integration of life cycle assessment
TRANSCRIPT
INTEGRATION OF LIFE CYCLE ASSESSMENT
AND CONCEPTUAL BUILDING DESIGN
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF
CIVIL AND ENVIRONMENTAL ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
John Paul Basbagill
December 2013
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/vt683fq0708
© 2013 by John Paul Basbagill. All Rights Reserved.
Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.
ii
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Michael Lepech, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Martin Fischer
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Douglas Noble
Approved for the Stanford University Committee on Graduate Studies.
Patricia J. Gumport, Vice Provost for Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.
iii
iv
Abstract
Conceptual building design involves decisions that have significant life cycle environmental
impact and life cycle cost implications. Designers are aware of the importance of creating
sustainable buildings, yet mechanisms are lacking that effectively inform designers of the
impacts of design decisions, specifically during the conceptual phase. Poor design choices
yielding carbon intensive, costly buildings are often discovered only late in the design process,
when design decisions are difficult and costly to modify.
A method is proposed that provides life cycle environmental impact and cost feedback to
designers during the conceptual design phase in a way that better enables designers to make
decisions leading to less carbon intensive and less costly buildings. Critical to the method is the
development of a set of heuristics which calculate the embodied impacts and costs of a full range
of building components. These heuristics require only three inputs typically known during the
conceptual design stage: gross floor area, building location, and building type. Multi-disciplinary
design optimization is then targeted as a point of departure for integrating the heuristics with
building design feedback using an automated approach. The heuristics are integrated specifically
with building information modeling, life cycle assessment, life cycle cost, and energy simulation
software and an automated feedback processor. The processor utilizes an algorithm to generate
life cycle impact feedback on many design alternatives, thereby allowing designers to understand
the full range of impacts possible for a given problem formulation. The method is also
configured to provide feedback specifically for design decisions made sequentially, as is typical
of the building design process, as opposed to providing feedback after all decisions have been
made, a limitation of design processes currently using multi-disciplinary design optimization. In
this way, designers understand the life cycle environmental impact and life cycle cost
v
implications of each decision for a range of building design alternatives and can easily modify
decisions in a way that better aligns with their desired performance objectives.
In summary, the proposed method’s primary contributions to knowledge are the reduction
of inputs required for life cycle assessment during the conceptual design stage to only three
inputs, the formalization of heuristics corresponding to the three inputs that approximate the life
cycle impact of conceptual design decisions, the integration of heuristics with automation in a
way that allows for rapid exploration of the full design space, and the sequential presentation of
life cycle environmental impact and life cycle cost feedback on conceptual building design
decisions.
vi
Acknowledgments
Gratitude is expressed to the numerous individuals and groups who have supported me
over the past four years. I would like to thank my family, including my parents, Pat and Paul
Basbagill, my sister, Marie Clark, my brother-in-law, Rob Clark, my nephew, Robby Clark, and
my niece and god-daughter, Emma Clark. I thank my advisors, Michael Lepech, Martin Fischer,
and Doug Noble, each a source of guidance and inspiration during my graduate studies at
Stanford University and the University of Southern California. I wish to thank the Leavell
Fellowship and the Center for Integrated Facility Engineering for supporting my research. I also
thank my friends and colleagues who have guided me and given me critical feedback throughout
the research process, in particular Forest Flager and members of my research group.
vii
Table of Contents
Abstract………………………………………………...…………………………………............iv
Acknowledgments…………………………………………..……………………………………vi
Table of Contents…………………………………………………………………………...…....vii
List of Figures……………………………………………………………….............……….......xii
List of Tables………………………………………………………………………….……......xvii
List of Appendices……………………………………………………………………………....xix
Chapter 1: Introduction………………………………………………………………………....…1
Observed Problem………………………………………………………………………...……1
Integration of Building Design and Performance Feedback…………………………...……3
Conceptual Design Stage…………………………………………………………...……….4
Life Cycle Assessment and Life Cycle Cost………………………...…………...………….6
Parametric Design……………………………………………………………..…………….9
Integration of Multi-Disciplinary Design Optimization and Sequential
Design Decision-Making Processes………………………………………………...…..13
Survey of Current Use of Life Cycle Assessment in the Building Design Industry…….....14
Defining the Theoretical Gap between LCA and LCC Feedback and Building Design.15
Defining the Practical Gap between LCA and LCC Feedback and Building Design.…17
Part I: Understanding General Trends of LCA and LCC Feedback and Design...…18
Part II: Understanding Firm-Specific Details of LCA and LCC Feedback and
Design……………………………………………………………………………21
Arup…………………………………………………………………………...…22
Atelier Ten……………………………………………………………………….24
Kohn Pedersen Fox………………………………………………………………26
viii
Summary of Key Themes……………………………………………………..…27
Reflection on Use of Life Cycle Assessment in Building Design……………….…...………28
Research Questions……………………………………………………………………..…….30
Organization of the Dissertation…………………………………………………………..….31
Chapter 2: Scope, Model Development, and Validation………………………………..……….33
Scope of the Method…………………………………………………………………..……...34
Building Life Cycle Phases……………………………………………………...………...34
Environmental Impact Indicators…………………………………………………….……35
Building Components, Materials, and Dimensions……………………………………….35
Software Integration Method…………………………………………………………………38
Embodied Impact Calculation……………………………………………………………..40
Operational Impact Calculation…………………………………………………………...42
Feedback Processor………………………………………………………………………..43
Software Used……………………………………………………………………………..43
Validation of Embodied Impact Heuristics…………………………………………….……..44
Chapter 3: Application of Life Cycle Assessment to Early Stage Building Design for Reduced
Embodied Environmental Impacts……………………………………………….…………...51
Abstract……………………………………………………………………………………….51
Introduction…………………………………………………………………………………...52
Literature Review……………………………………………………………………………..55
Methodology………………………………………………………………………………….57
Scope………………………………………………………………………………………57
Building Component Classification Framework………………………………………….60
Analysis Process…………………………………………………………………………..62
ix
Implementation……………………………………………………………………………….66
Problem Formulation……………………………………………………………………...66
Software Integration……………………………………………………………………….68
Results and Discussion…………………………………………………………………….…72
Conclusions…………………………………………………………………………………...78
Chapter 4: Evaluating Embodied Versus Operational Environmental Impact Trade-offs of
Conceptual Building Designs………………………………………………………………...80
Abstract……………………………………………………………………………………….80
Introduction…………………………………………………………………………………...81
Related Studies………………………………………………………………………………..84
Methodology………………………………………………………………………………….86
Case Study……………………………………………………………………………………91
Problem Formulation……………………………………………………………………...91
Results……………………………………………………………………………………..93
Comparison of Optimization Objectives……………………………………………....93
Analysis of Trade-off Variables………………………………………………………..97
Sensitivity Analysis…………………………………………………………………………103
Cladding Material………………………………………………………………………..104
Climate…………………………………………………………………………………...110
Building Size……………………………………………………………………………..117
Conclusions………………………………………………………………………………….124
Chapter 5: A Methodology for Providing Environmental Impact Feedback on Sequential
Conceptual Building Design Decisions……………………………………………………..126
Abstract……………………………………………………………………………………...126
Introduction………………………………………………………………………………….127
x
Related Studies………………………………………………………………………………131
Methodology…………………………………………………………………………….......133
Case Study…………………………………………………………………………………..138
Results……………………………………………………………………………………….140
Sequential Decision-Making Approach I: Minimization of Total Environmental
Impact………………………………………………………………………………...142
Sequential Decision-Making Approach II: Achievement of Carbon Performance
Value………………………………………………………………………………….144
Sequential Decision-Making Approach III: Maximization of Design Freedom………...146
Validation……………………………………………………………………………………148
Conclusions………………………………………………………………………………….150
Chapter 6: A Multi-Objective Feedback Approach to Evaluating Sequential Building Design
Decisions…………………………………………………………………………………….153
Abstract…………………………………………………………………………………...…153
Introduction………………………………………………………………………………….154
Related Studies………………………………………………………………………………158
Methodology………………………………………………………………………………...160
Scope……………………………………………………………………………………..160
Analysis Process…………………………………………………………………………162
Inspection of Results……………………………………………………………………..166
Software Implementation…………………………………………………………………167
Case Study…………………………………………………………………………………..167
Results…………………………………………………………………………………….....169
Conclusions………………………………………………………………………………….180
Chapter 7: Conclusions………………………………………………………………………....183
xi
Summary of Conclusions from Chapters 3 through 6………………………………………183
Contributions………………………………………………………………………..............185
Embodied Impact Heuristics……………………………………………………………..186
Integration of Automated Feedback and Sequential Decisions……………………….....188
Range of Control of Building Performance Alternatives………………………………..190
Answers to Research Questions…………………………………………………………......191
Challenges and Recommendations………………………………………………………….193
Limitations and Future Work……………………………………………………………..…194
Appendices……………………………………………………………………………………...198
References…………………………………………………………………………....................210
xii
List of Figures
Figure 1 – Design completion and uncertainty from pre-design through construction of a building
project (Struck and Hensen 2007)………………………………………………………...…....6
Figure 2 – Frequency that designers perform various sustainable design strategies on building
projects……………………………………………………………………………………..…19
Figure 3 – Stages during which designers make various building decisions compared with
American Institute of Architects guidelines……………………………………………….....20
Figure 4 – Barriers inhibiting design firms’ use of LCA on building projects…………………..21
Figure 5 – Building life cycle phases included in the BIM-enabled LCA-LCC feedback
method………………………………………………………………….……………………..35
Figure 6 – Architectural Design And Performance Tool (ADAPT): software integration tool
providing life cycle embodied impact, operational impact, and cost feedback on conceptual
building designs……………………………………………………………………………....38
Figure 7 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 1,358-m2 building …………..…………………………………...…….47
Figure 8 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 2,500-m2 building …………..……………………...……………...…..47
Figure 9 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 2,900-m2 building …………………………………………..…..……..48
Figure 10 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 8,458-m2 building……………..………………………..……….....…..48
Figure 11 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 22,982-m2 building…………………………………...……..………....49
xiii
Figure 12 - Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 34,910-m2 building…………………………………………...…….….49
Figure 13 - Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 47,250-m2 building………………………………………………....….50
Figure 14 - Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 85,000-m2 building…………………………………………...…….….50
Figure 15 – Building life cycle phases included in scope…..…………………………………....59
Figure 16 – Software integration for embodied impact feedback method…………….....………62
Figure 17 – Embodied impact reduction due to material decisions……………………………...76
Figure 18 – Embodied impact reduction due to thickness decisions………………………….....77
Figure 19 – Building life cycle phases included in scope………………………………......……87
Figure 20 – Software integration for optimized life cycle environmental impact feedback…….89
Figure 21 – Distribution of optimized design configurations………………………………...….94
Figure 22 – Distribution of life cycle environmental impacts……...……………………………95
Figure 23 – Distribution of window-to-wall ratio values………………………………………..98
Figure 24 – Distribution of glazing thickness values…………………………………………….99
Figure 25 – Distribution of presence of fins values……………………..……………………...100
Figure 26 – Distribution of presence of overhangs values………………..……………………101
Figure 27 – Distribution of fin depth values……………………………………...…………….102
Figure 28 – Distribution of overhang depth values…………………..……………………...…102
Figure 29 – Distribution of optimized design configurations for alternate cladding material….105
Figure 30 – Distribution of life cycle environmental impacts for alternate cladding material…106
Figure 31 – Distribution of window-to-wall ratio values for alternate cladding material…...…107
xiv
Figure 32 – Distribution of glazing thickness values for alternate cladding material………….108
Figure 33 – Distribution of presence of fins values for alternate cladding material……………108
Figure 34 – Distribution of presence of overhangs values for alternate cladding material…….109
Figure 35 – Distribution of fin depth values for alternate cladding material…………………...109
Figure 36 – Distribution of overhang depth values for alternate cladding material……………110
Figure 37 – Distribution of optimized design configurations for alternate climate……….…....112
Figure 38 – Distribution of life cycle environmental impacts for alternate climate….……...…113
Figure 39 – Distribution of window-to-wall ratio values for alternate climate…………..…….114
Figure 40 – Distribution of glazing thickness values for alternate climate………………….…115
Figure 41 – Distribution of presence of fins values for alternate climate……………………....115
Figure 42 – Distribution of presence of overhangs values for alternate climate…………….....116
Figure 43 – Distribution of fin depth values for alternate climate……………………….…..…116
Figure 44 – Distribution of overhang depth values for alternate climate……………………....117
Figure 45 – Distribution of optimized design configurations for alternate building size……....119
Figure 46 – Distribution of life cycle environmental impacts for alternate building size…...…119
Figure 47 – Distribution of window-to-wall ratio values for alternate building size……….….121
Figure 48 – Distribution of glazing thickness values for alternate building size………..….…..121
Figure 49 – Distribution of presence of fins values for alternate building size………...……....122
Figure 50 – Distribution of presence of overhangs values for alternate building size……........122
Figure 51 – Distribution of fin depth values for alternate building size………………......……123
Figure 52 – Distribution of overhang depth values for alternate building size………….…......123
Figure 53 – Three sequential decision-making approaches to which the environmental
impact feedback method may apply: (a) minimization of carbon footprint,
xv
(b) achievement of a carbon target value, and (c) maintenance of freedom and
flexibility…………………………………………………………………………………….130
Figure 54 – Building life cycle phases included in proposed method for providing environmental
impact feedback on sequential design decisions………….……..………………………..…134
Figure 55 – Three design alternatives generated by the building information modeling software
showing variations in several input parameters……………………………………………..135
Figure 56 – Method for providing probabilistic environmental impact feedback on sequential
building designs……………………………………………………………………………..137
Figure 57 – Probability mass function characterizing a design space size of 3.69x1023, showing
total environmental impacts for 8,689 selected designs prior to any design decisions……..141
Figure 58 – Probability impact distributions for four sequential decisions for the objective
minimizing total environmental impact.………………………………………...……...…...143
Figure 59 – Impact distributions for first four decisions for the objective achieving a carbon
performance value.………………………………………………………………...………...145
Figure 60 – Impact distributions for first four decisions for the objective maximizing
design freedom………………………………………...…………………….…………...….147
Figure 61 – Distribution of life cycle environmental impacts for alternate Latin hypercube
sampling algorithm…………………………………….……………………………..……..149
Figure 62 – Combined distribution of life cycle environmental impacts for original and
alternate sampling algorithms.…………………………………………………....………....150
Figure 63 – Three sequential decision-making design strategies to which designers might apply
the multi-objective feedback method: (a) minimization of carbon footprint,
(b) achievement of an environmental impact performance target, and (c) maintenance of
xvi
design freedom……………………………………………………………..……….……….156
Figure 64 – Building life cycle phases included in proposed method for providing multi-
objective feedback on sequential design decisions……………………………………….…162
Figure 65 – Automated method for providing life cycle environmental impact and life cycle
cost feedback on sequential building design decisions……………………...………………163
Figure 66 – Three design alternatives generated by the building information modeling
software showing variations in several input parameters………………………….………..164
Figure 67 – Distribution of building life cycle environmental impacts and life cycle costs
for a design space size of 6.07x1016
…………………….………………………….………171
Figure 68 – Distribution of life cycle environmental impacts and life cycle costs
after decision 1: number of buildings equals 3……………………………...………………173
Figure 69 – Distribution of life cycle environmental impacts and life cycle costs
after decision 2: low e-glazing……………………………...…………………………….…174
Figure 70 – Distribution of life cycle environmental impacts and life cycle costs
after decision 3: window-to-wall ratio equals 50……………………………………………176
Figure 71 – Distribution of life cycle environmental impacts and life cycle costs
after decision 3 (revised): window-to-wall ratio equals 15………………………..…..……177
Figure 72 – Distribution of life cycle environmental impacts and life cycle costs
after decision 4: orientation from 0° to 180°.……………………………………………….179
xvii
List of Tables
Table 1 – Building component classification framework………………….……..……………...37
Table 2 – Required inputs, variables, and assumptions for BIM-enabled automated feedback
method for life cycle assessment and life cycle cost…………………………………..……..39
Table 3 – Comparison of embodied impact values for eight building case studies: ADAPT
versus Arup (2013)……………………………………………...………………………....…45
Table 4 – Building component classification framework………...……………………………...61
Table 5 – Required inputs, variables, and assumptions for building information modeling-
enabled embodied impact feedback method………………………………………………….63
Table 6 – Problem formulation showing required inputs, variable values, and assumptions…....68
Table 7 – Impact allocation scheme and impact reduction scheme……………………………...74
Table 8 – Ranking scheme for material decisions achieving embodied impact reductions…......75
Table 9 – Ranking scheme for thickness decisions achieving embodied impact reductions…….75
Table 10 – Optimization problem formulation describing objectives and variable values……...92
Table 11 – Material, dimensional, and impact assumptions for building components with
embodied versus operational impact trade-offs………………………………………………93
Table 12 – Comparison of trade-off variable values and carbon impacts……………………….97
Table 13 – Comparison of trade-off variable values and carbon impacts for alternate
cladding material…………………..………………………………………………………...106
Table 14 – Comparison of trade-off variable values and carbon impacts for alternate
climate………………….………………………………………………………….……...…113
Table 15 – Comparison of trade-off variable values and carbon impacts for alternate
building size…………………………….…………………………………………………...120
xviii
Table 16 – Case study variables used to characterize building life cycle environmental
impacts………………………………………………………………………………………140
Table 17 – Metrics characterizing the design space for a design strategy minimizing
environmental impacts…………………………………….………………………………...144
Table 18 – Metrics characterizing the design space for a design strategy achieving a
performance value……………………………...………………………………………........146
Table 19 – Metrics characterizing the design space for a design strategy maximizing
design freedom………………………………………………...…………………………….148
Table 20 – Validation of impact distributions using alternate sampling method………………150
Table 21 – Required inputs and variables for automated life cycle environmental impact
and life cycle cost feedback method…………………………...……………………………168
Table 22 – Assumptions for automated life cycle environmental impact and life cycle cost
feedback method…………………………………………………………………………….168
Table 23 – Metrics characterizing the design space for two design strategies:
(a) achieving a life cycle environmental impact performance value and (b) minimizing life
cycle cost………………………………….………………………………………...……….180
xix
List of Appendices
Appendix 1 – Supporting Data for Chapter 1……………………………………………..……198
Survey Questions Provided to Design Firms to Gauge Trends in Use of LCA and LCC
Feedback in Design………………………………………………………………………199
Appendix 2 – Supporting Data for Chapters 2 and 3………………………………..……….....199
Material Alternatives Considered in Quantifying Embodied Impacts of Building
Components…………………………………………………………………………...…199
Material(s) Associated with Each Building Component and Material Properties Used to
Quantify Building Component Embodied Impacts……………………………………....200
Embodied Impact Heuristics Developed for Each Building Component…………………...201
Appendix 3 – Supporting Data for Chapter 5………………………………………..…………204
Variables and Variable Values Used as Inputs for Environmental Impact Feedback
Method…………………………………………………………………………………...204
Appendix 4 – Supporting Data for Chapter 6………………………………………..…………206
Variables and Variable Values Used as Inputs for Environmental Impact and Cost
Feedback………………………………………………………………………………....206
Sample Cost Formulas……………………………………………………………………….208
1
Chapter 1: Introduction
Observed Problem
Buildings consume considerable amounts of energy and materials and have significant
economic impacts. They account for 41% of the total energy consumption in the United States
(Fumo et al. 2010) and heavily contribute to greenhouse gas emissions levels (Sharma et al.
2011). In 2007, new residential and commercial construction expenditures in the United States
totaled $759 billion, and in 2010 the energy to operate new facilities in the United States
totaled over $431 billion (USDOE 2011a). Buildings also account for 72% of total electricity
consumption in the United States (USDOE 2011a), due for the most part to energy use during the
occupancy phase (Radhi 2010). Energy consumption during this phase typically accounts for 80-
85% of a building’s total life cycle energy (Adalberth et al. 2001). Embodied environmental
impacts generated by a building during its life cycle may also be significant (Fay et al. 2000,
Bribian et al. 2009) and, in cases where buildings have been designed for low- or net-zero
energy, can approach use phase impacts (Citherlet 2001, Thormark 2002, Winther and Hestnes
1999). Building lifetimes extend for many decades, and their environmental and economic
impacts occur at global, local, and indoor scales, suggesting the importance of considering
buildings’ entire life cycle when reducing their impacts (Tucker et al. 2003).
These impacts arise because buildings suffer from a host of poor performance issues.
Inefficient operation of building systems is common, such as wasteful conditioning of air, poor
air-tightness, lack of energy management, and a tendency for systems to default to “on” (Bordass
et al. 2001). Poor performance is also related to design flaws, such as redundant structural
elements, inefficient planning and circulation, ineffective shading devices, and other design
2
features that yield little reductions in energy use (Fay et al. 2000). Embodied energy
performance of buildings, which involves the energy required to mine raw materials,
manufacture building components, transport the components, and construct and demolish
buildings, may also be poor (Lawson et al. 1995). Strategies for reducing embodied impacts
include substituting low energy intensity materials for high energy intensity materials, reducing
construction waste, and using materials with a high recycled content (Fay et al. 2000).
In part, these performance issues are due to the lack of feedback designers receive,
particularly during the critical early stages of the design process (Chaszar 2003, Schlueter and
Thesseling 2009). Designers make many design decisions at this stage, and tools that provide
quantitative assessment of the impacts of these decisions are lacking. As a result, designers are
not fully informed of the life cycle impact implications of their design decisions, and this may
yield buildings with large carbon and cost footprints.
Fortunately, the architecture, engineering and construction (AEC) industry has shown
interest over the past few decades in improving the performance of buildings from a life cycle
impact standpoint (Bribian et al. 2009), in large part due to the many buildings that are designed,
built, renovated, and demolished each year. The total building stock in the United States is
approximately 300 billion square feet, and each year developers construct approximately five
billion square feet, demolish 1.75 billion square feet, and renovate five billion square feet of
commercial, residential, industrial, and institutional building space (Winter 2008). By the year
2035, approximately 75% of the built environment will be either new or renovated (USEIA
2010). Therefore, a tremendous opportunity exists for devising methods that provide designers
feedback on the life cycle performance of their building projects. The hope is that designers will
use this feedback to design less costly buildings with lower global warming potential, thereby
3
contributing towards the creation of a healthier and more sustainable global environment. The
challenge is to create a method that integrates life cycle assessment feedback – which typically
requires highly detailed inputs and may be difficult to interpret – with building design decisions
quickly, easily, and intuitively during the critical earliest stages of the design process.
Integration of building design and performance feedback
Large numbers of software tools integrating design with building performance have
existed in the AEC industry for over 30 years, and their shortcomings have been well
documented (Norton et al.1995, Augenbroe 1992, Augenbroe and Winkelmann 1990,
Papamichael 1991). The tools have been described as technically too complex for early stage
design, which has led to limited or ineffective use (van Hattem 1987). This is surprising, given
that architects’ involvement on projects typically precedes the involvement of other specialists,
such as lighting designers and mechanical, electrical, and plumbing engineers, and they would
therefore benefit from simple, intuitive feedback tools. Instead, the tools lack a designer-friendly
interface and have been more suitable for those with specialized knowledge of lighting,
structural, HVAC, or other building systems. In addition, the uniqueness and complexity of
building projects has prevented the creation of simplified design-feedback tools in the AEC
industry. This contrasts with such industries as aerospace or automotive, in which uniformity in
design of component parts has allowed design-feedback tools to more easily gain adoption
(AIAA 1991). Another reason performance feedback tools have lacked the ability to help create
low carbon, low cost buildings is their lack of analytical breadth: no one tool has integrated many
building performance aspects. Instead, each tool typically focuses on a single aspect of
improving building performance, yet often these aspects are related. In terms of environmental
4
performance feedback, the tools also typically quantify operational energy but often ignore
embodied energy considerations.
The US Department of Energy website currently lists 411 building software tools for
evaluating energy efficiency, each varying in terms of complexity and breadth (USDOE 2011b).
Yet only a few tools account for impacts due to both embodied and operational energy. Many of
the tools require detailed inputs and are more appropriate for use after the conceptual design
stage. This is problematic, since many building design parameters are determined by the late
design stages, and changes can be costly (Morbitzer et al. 2001). Designers have fewer design
alternatives to consider and designers’ role in the creation of sustainable buildings is weaker than
if tools existed that provided life cycle impact feedback specifically during the conceptual design
stage.
Conceptual design stage
The conceptual stage of the building design process has been recognized as a critical
stage for determining a building project’s life cycle impacts (Yohanis and Norton 2006).
Decisions made in early stages should be attended to with the most care, since early decisions –
such as shape, orientation, or distribution of glazing – largely determine a building’s
environmental impact or cost (Azhar et al. 2010). Unfortunately, energy simulation analysis is
often relegated to the post-design process (Kienzl et al. 2003), even though changing a building’s
shape, orientation, or envelope configuration during the early design stages has been shown to
reduce energy consumption by 30-40% at no extra cost (Baker and Steemers 2000, Cofaigh et al.
1999). Massing changes made in late design stages have also been shown to have considerable
environmental ramifications, but at high economic costs (Ellis et al. 2008).
5
Successful integration of life cycle feedback and building design requires designers and
engineers to collaborate during the conceptual design stage (Bribian et al. 2009), and this may
include owners, contractors, suppliers, building users, and design professionals (Alsamsam et al.
2008). This effort can be difficult to coordinate. In traditional architectural workflow, building
performance assessment is conducted by engineers or consultants subsequent to the design
(Schlueter and Thesseling 2009). This results in fragmented interactions and unnecessary lag
time between early design exploration and environmental feedback (Turpin 2007). Decisions
made in the early design stages are uninformed and result in potentially poor building
performance. Projects are divided into separate phases performed by different teams,
communication is often poor within the teams, exchange of information is limited, and when
information is dispersed it is often inaccurate and not current (Ospina-Alvarado and Castro-
Lacouture 2010). The transfer of information is cumbersome and prone to delays, and designers
often backtrack in their work, slowing design process momentum.
Another problem with current architectural workflow is that designers typically
investigate only a limited number of options during the early design stages before selecting one
for detailed consideration. Designers often fail to consider a breadth of options for a given
problem configuration, and this approach to conceptual design has been called an “exploitation”
rather than an “exploration” of design options (Grierson and Khajehpour 2002, p. 83). The lack
of consideration of many design alternatives is unfortunate, given the vast number of options
typically possible during the conceptual phase due to the openness of the problem definition.
Decisions are pushed to later design stages as a result of this lack of conceptual design feedback,
and this is likely to increase project impacts. The result is the creation of a building that is more
costly, both environmentally and economically, than if life cycle impact feedback on many
6
alternatives had been available at the conceptual design stage (Grierson and Khajehpour 2002).
Figure 1 describes the stages of the building design process in terms of uncertainty.
Opportunities for improving a building’s performance are greatest at the conceptual design stage,
when designers have the least understanding of, yet greatest ability to manipulate, the building
form. A large number of design scenarios are available to designers, and decisions made at the
early stages have the most significant environmental impact or cost implications. When these
decisions are made without performance feedback or deferred to later design stages,
environmental and cost impacts have the potential to be much greater (Schlueter and Thesseling
2009).
Figure 1 – Design completion and uncertainty from pre-design through construction of a building
project (Struck and Hensen 2007).
Life cycle assessment and life cycle cost
Efforts to improve building performance should consider the entire life cycle energy
requirements and costs of a building. Much research has focused on the operational phase, as
shown in the many simulation tools available on the US Department of Energy’s website.
Embodied energy has been largely neglected, even though embodied energy can comprise up to
7
40% of a building’s total impact (Winther and Hestnes 1999). In fact, research has shown that
integration of embodied impact feedback with design tools can be important to the creation of
sustainable architecture (Sandrolini and Franzoni 2009, Eštoková et al. 2011).
Life cycle assessment (LCA) and life cycle cost (LCC) are methods that account for a
building’s environmental, social, and economic impacts over its entire life cycle. LCA and LCC
have been recognized as important methods for improving a building’s energy and cost
performance during the early design stages (Bribian et al. 2009). Indicators are used to describe
where along an impact pathway a category has relevance: midpoint indicators relate to the
environmental mechanism, whereas endpoint indicators pertain to human health and natural
resources (LC-Impact 2013). Midpoint indicators used in LCA include global warming potential,
acidification potential, and carcinogens. Quantifying many building designs’ environmental
impacts using these indicators can help designers understand how a building compares to other
designs in terms of life cycle environmental impact performance.
The application of LCA and LCC methods to buildings is challenging. Reliable and
organized data on building materials and components is difficult to obtain, especially in the
United States (Yohanis and Norton 2006, Wang et al. 2005b). Material flows during building
construction, including purchasing, site storage, materials control, and wastage control, are
difficult to track, and construction waste is difficult to estimate (Yohanis and Norton 2006).
Accounting for energy used in the transportation of building materials can be especially
complex, since the country of origin of building materials may vary as designers modify material
choices for such building components as cladding, glazing, and structural members. Precise
material specifications for building components are typically unknown at the conceptual design
stage, yet current methods of performing LCAs and LCCs of conceptual building designs require
8
a high level of detail of information (Morbitzer et al. 2001).
Bribian et al. (2009) describe additional challenges to conducting LCA of buildings. LCA
was initially developed for designing products with low environmental impacts. In comparison to
most products, however, buildings have a relatively long life span, they often undergo
programmatic changes, spaces may have multiple functions, they contain many different
materials and components, they are typically unique, and system boundaries may be unclear. As
a result, performing LCA of buildings is less straightforward and more complex than for many
products. Presentation of LCA results to building designers in an easily understandable format is
another challenge.
The integration method described in chapter 2 does not attempt to address all of the
challenges of integrating LCA with conceptual building design. For example, chapter 2 describes
how the method assumes certain system boundaries, a building life span, and a building program;
the research does not intend to generalize broadly across these variables. The LCA data used by
the method also originates from two life cycle inventory databases, SimaPro and Athena, each of
which has a certain degree of uncertainty (SimaPro 2010, Athena 2011); the research is limited
in scope to this data. The method addresses other problems, in particular the fact that current
LCA methods require a high level of detail when evaluating building designs and therefore are
more suitable at later building design stages; do not incorporate sensitivity analysis of building
design parameters; and do not leverage automated processes to compare an initial design with
many design alternatives.
9
Parametric design
Parametric design tools are well suited for integration with LCA and LCC feedback of
buildings at the early design stages. Parametric design is a method of linking variables to
building geometry in such a way that when one parameter’s value changes, values for the linked
parameters automatically update. By defining dependencies between such parameters as
geometry, orientation, façade composition, and building materials, designers can significantly
reduce manual design modifications and quickly receive many design alternatives. Parametric
design’s integration with performance-based tools such as structural optimization (Shea et al.
2005), energy simulation (Schlueter and Thesseling 2009), and life cycle assessment and life
cycle costing (Wang et al. 2005a, Wang et al. 2005b) has been shown to rapidly increase the
number of analysis feedback loops, thereby allowing designers to evaluate many design
alternatives for a range of variables. These analyses have optimized a building’s performance in
terms of reduced structural weight, operational energy, or life cycle cost, resulting in lighter,
more energy efficient, and cheaper buildings.
The work of Wang et al. (2005b) is described as an example of how parametric design
and rapid generation of building design alternatives can be used to improve building
performance. The study presented a multi-objective optimization model, in which a genetic
algorithm varied eight conceptual design parameters in order to optimize life cycle cost and life
cycle environmental impact. By iterating across many design alternatives, the algorithm
generated a Pareto front defined by 29 non-dominated solutions, many of which improved upon
both the initial set of designs’ LCC and life cycle environmental impact. In terms of variable
performance, steel-frame walls were the cheapest option, and masonry walls achieved the lowest
environmental impact performance. The optimal building orientation converged to zero degrees
10
and the window-to-wall ratio converged to 20%, meaning the highest-performing designs
consistently contained these variable values. On the other hand, the building’s aspect ratio did
not converge to a single value, meaning several different shapes could be found in high-
performing designs. This example shows that parametric tools may be integrated with
optimization algorithms, thereby increasing the number of high-performing design options. By
specifying ranges for values of parameters, then combinatorially selecting from these values,
optimization algorithms can potentially evaluate millions of building designs. For example, a
designer may wish to determine which combination of building orientation and cladding material
yields the lowest life cycle environmental impact and cost. As Wang et al. (2005b) showed, an
optimization algorithm can evaluate all possible orientation angles and cladding materials, then
determine which combination(s) result in low values for both objectives. The method iterates
through many design possibilities in order to consider the environmental impact and cost
tradeoffs.
The example also illustrates the use of multi-disciplinary design optimization (MDO),
which is a method of evaluating many parametrically generated design alternatives when a
tradeoff exists between competing objectives, such as life cycle environmental performance and
cost. Pareto, or non-dominated, optimization performs many feedback loops in order to identify a
field of conceptual designs that are equal-rank optimal. The solution space is thoroughly
explored, and the designer is provided with a number of solutions that represent the set that best
satisfy both objectives. Increasing the number of feedback loops in this way can also show
designers relationships between the competing objectives, ranges of parameter values that
produce viable solutions, and the sensitivities of each of the parameters. Studies have used
MDO to computationally evaluate the tradeoffs between building materials’ thermal versus
11
lighting performance (Caldas 2008), buildings’ first cost versus operational energy cost
(Diakaki et al. 2008), and facades’ heat gains versus cost (Peippo et al. 1999).
Genetic algorithms are well suited for solving building MDO problems and have been
applied to many MDO studies optimizing environmental impact and cost performance (Wang
et al. 2005a, Wang et al. 2005b, Wang et al. 2006, Magnier and Haghighat 2010, Juan et al.
2010, Charron and Athienitis 2006, Geyer 2009). A distinguishing feature from other
optimization methods is that genetic algorithms operate on a population of solutions, which are
randomly generated in order to yield the first solution. This feature lends itself well to MDO
problems, since genetic algorithms can locate multiple Pareto optimal solutions in a single run.
Genetic algorithms also lend themselves well to single criterion building optimization problems
(Wright and Farmani 2001). Each individual in the population, or a chromosome, represents a
potential solution in the problem space. The chromosome is typically represented as a binary
string that can capture both discrete and continuous variables. This is advantageous for building
energy performance optimization problems, in which variables are both discrete (e.g., building
materials) and continuous (e.g., overhang angle). Conventional gradient-based optimization
methods, on the other hand, depend on initial guess values and are prone to being trapped in local
extrema (Deb 2001). Building optimization problems are often nonlinear, which leads to
discontinuous outputs, and gradient-based methods can only be applied to smooth and
continuous functions (Wetter and Wright 2004). Because genetic algorithms use a non-
dominated strategy to sample the solution space at many different points, they are well suited for
solving building energy optimization problems with many local minima (Coley and Schukat
2002). Genetic algorithms also process large quantities of data efficiently and can identify
solutions quickly (Rafiq et al. 2003), which is beneficial for building optimization problems with
12
complex search spaces. Despite genetic algorithms’ potential to optimize conceptual building
designs, only a few studies have done so for minimized life cycle environmental impacts and life
cycle costs (Wang et al. 2005a, Wang et al. 2005b, Wang et al. 2006). Other life cycle impact
building design optimization studies required detailed building inputs, typically known only at
late design stages, and focused on operational energy (Al-Homoud 1997, Wetter 2001, Coley and
Schukat 2002). A method is needed for integrating conceptual building design and life cycle
impact tools with a minimal number of inputs, so that designers can utilize the tool as early as
possible in the design process.
