Energy and Buildings 84 (2014) 426–441

Contents lists available at ScienceDirect

Energy and Buildings

journal homepage: www.elsevier.com/locate/enbuild

Evolutionary energy performance feedback for design:

Multidisciplinary design optimization and performance boundaries
for design decision support
Shih-Hsin Lin, David Jason Gerber ∗
School of Architecture, University of Southern California, 316 Watt Hall, Los Angeles, CA 90089, USA

a r t i c l e i n f o a b s t r a c t

Article history: In pursuit of including energy performance as feedback for architects’ early stage design decision mak-
Received 3 June 2014 ing, this research presents the theoretical foundation of a designer oriented multidisciplinary design
Received in revised form 20 August 2014 optimization (MDO) framework titled evolutionary energy performance feedback for design (EEPFD).
Accepted 21 August 2014
Through a comprehensive literature review and gap analysis EEPFD is developed into an MDO methodol-
Available online 30 August 2014
ogy that provides energy performance as feedback for influencing architects’ decision making more fluidly
and earlier than other approaches to date. Secondly, in response to the lack of an MDO best practice EEPFD
Keywords:
is investigated and evaluated through two experiments. The first experiment demonstrates the ability
Multidisciplinary design optimization
Energy performance feedback
to utilize EEPFD provided energy performance as feedback to pursue multiple architectural designs with
Early stage design competing objectives and tradeoffs. The second experiment identifies performance boundaries as a best
Design exploration practice for MDO applications to the early stage architectural design processes. The research synthe-
Performance-based design sizes the results into the basis for measuring these performance boundaries as a best practice in the
Parametric design context where architects must gauge multiple design concepts with varying complexity coupled with
Genetic algorithm performance objectives through EEPFD, thereby enhancing the influence of energy performance feed-
Multi-objective optimization back on the early stage design process. Finally, future research into the use of performance boundaries
Design decision support
for conceptual energy performance design exploration is discussed.
Design domain integration
© 2014 Elsevier B.V. All rights reserved.

1. Introduction decision-making synchronously during the early stages of architec-

In the current field of architecture the issue of sustainable design
and building performance, especially with respect to energy per-
formance, has become increasingly significant to the overall design
process. This growing emphasis is primarily attributed to the fact
that buildings account for a majority of all consumed energy, nearly
one half (48.7%) in the United States [1] and up to 40% of all energy
consumption in the European Union [2]. Research indicates that
incorporation of energy performance feedback at the early stage of
the design process can potentially increase the energy efficiency
over a building design’s entire lifecycle [3–8]. However, multiple
obstacles exist inhibiting seamless and timely inclusion of energy
performance feedback during this design phase. This leads to the
research pursuit of a “designing-in performance” methodology and
“best practice” where architects are able to utilize energy perfor-
mance as feedback to influence their design exploration and design
∗ Corresponding author. Tel.: +1 617 794 7367.
E-mail addresses: dgerber@usc.edu, djgerber@post.harvard.edu (D.J. Gerber).
tural design exploration. Our work is based in part on the simple
hypothesis that if the energy performance data, such as first and
second cost is available, designers will be influenced to choose and
pursue higher performing designs.
Despite the acknowledged beneficial impact of considering
energy performance early in design [9], obstacles between design
and energy simulation domains often prevent the inclusion of
energy performance during the early stage design process [10–12].
Multiple efforts have been made to resolve these issues, includ-
ing research into improved interoperability, platform integrations,
design automation, and multi-objective optimization techniques.
Among these efforts, multidisciplinary design optimization (MDO),
which combines multi-objective optimization (MOO) algorithms
with parametric design, demonstrates a great potential as an ini-
tial design exploration methodology that is capable of providing
rapid visual and analytical feedback for early stage design decision-
making. However, the application of MDO during the early stage
of the design process to support designers’ decision making has
not been adequately explored. Consequently, whether the provided
energy performance feedback from an MDO can actually support

http://dx.doi.org/10.1016/j.enbuild.2014.08.034
0378-7788/© 2014 Elsevier B.V. All rights reserved.
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
architectural design decision making is still a key research question.
Furthermore, due to the unique nature and extensive uncertainty of
the early stage architectural design process, arguably distinct from
other industries utilizing MDO, a best practice of MDO application
for building design is needed to be defined and explored.
