes down within the boundary, creating more enclosed spaces and channeling water to the inside space.

These anchor points are deﬁnite, in order for the catenary to generate properly.

Select Edge Points [5] Parameter 2 The second parameter selects the number and location of edge anchor points of the form. All points are turned to a short line of 3 points to increase the area the water can runoff down.

Randomly Select 3 Consecutive Points of the remaining edge anchor points Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

The existing corners of the footprint polygon already definitely have anchors in order for the form to work. The random edge points are selected from the remaining evenly divided points along each edge line. These are also a sequence of 3 consecutive points. These anchor points will have two functions •

To create more enclosure and obstructed views

•

To provide extra routes for the rainwater to dissipate and runoff down

# of Anchor Lines 0

Random Seed 1 2 3 4 5 6 7 8 9 10

We elected to limit the number of edge anchors to between 0 and 3 as beyond this reduced the water run off too much. NOTE: A ‘Seed’ is a speciﬁc variation of the randomisation.

These selected points inform where the catenary comes down at its edge, creating more enclosed spaces and channeling water to the ground.

These anchor points are deﬁnite, in order for the catenary to generate properly.

Randomly Select 3 Consecutive Points of the remaining edge anchor points Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

To create more enclosure and obstructed views

•

To provide extra routes for the rainwater to dissipate and runoff down

# of Anchor Lines 0

Random Seed 1 2 3 4 5 6 7 8 9 10

We elected to limit the number of edge anchors to between 0 and 3 as beyond this reduced the water run off too much. NOTE: A ‘Seed’ is a speciﬁc variation of the randomisation.

These selected points inform where the catenary comes down at its edge, creating more enclosed spaces and channeling water to the ground.

Catenary Construction [6]

Load Force

Construction of the form Having selected our footprint, anchors and a base surface Kangaroo can take these defined elements and construct a rational surface to meet design criteria. Initially this is just a fixed hight but other factors can be imparted into the process.

Edge Length

Mesh Density

Anchors

Catenary Construction [6]

Load Force

Edge Length

Mesh Density

Anchors

Water Run Off [7] Fitness Criteria 1 - Maximised This criteria determines how much rainwater would run on to the surface and how much would run off the edge. It simulates droplets, calculating their runoff path and determining their destination.

Step 1 Populate the surface with 500 randomly distributed points.

Step 2 Each point will then calculate in what direction it needs to move in order to most efficiently leave the surface. This is indicative of gravity.

A proportion is generated between those staying on the surface and ending up at the internal anchor points, and those ending up at the edge. This value is maximised to increase potential retention of rainwater.

For each point, it calculates which direction it has to move to have the greatest increase in Z vector. It repeats this 100mm at at a time, until all points have a Z value of 0.

The water droplets either end up at the edge anchors or internal.

Step 3 The points, after 8000 vector movements, will accumulate at the internal or external anchor points.

Step 4 The points, after 8000 vector movements, will accumulate at the internal or external anchor points.

# Particles Retained 500 Total Particles Rainwater Retained

Step 1 Populate the surface with 500 randomly distributed points.

Step 2 Each point will then calculate in what direction it needs to move in order to most efficiently leave the surface. This is indicative of gravity.

For each point, it calculates which direction it has to move to have the greatest increase in Z vector. It repeats this 100mm at at a time, until all points have a Z value of 0.

The water droplets either end up at the edge anchors or internal.

Step 3 The points, after 8000 vector movements, will accumulate at the internal or external anchor points.

Step 4 The points, after 8000 vector movements, will accumulate at the internal or external anchor points.

# Particles Retained 500 Total Particles Rainwater Retained

Views Area Variance [8]

NOTE: View variance and canopy height are directly proportional to proximity to water bodies on the roof canopy. This mass of water will have an atmospheric affect also.

Fitness Criteria 2 - Maximised An Isovist is run at each predetermined grid point under the canopy, excluding those used as anchors for that simulation. The area that can be seen out to from this point is calculated.

These area values are then put into a Variance calculation to identify how much this changes throughout the pavilion. This variance is maximised.

This is completed for a cross section of the canopy at the average eye height for an adult and also a child.

Eg 1

Eg 2

The isovists measure how much of the wider site can be seen at any one point. 50m

100m

50m

100m

50m

100m

Measure area at each grid point at both child and adult height.

Eg 3

Serpentine Gallery

Site Trees

Above is an example of the section cut lines of the mesh at each height.

1.8m- Average Adult Eye 1.2m- Average Child Eye

Eg 4

50m

100m

Views Area Variance [8]

NOTE: View variance and canopy height are directly proportional to proximity to water bodies on the roof canopy. This mass of water will have an atmospheric affect also.

These area values are then put into a Variance calculation to identify how much this changes throughout the pavilion. This variance is maximised.

This is completed for a cross section of the canopy at the average eye height for an adult and also a child.

Eg 1

Eg 2

The isovists measure how much of the wider site can be seen at any one point. 50m

100m

50m

100m

50m

100m

Measure area at each grid point at both child and adult height.

Eg 3

Serpentine Gallery

Site Trees

Above is an example of the section cut lines of the mesh at each height.

1.8m- Average Adult Eye 1.2m- Average Child Eye

Eg 4

50m

100m

Canopy Height Variance [9] Fitness Criteria 3 - Maximised An Height measurement is run at each predetermined grid point under the canopy. These height values are then put into a Variance calculation to identify how much this changes throughout the pavilion. This variance is maximised.

For each grid point it measures a height up to the canopy surface

NOTE: View variance and canopy height are directly proportional to proximity to water bodies on the roof canopy. This mass of water will have an atmospheric affect also.

For each grid point it measures a height up to the canopy surface

NOTE: View variance and canopy height are directly proportional to proximity to water bodies on the roof canopy. This mass of water will have an atmospheric affect also.

VARIANCE Range

A + B + C + D + E + F + G + Etc.

Standard Deviation

ME Standard Deviation

Number of Data Points

Range

Variance [10] Calculating the diversity of qualities of atmospheres

Variance

(A - MEAN)2 + (B - MEAN)2 + Etc Number of Data Points

From our atmospheres research we have ascertained one of the key features of the space we want to create is maximised contrast and dynamic change throughout. This feature can be applied to most qualities of the space. It can also be quantitatively calculated. To do this, all the values of views or height were input into the following calculation. VARIANCE

MEAN

High Range Low Mean Deviation

High Range High Mean Deviation

Low Range High Mean Deviation

VARIANCE Range

A + B + C + D + E + F + G + Etc.

Standard Deviation

ME Standard Deviation

Number of Data Points

Range

Variance [10] Calculating the diversity of qualities of atmospheres

Variance

(A - MEAN)2 + (B - MEAN)2 + Etc Number of Data Points

MEAN

High Range Low Mean Deviation

High Range High Mean Deviation

Low Range High Mean Deviation

08 | Generating Design Iterations Producing Initial Canopy Iterations of Varying Performance

Design Priorities

Maximum Water Retention

Qualitative & Subjective Assessment Approach

View Area Variance

The final solution will be selected based from a pool of high performing genomes, but will be chosen based also on its aesthetic value and how well it fulfils the qualitative elements of the design also.

Limitations & Biases Features Not Addressed By The System With any system there will be a number of qualities not addressed. Compromises will also have to be made between thoroughness and processing time. The following key elements, plus many others which would have less impact, were considered carefully for their implication.

Resilience is the fundamental element of the project, with atmospheres fulfilling an important secondary role. As such, this will be a priority when selecting options.

Irrelevant Views Analysis Views analysis only occurred at points that were not used as anchors. However very rarely some view points still positioned above the canopy so were irrelevant. This was unavoidable other than to increase the general height of the pavilion to reduce the rate it did.

Views Analysed

Still Sometimes Above Canopy

Canopy Height Variance

Anchor Automatically Excluded

Incurred Pool Volume While the system can assess how much water would pool on the roof, it does not calculate the volume the roof can hold at any one location. Points too close to the edge were therefore disregarded. Visual assessment would also assist this, plus our geometric strategy.

Proximity of Large open spaces to South

Visual quality of Form

Inferred circulation route interest

High Collection Rate Low Volume Capacity

High Collection Rate High Volume Capacity

Data Aggregation For a large number of parameters the aggregation of analysis has been reduced to limit the processing requirements of the script and reduce runtime to a more reasonable duration. These were decided to avoid major inaccuracy from the overall system. Examples include: •

View Area directions: Views area was calculated every 12 degrees out from each grid point.