Sampling algorithms can be integrated with parametric design tools to show designers the
full range of outputs possible for a given set of discrete input parameters. Probability mass
functions, or functions describing the probability that a variable will achieve each of its discrete
values, can be used to show the range of impacts possible for a given problem configuration. For
example, de Wit and Augenbroe (2002) generated a probability mass function to show the range
of thermal comfort options possible for a set of input parameters related to wind, temperature,
and solar transmission factors. In the example presented at the beginning of this section, Wang et
al. (2005b) discovered that one single value for building orientation and window-to-wall ratio is
found in all high-performing designs across all objectives, whereas certain values for wall type
perform well for one objective but not another, and values for other variables such as aspect ratio
exhibit no strong trend among high-performing designs. However the graphical results of the
study did not visually display these relationships very clearly. Probability mass functions can
clarify these relationships by showing quickly and easily which variable values are consistently
present in high-performing designs, which values are likely to be found in high-performing
designs for one objective but not other objectives, and which values exhibit no consistent
13
presence in high- or low-performing designs. Studies are lacking that show how automated
processes and probability mass functions can be leveraged in these ways to provide sensitivity
analysis on conceptual building design life cycle impact feedback.
Integration of multi-disciplinary design optimization and sequential
design decision-making processes
A final aspect of the problem designers face regarding life cycle environmental impact
and cost feedback is that MDO and design space sampling methods do not currently integrate
well with conceptual building design decision-making processes. Decisions are typically made
by architects in sequential fashion, such that for example once the orientation of the building is
known, the placement of shading devices can be determined for each façade in order to minimize
cooling loads. Designers may also wish to understand the life cycle environmental impacts and
costs associated with the wall assembly system before deciding upon the cladding system. Such a
multi-objective sequential feedback approach is typical in the architecture, engineering, and
construction industry in that project stakeholders often need to evaluate design decision trade-
offs for competing objectives. For example, a designer wishing to minimize both environmental
impact and cost may find that a certain window-to-wall ratio lowers carbon footprint at the
expense of greatly increased life cycle cost.
Existing MDO methods do not accommodate sequential decision-making processes.
MDO requires all design decisions to be made in parallel, instead of allowing designers to define
variable values sequentially and thereby understand the impacts for each successive decision.
Consequently designers utilizing MDO must decide on all building decisions before receiving
feedback on any single design choice. MDO methods do not integrate well with the AEC
14
industry, which relies on flexible and often-changing decision-making processes, especially
during the conceptual design phase.
Survey of current use of life cycle assessment in the building design
industry
In the last few decades, the architecture/engineering/construction (AEC) industry has
significantly evolved in terms of its use of computers in building design. Intelligent computer-
aided design (CAD) systems were first developed in the early 1970s as a way to rapidly create
forms and easily modify designs (Eastman 2008). The methods gained industry acceptance in the
1980s, as architects shifted from manual pencil and drafting board methods to computer-driven
form-based design methods (Terzidis 2006). Parametric design, or the defining of a CAD model
through the use of parameters, emerged in the 1980s with Lin et al.’s concept of variational
geometry (1981). The method allows designers to add geometrical constraints within computer
models, which automatically update as designers change parameters’ values (Chen et al. 2004).
Parametric design was taken to a powerful new level in the 1990s, when CAD modeling was
used to quickly generate many spatially novel, complicated forms. Variations in building designs
were generated through an algorithm, extending the role of the computer to both draftsperson
and performance analyst (Shea et al. 2005). This use of parametric design, also called generative
design, allowed for the creation of many design alternatives with complex curves and patterns
not easily visualized or manually drawn. Optimization algorithms programmed into generative
design programs quickly searched for solutions based on user-defined objectives.
Today, industry adoption of sustainable design techniques lags behind advancements in
design methods. Complicated designs are now easily possible but “often with little regard to the
15
cost in energy and material resources” (Taylor and Carper 2007, p. 152). The following two-part
glimpse into industry practice delineates the lack of sustainability-focused methods integrated
with design. The first part highlights the theoretical gap between LCA and LCC feedback and
building design by drawing from literature, certification programs, and software tools. The
second part highlights the practical gap between LCA and LCC feedback and building design by
reporting feedback from industry professionals. Results show that firms lack a quick and easy
method of receiving quantitative environmental impact and cost feedback on their designs at
early design stages. Such feedback would be of considerable use to industry in meeting the
challenge for life cycle impact feedback to maintain a similar technological pace as building
design methods.
Defining the theoretical gap between LCA and LCC feedback and building design
Building designers have been aware of sustainable design principles for decades, and
early design stages have been well established as critical elements in producing green buildings
(Gruman 2003). In early representative literature such as Climate Design (Watson and Labs
1983), 50 sustainable design strategies – such as site selection, microclimate control, and
massing – are outlined to architects and building energy consultants specifically for
consideration during early design stages. Drivers for incorporation of these strategies include
marketing benefits, environmental labeling of buildings, environmental targets for buildings and
nations, and subsidies for environmental impact reduction (Bribian et al. 2009).
The green building movement gained considerable traction with the inception of LEED,
or Leadership in Energy and Environmental Design, in 1998. LEED provides building owners
with a framework for implementing green building design strategies, and incentives for earning
16
LEED certification include growth in the local economy, tax breaks and abatements, and creation
of a healthier indoor and outdoor environment (USGBC 2011b). As of September 2011, 24,444
projects in the United States were LEED certified, and about 22 projects are certified daily
(USGBC 2011a). The meteoric rise of building energy simulation tools is another indication of
strong interest in sustainable design practices. Four hundred eleven tools are listed on the US
Department of Energy’s website (USDOE 2011b), and several are added every year.
Despite the heightened awareness of sustainable building design strategies in the past few
decades, additional efforts are needed to further engage the AEC industry with the sustainable
design movement. As one architect recently put it, “Architects’ increasing awareness of the
relevance and importance of developing a sustainable built environment has for the past few
decades been inversely proportional to architects’ direct efforts to create this environment” (de
Graaf 2011). Current CAD tools poorly integrate feedback with conceptual building design
(Drogemuller et al. 2004). Computational methods such as generative design are rarely used in
practice (Terzidis 2006). Performance feedback needs to be better integrated at early design
stages, when designers have the least understanding of, yet greatest opportunity to manipulate,
the building form. Initial forays during early design are critical to gaining a sense of a form’s
environmental and cost impacts yet remain problematic “for the novice and experienced architect
alike” (Fawcett 2003, p. 2).
Among the reasons for lack of adoption of sustainable methods are owners’ and
managers’ perceived increased financial risk of higher initial capital costs and perceived lack of
tenant demand (Wilson and Tagaza 2005). Designers are also reluctant to use energy simulation
tools because of the steep learning curve and extensive required data inputs (Jacobs and
Henderson 2002), poor interoperability with design tools and lengthy processing time (Krygiel
17
and Nies 2008), lack of known data on which to base energy simulations (O’Donnell et al.
2004), and difficulty in interpreting results (Schlueter and Thesseling 2009). Integration of
design and analysis is also seen as requiring extensive time and effort (Wilson and Tagaza 2005),
particularly during early design stages when designers lack full conceptual clarity. LCA
feedback integration with design is perceived as economically costly, complicated, and
inaccurate, techniques for improving building designs based on LCA feedback are unclear, and
LCA and LCC databases lack computational links with energy simulation programs, creating an
inefficient workflow (Bribian et al. 2009).
The literature points out that industry adoption of a method of incorporating LCA and
LCC feedback into early stage design would favor several features. The method should rapidly
evaluate many options, select solutions based on user preferences, and incorporate optimization
techniques for building energy consumption (Crosbie et al. 2010). The method should also
integrate, rather than separate, the roles of designers, engineers, and consultants. Complicated
analyses tend to foster fragmentation and compartmentalization among design-analysis teams
(Krygiel and Nies 2008). Fragmentation in turn causes a decline in efficiency – within both the
design and analysis processes as well as the buildings created (Krygiel and Nies 2008). A
method that fluidly integrates parametric design and environmental and cost analysis during the
early design stages needs to bridge this gap between efficiency and design of the sustainable
built environment.
Defining the practical gap between LCA and LCC feedback and building design
Feedback from AEC firms was solicited to validate these theoretical findings and help
understand the practical issues of implementing LCA and LCC feedback in building design. The
18
firms surveyed and interviewed perform design and environmental analysis to varying degrees.
The feedback consisted of two parts, and sustainability personnel at AEC firms were surveyed in
each part. In the first part, six firms completed an online survey so that general trends about LCA
feedback in the architecture industry could be understood. In the second part, three firms
conducted phone interviews so that in depth, firm-specific details on industry use of LCA in
design could be assessed. The objectives were twofold: to understand the degree to which
designers currently receive environmental impact and cost feedback on their designs and the
practical issues preventing industry adoption of an early stage LCA feedback parametric design
tool.
Part I – Understanding General Trends of LCA and LCC Feedback and Design
In Part I, AEC firms completed an online survey to gauge general trends in the use of LCA
and LCC feedback in design. Appendix 1 presents the survey questions, and the results are
presented here. Results showed consensus as far as firms’ methods of receiving environmental
impact feedback. In response to the first question asking how frequently firms use certain
strategies to reduce buildings’ environmental impact, all firms received some fort of feedback
but to different degrees. Figure 2 shows that firms perform LCA on about 25% of projects, which
is less often than other sustainable design strategies. Most firms only implemented strategies to
reduce operational energy.
19
Figure 2 – Frequency that designers perform various sustainable design strategies on building
projects.
As far as LCA software used in industry, Athena was the only program listed. Many firms
did not list any programs, and these firms acknowledged that in the rare case when a client
desired LCA feedback, the work is out-sourced to building energy consulting firms not involved
in the building design. Firms also did not list any database sources for conducting LCA. The
conclusion is that firms infrequently conduct LCA, both due to lack of client demand and out-
sourcing.
Firms were also asked at what stage in the design process they make design decisions.
Figure 3 shows that some decisions are made early in the design process and in some cases
earlier than American Institute of Architects’ (AIA) guidelines. Although it is well understood
the importance of making design decisions as early in the design process as possible (Struck and
Hensen 2007), this may be problematic if firms do so without understanding the life cycle
impacts associated with these choices. Changes to such decisions at later design stages can
greatly increase project impacts (Grierson and Khajehpour 2002). For example, Figure 3 shows
that the firms tend to make decisions on materials for finishes, partitions, cladding, and the
foundation, the window-to-wall ratio (“façade design”), the building form, as well as whether a
20
building will have shading (“envelope elements”) prior to AIA guidelines. Figure 2 showed that
firms tend not to receive LCA feedback on projects. These two results taken together suggest two
possible consequences associated with decisions made prior to AIA guidelines: 1) the decisions
may not have low life cycle impacts, 2) changes to these decisions at later designs stages may
exponentially increase project impacts. It would be beneficial if firms continued to make
decisions prior to AIA guidelines but with the guidance of a LCA feedback tool. In this way,
designers can make informed decisions of life cycle impacts during the conceptual design stage
with the aim of creating low-cost, low-carbon buildings.
Figure 3 – Stages during which designers make various building decisions compared with
American Institute of Architects guidelines.
Figure 4 shows that firms cited several barriers as far as implementing LCA feedback on
projects, including complexity and inaccessibility of the software and lack of clarity of results.
This suggests that firms would appreciate a tool that is simple to use, requires minimum inputs,
and presents results in an easy to understand fashion.
21
Figure 4 – Barriers inhibiting design firms’ use of LCA on building projects.
The final question asked about the usefulness of an early design-LCA feedback tool. All
responses were either “agree” or “agree strongly”, suggesting the high potential of providing
such a tool to the design industry.
Part II – Understanding firm-specific details of LCA and LCC feedback and design
In Part II, North American design personnel involved with sustainability initiatives at three
global AEC firms were interviewed on the phone for approximately 60 minutes each. Each firm
uses LCA and LCC in their design process to varying degrees. Objectives were to ascertain
challenges when using LCA and LCC to improve building designs as well as gauge opinions on
a tool that provides LCA and LCC feedback to generate optimal designs.
22
Arup
Founded in 1946, Arup is one of the world’s largest AEC firms and is comprised of
designers, planners, engineers, and technical specialists. The firm is headquartered in London
and has over 10,000 staff in 37 countries spanning southern Africa, Pacifica, and the Middle
East. Arup has designed such iconic buildings as the Sydney Opera House, Centre Pompidou in
Paris, the Bird’s Nest Stadium in Beijing, and the platinum LEED-rated California Academy of
Sciences in San Francisco. The latter solidified Arup’s reputation in green building design, as the
firm invests heavily in sustainability research and analysis. LCA and LCC have played an
increasingly valuable role in their design services, which range from optimizing the passive
performance of a building envelope to analyzing operational data in order to improve future
buildings’ life cycle performance.
Francis Yang, a structures and sustainability specialist in Arup’s San Francisco office,
provided the following insights on Arup’s use of LCA during a phone interview in May 2011.
The primary challenge Arup’s San Francisco office has encountered when integrating LCA
feedback with design is finding an LCA feedback tool appropriate for early stage design. The
office uses Athena’s Impact Estimator for Buildings (IE), since no other tool provides indicators
based on the Tool for the Reduction and Assessment of Chemical and Other Environmental
Impacts (TRACI). Arup has encountered three problems with the tool. First, Athena IE has
greatest utility at later design stages. The IE requires inputs, such as structural information,
known by engineers typically only at late design stages. Second, Athena IE is not very appealing
to architects, since its focus is on structure and materials rather than form generation,
visualization, and aesthetics. Form-generation capabilities and footprint shape options are absent
beyond basic rectangular configurations. Athena IE’s use of assemblies, or pre-defined wall,
23
window, and building envelope systems, is less useful than a method that makes use of material
quantities. More useful would be a way for designers and cost estimators to input quantities then
layer in environmental impacts. There are only a few choices for structural systems, concrete
assemblies, and many materials. More helpful would be an easy way to build up the concrete and
steel mixes, including recycled content, fuel inputs, and manufacturing method. The broad range
of wood options is not useful, as Arup rarely works in wood. Seismic and wind conditions are
also not included in the IE. TRACI midpoint indicator results are hard to interpret for designers,
and a major question is whether they are accurate for the project location. A final limitation is
how to compare results. The IE compares results to a universal baseline, which is typically
dissimilar in size, form, and function from the actual building. More useful would be a
comparison that is similar in terms of size, shape, and type to the building project under analysis.
Arup also believes that industry potential exists for a method that integrates LCA and
LCC feedback into early stage parametric design using techniques that generate many design
alternatives. The primary concern is validation of results and transparency of the degree of
accuracy of the results. High-performing results need to be clearly rationalized, including why
certain design variables were considered and others were not. Designers need to clearly
understand design tradeoffs, the biggest contributors to the building’s impacts, and those design
changes that will improve the design the greatest. The tool must also easily allow changes to
constraints. This is especially necessary for example if the building forms for high-performing
results are not constructible.
24
Atelier Ten
Founded in 1990 by a team of engineers, Atelier Ten is a London-based environmental
design firm with a focus on applying technological innovation to the design of sustainable built
environments. The firm has six offices in North America, Europe, and the Middle East, and
personnel represent a mixture of architects and engineers. Projects have included the tallest
LEED Gold certified commercial building in the United States, the first use of earth-duct
technology in the UK, and the use of layered facades, high performance ventilation, and a
“thermal labyrinth” for cooling of spaces in a large arts and media building in Melbourne,
Australia.
Emma Marchant in Atelier Ten’s San Francisco office was interviewed in May 2011 and
provided the following insights on their use of LCA in design. The firm performs environmental
design primarily after the early design stages. Typically the building form has been designed, and
Atelier Ten then analyzes climate data for the given building type. Their focus is applying design
strategies to make the operational phase more efficient. In the rare case when the firm is involved
in the early design stages, they take into account clients’ goals and constraints, climate data, and
code considerations before creating a carbon zero roadmap. Strategies for this roadmap include
daylighting, passive cooling, renewable energy sources, and water and waste reduction. Early
design decisions include cladding and other façade materials and floor and ceiling types. Most
other materials decisions are made late in the design process.
LCA is rarely used in Atelier Ten’s projects. Clients are not interested in whole building
LCAs because of perceptions over cost and efficiency: designers and building owners believe
LCA is synonymous with high cost and a protracted design process. Architects would like to be
25
as green as possible, but they would like complete information in a cheap, efficient, and easy to
understand manner.
Atelier Ten also questions the potential for a tool with form-generation capabilities. A
method of automating form changes is rarely used in practice, because designers like to maintain
control over the building form. Visualization of a building’s form evolution throughout the
design process is critical to architects. A challenge would be to design a form-generation tool
that appeals to designers in terms of form control, especially if an innovative form’s cost is
greatly different from a simple form. However, it would be useful for designers to understand the
environmental impact and cost tradeoffs of one building form over another.
The firm is also very interested in receiving environmental impact and cost performance
feedback for many design alternatives. Such feedback at the early design stages may outweigh
loss in control of the form’s generation. The firm would also like a clear understanding of the
pros and cons of these design alternatives in terms of environmental impact and cost. For
example, it would be useful to understand the tradeoffs of materials, such as whether to use a
heavyweight or lightweight material for the structural frame, cladding, rainscreen, and
foundation. As far as material choices, Atelier Ten would find it helpful to visualize the
embodied energy, cost, and availability of each material.
Atelier Ten also believes an early stage design LCA and LCC feedback tool should above
all be clear in its results. The architect should be able to easily interpret the results and trust that
the option or options presented are the best. Architects should be able to assign preference
weights to these issues, and results should then be ranked accordingly.
26
Kohn Pedersen Fox
Kohn Pedersen Fox Associates (KPF) is a large architecture firm founded in New York in
1976. Offices today include London, Shanghai, Seoul, Hong Kong, and Abu Dhabi. The firm’s
expansion in the 1980s to Europe, where LCA has historically had a stronger foothold than in the
United States, helped foster a sustainable design approach. KPF’s in-house environmental
specialists use tools such as Ecotect and Green Building Suite to consider operational energy
design strategies. Important projects include the World Bank Headquarters in Washington, D.C.,
redevelopment of the large-scale Canary Wharf in London, the Shanghai World Financial
Center, and several LEED gold and platinum certified projects in North America and China.
Brad Zuger in KPF’s New York office was interviewed in May 2011 to provide insights
on the firm’s use of LCA in design. KPF’s New York sustainability team considers both
operational and embodied energy considerations in the design process. Team members are
typically trained as architects rather than engineers, and specialized environmental analyses such
as LCA are out-sourced in the rare case when a client calls for them. Economic cost is just as
important, if not more important, a design consideration as environmental impact.
KPF believes an optimized form-finding LCA and LCC feedback tool has potential for
architects during the early design stages. Architects are always looking for new tools and are
willing to give up some of the creative process to generative design. However, architects should
be able to easily manipulate the results to generate new building forms. Architects would not
want to invest heavily in learning the tool and are cautious about the manual burden of sorting
through many generative solutions. The tool should also avoid narrow consideration of
objectives. Instead, the tool should be multi-disciplinary by showing tradeoffs between
environmental impacts and economic costs. These tradeoffs should relate not just to building
27
form but also material choices. An iterative manual approach between design and analysis is
much less desirable than an automated approach. In that sense, the tool’s ability to automatically
generate many options and present a few best examples – coupled with the analytical data –
could be of high interest to architects. Visualization of the form and aesthetics are important to
architects, but they are just as mindful of achieving their performance objectives. The analysis
should be rigorous, complete, and easily understandable in order for it to be useful.
Summary of Key Themes
The three AEC firms interviewed all performed design and environmental analysis to
varying degrees. This was due in large part to whether personnel included engineers and
architects (Arup, Atelier Ten) or architects alone (KPF). In the rare case when the firms
conducted LCA, it was almost always used in the late design stages when the building form was
established and firm design choices had been made. Such retrospective analyses were typically
used for marketing a project as green or sustainable, even though the LCA played no part in
guiding design choices. LCA feedback used specifically at the conceptual design stage could
therefore fill this gap by guiding design choices that help create buildings with lower life cycle
carbon footprint. The firms all agreed on several features of a LCA feedback tool. Data should be
rigorous and show design tradeoffs, especially the life cycle environmental impact and cost
implications of material choices and building orientation. Display of a design’s cost was
mentioned as a valuable feature. In order to overcome the perception that LCA is costly and
inefficient, the tool should be easy to learn and not require specialized knowledge. Results
should be easily understandable, presented quickly, and represent a manageable number of best
solutions. The utility of form-generation capabilities received mixed opinions. However, the
28
firms agreed that presenting a few snapshots of the best designs coupled with the analytical
results could be very useful, as long as the building form could be easily manipulated for those
best designs.
Reflection on Use of Life Cycle Assessment in Building
Design
The previous section of this chapter described the observed problem: life cycle
assessment lacks integration with conceptual building design. Instead, LCA is applied – if at all –
in the late stages of the design process, when the influence on a building’s performance is low.
The main reasons for this lack of integration are summarized here in order to motivate the
research questions presented in the next section.
The primary reason why LCA lacks integration with conceptual building design is that
LCA typically requires many detailed inputs. This relates to the top-ranked priority in a 2009
survey of designers and architects on the usability of building performance simulation tools and
their effectiveness in integrating with the building design process: analyses should be quick and
efficient, and assumptions and default values should be allowed in place of detailed inputs (Attia
et al. 2009). This sentiment was also shared in the survey given to design firms and as described
in the previous section, in which many respondents cited complexity of LCA as a barrier to
integrating LCA with their design processes (see Figure 4). Designers have formalized very little
information about a building project during the conceptual design stage, and therefore it is
difficult at this stage to apply LCA tools that require large amounts of precise information. More
useful would be a method that provides rapid LCA feedback based on a scaled-down number of
inputs.
29
A second problem with LCA feedback is that tools do not provide sensitivity analysis on
building design parameters in relation to design decisions. This was ranked the second-highest
priority in the survey (Attia et al. 2009) and could explain why firms completing the industry
survey cited “irrelevance of LCA to firm’s work” as a barrier to performing LCA (see Figure 4).
Users have no clear way of understanding which design parameters contribute significantly and
consistently to a project’s life cycle impact and which parameters are consistently less important.
A mechanism is needed that provides this sensitivity analysis feedback so that designers can
focus their efforts on those decisions that affect a building’s impacts. A related problem is that
designers do not have a clear way of understanding the impact tradeoffs of certain design
decisions. For example, a designer may wish to understand the embodied versus operational
impact tradeoffs of glazing materials in order to minimize a building’s life cycle impact. Without
such an understanding, a designer may choose a material with a low embodied impact but with a
high operational impact or vice versa. A method that shows such impact tradeoffs for a range of
design parameters would allow designers to make decisions yielding low life cycle impacts.
A third problem related to LCA and building design is that a tool does not exist that
integrates automated feedback with building design. This was the top priority cited by the survey
respondents when asked about information management of a building performance simulation
tool’s interface (Attia et al. 2009). Although this problem does not relate directly to any point
cited in the industry survey described in the previous section, use of automation can potentially
increase the relevance of LCA in firms’ design work, one of the top barriers preventing firms
from performing LCA as shown in Figure 4. As the survey by Attia et al. (2009) pointed out,
designers would like to compare their initial design’s performance with the performance of
multiple design alternatives. Such feedback would be especially useful if provided sequentially,
30
or immediately after each design decision is made, as opposed to after all decisions have been
made. Automated processes can be leveraged that provide designers with the ability to compare
the performance of many building design alternatives. Related to this problem is the ability for
designers to understand the full range of impacts possible across many design alternatives. This
would allow designers to easily compare the performance of a building design to the best and
worst designs possible as well as show designers the degree to which each design decision
improves, worsens, or does not affect a building design’s performance.
Research Questions
Development of an automated method for providing life cycle environmental impact and
life cycle cost feedback during the conceptual building design is motivated by the following five
research questions:
1. How many design inputs are required for a method that incorporates life
cycle assessment and life cycle cost feedback into sequential building design?
2. Which design decisions contribute most significantly to building embodied impacts?
3. How well does a design strategy minimizing only operational impact compare
with a strategy minimizing both operational and embodied impacts?
4. How well can a method that leverages automated feedback be used to support
sequential building design decision-making processes?
5. What is the range of control for a set of building design parameters in terms of
life cycle environmental impact and life cycle cost performance, and how can designers
make decisions within this range?
31
Organization of the Dissertation
The research is organized into five chapters, which build on each other in terms of
developing the method for providing conceptual design phase life cycle impact feedback.
Chapter 2 lays groundwork for the first research question by developing equations for an
important part of the feedback method. These heuristics calculate pre-operational embodied
impacts using a minimal number of inputs, in order that the method may be applied as early in
the design process as possible. Chapter 3 develops the methodology further by using a sampling
method to apply the equations to a case study. This chapter answers the second research question
by highlighting the degree to which design decisions contribute to embodied impacts, and
operational impacts are left out of the scope. Chapter 4 includes operational impacts within the
scope of the research, and the third research question is answered by applying an optimization
algorithm to the same case study. The chapter presents an application of the method, which
allows designers to understand the relative importance of embodied versus operational impacts.
Chapter 5 continues to build the method by using a sampling algorithm to integrate automated
feedback into sequential design. The chapter answers the fourth and fifth research questions by
applying probability mass functions to a range of sequential design strategies. Designers are
enabled to understand the range of control they have over a building project’s impacts for a given
problem formulation. Chapter 6 extends the integration of sequential decisions and LCA
feedback to include LCC feedback. The chapter introduces a new set of cost heuristics, which,
similar to the embodied impact heuristics introduced in Chapter 3, allow designers to receive
feedback on many designs given very few inputs. Probability mass functions are then applied as
in Chapter 5 to show designers the full range of possible costs and environmental impacts after
each design decision. This multi-objective feedback approach allows designers to evaluate the
32
life cycle environmental impact and life cycle cost trade-offs of sequential design decisions.
Based on this feedback, designers can understand how well each decision aligns with their
particular building performance strategy.
Chapters 3 through 6 are four journal papers, and each has been submitted to peer-
reviewed journal publications. Chapter 3 presents a method for calculating embodied impacts
and performing sensitivity analysis of building components and has been published in Building
and Environment (Basbagill et al. 2013). Chapter 4 presents a method for describing the
embodied versus operational impact tradeoffs of building component decisions and has been
submitted to Energy and Buildings. Chapter 5 presents a method for integrating automated life
cycle environmental impact feedback into sequential building design decisions and has been
submitted to The International Journal of Architectural Computing. Chapter 6 integrates life
cycle environmental impact and life cycle cost objectives into sequential building design
decisions and has been submitted to Automation in Construction. Each paper stands as an
independent publication, which means the papers’ background sections overlap somewhat in
content. The papers have been largely untouched in the dissertation, and references have been
aggregated into a list at the end of the dissertation.
Chapter 7 is the conclusion of the research and summarizes the primary contributions of
the research developed in Chapters 2 through 6. The chapter also summarizes answers to the
research questions, which are also developed in Chapters 2 through 6. Chapter 7 also includes
sections on challenges for researchers who may wish to extend the work presented here,
limitations of the method, and possible future research avenues. Finally, a set of appendices are
provided which include supplementary data tables.
33
Chapter 2: Scope, model development, and validation
This chapter describes the development of the model for integrating conceptual building
design with life cycle environmental impact and life cycle cost feedback. The model is intended
for application specifically during the early design stages, when the design problem is typically
not well defined, the number of design alternatives is large, and the potential to reduce
environmental impacts and cost is greatest. The approach for building the model leverages
automated processes. A computational feedback processor is integrated with building
information modeling (BIM), life cycle assessment, and life cycle cost software tools. The
processor performs sensitivity analysis on building design parameters by quickly evaluating the
impacts of thousands of building design alternatives. The feedback processor applies an
optimization algorithm to a given problem formulation, which yields a set of improved designs.
Alternatively, the processor can apply a sampling algorithm to the design space and build a
profile of the full range of impacts. Such feedback can be used to show designers which building
components are consistently large contributors to a building’s impacts, and which are less
important.
As part of the development of the software integration method, two sets of heuristics
were developed for calculating pre-operational impacts. These heuristics require few inputs and
allow for quick calculation of pre-operational 1) embodied impacts and 2) costs for a range of
building components. Use of the heuristics within the BIM-enabled LCA-LCC feedback method
allows designers to understand the relative importance of building components’ life cycle
embodied impacts and costs for thousands of building design alternatives. The sections in this
chapter outline the steps required to create the BIM-enabled LCA-LCC feedback method and
34
describe the development of the pre-operational embodied impact heuristics. Chapter 6 provides
detail on the development of the pre-operational cost heuristics.
Scope of the method
The goal of the methodology is to use an automated approach to provide LCA and LCC
feedback to designers for many building designs at the conceptual design stage. Central to
creating the method is deciding on the scope for building life cycle phases, environmental impact
indicators, and building components. Several phases and components are included in the method
in an effort to maintain as broad a scope as possible.
Building life cycle phases
Figure 5 schematically shows the complete physical life cycle of a typical building. The
shaded area shows those phases that are included in the method. Embodied impacts prior to the
operational phase include raw material acquisition and material production, and embodied
impacts during the operational phase include the maintenance, repair, and replacement (MRR) of
building components. Utility impacts during the operational phase include impacts due to
HVAC, lighting, plug loads, and water use. Demolition and on-site construction have been
excluded, since impacts associated with these phases have been shown to be difficult to calculate
(Schoch et al. 2011) and small when compared with other phases (Scheuer et al. 2003). The
scope of the building phases included in the method is identical to the scope of the building
phases included in the validation of the method. See the section “Validation of embodied impact
heuristics” for details on the validation.
35
Figure 5 – Building life cycle phases included in the BIM-enabled LCA-LCC feedback method.
Environmental impact indicators
Researchers have identified several impact categories that are useful in measuring the
environmental impact of buildings. These impact categories include global warming potential,
non-renewable energy consumption, human toxicity, acidification, and eutrophication, among
others (Jolliet et al. 2003). Although the importance of all of these categories in comprehensively
assessing environmental impact is recognized, the proposed method is demonstrated for only
global warming potential. The metric used for this purpose is carbon dioxide equivalents (CO2e)
using the relevant 100-year global warming potential (Wright et al. 2011), which measures the
total amount of greenhouse gas emissions of the building, considering all relevant sources. The
building owner or designer could add other impact categories to the analysis as required.
Building components, materials, and dimensions
The framework used to determine the scope of the building components and materials is
based on Uniformat 2010. This classification system is used in the AEC industry to classify
building components within building element categories (Construction Specifications Institute
2010). Elements within the system refer broadly to the parts of a building. Uniformat elements
36
within the project scope are: Substructure (A), Shell (B), Interiors (C), and Services (D). The
remaining elements (Equipment and Furnishings (E), Special Construction and Demolition (F),
and Sitework (G)) are not considered, since these decisions relate to interior aesthetics, require
specialized knowledge of site conditions, or otherwise involve decisions that would be
impractical to make by designers before the design development stage.
Table 1 outlines the assemblies and their sub-components for each of the four Uniformat
elements. Material choices for each component are determined using choices for building
component assemblies available in RSMeans (RSMeans 2007) and Athena EcoCalculator
(Athena 2011). These choices are not meant to be exhaustive but rather representative of
common materials for each component. Appendix 2 enumerates the material choices and their
properties for each building component. These properties include material densities and
embodied CO2e factors, or the amount of carbon dioxide equivalents associated with materials’
feedstock energy, energy required to process the materials into building components, and fuel
cycle energy for all pre-operational processes.
The building component classification framework includes thickness as a dimensioning
variable. Specifications from several construction material and equipment supplier sources are
used to determine thickness ranges. These sources are listed in the description for Table 1. The
smallest minimum value and largest maximum value are identified across all sources for each
sub-component and placed into the table. Thickness ranges are not articulated for every
component, namely those whose size determinations are difficult to reduce to one single
thickness parameter and/or best quantified by structural analysis methods applicable to later
design stages.
37
Table 1 – Building component classification framework. Sources used for thickness ranges (by
Uniformat element): A: (RSMeans 2007), (ACI 2004); B: (RSMeans 2007); C: (RSMeans 2007).
bThickness
Uniformat element Assembly aSub-components Number of
material choices Minimum (m) Maximum (m)
cA: Substructure piles piles, vapor barrier, caps, slab-on-
grade, grade beam, rebar, formwork
2, 2, 1, 1, 1, 1, 1 0.1 0.4
footings footings, vapor barrier, slab-on-grade,
grade beam, rebar, formwork
1, 2, 1, 1, 1, 1 0.1 0.4
mat foundation foundation, vapor barrier 1, 2 0.2 1.8
B: Shell columns and beams 10 n/a n/a
floor structure 12 n/a n/a
roof roof structure, membrane, insulation,
paint
10, 5, 1, 1 n/a n/a
stairs stairs, railings 3, 3 n/a n/a
cladding 7 0.02 0.08
exterior walls wall structure, insulation, membrane,
gypsum, paint
5, 1, 1, 1, 1 n/a n/a
glazing glass, polyvinyl butyral, frame,
hardware
1, 1, 5, 1 0.007 0.02
doors door, hardware 3, 1 n/a n/a
C: Interiors partitions partition structure, gypsum, paint 2, 1, 1 0.2 0.6
doors door, hardware 2, 1 n/a n/a
wall finishes covering, paint 2, 1 0.005 0.02
flooring surface, insulation 9, 13 0.1 0.2
ceiling plaster, gypsum, paint 1, 1, 1 0.006 0.02
dD: Services mechanical 17 sub-components e13 n/a n/a
electrical 16 sub-components 1 n/a n/a
plumbing 23 sub-components 1 n/a n/a
fire 4 sub-components 1 n/a n/a
conveying elevator 1 n/a n/a aTotal number of sub-components listed in the table equals 107. Of these, 102 are distinct sub-components. Five are double counted and occur in multiple components:
vapor barrier, slab-on-grade, grade beam, rebar, and formwork. This double counting occurs because the substructure consists either of piles (seven sub-components),
or footings (six sub-components, five of which are present in the piles sub-components), or mat foundation (two sub-components, one of which is present in piles and
footings).
b Thickness ranges correspond to bold sub-component and all material choices for that sub-component. For assemblies with multiple bold sub-components, ranges
represent combined thicknesses.
c Substructure consists of one of the three listed assemblies. Remaining three elements consist of all listed assemblies.
d Large numbers of services sub-components preclude enumeration.
e Duct insulation is a mechanical sub-component with 13 material choices. Remaining mechanical sub-components have one material choice.
38
Software Integration Method
Figure 6 is a schematic showing the software and data dependencies in the automated
conceptual building design and LCA and LCC feedback method. The method is called the
Architectural Design And Performance Tool, or ADAPT. The arrows in the figure represent the
data dependencies. A broad overview of the method is presented here, whereas further details on
the specific software used are presented in the method section of Chapters 3, 4, 5, and 6.
Figure 6 – Architectural Design And Performance Tool (ADAPT): software integration tool
providing life cycle embodied impact, operational impact, and cost feedback on conceptual
building designs.
The process of receiving automated impact feedback on building design alternatives
begins with a designer manually creating a building information model (BIM). The BIM inputs
consist of the building’s size, type, and location as well as ranges for each of the variables. Table
2 lists these inputs in terms of constraints, variables, and assumptions. The constraints are
39
necessary for determining the maintenance, repair, and replacement schedule. For example, a
building of one type and size located in a hot and dry climate may have very different MRR
impacts than a building of another type and size located in a cold and wet climate. Appendix 2
lists the materials considered for each of the building components. Table 1 lists the ranges of
dimensions considered for each building component. A broad overview of all variables
considered is presented here. The specific variables considered for each of the studies performed
in Chapters 3, 4, 5, and 6 depends on the particular objectives of the study. Details for each
specific problem configuration are provided in the chapters.