This research hypothesizes that by providing architectural
designers with a designer-oriented MDO framework, enabling
rapid access to improved energy performance feedback with
visualized and quantified trade off analysis, will allow energy per-
formance feedback to influence early stage design decision-making.
Consequently, higher performing design can be more efficiently
realized while pursuing the complex set of coupled and indepen-
dent design goals. For these reasons, a design and energy centric
MDO framework, evolutionary energy performance feedback for
design (EEPFD), is first introduced and manifested through a pro-
totype tool developed to fulfill the unique needs of the early stage
architectural design process, where diversity in ideation, geome-
try, and programming are needed [13]. In addition, the framework
provides improved results through genetic algorithm based opti-
mization irrespective of the exploration stopping point, which is
typically determined by the dominant limiting factor of time in
early stage design exploration [14]. This paper continues to use the
developed framework, EEPFD, to addresses two critical gaps found
in precedent and literature research. The first is the lack of evidence
that supports the effectiveness of using MDO for providing energy
performance as feedback for influencing architectural designers’
decision making. The second gap is that, due to the idiosyncratic
nature of architectural design problems, a best practice for apply-
ing MDO to the early stage of the architectural design process is
needed but has yet to be adequately explored and validated. In
response this paper addresses these gaps through the following
two objectives: (1) to observe whether the provided energy perfor-
mance feedback from EEPFD influences designers’ design decision
making; and (2) to observe the use of performance boundaries
to gauge competing design concepts. A performance boundary is
hypothesized as advantageous as it is found automatically within
the solution space being generated by EEPFD more rapidly than
the mathematical convergence of a single design concept, as is
the current norm. Therefore, this paper first provides a literature
review identifying gaps in current efforts for bridging design and
energy simulation domains followed by current applications and
gaps related to energy performance optimization. Second, EEPFD
is introduced and explained in brief. Thirdly, the experiment to
observe the influence of the provided energy performance feedback
from EEPFD on the designers’ decision making is outlined along
with the experiment to support the development of a best practice
for using performance boundaries to gauge design concepts more
rapidly while maintaining early stage ideation and design diversity.
Finally, the research findings from these two experiments in the
context of the literature, limitations and future work are concluded
and discussed.
2. Bridging architectural design and energy simulation
It has been widely recognized that incorporating building per-
formance feedback to support design decision making can improve
the building performance throughout the building lifecycle when
such goals are set as a priority [3–8]. However, due to an array
of challenges the obstacles between design and energy perfor-
mance still remain [4,15–21]. Extensive prior research provides
an exhaustive comparison analysis of the variety of energy sim-
ulation tools available along with their pros and cons [10,22–26].
Separately, additional reviews have focused on the specific needs
of architectural designers looking to utilize energy simulation tools
[8,16,27–29]. However, an overall holistic approach to bridge the
427
gap between the design and energy simulation domains is still
undefined. In order to provide a different lens and isolate a poten-
tial holistic approach to bridging the gap, this research reviews
and synthesizes current efforts from a design process standpoint
and groups the current obstacles and potential solutions into three
major categories for discussion: (1) tools and tool interoperability;
(2) domain knowledge integration; and (3) design decision systems
support.
2.1. Tools and tool interoperability
Previous researchers reveal that currently available energy sim-
ulation tools are not considered to be “architect-friendly” for use
by designers during the early phases of design [16,27–29]. More
specifically, only 10% of the tools available are intended for archi-
tects’ use and only 1% of these are able to support the early
design stage [30]. In addition to the limited tool availability, seam-
less integration between software programs is typically lacking.
Thus, the necessary data transfer between tools leads to the loss
of information and knowledge capture, incurring inefficient man-
ual modification of models between design and energy simulation
tools [20,21]. While some of these issues can be resolved by a
standardized data exchange format such as IFC or gbXML, trans-
lating the solid building element geometry into space boundaries
for energy simulations, these formats are presently still limited in
supporting complex geometrical exploration, a critical requirement
for assisting contemporary architectural design [31–33]. Conse-
quently, despite the resolution of the data translation issues, these
tools are ill suited for architectural designers’ use during the early
stage design process [16,27,34,35].