•

Mesh subdivisions: The pavilion mesh is subdivided 12 times per polygon. This had major reductions in failures.

•

Water runoff Points: Water run off was only calculated from 500 points.

Continuous Headway

Partially Randomised

Whilst a proportion of the canopy can be made headroom accessible, the system cannot be coded to ensure it is continuous and doesn’t create dead ends. Outputs will therefore have to be assessed visually to ensure the solution is viable for circulating.

2 of the 4 variables that define the anchors are based on randomisation. This is unavoidable in order to produce a variety of options without being prescriptive, when the number of solutions must be limited. This means the solver may struggle to find high degrees of correlation.

Dead end with no headroom

Balance

Design Priorities

Maximum Water Retention

Qualitative & Subjective Assessment Approach

View Area Variance

Resilience is the fundamental element of the project, with atmospheres fulfilling an important secondary role. As such, this will be a priority when selecting options.

Views Analysed

Still Sometimes Above Canopy

Canopy Height Variance

Anchor Automatically Excluded

Proximity of Large open spaces to South

Visual quality of Form

Inferred circulation route interest

High Collection Rate Low Volume Capacity

High Collection Rate High Volume Capacity

View Area directions: Views area was calculated every 12 degrees out from each grid point.

•

Mesh subdivisions: The pavilion mesh is subdivided 12 times per polygon. This had major reductions in failures.

•

Water runoff Points: Water run off was only calculated from 500 points.

Continuous Headway

Partially Randomised

Dead end with no headroom

Balance

Analysing Iterations

Anchor Points (Internal) 13

Initial evaluation of Footprint Geometry A constrained analysis of each footprint gave us a first indication of how these each performed based on a pool of 12 iterations each.

Random Seed 10

6 8 4

Height Variance MAX 9.2

View Area Variance MAX 280

Water Retention MAX 0.62

Analysing Iterations

Anchor Points (Internal) 13

Initial evaluation of Footprint Geometry A constrained analysis of each footprint gave us a first indication of how these each performed based on a pool of 12 iterations each.

Random Seed 10

6 8 4

Height Variance MAX 9.2

View Area Variance MAX 280

Water Retention MAX 0.62

High Performers

Neural Network

High Performers

Refined Footprint Selection 1

Refined Footprint Selection 2

From the previous analysis, we could refine our options down to 6. Decision on this was on a much more subjective grounds. However the data allowed us to have conviction that we were choosing an objectively strong option.

The footprints with a low number of edge vertices performed very highly in terms of water retention. However they covered only a low portion of the site and were inarticulate in their form.

Anchor Locations 52% Site Coverage 62% Water Retention

5.4

Height Variance

199.3 Views Variance 62% Water Retention

Anchor Locations 44% Site Coverage 63% Water Retention

Neural Network

5.9

Height Variance

195.2 Views Variance 63% Water Retention

High Performers

Neural Network

High Performers

Refined Footprint Selection 1

Refined Footprint Selection 2

The footprints with a low number of edge vertices performed very highly in terms of water retention. However they covered only a low portion of the site and were inarticulate in their form.

Anchor Locations 52% Site Coverage 62% Water Retention

5.4

Height Variance

199.3 Views Variance 62% Water Retention

Anchor Locations 44% Site Coverage 63% Water Retention

Neural Network

5.9

Height Variance

195.2 Views Variance 63% Water Retention

High Performers

Neural Network

High Performers

Refined Footprint Selection 3

Refined Footprint Selection 4

This one had low site coverage, but avoided enclosing the footpath through the site. It has high retention and very high views variance from its highest performer.

This option filled the site the most but as a result of its high number of anchor points it had a low water retention. It had high height variance potential because of its size also.

Anchor Locations 44% Site Coverage 48% Water Retention

7.2 Height Variance 239.9 Views Variance 48% Water Retention

Anchor Locations 61% Site Coverage 34% Water Retention

Neural Network

8.4 Height Variance 221.7 Views Variance 34% Water Retention

High Performers

Neural Network

High Performers

Refined Footprint Selection 3

Refined Footprint Selection 4

This one had low site coverage, but avoided enclosing the footpath through the site. It has high retention and very high views variance from its highest performer.

This option filled the site the most but as a result of its high number of anchor points it had a low water retention. It had high height variance potential because of its size also.

Anchor Locations 44% Site Coverage 48% Water Retention

7.2 Height Variance 239.9 Views Variance 48% Water Retention

Anchor Locations 61% Site Coverage 34% Water Retention

Neural Network

8.4 Height Variance 221.7 Views Variance 34% Water Retention

High Performers

Neural Network

High Performers

Refined Footprint Selection 5

Refined Footprint Selection 6

This option interfaced with the 3 entry points well. It had high water retention but low site coverage and low views variance.

This option had the second highest site coverage and a high water retention. It also interfaced well with the entry points well, with a high potential view and height variance. We also appreciated the aesthetic of the outputs.

Anchor Locations 41% Site Coverage 57% Water Retention

6.9 Height Variance 200.6 Views Variance 57% Water Retention

Anchor Locations 58% Site Coverage 55% Water Retention

Neural Network

7.0 Height Variance 242.2 Views Variance 55% Water Retention

High Performers

Neural Network

High Performers

Refined Footprint Selection 5

Refined Footprint Selection 6

This option interfaced with the 3 entry points well. It had high water retention but low site coverage and low views variance.

Anchor Locations 41% Site Coverage 57% Water Retention

6.9 Height Variance 200.6 Views Variance 57% Water Retention

Anchor Locations 58% Site Coverage 55% Water Retention

Neural Network

7.0 Height Variance 242.2 Views Variance 55% Water Retention

58%

Selected Footprint Identifying the Base Footprint From these 6 options our final footprint was chosen, with the data in mind and the qualitative features appraised. It found a good balance of limiting the number of vertices whilst still filling a large percentage of the site. It also interfaced well with a number of entry points creating an enticing entrance to draw visitors in. Whilst this is not the most high performing on proportion of water retained, this was only by a small degree, plus, its large footprint will allow it to generally accumulate more water.

58%

09 | Iteration Optimisation Refining Canopy Iterations for Highest Performers

2 2

6.5

6.0

Most Water Retention

Sort by:

400

0.5

0.0

Water Run On/Off

0.1 350



15/01/2020

Design Explorer 2

  Reset Selection Exclude Selection Zoom to Selection Save Selection to File My Static Link Tutorial (http://www.mpendesign.com/category/tutorial/) Services (https://www.thorntontomasetti.com/services/sustainability/) Info  (https://github.com/tt-acm/DesignExplorer/wiki)

Number of anchors (Internal)

15/01/2020

Design Explorer Exploration

Seed (Internal)

Number of anchors (Edge)

Seed (Edge)

3.0

8.5

2.5

Height Variance

Design Explorer 2

View Area Variance

550

8.0

5 2.0 500 6 7.5 Info  (https://github.com/tt-acm/DesignExplorer/wiki)   Reset Selection Exclude Selection Zoom to Selection Save Selection to File My Static Link Tutorial (http://www.mpendesign.com/category/tutorial/) Services (https://www.thorntontomasetti.com/services/sustainability/) 4 1.5 3 450 4 7.0 3 1.0 Number of anchors (Internal) Seed (Internal) Number of anchors (Edge) Seed (Edge) Height Variance View Area Variance 2 400 7 10 3.0 5 2 6.5 2 0.5 8.5 550 0 6.0 6 2.5 8 350 4 1 0.0 1 8.0 5 2.0 500 6 7.5 Sort by: Water Run On/Off  4 1.5 3 450 4 7.0 3 1.0 2 400 2 6.5 2 0.5 0 6.0 350 1 0.0 1 Number of anchors (Internal) Seed (Internal) Height Variance View Area Vari… Water Run On/… img Rating scid 1 Sort by: 1

Water Run On/Off

-1 3



9.171116

354.60162

0.186

Numberofanch… 0

11.899018

335.279197

0.096

Numberofanch… 0

9.251827

343.174129

0.134

Numberofanch… 0

7.411818

335.521255

0.144

Numberofanch… 0

-1

12.316014

538.761235

0.506

Numberofanch… 0

Second wave of iterations

11.909349

500.461175

0.524

Numberofanch… 0

13.627861

512.733879

0.444

Numberofanch… 0

15.071996

536.292243

0.502

Numberofanch… 0

15.089412

629.012985

0.666

Numberofanch… 0

Design Explorer is a system for documenting, visualising and evaluating design options. It is not a solver, instead capturing images and data for each option which can then be uploaded to an interface designed to explore the design space.