Table 2 – Required inputs, variables, and assumptions for BIM-enabled automated feedback
method for life cycle assessment and life cycle cost.
Required Inputs
Location
Building type
Gross floor area
Variables
Number of buildings
Number of floors
Shape parameters
Window-to-wall ratio
Orientation
Building component materials
Presence of shading devices
Substructure type
Building materials
Building component sizes
Assumptions
Footing depth = 2m
Pile depth = 15m
Bay spacing = 9m
Floor-to-floor height = 4m
Wall assembly R-value = 16.81 K·m2/W
Roof R-value = 21.84 K·m2/W
Service life = 30 years
40
Embodied impact calculation
Once the designer has specified inputs in the BIM, the automated design-feedback
process begins. Calculation of pre-operational carbon footprint is the first step in this process.
The use of embodied impact heuristics is an essential part of this calculation. These heuristics
answer the first research question and are a theoretical contribution of the work and can be
applied to many types of LCA and LCC building analyses. These formulas depend on the BIM
inputs outlined in Table 2, and Appendix 2 provides all the formulas. The building phases
included in the scope of these formulas are shown in Figure 5 and are identical to the scope of
the building phases included in the validation of these formulas in the next section. The impact
of each building component material can be calculated from these formulas. Inputs include
material choices and average thickness values from Table 1 as well as gross floor area and all
required inputs, variables, and assumptions outlined in Table 2. Pre-operational carbon footprint
is calculated by multiplying each quantity by the embodied CO2e factors in Appendix 2 and
summing the resulting impacts from the 102 sub-components.
These heuristics can be contrasted with embodied impact formulas used in Athena
EcoCalculator, which have been used by several recent studies to calculate embodied impacts of
building components (Wang et al. 2005b, Crawford et al. 2010, Xiong and Zhao 2011, Pidgeon
2012). Athena requires ten inputs, including building location, building type, and gross area for
the foundation wall, slab, supported floors and roof, intermediate floors, exterior cladding,
windows, partitions, and roofs. Building size and type are also required in ADAPT, whereas the
eight area inputs are reduced to a single input, the gross floor area of the entire building. This
reduction in inputs was accomplished by developing heuristics in consultation with senior
estimators at Beck Technology, an AEC firm. Beck aggregated data from bill of material
41
quantities on approximately one dozen of their building projects. The number of formulas totals
102 and equals the number of distinct building sub-components outlined in Table 1.
Calculation of pre-operational embodied impacts due to the service equipment (e.g.,
HVAC equipment, plumbing equipment) is an important sub-step of this process. These impacts
comprise 61 of the 102 building component impacts and are calculated by sizing each piece of
service equipment according to peak building load. An energy simulation program performs this
step as follows (eQUEST 2010). Inputs from Table 2 are automatically passed from the BIM to
the energy simulation program. Thermal zones are defined in the resulting energy simulation
model as well as standard assumptions regarding building occupancy and HVAC system controls
(ASHRAE 2009). The energy simulation software then calculates peak building load from these
inputs and assumptions. The result becomes the input to the 61 material quantity formulas for the
service equipment. Equipment supplier documentation is used to determine whether each piece
of service equipment typically increases in size as peak building load increases. Material
quantities for those pieces of equipment that typically increase in size are scaled linearly
according to peak building load. The resulting scaled and non-scaled material quantities are then
multiplied by the CO2e impact factors in Appendix 2 to determine the service equipment’s pre-
operational embodied impact.
The remaining 41 embodied impacts include building components in the substructure,
shell, and interiors. These components include all parts of the building envelope as well as the
structural systems. Calculation of the material quantities of these components is more
straightforward than the calculation of the service equipment material quantities, as no scaling
factor is applied according to peak building load. The formulas are given in Appendix 2 and
include values for the required inputs, variables, and assumptions in Table 2. The material
42
quantities are then multiplied by a CO2e impact factor in Appendix 2 to determine the embodied
impact.
Operational impact calculation
An energy simulation model is used to calculate the annual energy consumption of the
building in terms of electricity (kWh) and natural gas (kBtu). The energy simulation model is
created based on the geometry and building material information contained in the BIM. Thermal
zones as well as standard assumptions regarding building occupancy and HVAC system controls
are defined in the BIM (ASHRAE 2009).
A maintenance, repair and replacement (MRR) schedule is used to determine the impacts
associated with maintaining service equipment during the operational phase of the building.
Impacts associated with the production of materials in this schedule are grouped with the
building’s embodied impact. The schedule is determined by the gross floor area, building type,
location, and structural and mechanical details defined in the building information model, which
are entered into an online facility operations reference database (CostLab 2011).
Operational carbon footprint calculations have two components. The first depends on the
building’s electricity and natural gas consumption as calculated by the energy simulation model.
These quantities are multiplied by a unit impact to calculate carbon footprint. The second
component is associated with the maintenance, repair, and replacement of the service equipment.
The MRR schedule is determined by inputting all the constraints as well as the service life
assumption from Table 2 into an online facility operations reference database (CostLab 2011).
The program returns each component’s MRR dollar costs for every year of the building’s
operation. Equipment supplier documentation is then used to look up a typical material, material
43
quantity, and cost for each component. Material quantities are then calculated by combining the
MRR cost outputs from the operations database with the data from the supplier documentation.
These quantities are multiplied by a CO2e impact factor in a similar fashion to the pre-
operational impact calculation to determine the MRR operational carbon footprint. The life cycle
embodied carbon footprint is then calculated by summing the pre-operational and MRR
operational CO2e impacts. Life cycle environmental impact is equal to the sum of the embodied
and operational impacts.
Feedback processor
A genetic algorithm is used to automatically iterate the carbon footprint analyses
described above across a defined range of design variables. The algorithm may be used for such
objectives as minimization of embodied energy, minimization of operational energy, or
minimization of total energy. A sampling algorithm is used to understand the full range of total
impacts possible for a given set of design variables. The algorithm generates a probability mass
function, which is useful for showing how designs generated by the genetic algorithm compare
to the full range of total impacts possible for the design problem under consideration.
Software used
Eight software components are used to implement the proposed method illustrated in
Figure 6. DProfiler is used for building information modeling (DProfiler 2012). SimaPro and the
Athena EcoCalculator are used for environmental impact data and for calculating the building’s
carbon footprint (SimaPro 2010, Athena 2011). RSMeans is used to calculate building life cycle
44
cost (RSMeans 2007). The energy simulation software eQUEST is used to calculate operational
energy (eQUEST 2010), and CostLab is used to estimate the service schedules (CostLab 2011).
Excel is used to calculate the carbon footprint metrics based on the data provided by the previous
components (Excel 2007). The optimization and sampling processes are implemented using
ModelCenter (ModelCenter 2008), a program that allows users to bring commercial software
tools into a common environment using software “wrappers” to facilitate the application of
automated design problem exploration techniques. The genetic algorithm chosen is the Darwin
algorithm, and the sampling algorithm chosen is an orthogonal array for 90% of the designs and
a Latin hypercube for 10% of the designs.
Validation of Embodied Impact Heuristics
Eight buildings were analyzed to validate the embodied impact heuristics developed in
ADAPT. Data from these buildings were obtained from the Arup Project Embodied Carbon
Database (Arup 2013). Various sources conducted highly detailed life cycle assessments of
embodied impacts requiring considerable effort and information on each of these buildings, and
the scope of these LCAs was identical to the scope of the embodied impact heuristics developed
for ADAPT (Figure 5). The methods for performing these LCAs were independent of the
methods performed by ADAPT. Table 3 provides information on each of these buildings,
including the source that conducted the LCA, building type, building size, and the building’s
construction completion date. Each of the eight LCAs calculated a single value for life cycle
embodied impact in terms of kg CO2e/m2. This value is provided in Table 3 under “mean life
cycle embodied impact”.
45
Table 3 – Comparison of embodied impact values for eight building case studies: ADAPT versus
Arup (2013).
Method of Calculating
Embodied Impacts
Building Type Building
Size (m2)
Construction
Completion
Date
# of
Building
Designs
Analyzed
aMean Life
Cycle
Embodied
Impact
(kg CO2e/m2)
Standard
Deviation
(kg CO2e/m2)
ADAPT Office 1,358 1989
968 1,994 1,036 bSuzuki and Oka 1998 1 1,100 n/a
ADAPT Religious 2,500
Summer
2013
920 1,494 668
Arup 1 1,229 n/a
ADAPT Office 2,900 (completed)
738 1,763 812
Arup 1 1,059 n/a
ADAPT Office 8,458 1989
922 1,363 714 bSuzuki and Oka 1998 1 790 n/a
ADAPT Office 22,982 1987
952 944 239 bSuzuki and Oka 1998 1 780 n/a
ADAPT Civic 34,910 (completed)
980 906 229
Arup 1 818 n/a
ADAPT Train Station 47,250
Summer
2015
995 873 215
Arup 1 1,089 n/a
ADAPT Office, Retail,
and Restaurant 85,000 2014
986 853 207
Arup 1 776 n/a aScope of all LCAs is identical to ADAPT and illustrated in Figure 5.
bStudy was conducted independent of Arup and included in the Arup database.
The embodied impact values for the eight detailed LCAs were compared with
distributions of impact values generated using ADAPT. Eight models were created in ADAPT,
and the only input was the gross floor area for each of the eight case studies. Ranges were
entered for the variables listed in Table 2. The number of buildings was held constant for each
case study and equaled one, except for the 2,900-m2 office building which consisted of five
buildings. The range of values for the shape parameters and number of floors was adjusted based
on the gross floor area, with the larger buildings having slightly greater minimum and maximum
possible values than the smaller buildings. Table 3 shows the mean embodied impact and
standard deviation for each of the eight distributions. Operational impact and cost were not
calculated as part of the embodied impact validation method.
46
Figures 7 through 14 compare the embodied impacts generated using ADAPT with the
embodied impact value obtained from the Arup database for each of the eight case studies. The
results are presented from smallest to largest building. The results show that ADAPT captures
reasonably well embodied impact values of highly detailed LCAs for small, medium, and large
buildings. All eight distributions captured the detailed LCA value, and this value was within one
standard deviation of the mean for seven out of the eight distributions. The train station case
study was the exception, and this embodied impact value was within 1% of the upper value of
the distribution’s standard deviation. Mean values of the distributions for large buildings were
generally much closer to the detailed LCA values than small buildings. Embodied impacts of
small building sizes measured using ADAPT were consistently greater than embodied impacts of
detailed LCAs. The standard deviations of the distributions are also lower for larger building
sizes, suggesting the heuristics are able to predict impacts more precisely as building size
increases. These observations are consistent with the fact that the underlying embodied heuristics
in ADAPT were developed from data on large building projects. ADAPT is also better suited for
larger projects, since such projects have a larger number of variables and variable ranges and a
larger design space size than smaller projects. However, the results suggest that the heuristics
may be applied to both small and large building projects.
47
Figure 7 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 1,358-m2 building.
Figure 8 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 2,500-m2 building.
48
Figure 9 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 2,900-m2 building.
Figure 10 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 8,458-m2 building.
49
Figure 11 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 22,982-m2 building.
Figure 12 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 34,910-m2 building.
50
Figure 13 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 47,250-m2 building.
Figure 14 – Comparison of embodied impact values generated from a detailed LCA case study
versus ADAPT for a 85,000-m2 building.
51
Chapter 3: Application of life cycle assessment to early
stage building design for reduced embodied
environmental impacts1
Abstract
Decisions made during a building’s early design stages critically determine its environmental
impact. However, designers are faced with many decisions during these stages and typically lack
intuition on which decisions are most significant to a building’s impact. As a result, designers
often defer decisions to later stages of the design process. Life cycle assessment (LCA) can be
used to enable better early stage decision-making by providing feedback on the environmental
impacts of building information modeling (BIM) design choices. This paper presents a method
for applying LCA to early stage decision-making in order to inform designers of the relative
environmental impact importance of building component material and dimensioning choices.
Sensitivity analysis is used to generalize the method across a range of building shapes and design
parameters. An impact allocation scheme is developed that shows the distribution of embodied
impacts among building elements, and an impact reduction scheme shows which material and
thickness decisions achieve the greatest embodied impact reductions. A multi-building
residential development is used as a case study for introducing the proposed method to industry
practice. Results show that the method can assist in the building design process by highlighting
those early stage decisions that frequently achieve the most significant reductions in embodied
carbon footprint.
52
1This paper was co-authored with postdoctoral fellow Forest Flager, Assistant Professor Michael Lepech, and
Professor Martin Fischer in the journal Building and Environment. The citation is as follows: Basbagill J, Flager F,
Lepech M, and Fischer, M. (2013). Application of life cycle assessment to early stage building design for reduced
embodied environmental impacts. Building and Environment 60: 81-92.
Keywords: Life cycle assessment; Sustainable design; Embodied environmental impact;
Sensitivity analysis
Introduction
Buildings consume significant amounts of energy and materials. They account for 41% of
the total energy consumption in the United States (Fumo et al. 2010) and 38% of the nation’s
greenhouse gas emissions (USDOE 2011c). Buildings’ embodied energy, which includes
feedstock and process energy for production of building materials as well as the total fuel cycle
energy for all processes required to construct a building, may be particularly significant (Fay et
al. 2000, Bribian et al. 2009). In cases where buildings have been designed for low- or net-zero
energy, embodied environmental impacts can approach the magnitude of impacts due to
operational energy use (Citherlet 2001, Thormark 2002, Winther and Hestnes 1999).
A significant portion of a building’s life cycle impacts are determined by decisions made
in the early design stages (Cofaigh et al. 1999, Ellis et al. 2008, Kotaji et al. 2003). Choosing
materials with low embodied impacts at this stage therefore has potential to significantly reduce
a building’s life cycle impact (Lawson et al. 1995). However, evaluation of the environmental
performance of these decisions and strategies for generating alternatives that improve upon the
performance of designs are typically not performed until the design development stage
(Schlueter and Thesseling 2009). Life cycle Assessment (LCA), when applied to buildings, is a
method for predicting how a facility will perform over its lifetime, which includes raw material
extraction, manufacturing, construction, operation, maintenance, repair, replacement, and
53
demolition (ECDGEI 2007). LCA considers environmental and social impacts and is often
coupled with life cycle cost assessment methods that consider economic impacts (Norris 2001).
Commonly applied environmental indicators include global warming potential, carcinogenicity,
and resource consumption.
Life cycle assessment is commonly used in such industries as automotive design,
equipment manufacturing, and consumer product design (Spitzley et al. 2005, DeKleine et al.
2011, Keoleian et al. 2004). Compared to products produced in these industries, buildings are
unique, their lifetime is decades long, they have multiple functions, and they are locally
assembled. Adoption of LCA methods to architecture, engineering and construction (AEC)
projects has been limited due to these features. In addition, LCA methods typically require
significant time and effort for implementation. The difficulties in applying LCA to the AEC
industry have been noted by others, including obtaining complete environmental impact data for
building components, tracking material flows, and clearly defining system boundaries (Yohanis
and Norton 2006, Gluch and Baumann 2004, Lee et al. 2009). In addition, building information
modeling (BIM), which is increasingly used by AEC designers to digitally represent a facility
during the early design stages, currently lacks interoperability with LCA software (Bribian et al.
2009). An additional challenge of performing LCA during the early stages of a building project
is the complexity and large number of decisions that a designer faces. For example, the building
design process requires material and dimensioning specifications for hundreds of components.
Yet the design process is highly fragmented, with professionals working in an uncoordinated
fashion on such solutions as safety, health, serviceability, and aesthetics. Applying LCA to early
building design is therefore not straightforward, and material and dimensioning specifications are
typically deferred to engineering and construction teams in the design development stage (Kienzl
54
et al. 2003). Postponing or making changes to such decisions during this stage has been shown to
lead to significant increases in building impact (Schlueter and Thesseling 2009).
In order for LCA to be an effective early stage decision-making tool for the AEC
industry, designers must therefore be better enabled to understand which material and
dimensioning decisions most significantly determine a building’s environmental impact and
which decisions are less important. This knowledge can be part of an integrated, BIM-enabled
environmental impact feedback process, with designers focusing on decisions with large impact
during the early design stages and deferring decisions with marginal impact to later design
stages. This paper introduces a framework for providing designers with intuition on how
buildings’ embodied impacts are distributed throughout building elements. The framework is
intended for application specifically during the early design stages, when the design problem is
typically not well defined, the number of design alternatives is large, and the potential to reduce
environmental impacts is greatest. The framework utilizes a computational method that
integrates BIM software with LCA and energy analysis software, in order to quickly evaluate the
embodied impacts of thousands of building designs. Sensitivity analysis is then performed on
these results in order for designers to understand which building components’ embodied impacts
consistently contribute the largest to a building’s environmental impact across the designs.
Material choice and component dimensions, in the case of this paper a surface component’s
thickness, are selected as the two bases for demonstrating how designers can reduce a building’s
environmental impact. The framework is applied to a case study to show how impacts are
distributed throughout a building in the early design stages as well as which building component
decisions are the most important in terms of environmental impact. The framework requires a
minimal number of inputs and accommodates a range of values for massing parameters and other
55
design inputs. The inherent flexibility of the method and minimal required inputs therefore make
it useful for the early design stages.
Literature Review
Researchers and practitioners have recognized the importance of early design stages
when reducing buildings’ life cycle environmental impact. Numerous researchers have shown
that the earlier decisions are made in the design process and the fewer the changes to these
decisions at later stages, the greater is the potential for reducing the building’s environmental
impact. For example, by selecting an environmentally preferred building shape and orientation
during the early design stages, Cofaigh et al. (1999) were able to reduce a baseline design’s
environmental impact by 40%. Massing changes made during late design stages were shown by
Ellis et al. (2008) to have considerable environmental and economic cost ramifications.
Providing designers with early stage environmental impact performance feedback was
demonstrated by Schlueter and Thesseling (2009) to have strong effects on design choices,
resulting in less energy intensive buildings and increasing awareness of ways to reduce energy
consumption.
Building on this early stage design work, others have integrated BIM software with LCA
methodology and optimization techniques in order to minimize buildings’ environmental impacts
during early stage design. Wang et al. (2005b) computationally integrated BIM, LCA, energy
analysis, and optimization software in order to evaluate the environmental impact consequences
of various early stage building design parameters. A multi-objective genetic algorithm was used
to identify Pareto optimal solutions for minimized cost and environmental impact performance,
resulting in a significant reduction in global warming potential. Similarly, Hauglustaine and Azar
56
(2001) developed a computational integration method for providing BIM energy performance
feedback prior to the sketch design phase. A limited number of design variables were included,
such as those relating to a building’s geometry and thermal performance. A genetic algorithm
was used to optimize cost and energy performance, and sensitivity analysis was performed to
show the relationship between performance characteristics and changes to the design variables.
Coley and Schukat (2002) integrated BIM and thermal analysis software in developing a method
for optimizing early-stage building designs for energy performance. The method gives designers
the flexibility to choose from a set of high-quality designs based on non-optimized criteria.
LCA has also been used to estimate impacts of buildings at the early design stages.
Pushkar et al. (2005) used LCA methodology to group design variables into four clusters then
show each variable’s environmental impact bounds for each phase in a building’s life cycle.
Common building material and dimensioning alternatives were considered. Sensitivity analysis
was conducted using different fuel sources and production methods, in order to show the range
of material quantity impacts for each life cycle stage. Bribian et al. (2011) also applied LCA to
early stage building design using common building component materials and sizes, in order to
provide material selection guidelines based on minimized embodied impacts.
A number of software tools have also been developed for using LCA to assess buildings’
environmental impact at the early design stages. For example, the Athena EcoCalculator
provides environmental impact estimates of buildings based on minimal inputs (Athena 2011).
However, these tools provide no sensitivity analysis showing how building components’
environmental impacts vary over a range of design alternatives. Lack of integration with BIM
tools also reduces their utility during the early design stages.
57
Prior research has also neglected to conduct sensitivity analysis on the embodied impacts
of building component materials and dimensions for a range of design alternatives. An early
stage decision support method is lacking that shows the degree to which design choices achieve
embodied impact reductions for thousands of design variations. Massing parameters, such as
building shape and height, typically are not varied in environmental impact analyses of
buildings. Prior studies also do not include impacts due to mechanical, electrical, and plumbing
(MEP) or other service equipment or the maintenance, repair and replacement (MRR) of building
components. These limitations are addressed by the proposed method.
Methodology
Scope
The goal of the proposed methodology is to enable designers to understand the relative
environmental impact implications of building component decisions. The choices of building
component material and building component dimensions (e.g., thickness) have been shown to be
important in terms of contributing to a building’s life cycle environmental impact (Venkatarama
and Jagadish 2003, Gustavsson and Sathre 2006, Comakli and Yuksel 2004). In addition,
material and thickness choices extend to many building components, such as the foundation,
cladding, walls, floors, and duct insulation. Broadness of scope and degree of impact therefore
drive the selection of material and thickness as the two types of decisions used to determine
which building components contribute most significantly to a building’s embodied impact.
A second goal of the methodology is to create an automated or semi-automated process
that provides environmental impact feedback on many building designs. Central to the method is
the integration of BIM software with LCA, energy simulation, and sensitivity analysis software.
58
Designers manually input a range of values into the BIM for a limited number of design
variables. The BIM also requires input values for a limited number of constraints. The method
then computationally iterates through all possible design variable values, thereby creating a large
set of building designs each with an environmental impact. Data is then aggregated and analyzed
in order to determine which building components consistently contribute the most to a building’s
embodied impact. The method is not a full integration, since LCA and energy analysis data are
manually extracted from various programs prior to the iterations. Further detail on the degree of
automation used in the process is presented later in the paper.
The Uniformat 2010 classification system is used in the AEC industry to classify building
components within building element categories (Construction Specifications Institute 2010).
These elements refer broadly to the parts of a building. Uniformat elements within the project
scope are: Substructure (A), Shell (B), Interiors (C), and Services (D). The remaining elements
(Equipment and Furnishings (E), Special Construction and Demolition (F), and Sitework (G)) are
not considered, since these decisions relate to interior aesthetics, require specialized knowledge
of site conditions, or otherwise involve decisions that would be impractical to make by designers
before the design development stage. A detailed description of this classification framework is
presented in the next section.
Figure 15 schematically shows the complete life cycle of a typical building. The shaded
area shows those phases that are included in this method. Operational phase impacts due to
HVAC, lighting, plug loads, and water use have been excluded in the scope. Rather, the research
focuses on embodied impacts due to building component material and dimension choices. The
operational phase is limited to embodied impacts due to maintenance, repair, and replacement
(MRR) of building components. As such, decisions determining a building’s impact due to
59
operational energy use from utilities are assumed to be independent from decisions determining
embodied impacts. The intention of the study is to make as granular as possible those decisions
determining building embodied impacts. Including the effects of building materials and
thicknesses on operational energy would dilute this feedback. Therefore, operational energy use
beyond MRR is excluded from the study. Coupling of these inter-life cycle-phase design
decisions is a topic of future planned research. Demolition and on-site construction have also
been excluded, since impacts associated with these phases have been shown to be difficult to
calculate (Schoch et al. 2011) and small when compared with other phases (Scheuer et al. 2003).
Researchers have identified several impact categories that are useful in measuring the
environmental impact of buildings. These impact categories include global warming potential,
non-renewable energy consumption, human toxicity, acidification, and eutrophication, among
others (Jolliet et al. 2003). Although the authors recognize the importance of all of these
categories in comprehensively assessing environmental impact, this methodology considers only
global warming potential. The metric used for this purpose is carbon dioxide equivalents (CO2e)
using the relevant 100-year global warming potential, which measures the total amount of
greenhouse gas emissions of the building, considering all relevant sources (Wright et al. 2011).
The building owner or designer could add other impact categories to the analysis as required.
Figure 15 – Building life cycle phases included in scope.
60
Building component classification framework
The framework used to structure the building component decision-making process is
based on Uniformat 2010. Table 4 outlines the assemblies and their sub-components for each of
the four Uniformat elements. Material choices for each component are determined using
RSMeans (RSMeans 2007) and Athena EcoCalculator (Athena 2011). These choices are not
meant to be exhaustive but rather representative of common materials for each component. The
appendices enumerate the material choices and their properties for each building component.
These properties include material densities and embodied CO2e factors, or the amount of carbon
dioxide equivalents associated with materials’ feedstock energy, energy required to process the
materials into building components, and fuel cycle energy for all pre-operational processes. The
“Software Integration” sub-section of the “Implementation” section describes the software from
which these factors are obtained and why these programs are chosen.
The building component classification framework includes thickness as a dimensioning
variable. Specifications from several construction material and equipment supplier sources are
used to determine thickness ranges. These sources are listed in the description for Table 4. The
smallest minimum value and largest maximum value are identified across all sources for each
sub-component then placed into the table. Thickness ranges are not articulated for every
component, namely those whose size determinations are difficult to reduce to one single
thickness parameter and/or best quantified by structural analysis methods applicable to later
design stages.
61
Table 4 – Building component classification framework. Sources used for thickness ranges (by
Uniformat element): A: (RSMeans 2007), (ACI 2004); B: (RSMeans 2007); C: (RSMeans 2007).
bThickness
Uniformat element Assembly aSub-components Number of
material choices Minimum (m) Maximum (m)
cA: Substructure piles piles, vapor barrier, caps, slab-on-
grade, grade beam, rebar, formwork
2, 2, 1, 1, 1, 1, 1 0.1 0.4
footings footings, vapor barrier, slab-on-grade,
grade beam, rebar, formwork
1, 2, 1, 1, 1, 1 0.1 0.4
mat foundation foundation, vapor barrier 1, 2 0.2 1.8
B: Shell columns and beams 10 n/a n/a
floor structure 12 n/a n/a
roof roof structure, membrane, insulation,
paint
10, 5, 1, 1 n/a n/a
stairs stairs, railings 3, 3 n/a n/a
cladding 7 0.02 0.08
exterior walls wall structure, insulation, membrane,
gypsum, paint
5, 1, 1, 1, 1 n/a n/a
glazing glass, polyvinyl butyral, frame,
hardware
1, 1, 5, 1 0.007 0.02
doors door, hardware 3, 1 n/a n/a
C: Interiors partitions partition structure, gypsum, paint 2, 1, 1 0.2 0.6
doors door, hardware 2, 1 n/a n/a
wall finishes covering, paint 2, 1 0.005 0.02
flooring surface, insulation 9, 13 0.1 0.2
ceiling plaster, gypsum, paint 1, 1, 1 0.006 0.02
dD: Services mechanical 17 sub-components e13 n/a n/a
electrical 16 sub-components 1 n/a n/a
plumbing 23 sub-components 1 n/a n/a
fire 4 sub-components 1 n/a n/a
conveying elevator 1 n/a n/a aTotal number of sub-components listed in the table equals 107. Of these, 102 are distinct sub-components. Five are double counted and occur in multiple components:
vapor barrier, slab-on-grade, grade beam, rebar, and formwork. This double counting occurs because the substructure consists either of piles (seven sub-components),
or footings (six sub-components, five of which are present in the piles sub-components), or mat foundation (two sub-components, one of which is present in piles and
footings).
b Thickness ranges correspond to bold sub-component and all material choices for that sub-component. For assemblies with multiple bold sub-components, ranges
represent combined thicknesses.
c Substructure consists of one of the three listed assemblies. Remaining three elements consist of all listed assemblies.
d Large numbers of services sub-components preclude enumeration.
e Duct insulation is a mechanical sub-component with 13 material choices. Remaining mechanical sub-components have one material choice.
62
Analysis process
The general steps of the proposed integrated, BIM-enabled embodied impact feedback
method are shown in Figure 16. The arrows in the figure represent data dependencies between
process steps.
Figure 16 – Software integration for embodied impact feedback method.
The process begins with a designer manually creating a building information model
(DProfiler 2012). Table 5 lists the BIM inputs in terms of constraints, variables, and
assumptions. The constraints are necessary for determining the maintenance, repair, and
replacement (MRR) schedule. For example, a building of one type and size located in a hot and
dry climate may have very different MRR impacts than a building of another type and size
located in a cold and wet climate. Minimum and maximum values are required for the variable
inputs, and no material or size specifications are required. Assumptions are automatically
programmed into the BIM but may be modified by the designer.
63
Table 5 – Required inputs, variables, and assumptions for building information modeling-
enabled embodied impact feedback method.
Required Inputs
Location
Building type
Gross floor area
Variables
Number of buildings
Number of floors
Length and width parameters determining
building footprint
Window-to-wall ratio (WWR)
Assumptions
Footing depth
Bay spacing
Floor-to-floor height
Service life
Once the BIM is created, the automated design-feedback process begins. Calculation of
pre-operational carbon footprint is the first step in this process. The use of material quantity
formulas is an essential part of this calculation. These formulas depend on the BIM inputs
outlined in Table 5, and many were developed in consultation with senior estimators at Beck
Technology, an AEC firm. Beck aggregated data from bill of material quantities on
approximately one dozen of their building projects. The number of formulas totals 102 and
equals the number of distinct building sub-components outlined in Table 4. Appendix 2 provides
all formulas for each of the four building elements. The formulas are used to calculate the
minimum and maximum possible quantities for each building component material. Inputs are
material choice, minimum thickness, and maximum thickness as given in Table 4 as well as
gross floor area and all variables and assumptions outlined in Table 5. Pre-operational carbon
footprint is calculated by multiplying each quantity by the embodied CO2e factors in the
Appendix 2 and summing the resulting impacts from the 102 sub-components.
64
Calculation of pre-operational embodied impacts due to the service equipment is an
important sub-step of this process. These impacts comprise 61 of the 102 building component
impacts and are calculated by sizing each piece of service equipment according to peak building
load. An energy simulation program performs this step as follows (eQUEST 2010). Inputs from
Table 5 are automatically passed from the BIM to the energy simulation program. Thermal zones
are defined in the resulting energy simulation model as well as standard assumptions regarding
building occupancy and HVAC system controls (ASHRAE 2009). The program then calculates
peak building load from these inputs and assumptions, and the result is an input to the 61
material quantity formulas. Equipment supplier documentation is used to determine whether each
piece of service equipment typically increases in size as peak building load increases. Material
quantities for those pieces of equipment that typically increase in size are scaled linearly
according to peak building load. The resulting scaled and non-scaled material quantities are then
multiplied by the CO2e impact factors in Appendix 2 to determine the service equipment’s pre-
operational embodied impact.
A maintenance, repair and replacement schedule is used to determine the operational
phase impacts associated with the building components. The MRR schedule is determined by
manually entering all the constraints as well as the service life assumption from Table 5 into an
online facility operations reference database (CostLab 2011). The program returns each
component’s MRR dollar costs for every year of the building’s operation. Equipment supplier
documentation is then used to look up a typical material, material quantity, and cost for each
component. Material quantities are then calculated by combining the MRR cost outputs from the
operations database with the data from the supplier documentation. Material quantities are then
scaled according to peak building load as described in the previous paragraph. These quantities
65
are multiplied by a CO2e impact factor in a similar fashion to the pre-operational impact
calculation to determine the minimum and maximum MRR operational carbon footprint. The life
cycle embodied carbon footprint is then calculated by summing the pre-operational and MRR
operational CO2e impacts.
Sensitivity analysis software is then used to search the design space for the minimum and
maximum possible embodied impacts due to each building component by varying the input
variables from Table 5 using the pre-defined ranges (ModelCenter 2008). The entire space is
searched and the number of designs generated equals the product of the number of choices for
each variable in Table 5. By generating a large number of designs, the method therefore shows
how each building component’s embodied impact varies across a wide range of input parameters.
The results present designers with an impact allocation scheme, which shows the
minimum and maximum embodied impacts possible for each of the building components across
all designs considered. The maximum possible embodied impact for a given design is first
determined by selecting the material and thickness with the largest impact. A building
component’s minimum impact is then determined by selecting the material and thickness with
the smallest impact for that component. The material and size with the largest impact are chosen
for all other components. Maximum impacts are determined in a similar fashion. Minimum and
maximum impacts are expressed as a percentage of a given design’s maximum possible
embodied impact.
Designers are also presented with an impact reduction scheme, which shows the degree to
which each building component achieves reductions in embodied impact due to changes in both
material and thickness. The maximum embodied impact reduction due to a change in material is
calculated by subtracting the smallest possible impact from the largest possible impact. The
66
maximum embodied impact reduction due to a change in thickness is calculated in a similar
fashion. The reductions are expressed as a percentage of the entire building’s maximum possible
embodied impact for a given design.
The material and thickness impact reductions for each building component are calculated
for all designs. The lowest and highest of these maximum values represent the range of
maximum impact reductions possible across all designs for each building component. Each
design’s building component impact reductions are then summed, and the designs with the
minimum and maximum sums represent the lowest and highest maximum impact reductions for
the whole building. Histograms showing the distribution of maximum impact reductions for the
building are then generated. One histogram is created for material decisions. Another histogram
is created for thickness decisions. Distributions for building component material and thickness
impact reductions are overlayed showing the degree to which each decision reduces a building’s
mean embodied impact. These distributions are generated in order of those that reduce the
histogram’s mean embodied impact the most to those that reduce the mean the least.
Implementation
Problem formulation
A mid-rise multi-building residential development is used as a case study to demonstrate
the utility of the proposed method to industry practice. The development plan was provided by
Beck Technology as a retrospective case study. At the time of publication, the complex was in
the early design stage. The case study thus provided an opportunity to show which decisions
67
could reduce the embodied environmental impact of the development the most in terms of
building component material and thickness changes.
The problem formulation is described in Table 6 in terms of required inputs, variables,
and assumptions. Values for the required inputs and assumptions are given as well as choices for
the variable inputs. The total number of possible designs is 5,832.
The development has a total floor area of 50,000 m2. The variable “Number of
buildings” refers to the choices for the number of individual buildings in the housing
development. For a given design, each building is identical in terms of values selected for the
variables as well as the material and thickness selected for each building component. Six shape
parameters determine the form of the building. Peak building load is used to size the service
equipment and determine the MRR impacts for every possible combination of number of floors
and number of buildings, using the process described in the previous section. Embodied carbon
footprint ranges and reductions due to building component material and thickness changes are
calculated in terms of CO2e as described in the previous section. Internal loads and the weekly
operating schedule are determined from the 2009 ASHRAE Fundamentals (ASHRAE 2009).
Since orientation is constrained in this study to a single value, orientation is not included as a
variable in the problem formulation and therefore does not influence changes in embodied
impact.
68
Table 6 – Problem formulation showing required inputs, variable values, and assumptions.
Required Inputs
Location = confidential
Building type = residential mid-rise
Gross floor area = 50,000 m2
Variables
Number of buildings = 3, 4
Number of floors = 5, 6, 7, 8
Shape parameters:
a = 0m, 15m, 30m
b = 10m, 20m, 30m
c = 0m, 15m, 30m
d = 10m, 30m, 50m
e = 0m, 15m, 30m (a)
f
Window-to-wall ratio = 0.15, 0.325, 0.50
Assumptions
Footing depth = 2m
Bay spacing = 9m
Floor to floor height = 3.6m
Service life = 30 years (a)
Shape parameter “f” is dependent on the other five
shape parameters. Minimum possible value for “f” is 15,
and maximum possible value is 30.
Software integration
The various software components used to implement the BIM-enabled embodied impact
feedback method are shown in Figure 16. Descriptions of the software, pros, cons, and
alternatives are discussed in the following section.