Others attempt to improve the user interface to expand the
usability of the tools or facilitate greater data translation among
different platforms. The majority of these efforts have focused on
addressing the critical obstacle of interoperability across different
software platforms, applications, or user groups, including soft-
ware developers, researchers, and variable members of the building
industry. These aforementioned efforts have explored scripting
interfaces and self-developed plugins attempting to solve exist-
ing interoperability issues between the design and performance
analysis domains. While solutions to tools and tool interoperability
would ease the process in the generation and evaluation of design
alternatives, this effort is arguably insufficient [36]. The informa-
tion transferred from a design tool to an energy analysis tool is
still a one-way trip. In other words, there is no way of implemen-
ting the knowledge gained or the changes identified to the design
performed in the energy simulation tool, except through manually
re-entering the information.
2.2. Domain knowledge integration
In addition to the limited number and availability of designer-
oriented energy simulation tools, design professionals are often
unfamiliar with energy simulation tools as the functionalities are
often outside of their expertise [27,31,33]. This is reported as one
of the major reasons preventing architects from using energy sim-
ulation tools [34]. As a result, environmental simulation software
is routinely operated by domain experts who are familiar with the
specialized nature of these tools [37]. Consequently, domain expert
engineers regularly conduct energy simulations and their work is
typically used for performance validation post priori or later in the
design development phase rather than to support earlier design
decision making [26,38]. In addition, designers are faced with chal-
lenges in interpreting the results and successfully incorporating the
feedback into their design process [21].
Another stream of research has focused on the knowledge
exchange between different expertise domains. Examples of these
428 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
efforts can be found in data model and process standardizations
[39–41], along with collaborative framework developments. A
selection of the representative efforts is found here [26,42–50]. In
summary, these efforts focus on collaborative frameworks to over-
come interoperability between expert domains and to facilitate the
inclusion of performance feedback. These efforts all acknowledge
that solving the domain knowledge exchange issues across differ-
ent platforms is a prerequisite for further progress in the fields of
architecture and engineering disciplines related to the built envi-
ronment. In addition, they all conclude the need for an integrated
platform to provide a supportive environment and acknowledge
the needs of a designer-specific and designer friendly user inter-
face. However, most of the frameworks discussed above are still
very conceptual and in need of further development. While the
intent of these frameworks is to support design decision-making
in early project phases, with expert knowledge support to over-
come insufficient domain knowledge, process evaluation of these
frameworks adopted either by researchers or by practitioners is
still lacking [51,52].
2.3. Design decision systems support
Of equal importance are the findings that (1) simulation results
require expert set up and analysis cycles, and (2) therefore do
not provide immediate feedback necessary for conceptual design
exploration. Examples of these simulation failures include inabil-
ity to deliver informative context to support designers’ early stage
design decision-making, i.e. quantifiable evaluation of design solu-
tions, context specific analysis, design strategy comparisons, or
optimization of design solutions [8,32,53]. Furthermore, precedent
research suggests that tradeoff studies are necessary in order to
provide adequate feedback for design decision-making [54,55].
However, conventionally adopted performance-based analysis
methods have been shown by prior studies to be ill suited in
providing either the desired context or necessary information for
trade-off studies due to design cycle latency intrinsic in norma-
tive methodologies for early stage design process [56]. In particular
there is an implication that a system and user interface capa-
ble of visualizing analysis results and corresponding geometric
configurations is a necessary prerequisite to assist in design deci-
sion making [24]. Unfortunately, current energy simulation tools
often lack the ability to provide results with these types of visu-
alizations. Therefore, 3D geometric visualization in parallel with
the analysis data graphical representation is seen as desirable
for providing the context for enhancing design decision making
[21,24].
In support of design decision making, there are efforts focused
on the development of user interfaces for facilitating the pro-
cess, such as providing 2D and 3D design and data visualization
[16,24,57]. However, an essential obstacle to overcome is the
insufficient feedback and the absence of a cost–benefit (trade-
off) analysis, which would enable designers to understand the
impact of their decisions. In addition, in order to better incorpo-
rate energy simulation results for design decision making, rapid
evaluation of design alternative is needed [58]. To overcome these
limitations, a group of efforts focus on enabling rapid design alter-
native generation, evaluation, and a comparison of analysis results
with tradeoff studies for identifying “best fits” across competing
objectives [13,59–69]. Another focus area is the promotion of the
sensitivity analysis (SA) and uncertainty analysis (UA) to support
the decision maker in identifying the most sensitive parameters
to make the designer aware of the risks associated with each
option, in particular if they affect a specific performance aspect
[70–74]. Lastly, a group of efforts focus on the application of multi-
criteria decision making (MDCM) as the decision-making process
for addressing multiple competing criteria and uncertainty inclu-
sive of energy performance feedback [75–79].