12.381195

591.738595

0.602

Numberofanch… 0

16.266078

612.323185

0.562

Numberofanch… 0

Setting L M S 

Water Run On/Off

0.5 0.4 Setting L M S  0.3 0.2Run On/Off Water 0.5 0.1 0.4 0.3 0.2 0.1

tt-acm.github.io/DesignExplorer/

As can be seen from the range of Anchor points and Seeds that were active in the neural network (Right) arbritrary decisions had to be made by us about limiting the scope of the design space. Design Explorer has no evolution mechanic so is not optimal for evolving solutions. It is, however, very efficient at visualising and ordering options. This system also starkly demonstrated the positive correlation between internal anchor points and performance.

1/1

Anchor Points Random Seed Number of anchors (Internal) Seed (Internal) Height Variance (Internal) 1

13 1

-1

Anchor Points (External)

View Area Vari… Water Run On/… img

scid

Numberofanch… 0

0.134

Numberofanch… 0

0.144

Numberofanch… 0

538.761235

0.506

Numberofanch… 0

500.461175

0.524

Numberofanch… 0

15.071996 9.171116 15.089412 11.899018 12.381195 9.251827 16.266078 7.411818

512.733879 View Area Vari… 536.292243 354.60162 629.012985 335.279197 591.738595 343.174129 612.323185 335.521255

0.444 Water Run On/… 0.502 0.186 0.666 0.096 0.602 0.134 0.562 0.144

Numberofanch… img Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch…

7 scid 10 0 5 3 8 6 11 9

12.316014

538.761235

0.506

Numberofanch… 0

9.171116

10 11.899018

354.60162

0.186

335.279197

0.096

9.251827

343.174129

7.411818

335.521255

-1

12.316014

11.909349

4 Number of anchors (Internal) 4 1 7 1 7 1 7 1

6 Seed (Internal) 10 -1 3 3 6 6 10 10

-1

4 7

Random Seed Rating

613.627861 Height Variance 5

411.909349

0 Rating 0 0 0 0 0 0 0 0

500.461175

0.524

Numberofanch… 0

13.627861

512.733879

0.444

Numberofanch… 0

15.071996

536.292243

0.502

Numberofanch… 0

15.089412

0.666

Numberofanch… 0

112.381195

629.012985 591.738595

0.602

Numberofanch… 0

16.266078

612.323185

0.562

Numberofanch… 0

tt-acm.github.io/DesignExplorer/

Height Variance MAX 7.2

View Area Variance MAX 253.4

Water Retention MAX 0.56

High performance correlated with high internal anchor points.

1/1

2 2

6.5

6.0

Most Water Retention

Sort by:

400

0.5

0.0

Water Run On/Off

0.1 350



15/01/2020

Design Explorer 2

Number of anchors (Internal)

15/01/2020

Design Explorer Exploration

Seed (Internal)

Number of anchors (Edge)

Seed (Edge)

3.0

8.5

2.5

Height Variance

Design Explorer 2

View Area Variance

550

8.0

Water Run On/Off

-1 3



9.171116

354.60162

0.186

Numberofanch… 0

11.899018

335.279197

0.096

Numberofanch… 0

9.251827

343.174129

0.134

Numberofanch… 0

7.411818

335.521255

0.144

Numberofanch… 0

-1

12.316014

538.761235

0.506

Numberofanch… 0

Second wave of iterations

11.909349

500.461175

0.524

Numberofanch… 0

13.627861

512.733879

0.444

Numberofanch… 0

15.071996

536.292243

0.502

Numberofanch… 0

15.089412

629.012985

0.666

Numberofanch… 0

12.381195

591.738595

0.602

Numberofanch… 0

16.266078

612.323185

0.562

Numberofanch… 0

Setting L M S 

Water Run On/Off

0.5 0.4 Setting L M S  0.3 0.2Run On/Off Water 0.5 0.1 0.4 0.3 0.2 0.1

tt-acm.github.io/DesignExplorer/

1/1

Anchor Points Random Seed Number of anchors (Internal) Seed (Internal) Height Variance (Internal) 1

13 1

-1

Anchor Points (External)

View Area Vari… Water Run On/… img

scid

Numberofanch… 0

0.134

Numberofanch… 0

0.144

Numberofanch… 0

538.761235

0.506

Numberofanch… 0

500.461175

0.524

Numberofanch… 0

15.071996 9.171116 15.089412 11.899018 12.381195 9.251827 16.266078 7.411818

512.733879 View Area Vari… 536.292243 354.60162 629.012985 335.279197 591.738595 343.174129 612.323185 335.521255

0.444 Water Run On/… 0.502 0.186 0.666 0.096 0.602 0.134 0.562 0.144

Numberofanch… img Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch… Numberofanch…

7 scid 10 0 5 3 8 6 11 9

12.316014

538.761235

0.506

Numberofanch… 0

9.171116

10 11.899018

354.60162

0.186

335.279197

0.096

9.251827

343.174129

7.411818

335.521255

-1

12.316014

11.909349

4 Number of anchors (Internal) 4 1 7 1 7 1 7 1

6 Seed (Internal) 10 -1 3 3 6 6 10 10

-1

4 7

Random Seed Rating

613.627861 Height Variance 5

411.909349

0 Rating 0 0 0 0 0 0 0 0

500.461175

0.524

Numberofanch… 0

13.627861

512.733879

0.444

Numberofanch… 0

15.071996

536.292243

0.502

Numberofanch… 0

15.089412

0.666

Numberofanch… 0

112.381195

629.012985 591.738595

0.602

Numberofanch… 0

16.266078

612.323185

0.562

Numberofanch… 0

tt-acm.github.io/DesignExplorer/

Height Variance MAX 7.2

View Area Variance MAX 253.4

Water Retention MAX 0.56

High performance correlated with high internal anchor points.

1/1

Octopus Solution Solver Experimentation

Pros

The system maps outputs along 3 axes as well as colour gradients allowing the visualisation of numerous fitness criteria. The phenomes are also each represented in 3d within the matrix.

•

Has highly parametric mutation and elitism

•

More prone to crashing randomly

•

Visual exploration of design, relative to one another.

•

Allows you to select high performing Genomes and Phenomes to replicate.

Has a number of glitches including obscured data values in the visualisation.

•

Generates results quickly.

Evolution Parameters ce

rian

s Va

View

The mutation and elitism parameters are controllable within the system to vary the outputs visualised in the matrix.

Experiment 1 Each iteration during the experiment phase was run for 20 minutes to give an indication of what it would output. The first was completed with the baseline advised parameters.

Heig

ht V aria

nce

Elitism

n Water Retentio

3d Matrix Visualisation

Water Retention

Octopus is a multi-objective solution visualiser. It generates 3d geometry for a selection of the potential solutions, as well as documenting their fitness values for the objectives. It then visualises them on a 3d matrix.

Cons

Elitism

Marked Output visualised in 3d

Mutation Probability

Mutation Rate

Crossover Rate

Population Size

XXX

0.54

Vie

Mutation Probability

ari sV

.52

186

.99

9.2

Population Size Diversify Parameters

igh

tV ari

200

5.6

Octopus Solution Solver Experimentation

Pros

The system maps outputs along 3 axes as well as colour gradients allowing the visualisation of numerous fitness criteria. The phenomes are also each represented in 3d within the matrix.

•

Has highly parametric mutation and elitism

•

More prone to crashing randomly

•

Visual exploration of design, relative to one another.

•

Allows you to select high performing Genomes and Phenomes to replicate.

Has a number of glitches including obscured data values in the visualisation.

•

Generates results quickly.

Evolution Parameters ce

rian

s Va

View

The mutation and elitism parameters are controllable within the system to vary the outputs visualised in the matrix.

Experiment 1 Each iteration during the experiment phase was run for 20 minutes to give an indication of what it would output. The first was completed with the baseline advised parameters.