DProfiler is used as the selected BIM software (DProfiler 2012). The program is a
conceptual level building modeler and is useful for early design stages. The program provides
69
detailed material quantity and energy analysis feedback given minimal building design inputs.
Heuristics programmed into the software compute building component material quantities. The
program outputs a detailed BIM with fewer inputs than is typically required by alternative BIM
programs such as Revit (Revit 2013). Another advantage of DProfiler is that the BIM created by
the program is automatically exported to eQUEST, an energy simulation program (eQUEST
2010). This export occurs entirely within DProfiler without any user interface with eQUEST.
Linkage of the BIM and energy model configuration is therefore useful for the method developed
in this paper since users receive energy analysis feedback on BIM inputs without having to
separately create an energy simulation model.
A number of limitations exist with the computational architecture shown in Figure 16.
DProfiler has limitations in terms of the range of geometric forms that it can create. DProfiler’s
building wizard tool is used to implement the method in Figure 16, and BIM geometries are
limited to fairly simple orthogonal building shapes such as the H-shaped figure shown in Table
6. Complex or freeform architectural shapes are currently not accommodated by the software.
Limitations of eQUEST include long run times, meaning it may take several days to execute the
method shown in Figure 16 for thousands of runs. The program also does not model natural
ventilation, operable windows, or thermal comfort. However, the proposed method includes only
MRR operational impacts and so is not affected by these limitations.
SimaPro is the LCA software used to obtain many of the CO2e impacts outlined in
Appendix 2 (SimaPro 2010). As described in the analysis process section, these factors are
critical for converting building component material quantities into embodied impacts. SimaPro is
chosen because the software contains impact factors for many different building materials, and
these impacts correspond to the life cycle phases scoped in the method and outlined in Figure 15.
70
SimaPro is provided the quantity of building component material as an input and returns kg
CO2e per kg of material. A limitation of the program is that the software does not
computationally integrate into the method proposed in Figure 16. Instead, the CO2e impact
factors must be manually extracted and placed into Excel.
Athena EcoCalculator is also used to obtain some of the CO2e impact factors in Appendix
2 (Athena 2011). The program is helpful for estimating impacts of structural assemblies, such as
columns and beams, interior walls, or the roof structure. The tool scales impacts by building
gross floor area rather than a building component thickness parameter. This scaling method is
especially useful during the early stages of building design when very limited information is
known about a building such as structural requirements. The tool integrates well with the
proposed method, since gross floor area is a required input in Table 5. As with SimaPro,
however, the program does not computationally integrate into the proposed method and CO2e
impact factors are manually extracted and placed into Excel.
CostLab is the online facility operations reference database used to calculate MRR
impacts (CostLab 2011). The software creates an MRR schedule comprised of a detailed list of
building components within its database. LCA tools do not provide such detailed MRR
information. CostLab works well with the proposed method since the program takes as inputs the
constraints outlined in Table 5. Since the program is not an LCA tool, impacts are outputted in
terms of dollar costs instead of CO2e for every component for every year of a building’s
operation. Therefore, the program is cumbersome to work with since dollar costs must be
converted to embodied impacts as described in the analysis process section. In addition, since the
program does not computationally integrate into the proposed method, BIM inputs are manually
entered and cost outputs are manually extracted and placed into Excel.
71
Excel is used to perform simple mathematical operations on the material quantities that
determine the embodied carbon footprint of each building design (Excel 2007). Data is manually
extracted from the various software sources as described above and Excel formulas are used to
perform the operations for each iteration.
ModelCenter is the sensitivity analysis software used to integrate DProfiler and Excel
into a common environment (ModelCenter 2008). The program’s ScriptWrapper utility enables
users to integrate or “wrap” different software components using a scripting language. In this
work, DProfiler is wrapped with ModelCenter to allow the programs to communicate with each
other in terms of inputs and outputs. Phoenix Integration, Inc. has wrapped Excel with
ModelCenter, thereby allowing DProfiler and Excel to communicate with each other. The
program’s “Design Explorer” tool allows sensitivity analysis to be performed on a large design
space by computationally iterating through all possible designs. Ranges for design variables are
articulated either as a continuous range or as a discrete set. Table 6 lists the discrete set of
variable choices used for the case study. The program iterated through all 5,832 designs in order
to determine the minimum and maximum embodied impacts for each building component.
Choosing every single integer value for each variable in Table 6 between the minimum and
maximum given values would require significant memory and several weeks of computer
processing time. Due to these constraints, only the lower bound, upper bound, and mid-point
value of each variable range were chosen as shown in the table. Limiting the selection of
parameter values in this way does not affect results, since sensitivity analysis is concerned with
determining only minimum and maximum possible embodied impacts.
72
Results and Discussion
The proposed method was applied to the case study project to determine in which
building components embodied impacts are concentrated as well as which design decisions
achieve the greatest reductions in embodied impact.
Table 7 presents the impact allocation and impact reduction schemes described in the
analysis process section. Tables 8 and 9 regenerate the impact reductions after each material and
thickness decision, respectively, are made. The results show how the range of embodied impacts
is steadily reduced as decisions are made in order from those achieving the greatest embodied
impact reductions to those achieving the least reductions. The results are not meant to suggest
that making decisions in a certain sequence – from those achieving the greatest impact reduction
to those achieving the least – can help designers arrive at a best or improved design in terms of
lowered embodied impact. Rather, the results are meant to help designers visualize the potential
reductions for each building component so that they can understand which decisions consistently
contribute to a building’s embodied impact then focus on making choices for those decisions that
matter the most.
The impact allocation scheme shows that the range of impacts is very large for building
components in all four of the elements. The total embodied impact can potentially be
concentrated in any of these elements, as long as materials and thicknesses with minimum
embodied impact are chosen for components in each of the other elements. Each element may
contribute over 50% of the development’s embodied impact, depending on the design under
consideration.
In terms of the impact reduction scheme shown in Table 7, both material and thickness
changes can potentially lower embodied impacts by large amounts for many building
73
components in the substructure (Uniformat Element A), the shell (Element B), and the interiors
(Element C). Fourteen of the 21 components’ maximum impacts are greater than 10% of the total
embodied impact, and six are greater than 40%. The largest impact changes are seen in cladding
material (reduced from 35% to approximately 7% of the maximum possible total embodied
impact), cladding thickness, piles material, glazing material, and flooring material. Designers
should be aware of the material and thickness choices for these building components during the
early design stages. In contrast, changes to materials and thicknesses are not important for all of
the services components within Uniformat Element D, suggesting designers need not focus on
these decisions during the early design stages.
Tables 8 and 9 regenerate the impact reductions after each material and thickness
decision, respectively, is made. The results show how the range of embodied impacts is steadily
reduced as decisions are made in order from those achieving the greatest embodied impact
reductions to those achieving the least reductions. The tables confirm the results from Table 7.
Tables 8 and 9 show that large reductions can be achieved for both material and thickness
decisions for building components in Uniformat Elements A, B, and C. In terms of material
choice, the greatest reductions are achieved for cladding, substructure, partitions, and flooring
surface. In terms of thickness choice, the greatest reductions are achieved for cladding, flooring
surface, ceiling, and wall finishes.
74
Table 7 – Impact allocation scheme and impact reduction scheme.
Impact Allocation Scheme Impact Reduction Scheme
(as % of total embodied impact)
Material change
(as % of max embodied impact)
Thickness change
(as % of max embodied impact)
Uniformat
element
Assembly Minimum
impact
Maximum
impact
Min impact
reduction
Max impact
reduction
Min impact
reduction
Max impact
reduction
Whole building 62.95 74.94 19.98 37.33
A: Substructure 0.21 55.07 7.77 19.65 0.33 2.03
piles 1.35 51.51 7.77 19.65 0.33 0.63
footings 11.13 55.07 n/a n/a 0.33 0.63
mat foundation 0.21 10.68 n/a n/a 0.85 2.03
B: Shell 2.22 78.97 21.17 49.64 6.03 27.27
columns and
beams 0.27 19.18 2.59 4.50 n/a n/a
floor 0.29 25.77 3.74 7.31 n/a n/a
roof 0.02 2.78 0.36 0.47 n/a n/a
stairs 0.00 2.13 0.25 0.50 n/a n/a
cladding 0.01 68.47 6.78 35.08 5.10 26.39
exterior walls 0.24 30.38 0.97 5.15 n/a n/a
glazing 0.35 55.61 0.60 7.52 0.22 3.17
doors 0.00 0.13 0.01 0.03 n/a n/a
C: Interiors 5.42 70.27 14.98 26.23 8.61 13.82
partitions 0.92 39.40 6.81 12.20 n/a n/a
doors 0.00 0.44 0.05 0.09 n/a n/a
wall finishes 0.89 23.40 1.35 1.81 2.02 2.5
flooring 0.18 44.69 6.77 12.13 3.76 6.73
ceiling 1.95 30.78 n/a n/a 2.69 4.81
D: Services 8.06 70.27 1.09 1.94 0.00 0.78
mechanical 4.12 42.62 1.09 1.94 0.00 0.57
electrical 2.96 24.10 n/a n/a 0.00 0.2
plumbing 0.84 6.72 n/a n/a 0.00 0.01
fire 0.03 0.21 n/a n/a n/a n/a
conveying 0.05 0.38 n/a n/a n/a n/a
75
Table 8 – Ranking scheme for material decisions achieving embodied impact reductions.
Material
decision number Assembly (specific component)
(a) Minimum
impact reduction
(a) Maximimum
impact reduction
1 Cladding 38.86 59.70
2 Substructure (vapor barrier, piles) 29.71 44.23
3 Partitions 22.89 32.03
4 Flooring surface 16.11 21.75
5 Floor structural assembly 10.76 17.28
6 Column and beam
structural assembly
6.50 14.52
7 Window assembly 5.39 8.37
8 Wall assembly 3.14 4.78
9 Wall finishes 1.41 2.97
10 Mechanical system
(duct insulation)
0.71 1.03
11 Roof assembly 0.32 1.00
12 Stairs 0.07 0.12
13 Interior doors 0.01 0.03
14 Exterior doors 0 0 (a) Material impact reduction ranges reflect the reduction in embodied impact possible after each decision has been made.
Therefore, no impact reduction is possible once the final decision has been made.
Table 9 – Ranking scheme for thickness decisions achieving embodied impact reductions.
Thickness
decision number Assembly
(a)Minimum
impact reduction
(a)Maximum
impact reduction
1 Cladding 10.67 16.67
2 Flooring surface 6.69 10.09
3 Ceiling 3.82 6.98
4 Wall finishes 1.56 4.69
5 Substructure 0.24 3.66
6 Window assembly 0.00 0.78
7 Mechanical system 0.00 0.21
8 Electrical system 0.00 0.01
9 Plumbing system 0 0 (a)
Thickness impact reduction ranges reflect the reduction in embodied impact possible after each decision has been made.
Therefore, no impact reduction is possible once the final decision has been made.
Figures 17 and 18 are histograms illustrating the distributions of embodied impact
reductions across all 5,832 designs before any decisions have been made, after the first decision
achieving the greatest reduction has been made, and after the second decision achieving the next
greatest reduction has been made for material and thickness choices, respectively. The “Analysis
process” section described how these histograms are generated. The impact ranges for each
decision correspond to the minimum and maximum impact reduction values in Tables 8 and 9.
76
Impact Reduction Due to Material Decisions
Impact reduction (as % of max embodied impact)
Num
ber
of
desi
gns
No material decisions made
Decision achieving greatest reduction: Cladding
Decision achieving second greatest reduction: Substructure
The rightmost distributions in Figures 17 and 18 show the potential for reducing a building’s
maximum total embodied impact before any material or thickness decisions have been made. For
Figure 17, anywhere from a 63% to 75% reduction in the building’s maximum total embodied
impact is possible, depending on the particular set of design parameters selected from Table 6.
Once the cladding material decision has been made, the impact due to cladding is substracted
from the total impact, and between 39% to 60% of the remaining embodied impact can be
reduced. This process continues in a similar fashion for the remaining 13 material decisions until
no impact reduction is possible once the final decision has been made. Results for the thickness
decisions are presented in a similar fashion, with anywhere from a 20% to 37% reduction in the
building’s total embodied impact possible depending on the design configuration. The
histograms allow designers to visually determine during the early design stages which building
components are most important in terms of achieving embodied impact reductions through
material and thickness choices.
\
Figure 17 – Embodied impact reduction due to material decisions.
77
Impact Reduction Due to Thickness Decisions
Impact reduction (as % of max embodied impact)
Num
ber
of
desi
gns
No thickness decisions made
Decision achieving greatest reduction: Cladding
Decision achieving second greatest reduction: Flooring Surface
Figure 18 – Embodied impact reduction due to thickness decisions.
Tables 7-9 and Figures 17 and 18 together show that significant embodied impact
reductions can be achieved in the substructure, shell, and interiors of the case study building. The
largest reductions for material changes are cladding, substructure, and partitions, and the least
important material decisions are related to the doors, stairs, and service equipment. The largest
reductions for thickness dimension changes are cladding, flooring surface, and the ceiling, and
the smallest reductions are for the window assembly and service equipment. The impact
reduction schemes for material and thickness dimensions taken together suggest that significant
reductions in the building development’s embodied impact cannot be achieved by making
decisions for the wall finishes or service equipment.
78
The distributions show the relative importance of making a decision for one building
component over another. Designers are provided with intuition on which decisions frequently
achieve significant embodied impact reductions. By postponing material and thickness decisions
achieving smaller reductions to the design development stage, designers would avoid expending
effort on inconsequential decisions during the critical early design stages. The method ultimately
allows designers to focus their efforts during the early design stages on those decisions that are
most likely to decrease a building’s life cyle embodied environmental impact.
Conclusions
A BIM-enabled decision support method is proposed that helps designers predict which
decisions most critically determine a building’s embodied impact. The automated method
integrates BIM, LCA, energy simulation, MRR scheduling, and sensitivity analysis software.
The framework is well suited for the early design stages as very few inputs are required, and the
method can quickly iterate across many building designs thereby presenting a number of design
alternatives.
A case study analysis is presented in order to show how designers can understand which
building component decisions consistently contribute the largest to a building’s embodied
impact. Results are presented in the form of an impact allocation scheme, an impact reduction
scheme, and histograms showing the distributions of embodied impacts for many designs. The
results rank building components from those achieving the greatest to least embodied impact
reductions. A building’s embodied impact can potentially be concentrated in the substructure,
shell, or interiors. Embodied impacts due to service equipment are small, whereas cladding
material and thickness choices are consistently the most significant, regardless of building design
79
configuration. A designer should focus during the early design stages on these decisions that
achieve a large embodied impact reduction and defer less important decisions to the design
development stage.
The scope of this method is limited to building components for which dimensional
thickness ranges can be predicted at the early design stages. Future work will consider additional
sizing parameters besides thickness in order that structural components and service equipment
may be included in the sizing decisions. Operational impacts do not include impacts from HVAC
or lighting equipment, plug loads, or water use. Future research will consider the effects that
orientation and thermal properties of the building envelope have on operational energy use in
order to develop a more comprehensive understanding of the relationship between early stage
design decisions and life cycle environmental impacts. Finally, the validation of the method is
currently limited to a single case study involving a particular building type, size, location, and
geometry. Additional case study applications will be required to comment more generally on the
performance and robustness of the proposed decision support method.
80
Chapter 4: Evaluating embodied versus operational
environmental impact trade-offs of conceptual
building designs1
Abstract
The conceptual building design phase involves decisions that have significant life cycle
environmental impact implications. Decisions made during this phase may be costly to change
during later design stages. By understanding how design alternatives compare in terms of carbon
footprint, designers can make choices resulting in buildings with lower impacts. An important
part of this process is understanding the embodied versus operational impact trade-offs of design
decisions. Improving the operational performance of a building may negatively impact its
embodied performance. Similarly, decisions weighted towards embodied energy reduction may
greatly increase operational energy. This paper presents an automated method that employs a
multi-objective genetic algorithm to analyze the trade-offs between embodied and operational
impacts for a range of building design alternatives. A residential case study is used to evaluate
trade-offs for several decisions, including window-to-wall ratio, shading devices, and glazing
thickness. Solutions yielding significant reductions in the building’s carbon footprint are
identified. The method also determines the degree to which strategies weighted towards
minimizing embodied or operational impacts yield lower total impact. Designers are informed of
the relative importance of embodied versus operational impacts for several decisions and enabled
to make decisions leading to less energy intensive and more sustainable buildings.
81
1This paper was co-authored with Assistant Professor Michael Lepech and submitted to Energy and Buildings in
October 2013.
Keywords: Conceptual building design, embodied energy, operational energy, environmental
impact, optimization
Introduction
The building industry is a primary consumer of energy and natural resources, and
building energy consumption has reached between 20% and 40% of total energy use among all
industry sectors in developed countries (Perez-Lombard et al. 2008). The problem extends to
developing countries in such regions as Asia and the Middle East, where energy consumption
has risen dramatically in recent decades due in large part to the increase in building construction
(Aboulnaga 2006). High rise buildings built in such urban centers as Shanghai and Doha have
contributed to developing countries’ carbon emissions per capita rising to twice those of western
developed countries in the past two decades (Kazim 2007).
Buildings’ life cycle environmental impacts are determined by embodied and
operational energy use (Ding 2004, Crowther 1999). Embodied energy is energy sequestered in
building materials during production processes, on-site construction, maintenance, repair,
replacement, and end-of-life scenarios. Operational energy is energy expended during the life of
the building, and this includes energy for heating and cooling, lighting, water use, and operation
of appliances (Dixit 2010).
Operational energy typically dominates the energy profile of buildings, contributing
80% or more of the life cycle energy consumption (RAIA 2004). Research has primarily focused
on ways to reduce operational energy by improving the thermal efficiency of the building
envelope (Der-Petrossian 2000, Scheuer and Keolian 2002). One problem with this approach is
82
that building designs with highly insulated wall and window assemblies improve the operational
energy efficiency of buildings at the expense of using materials with high embodied energy
and/or increasing building material quantities (Huberman and Pearlmutter 2008). Such energy
efficient strategies increase the energy requirements of the material production phase, and these
may outweigh the operational energy savings and ultimately increase the life cycle
environmental impact. The role of embodied energy and its relationship to operational energy
becomes increasingly more important, especially since embodied energy has been found to
contribute up to 60% of a building’s life cycle energy use (Langston and Ding 2001, Thormark
2002, Yohanis and Norton 2002). This is also apparent as strategies for reducing operational
energy become more prominent in building design (RAIA 2004, Cole and Kernan 1996, Mumma
1995).
Building designers often make several decisions exhibiting embodied versus operational
energy trade-offs. The decisions typically involve components in the building envelope, since
components here have been found to account for nearly 30% of buildings’ embodied energy
(Atkinson et al. 1996, Yohanis and Norton 2002). Here, embodied and operational energy often
have an inverse relationship with each other. For example, increasing glazing thickness will
improve a window’s thermal resistance, thereby decreasing the amount of energy needed to cool
a building. This increased thickness comes at the expense of reducing the glazing’s thermal
transmissivity, which increases the amount of energy needed to heat the building. In addition, the
increased thickness increases the building’s embodied energy, since a greater amount of glazing
material is produced. Similarly, energy efficient strategies may reduce a building’s window-to-
wall ratio (WWR) may improve a façade’s thermal performance during the operation of the
building. However, the energy consumed during the material production phase may increase if a
83
building’s cladding material has a greater embodied energy per kilogram than its glazing
material. The net embodied energy will be greater than for the higher WWR. If this increase in
embodied energy is greater than the embodied energy savings, then the net effect of reducing the
WWR is to increase the life cycle environmental impact of the building. Given Earth’s surface
temperature has increased by between 0.3 and 0.6°C over the last 100 years (Houghton et al.
1995), such trade-offs may become increasingly important in regions with extremely hot climate
conditions. For example, cities on the Arabian Peninsula are likely to face summer mean
temperatures often exceeding 40° C (Jentsch et al. 2010). In regions with such harsh climatic
conditions, the cooling needs of poorly designed buildings may be considerable (Al-Homoud
2005). Highlighting the energy trade-offs associated with building envelope decisions can avert
this problem by allowing designers to make design decisions optimized for building energy
reductions, thereby leading to lower carbon emissions.
Designers also need a systematic method of exploring the large number of design
possibilities associated with these energy trade-offs. Design teams typically explore building
design alternatives using an iterative, trial and error process that is performed manually.
Exploring the design space in this way can be difficult, given the large number of alternatives
and the complicated interactions between many design variables. Designers also often reason
backwards using a deductive approach to make large problems more manageable (Ahmed et al.
2003). This approach can leave large areas of the design space unexplored. Automated
procedures to optimize aspects of building design can reduce the effort needed to systematically
search the complete design space for high-performing design solutions (Flager et al. 2012).
This paper introduces an optimization method that provides designers with information
on the embodied versus operational energy trade-offs of building design decisions for a range of
84
design alternatives. The method calculates impacts for several conceptual design decisions,
including shape, massing, site orientation, and building materials. The method also generates
many design alternatives, and probabilistic analysis of these alternatives determines whether
minimizing only for operational energy and/or embodied energy can serve as a proxy for
minimizing for both embodied and operational energy. In this way the design problem may be
simplified, and designers can focus effort on decisions that minimize only operational or
embodied energy.
Related Studies
A number of studies have looked at methods for evaluating the environmental impact
implications of building design decisions. Radford and Gero’s (1980) research introduced the
idea of trade-off diagrams using a Pareto optimality approach. By moving beyond trial and error
methods and quantitatively relating the performance of design variables to each other, the
research offered a prescriptive method for improving building performance. This approach was
used to show trade-offs between day-lighting and peak summer temperature as well as associated
values for a set of design variables, including glass type, window size, and sun shade projection
(Gero et al. 1983).
Recent studies have applied this method when looking specifically at the embodied
versus operational energy trade-offs of building design decisions (Hacker et al. 2008, Thormark
2006, Rai et al. 2011, Ardente et al. 2008). Huberman and Pearlmutter (2008) looked at
embodied and operational energy consumption of a building in a desert region in Israel when
optimizing building materials according to minimum life cycle energy requirements. Tuhus-
Dubrow and Krarti (2010) included building shape variation in using a genetic algorithm to
85
optimize a building envelope for minimized energy use. Pierquet et al. (1998) looked at wall
system trade-offs for buildings in cold climates in the United States and found that systems made
from non-renewable materials often performed more poorly than systems made with natural
materials. Radhi (2010) also looked at wall systems and incorporated sensitivity analysis to show
the economic, operational, and embodied energy performance trade-offs. The research concluded
that trade-offs are not straightforward and that a careful approach should be taken when choosing
building materials to reduce carbon emissions.
Wang et al. (2005a, 2005b) also evaluated building design alternatives in terms of both
embodied and operational energy consumption. The variables in the study included building
orientation, aspect ratio, window-to-wall ratio and wall construction type. A multi-objective
genetic algorithm was applied to identify Pareto optimal solutions considering both energy
consumption and life cycle cost objectives. The energy simulation performed did not consider
natural lighting or define multiple thermal zones within the building. Maintenance, repair and
replacement impacts were also not included in the analysis.
Although many of the studies used a prescriptive approach to evaluate building design
decisions, none explicitly optimized embodied versus operational energy impacts for a broad set
of design variables. The research here fills this gap by using an automated Pareto optimization
approach to consider a comprehensive set of variables exhibiting this trade-off. The method
evaluates a range of building shapes and massing alternatives, in order to increase the utility of
the method specifically during the conceptual design phase as well as to generalize the findings
across a range of building prototypes. The research relies on probability distributions in
conveying the results, in order to show the likelihood of a given design choice associated with a
trade-off variable being found among designs optimized for environmental impacts. Such a
86
method allows designers to quickly determine how strongly a design choice yielding either an
embodied or operational impact reduction correlates with a reduction in the building’s overall
impact. The research intends to help designers understand the complicated energy trade-offs of
building design choices and lend careful calibration to conceptual design decision-making in a
way that will help minimize buildings’ life cycle environmental impacts.
Methodology
The goal of the proposed methodology is to enable designers to accurately assess
trade-offs between embodied and operational impacts for a predefined range of building design
alternatives. To illustrate the potential to optimize for environmental performance across a large
number of building systems, the building’s substructure, façade, interior, and service equipment
are included in the analysis.
The shaded area in Figure 19 shows the phases of the building life cycle that are
considered in the research. Evidence from previous research suggests the included phases,
namely raw material acquisition, building material production, maintenance, repair, and
replacement, and operation account for over 95% of a building’s life cycle environmental impact
(Cole and Kernan 1996). Demolition has not been included since impacts associated with this
phase have been shown to be difficult to calculate (Pushkar et al. 2005, Schoch et al. 2011) and
small when compared with other phases (Scheuer et al. 2003).
Researchers have identified several impact categories that are useful in measuring the
environmental impact of buildings, including global warming potential, human toxicity, and
acidification, among others (Jolliet et al. 2003). Although the authors recognize the importance
of all of these categories in assessing the life cycle environmental impact of buildings, the
87
proposed method considers only global warming potential. The metric used for this indicator is
carbon dioxide equivalents (CO2e), which measures the total amount of greenhouse gas
emissions of the building.
Figure 19 – Building life cycle phases included in scope.
The general steps involved in the proposed method for evaluating the embodied versus
operational environmental impact trade-offs of conceptual building design decisions are shown
in Figure 20. The arrows in the figure represent data dependencies between process steps. The
analysis process begins with a building information model representing a given design
configuration. This model describes the building’s geometry, materials and components as well
as the project’s geographic position and orientation. The embodied carbon footprint is calculated
based on the building material and component quantities extracted from the model. Additional
details on how the embodied carbon footprint is calculated can be found at Basbagill et al.
(2013).
An energy simulation model is used to calculate the annual energy consumption of the
building in terms of electricity (kWh) and natural gas (kBtu). The energy simulation model is
created based on the geometry and building material information contained in the building
88
information model. Thermal zones are also defined in the model as well as standard assumptions
regarding building occupancy and HVAC system controls (ASHRAE 2009).
A maintenance, repair and replacement schedule is used to determine the impacts
associated with service equipment during the operational phase of the building. Impacts
associated with the production of materials in this schedule are grouped with the building’s
embodied impact. The schedule is determined by the gross floor area, building type, location,
and structural and mechanical details defined in the building information model, which are
entered into an online facility operations reference database (CostLab 2011).
Operational carbon footprint calculations have two components. The first depends on the
building’s electricity and natural gas consumption as calculated by the energy simulation model.
These quantities are multiplied by a unit impact to calculate carbon footprint. The second
component is associated with the maintenance, repair, and replacement of the service equipment.
The carbon impact of the mechanical, electrical, and plumbing equipment is determined by
looking up a typical material, material quantity, and cost for each component using equipment
supplier documentation. Each quantity is then multiplied by a unit impact in a similar fashion to
the pre-operational impact calculations. Total environmental impact is calculated by summing
embodied and operational CO2e totals.
89
Figure 20 – Software integration for optimized life cycle environmental impact feedback.
A genetic algorithm is applied to automatically iterate the carbon footprint analyses
described above across a defined range of design variables. The algorithm is used for three
separate objectives: minimization of embodied energy, minimization of operational energy, and
minimization of total energy, which is the sum of embodied and operational energies. The goals
of the optimization process are twofold: first, to determine whether a strategy that focuses on
minimizing only embodied energy and/or a strategy that focuses on minimizing only operational
energy can consistently yield designs with low total carbon footprint. This involves determining
whether either strategy yields designs with total carbon footprint close to the carbon footprint of
designs minimized for total carbon footprint. In this way, it is determined whether minimizing
embodied and/or operational impact can act as a proxy for minimizing total impact. The second
goal is to highlight the embodied versus operational impact tradeoffs for six variables: WWR,
glazing thickness, presence of fins, presence of overhangs, fin depth, and overhang depth.
Probability mass functions are constructed in order to characterize the impacts associated with
each variable value then determine which values are consistently found in high-performing
90
designs. Inspection of these functions determines whether minimizing embodied or operational
energy alone may reduce a building’s life cycle environmental impact to a similar degree as a
strategy that minimizes embodied and operational energy together.
A sampling algorithm is used to understand the full range of total impacts possible for a
given set of design variables. The algorithm generates a probability mass function, which is
useful for showing how designs generated by the genetic algorithm compare to the full range of
total impacts possible for the problem definition under consideration.
Seven software components are used to implement the proposed method illustrated in
Figure 20. DProfiler is used for building information modeling (DProfiler 2012). SimaPro and
the Athena EcoCalculator are used for life cycle environmental impact data and for calculating
the building’s carbon footprint (SimaPro 2010, Athena 2011). The energy simulation software
eQUEST is used to calculate operational energy (eQUEST 2010), and CostLab is used to
estimate the service schedules (CostLab 2011). Excel is used to calculate the carbon footprint
metrics based on the data provided by the previous components (Excel 2007). The optimization
and sampling processes are implemented using ModelCenter, a program that allows users to
bring commercial software tools into a common environment using software “wrappers” to
facilitate the application of automated design space exploration techniques (ModelCenter 2008).
The genetic algorithm chosen is the Darwin algorithm, and the sampling algorithm chosen is an
orthogonal array for 90% of the designs and a Latin hypercube for 10% of the designs.
91
Case Study
A residential complex of four eight-story buildings located in the Middle East is used as a
case study to evaluate the embodied versus operational impact trade-offs of several building
design decisions. The proposed design has buildings of identical size, shape, orientation, and
building materials. At the time this paper was submitted, the initial design scheme for the
buildings had been determined. The case study thus provided an opportunity to show how
retrospective changes in design could reduce the environmental impact of the building complex.
The proposed design has a total floor area of 50,468 m2, and the service life of the
building was assumed to be 30 years. The floor-to-floor height is 4.0 m. The building envelope
consists of a uniform cladding pattern consisting of steel and a translucent glazing material. The
mechanical system is a variable air volume forced air system with direct-expansion coils for
cooling and a central furnace for heating. Internal loads and the weekly operating schedule were
determined for a residential building using the 2009 ASHRAE Fundamentals (ASHRAE 2009).
Problem formulation
Table 10 summarizes the objectives and variables used in the optimization study. Carbon
footprint was calculated in terms of CO2e as described in the Methodology section. The
following energy conversions were used to perform the analysis: electricity impact: 0.664 kg
CO2e/kWh, natural gas: 0.251 kg CO2e/kBtu (SimaPro 2010).
Fourteen design variables were manipulated to minimize the environmental impact of the
building complex: (1) amount of glazing on the facade as a percentage of total façade area, (2)
glazing thickness, (3) presence or absence of fins and overhangs, (4) fin and overhang depth, (5)
92
building “H” shape determined by six shape parameters, (6) number of buildings, and (7) number
of floors. Total square footage remained constant. Table 11 lists the materials and impact factors
for each of the variables having an embodied versus operational energy impact trade-off.
Table 10 – Optimization problem formulation describing objectives and variable values.
Objectives
Minimize embodied impact (kg CO2e/m2)
Minimize operational impact (kg CO2e/m2)
Minimize total impact (kg CO2e/m2)
Variables Possible values
Window-to-wall ratio 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%
a,b
Glazing thickness (m) 0.0032, 0.0064, 0.0095, 0.0127, 0.0159, 0.0191
Has fins? true, false
Has overhangs? true, false
Fin depth (m) 0.3048, 0.9144, 1.524, 2.134
Overhang depth (m) 0.3048, 0.9144, 1.524, 2.134
Building shape a: 10, 15, 20, 25, 30
b: 10, 15, 20, 25, 30
c: 0, 5, 10, 15, 20, 25, 30
d: 10, 15, 20, 25, 30, 35, 40, 45, 50
e: 5, 10, 15, 20, 25
cf
Number of buildings 3, 4
Number of floors 5, 6, 7, 8
Orientation 0, 5, 10, …, 345, 350, 355
aU-factor (W/m
2*K) associated with each glazing thickness: 0.46, 0.23, 0.16, 0.12, 0.092, 0.077.
bSolar heat gain coefficient is 0.32 and visible transmittance is 0.62 for each glazing thickness.
cShape parameter “f” is dependent on the values for a, b, c, d, and e and ranges from 0m to 30m.
93
Table 11 – Material, dimensional, and impact assumptions for building components with
embodied versus operational impact trade-offs.
Building Material Thickness (m) Area (m2) Embodied Impact
Component (kg CO2e/kg material)
Glazing aVNE 1-63
bvaries
cvaries 1.06
Cladding steel 0.0677 cvaries 1.89
Fins concrete b,varies 1% of glazing area 0.121
Overhangs concrete bvaries 3% of glazing area 0.121
aGlazing supplied by Viracon (2013).
bValues are given in Table 10.
cDepth values are proportional to WWR, the values of which are given in Table 10.
Results
Comparison of optimization objectives
The distribution of design configurations generated by the proposed optimization process
in the performance space is shown in Figure 21. The graph represents a total of 10,297 designs
generated over 94 hours. The final design proposed by the design team for the building complex
is shown as well as the best design for each of the three objectives. A Pareto front shows the
relationship of each of these designs to the set of non-dominated best designs, and designers can
quantitatively see the tradeoffs between embodied and operational impacts. Operational energy
dominates the life cycle energy consumption and ranges from 76% to 99% of the total impact.
Eighty-seven percent of the design configurations generated by the proposed optimization
process exhibit improved performance with respect to total environmental impact compared to
the design chosen by the design team.
94
Figure 21 – Distribution of optimized design configurations.
The distribution of total impacts generated by the sampling algorithm is shown in Figure
22. The results consist of 5,522 designs and have a mean total impact of 19,170 kg CO2e per m2.
The most frequent range occurs between 16,000 to 17,000 kg CO2e per m2, the minimum impact
is 4,589 kg CO2e per m2, and the maximum impact is 50,888 kg CO2e per m
2. In comparing
Figures 21 and 22, 91% of the designs generated by the optimization algorithm have total
impacts less than 90% of the total impacts for the designs generated by the sampling algorithm.
95
Figure 22 – Distribution of life cycle environmental impacts.
Table 12 compares the best designs for the three objectives with the proposed design
configuration. The mean total impact for the top-performing designs for the strategy that
minimized operational impact was only 0.51% greater than the strategy that minimized both
operational and embodied impacts. This suggests that a strategy for minimizing carbon impact
that takes into account only operational impact may yield just as favorable results as when both
embodied and operational impacts are considered. This may be useful for a design team which
lacks the knowledge of material choices for building components and would like to focus on
operational impact. The strategy minimizing for operational impact also discovered the design
with the lowest total impact, which was lower than the proposed design’s total impact by 62%.
This value was within 1% of the best design found by the algorithm minimizing for total impact,
further suggesting minimizing for operational impact may act as a proxy for minimizing total
impact. The mean total impact for the top-performing designs minimized for embodied impact
96
was 52% greater than the strategy that minimized both operational and embodied impacts,
suggesting minimizing for embodied impact is not a good proxy for minimizing total impact.
The best designs for all three optimization objectives were better than the best design
discovered by the sampling algorithm in Figure 22, showing the utility of the genetic algorithm
in discovering high-performing designs regardless of the objective. This suggests a strategy
utilizing optimization algorithms are more effective at discovering high-performing designs than
sampling algorithms. Designs with low embodied impact often had a high operational impact,
further suggesting that minimizing for embodied impact may not approximate well high-
performing designs minimized for total impact. For example, the lowest embodied impact for the
minimized embodied impact objective was 539 kg CO2e/m2, or 38% lower than the embodied
impact of the design with the lowest total impact for this objective. Yet the operational impact
corresponding to this embodied impact was 8,015 kg CO2e/m2, or 119% higher than the
operational impact for the best design for this objective.