2.4. State of the art and gap analysis of MDO for building
performance
MDO has drawn the attention of the Architecture Engineering
and Construction (AEC) industry due to being capable of providing
potential solutions to overcome barriers that exist between design
and engineering analysis i.e. performance domains. The applica-
tion of MDO can be found in explorations of space layout, façade
configuration, building shape, HVAC optimization, structural sizing
and member complexity optimization, thermal comfort, daylight-
ing, and energy performance [11,80–88]. Based on the precedent
research into relevant optimization techniques building design
researchers have identified the use of genetic algorithm (GA) based
multi-objective optimization methods as having strong potential
for tackling design problems with varying degrees of parame-
ter coupling and complexity and large degrees of uncertainty
[51,69,89]. In addition, the majority of applications of optimization
techniques for building performance purposes is primarily from
within engineering disciplines with a focus on research rather than
accommodating design practitioner needs, such as ease of use, short
analysis time, high returns on investment, and flexibility for varying
design problems [13,51,89].
A comprehensive review of 166 related research papers from
115 research groups from the fields of architecture, building sci-
ence, engineering, and design and computation was conducted.
This selection of research papers extends from 1978 to 2013
and is focused on topics relating to computational optimiza-
tion techniques applied in architecture, engineering and building
performance. Although the works within the building energy sim-
ulation and design domain are the focus, the study included other
performance analysis domains in the review. The review sought
evidence and criteria inclusive of energy performance as feedback,
design phase relevance, designer centricity, building typology,
complexity of geometry explored, and simultaneously considered
objective counts during the optimization process, as these charac-
teristics are the crucial criteria for designer use of MDO during early
stages of the design process. The list in its entirety can be found in
Lin [90]. Among the reviewed works, more than half (54%) of the
works include energy performance as part of their objective. Among
these efforts, only 15% of the works reviewed are for designer use
with only 3% specifically intended for the early stages of design.
When all the reviewed researches are considered only 8% con-
tribute to early stage design. While half of these are designated for
designer use, none of these support exploration of complex geom-
etry, which we consider an important objective in order to provide
relevance to contemporary architectural design and practice. When
considering all the reviewed works that do support the exploration
of complex geometry, none are found to support the early design
phase with more than 2 simultaneously considered objective func-
tions. Among the reviewed 166 research papers, only 12 papers
from 3 research groups, apply MDO from the designers’ view point
and explore building geometry beyond a simplified box. Among
these 3 research groups, all 3 include whole building energy sim-
ulation as one of the performance objectives. However, none of
these three evaluate more than 2 objectives at the same time. In
addition, these examples have some acknowledged issues, such as
the flexibility of the framework and the lacking of framework eval-
uation by and for designers [13]. Furthermore, the applications of
building performance optimization in the actual design process by
designers and practitioners need to be explored to understand the
actual benefits and difficulties a MDO methodology lends to design
decision-making, as identified by prior review [51].
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
3. Motivation and objectives
In order to enable energy performance feedback to influence
design decision making our work is focused on closing the gap
between design and energy simulation silos. Based on the liter-
ature and analysis, an integral approach to achieve designing-in
energy performance should be inclusive of the following criteria:
(1) designer friendly user interface; (2) should include platform
integration and automation alleviating interoperability issues and
reducing design cycle latency; (3) should provide rapid generation
and evaluation of design alternatives: (4) should provide trade-
off analysis for competing criteria; (5) should provide sensitivity
analysis and uncertainty analysis to inform the impact of decisions
made; (6) should provide data interpretation guidance to overcome
domain knowledge gaps; and (7) provide user friendly visualiza-
tion of geometric configurations in parallel with data to illustrate
performance comparisons for design alternatives. The review has
identified a promising applicability of MDO to meet the above
criteria and therefore enable a “designing-in performance” envi-
ronment [11,20]. However, the relationship between design form
exploration and energy performance has been largely excluded
from previous research efforts. Furthermore, current attempts have
yet to fully explore the applicability of this approach in the con-
text of the early stage design, where rapid exploration of variety
and alternatives is desired. In response to this gap in existing
research, there is a need for a designer-oriented MDO framework
that incorporates both conceptual energy analysis and the explo-
ration of overall building geometric configurations for the purpose
of providing early stage energy performance feedback. The use of a
designer-centric MDO framework can then be explored regarding
its usability and its effectiveness to enable energy performance
feedback to support designers design decision making.