Heig

ht V aria

nce

Elitism

n Water Retentio

3d Matrix Visualisation

Water Retention

Cons

Elitism

Marked Output visualised in 3d

Mutation Probability

Mutation Rate

Crossover Rate

Population Size

XXX

0.54

Vie

Mutation Probability

ari sV

.52

186

.99

9.2

Population Size Diversify Parameters

igh

tV ari

200

5.6

Data Ranges Water Retention: 61% View Variance: 271.9

0.16

Height Variance: 7.6

Best-Fit Geometry Selected

Mutation Probability

For the final run the time was not constrained, and the parameters were applied in two phases.

Mutation Rate

The first phase had higher mutation and lower elitism to explore as much of the design space as possible.

Crossover Rate Population Size Diversify Parameters

200

High Performing Phenome ce

ari sV

Vie 0.55

The second phase had high elitism to evolve higher performing solutions

n Water Retentio

Final Selection

Elitism

8.2

Conclusions

9.4

Mutation Probability

igh

tV ari

High Performing Genome

Mutation Rate

Crossover Rate 7.0

Population Size Diversify Parameters

200

The final output was one of the highest performing on two fronts, with the maximum water retention which was a design priority. The system struggled to evolve solutions. If re-done, and if more computer power was available, a more advanced process for determining anchors would have been better.

Data Ranges Water Retention: 61% View Variance: 271.9

0.16

Height Variance: 7.6

Best-Fit Geometry Selected

Mutation Probability

For the final run the time was not constrained, and the parameters were applied in two phases.

Mutation Rate

The first phase had higher mutation and lower elitism to explore as much of the design space as possible.

Crossover Rate Population Size Diversify Parameters

200

High Performing Phenome ce

ari sV

Vie 0.55

The second phase had high elitism to evolve higher performing solutions

n Water Retentio

Final Selection

Elitism

8.2

Conclusions

9.4

Mutation Probability

igh

tV ari

High Performing Genome

Mutation Rate

Crossover Rate 7.0

Population Size Diversify Parameters

200

Low Performer Minimum Fitness Values This option shows the worst of the options Octopus returned to us within its interface.

Data

Height

Views

Rain Retention

The low performer produced the following data. Water Retention: 6.2% View Variance: 144.54 Height Variance: 6.2

Analysis

Properties

It has low anchor points meaning it has limited potential for variance. It also retains almost no water.

The data output was severely low showing the range of the design options available.

Height Variance Views Variance Water Retention

Low Performer Minimum Fitness Values This option shows the worst of the options Octopus returned to us within its interface.

Data

Height

Views

Rain Retention

The low performer produced the following data. Water Retention: 6.2% View Variance: 144.54 Height Variance: 6.2

Analysis

Properties

It has low anchor points meaning it has limited potential for variance. It also retains almost no water.

The data output was severely low showing the range of the design options available.

Height Variance Views Variance Water Retention

Average Performer Median Fitness Values This option shows the worst of the options Octopus returned to us within its interface.

Data

Height

Views

Rain Retention

The average performer produced the following data. Water Retention: 31.2% View Variance: 175.2 Height Variance: 8.1

Analysis

Properties

This option created good height variance as a result of its low anchor points with a high amount of spacing. However it had low water retention.

The data output shows mediocre values, though still a considerable improvement from the previous. It showed insufficient enclosed spaces with a priority towards vast volumes.

Height Variance Views Variance Water Retention

Average Performer Median Fitness Values This option shows the worst of the options Octopus returned to us within its interface.

Data

Height

Views

Rain Retention

The average performer produced the following data. Water Retention: 31.2% View Variance: 175.2 Height Variance: 8.1

Analysis

Properties

This option created good height variance as a result of its low anchor points with a high amount of spacing. However it had low water retention.

The data output shows mediocre values, though still a considerable improvement from the previous. It showed insufficient enclosed spaces with a priority towards vast volumes.

Height Variance Views Variance Water Retention

Optimum Selection Selected and Articulated The optimum output was selected. The form was then regenerated with double the mesh subdivision. It was also re-analysed to much higher tolerances. Adjustments of the anchors based on aesthetic and improving the mesh properties concluded with the following outputs.

Data

Height

Views

Rain Retention

Once adjusted the data outputs changed. This was considered through the adjustment with live feedback to tailor the output to our requirements. Water Retention: 59% View Variance: 280.2 Height Variance: 7.3

Analysis

Properties

The anchors clustered together to create low level spaces in regions. Whilst there are very high volumes there is a high proportion of tightly enclosed space with substantial pool volumes above.

The final output showed substantial improvement from the original best output extracted through Design Explorer in order to select the footprint. This can be seen in the orange bars (Left).

Some anchors were adjusted inwards from the edge slightly to improve water capacity. The output was selected based on higher volume spaces being localised towards to South face with high light potential. The edge anchors were redistributed to aid this.

Large volume for cafe space.

The form inferred a long winding and dynamically changing circulation route through the space, deviating inward and out to the edge of the canopy.

Height Variance Views Variance Water Retention

Data

Height

Views

Rain Retention

Analysis

Properties

The anchors clustered together to create low level spaces in regions. Whilst there are very high volumes there is a high proportion of tightly enclosed space with substantial pool volumes above.

The final output showed substantial improvement from the original best output extracted through Design Explorer in order to select the footprint. This can be seen in the orange bars (Left).

Large volume for cafe space.

The form inferred a long winding and dynamically changing circulation route through the space, deviating inward and out to the edge of the canopy.

Height Variance Views Variance Water Retention

10 | Landscaping The Leviathan Applying Field Theory For Circlation Routes

Field Conditions Enriching Emergent Properties of the Pavilion Through the generative design process of inducing high spatial and views variance across the site, the system has generated maximums and minimums within the pavilion, that collaborate with the existing features of the site. Using the canopy geometry and properties, key nodes will be selected as stopping points in the circulation through the space and synthesising the legibility of the design.

Canopy Height MIN

MAX

View Area -ve

+ve

Selected Nodes Selected Anchors

Field Theory

Node Selection

Field theory relies on systems of organisation capable of producing vortexes peaks and protuberances out of individual elements. Gradients and transitions define the properties of spaces.

The nodes are selected to aid in amplifying the emergent properties of the pavilion. They are selected on a range of criteria.

Field Generation Each of these nodes will become critical features of the design, key spaces to pause in. Field lines generated by giving the nodes charges will inform our geometry.

Intensity of activity MIN

MAX

Flow towards Nodes WayďŹ nding and Focuses

Canopy Height MIN

MAX

View Area -ve

+ve

Selected Nodes Selected Anchors

Field Theory

Node Selection

Field theory relies on systems of organisation capable of producing vortexes peaks and protuberances out of individual elements. Gradients and transitions define the properties of spaces.

The nodes are selected to aid in amplifying the emergent properties of the pavilion. They are selected on a range of criteria.

Field Generation Each of these nodes will become critical features of the design, key spaces to pause in. Field lines generated by giving the nodes charges will inform our geometry.

Intensity of activity MIN

MAX

Flow towards Nodes WayďŹ nding and Focuses

Orchestrating Milieu Field Informed Elements and Landscaping The fields generated can inform the distribution and geometry of different materials and features throughout the site.

GSEducationalVersion

Orchestrating Milieu Field Informed Elements and Landscaping The fields generated can inform the distribution and geometry of different materials and features throughout the site.

GSEducationalVersion

Pavilion

Without Solar Strategy

Opening 24th June

Environmental Optimisation

Midway 10th August

Closing 24th November

Base Analysis

The base solar analysis with the whole canopy, demonstrates the extensive dark environment within the pavilion, into which we can bring light.

Manipulating Natural Light

Panel Selection

We also used our fields outputs to manipulate the natural light penetrating into the pavilion. Our method for this was to replace opaque panels with translucent ones. This would visually communicate the channeling strategy of the canopy to inhabitants also.

We ran the Ladybug solar analysis, adjusting our roof panel selection in order to align changes in light and dark with the nodes of our fields.

Pavilion

With Solar Strategy

Opening 24th June

Midway 10th August

Closing 24th November

Pavilion

Without Solar Strategy

Opening 24th June

Environmental Optimisation

Midway 10th August

Closing 24th November

Base Analysis

The base solar analysis with the whole canopy, demonstrates the extensive dark environment within the pavilion, into which we can bring light.

Manipulating Natural Light

Panel Selection

We ran the Ladybug solar analysis, adjusting our roof panel selection in order to align changes in light and dark with the nodes of our fields.