97
Table 12 – Comparison of trade-off variable values and carbon impacts.
Selected design configurations
●
Baseline Minimized
total impact
Minimized
embodied impact
Minimized
operational impact
Mean total impact
(kg CO2e/m2)
a
-- 4,285 6,520 4,307
Variables
Glazing thickness (m) 0.013 0.0032 0.0032 0.0032
Window-to-wall ratio 30% 15% 15% 15%
Has fins? true true false true
Has overhangs? true false false false
Fin depth (m) 0.30 0.30 -- 0.30
Overhang depth (m) 0.30 -- -- --
Best design (kg CO2e/m2)
Total 8,783 3,424 4,473 3,348
Relative difference -˗ -61% -49% -62%
Embodied impact 741 893 814 814
Operational impact 8,042 2,531 3,659 2,534
Lowest impacts (kg CO2e/m2)
Embodied n/a 558 540 558
Operational n/a 2,531 3,659 2,534 a
Calculated among top 10% of designs with lowest total impact.
Analysis of trade-off variables
Figures 23 through 26 represent probability mass functions for the six variables
associated with embodied versus operational impact trade-offs. The figures show the likelihood
of each variable value appearing in the top 10% of designs minimized for either embodied
impact, operational impact, or total impact. The graphs make clear which variable values are
likely to be found in high-performing designs and which are less likely. They also determine
whether a strategy weighted towards minimizing embodied and/or operational impacts correlates
with a strategy minimizing total impact, is likely to achieve high-performing designs, and can
therefore serve as a proxy for a strategy that minimizes both embodied and operational impacts.
98
Figure 23 shows the distributions for WWR. The minimized operational impact objective
correlates strongly with the minimized total impact objective, as all top designs for both
optimization strategies converge to the minimum value of 15%. The minimized embodied impact
showed a strong preference for the maximum value of 50%, which is expected given the lower
impact factor for glazing in Table 2 compared with the steel cladding impact factor. The results
suggest that for WWR a strategy minimizing only operational impact is a good approximation of
total impact and likely to achieve high-performing designs, whereas a strategy that minimizes
only embodied impact may not serve as a good proxy and may not yield high-performing
designs. Future work will consider alternate cladding materials to determine the extent to which
these results are specific to steel cladding.
Figure 23 – Distribution of window-to-wall ratio values.
99
Figure 24 shows the distributions for glazing thickness. The minimized embodied impact
objective strongly favored a low value, which is to be expected since glazing thickness is
proportional to embodied impact. The minimized operational objective also favored the low
value but to a lesser degree. The distribution for the minimized operational objective correlated
fairly well with the total impact distribution and better than did the embodied impact distribution,
suggesting that for glazing thickness a strategy minimizing only operational impact is also a
fairly good approximation of total impact.
Figure 24 – Distribution of glazing thickness values.
Figures 25 and 26 are the probability distributions for the presence of fins and overhangs
and show that each objective strongly preferred no fins or overhangs. The results for minimized
100
embodied impact are predictable, since embodied impact increases when shading devices are
present. As for total impact, the presence of the devices may have blocked a significant amount
of solar energy during the cooler portions of the day, thus requiring an amount of heating energy
that was greater than the difference between the cooling energy saved and the embodied energy
of the shading devices. A strategy that minimizes either embodied or operational energy is
therefore likely to achieve high-performing designs for these variables. A moderate percentage
of high-performing designs also had fins and overhangs for all three objectives. Careful
consideration must therefore be given to the placement of these components, since they have
noticeable effects on buildings’ embodied and operational impacts. Future work will consider
alternate climates, shading materials, and building sizes and will optimize the placement of
shading devices by building face in order to determine under which conditions the presence of
fins and overhangs will minimize a building’s life cycle environmental impact.
Figure 25 – Distribution of presence of fins values.
101
Figure 26 – Distribution of presence of overhangs values.
Figures 27 and 28 are the distributions for fin and overhang depth. Few trends emerged,
although the minimized total impact objective favored low overhang depth values. Given that
fewer than 35% of the top designs favored shading devices across all three objectives, it is not
surprising that preferences did not emerge. This is likely because changes to fin and overhang
depth were found to minimally affect embodied and operational impacts when no other variables
changed. As with the presence of fins and overhangs, future work will consider other climates,
building sizes, and shading materials in order to determine under what conditions changes to
shading depth may significantly alter a building’s embodied and operational impacts.
102
Figure 27 – Distribution of fin depth values.
Figure 28 – Distribution of overhang depth values.
103
Sensitivity Analysis2
Additional analyses were carried out in order to understand the degree to which the
results may be generalized across a broad set of variables. Previous research performed by
Basbagill et al. (2013) demonstrated that changes to cladding material can significantly affect a
building’s embodied impact. Changes to geographic location (Pulselli et al. 2009, Christenson et
al. 2006, Krarti et al. 2005) and gross floor area (Chung et al. 2006) have also been shown to
significantly affect a building’s energy performance. Therefore sensitivity analysis was
conducted on these three variables in order to comment on the degree to which the results may
be generalized to variables potentially significantly influencing the results. One alternative was
analyzed for each variable. Concrete was chosen to replace steel as the cladding material, since
concrete’s unit impact of 0.121 kg CO2e/kg material is significantly lower than steel’s value of
1.89 kg CO2e/kg material. In addition, concrete’s unit impact is significantly lower than the
glazing unit impact of 1.06 kg CO2e/kg material given in Table 11, whereas steel’s unit impact is
nearly twice that of the glazing material. Therefore it was believed that for concrete cladding the
minimized embodied impact objective would likely favor a low WWR value that better aligns
with the minimized total impact and minimized operational impact objectives than the high value
favored in Figure 23 by steel. Chicago was chosen as the alternate location, since Chicago is
located in climate zone 5 (cool, humid). The weather here is significantly different than the
weather for climate zone 2a (hot, humid) used in the results. A building gross floor area of
30,000 m2 was chosen as the alternate building size.
2This section provides additional analysis which was not included in the submission to Energy and Buildings.
104
Cladding material
Concrete was chosen as the alternate cladding material. Results are presented in terms of
the Pareto plot (Figure 29) and corresponding table of metrics (Table 13), distribution of impacts
(Figure 30), and favored values for the six tradeoff variables for the three optimization strategies
(Figures 31-36).
The Pareto plot and Table 13 show that substituting concrete for steel cladding
consistently achieves significant reductions in embodied impact. The average embodied impact
across all three optimization strategies for concrete cladding is 500 kg CO2e/m2, whereas the
average impact for steel cladding is 710 kg, a reduction of 30%. This reduction falls to 24%
when the probability mass functions (Figures 22 and 30) are compared for the two materials,
which is still a significant reduction. Comparison of the probability mass functions also shows
that substituting concrete for steel cladding filters out the worst-performing designs. These
results are expected, given concrete’s significantly lower carbon impact factor.
When operational impact is included in the analysis, Table 13 shows that the mean
impact for concrete cladding for the top 10% of high-performing designs is lower than the mean
impact for steel cladding for all three optimization strategies. The best designs are also achieved
with steel cladding. Since a constant R-value is assumed for both steel and concrete cladding, no
embodied versus operational impact tradeoff exists for cladding material. Therefore, these results
are likely due to the randomness of the genetic algorithm in discovering better designs with
lower operational energy for concrete cladding.
The results can also generalize the answer to the research question asking whether an
optimization strategy that minimizes only operational impact can act as a proxy for total impact.
As with steel cladding, the mean impact for the top 10% of high-performing designs for the
105
minimized operational impact objective is very close to the mean impact for the top-performing
designs for the minimized total impact objective. This alignment of values for a cladding
material with a high impact factor (steel) and a lower impact factor (concrete) for these two
optimization strategies suggests that operational impact may act as a good proxy for total impact
when implementing environmental impact optimization strategies regardless of cladding
material.
Figure 29 – Distribution of optimized design configurations for alternate cladding material.
106
Figure 30 – Distribution of life cycle environmental impacts for alternate cladding material.
Table 13 – Comparison of trade-off variable values and carbon impacts for alternate cladding
material.
Selected design configurations
●
Baseline Minimized
total impact
Minimized
embodied impact
Minimized
operational impact
Mean total impact
(kg CO2e/m2)
a
-- 3,960 5,346 3,959
Variables
Glazing thickness (m) 0.013 0.0032 0.0032 0.0032
Window-to-wall ratio 30% 15% 15% 15%
Has fins? true false false false
Has overhangs? true false false false
Fin depth (m) 0.30 -- -- --
Overhang depth (m) 0.30 -- -- --
Best design (kg CO2e/m2)
Total 8,783 3,934 4,107 3,934
Relative difference -˗ -55% -53% -55%
Embodied impact 741 497 492 497
Operational impact 8,042 3,437 3,615 3,437
Lowest impacts (kg CO2e/m2)
Embodied n/a 470 470 471
Operational n/a 3,437 3,615 3,437 a
Calculated among top 10% of designs with lowest total impact.
107
Analysis of the tradeoff variables shows that WWR (Figure 31) converges to the
minimum value for all three optimization strategies. This is expected, given that concrete’s
carbon factor is lower than the glazing material. Distribution of values for glazing thickness
(Figure 32) and presence of fins (Figure 33) or overhangs (Figure 34) remain the same for both
concrete and steel cladding, and this is expected given that that no coupled effects between
embodied and operational impact were set up in the problem formulation between cladding
material and these variables. The shape of the distributions for fin depth (Figure 35) and
overhang depth (Figure 36) are different between the two cladding types. Given shading depth
was found to minimally influence total impact as discussed in the results, it is likely that the
shading depth distributions are due to randomness in the genetic algorithm’s choices for shading
depth.
Figure 31 – Distribution of window-to-wall ratio values for alternate cladding material.
108
Figure 32 – Distribution of glazing thickness values for alternate cladding material.
Figure 33 – Distribution of presence of fins values for alternate cladding material.
109
Figure 34 – Distribution of presence of overhangs values for alternate cladding material.
Figure 35 – Distribution of fin depth values for alternate cladding material.
110
Figure 36 – Distribution of overhang depth values for alternate cladding material.
Climate
Chicago is located in climate zone 2a (cool, humid) and is chosen as an alternate climate
zone to compare with Houston in climate zone 5 (hot, humid). Results present the Pareto plot
(Figure 37) and corresponding table of metrics (Table 14), distribution of impacts (Figure 38),
and favored values for the six tradeoff variables (Figures 39-44).
The Pareto plot, Table 14, and the distribution of impacts show that operational impacts
are significantly greater in Chicago than in Houston. The average operational impact in Chicago
is 23,335 kg CO2e/m2, or 26% greater than Houston’s average operational impact of 18,451 kg
CO2e/m2. Minimizing only for embodied impact is a clear poor strategy yielding top designs
with average impacts 74% higher than designs minimized either for operational impact only or
111
total impact. The distributions of impacts also have higher values and are much more spread out
with lower probabilities in general than the Houston impact values. The average total impact
values in Chicago and Houston are 24,053 kg CO2e/m2 and 19,170 kg CO2e/m
2, respectively, and
the standard deviation values in Chicago and Houston are 19,793 kg CO2e/m2
and 7,684 kg
CO2e/m2, respectively.
The research question asks whether considering only operational impact in optimization
strategies can serve as a proxy for optimizing the total impact. Results show that both Chicago
and Houston climates yield mean impacts for the top 10% of high-performing designs for the
minimized operational impact objective very close (less than 1%) to the mean impact for the top-
performing designs for the minimized total impact objective. Therefore, the answer can be
generalized and operational impact may act as a good proxy for total impact when implementing
environmental impact optimization strategies in humid climates with extreme hot and/or cold
weather.
112
Figure 37 – Distribution of optimized design configurations for alternate climate.
113
Figure 38 – Distribution of life cycle environmental impacts for alternate climate.
Table 14 – Comparison of trade-off variable values and carbon impacts for alternate climate.
Selected design configurations
●
Baseline Minimized
total impact
Minimized
embodied impact
Minimized
operational impact
Mean total impact
(kg CO2e/m2)
a
-- 11,404 19,843 11,419
Variables
Glazing thickness (m) 0.013 0.0032 0.0032 0.0032
Window-to-wall ratio 30% 15% 50% 15%
Has fins? true false true false
Has overhangs? true false false false
Fin depth (m) 0.30 -- 2.1 --
Overhang depth (m) 0.30 -- -- --
Best design (kg CO2e/m2)
Total 8,783 11,367 12,165 11,367
Relative difference -˗ +29% +39% +29%
Embodied impact 741 722 603 722
Operational impact 8,042 10,646 11,563 10,646
Lowest impacts (kg CO2e/m2)
Embodied n/a 558 543 558
Operational n/a 10,646 11,563 10,646 a
Calculated among top 10% of designs with lowest total impact.
114
Analysis of the tradeoff variables for the alternate climate shows that the distributions for
WWR, glazing thickness, and presence of shading devices are very similar to Houston. Therefore
minimizing only for operational impact is likely a good strategy for achieving high-performing
designs when determining WWR or presence of shading devices. For glazing thickness, such a
strategy aligns fairly well with minimizing total impact. For shading depth, no clear trends
emerge due likely to the relative unimportance of these variables on total impact.
Figure 39 – Distribution of window-to-wall ratio values for alternate climate.
115
Figure 40 – Distribution of glazing thickness values for alternate climate.
Figure 41 – Distribution of presence of fins values for alternate climates.
116
Figure 42 – Distribution of presence of overhangs values for alternate climate.
Figure 43 – Distribution of fin depth values for alternate climate.
117
Figure 44 – Distribution of overhang depth values for alternate climate.
Building size
An alternate building size of 30,000 m2 is chosen to compare with the proposed design of
50,468 m2. Results present the Pareto plot (Figure 45) and corresponding table of metrics (Table
15), distribution of impacts (Figure 46), and favored values for the six tradeoff variables (Figures
47-52).
The Pareto plot, Table 15, and the distribution of impacts show that operational impacts
are significantly greater for the smaller building, although not to the same degree as the Chicago
analysis. The average operational impact for the smaller building is 21,010 kg CO2e/m2, an
increase of 14% over the larger building. As with cladding material and climate, minimizing only
118
for embodied impact is a poor strategy yielding top designs with average impacts 48% higher
than designs minimized either for operational impact only or total impact. The average total
impact values for the smaller and larger buildings are 21,817 kg CO2e/m2
and 19,170 kg
CO2e/m2, respectively, and the standard deviation values are 7,924 kg CO2e/m
2 and 7,684 kg
CO2e/m2.
The research question asks whether considering only operational impact in optimization
strategies can serve as a proxy for optimizing the total impact. Both building sizes yield mean
impacts for the top 10% of high-performing designs for the minimized operational impact
objective very close to the mean impact for the top-performing designs for the minimized total
impact objective. Therefore, the answer can be generalized and operational impact may act as a
good proxy for total impact when implementing environmental impact optimization strategies for
building sizes between 30,000 m2 and 50,468 m
2.
119
Figure 45 – Distribution of optimized design configurations for alternate building size.
Figure 46 – Distribution of life cycle environmental impacts for alternate building size.
120
Table 15 – Comparison of trade-off variable values and carbon impacts for alternate building
size.
Selected design configurations
●
Baseline Minimized
total impact
Minimized
embodied impact
Minimized
operational impact
Mean total impact
(kg CO2e/m2)
a
-- 9,079 13,430 9,092
Variables
Glazing thickness (m) 0.013 0.0032 0.0095 0.0032
Window-to-wall ratio 30% 15% 15% 15%
Has fins? true False false false
Has overhangs? true False true false
Fin depth (m) 0.30 -- -- --
Overhang depth (m) 0.30 -- 1.5 --
Best design (kg CO2e/m2)
Total 8,783 9,026 9,387 9,026
Relative difference -˗ +2.8% +6.9% +2.8%
Embodied impact 741 773 973 773
Operational impact 8,042 8,253 8,414 8,253
Lowest impacts (kg CO2e/m2)
Embodied n/a 680 648 687
Operational n/a 8,253 8,414 8,253 aCalculated among top 10% of designs with lowest total impact.
Analysis of the tradeoff variables shows similar patterns for all six variables for the two
building sizes. Minimization of only operational impact is a good strategy when selecting WWR
values, since this strategy aligns well with top-performing designs for the minimization of total
impact objective. This strategy also yields strong alignment in glazing thickness and shading
presence values in top-performing designs for these variables for the minimization of operational
impact and the minimization of total impact objectives. No trends can be drawn from shading
depth, again likely due to the relative unimportance of these variables in contributing to total
environmental impact.
121
Figure 47 – Distribution of window-to-wall ratio values for alternate building size.
Figure 48 – Distribution of glazing thickness values for alternate building size.
122
Figure 49 – Distribution of presence of fins values for alternate building size.
Figure 50 – Distribution of presence of overhangs for alternate building size.
123
Figure 51 – Distribution of fin depth values for alternate building size.
Figure 52 – Distribution of overhang depth values for alternate building size.
124
Conclusions
Evaluating the embodied versus operational impact trade-offs of building design
decisions can play an important part in creating a sustainable built environment. The proposed
optimization method allows designers to evaluate these trade-offs over a range of design
alternatives for a specified set of design variables. The method shows whether a design strategy
that minimizes only embodied or operational impact can serve as a proxy for minimizing total
impact. Design efforts can then focus on minimizing only this impact. Variation in building
shape, orientation, and massing extends the method’s scope to broad building typologies.
Results of the proposed method applied to the case study show that a design strategy
minimizing only operational impact consistently yields high-performing designs. Such a strategy
resulted in a mean total impact for the top-performing designs that was only 0.51% less than the
mean total impact for designs minimized for both operational and embodied impacts. In
particular, this strategy yielded high-performing design values for WWR, glazing thickness, and
presence of shading devices. A strategy minimizing only embodied impacts did not consistently
yield high-performing designs and resulted in a mean impact that was 52% greater than the mean
impact for designs minimized for total impact. The method also has the ability to improve upon
designs, as 87% of the designs had total impacts lower than the actual design proposed by the
design team.
The method allows designers to understand whether minimizing embodied and/or
operational impacts can guide their decision making for variable choices and yield reasonable
approximations of the minimized total impact, or whether minimization of both impacts together
is a better strategy. The scope of this method is limited to six trade-off variables for one building
size and type in a hot and humid climate with only one material considered for each variable.
125
Future research will evaluate case studies in additional climates with alternative building sizes
and materials. Additional trade-off variables will also be incorporated, such as R-values of wall
assemblies and roof slabs and cladding material thicknesses. Several of the variables can also be
optimized by façade to increase the utility of the method.
126
Chapter 5: A methodology for providing
environmental impact feedback on sequential
conceptual building design decisions1
Abstract
Conceptual design decisions play a critical role in determining the life cycle environmental
impact performance of buildings. Stakeholders often make these decisions without a quantitative
understanding of how a particular decision will impact future choices or the ultimate
performance of a project. A sequential decision support methodology is developed to provide
stakeholders with precise information on the relative influence that design decisions have on a
project’s environmental impact performance. Sensitivity analysis is performed on the impacts
associated with thousands of building design alternatives. A case study is presented showing how
the proposed methodology may be used by designers with various design strategies in mind.
Results are presented in the form of probabilistic distributions, which show the degree to which
each decision helps achieve a given strategy’s objective. The method provides environmental
impact feedback throughout sequential decision-making processes, thereby aiding designers in
achieving various building performance objectives during the conceptual design phase.
1This paper was co-authored with Assistant Professor Michael Lepech and submitted to The International
Journal of Architectural Computing in October 2013.
127
Introduction
Multidisciplinary design optimization (MDO) methods exist that allow designers to
explore very large design spaces, quickly evaluate many design alternatives, and find optimal or
near optimal solutions for various performance criteria. The benefits of MDO methods are well
documented in such industries as aerospace, automotive, and electronics. Within the architecture,
engineering, and construction industry, application of MDO methods has been shown to yield
significant reductions in building life cycle cost and environmental impacts compared to
conventional methods (Flager et al. 2012, Wang et al. 2005b).
Although MDO has potential to improve design process efficiency and the quality of the
resulting product, MDO methods are not widely used within the building design industry,
particularly during conceptual design. The conceptual design stage has been recognized as a
critical determinant of project cost and environmental impact (Ellis et al. 2008, Schlueter and
Thesseling 2009). At the conceptual design stage, many choices exist for building decisions,
such as building shape, massing, and dimensioning and materials for each building component.
These decisions are typically made by architects in sequential fashion, such that for example
once the orientation of the building is known, the placement of shading devices can be
determined for each façade in order to minimize cooling loads. Designers may also wish to
understand the life cycle cost and environmental impacts associated with the wall assembly
system before deciding upon the cladding system. Such a multi-objective sequential feedback
approach is typical in the architecture, engineering, and construction industry in that project
stakeholders often need to evaluate design decision trade-offs for competing objectives. For
example, a designer wishing to minimize both environmental impact and cost may find a certain
window-to-wall ratio lowers carbon footprint at the expense of greatly increased life cycle cost.
128
Existing MDO methods do not accommodate sequential decision-making processes.
MDO requires all design decisions to be made in parallel, instead of allowing designers to define
variable values sequentially and thereby understand the impacts for each successive decision.
Consequently designers utilizing MDO must decide on all building decisions before receiving
feedback on any single design choice. MDO methods do not integrate well with the architecture,
engineering, and construction industry, which relies on flexible and often-changing decision-
making processes, especially at the early stages.
A new method is proposed that integrates MDO methods with conceptual building design
in a way that provides quantitative feedback for a range of design strategies reliant on sequential
decision-making processes. Building information modeling software is integrated with life cycle
assessment and energy simulation software, allowing a sampling algorithm to generate thousands
of building design alternatives across the design space and compute life cycle environmental
impact feedback. Probability mass functions are then used to characterize the environmental
impacts of decisions as they are made in sequential fashion, thereby providing designers with
visual quantitative feedback on each of many alternatives.
Figure 53 shows how the method can be applied to three different sequential decision-
making strategies often used by designers. In each scenario, probabilistic distributions show the
range of life cycle environmental impacts possible for all design alternatives before any
decisions have been made. Once a decision is made, a new probabilistic distribution is generated
showing the range of impacts possible for the remaining design decisions. Designers are able to
understand the full range of control of environmental impact performance as well as the relative
influence of design decisions throughout the sequential decision-making process.
129
In Figure 53(a), a designer would like to minimize a building design’s life cycle
environmental impact. This strategy relies on single-objective optimization, which several
studies have shown can be an effective strategy for helping designers minimize the
environmental impact of buildings (Coley and Schukat 2002, Al-Homoud 1997, Wetter 2001).
As each sequential decision is made, the designer understands whether a decision improves upon
the previous decision in terms of either reducing or increasing the building’s life cycle
environmental impact. The designer also understands with each new decision whether chances
improve, worsen, or have been eliminated of achieving the design with the lowest possible
carbon footprint. Throughout the process the designer knows the full range of control for each
design decision as well as how each decision relates to the initial range of environmental impacts
before any decisions were made.
A second sequential decision-making approach employed by designers is designing for
environmental performance values (Figure 53(b)). Andreu and Oreszczyn (2004) discuss how
this strategy can be effective in creating designs with low life cycle environmental impact. Such
a strategy caters to designers interested in building rating systems and assessment tools, such as
the Green Building Challenge and the United States Green Building Council’s Leadership in
Energy and Environmental Design program. Dozens of building performance tools exist, which
are designed to provide indicators on the environmental performance of design alternatives or
rate the environmental performance level of a building (Ding 2008). In Figure 53(b), a designer
has an environmental impact performance target in mind. As in Figure 53(a), probabilistic
distributions are generated with each new design decision, and designers can understand whether
each decision helps or hurts in achieving the specified target value.
130
A third sequential decision-making approach employed by designers is the maintenance
of flexibility and adaptability (ALwaer and Clements-Croome 2010). This objective is
particularly relevant for designers with competing objectives in mind, such as optimizing designs
for both economic as well as environmental sustainability. In Figure 53(c), a designer wishes to
preserve freedom and flexibility throughout the design process by maximizing the number of
remaining designs as decisions are made. The designer does not want to be confined to a narrow
subset of designs, regardless of whether they have low, medium, or high environmental impacts.
Such a strategy maximizes the tradeoff options to be considered among the competing
objectives.
(a) (b) (c)
Figure 53 – Three sequential decision-making approaches to which the environmental impact
feedback method may apply: (a) minimization of carbon footprint, (b) achievement of a carbon
target value, and (c) maintenance of freedom and flexibility.
The decision support methodology is a quantitative approach that supplies stakeholders
with precise information about the relative influence that each design decision has on
environmental impacts. Three sequential decision-making strategies – minimizing environmental
impact, achieving performance values, and maximizing conceptual design freedom – are
described to which the method applies, although the method can easily be adapted to other
131
approaches. Designers quickly assess which design variables are significant contributors to a
building’s carbon footprint and which are less important. The use of probability mass functions
allows designers to predict with each successive decision the probability of achieving a given
impact value, and decisions may easily be adjusted in order to increase or decrease this
probability. The method provides visual understanding of the range of control of the entire
design space’s environmental performance, and by accommodating various sequential design-
making approaches, the method enhances its utility as a conceptual design stage decision-making
tool.
Related Studies
Research in multidisciplinary design optimization (MDO) is used as a point of
departure in order to present the proposed research methodology. MDO involves the
formalization of design coordination and iteration for groups working on complex engineering
systems such as buildings and civil infrastructure. Computational optimization techniques are
applied to systemically search through a range of design options defined by the design team to
find solutions that best meet the objectives and constraints of project stakeholders. MDO
methods were first developed in the aerospace industry in the 1970’s and are now successfully
used in a number of fields including automotive, naval architecture, and electronics design
(AIAA 1991).
A number of studies have used MDO as a method for providing environmental impact
feedback on conceptual building designs (Hauglustaine and Azar 2001, Caldas and Norford
2002, Wright et al. 2002, Geyer 2009). Wang et al. (2005b) integrated building information
modeling, life cycle assessment, energy analysis, and MDO software in order to evaluate the
132
environmental impact consequences of various conceptual building design parameters, such as
shape, orientation, and building materials. A multi-objective genetic algorithm was used to
identify Pareto optimal solutions for minimized cost and environmental impact performance,
resulting in reductions in cost and global warming potential. Al-Homoud (1997) applied a direct
search optimization technique in order to minimize the annual energy consumption of an office
building for different climates. The method provided optimized thermal performance feedback
on several hundred design alternatives, and variables included building orientation and thermal
properties of glazing materials. Coley and Schukat (2002) integrated building information
modeling and thermal analysis in developing a method for optimizing the energy performance of
a community hall. The method applied a genetic algorithm during the conceptual design phase,
and the results displayed a range of architecturally distinct designs minimized for operational
energy.
Prior MDO research has several limitations as far as its ability to integrate with AEC
industry design practices. MDO has limited ability to provide useful feedback on design
decisions. MDO feedback typically relies on Pareto fronts for evaluating the tradeoffs between
objectives. Such diagrams provide feedback only when designs are fully articulated. They offer
no performance evaluation on partially defined designs and fail to show the sensitivities of
performance criteria to changes in design variables. The feedback does not align well with a
design process in which decisions are made in sequential fashion and environmental impacts are
in flux with each successive design change. MDO feedback has also often been limited to the
operational phase of a building rather than its entire life cycle, and design variables are usually
concentrated in the building envelope. Embodied impacts of building components are not always
133
included, and variables relating to building systems’ materials and dimensions, particularly
structural assemblies, walls, floors, and finishes are often excluded.
The proposed method fills these gaps by providing environmental impact feedback on a
comprehensive set of design variables throughout the envelope, substructure, interiors, and
structural components. A design space consisting of a very large number of design alternatives is
evaluated by the method. A sampling algorithm generates probability mass functions, which
dynamically show designers how the range of control over environmental impacts changes with
each sequential decision. The distributions also show the sensitivity of environmental impacts to
changes to these variables. Embodied and operational impacts are included in the scope, as well
as a range of shape, massing, building material, and dimensioning parameters. The method is
well suited to a range of decision-making strategies, since in all cases designers understand the
likelihood of achieving a certain carbon footprint after each new design decision.
Methodology
The goal of the proposed methodology is to provide MDO feedback to designers at the
conceptual design stage in such a way that designers can understand the environmental impact
implications of design alternatives for each sequential decision. To illustrate the potential to
provide environmental performance feedback across a large number of building systems, the
building’s substructure, façade, interior, and service equipment are included in the analysis.
The shaded area in Figure 54 shows the phases of the building life cycle that are
considered in the research. Evidence from previous research suggests the included phases,
namely raw material acquisition, building material production, maintenance, repair, and
replacement, and operation account for over 95% of a building’s life cycle environmental impact
134
(Cole and Kernan 1996). Demolition has not been included since impacts associated with this
phase have been shown to be difficult to calculate (Pushkar et al. 2005, Schoch et al. 2011) and
small when compared with other phases (Scheuer et al. 2003).
Researchers have identified several impact categories that are useful in measuring the
environmental impact of buildings, including global warming potential, human toxicity, and
acidification, among others (Jolliet et al. 2003). Although the authors recognize the importance
of all of these categories in assessing the life cycle environmental impact of buildings, the
proposed method considers only global warming potential. The metric used for this indicator is
carbon dioxide equivalents (CO2e), which measures the total amount of greenhouse gas
emissions of the building.
Figure 54 – Building life cycle phases included in proposed method for providing environmental
impact feedback on sequential design decisions.
The proposed methodology integrates several building design and energy analysis
software packages. Figure 55 presents images from the building information modeling software
DProfiler (DProfiler 2012). Each image represents a unique design configuration for a set of
input parameters including percentage of glazing, building orientation, building shape, number of
floors, and presence of fins or overhangs.
135
Figure 55 – Three design alternatives generated by the building information modeling software
showing variations in several input parameters.
The general steps involved in the proposed method for providing environmental impact
feedback on sequential design decisions are shown in Figure 56. The arrows in the figure
represent data dependencies between process steps. The analysis process begins with an initial
seed design manually inputted into the building information model. This may represent a design
team’s proposed solution to which they would like to see how alternatives compare, or the initial
design may be a random configuration. The building information model describes the building’s
geometry, materials, and components as well as the project’s geographic position and
orientation. The embodied carbon footprint is calculated based on the building material and
component quantities extracted from the building information model. Each quantity is multiplied
by a unit impact (kg CO2e) to determine the carbon footprint.
An energy simulation model is used to calculate the annual energy consumption of the
building in terms of electricity (kWh) and natural gas (kBtu). The energy simulation model is
created based on the geometry and building material information contained in the building
information model. Thermal zones are also defined in the model as well as standard assumptions
regarding building occupancy and HVAC system controls (ASHRAE 2009).
136
A maintenance, repair and replacement schedule is used to determine the impacts
associated with service equipment during the operational phase of the building. Impacts
associated with the production of materials in this schedule contribute to the building’s embodied
impact. The maintenance, repair, and replacement schedule is determined by the gross floor area,
building type, location, and structural and mechanical details defined in the building information
model, which are entered into an online facility operations reference database (CostLab 2011).
Operational carbon footprint calculations have two components. The first depends on
the building’s electricity and natural gas consumption as calculated by the energy simulation
model. These quantities are multiplied by a unit impact to calculate carbon footprint. The second
component is associated with the maintenance, repair, and replacement of the service equipment.
The carbon impact of the mechanical, electrical, and plumbing equipment is determined by
looking up a typical material, material quantity, and cost for each service component using
equipment supplier documentation. Each quantity is then multiplied by a unit impact in a similar
fashion to the pre-operational impact calculations. Total environmental impact is calculated by
summing embodied and operational CO2e totals. The sampling algorithm is then used to generate
thousands of design alternatives, and feedback on these alternatives is inspected as described in
the following section.
137
Figure 56 – Method for providing probabilistic environmental impact feedback on sequential
building designs.
The goal of the proposed building information model-environmental impact feedback
integration method is to allow designers continuous probabilistic visualization of the
environmental impact performance of their design choices for a range of sequential decision-
making strategies. A sampling algorithm is used to characterize the life cycle environmental
impacts of sequential design decisions across a broad range of design variables. Probability mass
functions are constructed from the environmental impact data received from each of the
thousands of designs sampled. Inspection of these distributions allows designers to achieve their
particular objective, whether to minimize carbon footprint, achieve a carbon performance value,
or preserve design freedom. These strategies are not mutually exclusive, as a designer interested
in minimizing carbon footprint and meeting a performance value can test out different options
for a single design variable and see how the resulting distribution relates to the full range of
environmental impacts. The designer can then choose the value that preserves the most number
of desirable designs or, if no value yields a preferred distribution, the designer can backtrack,
modify a prior decision, and consider the new options.
138
Eight software components are used to implement the proposed method illustrated in
Figure 56. DProfiler is used for the building information modeling software (DProfiler 2012).
SimaPro and the Athena EcoCalculator are used for environmental impact data and for
calculating the building’s carbon footprint (SimaPro 2010, Athena 2011). The energy simulation
software eQUEST is used to calculate operational energy (eQUEST 2010), and CostLab is used
to estimate the service schedules (CostLab 2011). Excel is used to calculate the carbon footprint
metrics based on the data provided by the previous components (Excel 2007). The sampling
distributions are generated using the software ModelCenter, an MDO program that allows users
to bring commercial software tools into a common environment using software “wrappers” to
facilitate the application of automated design space exploration techniques (ModelCenter 2008).
The sampling algorithm chosen is an orthogonal array for 90% of the designs and a Latin
hypercube for 10% of the designs.
Case Study
A residential complex of four eight-story buildings located in a hot and humid climate
is used as a case study to illustrate the proposed environmental impact feedback method. The
buildings are of identical size, shape, orientation, and building materials. At the time this paper
was submitted, the initial design scheme for the buildings had been determined. In addition to the
proposed objective mentioned in the previous section, the case study thus provided an
opportunity to show how retrospective changes in design could reduce the environmental impact
of the buildings.
139
The case study building has a total floor area of 50,468 m2, and the service life of the
building was assumed to be 30 years. The floor-to-floor height is 4.0 m. The building envelope
consists of a uniform cladding pattern and a translucent glazing material. The mechanical system
is a variable air volume forced air system with direct-expansion coils for cooling and a central
furnace for heating. Internal loads and the weekly operating schedule were determined for a
residential building using the 2009 ASHRAE Fundamentals (ASHRAE 2009).
Thirty-three design variables could be manipulated in the design problem, and the
scope of these variables included the building’s substructure, envelope, and interiors. Table 16
categorizes the variables into three groups: materials, thicknesses, and design variables, which
represent variables not related to the previous two categories. Appendix 3 lists all possible input
values for each variable. Carbon footprint was calculated in terms of CO2e as described in the
Methodology section. The following energy conversions were used to perform the analysis:
electricity impact = 0.664 kg CO2e/kWh and natural gas impact = 0.251 kg CO2e/kBtu (SimaPro
2010). Total square footage remained constant.
140
Table 16 – Case study variables used to characterize building life cycle environmental impacts.
Variable Type
Design Material Thickness
Window-to-wall ratio Cladding Ceiling
Has fins? Roof Cladding
Has overhangs? Wall Floor finishes
Fin depth Partitions Floor insulation
Overhang depth Columns and beams Mat foundation
aBuilding shape Floor finishes Wall finishes
Number of buildings Floor insulation Glazing
Number of floors Floor structure
Orientation Piles
Substructure system Shading device
Wall finishes
Window frame aShape parameters defined as follows. Note “f” is dependent on “a” through “e”:
Results
Results are presented for each of the three sequential decision-making approaches
described in Figure 53. For each approach, a probability mass function is generated after each of
four building design decisions. A range of decisions is evaluated to illustrate the breadth of
design choices the method can accommodate, and many others can be substituted in their place.