As a result, the overall aim of the research is to test and validate a
designer-centric MDO design framework which includes the above
outlined criteria as its theoretical foundation. In order to pursue
this overall goal, a designer appropriate MDO framework, EEPFD, is
first developed then validated against the needs of designers dur-
ing the early stage of design [13]. Next, process evaluation metrics
and measurements are established as a means to understand the
effectiveness of the developed framework [69]. The research then
uses the established metrics to evaluate the processes designers
engage in both pedagogical and professional settings to initially
understand the applicability and impact of EEPFD on the design
process and the comparability of these processes when using the
established metrics. Prior publications of the authors address the
detailed development of the designer-oriented MDO design frame-
work and are validated against the needs of the early stage design
process [13]. In addition, a series of process evaluations through
pedagogical and design simulation experiments using the estab-
lished evaluation metrics are conducted to benchmark evaluation
of the framework [69,91–93]. Here, the focus of this paper is to
present new results of case-based experiments to further under-
stand the effectiveness of EEPFD in providing energy performance
as feedback to influence designers’ decision-making and energy
savings behavior of the designer and their selected architectural
designs.
4. Introduction to evolutionary energy performance
feedback for design (EEPFD)
EEPFD incorporates both conceptual energy analysis and the
exploration of building geometric configurations for the purpose of
providing early stage designs with energy performance feedback.
While it is acknowledged that MDO methodology is applicable to
various objectives and types of performance feedback, this research
429
focuses on investigating whether EEPFD is capable of enabling
a “designing-in performance” environment where energy perfor-
mance feedback can synchronously and directly influence design
decision-making. The theoretical foundation of the EEPFD is built
upon the seven identified needs, as summarized above. As illus-
trated in Fig. 1, the overall framework can be grouped into three
major components; problem formulation, automation, and design
decision support. (1) Problem formulation and building the design
model parametrically is the first interaction between the user
and the automated system. In order to ensure usability by the
intended audience, a designer friendly environment capable of
accommodating the varying complexity of diverse design prob-
lems is necessary and enabled. (2) Automating the translation of
a design model to an energy model helps to overcome inconsis-
tencies, input redundancy, and errors commonly occurring during
the model translation, simulation, and evaluation process, thereby
reducing design cycle latency and increasing accuracy. A genetic
algorithm (GA) based multi-objective optimization acts as the driv-
ing engine of the automation system to systematically explore and
search for the best-fit solutions among the competing objectives. In
addition, it enables the inclusion of other performance objectives as
part of the tradeoff analysis, which has been identified through the
literature review as necessary to support design decision-making.
(3) Design decision support is the final component of the frame-
work which provides an informative context to support designers’
decision making through user friendly visualization of geometric
configurations in parallel with the performance data and sensitiv-
ity to inform energy and competing performance and their impacts
relative to design decision choices. In addition, an interactive inter-
face to provide design guidance, sorting, and filtering functionality
based on designer’s design goal further facilitates designers’ deci-
sion making process.
As illustrated in Fig. 1, the framework provides two interactions
between designers and the EEPFD. Prior to the design problem
being explored through the EEPFD methodology designers are
required to define their design problem mathematically, associa-
tively, and logically, i.e., through parameterization. This provides
the basis for the system to automatically generate, analyze, and
evaluate design alternatives until the automated process is termi-
nated. The second interaction with EEPFD occurs at the end of the
automation loop, where a solution space containing all the explored
design alternatives is provided visually and analytically, through
each alternative’s 3D model, performance evaluations, and multi-
objective ranked results necessary for design decision support. The
three technologies utilized are Revit, Excel, and GBS. Revit serves
as the parametric design platform, while Excel serves as the plat-
form for calculation of the three design objective scores. Finally,
GBS is accessed through Revit’s conceptual energy analysis, and
used as the energy simulation engine, capable of generating all the
necessary energy analysis results. EEPFD is tested through using
the prototype tool, H.D.S. Beagle [13], a plug-in built on top of the
Autodesk® Revit® [94] (Revit) API. The Beagle integrates Autodesk®
Green Building Studio® [95] and (GBS) and Microsoft® Excel® [96]
with Revit as the design interface and parametric design engine.