Pavilion

With Solar Strategy

Opening 24th June

Midway 10th August

Closing 24th November

11 | Water Runoff Study Methods Methods For Testing flood mitigation Strategies

Water Study Goals

SLOWING THE RATE OF DRAINAGE OFF PAVILION

Site saturation is a major issue to mitigate when considering rain induced flash flooding. This can be controlled with the nature of the pavilions skin, which will be considered as a test ‘surface’ for the sake of this experiment. Flood risks can be reduced by slowing the rate of the rainwater draining off these ‘surfaces,’ and so this experiment aims to test shapes to formulate efficient strategies to slow the draining of water.

WATER FLOWING OFF SURFACE

PERMANENT WATER POOLING

In this experiment we want to avoid permanent water pooling as our primary research driven experiment is framed around the distribution of water, not the collection. Pooling is still possible, however a strategy would need to be employed to eventually dissipate this as well. WATER-SATURATED GROUND

LITRES OF RAIN TO FALL

SITE SATURATION-LITRES

AVOID PERMANENT WATER POOLING

LITRES OF WATER TO DRAIN OFF PAVILION PER SECOND (RATE)

Fundamental Goal: slow the rate of drainage off our pavilion (l/s)

Water Study Goals

SLOWING THE RATE OF DRAINAGE OFF PAVILION

WATER FLOWING OFF SURFACE

PERMANENT WATER POOLING

LITRES OF RAIN TO FALL

SITE SATURATION-LITRES

AVOID PERMANENT WATER POOLING

LITRES OF WATER TO DRAIN OFF PAVILION PER SECOND (RATE)

Fundamental Goal: slow the rate of drainage off our pavilion (l/s)

Water Study Overview Staged Process Summary The diagram below illustrates the process of developing a strategy to slow the rate of water flow off a pavilion. The process begins with generating methods to calculate metrics to input and output. This in turn aids us to assess the performance of strategies we formulate to achieve our studies goal.

[1]

1. Set up raining / channelling simulation

[2]

[5]

[6]

2. Calculate input data for simulation

5. Iterate basic strategies

6. Evaluate best performing strategies

[3]

[4]

3. Calculate output data for simulation to find performance

4. Define performance with metrics

[10]

[7]

[8]

[9]

7. Fuse best performing strategies

8. Iterate new fused strategy

9. Define best performing strategy

10. Apply best strategy to pavilion

[11]

11. Test final pavilionâ&#x20AC;&#x2122;s performance

[1]

1. Set up raining / channelling simulation

[2]

[5]

[6]

2. Calculate input data for simulation

5. Iterate basic strategies

6. Evaluate best performing strategies

[3]

[4]

3. Calculate output data for simulation to find performance

4. Define performance with metrics

[10]

[7]

[8]

[9]

7. Fuse best performing strategies

8. Iterate new fused strategy

9. Define best performing strategy

10. Apply best strategy to pavilion

[11]

11. Test final pavilionâ&#x20AC;&#x2122;s performance

Water Study Animation [1] Particle Simulation For Rain

Water Study Methods [1] Conditions

STAGE 2 PARTICLE PROPERTIES CONSIDERED

MIMICKING WATER The properties of the particles are considered to create a grouped behaviour as close to water as allowed by the systems capacity.

PHYSICS ENGINE

STAGE 1 -

FRICTION

TESTING SURFACES

SMALL PARTICLE GENERATION

PARTICLE SYSTEM BREAKDOWN COHESION

TIME FACTORED

We will generate basic test surfaces to allow us to begin to formulate strategies towards our goal by generating metrics to assess.

COLLISION

FUNDAMENTAL FORCES

RADIUS

PARTICLE SIMULATION SET UP To achieve our experiment goals underlines previously, we are using a particle simulation system in order to replicate water flowing onto and off the surface. This will help us achieve our goal metric of rate of drainage off our pavilion.

STAGE 3 TEST SURFACES

STAGE 4 ANALYSE METRICS FOR FITNESS

Water Study Methods [1] Conditions

STAGE 2 PARTICLE PROPERTIES CONSIDERED

MIMICKING WATER The properties of the particles are considered to create a grouped behaviour as close to water as allowed by the systems capacity.

PHYSICS ENGINE

STAGE 1 -

FRICTION

TESTING SURFACES

SMALL PARTICLE GENERATION

PARTICLE SYSTEM BREAKDOWN COHESION

TIME FACTORED

We will generate basic test surfaces to allow us to begin to formulate strategies towards our goal by generating metrics to assess.

COLLISION

FUNDAMENTAL FORCES

RADIUS

STAGE 3 TEST SURFACES

STAGE 4 ANALYSE METRICS FOR FITNESS

Start

Stage 2 1m³

Water Study Methods [2]

STAGE 1 PARTICLES TO LITRES

1M³ BOX = 842 560 PARTICLES

X 842 560

In order to obtain and manipulate data for strategy generation, particles must be converted into a universal metric to analyse outputs and determine performance levels for each experiment. Stage 1 illustrates our methodology in converting the particles into litres.

Input Metrics Stage 1 This experiment has a 4 stage process in establishing our methodology to reach our goal. Stage 1 is the process in which we convert our particles in the simulation into a known metric, which in this case is litres. The methods taken to do this started with filling a 1m³ empty box with particles. As 1m³ is equivalent to 1000Litres of water we can deduce that the amount of particles that fill the box is equal to this. This then gives us a formula to convert particles to litres throughout the experiment.

FILL A 1M³ BOX WITH PARTICLES

1M³ BOX = 1000 LITRES

1L FINAL CONVERSION

842.56 Particles

Start

Stage 2 1m³

Water Study Methods [2]

STAGE 1 PARTICLES TO LITRES

1M³ BOX = 842 560 PARTICLES

X 842 560

FILL A 1M³ BOX WITH PARTICLES

1M³ BOX = 1000 LITRES

1L FINAL CONVERSION

842.56 Particles

To maintain accuracy and realism within the experiment, a recent metric from the met office is used to represent heavy rain. Using the metric conversion from Stage 1, stage 2 illustrates how we translate heavy rain in litres back into particles as a metric to simulate as well as considering time in seconds.

Stage 3 32L PER HOUR PER Sqm

STAGE 2 -

DETERMINING HEAVY RAIN

2019 HEAVIEST RAIN = 32mm PER HOUR

Water Study Methods [2] Input Metrics Stage 2

60² ÷

7.49 PARTICLES FALLING PER SECOND

0.0088L Per Sec Per Sqm

CONVERT HOURS TO SECONDS FOR EXPERIMENT TIME METRIC HEAVY RAIN RATE

Stage 2 follows on from stage 1 by using the conversion formula to determine the rate at which the experiment particles should be falling. Using weather data from the Met Office we can convert the most recent metric for heavy rain the UK into a particle format, so the experiment accurately resembles heavy rain. This completes the translation from particles to litres and back again to produce ‘rain.’

Stage 3 32L PER HOUR PER Sqm

STAGE 2 -

DETERMINING HEAVY RAIN

2019 HEAVIEST RAIN = 32mm PER HOUR

Water Study Methods [2] Input Metrics Stage 2

60² ÷

7.49 PARTICLES FALLING PER SECOND

0.0088L Per Sec Per Sqm

CONVERT HOURS TO SECONDS FOR EXPERIMENT TIME METRIC HEAVY RAIN RATE

Water Study Methods [3] Output Metrics EXPERIMENT COMPLETE : APPLY STRATEGY TO FORM

INPUT HEAVY RAIN METRIC

REPEAT STAGE 3 TO ITERATE STRATEGIES UNTIL OUTPUT METRIC IS OPTIMISED (MINIMISED)

TIME ELAPSED (SECONDS) - t

STAGE 3 -

00 : 10 : 00

PRODUCING SIMULATION OUTPUT METRIC

STAGE 4 TESTING STRATEGIES

BREED HIGHEST PERFORMING STRATEGIES

PARTICLES BELOW LOWEST POINT (LITRES) - x

RUN 10 MINUTE SIMULATION

To test the result for ‘performance’ an output metric is required. Stages 3 & 4 illustrate the process we undertake to acquire and assess this metric to produce an optimal water channelling strategy

GENERATE BASIC STRATEGIES (SKIP STEP AFTER ITERATION 1)