The process of generating a new probability mass function is repeated after each new design
decision, although length constraints preclude showing distributions here for all 33 decisions.
141
Figure 57 shows the probability mass function for the entire design space before any
decisions have been made. The size of the design space is equal to the product of the number of
design choices for each design decision, or 3.69x1023
. From this design space, a sampling
algorithm selected 8,689 designs over 118 hours and computed the environmental impacts
according to the process shown in Figure 56. The mean total impact of the selected designs was
18,237 kg CO2e/m2, and the standard deviation was 6,907 kg CO2e/m
2. The global minimum was
3,826 kg CO2e/m2, and the global maximum was 47,207 kg CO2e/m
2. The following three
sections present the results for each of the three decision-making approaches after the first four
decisions have been made.
Figure 57 – Probability mass function characterizing a design space size of 3.69x1023
, showing
total environmental impacts for 8,689 selected designs prior to any design decisions.
142
Sequential decision-making approach I: minimization of total
environmental impact
Figure 58 presents the probability mass functions for the first four decisions for the
objective minimizing total environmental impact, and Table 17 presents relevant metrics for each
decision. The percentage of remaining designs is reduced considerably after the first decision,
since the designer has selected only four values for the orientation out of 72 possible values. This
decision has also reduced the mean considerably, showing that orthogonal orientations
significantly lower the total impact. The standard deviation is also much lower, as the
distribution shows that the remaining designs are clustered at the low end of the distribution.
Decisions two through four also result in distributions clustered near the low end. The global
minimum of 3,826 kg CO2e/m2 is maintained throughout each decision, and the global maximum
is steadily reduced to a value after the fourth decision that is 73% lower than the original global
maximum. The designer has achieved their objective by successfully choosing a set of decisions
that eliminates high impact designs and leads to a relatively high probability that the final design
will be in close proximity to the global minimum.
143
Figure 58 – Probability impact distributions for four sequential decisions for the objective
minimizing total environmental impact.
144
Table 17 – Metrics characterizing the design space for a design strategy minimizing
environmental impacts.
Objective: minimize total environmental impact
No Decisions Decision 1 Decision 2 Decision 3 Decision 4
Decision Orientation = 90°, 180°,
270°, or 360°
Cladding material
= concrete or
limestone
Wall material =
steel stud 16” o.c.
or steel stud 24”
o.c.
Number of
buildings = 3
Mean
(kg CO2e/m2)
18,237 7,663 7,669 7,857 7,797
Standard
deviation
(kg CO2e/m2)
6,907 2,015 1,991 1,964 2,292
Remaining
designs
(% of total)
100 6.2 1.8 0.86 0.45
Global
minimum
(kg CO2e/m2)
3,826 3,826 3,826 3,826 3,826
Global
maximum
(kg CO2e/m2)
47,207 14,848 12,906 12,748 12,748
Sequential decision-making approach II: achievement of carbon
performance value
Figure 59 presents the probability mass functions for the first four decisions for the
objective that aims to achieve a carbon performance value, and Table 18 presents the relevant
metrics after each decision has been made. An arbitrary performance value of 15,000 kg
CO2e/m2 has been marked on each distribution with a dashed line. The designer following this
approach may have a competing objective and so should maintain as high a number of designs as
possible within as close proximity to the performance value as possible. The distributions and
values in Table 18 show that the mean stays within 38% of the performance value throughout the
145
decision-making process. The percent of designs meeting the performance value and staying
within one standard deviation of this value is 32% or less after all the decisions. The fourth
decision results in the highest mean and standard deviation and lowest percent within one
standard deviation for the four decisions, and this may motivate a designer to backtrack and
revise the decision in order to reduce the mean and standard deviation, thereby increasing the
number of designs close to the performance value. Such a tactic highlights the ability of the
feedback method to accommodate decisions that are in flux throughout the conceptual design
stage.
Figure 59 – Impact distributions for first four decisions for the objective achieving a carbon
performance value.
146
Table 18 – Metrics characterizing the design space for a design strategy achieving a carbon
performance value.
Objective: achieve environmental impact performance value
No Decisions Decision 1 Decision 2 Decision 3 Decision 4
Decision Number of
floors = 7
Orientation =
130-210°
Glazing
thickness ≤
0.0064m
Shape
parameters
“C”, “E” ≤ 15m
Mean
(kg CO2e/m2)
18,237 18,445 18,682 18,358 20,590
Standard
deviation
(kg CO2e/m2)
6,907 6,901 6,685 6,854 8,021
Remaining
designs
(% of total)
100 24 5.9 2.1 0.73
% of designs
within 1 σ of
performance
value
30 30 30 32 27
Global
minimum
(kg CO2e/m2)
3,826 3,935 5,348 5,348 6,711
Global
maximum
(kg CO2e/m2)
47,207 47,114 44,254 44,254 44,254
Sequential decision-making approach III: maximization of design
freedom
Figure 60 presents the probability mass functions for the first four decisions for the
objective that aims to maximize freedom and flexibility and preserve a large number of design
options. A designer in this case would like to maintain a high standard deviation and percentage
of remaining designs, in order to avail the most number of options in the design space after each
decision. As with the strategy that aims to achieve a performance value, this strategy may also
have a competing objective, such as preserving as many aesthetic differences in the building
147
shape or façade materials. Table 19 presents the relevant metrics after each decision has been
made and shows that nearly ten times as many designs are still available to the designer after the
fourth decision than in either of the previous two decision-making strategies. Most importantly,
the standard deviation remains high throughout all the decisions. The mean, standard deviation,
and global minimum and maximum all stay within 3% of the mean, standard deviation, and
global values for the distribution generated before any decisions have been made. A designer
would have many options at this point from which to compare environmental impacts with other
objectives.
Figure 60 – Impact distributions for first four decisions for the objective maximizing design
freedom.
148
Table 19 – Metrics characterizing the design space for a design strategy maximizing design
freedom.
Objective: maximize freedom, flexibility, and design options
No Decisions Decision 1 Decision 2 Decision 3 Decision 4
Decision Has overhangs = true
Overhang depth
≤ 0.9144m
Wall finish =
vinyl
Number of
buildings = 3
Mean
(kg CO2e/m2)
18,237 18,253 18,100 18,145 17,801
Standard
deviation
(kg CO2e/m2)
6,907 7,000 7,009 7,111 7,037
Remaining
designs
(% of total)
100 50 24 12 6.8
Global
minimum
(kg CO2e/m2)
3,826 3,935 3,935 3,935 3,935
Global
maximum
(kg CO2e/m2)
47,207 47,132 47,093 46,894 46,894
Validation2
Validation of the method was conducted by determining whether enough data points had
been generated by the sampling algorithm. The simulation was run again with an identical
problem formulation but for a different sampling algorithm. The resulting distribution was then
combined with the original distribution to yield an aggregate distribution. Comparison of the
mean and standard deviation among the three distributions shows that enough data points had
been generated.
Figure 61 is the impact distribution generated using the same problem formulation as
2This section provides additional analysis not included in the submission to The International Journal of
Architectural Computing.
149
described in the methodology. The only difference is that instead of using an orthogonal array
algorithm for 90% of the points and a Latin hypercube algorithm for 10% of the points, all points
were generated using a Latin hypercube sampling algorithm. Figure 62 is the aggregate impact
distribution when Figures 57 and 61 are combined. Table 20 shows that the mean of the Latin
hypercube distribution is within 2% of the mean of the distribution from the results section, and
the mean of the aggregate distribution is within 1%. The standard deviations are also within
10%. These metrics suggest that the impact distribution generated in the results section is
reasonably stable and valid.
Figure 61 – Distribution of life cycle environmental impacts for alternate Latin hypercube
sampling algorithm.
150
Figure 62 – Combined distribution of life cycle environmental impacts for original and alternate
sampling algorithms.
Table 20 – Validation of impact distributions using alternate sampling method.
Sampling
method
Mean
(kg CO2e/m2)
Relative
Difference
Standard
deviation
# of designs Minimum impact
(kg CO2e/m2)
Maximum
impact
(kg CO2e/m2)
90% orthogonal
array+10% Latin
hypercube
18,237 -- 6,907 8,689 3,826 47,207
Latin hypercube 17,874 -1.99% 6,240 7,693 4,665 49,602
Combined 18,066 -0.94% 6,604 16,382 3,826 49,602
Conclusions
The application of MDO to conceptual building design is currently not well suited to
conventional architectural, engineering, and construction design practices. A method is needed
that accommodates the flexible, often-changing nature of sequential decision-making processes.
An environmental impact feedback methodology is proposed that relies on probabilistic
151
distributions to describe the full range of control of environmental impacts for sequential
building design decisions during the conceptual design stage. The method allows designers to
easily make changes to a set of decisions after viewing the probabilistic impacts associated with
each decision. The inclusion of a range of variables, including shape, massing, materials, and
dimensions, increases the method’s flexibility and applicability to a broad range of building
designs.
Results of the case study show how the method can successfully accommodate several
sequential decision-making strategies, including the minimization of environmental impact,
achievement of a performance value, and maximization of design freedom. Four sample
decisions are provided for each design strategy, and inspection of the mean, standard deviation,
number of remaining designs, and other metrics after each decision shows the likelihood of each
strategy meeting its objective. In several examples, decisions were made that increased the
likelihood of meeting the objective. The feedback method also shows when a decision does not
help in achieving an objective, such as the case study’s restriction of the shape to a square layout
instead of an “H” pattern, which in turn yielded a distribution that increased the mean and
standard deviation and reduced the likelihood of meeting the performance value.
The scope of the method is limited to 33 design variables and a design space size of
3.69x1023
that includes the substructure, façade, and interiors, and each variable has a limited
number of options. The case study validates the method for one building size and type in one
climate. Future research will consider alternative building sizes and types as well as alternate
climates. Designers must rely on intuition and use trial and error in determining which design
choices improve or worsen the chances of meeting an objective. Future research will employ
152
probability mass functions to show which variable values consistently aid in the achievement of
a particular objective and which values consistently lessen the chances of achieving an objective.
153
Chapter 6: A multi-objective feedback approach to
evaluating sequential building design decisions1
Abstract
Conceptual design decision-making plays a critical role in determining life cycle environmental
impact and cost performance of buildings. Stakeholders often make these decisions without a
quantitative understanding of how a particular decision will impact future choices or a project’s
ultimate performance. A sequential decision support methodology is developed to provide
stakeholders with quantitative information on the relative influence design decisions have on a
project’s life cycle environmental impact and life cycle cost. A case study is presented showing how
the proposed methodology may be used by designers considering these performance criteria.
Sensitivity analysis is performed on thousands of computationally generated building alternatives.
Results are presented in the form of probabilistic distributions showing the degree to which each
decision helps in achieving a given performance criterion. The method provides environmental
impact and cost feedback throughout the sequential building design process, thereby guiding
designers in creating low-carbon, low-cost buildings at the conceptual design phase.
1This paper was co-authored with postdoctoral fellow Forest Flager and Assistant Professor Michael Lepech and
submitted to Automation in Construction in October 2013. This is the capstone chapter that builds on the
contributions developed in Chapters 2-5.
154
Introduction
Multidisciplinary design optimization (MDO) methods exist that allow designers to
explore very large design spaces, quickly evaluate many design alternatives, and find optimal or
near optimal solutions for various performance criteria. The benefits of MDO methods are well
documented in such industries as aerospace, automotive, and electronics. Within the architecture,
engineering, and construction industry, application of MDO methods has been shown to yield
significant reductions in building life cycle environmental impact and cost compared to
conventional design methods (Flager et al. 2012, Wang et al. 2005b).
Although MDO has potential to improve design process efficiency and the quality of the
resulting product, MDO methods are not widely used within the building design industry,
particularly during conceptual design. The conceptual design stage has been recognized as a
critical determinant of project environmental impact and cost (Ellis et al. 2008, Schlueter and
Thesseling 2009). At the conceptual design stage, many choices exist for building decisions,
such as shape, orientation, massing, and materials for each building component. These decisions
are typically made by architects in sequential fashion, such that for example once the orientation
of a building is known, the placement of shading devices can be determined for each façade in
order to minimize cooling loads and life cycle costs. Designers may also wish to understand a
project’s environmental impact and cost once the wall assembly system has been chosen but
before deciding upon the cladding system. Such a multi-objective sequential feedback approach
is typical in the architecture, engineering, and construction industry in that project stakeholders
often need to evaluate design decision trade-offs for competing objectives. For example, a
designer wishing to minimize both environmental impact and cost may find a certain window-to-
wall ratio lowers carbon footprint at the expense of greatly increased life cycle cost.
155
Existing MDO methods do not accommodate sequential decision-making processes.
MDO requires all design decisions to be made in parallel, instead of allowing designers to define
variable values sequentially and thereby understand the impacts for each successive decision.
Consequently designers utilizing MDO must decide on all building decisions before receiving
feedback on any single design choice. MDO methods do not integrate well with the architecture,
engineering, and construction industry, which relies on flexible and often-changing decision-
making processes, especially at the early stages.
A new method is proposed that integrates MDO methods with conceptual building design
in a way that provides life cycle environmental impact and life cycle cost feedback for sequential
decision-making processes. Building information modeling software is integrated with life cycle
assessment and energy simulation software, a sampling algorithm generates thousands of
building design alternatives across the design space, and life cycle environmental impact and
cost feedback is computed for each alternative. Probability mass functions are then used to
characterize the environmental impact and cost of decisions as they are made in sequential
fashion. Designers are provided with visual quantitative feedback on many alternatives and can
determine the degree to which each decision helps or hurts in achieving each of their objectives.
Figure 63 illustrates how the method can be applied to three different sequential decision-
making strategies often used by designers. Environmental impact is displayed here as the
feedback type, although distributions can be simultaneously provided for cost feedback as well.
In each scenario, probabilistic distributions show the range of impacts possible for all design
alternatives before any decisions have been made. Once a decision is made, new probabilistic
distributions are generated showing the range of impacts possible for the remaining design
decisions. Designers are able to understand the full range of control of life cycle environmental
156
impact and cost performance as well as the relative influence of design decisions on both of these
objectives throughout the sequential decision-making process.
(a) (b) (c)
Figure 63 – Three sequential decision-making design strategies to which designers might
apply the multi-objective feedback method: (a) minimization of carbon footprint,
(b) achievement of an environmental impact performance target, and (c) maintenance of
design freedom.
In Figure 63(a), a designer would like to minimize a building design’s life cycle
environmental impact. This strategy relies on single-objective optimization, which studies have
shown can be an effective strategy for helping designers minimize the environmental impact of
buildings (Coley and Schukat 2002, Al-Homoud 1997, Wetter 2001). As each sequential
decision is made, the designer understands whether a decision improves upon the previous
decision in terms of either reducing or increasing the building’s remaining life cycle
environmental impacts. The designer also understands with each new decision whether chances
improve, worsen, or have been eliminated of achieving the design with the lowest possible
carbon footprint. Throughout the process the designer knows the full range of control for each
design decision as well as how each decision relates to the initial range of building impacts
before any decisions were made.
157
A second sequential decision-making approach employed by designers is designing for a
specific environmental impact performance target. Andreu and Oreszczyn (2004) discuss how
this strategy can be effective in creating designs with low life cycle environmental impact. Such
a strategy caters to designers interested in building rating systems and assessment tools, such as
the Green Building Challenge and the United States Green Building Council’s Leadership in
Energy and Environmental Design program. Designers employing this strategy may have a
secondary objective which they would like to optimize. In Figure 63(b), a designer has an
environmental impact performance target in mind and may secondarily like to minimize life
cycle cost. As in the first scenario, probability mass functions are generated with each new
design decision for both environmental impact and cost. Designers can simultaneously visualize
the range of control for each objective as well as evaluate the degree to which each decision
helps in achieving both objectives.
A third sequential decision-making approach employed by designers is the maintenance
of flexibility and adaptability throughout the design process (ALwaer and Clements-Croome
2010). As with the previous strategy, this approach is particularly relevant for designers with
competing objectives in mind, such as optimizing designs for both environmental impact and
cost. In Figure 63(c), a designer wishes to preserve freedom and flexibility throughout the design
process by maximizing the number of remaining designs as each decision is made. The designer
has a cost objective in mind that competes with environmental impact and does not want to be
confined to a narrow subset of designs, regardless of whether they have low, medium, or high
environmental impacts. Such a strategy maximizes the trade-off options to be considered
between the competing objectives.
158
The proposed decision support methodology is a quantitative approach that supplies
stakeholders with information about the relative influence that each design decision has on both
life cycle environmental impact and life cycle cost. The method is applied to a multi-objective
sequential decision-making strategy – achieving an environmental impact performance target and
minimizing cost – although the method can be adapted to other strategies. Designers quickly
assess which design variables are significant contributors to a building’s carbon footprint and
cost and which are less important. The use of probability mass functions allows designers to
predict with each successive decision the likelihood of achieving a given impact value, and
decisions may easily be adjusted in order to increase or decrease this likelihood. The method
provides visual understanding of the range of control of the entire design space’s environmental
impact and cost performance, and by accommodating various sequential design strategies, the
method enhances its utility as a conceptual design stage decision-making tool.
Related Studies
Research in Multidisciplinary Design Optimization (MDO) is used as a point of departure
in order to present the proposed research methodology. MDO involves the formalization of
design coordination and iteration for groups working on complex engineering systems such as
buildings and civil infrastructure. Computational optimization techniques are applied to
systemically search through a range of design options defined by the design team to find
solutions that best meet the objectives and constraints of project stakeholders. MDO methods
were first developed in the aerospace industry in the 1970’s and are now successfully used in a
number of fields including automotive, naval architecture and electronics design (AIAA 1991).
A number of studies have used MDO as a method for analyzing the trade-offs between
159
life cycle environmental impact and cost feedback on conceptual building design alternatives
(Hauglustaine and Azar 2001, Caldas and Norford 2002, Wright et al. 2002, Geyer 2009). Wang
et al. (2005b) integrated building information modeling, life cycle assessment, energy analysis,
and MDO software in order to minimize two objective functions: life cycle environmental
impact and life cycle cost. Several building design parameters were included in the analysis,
including shape, orientation, and building materials. A multi-objective genetic algorithm was
used to identify Pareto optimal solutions, or those solutions not dominated by any others in the
decision variable space for both life cycle carbon footprint and cost. Al-Homoud (1997) applied
a direct search optimization technique in order to minimize the annual energy consumption of an
office building for different climates. The method provided optimized thermal performance
feedback on several hundred design alternatives, and variables included building orientation and
thermal properties of glazing materials. Coley and Schukat (2002) integrated building
information modeling and thermal analysis in developing a method for optimizing the energy
performance of a community hall. The method applied a genetic algorithm during the conceptual
design phase, and the results displayed a range of architecturally distinct designs minimized for
operational energy.
Prior MDO research has several limitations as far as its ability to integrate with
architecture, engineering, and construction industry design practices. MDO has limited ability to
provide useful feedback on sequential design decisions. MDO feedback typically relies on Pareto
fronts for evaluating the tradeoffs between objectives. Such diagrams provide feedback only
when designs are fully articulated. They offer no performance evaluation on partially defined
designs and fail to show the sensitivities of performance criteria to changes in design variables.
The feedback does not align well with a design process in which decisions are made in sequential
160
fashion and life cycle environmental impacts and costs are in flux with each successive design
change. MDO feedback has also often been limited to the operational phase of a building rather
than its entire life cycle, and design variables are usually concentrated in the building envelope.
Embodied impacts of building components are not always included, and variables relating to
building systems’ materials and dimensions, particularly structural assemblies, walls, floors, and
finishes are often excluded.
The proposed method fills these gaps by providing life cycle environmental impact and
cost feedback on a comprehensive set of design variables throughout the envelope, substructure,
interiors, and structural components. A design space consisting of a very large number of design
alternatives is evaluated by the method. A sampling algorithm generates probability mass
functions, which dynamically show designers how the range of control over environmental
impacts and costs changes with each sequential decision. The distributions also show the
sensitivity of impacts to changes to these variables. Embodied and operational impacts are
included in the scope, as well as a range of shape, massing, building material, and dimensioning
parameters. The method accommodates a range of decision-making strategies, since in all cases
designers understand the likelihood of achieving a certain carbon or cost footprint after each new
design decision.
Methodology
Scope
The goal of the proposed methodology is to provide multi-objective feedback to
designers at the conceptual design stage in such a way that designers can understand the life
161
cycle environmental impact and cost implications of design alternatives for each sequential
decision. To illustrate the potential to provide this feedback across a large number of building
systems, the building’s substructure, façade, interior, and service equipment are included in the
analysis.
The Uniformat 2010 classification system is used in the AEC industry to classify building
components within building element categories (Construction Specifications Institute 2010).
Uniformat elements within the project scope are: Substructure (A), Shell (B), Interiors (C), and
Services (D). The remaining elements (Equipment and Furnishings (E), Special Construction and
Demolition (F), and Sitework (G)) are not considered, since these decisions relate to interior
aesthetics, require specialized knowledge of site conditions, or otherwise involve decisions that
would be impractical to make by designers before the design development stage. Further detail
on the classification framework is found at Basbagill et al. (2013). Appendix 4 enumerates the
material choices and their properties considered for each building component. These properties
include material densities and embodied CO2e factors, or the amount of carbon dioxide
equivalents associated with materials’ feedstock energy, energy required to process the materials
into building components, and fuel cycle energy for all pre-operational processes.
The shaded area in Figure 64 shows the phases of the building life cycle that are
considered in the research. Evidence from previous research suggests the included phases,
namely raw material acquisition, building material production, maintenance, repair, and
replacement, and operation account for over 95% of a building’s life cycle environmental impact
(Cole and Kernan 1996). Demolition has not been included since impacts associated with this
phase have been shown to be difficult to calculate (Pushkar et al. 2005, Schoch et al. 2011) and
small when compared with other phases (Scheuer et al. 2003).
162
Figure 64 – Building life cycle phases included in proposed method for providing multi-
objective feedback on sequential design decisions.
Researchers have identified several impact categories that are useful in measuring the
environmental impact of buildings, including global warming potential, human toxicity, and
acidification, among others (Jolliet et al. 2003). Although the authors recognize the importance
of all of these categories in assessing the life cycle environmental impact of buildings, the
proposed method is demonstrated for global warming potential. The metric used for this
indicator is carbon dioxide equivalents (CO2e), which measures the total amount of greenhouse
gas emissions of the building.
Analysis process
The general steps involved in the proposed method for providing multi-objective
feedback on sequential design decisions are shown in Figure 65. The proposed methodology
integrates several building design, energy analysis, life cycle assessment, and life cycle cost
software packages. The arrows in the figure represent data dependencies between process steps.
163
Figure 65 – Automated method for providing life cycle environmental impact and life cycle cost
feedback on sequential building design decisions.
The analysis process begins with an initial seed design manually inputted into the
building information model (DProfiler 2012). This may represent a design team’s proposed
solution to which they would like to see how alternatives compare, or the initial design may be a
random configuration. The building information model describes the building’s geometry,
materials, and components as well as the project’s geographic position and orientation. The
constraints are necessary for determining the operational environmental impacts and costs.
Assumptions are automatically programmed into the model but may be modified by the designer.
Figure 66 presents images from the building information modeling software. Each image
represents a unique design configuration for a set of input parameters including percentage of
glazing, building orientation, building shape, number of floors, and presence of fins or
overhangs.
164
Figure 66 – Three design alternatives generated by the building information modeling software
showing variations in several input parameters.
The embodied carbon footprint is calculated based on building component material
quantities extracted from the building information model. Each quantity is multiplied by a unit
impact (kg CO2e) to determine the carbon footprint. Appendix 4 provides sample formulas for
material quantities associated with each of the four building elements. Further details on how
these material quantities were derived can be found at Basbagill et al. (2013).
An energy simulation model is used to calculate the annual energy consumption of the
building in terms of electricity (kWh) and natural gas (kBtu). The energy simulation model is
created based on the geometry and building material information contained in the building
information model. Thermal zones are also defined in the model as well as standard assumptions
regarding building occupancy and HVAC system controls (ASHRAE 2009).
Pre-operational costs are calculated using formulas derived from RSMeans (RSMeans
2007). Many of these costs are based on the building’s gross floor area. Pre-operational costs not
dependent solely on gross floor area are calculated from variables defined by designers. Default
165
values are provided for those variable values not defined by designers. Sample cost formulas for
the four building elements considered in the scope of the research are given in Appendix 4.
A maintenance, repair and replacement schedule is used to determine the impacts
associated with service equipment during the operational phase of the building. Impacts
associated with the production of materials in this schedule contribute to the building’s embodied
impact. The maintenance, repair, and replacement schedule is determined by the gross floor area,
building type, location, and structural and mechanical details defined in the building information
model, which are entered into an online facility operations reference database (CostLab 2011).
Operational carbon footprint and cost calculations have two components. The first
depends on the building’s electricity and natural gas consumption as calculated by the energy
simulation model. These quantities are multiplied by a unit impact and a unit cost to calculate
carbon footprint and cost, respectively. The second component is associated with the
maintenance, repair, and replacement of the service equipment. The cost data is obtained directly
from the maintenance, repair, and replacement schedule. The carbon impact of the mechanical,
electrical, and plumbing equipment is determined by looking up a typical material, material
quantity, and cost for each service component using equipment supplier documentation. Each
quantity is then multiplied by a unit impact in a similar fashion to the pre-operational impact
calculations.
Life cycle environmental impact and cost are then calculated by summing the pre-
operational and operational CO2e and cost totals. The life cycle cost is calculated as a net present
value assuming a specified discount rate and escalation rate for electricity and natural gas prices.
The pre-operational costs are incurred at year zero and the operational costs are accounted for on
an annual basis over the service life of the facility.
166
A sampling algorithm is then applied to automatically iterate the life cycle carbon and
cost analyses described above across a defined range of design variables. The method generates
thousands of design alternatives, and feedback on these alternatives is inspected as described in
the next section.
Inspection of results
The goal of the proposed method for providing life cycle environmental impact and cost
feedback is to allow designers continuous visualization of the environmental impact and cost
performance of their design choices for sequential decision-making strategies. A sampling
algorithm is used to characterize the life cycle impacts of sequential design decisions across a
broad range of design variables. Two sets of probability mass functions – one set for life cycle
environmental impact and one set for life cycle cost – are constructed from each of the thousands
of designs analyzed. Inspection of these distributions aids designers in achieving their particular
objective, whether to minimize both carbon footprint and cost, attain a performance target, or
maximize design freedom. The method accommodates multiple objectives, as a designer
interested in meeting an environmental impact performance target and minimizing cost can test
out different options for a single design variable and see how the resulting distributions improve
upon the full range of carbon and cost impacts. The designer can then choose the value that
preserves the most number of desirable designs for both objectives or, if no value yields a
preferred distribution, the designer can backtrack, modify a prior decision, and consider the new
options.
167
Software implementation
Eight software components are used to implement the proposed method illustrated in
Figure 65. DProfiler is used for the building information modeling software (DProfiler 2012).
SimaPro and the Athena EcoCalculator are used for environmental impact data and for
calculating the building’s carbon footprint (SimaPro 2010, Athena 2011). RSMeans is used to
calculate building life cycle cost (RSMeans 2007). The energy simulation software eQUEST is
used to calculate operational energy (eQUEST 2010), and CostLab is used to estimate the service
schedules (CostLab 2011). Excel is used to calculate the carbon footprint metrics based on the
data provided by the previous components (Excel 2007). The sampling distributions are
generated using the software ModelCenter, an MDO program that allows users to bring
commercial software tools into a common environment using software “wrappers” to facilitate
the application of automated design space exploration techniques (ModelCenter 2008). The
sampling algorithm chosen is an orthogonal array for 90% of the designs and a Latin hypercube
for 10% of the designs.
Case Study
A residential complex of four eight-story buildings located in a hot and humid climate is
used as a case study to illustrate the proposed method. The buildings are of identical size, shape,
orientation, and building materials. Table 21 lists the building information model inputs in terms
of required inputs and variables, and Table 22 lists the assumptions. Variables not defined by the
designer are selected from the values given in Appendix 4. The case study building has a total
floor area of 50,468 m2.
168
Table 21 – Required inputs and variables for automated life cycle environmental impact
and life cycle cost feedback method. Required Inputs
Location
Building type
Gross floor area
Variables
Design Material Thickness
Number of buildings Cladding Ceiling
Number of floors Roof Cladding aBuilding shape Partitions
Window-to-wall ratio Columns and Beams
Has fins?
Has overhangs?
Floor finishes
Floor insulation
Fin depth Floor structure
Overhang depth Piles
Orientation Shading device
Substructure system Wall finishes
Glazing material aShape parameters defined as follows, with “f” dependent on “a” through “e”:
Table 22 – Assumptions for automated life cycle environmental impact and life cycle cost
feedback method.
Assumptions Value
Footing depth (m) 2.0
Bay spacing (m) 9.0
Floor-to-floor height (m) 4.0
Service life (years) 30
Discount rate
Electricity
5%
Cost (USD/kWh) a0.20
Impact (kg CO2e/kWh) b0.66
Escalation rate 3.0%
Natural gas
Cost (USD/kBtu) a0.03
Impact (kg CO2e/kWh) b0.25
Escalation rate 3.0% aRSMeans (2007)
bSimaPro (2010)
169
Twenty-seven design variables are manipulated in the design problem, and the scope of
the variables includes the building’s substructure, envelope, and interiors. Table 21 categorizes
the variables into three groups: materials, thicknesses, and design variables, which represent
variables not related to the previous two categories. The building envelope consists of a uniform
cladding pattern and a translucent glazing material. The mechanical system is a variable air
volume forced air system with direct-expansion coils for cooling and a central furnace for
heating. Internal loads and the weekly operating schedule are determined for a residential
building using the 2009 ASHRAE Fundamentals (ASHRAE 2009). Cost is calculated in terms of
US dollars, and carbon footprint is calculated in terms of CO2e as described in the “Analysis
process” section of the Methodology.
Results
Results are presented in the form of probability mass functions for a designer interested in
multiple objectives: minimizing cost while also attaining an environmental impact performance
target. These two objectives correspond to the schematics shown in Figures 63(a) and 63(b),
respectively, and the probability mass functions for both objectives are presented together in the
same figure. Such side-by-side presentation of the results allows designers to easily compare the
two objectives, determine whether undesired trade-offs exist for any design decisions, then either
make a new decision or revise a previous decision. For example, a decision may have the
undesired effect of lowering a building’s mean environmental impact at the expense of
increasing the mean cost. A designer can easily visualize this effect by comparing the two
objectives’ probability mass functions when placed next to each other as opposed to in separate
figures. Designers view these sets of distributions as a sequence: one set before making the
170
decision and the next set after the decision has been made. In this way, a designer receives
sequential feedback one design decision at a time for multiple objectives and can visually
determine whether each decision helps attain one, both, or neither of their objectives.
Probability mass functions are generated here for both objectives after each of four
sample building design decisions. A range of decisions is evaluated to illustrate the breadth of
design choices the method can accommodate, and many others can be substituted in their place.
In addition, one of these decisions is revised to show the ability of the method to improve upon
decisions that do not help attain both objectives. The process of generating new probability mass
functions is repeated after each design decision; length constraints preclude showing
distributions here for all 27 sequential decisions.
Figure 67 shows the probability mass functions for the entire design space for
environmental impact and cost before any decisions have been made. The size of the design
space is equal to the product of the number of choices for each variable, or 6.07x1016
. From this
design space, the sampling algorithm selected 7,623 designs and computed the environmental
impacts and costs over 118 hours according to the automated process shown in Figure 65. The
mean life cycle environmental impact of the selected designs is 20,981 kg CO2e per m2, and the
mean life cycle cost is $1,481 per m2. An environmental impact performance target of 13,000 kg
CO2e/m2 is marked with a dashed line. This value is arbitrarily chosen to represent a value that a
designer wishes to attain, perhaps as prescribed by a green building rating system or a design
firm’s sustainability benchmarks. A designer with the two stated objectives in mind –
minimization of cost and attainment of an environmental impact target – would try to achieve as
low a mean cost as possible while simultaneously maintaining as many designs as possible close
171
to the target value of 13,000 kg CO2e/m2. Additional relevant metrics for the two distributions
are presented in the first column of Table 23.
Figure 67 – Distribution of building life cycle environmental impacts and life cycle costs for a
design space size of 6.07x1016
.
Figures 68-72 show the resulting probability mass distributions after each of the four
decisions have been made, and Table 3 presents relevant metrics for each of the distributions.
172
The first decision sets the number of buildings equal to 3, and Figure 68 shows the
resulting probability mass distributions. Figure 68 and the metrics for this decision in Table 23
show that this decision filters out low-performing designs for the environmental impact objective
and high-performing designs for the cost objective. The percent of designs whose environmental
impact is within one standard deviation of the performance target increases slightly by 1%, and
the mean cost impact increases by 3%. The decision therefore shows a trade-off between
lowering carbon impact at an increased cost. A designer may accept the result if they favor the
environmental impact objective more than the cost objective, or if not a designer may revise the
decision.
173
Figure 68 – Distribution of life cycle environmental impacts and life cycle costs after decision 1:
number of buildings equals 3.
Assuming the first decision remains unchanged, the second hypothetical decision is low-e
glazing. Figure 69 shows the probability mass functions for life cycle environmental impact and
cost, and Table 23 shows relevant metrics after this decision. The decision has little effect from
the previous decision, as the means for both objectives decrease by less than 1%, the percent of
174
designs within one standard deviation from the performance target remains unchanged, and the
extreme values remain for both distributions. Still, the decision promotes both objectives, and a
designer would likely retain the decision after receiving this feedback.
Figure 69 – Distribution of life cycle environmental impacts and life cycle costs after decision 2:
low-e glazing.
175
Figure 70 shows the environmental impact and cost distributions and Table 23 shows the
relevant metrics after the third decision, window-to-wall ratio equal to 50. The decision filters
out high-performing designs and increases the mean environmental impact of remaining designs
by 30% and cost by 10%. The percent of designs falling within one standard deviation of the
environmental impact target also decreases from 27% to 4.2%. The decision clearly does not
help a designer achieve either of the two objectives, and the decision would likely be revised.
176
Figure 70 – Distribution of life cycle environmental impacts and life cycle costs after decision 3:
window-to-wall ratio equals 50.
Figure 71 and Table 23 show the effect of modifying the third decision’s window-to-wall
ratio value from 50 to 15. The decision is highly favorable to both objectives, as the mean
environmental impact is reduced by 43% from the second decision, and the mean cost is reduced
by 15%. The worst-performing designs for both objectives have carbon impacts and costs that
are 56% and 32% lower, respectively, than the worst-performing designs after the second
decision. The method can easily accommodate changes like this that better satisfy multiple
177
objectives, and a designer would likely retain this decision over the previous window-to-wall
ratio value.
Figure 71 – Distribution of life cycle environmental impacts and life cycle costs after decision 3
(revised): window-to-wall ratio equals 15.
The final decision limits the orientation to values between 0° and 180° (inclusive), as
shown in Figure 72 and Table 23. This decision shows the ability of the method to accommodate
multiple discrete choices for a single decision, thereby expanding its utility to designers who
178
may be able to filter out some values but not all. The decision promotes both objectives, as the
mean environmental impact is reduced by 4% and the percent of designs within one standard
deviation of the environmental impact performance target remains at 32%. The mean cost is
reduced slightly by 0.4%, and a designer with the stated objectives in mind would likely accept
the results and continue on to the next decision. The process would continue for the remaining 23
decisions until a final design is reached.