INPUT PARTICLE TO LITRE METRIC

OUTPUT METRIC : LITRES PER SECOND TO FLOW OFF SURFACE (x ÷ t)

ASSESS HIGHEST PERFORMING STRATEGIES

Water Study Methods [3] Output Metrics EXPERIMENT COMPLETE : APPLY STRATEGY TO FORM

INPUT HEAVY RAIN METRIC

REPEAT STAGE 3 TO ITERATE STRATEGIES UNTIL OUTPUT METRIC IS OPTIMISED (MINIMISED)

TIME ELAPSED (SECONDS) - t

STAGE 3 -

00 : 10 : 00

PRODUCING SIMULATION OUTPUT METRIC

STAGE 4 TESTING STRATEGIES

BREED HIGHEST PERFORMING STRATEGIES

PARTICLES BELOW LOWEST POINT (LITRES) - x

RUN 10 MINUTE SIMULATION

GENERATE BASIC STRATEGIES (SKIP STEP AFTER ITERATION 1)

INPUT PARTICLE TO LITRE METRIC

OUTPUT METRIC : LITRES PER SECOND TO FLOW OFF SURFACE (x ÷ t)

ASSESS HIGHEST PERFORMING STRATEGIES

Identifying Bias - Surface Size

To remove the bias from the experiment, the output is adjusted to consider the ratio between the rate of flow per second and the input surface’s footprint area, this is our revised output metric:

Adjusted Output Metric: LITRES PER SECOND PER M² TO FLOW OFF SURFACE (l/s/m²)

1m²

RATIO OF RATE OF FLOW TO SURFACE AREA

MORE FLOW OFF SURFACE (L) LARGER FOOTPRINT SURFACE AREA MORE PARTICLES / LITRES ‘RAINING’

Water Study Methods [3] Eliminating Bias Although our experiment’s main goal is accomplished following our methodology, it holds a bias against larger test surfaces. Larger surfaces require a larger area for rain to be distributed, and as such more particles are involved with that particular experiment. This results in a larger final output for larger surfaces and smaller ones for smaller surfaces. This bias detracts from the goal of developing high performance strategies to deal with heavy rain. The diagram to the right illustrates the bias and how we overcame it by creating a ration based on size and the original metric of rate of flow off the surface in litres per second.

The experiment’s output metric is currently bias to surface footprint size. The larger the surface the more rain will fall onto it, and so more litres will flow off per second. A strategy must be formulated to maintain consistency to focus the experiment solely on water channelling.

Eliminating Bias - Standardising Size

Identifying Bias - Surface Size

To remove the bias from the experiment, the output is adjusted to consider the ratio between the rate of flow per second and the input surface’s footprint area, this is our revised output metric:

Adjusted Output Metric: LITRES PER SECOND PER M² TO FLOW OFF SURFACE (l/s/m²)

1m²

RATIO OF RATE OF FLOW TO SURFACE AREA

MORE FLOW OFF SURFACE (L) LARGER FOOTPRINT SURFACE AREA MORE PARTICLES / LITRES ‘RAINING’

Eliminating Bias - Standardising Size

Final Data Output :

0.93 ml/s/m2

Low Performance

Millilitres to Flow off Surface per Second per m2

Final Data Output :

High Performance

Low Performance

Medium-High Performance

LITRES OF RAIN (TOTAL)

2.64 ml/s/m2

Higher Final Data Outputs are Lower Performing

High Performance Low Performance

LITRES OF RAIN (TOTAL)

MILLILITRES TO FLOW OFF SURFACE PER SECOND

71.7 l

55.1 l

High-Medium Value

Medium-High Value

BASIC SLOPE

UMBRELLA 37 ml/s

81 ml/s

Medium-High Value

Low Value TIME ELAPSED (Minutes)

TIME ELAPSED (Minutes)

FOOTPRINT SURFACE AREA (mÂ˛)

10 Minutes

Standardised Value

40.24m2

Medium Value

Water Study Methods [4] Defining High Performance Data Now the metrics and methodology for conducting the experiment are established, it is important to display the data produced in a uniform template to visually and numerically compare strategies. This example illustrates a medium - high performing strategy of a basic slope. The final data output requires minimising (maintaining above 0) to achieve high performance. This is due to our goal of slowing the rate of water flow off generated strategies down.

30.63m2

Medium-Low Value

Water Study Methods [4] Defining Low Performance Data This example illustrates a low performing strategy, modelled after an umbrella. This strategy clearly allows water to flow off it at a much faster rate than the slope example to the left. This is why higher final data outputs represent lower performances. The remainder of the data is present to visualise all the inputs of the experiment established in the methodology, and understand how these values culminate to produce the final output.

Final Data Output :

0.93 ml/s/m2

Low Performance

Millilitres to Flow off Surface per Second per m2

Final Data Output :

High Performance

Low Performance

Medium-High Performance

LITRES OF RAIN (TOTAL)

2.64 ml/s/m2

Higher Final Data Outputs are Lower Performing

High Performance Low Performance

LITRES OF RAIN (TOTAL)

MILLILITRES TO FLOW OFF SURFACE PER SECOND

71.7 l

55.1 l

High-Medium Value

Medium-High Value

BASIC SLOPE

UMBRELLA 37 ml/s

81 ml/s

Medium-High Value

Low Value TIME ELAPSED (Minutes)

TIME ELAPSED (Minutes)

FOOTPRINT SURFACE AREA (mÂ˛)

10 Minutes

Standardised Value

40.24m2

Medium Value

30.63m2

Medium-Low Value

12 | Water Runoff Rate Experiment Adapting Pavilion Form Through Iterations Based off Experiment Data

Breed Best Performing Strategies

2.16 ml/s/m2

0.903 ml/s/m2

0.578 ml/s/m2

0.375 ml/s/m2

Roof & Gutters

Branches

Half Pipe

Valley

Low Performance Strategy

Low-Medium Performance Strategy

Medium-Low Performance Strategy

Medium Performance Strategy

2.66 ml/s/m2

0.261 ml/s/m2

0.313 ml/s/m2

0.093 ml/s/m2

Cone

Meander

Stepped

Spiral

Low Performance Strategy

Medium-High Performance Strategy

High Performance Strategy

Water Study Iteration 1 [5] Basic Geometry As stated in the methodology, testing basic geometry for performance is our first step towards creating a water channelling strategy. The data graphs establish these performance levels. As the Spiral is the highest performing strategy this will be key in the breeding process. The stepped and Meander strategies also perform well and so will be factored in with less weight. Other strategies are not carried through but may have merit down the line.

Key :

Low Performance Strategy

2nd Tier Breeding Candidate Strategy

Top Tier Breeding Candidate Strategy

Breed Best Performing Strategies

2.16 ml/s/m2

0.903 ml/s/m2

0.578 ml/s/m2

0.375 ml/s/m2

Roof & Gutters

Branches

Half Pipe

Valley

Low Performance Strategy

Low-Medium Performance Strategy

Medium-Low Performance Strategy

Medium Performance Strategy

2.66 ml/s/m2

0.261 ml/s/m2

0.313 ml/s/m2

0.093 ml/s/m2

Cone

Meander

Stepped

Spiral

Low Performance Strategy

Medium-High Performance Strategy

High Performance Strategy

Key :

Low Performance Strategy

2nd Tier Breeding Candidate Strategy

Top Tier Breeding Candidate Strategy

13.8m

Path Length

Performance

0.261 ml/s/m2

0.20 ml/s/m2 28.1m Path Length

Increase Path Length

Improved Performance

1.73m

Height & Path Angle

Performance

Adjusted Output Metric

0.186 ml/s/m2 0.87m & 5° Height & Path Angle

Meander Evaluation [6]

Decrease Height & Path Angle

9.78°

High Performance Attributes

Graph Key :

Median Angle Original Attribute Data

New Iteration Data

Improved Performance

45.3° Performance

The meander strategy was determined to be a high performing strategy in iteration 1. We must now deconstruct the attributes of this shape to assess what contributes to high performance and what doesn’t. As seen in the diagram the change of these attributes produces a differing output metric. All of these altering attributes produce a better performing strategy when increased or decreased. This will be factored into the breeding process.

Adjusted Attribute Metric

0.17 ml/s/m2 26.1°

Median Angle

Decrease Median Angle

Improved Performance

13.8m

Path Length

Performance

0.261 ml/s/m2

0.20 ml/s/m2 28.1m Path Length

Increase Path Length

Improved Performance

1.73m

Height & Path Angle

Performance

Adjusted Output Metric

0.186 ml/s/m2 0.87m & 5° Height & Path Angle

Meander Evaluation [6]

Decrease Height & Path Angle

9.78°

High Performance Attributes

Graph Key :

Median Angle Original Attribute Data

New Iteration Data

Improved Performance

45.3° Performance

Adjusted Attribute Metric

0.17 ml/s/m2 26.1°

Median Angle

Decrease Median Angle

Improved Performance

7.41m

Path Length

Performance

0.313ml/s/m2

0.08 ml/s/m2 19.4m Path Length

Increase Path Length

Improved Performance

1.83m

Height & Path Angle

Performance

Adjusted Output Metric

0.111 ml/s/m2 1.47m & 15° Height & Path Angle

Stepped Evaluation [6]

Decrease Height & Path Angle

18.2°

High Performance Attributes

Graph Key :

Median Angle Original Attribute Data

New Iteration Data

Improved Performance

45° Performance

The stepped strategy was determined to be a high performing strategy in iteration 1. The change of the strategy’s attributes produces a differing output metric. These altering attributes either improve or retain the output metric, in this case, the best performing iterating of the experiment yet is attained. This will be factored into the breeding process.

Adjusted Attribute Metric

0.313 ml/s/m2 45°

Median Angle

Decrease Median Angle

Same Performance

7.41m

Path Length

Performance

0.313ml/s/m2

0.08 ml/s/m2 19.4m Path Length

Increase Path Length

Improved Performance

1.83m

Height & Path Angle

Performance

Adjusted Output Metric

0.111 ml/s/m2 1.47m & 15° Height & Path Angle

Stepped Evaluation [6]

Decrease Height & Path Angle

18.2°

High Performance Attributes

Graph Key :

Median Angle Original Attribute Data

New Iteration Data

Improved Performance

45° Performance

Adjusted Attribute Metric

0.313 ml/s/m2 45°

Median Angle

Decrease Median Angle

Same Performance

20.0m

Path Length

Performance

0.093 ml/s/m2

0.20 ml/s/m2 23.2m Path Length

Increase Path Length

2.29m

Adjusted Output Metric

Performance

Height & Path Angle

0.098 ml/s/m2 3.23m & 8째 Height & Path Length

Spiral Evaluation [6]

Increase Height & Path Angle

5.92 째

High Performance Attributes

Graph Key :

Original Attribute Data

New Iteration Data

Adjusted Attribute Metric

Similar Performance

9.25 째

Median Angle

Performance

The spiral strategy was the best performing in iteration 1. As seen in the diagram the change of these attributes produces a differing output. In this case the increase of these values results in a worst performance, suggesting the spiral assessed in iteration 1 is already near optimal for its typology. Moving forward this means the properties of the spiral should be adopted without the literal implementation of the typology.

Decreased Performance

0.13 ml/s/m2 23.9째 Median Angle

Increase Median Angle

Decreased Performance

20.0m

Path Length

Performance

0.093 ml/s/m2

0.20 ml/s/m2 23.2m Path Length

Increase Path Length

2.29m

Adjusted Output Metric

Performance

Height & Path Angle

0.098 ml/s/m2 3.23m & 8째 Height & Path Length

Spiral Evaluation [6]

Increase Height & Path Angle

5.92 째

High Performance Attributes

Graph Key :

Original Attribute Data

New Iteration Data

Adjusted Attribute Metric

Similar Performance

9.25 째

Median Angle

Performance

Decreased Performance

0.13 ml/s/m2 23.9째 Median Angle

Increase Median Angle

Decreased Performance

Input Best Attributes

Height

Median Angle

Height

Path Angle

Path Length

Input Primary Attributes

Breeding Strategies : i1 [7] Fusing Attributes After generating the first generation iterations of the strategies, assessing their attributes and iterating to improve performance, the fusion process can begin. As the spiral cannot be improved by adjusting attributes, its primary attributes will be fused instead. The stepped strategy produces the best performing output metric, but it limited by its nature. The meander strategy can perform well when adjusted, and has much room to perform better with further adjustments. As such the meander strategy will be the new primary strategy when fusing.

Fused Strategy Output [7] Iteration 2 : Stepped Meander The resultant strategy from the fusion process resembles a â&#x20AC;&#x2DC;stepped meanderâ&#x20AC;&#x2122;. The attributes listed have been considered in formulating a new high performance strategy. These have been inherited from the meander and spiral strategies, manifesting in the form of a longer path length and a lower median angle. The stepped aesthetic also provides a higher path length to median angle ratio. As the spiral offers only its primary attributes, the path length is inherited. This new strategy becomes the new foundation for the 2nd iteration.

Stepped

Input Best Attributes

Spiral : Cant Be Improved

Spiral

Stepped : New Highest Performing

Meander

Meander : Most Room To Improve

Input Best Attributes

Height

Median Angle

Height

Path Angle

Path Length

Input Primary Attributes

Stepped

Input Best Attributes

Spiral : Cant Be Improved

Spiral

Stepped : New Highest Performing

Meander

Meander : Most Room To Improve

Final Strategy To Implement [9]

0.091ml/s/m2

0.093 ml/s/m2

0.092 ml/s/m2

0.069 ml/s/m2

Stepped Meander

Medium-High Performance Strategy

0.081 ml/s/m2

Flattened Stepped Meander

Curved Stepped Meander

Medium-High Performance Strategy

0.079 ml/s/m2

0.069 ml/s/m2

Tight Stepped Meander

Tight-Maze Stepped Meander

Water Study Iteration 2 [8] Strategy Development To iterate the stepped meander, we have sub-categorised typologies to assess performance. All typologies output high performance output metrics, so the rating system is adjusted to penalise lower performance more. The tight maze strategy produces the highest perofrmance by a significant margin. This is our final strategy, so the next stage is to apply it to our pavilion and test to see if the pavilion perofrmance is improved.

Key :

Low Performance Strategy

High Performance Strategy

Highest Performing Strategy

Mazed Stepped Meander Very High Performance Strategy

Very High Performance Strategy

Highest Performance Strategy

Final Strategy To Implement [9]

0.091ml/s/m2

0.093 ml/s/m2

0.092 ml/s/m2

0.069 ml/s/m2

Stepped Meander

Medium-High Performance Strategy

0.081 ml/s/m2

Flattened Stepped Meander

Curved Stepped Meander

Medium-High Performance Strategy

0.079 ml/s/m2

0.069 ml/s/m2

Tight Stepped Meander

Tight-Maze Stepped Meander

Key :

Low Performance Strategy

High Performance Strategy

Highest Performing Strategy

Mazed Stepped Meander Very High Performance Strategy

Very High Performance Strategy

Highest Performance Strategy

Channelling Fins ‘Steps’

Tight-Maze Stepped Strategy Taking The longest downwards path with the most deviations.

Strategy To Pavilion Translation

Final Smooth Skin

Rigid Section: Untestable

Smooth Section: Testable

The rigid implementation of the strategy results in several block-like masses which is impossible to manipulate to test on. Therefore it must be merged into 1 complete mass for testing.

To manipulate the pavilion for testing its performance it requires smoothing. However when this is carried out, the steps ‘melt’ into the surface, leaving only the fins.

Applying the Highest performing strategy attributes to the pavilion form. The channel route is dictated by the positions of the anchors and peaks.

Optimised Form The form in which we apply the strategy to, which has been derived and optimised in previous sections.

Pavilion Implementation [10] Applying Strategy to Pavilion Now iteration 2 is complete and has outputted a highly performing strategy, we must apply it to the pavilion. To adapt the ‘tight-maze stepped meander’ strategy to function and perform highly for the pavilion, we must take its attributes and apply the same logic to extract channelling routes for the water. There are points which we want for channel water towards (anchors) from all points on the pavilions skin.

Final Integrated Pavilion [10] Modelling Constraints

Finalising the strategy onto the pavilion is a complex process requiring the use of multiple modelling tools. As illustrated above, the rigid manifestation of the strategy is not testable, and so requires editing. This edit successfully merges all elements into one, rendering the pavilion testable. However, by doing this the stairs dissapear, absorbed into the pavilion. This modelling constraint means we have to sacrifice clearly defined steps in order to complete the water channeling studies.