At this point in the design process the results can also be compared with the full range of
impacts possible prior to any design decisions. The mean life cycle environmental impact has
decreased by 46%, the percent of designs within one standard deviation of the environmental
impact performance target has increased by 6%, and the mean life cycle cost has decreased by
12%.
179
Figure 72 – Distribution of life cycle environmental impacts and life cycle costs after decision 4:
orientation from 0° to 180°.
180
Table 23: Metrics characterizing the design space for two design strategies: (a) achieving a life
cycle environmental impact performance value and (b) minimizing life cycle cost.
Objectives: (a) achieve life cycle environmental impact performance value
(b) minimize life cycle cost
No Decisions Decision 1 Decision 2 Decision 3 Decision 3 (revised) Decision 4
Decision # buildings = 3 Glazing = low-e WWR = 50 WWR = 15 Orientation = 0° to
180°
Objective (a) (b) (a) (b) (a) (b) (a) (b) (a) (b) (a) (b)
aMean 20,981 1,481 20,958 1,530 20,896 1,528 27,124 1,689 11,842 1,302 11,426 1,297
aStandard
deviation
7,517 180 7,558 195 7,592 196 7,570 191 3,208 92 3,112 83
Remaining
designs
(% of total)
n/a n/a 50 50 25 25 4.9 4.9 1.5 1.5 0.85 0.85
b% of
designs
within 1 σ of
target
26 n/a 27 n/a 27 n/a 4.2 n/a 32 n/a 32 n/a
aGlobal
minimum
5,042 1,027 5,178 1,084 5,258 1,084 7,467 1,134 5,258 1,084 5,258 1,084
aGlobal
maximum
50,692 2,328 50,351 2,328 50,351 2,328 50,351 2,328 22,100 1,579 21,127 1,505
aUnits for (a) are kg CO2e/m
2, and units for (b) are USD/m
2.
bThis metric is relevant only to the strategy achieving an environmental impact performance target and so is not
calculated for the strategy minimizing life cycle cost.
Conclusions
The application of multi-disciplinary design optimization to conceptual building design is
currently not well suited to conventional architectural, engineering, and construction design
practices. A method is needed that accommodates the flexible, often-changing nature of
sequential decision-making processes. A multi-objective feedback methodology is proposed that
relies on probabilistic distributions to describe the full range of control of impacts for sequential
building design decisions during the conceptual design stage. The method allows designers to
easily make changes to a set of decisions after viewing the probabilistic impacts associated with
181
each decision. The inclusion of a range of variables, including shape, massing, materials, and
dimensions, broadens the method’s applicability to a variety of building typologies.
Results of the case study show how the automated sequential feedback method can
successfully accommodate multiple objectives, in this case minimization of life cycle cost and
achievement of a life cycle environmental impact performance target. Four sample decisions are
provided, and visual inspection of the resulting probability mass functions shows the degree to
which each decision helps promote the objective. The results show that high-performing designs
are retained for many of the scenarios. When the number of buildings is set to three, a trade-off
arises between the two objectives, as the mean environmental impact decreases but the mean cost
increases. This trade-off may be due for example to a material’s higher unit cost and lower unit
embodied environmental impact than another material; in this way, the method can help
illuminate complex carbon versus cost trade-offs. For decisions involving such a trade-off, a
designer would weight their objectives, determine whether the net result is beneficial, then either
maintain or revise the decision. The case study results also show that high-performing designs
for both objectives are filtered out when the window-to-wall ratio equals 50, and a designer
would likely revise the decision. On the other hand, a window-to-wall ratio value of 15 greatly
improves both objectives. In this way, the method is flexible enough to easily accommodate
changes to design decisions that may better promote a designer’s objectives.
The scope of the method is limited to 27 design variables in the substructure, façade,
and interiors, and each variable has a limited number of options. The case study demonstrates the
method for one building size and type in one climate. Future research will consider alternative
building sizes and types as well as alternate climates. Designers must rely on intuition and use
trial and error in determining which design choices improve or worsen the chances of meeting
182
each objective. Future research will develop a method that more systematically evaluates the
trade-offs between the objectives, so that designers can make choices that yield results optimal to
both objectives. Further development of an integrated approach to evaluating multiple objectives
in this way during the conceptual design process can help designers analyze very large design
spaces with less effort leading to the creation of low-carbon, low-cost buildings.
183
Chapter 7: Conclusions
This chapter summarizes the conclusions from chapters 3 through 6 as well as the
primary contributions to theory. Answers to the research questions presented in Chapter 1 are
summarized. The chapter ends with a discussion on the challenges and recommendations related
to the research as well as limitations and possible avenues for future work.
Summary of Conclusions from Chapters 3 through 6
Chapter 3 presented an automated method of integrating life cycle assessment with
conceptual building design in a way that reduces embodied impacts. The method integrates BIM,
LCA, energy simulation, and MRR scheduling software with a feedback processor. Very few
inputs are required to generate many design alternatives. Sensitivity analysis was performed
using a case study and showed which building components consistently contribute to a building’s
embodied impact. Embodied impacts can potentially be large in the substructure, shell, or
interiors but not the service equipment. Cladding material and cladding thickness are consistently
large contributors to embodied impact. Therefore, a designer should focus on these decisions that
can potentially achieve a large embodied impact reduction and defer less significant decisions to
later design stages.
Chapter 4 builds on the work presented in Chapter 3 by including operational impacts in
the scope of the analysis. The method evaluates embodied versus operational environmental
impact trade-offs of conceptual building design decisions. Designers can evaluate these trade-
offs over a range of design alternatives for a specified set of design variables. The method also
shows whether a design strategy that minimizes only embodied or operational impact can serve
184
as a proxy for minimizing total impact. Design efforts can then focus on minimizing only this
impact. Results of the method applied to a case study show that a design strategy minimizing
only operational impact consistently yields high-performing designs. This strategy yields high-
performing design values for window-to-wall ratio, glazing thickness, and presence of shading
devices. A strategy minimizing only embodied impacts does not consistently yield high-
performing designs. The method therefore allows designers to understand whether minimizing
embodied and/or operational impacts can guide their decision making for variable choices and
yield reasonable approximations of the minimized total impact.
Chapter 5 presented a method for providing life cycle environmental impact feedback on
sequential building design decisions. The method accommodates the flexible, often-changing
nature of sequential decision-making processes in the AEC industry. The method relies on
probability mass distributions to describe the full range of control of environmental impacts for
sequential building designs during the conceptual design stage. Designers can easily make
changes to a set of decisions after viewing the probabilistic impacts associated with each
decision. Results of a case study show how the method can accommodate a range of sequential
decision-making strategies, including minimizing environmental impact, achieving a
performance target value, and maximizing design freedom. Four sample decisions were made,
and the results show the degree to which each decision helps or does not help achieve a given
objective. In this way, designers can understand how likely a given decision will help them
successfully execute a sequential decision-making strategy.
Chapter 6 extends the work presented in Chapter 5 by including other objectives in the
analysis besides life cycle environmental impact. The method presented provides multi-objective
feedback when evaluating sequential building design decisions. Probability mass distributions
185
are used again to provide probabilistic feedback on multiple objectives for a range of sequential
decision-making design strategies. Changes to decisions can easily be made after designers view
these sets of distributions. A case study applied the method to two strategies, minimization of life
cycle cost and achievement of a life cycle environmental impact performance target value,
although the method is flexible enough to accommodate additional strategies. Results of the case
study show the degree to which four sample decisions help achieve both strategies. Trade-offs
are highlighted showing when a decision helps achieve one strategy but not the other strategy.
Results also show when a decision filters out high-performing designs for both strategies. In this
way, the method gives designers the flexibility and ease to make changes to their design choices
in a way that better promotes both objectives.
Contributions
A methodology is presented that integrates life cycle assessment and conceptual building
design. The methodology provides automated environmental impact and cost feedback for many
design alternatives by integrating building information modeling, life cycle assessment, life cycle
cost, and energy simulation software with a feedback processor. Multi-disciplinary design
optimization (MDO) is a primary point of departure for the research, and part of the
methodology integrates automated feedback with sequential design decision-making processes.
The method allows designers to understand the full range of control they have in the detailed
decisions ahead over life cycle environmental impacts and life cycle costs for a given set of
inputs. Designers also understand which design decisions are significant contributors to these
impacts and which are less important. Various algorithms are used to iterate over the
computational method and generate the design alternatives, and these include a genetic algorithm
186
and a sampling algorithm. The method requires as few inputs as possible, which makes the
method appropriate for use during the critical early stages of the design process.
Three theoretical contributions are developed in the research, and each is discussed in the
following sections.
Embodied impact heuristics
A set of heuristics is developed that is used to calculate the pre-operational embodied
impacts of a full range of building components given as few inputs as possible. The required
inputs are gross floor area, building location, and building type. The heuristics also include
several assumptions and variables, the ranges of which may or may not be known by a designer.
The heuristics are integrated within the computational methodology so that embodied impacts
can be quickly calculated for many building designs. Optimization algorithms can then be
applied to minimize impacts or costs, and sampling algorithms can be applied to allow designers
to understand the full range of impacts.
The heuristics depart from existing methods for calculating embodied impacts in that
only three inputs are required. Athena EcoCalculator has been used by other studies to calculate
embodied impacts of building components within a computational framework, and Athena
requires ten inputs. Therefore, the heuristics require less information than Athena, and they may
therefore be used even earlier in the conceptual design stage. This allows designers to receive
environmental impact feedback at a more critical point in the design process in terms of ability to
determine life cycle impact. Designers can therefore be more informed of their decisions earlier,
and the hope is that this will lead to a final design with lower life cycle environmental impacts
than had the method not been used.
187
Chapter 2 introduces the embodied impact heuristics, and Chapter 3 provides detail on
their development and gives some examples. Chapter 3 then integrates the heuristics within the
computational methodology for providing environmental impact feedback. The heuristics are
applied to show which building components are consistently significant contributors to a
building’s embodied impact and which are less important. Chapter 4 applies the heuristics in a
way that shows the embodied versus operational impact tradeoffs of various design decisions.
Designers can understand the relative importance of these decisions’ embodied and operational
impacts on a building’s total impact. Designers also understand whether an optimization strategy
that focuses on minimizing only embodied impact and/or a strategy that minimizes only
operational impact can approximate a strategy that minimizes embodied and operational impacts
together. In this way, designers can understand which decisions to focus on during the
conceptual design phase and which information can be deferred to later design stages. Chapter 5
integrates the heuristics within a sequential decision-making process typically used by designers.
Contributions of this integration process are described in the next two sections.
Chapter 6 uses the embodied impact heuristics as a basis for coming up with a new set of
heuristics for calculating the pre-operational costs of a comprehensive set of building
components. As with the embodied impact heuristics, only three inputs are required, and the
heuristics are integrated within the computational feedback framework. When applied to
sequential decision-making processes, the heuristics inform designers of the degree to which
each design decision achieves cost-effective buildings. The environmental impact and cost
feedback can be used in parallel to achieve high-performing designs for multiple objectives.
The heuristics are validated in Chapter 2. Life cycle assessment data were obtained on
eight buildings of varying sizes in the Arup Project Embodied Carbon Database (Arup 2013).
188
These eight LCAs are highly detailed and were performed after most of the design details of each
building had been determined. The LCAs therefore required considerable effort and detailed
information. ADAPT also performed LCAs on the same eight buildings but required only one
input, building size, as well as ranges for each variable to calculate life cycle embodied impact.
The embodied impact values for the eight detailed LCAs were then compared with distributions
of impact values generated using ADAPT. The impacts for the two methods are consistently
close – within one standard deviation in seven cases, and less than 1% away from one standard
deviation in the eighth case – with values for large buildings generally much closer to the
detailed LCA values. ADAPT therefore captures reasonably well embodied impact values of
highly detailed LCAs for small, medium, and large buildings. These observations suggest that
the embodied impact heuristics are validated for a range of building sizes, at least to the level of
precision that can be expected in conceptual design.
In summary, Chapters 2 and 3 develop the embodied impact heuristics, Chapter 2
presents validation of the heuristics, and Chapters 3 through 6 show how the heuristics can be
used for a number of useful applications relating to the creation of low-carbon, low-cost
buildings.
Integration of automated feedback and sequential decisions
A second contribution to theory is a new method for integrating automated feedback and
sequential design decisions. One of the shortcomings of multi-disciplinary design optimization is
that parameters and decisions must be declared before any feedback is received. Chapter 5
describes a reconfiguration of the design process such that automated feedback is provided
sequentially after every single design decision. This feedback can come in the form of life cycle
189
environmental impact (Chapter 5), life cycle cost (Chapter 6), or other building performance
criteria. The use of probability mass functions allows designers to visually see the probabilistic
effects of each design decision relative to the full range of impacts possible before any decisions
have been made. In this way, designers can easily tailor their decisions to their precise design
strategy, be this minimization of a performance criterion, achievement of a performance target,
or maximization of design freedom. Designers can understand whether each decision they make
helps or hurts in achieving the goal of their strategy, and designers can easily backtrack and
modify design decisions so that they are more likely to achieve their objective.
Chapter 5 validates this method for integrating automated feedback with sequential
design decisions by examining a single building case study. The validation approach was to
determine whether enough data points had been generated by the sampling algorithm. The
method was then applied to an identical problem formulation but using a different sampling
algorithm. The resulting distributions – one set of distributions for the first sampling algorithm
and another set of distributions for the second sampling algorithm – were then combined to yield
an aggregate distribution. Comparison of the mean and standard deviation of the two
distributions in isolation as well as the aggregate distribution showed that the three means were
all within 2% and the standard deviations were within 10%. Thus the distributions were
reasonably stable. This comparison suggests that enough data points have been generated and
that the method for integrating LCA feedback with sequential designs is validated.
The method departs from existing methods for integrating MDO with design in that
MDO is not currently configured to accommodate sequential design processes. The contribution
is a novel way of making sequential design decisions based on automated feedback for a range of
building performance objectives.
190
Range of control of building performance alternatives
A third contribution is the ability for designers to visualize the full range of control of
values for a given building performance objective. The contribution is the plot showing the range
of performance values possible versus the likelihood of achieving a particular performance value
for a given problem formulation. Each performance value is plotted on the x-axis, and the
probability of achieving each value is plotted on the y-axis. Possible performance objectives
include life cycle environmental impact, life cycle cost, and schedule performance, although the
method may accommodate other building performance objectives. The contribution builds off of
the previous two contributions, in that the embodied impact heuristics are required to calculate
the embodied impacts of building components. Automated feedback presented in sequential
fashion then allows designers to understand how each decision relates to the full range of control
of all possible building performance alternatives. Only three inputs are required to generate the
full range of control of values, making the contribution highly applicable during the conceptual
design stage. By knowing very limited information, a designer can understand the minimum and
maximum possible impacts, the mean and standard deviation, whether a given design decision is
favorable to achieving a designer’s objective, and how the decision compares to the full scope of
impact values before any decisions have been made. Chapter 5 presents the automated feedback
in the form of life cycle environmental impact, and chapter 6 presents multi-objective feedback:
life cycle environmental impact and life cycle cost.
The method departs from existing methods for understanding the full range of control of
values for a given building performance objective, in that the method is applicable specifically at
the conceptual design stage and requires only three inputs. No other methods are known that
191
provide feedback on the full range of control of life cycle environmental impact and life cycle
cost feedback values for so few inputs.
Answers to Research Questions
Five research questions were proposed in Chapter 1, and a summary of the answers
follows in this section.
The first question asked how many design inputs are required for a method that integrates
MDO into sequential building design? Chapters 2 and 3 discussed development of several
embodied impact heuristics and demonstrated how the number of required inputs can be reduced
to three. These heuristics can then be used in a number of applications that provide building
performance feedback on design decisions during the conceptual design phase. One of these
applications is a method that integrates automated feedback into sequential building design
processes.
The second research question asked which design decisions contribute most significantly
to embodied impacts of buildings? Chapter 3 applied the embodied impact heuristics to a given
problem formulation in order to answer the questions. A ranked set of material and dimensioning
decisions was generated for a range of building components showing which decisions can achieve
significant embodied impact reductions and which are less important.
The third research question asked how a design strategy minimizing only operational
impact compares with a strategy minimizing both operational and embodied impacts? Chapter 4
looked at various optimization strategies in order to understand the relative importance of
embodied and operational impacts for a range of decisions exhibiting a tradeoff between
embodied and operational impacts. In nearly all cases, it was found that a design strategy that
192
minimizes only operational impact compares favorably with and can approximate well a design
strategy that minimizes embodied and operational impacts together. Sensitivity analysis was also
carried out on an alternate cladding material, location, and building size, and results in these
scenarios were consistent with the original analyses.
The fourth research question asked how well can a method that leverages automated
feedback be used to support sequential building design decision-making processes? Chapter 5
presented a method for integrating automated feedback into sequential design decision-making
processes. The feedback is provided after every decision for a range of design strategies,
including minimization of a building performance objective, achievement of a performance value,
and maximization of design freedom. A sample set of decisions was made for each strategy
showing how the method can be successfully used to provide feedback on each decision in such a
way that a designer understands how much each decision helps in achieving their design
objective.
The fifth research question asked about the range of control for a set of building design
parameters in terms of life cycle environmental impact and life cycle cost performance, and about
how designers make decisions within this range. Chapter 5 generated impact distributions for a
given problem formulation showing the full range of control of life cycle environmental impact
performance values for a very large design space. Chapter 6 generated impact distributions for
both life cycle environmental impact and life cycle cost. Designers are able to see the probability
of achieving any one particular value along these distributions. When integrated with sequential
design decision-making processes, this range of control serves as a point of reference allowing
designers to understand how each decision they make compares with the full range of possible
impact values for each objective before any decisions have been made.
193
Challenges and Recommendations
This section presents general challenges and recommendations which may be useful for
researchers who wish to extend the work presented here. The first set of challenges relates to use
of the software. New software versions may be incompatible with software that was compatible
with older versions. It is suggested that new versions be tested on a secondary machine to ensure
compatibility before installing on a primary machine. It is also important to be aware that
software licenses may have an expiration date. A suggestion is to avoid running models in close
proximity to these dates. Another point is that integrating multiple software platforms inevitably
means working with software developers at different firms, and the developers may not be
familiar with all software platforms. A suggestion is to plan group meetings with all developers
to ensure everyone understands the functionality and limitations of each software platform. A
final suggestion relating to the software is to clarify the functionality of “wrapper” inputs and
outputs with software developers. This will help understand precisely how building information
modeling inputs affect energy analysis outputs.
A second set of challenges relates to data sources. A suggestion is to determine early on
in the project the sources from which data will be obtained. This allows understanding of the
scope and limitations of the project. Another suggestion is to determine early on the feasibility of
integrating the data into a computational framework. If the data are embedded in software, a
suggestion is to discuss with software developers early on the feasibility of integrating the
software into the framework using automated integration techniques. If automated integration is
infeasible, researchers should determine early on whether manually extracting the data will
accomplish the project objectives.
194
Recommendations are also provided relating to the project scope and organization. It is
recommended that an initial research scope be set early on and modified accordingly throughout
the project. The scope should be narrow at first and widened as progress is made. In terms of file
organization, a suggestion is to avoid saving files in multiple locations. When using a remote
computer, a recommendation is to create a system at the beginning of the project that establishes
which files will be saved on the remote computer and which files will be saved on the primary
computer. It is recommended that an organized file management system be created at the
beginning of a project. A naming convention system should be maintained for different versions
of files, and all versions of files should be saved in the same folder.
A final set of recommendations involves the use of human resources. It is recommended
that key contacts be established early on at architecture, building information modeling, energy
analysis consulting, and software firms. The relationships can be strengthened by maintaining
contact often, informing partners of problems and project successes, and collaborating on
publications. It is also suggested that researchers offer to help contacts with their own research
and share relevant resources including publications and other reference material.
Limitations and Future Work
The method presented for integrating life cycle assessment and conceptual building
designs has a number of limitations, each of which offers avenues for possible future research.
The first avenue for future work relates to expanding the scope of the project. Additional
software can be computationally integrated into the framework developed in the research.
Additional objectives beyond life cycle environmental impact and cost could then also be
considered in the analysis. Additional work may look at climate zones beyond the two analyzed
195
in the research (climate zones 3 and 5). The only building type for which the method has been
tested is a residential mid-rise building, and additional work can determine the degree to which
the method can accommodate commercial and industrial building types. The geometry is also
restricted to orthogonal H-shapes, and the work could be extended to include more generalized
forms including curved shapes. The case studies only considered building heights from one to
eight stories and WWR values from 15 to 50%. Additional case studies could be analyzed to
comment more generally on the range of possible values for input parameters such as these.
Variation in R-value for different façade, roof, and floor materials is a possible extension of the
work to show additional embodied versus operational impact tradeoffs. Additional materials and
sizes can be included as well for each of the building components beyond those listed in the
Appendices. Another possible extension is optimizing variables by façade, in order to account
for variation in heat gain along each face of a building. Additional functionality can be built in to
the method to accommodate partial design decisions, such as when a designer may know a
façade R-value is not greater than a certain value but is uncertain about the lower bound.
A second avenue for future work involves building on the method for integrating
automated feedback with sequential building design decisions. The method uses probability mass
functions as the primary feedback mechanism to aid designers in understanding the
environmental impact implications of each design decision. These functions are provided only
after a decision has been made. Additional probability mass functions can be generated and
shown to designers prior to each decision. This pre-processing step would generate all possible
probability mass functions for every single variable for every single possible ordering of design
decisions. Considering the very large design space and large number of permutations such a
strategy would typically involve, an a priori tool would require considerable computing power in
196
order to accurately characterize the distributions for each variable value across all possible
design decision sequences. One way to generate such functions would be brute-force
computation, although further research may illuminate more efficient ways to provide such
information and insight. Designers could then refer to these functions prior to making a decision,
visually determine which variable values best align with the achievement of their performance
objective, select a value, then receive feedback as described in Chapters 5 and 6 to see the
outcome of this decision. In this way, designers would have two separate tools for guiding their
decisions, one useful for determining the precise variable value to choose, and the other for
determining how well the outcome of the decision aligns with their performance objective.
Development of the method so far allows for backtracking and easy changes to design decisions.
However by including this additional tool, the process for selecting variable values is less reliant
on trial and error. Use of this additional tool should reduce the number of iterations and amount
of time needed to arrive at a desired design configuration.
Integration of such avenues for future work with the method’s current functionality can
yield a powerful tool for the sustainable design community. Designers, architects, contractors,
developers, engineers, and building owners can quickly receive life cycle environmental impact
and cost feedback on a range of building design alternatives. The method empowers stakeholders
with quantitative feedback on building designs, and integrating such information specifically
during the conceptual design phase should contribute to less fragmented interactions and more
streamlined workflows in the AEC industry. The method requires only a few inputs, which may
consist of nothing more than a project’s gross floor area. Alternatively, the method can
accommodate highly detailed building information models with unconventional geometries.
Thus, the method integrates with however much information is known and can adapt as new
197
information is learned throughout the design process. The method’s ability to easily provide life
cycle assessment feedback based on very little information should also hopefully change
prevailing industry attitudes that life cycle assessment is too complex or not important enough to
perform on building designs. Sensitivity analysis is a powerful mechanism built into the method
that provides information on a full range of design alternatives, and such feedback enables
designers to forecast which design decisions affect cost and carbon footprint the most.
Stakeholders are therefore empowered to focus only on those decisions that have the greatest
bearing on their objectives of interest, and this can save the AEC industry significant time and
money as well as ultimately contribute towards the creation of a more sustainable built
environment.
198
Appendices
Appendix 1 – Supporting data for chapter 1
Survey questions provided to design firms to gauge trends in use of
LCA and LCC feedback in design
(1) For new building projects, which feedback strategies does your firm use to reduce the
building’s environmental impact, and how often do you use these strategies? Choose
from the following strategies: LCA, strategies to reduce operational energy, and
strategies to reduced embodied energy.
(2) Which LCA software programs do you use?
(3) When in the design process (early stages, design development, late stages) does your firm
make massing and material decisions for each building component? Consider the
following components: foundation, structural elements, cladding, insulation, roofing,
partitions, finishes, and HVAC and mechanical systems.
(4) What factors prevent your firm from conducting LCA on projects?
(5) How helpful would an early stage design-LCA feedback tool be for your firm?
199
Appendix 2 – Supporting data for chapters 2 and 3
Material alternatives considered in quantifying embodied impacts of
building components
Building component Material alternatives
Piles Steel pipe, precast concrete
Vapor barrier Polyethylene plastic sheeting, polypropylene cloth
Columns and beams Concrete column/concrete beam, concrete column/glulam beam, concrete column/LVL beam,
concrete column/WF beam, HSS column/glulam beam, HSS column/LVL beam, HSS column/WF
beam, WF column/glulam beam, WF column/LVL beam, WF column/WF beam
Floor structure Glulam/plywood decking, precast hollowcore/concrete topping, wood I-joist/plywood decking,
open-web steel joist/concrete topping, open-web steel joist/plywood decking, precast double-
T/concrete topping, steel joist/concrete topping, steel joist/plywood decking, suspended concrete
slab/concrete topping, wood joist/plywood decking, wood chord and steel web truss/plywood
decking, wood truss/plywood decking
Roof Precast hollow-core concrete, precast concrete double-T, suspended concrete slab, open-web steel
joist w/steel decking, open-web steel joist w/wood decking, glulam joist with plank decking, wood
I-joist w/WSP decking, solid wood joist w/WSP decking, wood chord/steel web truss with WSP
decking, wood truss (flat) with WSP decking
Roof membrane EPDM, PVC, modified bitumen, 4-ply built-up roofing system, steel roofing system
Stairs Precast concrete, wood, steel
Railings Wood, precast concrete, aluminum
Cladding Brick, steel, stucco, vinyl, wood, limestone, concrete
Wall structure Concrete block, cast-in-place concrete, steel stud wall, curtainwall: opaque glazing, curtainwall:
metal spandrel panel
Window frame Aluminum, steel, PVC, wood-metal, wood
Exterior doors Steel, wood-glass-steel, wood-steel
Partitions Steel stud wall, concrete block
Interior Doors Wood, steel
Wall coverings Ceramic wall tile, vinyl
Flooring surface Ceramic floor tile, stone floor tile, cement facing tile with fiber, cement cast plaster floor, neoprene,
polyacrylate with ground granite, polyurethane with vinyl chips, carpet tile with nylon, laminated
veneer wood
Floor insulation Blown cellulose, extruded polystyrene, polystyrene foam slab, cork slab, foam glass, glass wool
mat, glass wool (fleece), rock wool, rock wool (fleece), rock wool (packed), tube insulation
(elastomere), urea formaldehyde foam slab, urea formaldehyde in situ foaming
Duct insulation Blown cellulose, extruded polystyrene, polystyrene foam slab, cork slab, foam glass, glass wool
mat, glass wool (fleece), rock wool, rock wool (fleece), rock wool (packed), tube insulation
(elastomere), urea formaldehyde foam slab, urea formaldehyde in situ foaming
200
Material(s) associated with each building component and material
properties used to quantify building component embodied impacts
Material Building Component(s) Density (kg/m3) Embodied impact factor (kg
CO2e/kg material)
Concrete Mat foundation 2400 .050
Concrete Footing 2400 5.15
Concrete Piles, pile caps 2400 0.12
Concrete (precast) Grade beams, stairs, railings,
Cladding
2400 0.121
Concrete Slab on grade 2400 .065
Steel Rebar 8000 1.03
Steel Piles, stairs, cladding 8000 1.89
Steel (stainless) Gutter 8000 3.38
Brass Service equipment (various) 8400 2.34
Cast iron Service equipment (various) 7000 2.34
Limestone Cladding 2500 0.019
Limestone Flooring 2500 0.37
Stucco Cladding 800 0.070
Copper Gutter 8930 1.81
Brick Cladding 2403 1.13
Neoprene Flooring 1230 2.46
Polyethylene Vapor barrier 950 1.58
Fiberglass Pipe insulation 26 1.50
Flat glass Glazing 2310 1.06
Polyvinyl butyral Glazing 1100 6.98
Vinyl Gutter, cladding 580 2.37
Vinyl Wall coverings 1360 1.80
Ceramic tile Wall coverings, flooring 1360 0.74
Acrylic paint Wall coverings 1200 2.66
Aluminum Gutter, railings 2700 11.4
Wood (Douglas fir) Formwork 600 1.02
Wood (Douglas fir) Gutter 600 0.29
Wood Stairs, railings, cladding, flooring 600 0.33
Polyurethane foam (high density) Flooring 400 4.04
Polypropylene (vapor retarder) Vapor barrier 946 1.68
Polyacrylate terrazzo Flooring 1054 3.23
Carpet tile Flooring 74 6.78
Blown cellulose Floor insulation, duct insulation 56 0.35
Extruded polystyrene Floor insulation, duct insulation 30 10.1
Polystyrene foam slab Floor insulation, duct insulation 30 3.36
Cork slab Floor insulation, duct insulation 110 1.09
Foam glass Floor insulation, duct insulation 110 1.50
Glass wool mat Floor insulation, duct insulation 40 1.40
Glass wool, fleece Floor insulation, duct insulation 40 2.79
Rock wool Floor insulation, duct insulation 46 1.00
Rock wool, fleece Floor insulation, duct insulation 28 1.12
201
Rock wool, packed Floor insulation, duct insulation 100 1.05
Tube insulation, elastomere Floor insulation, duct insulation 75 4.16
Urea formaldehyde foam slab Floor insulation, duct insulation 20 2.71
Urea formaldehyde foam, in situ
foaming
Floor insulation, duct insulation 20 2.85
Embodied impact heuristics developed for each building component
Uniformat element Assembly Sub-component (a)Material quantity formula
A: Substructure (9) piles piles density * (b)slab area * slab depth
vapor barrier density * thickness * ((b)slab area + (c)perimeter * (d)footing depth)
caps 48 * density *((c)number of interior grid intersections + (c)number of
exterior grid intersections)
slab-on-grade (b)slab area * thickness
grade beam 4 * density * (c)perimeter
rebar 8 * (b)slab area
formwork 4 * thickness * (c)perimeter
footings footings 0.2 * density * slab area
mat foundation foundation (b)slab area * slab depth
B: Shell (20) columns and beams columns and beams (e)gross floor area + roof area
floor floor structure gross floor area – (b)slab area
roof roof structure (b)slab area
membrane (b)slab area
insulation (b)slab area
paint (b)slab area
stairs stairs 0.00017 *density *slab area
railing 0.91 *density *slab area * (number of floors – 1)
cladding cladding density * thickness * (1-WWR) * (c)perimeter * (e)height
exterior walls wall structure (1-WWR) * (c)perimeter * (e)height
insulation (1-WWR) * (c)perimeter * (e)height
membrane (1-WWR) * (c)perimeter * (e)height
gypsum (1-WWR) * (c)perimeter * (e)height
paint (1-WWR) * (c)perimeter * (e)height
glazing glass 2 * density * thickness * (f)glazing area
polyvinyl butyral density * thickness * (f)glazing area
frame (f)glazing area
hardware (f)glazing area
doors door 0.00098 * gross floor area
hardware 0.00098 * gross floor area
C: Interiors (12) partitions partition structure 1.5 * gross floor area
gypsum 1.5 * gross floor area
paint 1.5 * gross floor area
doors door 0.0039 * gross floor area
hardware 0.0039 * gross floor area
wall finishes covering density * thickness * (2.55 * gross floor area + (c)perimeter *
(e)height)
paint density * thickness * (c)perimeter * (e)height
flooring surface density * thickness * gross floor area
insulation density * thickness * gross floor area
ceiling plaster gross floor area + roof area
gypsum gross floor area + roof area
202
paint gross floor area + roof area (g) D: Services (61) mechanical (17) air conditioner 29,687 air handler 8,347 gland ball valve 164 boiler 8,430 chiller 26,466 condenser 6,788 direct digital
controls
720
ductwork 11,425 duct insulation 20,026 fan and motor
exhaust fan
(ceiling)
2,020
exhaust fan (roof
mounted)
277
flow control valve 109 HVAC control
panel
1,120
thermostat 22.5 VAV control box 103 pipe and fittings 120 gages and valves 200
electrical (16) bus duct and fittings 26,113
circuit breaker 0.45
disconnect switch 48
emergency lighting
pack, 2 lights
with battery
(entire)
3,463
emergency lighting
pack, 2 lights
with battery
(lamps)
413
exit lighting fixture,
w/battery
307
ballast and lamps,
fluorescent
lighting fixture,
T8, 32W (glass
tube)
364
ballast and lamps,
fluorescent
lighting fixture,
T8, 32W (steel)
57,543
heat detector 267
main switchgear 18,084
motor starter 1165.5
power panel board 1194
receptacle 1,448
secondary
transformer
9,059
TV cable outlet 80
203
wiring device
(switch)
357
plumbing (23) pipe and fittings,
cast iron
879
pipe and fittings,
copper
5,622
pipe and fittings,
PVC
1,402
pipe and fittings,
steel
879
gasket and bolts, 1” 80
gasket and bolts, 3” 24
pipe insulation 820.5
backflow preventer 1,295
bathtub and shower
enclosure –
diverter valve
57.12
bathtub and shower
enclosure
3,234
bathtub and shower
enclosure –
faucet washer
and clean shower
heat
8.4
bathtub and shower
enclosure – valve
set
42
circulator pumps 125.3
floor drain 755
flush tank water
closet
8,916
lavatory 9,010
lavatory valve 901
lavatory faucet
washer and clean
trap
270.3
lavatory washer and
spud connection
180.2
strainer service sink 321
service sink faucet
washer and clean
trap
9.63
service sink valve
set
32.1
water heater 79.8
fire (4) manual pull station 145
smoke detector 349
fire extinguisher 800
fire sprinkler head 250
conveying (1) elevator 850 (a) Material quantities may be multiplied by number of buildings in a project. This assumes each building is uniform in terms of all variables. (b) Slab area = gross floor area / number of floors.
204
(c) Variable determined by length and width parameters. (d) Constraint values are given in Chapter 2. (e) Height = number of floors * floor to floor height. (f) Glazing area = building length * building height * WWR. Note glazing area is dependent on building length for each façade. (g) Chapter 3 describes how an online facility operations reference database takes as inputs gross floor area, constraints,
and service life assumptions outlined in Chapter 2, scales the given numerical quantities presented here according to peak building load, and
returns impacts.
Appendix 3 – Supporting data for chapter 5
Variables and variable values used as inputs for environmental impact
feedback method
Design variable Values
Window-to-wall ratio 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%
Has fins? true, false
Has overhangs? true, false
Fin depth (m) 0.30, 0.91, 1.5, 2.1
Overhang depth (m) 0.30, 0.91, 1.5, 2.1
Building shape a: 10, 15, 20, 25, 30
b: 10, 15, 20, 25, 30
c: 5, 10, 15, 20, 25, 30
d: 10, 15, 20, 25, 30, 35
e: 5, 10, 15, 20, 25
af
Number of buildings 3, 4
Number of floors 5, 6, 7, 8
Orientation 0, 5, 10, …, 345, 350, 355
Substructure system mat foundation, piles, footings
Material variable Values
Cladding brick, steel, stucco, vinyl, wood, limestone, concrete
Roof precast hollow-core concrete, precast concrete double-T,
suspended concrete slab, open-web steel joist with steel
decking, open-web steel joist with wood decking, glulam joist
with plank decking, wook I-joist with WSP decking, solid
wood joist with WSP decking, wood chord/steel web truss with
205
WSP decking, wood truss (flat) with WSP decking
Wall concrete block, cast-in-place concrete, 2x4 steel stud 16”
o.c., 2x4 steel stud 24” o.c.