Channelling Fins ‘Steps’

Tight-Maze Stepped Strategy Taking The longest downwards path with the most deviations.

Strategy To Pavilion Translation

Final Smooth Skin

Rigid Section: Untestable

Smooth Section: Testable

The rigid implementation of the strategy results in several block-like masses which is impossible to manipulate to test on. Therefore it must be merged into 1 complete mass for testing.

To manipulate the pavilion for testing its performance it requires smoothing. However when this is carried out, the steps ‘melt’ into the surface, leaving only the fins.

Applying the Highest performing strategy attributes to the pavilion form. The channel route is dictated by the positions of the anchors and peaks.

Optimised Form The form in which we apply the strategy to, which has been derived and optimised in previous sections.

Final Integrated Pavilion [10] Modelling Constraints

Strategised Pavilion - Post-Water Study

Plain Pavilion - Pre-Water Studies

Final Performance : 0.642 ml/s/m2

Final Performance : 0.774 ml/s/m2

Testing Pavilion [11] Assessing Experiment Output Now a strategy has been developed and implemented, it is important to test the result, to determine if the study was a success or not. As expected, the strategised pavilion outperforms the standard catenary form significantly. In addition, considering the modelling constrains of the implemented pavilion, as it was not fully integrated, we may speculate that a genuine â&#x20AC;&#x2DC;steppedâ&#x20AC;&#x2122; meander translation could well be the optimum strategy. To conclude, this study explored many options of water channelling methods, to mitigate flash flooding caused from ground saturation. Considering our goals at the start, this study can be considered an overall success.

Strategised Pavilion - Post-Water Study

Plain Pavilion - Pre-Water Studies

Final Performance : 0.642 ml/s/m2

Final Performance : 0.774 ml/s/m2

13 | Detailing Finalising the Pavilion Structure

1:200 Plan

Section

Base Structural Grid

Voronoi Grid

Karamba3D Analysis

Structural Reduction

Structural Evaluation

Result of the karamba Analysis

Karamba is a design tool used to evaluate and refine the structure of an element, this is particularly useful in none standard implication like the multi-domed structure of the pavilion.

Starting with a standard square structural grid this is then attracted to the areas of high stress in the structure and a voronoi grid is then applied based of the deformed points.

This allows the evaluation of where more or reduced structure is needed.

This system allows for dense structure where it is needed allowing of reduced structure where it could cause more stress than support.

Anchor Points

Base Structural Grid

Voronoi Grid

Karamba3D Analysis

Structural Reduction

Structural Evaluation

Result of the karamba Analysis

Karamba is a design tool used to evaluate and refine the structure of an element, this is particularly useful in none standard implication like the multi-domed structure of the pavilion.

Starting with a standard square structural grid this is then attracted to the areas of high stress in the structure and a voronoi grid is then applied based of the deformed points.

This allows the evaluation of where more or reduced structure is needed.

This system allows for dense structure where it is needed allowing of reduced structure where it could cause more stress than support.

Anchor Points

High Rainfall is collected in the pools Recycled ABS Plastic

Plastic is ground

Small pellets are formed

Pellets are melted

6mm sheets are produced

A sheet is heated above a form

A thin light weight but strong panel is produced

Water leaves the pool at a rate that is equal to the average rainfall of the area feeding the pool per second The top clip allows the external panels to be attached Once the structure in connected to the ring the clip is tightened to hold the structure

Pools & Panelling Envrionmental & Structural Strategy The basis of the pavilions structure comes from cutting flat patterns from recycled aluminium panels. These flat patterns can be folded and joined together the create a composite 3d structure that provided adequate structural support without adding additional weight. Water flows off the surface and is collected in the pools, these pools slow the release of water into the ground water system during heavy rain to be water the covered area would experience under usual weather conditions.

High Rainfall is collected in the pools Recycled ABS Plastic

Plastic is ground

Small pellets are formed

Pellets are melted

6mm sheets are produced

A sheet is heated above a form

A thin light weight but strong panel is produced

Bibliography Allan, S. (2008). From Objects to Fields. Carpo, M. (2012). The digital turn in architecture 1990-2012. Erioli, A. (2015). Intensive Aesthetics / Intensive Surfaces. Hollings, C. S. (1996). Engineering Resilience versus Ecological Resilience. Iain Maxwell, D. P. (2010). Inorganic Speciation. Kendall, A. (2019). Optimised Material Fabrication in the Digital Age. Lavin, S. (2012). Vanishing Point: Sylvia Lavin on the Contemporary Pavilion. Nagy, D. (2017). The Problem of Learning. Reiser, J. (2005). Atlas of Novel Techtonics. Zumpthor, P. (2006). Atmospheres: Architectural Environments - Surrounding Objects. Baan, I., 2015. dezeen - Serpentine Pavilion 2015 by SelgasCano photographed by Iwan Baan. [Online] Available at: https://www.dezeen.com/2015/06/22/serpentine-pavilion-2015-iwan-baan-london-selgascano/ [Accessed 02 10 2019]. Crow, H. +., 2011. dezeen - Serpentine Gallery Pavilion 2011 by Peter Zumthor photographed by Hufton + Crow. [Online] Available at: https://www.dezeen.com/2011/07/06/serpentine-gallery-pavilion-2011-by-peter-zumthor-photographed-by-hufton-crow/ [Accessed 02 10 2019]. Shepherd, J., 2016. Independant - Serpentine Gallery Pavilion 2016: Bjarke Ingelsâ&#x20AC;&#x2122; summer house unveiled. [Online] Available at: https://www.independent.co.uk/arts-entertainment/art/news/serpentine-gallery-pavilion-2016-bjarke-ingels-summer-house-unveiled-a7069446.html [Accessed 02 10 2019]. Stephenson, J., 2017. Area - Serpentine Pavilion 2017. [Online] Available at: https://www.area-arch.it/en/serpentine-pavilion-2017/ [Accessed 02 10 2019]. https://tomwiscombe.com/MOCA-PAVILION https://www.mathsisfun.com/data/standard-deviation.html https://www.theguardian.com/uk-news/2019/jun/12/burst-thames-water-pipe-cuts-off-up-to-100000-london-homes https://www.bbc.co.uk/news/uk-england-london-48529484 https://www.london.gov.uk/sites/default/files/running_out_or_flooded_out_-_londons_water_crisis_by_leonie_cooper_am.pdf https://www.metoffice.gov.uk/binaries/content/assets/metofficegovuk/pdf/weather/learn-about/uk-past-events/interesting/2019/2019_008_july_rainfall.pdf?fbclid=IwAR3CdkCWqpypBre1H4e3qSrT7spETNqQkRMZBuC_ YHKlJW0BvJx_Ubtjr-o https://www.archdaily.com/91273/ad-classics-jewish-museum-berlin-daniel-libeskind/5afa4a49f197cc297f000003-ad-classics-jewish-museum-berlin-daniel-libeskind-photo https://libeskind.com/work/jewish-museum-berlin/ https://www.archdaily.com/91273/ad-classics-jewish-museum-berlin-daniel-libeskind https://www.google.co.uk/imgres?imgurl=https%3A%2F%2Fi.pinimg.com%2Foriginals%2F11%2F7e%2F37%2F117e370b906b5f8f1dcf17cf228411ab.jpg&imgrefurl=https%3A%2F%2Fwww.pinterest.com%2Fpin%2F49349607 1652338554%2F&docid=AvP-9SoICWLebM&tbnid=1-3qnex6JvH2-M%3A&vet=10ahUKEwit8uXjsp7mAhXbi1wKHeaQCGAQMwiGASghMCE..i&w=796&h=1200&hl=en&bih=938&biw=1840&q=jewish%20museum%20berlin%20 inside&ved=0ahUKEwit8uXjsp7mAhXbi1wKHeaQCGAQMwiGASghMCE&iact=mrc&uact=8 https://www.serpentinegalleries.org/explore/pavilion

The Leviathan Located on the site of the serpentine gallery, The Leviathan is a conceptual proposal for the annual serpentine pavilion. â&#x20AC;&#x2DC;Leviathanâ&#x20AC;&#x2122; is Hebrew for sea serpent, which is inspired by two facets of our project: The site & Water. Our proposal directly addresses the continual climate crisis through data driven design, both qualitatively and quantitatively through a computational methodology.