Partitions steel stud 16” o.c. + gypsum board, steel stud 24” o.c. +
gypsum board, concrete block + gypsum board, concrete block
Columns and beams concrete column/concrete beam, concrete column/glulam
beam, concrete column/LVL beam, concrete column/WF
beam, HSS column/glulam beam, HSS column/LVL beam,
HSS column/WF beam, WF column/glulam beam, WF
column/LVL beam, WF column/WF beam
Floor finishes ceramic floor tile, stone floor tile, cement facing tile with fiber,
cement cast plaster floor, neoprene, polyacrylate with ground
granite, polyurethane with vinyl chips, carpet tile + nylon,
laminated veneer wood
Floor insulation blown cellulose, extruded polystyrene, polystyrene foam slab,
cork slab, foam glass, glass wool mat, glass wool (fleece), rock
wool, rock wool (fleece), rock wool (packed), tube insulation
(elastomere), urea formaldehyde foam slab, urea formaldehyde
foam
Floor structure glulam + plywood decking, precast hollowcore + concrete
topping, wood I-joist + plywood decking, open-web steel joist
+ concrete topping, open-web steel joist + plywood decking,
precast double-T + concrete topping, steel joist + concrete
topping, steel joist + plywood decking, suspended concrete
slab + concrete topping, wood joist + plywood decking, wood
chord and steel web truss + plywood decking, wood truss +
plywood decking
Piles steel pipe, precast concrete
Shading device steel, aluminum, concrete
Wall finishes ceramic wall tile + acrylic paint, vinyl wall covering + acrylic
paint
Window frame aluminum, steel, PVC, wood-metal, wood
Thickness Variable Values
Ceiling (m) 0.0064, 0.011, 0.015, 0.019
Cladding (m) 0.025, 0.067, 0.11, 0.15
Floor finishes (m) 0.0064, 0.011, 0.015, 0.019
Floor insulation (m) 0.089, 0.13, 0.17, 0.22
206
Mat foundation (m) 0.20, 0.44, 0.67, 0.90
Wall finishes (m)
Wall tile 0.0064, 0.011, 0.015, 0.019
Vinyl wall covering 0.0016, 0.0021, 0.0026, 0.0032
b,c
Glazing thickness (m) 0.0032, 0.0064, 0.0095, 0.0127, 0.0159, 0.0191
aShape parameter “f” is dependent on the values for a, b, c, d, and e and ranges from 2m to 30m
bU-factor (W/m
2*K) associated with each glazing thickness is: 0.46, 0.23, 0.16, 0.12, 0.092, 0.077
cSolar heat gain coefficient is 0.32 and visible transmittance is 0.62 for each glazing thickness
Appendix 4 – Supporting data for chapter 6
Variables and variable values used as inputs for environmental
impact and cost feedback
Design variable Values
Window-to-wall ratio 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50% Has fins? true, false Has overhangs? true, false Fin depth (m) 0.30, 0.91, 1.5, 2.1 Overhang depth (m) 0.30, 0.91, 1.5, 2.1 Building shape a: 10, 15, 20, 25, 30
b: 10, 15, 20, 25, 30
c: 5, 10, 15, 20, 25
d: 10, 15, 20, 25, 30, 35, 40, 45, 50
e: 5, 10, 15, 20, 25
af
Number of buildings 3, 4 Number of floors 5, 6, 7, 8 Orientation 0, 5, 10, …, 345, 350, 355 Substructure system mat foundation, piles, footings
Material variable Values
Cladding brick, steel, stucco, vinyl, wood, limestone, concrete Roof precast hollow-core concrete, suspended concrete slab, open-web steel joist with
steel decking with steel decking Partitions steel stud 24” o.c. + gypsum board, concrete block + gypsum board Columns and beams concrete column/concrete beam, concrete column/WF beam, WF column/WF
beam beam Floor finishes ceramic floor tile, stone floor tile, cement facing tile with fiber, cement cast plaster floor, neoprene, polyacrylate with ground granite, polyurethane with vinyl chips, carpet tile + nylon, laminated veneer wood Floor insulation blown cellulose, extruded polystyrene, polystyrene foam slab, foam glass, glass
207
wool mat, rock wool Floor structure precast hollowcore + concrete topping, open-web steel joist + concrete topping, steel joist + concrete topping, suspended concrete slab + concrete topping Piles steel pipe, precast concrete Shading device steel, aluminum, concrete Wall finishes ceramic wall tile + acrylic paint, vinyl wall covering + acrylic paint Glazing material regular, low-e Thickness Variable Values
Ceiling (m) 0.0064, 0.011, 0.015, 0.019 Cladding (m) 0.025, 0.067, 0.11, 0.15
Material properties used to quantify building component embodied
impacts
Material Building Component(s) Density (kg/m3) Embodied impact factor (kg
CO2e/kg material)
Concrete Mat foundation 2400 .050
Concrete Footing 2400 5.2
Concrete Piles, grade beams, cladding,
shading
2400 0.12
Concrete Slab on grade 2400 .065
Concrete Columns + beams n/a 7.9
Concrete (precast hollowcore) Roof structure n/a 14
Concrete (precast hollowcore) +
concrete topping
Floor structure n/a 8.7
Concrete block (6”) + gypsum
board
Partitions n/a 1.7
Concrete (suspended slab) Roof structure n/a 21
Concrete (suspended slab) +
concrete topping
Floor structure n/a 16
Concrete + steel Columns + beams (wide
flange)
n/a 5.0
Cement facing tile with fiber Floor finishes 2400 1.1
Cement cast plaster floor Floor finishes 2400 0.37
Steel Rebar 8000 1.0
Steel Piles, cladding, shading 8000 1.9
Steel Columns (wide flange) +
beams (wide flange)
n/a 2.4
Steel stud (24” o.c.) + gypsum
board
Partitions n/a 15.2
Steel joist + concrete topping Floor structure n/a 5.9
Steel joist (open-web) + steel
decking
Roof structure n/a 11
Steel joist (open-web) + concrete
topping
Floor structure n/a 5.4
Limestone Cladding 2500 0.019
Limestone Floor finishes 2500 0.37
Stucco Cladding 800 0.070
Brick Cladding 2403 1.1
Neoprene Floor finishes 1230 2.5
Polyethylene Vapor barrier 950 1.6
208
Regular glass (flat glass) Glazing 2310 0.95
Low-e glass Glazing 2310 1.1
Vinyl Cladding 580 2.4
Vinyl Wall coverings 1360 1.8
Ceramic tile Wall coverings, floor finishes 1360 0.74
Acrylic paint Wall coverings 1200 2.7
Aluminum Shading devices 2700 11
Wood Cladding, floor finishes 600 0.33
Polyurethane foam (high density) Floor finish 400 4.0
Polypropylene (vapor retarder) Vapor barrier 946 1.7
Polyacrylate terrazzo Floor finishes 1054 3.2
Carpet tile Floor finishes 74 6.8
Blown cellulose Floor insulation 56 0.35
Extruded polystyrene Floor insulation 30 10
Polystyrene foam slab Floor insulation 30 3.4
Foam glass Floor insulation 110 1.5
Glass wool mat Floor insulation 40 1.4
Rock wool Floor insulation 46 1.0
Sample cost formulas
Uniformat element Assembly Sub-component (a)Cost formula
A: Substructure pile piles (steel) $58 * ((c)# interior grid intersections + (c)# exterior grid intersections)
* (d)pile depth
piles (concrete) $60.5 * ((c)# interior grid intersections + (c)# exterior grid
intersections) * (d)pile depth
vapor barrier $0.088 * ((b)slab area + (c)perimeter * (d)footing depth)
slab-on-grade $3.75 * (b)slab area
grade beam $70 * (c)perimeter
footings concrete footings +
rebar
$1282.5 * ((c)# interior grid intersections + (c)# exterior grid
intersections)
mat foundation foundation $142 * (b)slab area
B: Shell columns and beams concrete columns
and beams
$143 * ((c)# interior grid intersections + (c)# exterior grid
intersections) * (d)floor-to-floor height * number of floors
floor floor structure
(precast
hollowcore +
concrete topping)
$11.35 * (b)slab area * number of floors
roof roof structure
(precast
hollowcore
concrete)
$13.26 * (b)slab area
cladding cladding (steel) $16.85 * thickness * (1-WWR) * (c)perimeter * (e)height
C: Interiors partitions partition structure
(steel stud +
gypsum
board)
$3.8 * (c)perimeter * (e)height
wall finishes covering (ceramic
wall tile)
$3.99 * (c)perimeter * (e)height
floor finishes surface (stone floor
tile)
$12.05 * gross floor area
209
ceiling gypsum + paint $42.50 * (gross floor area + (b)slab area) (f) D: Services mechanical air conditioner gross floor area
electrical compact fluorescent
lighting fixture
ballast and lamps
(glass tube)
gross floor area
plumbing circulator pump gross floor area
fire fire extinguisher gross floor area
conveying elevator gross floor area (a) All formulas multiplied by number of buildings, a parameter given in Table 21. (b) Slab area = gross floor area / number of floors. (c) Perimeter and # grid intersections determined by length and width parameters, as given in Table 21. (d) Assumption described in “Case Study” section of Chapter 6. (e) Height = number of floors * floor to floor height. (f) The “Analysis process” sub-section of the “Methodology” section in Chapter 6 describes how an online facility operations reference
database takes as inputs gross floor area, a service life assumption, and other assumptions described in the “Case Study” section and returns
service costs.
210
References
Aboulnaga, M. (2006). Towards green buildings: Glass as a building element – the use and
misuse in the gulf region. Renewable Energy, 31(5), 631-653.
ACI Committee 302. (2004). Guide for concrete floor and slab construction. PDF file.
Adalberth, K., Almgren, A., and Peterson, E. (2001). Life cycle assessment of four
multi-family buildings. International Journal of Low Energy and Sustainable
Buildings, 2, 1-21.
Ahmed, S., Wallace, K., and Blessing, T. (2003). Understanding the differences between how
novice and experienced designers approach design tasks. Research in Engineering
Design, 14, 1-11.
Al-Homoud, M. (1997). Optimum thermal design of office buildings. International Journal
of Energy Research, 21(10), 941-957.
Al-Homoud, M. (2005). Performance characteristics and practical applications of common
building thermal insulation materials. Building and Environment, 40(3), 353-366.
Alsamsam, I., Lemay, L., and VanGeem, M. (2008). Sustainable high performance concrete
buildings. In Proceedings of the 2008 Structures Congress. American Society of
Civil Engineers.
ALwaer, H. and Clements-Croome, D. (2010). Key performance indicators (KPIs) and priority
setting in using the multi-attribute approach for assessing sustainable intelligent
buildings. Building and Environment, 45(4), 799-807.
American Institute for Aeronautics and Astronautics Inc. (AIAA). (1991). Current state of
the art in multidisciplinary design optimization. MDO Technical Committee.
Andreu, I. and Oreszczyn, T. (2004). Architects need environmental feedback. Building
Research and Information, 32(4), 313-328.
Arup, Inc. (2013). Arup Project Embodied Carbon Database. https://reporting.climateearth.com.
Accessed July 15, 2013.
American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE).
(2009). ASHRAE Handbook of Fundamentals. Atlanta: ASHRAE.
Ardente, F., Beccali, M., Cellura, M., and Mistretta, M. (2008). Building energy performance: A
LCA case study of kenaf-fibres insulation board. Energy and Buildings, 40(1), 1-10.
211
Athena EcoCalculator for Commercial Assemblies. (2011). Version 3.6. Athena Sustainable
Materials Institute.
Atkinson, C., Hobbs, S., West, J., and Edwards, S. (1996). Life cycle embodied energy and
carbon dioxide emissions in buildings. Industry and Environment, 2, 29-31.
Attia, S., Beltran, L., Herde, A., and Hensen, J. (2009). “Architect friendly”: A comparison of
ten different building performance simulation tools. In Proceedings of the 11th
International Building Performance Simulation Association Building Simulation
Conference. Glasgow, 204-211.
Augenbroe, G. (1992). Integrated building performance evaluation in the early design
stages. Building and Environment, 27(2), 149-161.
Augenbroe, G. and Winkelmann, F. (1990). Integration of simulation into building design;
the need for a joint approach. In Proceedings of ASME International Solar
Energy Conference. Miami, 247-255.
Azhar, S., Brown, J., and Sattineni, A. (2010). A case study of building performance
analyses using building information modeling. In Proceedings of the 27th
International
Symposium on Automation and Robotics in Construction. Bratislava, Slovakia.
Baker, N. and Steemers, K. (2000). Energy and environment in architecture: a technical
design guide. New York: E&FN Spon.
Basbagill J., Flager F., Lepech M., and Fischer, M. (2013). Application of life cycle assessment
to early stage building design for reduced embodied environmental impacts. Building
and Environment, 60, 81-92.
Bordass, B., Cohen, R., Standeven, M., and Leaman, A. (2001). Assessing building
performance in use 3: Energy performance of the Probe buildings. Building Research and
Information, 29(2), 114-128.
Bribian I., Capilla, A., and Uson, A. (2011). Life cycle assessment of building materials:
Comparative analysis of energy and environmental impacts and evaluation of the eco-
efficiency improvement potential. Building and Environment, 46(5), 1133-1140.
Bribian, I., Uson, A., and Scarpellini, S. (2009). Life cycle assessment in buildings: State-of-the-
art and simplified LCA methodology as a complement for building certification. Building
and Environment, 44(12), 2510-2520.
Caldas, L. and Norford, L. (2002). A design optimization tool based on a genetic algorithm.
Automation in Construction, 11(2), 173-184.
212
Caldas, L. (2008). Generation of energy-efficient architecture solutions applying GENE_ARCH:
An evolution-based generative design system. Advanced Engineering Informatics, 22(1),
59-70.
Charron, R. and Athienitis, A. (2006). The use of genetic algorithms for a net-zero energy solar
home design optimization tool. ASHRAE Transactions, 112(2), 285–295.
Chaszar, A. (2003). Bridging the gap with collaborative design programs. In Chaszar A., ed.
Blurring the lines: computer-aided design and manufacturing in contemporary
architecture. West Sussex: John Wiley & Sons.
Chen, M., Huang, D., Zhang, D., and Zhang, J. (2004). Three-dimensional geometric modeling
based on parametric design and its applications in enterprise production. International
Conference on Intelligent Mechatronics and Automation, 1:733-738.
Christenson, M., Manz, H., and Gyalistras, D. (2006). Climate warning impact on degree-days
and building energy demand in Switzerland. Energy Conversion and Management, 47(6),
671-686.
Chung, W., Hui, Y., and Lam, Y. (2006). Benchmarking the energy efficiency of commercial
buildings. Applied Energy, 83(1), 1-14.
Citherlet, S. (2001). Towards the holistic assessment of building performance based on an
integrated simulation approach. Lausanne: Swiss Federal Institute of Technology EPFL.
Cofaigh, E., Fitzgerald, E., Alcock, R., McNicholl, A., Peltonen, V., and Marucco, A. (1999). A
green Vitruvius: Principles and practice of sustainable architectural design. London:
James & James (Science Publishers) Ltd.
Cole, R. and Kernan, P. (1996). Life cycle energy use in office buildings. Building and
Environment, 31(4), 307-317.
Coley, D. and Schukat, S. (2002). Low-energy design: combining computer-based optimization
and human judgment. Building and Environment, 37(12), 1241-1247.
Comakli, K. and Yuksel, B. (2004). Environmental impact of thermal insulation thickness in
buildings. Applied Thermal Engineering, 24(5-6), 933-940.
The Construction Specifications Institute. (2010). Uniformat: A uniform classification of
construction systems and assemblies.
CostLab. (2011). Whitestone Research, Santa Barbara, CA, USA.
Crawford, R., Czerniakowski, I. and Fuller, R. (2010). A comprehensive framework for
assessing the life cycle energy of building construction assemblies. Architectural
Science Review, 53(3), 288-296.
213
Crosbie, T., Dawood, N, and Dean, J. (2010). Energy profiling in the life-cycle assessment of
buildings. Management of Environmental Quality: An International Journal, 21(1): 20-
31.
Crowther, P. (1999). Design for disassembly to recover embodied energy. The 16th
Annual
Conference on Passive and Low Energy Architecture. Melbourne/Brisbane/Cairns,
Australia.
de Graaf, Reinier. “Sustainable growth”. Lecture: Architecture of Consequence. American
Institute of Architects. San Francisco. September 12, 2011.
Deb, K. (2001). Multi-objective optimization using evolutionary algorithms. New York: John
Wiley & Sons.
DeKleine, R., Keoleian, G., and Kelly, J. (2011). Optimal replacement of residential air
conditioning equipment to minimize energy, greenhouse gas emissions, and consumer
cost in the US. Energy Policy, 39(6), 3144-3153.
Der-Petrossian, B. (2000). Environment-friendly construction practices. United Nations Center
for Human Settlements – UNCHS (Habitat), HS/596/00E, Vienna, 143-169.
de Wit, S. and Augenbroe, G. (2002). Analysis of uncertainty in building design evaluations and
its implications. Energy and Buildings. 34(9), 951-958.
Diakaki, C., Grigoroudis, E., and Kolokotsa, D. (2008). Towards a multi-objective optimization
approach for improving energy efficiency in buildings. Energy and Buildings, 40(9),
1747-1754.
Ding, G. (2004). The development of a multi-criteria approach for the measurement of
sustainable performance for built projects and facilities. Ph.D. Thesis. University of
Technology, Sydney, Australia.
Ding, G. (2008). Sustainable construction – The role of environmental assessment tools. Journal
of Environmental Management, 86(3), 451-464.
Dixit, M. (2010). Identification of parameters for embodied energy measurement: A literature
review. Energy and Buildings, 42(8), 1238-1274.
DProfiler. (2012). Rev. 2.0. Beck Technology, Ltd., Dallas, TX, USA.
Drogemuller, R., Crawford, J., and Egan, S. (2004). Linking early design decisions across
multiple disciplines. In Proceedings of the Conference on Product and Process
Modelling in the Building and Construction Industry. Istanbul: Taylor & Francis.
214
Eastman, C., Teicholz, P., Sacks, R., and Liston, K. (2008). BIM handbook: A guide to building
information modeling for owners, managers, engineers, and contractors. Hoboken: John
Wiley & Sons.
Ellis, P., Torcellini, P., and Crawley, D. (2008). Energy design plugin: an EnergyPlus Plugin for
SketchUp. IBPSA-USA SimBuild Conference. Berkeley.
eQUEST. (2010). Version 3.64. James J. Hirsch & Associates (http://www.doe2.com/eQUEST/).
Eštoková, A., Porhincák, M., and Ruzbacký, R. (2011). Minimization of CO2 emissions and
primal energy by building materials’ environmental evaluation and optimization.
Chemical Engineering Transactions, 25, 653-658.
European Commission’s Directorate General for Enterprise and Industry (ECDGEI). (2007). Life
cycle costing (LCC) as a contribution to sustainable construction: a common
methodology – final report. Davis Langdon Management Consulting.
Excel. (2007). Microsoft Corporation, Redmond, WA, USA.
Fawcett, A. (2003). Architecture design notebook. Burlington, Massachusetts: Architectural
Press.
Fay, R., Treloar, G., and Iyer-Raniga, U. (2000). Life cycle energy analysis of buildings: a case
study. Building Research and Information, 28(1), 31-41.
Flager, F., Basbagill, J., Lepech, M., and Fischer, M. (2012). Multi-objective building envelope
optimization for life-cycle cost and global warming potential. In: Gudnason, G. and
Scherer, R., eds., eWork and eBusiness in Architecture, Engineering and Construction:
Proceedings of the European Conference on Product and Process Modelling 2012.
London: Taylor & Francis, 193-200.
Fumo, N., Mago, P., and Luck, R. (2010). Methodology to estimate building energy consumption
using EnergyPlus benchmark models. Energy and Buildings, 42(12), 2331-2337.
Gero, J., D’Cruz, N., and Radford, A. (1983). Energy in context: A multicriteria model for
building design. Building and Environment, 18(3), 99-107.
Geyer, P. (2009). Component-oriented decomposition for multidisciplinary design optimization
in building design. Advanced Engineering Informatics, 23(1), 12-31.
Gluch, P. and Baumann, H. (2004). The life cycle costing (LCC) approach: a conceptual
discussion of its usefulness for environmental decision-making. Building and
Environment, 39(5), 571-580.
Grierson, D. and Khajehpour, S. (2002). Method for conceptual design applied to office
buildings. Journal of Computing in Civil Engineering, 16(2), 83-103.
215
Gruman, D. (2003). ASHRAE Green Guide. Atlanta: American Society of Heating Refrigerating
and Air-Conditioning Engineers, Inc.
Gustavsson, L. and Sathre, F. (2006). Variability in energy and carbon dioxide balances of wood
and concrete building materials. Building and Environment, 41(7), 940-951.
Hacker, J., De Saulles, T., Minson, A., and Holmes, M. (2008). Embodied and operational
carbon dioxide emissions from housing: A case study on the effects of thermal mass and
climate change. Energy and Buildings, 40(3), 375-384.
Hauglustaine, J. and Azar, S. (2001). Interactive tool aiding to optimize the building envelope
during the sketch design. In Proceedings of the Seventh International IPBSA Conference,
387-394.
Houghton, J., Ding, Y., Griggs, D., Noguer, M., van der Linden, P., Dai, X., Maskell, K.,
Johnson, C., (eds.). (1995). Climate Change 1995 – The Science of Climate Change.
Cambridge: Cambridge University Press.
Huberman, N. and Pearlmutter, D. (2008). A life cycle energy analysis of building materials in
the Negev desert. Energy and Buildings, 40(5), 837-848.
Jacobs, P. and Henderson, H. (2002). State-of-the-art review: whole building, building envelope,
and HVAC component and system simulation and design tools. Project 605-30010,
Architectural Energy Corporation, Boulder, Colorado.
Jentsch, M., James, P., and Bahaj, A. (2010). Climate change adapted simulation weather data:
implications for cities in hot, arid climates of the Middle East. World Renewable Energy
Congress XI, Abu Dhabi, UAE.
Jolliet, O., Margni, M., Charles, R., Humbert, S., Payet, J., and Rebitzer, G. (2003). IMPACT
2002+: A new life cycle impact assessment methodology. The International Journal of
Life Cycle Assessment, 8(6), 324-330.
Juan, Y., Gao, P., and Wang, J. (2010). A hybrid decision support system for sustainable office
building renovation and energy performance improvement. Energy and Buildings, 42(3),
290-297.
Kazim, A. (2007). Assessments of primary energy consumption and its environmental
consequences in the United Arab Emirates. Renewable and Sustainable Energy Reviews,
11(3), 426-446.
Keoleian, G., Phipps, A., Dritz, T., and Brachfeld, D. (2004). Life cycle environmental
performance and improvement of a yogurt product delivery system. Packaging
Technology and Science, 17(2), 85-103.
216
Kienzl, N., Chaszar, A., and Stoller, P. (2003). Environmental engineering: integrating computer
simulation into the design process. In Chaszar A., ed. Blurring the lines: computer-aided
design and manufacturing in contemporary architecture. West Sussex: John Wiley &
Sons.
Kotaji, S., Schuurmans, A., and Edwards, S. (2003). Life cycle assessment in building and
construction: a state-of-the-art report of SETAC-Europe. Brussels: Society of
Environmental Toxicology & Chemist.
Krarti, M., Erickson, P., and Hillman, T. (2005). A simplified method to estimate energy savings
of artificial lighting use from daylighting. Building and Environment, 40(6), 747-754.
Krygiel, E. and Nies, B. (2008). Green BIM: Successful sustainable design with building
information modeling. Indianapolis: Wiley Publishing.
Langston, C. and Ding, G. (2001). Sustainable Practices in the Built Environment, Butterworth
Heinemann.
Lawson, B., Partridge, H., and Gelder, J. (1995). Assessing the environmental impact of building
materials. PRO 1. RAIA Environment Design Guide. Redhill, Australia.
LC-Impact. (2013). Life cycle impact assessment (LCIA). http://www.lc-impact.eu/life cycle-
impact-assessment-lcia. Accessed November 2013.
Lee, K., Tae, S., and Shin, S. (2009). Development of a life cycle assessment program for
building (SUSB-LCA) in South Korea. Renewable and Sustainable Energy Reviews,
13(8), 1994-2002.
Lin, V., Gossard, D., and Light, R. (1981). Variational geometry in computer-aided design.
Computers & Graphics, 15(3): 171-177.
Magnier, L. and Haghighat, F. (2010). Multiobjective optimization of building design
using TRNSYS simulations, genetic algorithm, and Artificial Neural Network.
Building and Environment, 45(3), 739-746.
ModelCenter. (2008). Version 10.0., Phoenix Integration, Inc., Blacksburg, VA, USA.
Morbitzer, C., Strachan, P., Webster, J., Spires, B., and Cafferty, D. (2001). Integration of
building simulation into the design process of an architecture practice. Proceedings of the
Seventh International IBPSA Conference.
Mumma, T. (1995). Reducing the embodied energy of buildings. Home Energy Magazine, 12(1).
Norris G. (2001). Integrating life cycle cost analysis and LCA. The International Journal of Life
Cycle Assessment, 6(2), 118-120.
217
Norton, B., Lo, S., Kronin, K., and Yohanis, Y. (1995). Towards the harmonius integration of
information technology in building design and construction. International Journal of
Ambient Energy, 16(2), 95-109.
O’Donnell, J., Morrissey, E., Keane, M., and Bazjanac, V. (2004). BuildingPI: A future tool for
building life cycle analysis. Boulder, CO. doi: DE-AC02-05CH11231.
Ospina-Alvarado, A. and Castro-Lacouture, D. (2010). Interaction of processes and phases in
project scheduling using BIM for A/E/C/FM integration. Construction Research
Congress 2010: Innovation for Reshaping Construction Practice, 1, 939-948.
Papamichael, M. (1991). Design process and knowledge, possibilities and limitations of
computer aided design. Ph.D. Thesis. University of California, Berkeley.
Peippo, K., Lund, P., and Vartiainen, E. (1999). Multivariate optimization of design trade-offs
for solar low energy buildings. Energy and Buildings, 29(2), 189-205.
Perez-Lombard, L., Ortiz, J., and Pout, C. (2008). A review on buildings energy consumption
information. Energy and Buildings, 40(3), 394-398.
Pidgeon, A. (May 2012). Improved life cycle performance for construction of big box retail.
Master’s Thesis. Illinois Institute of Technology, http://hdl.handle.net/10560/2744.
Accessed August 14, 2013.
Pierquet, P., Bowyer, J. and Huelman, P. (1998). Thermal performance and embodied energy of
cold climate wall systems. Forest Products Journal, 48(6), 53-60.
Pulselli, R., Simoncini, E., and Marchettini, N. (2009). Energy and emergy based cost-benefit
evaluation of building envelopes relative to geographical location and climate. Building
and Environment, 44(5), 920-928.
Pushkar S., Becker, R., and Katz, A. (2005). A methodology for design of environmentally
optimal buildings by variable grouping. Building and Environment, 40(8), 1126-39.
Radford, A. and Gero, J. (1980). Trade-off diagrams for the integrated design of the physical
environment in buildings. Building and Environment, 15(1), 3-15.
Radhi, H. (2010). On the optimal selection of wall cladding system to reduce direct and indirect
CO2 emissions. Energy, 35(3), 1412-1424.
Rafiq, M., Mathews, J., and Bullock, G. (2003). Conceptual building design – evolutionary
approach. Journal of Computing in Civil Engineering, 17(3), 150-158.
Rai, D., Sodagar, B., Fieldson, R., and Hu, X. (2011). Assessment of CO2 emissions reduction in
a distribution warehouse. Energy, 36(4), 2271-2277.
Revit. (2013). San Rafael, California: Autodesk, Inc.
218
The Royal Australian Institute of Architects (RAIA). (2004). Towards a national framework for
energy efficiency – Issues and challenges: 1-36.
RSMeans Facilities Construction Cost Data 2008. (2007). Kingston, MA: RSMeans Construction
Publishers & Consultants.
Sandrolini, F. and Franzoni, E. (2009). Embodied energy of building materials: A new
parameter for sustainable architectural design. International Journal of Heat and
Technology, 27(1), 163-167.
Scheuer, C. and Keolian, G. (2002). Evaluation of LEED using life cycle assessment
methods. National Institute of Standards and Technology (NIST), US Department
of Commerce, Michigan.
Scheuer, C., Keoleian, G., and Reppe, P. (2003). Life cycle energy and environmental
performance of a new university building: modeling challenges and design implications.
Energy and Buildings, 35(10), 1049-1064.
Schlueter A., and Thesseling, F. (2009). Building information model based energy/exergy
performance assessment in early design stages. Automation in Construction, 18(2), 153-
163.
Schoch, M., Prakasvudhisarn, C., and Praditsmanont, A. (2011). Building-volume designs with
optimal life cycle costs. International Journal of Architectural Computing, 9(1), 55-76.
Sharma, A., Saxena, A., Sethi, M., Shree, V., and Varun. (2011). Life cycle assessment of
buildings: A review. Renewable and Sustainable Energy Reviews, 15(1), 871-875.
Shea, K., Aish, R., and Gourtovaia, M. (2005). Toward integrated performance-driven
generative design tools. Automation in Construction, 14(2), 253-264.
SimaPro. (2010). Version 7.3.2. PRe Consultants, The Netherlands.
Spitzley D., Grande, D., Keoleian, G., and Kim, H. (2005). Life cycle optimization of ownership
costs and emissions reductions in US vehicle retirement decisions. Transportation
Research Part D: Transport and Environment, 10:2(March), 161-175.
Struck, C. and Hensen, J. (2007). On supporting design decisions in conceptual design
addressing specification uncertainties using performance simulation. In Proceedings of
the 10th
IBPSA Building Simulation Conference.
Suzuki, M. and Oka, T. (1998). Estimation of life cycle energy consumption and CO2 emission
of office buildings in Japan. Energy and Buildings, 28(1), 33-41.
219
Taylor, M. and Carper, K. (2007). Designing with technology: a collaborative and creative
process. In: McClure, W. and Bartuska, T., eds., The built environment: A collaborative
inquiry into design and planning. Hoboken: John Wiley & Sons.
Terzidis, K. (2006). Algorithmic architecture. Oxford: Architectural Press.
Thormark, C. (2002). A low energy building in a life cycle – its embodied energy, energy needed
for operation and recycling potential. Building and Environment, 37(4), 429-435.
Thormark, C. (2006). The effect of material choice on the total energy need and recycling
potential of a building. Building and Environment, 41(8), 1019-1026.
Tucker, S., Ambrose, M., Johnston, D., Newton, P., Seo, S., and Jones, D. (2003). LCADesign:
An integrated approach to automatic eco-efficiency assessment of commercial buildings.
In CIB W78 Conference on Construction IT proceedings.
Tuhus-Dubrow, D. and Krarti, M. (2010). Genetic-algorithm based approach to optimize
building envelope design for residential buildings. Building and Environment, 45(7), 1574-
1581.
Turpin, J. (2007). Cultural blueprints of human existence. In The built environment: a
collaborative inquiry into design and planning, McClure, W. and Bartuska, T., eds.
Hoboken: John Wiley & Sons.
US Department of Energy (USDOE). (2011a). 2010 buildings energy data book.
http://buildingsdatabook.eere.energy.gov. Accessed April 2011.
US Department of Energy (USDOE). (2011b). Building energy software tools directory.
http://apps1.eere.energy.gov/buildings/tools_directory. Accessed November 2011.
US Department of Energy (USDOE). (2011c). Emissions of greenhouse gases in the United
States 2009. Washington, DC: US Energy Information Administration.
US Energy Information Administration. (2010). Architecture 2030.
http://architecture2030.org/the_solution/buildings_solution_how. Accessed September
2011.
US Green Building Council. (2011a). http://www.usgbc.org. Accessed September 2011.
US Green Building Council. (2011b).
http://www.usgbc.org/DisplayPage.aspx?CMSPageID=2078. Accessed
September 2011.
van Hattem, D. (1987). Proceedings of the Workshop on Future Building Energy Modelling.
Commission of the European Communities, Ispra.
220
Venkatarama, B. and Jagadish, K. (2003). Embodied energy of common and alternative building
materials and technologies. Energy and Buildings, 35(2), 129-137.
Viracon. 2013 product guide. Insulating glass.
http://viracon.com/images/pdf/ProductGuide2013.pdf. Accessed July 12, 2013.
Wang, W., Rivard, H., and Zmeureanu, R. (2005a). An object-oriented framework for
simulation-based green building design optimization with genetic algorithms.
Advanced Engineering Informatics, 19(1), 5-23.
Wang, W., Rivard, H., and Zmeureanu, R. (2005b). Applying multi-objective genetic
algorithms in green building design optimization. Building and Environment,
40(11), 1512-1525.
Wang, W., Zmeureanu, R., and Rivard, H. (2006). Floor shape optimization for
green building design. Advanced Engineering Informatics, 20(4), 363-378.
Watson, D. and Labs, K. (1983). Climate design: Energy-efficient building principles and
practices. Texas: McGraw-Hill.
Wetter, M. (2001). GenOpt – A Generic Optimization Program. In: Lamberts, R., Negarao, C.,
and Hensen, J., eds. Proceedings of the Seventh International IBPSA Conference, 601-
608.
Wetter, M. and Wright, J. (2004). A comparison of deterministic and probabilistic optimization
algorithms for nonsmooth simulation-based optimization. Building and Environment,
39(8), 989-999.
Wilson, J. and Tagaza, E. (2005). Green building in Australia: Drivers and barriers. In
Proceedings of the Australian Structural Engineering Conference. New Castle,
Australia: Swinburne University of Technology.
Winter, S. (2008). Green residential building in North America: A perspective from
the United States (Commission for Environmental Cooperation Paper 4b). In Secretariat
Report to Council under Article 13 of the North American Agreement on Environmental
Cooperation. http://www.cec.org/Storage/61/5380_GBPaper4b_en.pdf. Accessed
September 30, 2011.
Winther, B. and Hestnes, A. (1999). Solar versus green: the analysis of a Norwegian row house.
Solar Energy, 66(6), 387-393.
Wright, J. and Farmani, R. (2001). In Proceedings of the Seventh International IBPSA
Conference on Simultaneous Optimization of Building Fabric Construction, HVAC
System Size, and the Plant Control Strategy. Rio de Janeiro, 865-872.
221
Wright, J., Loosemore, H., and Farmani, R. (2002). Optimization of building thermal design and
control by multi-criterion genetic algorithm. Energy and Buildings, 34(9), 959-972.
Wright, L., Kemp, S., and Williams, I. (2011). 'Carbon footprinting': towards a universally
accepted definition. Carbon Management, 2(1), 61-72.
Xiong, H. and Zhao, Y. (2011). Environment impact comparison of different structure systems
based on life cycle assessment methodology. Advanced Materials Research, 243-249,
5275-5279.
Yohanis, Y. and Norton, B. (2002). Life cycle operational and embodied energy for a generic
single-storey office building in the UK. Energy, 1, 77-92.
Yohanis, Y. and Norton, B. (2006). Including embodied energy considerations at the conceptual
stage of building design. Proceedings of the Institution of Mechanical Engineers, Part A:
Journal of Power and Energy, 220(3), 271-288.