The three objective functions for the experimental framework of
EEPFD include the whole building energy use intensity (EUI), net
present value (NPV), and a spatial programming compliance score
(SPC) to represent the energy, financial and design compliance per-
formance respectively. This selection first echoes the overall goal
of the research to understand the effectiveness of MDO to provide
energy performance as feedback to influence designers’ design
decision making, and secondly addresses the identified gap in cur-
rent efforts to consider multiple objectives simultaneously, more
relevant to realistic design practices. The usability issues found in
current MDO applications are addressed through integrating plat-
forms, the problem formulation method, and the encoding method
430 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
Fig. 1. The MDO framework, evolution energy performance feedback for design.
for optimization within the prototyped tool. The design decision
support is addressed through the provision of semi-automated 2D,
3D data plots combined with the automated 3D geometric model
and image output. The detailed algorithm, problem formulation
method and processes, and previous experiments are described in
detail in prior publications of the authors [13,69].
5. Experiments and research methodology
In order to measure effectiveness of EEPFD a series of experi-
ments were administered in pedagogical and workshop settings.
Due to the inherent unique nature of the design process a huge
variance will be observed between designers following the same
design protocol with the same design question. How designers’
cognition and response to feedback is highly dependent on an
individual’s background, experience, their psychological state, and
personal preference. As a result, prior research focusing on measur-
ing the design process typically are confined to small sample sizes
in order to dissect in detail each activity during the process to fully
understand all the variable and interactions of the design frame-
work/protocol or the feedback collected during the process. This
unique difficulty in measuring design in process is observed in vari-
ous design fields, i.e. industrial design [97], engineering design [98],
software design [99]. Architectural design present further chal-
lenges as the design goal for each individual can vary widely even
when following the same protocol [100,101]. In order to understand
the influence and impact of the provided feedback from EEPFD, ini-
tial experiments with reduced sample sizes are conducted. These
experiments are conducted through a series of set design activities
where responses and the process of each participant are recorded
in detail [13,69,91]. A detailed documentation of their activities
and unedited responses can be found in [90]. Here the research
presents the initial results and findings from two experiments run
with designers acting as the subjects.
In both cases the objective of the experiments was to observe the
impact and effectiveness of EEPFD on energy performance feedback
in the early stage design exploration process. For both experi-
ments the participants were considered normalized as none were
significantly professionally experienced or experts in the domain.
Rather, all participants for each experiment were in fact students,
the majority of which were undergraduate and/or early within
their graduate degree programs. Both experiments and the chosen
subjects are designed to act as the initial test bed to understand
whether the design of the experiment is conducive to observations
and measurements of the influence of EEPFD and the incorporation
of energy performance feedback. It is the intention of the research
to repeat the experiment on a larger scale and sample size as well
as collect more data in a professional setting, discussed further in
Section 6.
5.1. Experiment I: energy performance feedback using EEPFD
Experiment I is designed as a computational design workshop
that gathers data to understand the impact of the EEPFD process
in comparison to a manual exploration process. The primary goal
of this experiment is to understand whether the provided energy
performance feedback via EEPFD influences designers’ design deci-
sion making [69]. The experiment is confined to fit within a one-day
workshop. During Experiment I a total of seven participants were
observed and recorded. Their basic data is summarized in Table 1
and in summation can be understood as novice designers qualita-
tively similar in terms of experience and knowledge.
For quantitative and qualitative comparative studies, two design
processes are utilized: (1) a benchmarking manual exploration
process and (2) a design exploration process using EEPFD pre-
computed models and evaluation data and visualization. These
processes are illustrated in Fig. 2. For all activities during the
benchmarking manual exploration process the same technolo-
gies within EEPFD are used so participant exploration results can
be compared to those obtained through EEPFD. The experiment
involves three major components: (1) curriculum design; (2) lec-
ture, experiment data collection; and (3) data analysis following
the two experimental procedures. For both procedures, the first
major activity introduces the participants to the provided paramet-
ric conceptual mass compatible with EEPFD. The participants learn
the parameterization and problem formulation method by follow-
ing provided step-by-step instructions. The next section separates
into two procedures: (1) the manual exploration process and (2)
the EEPFD pre-computed models and evaluation data during the
same exploration time period. These two processes were conducted
consecutively by the same group of participants as illustrated by
Fig. 3. In order for the participants’ manually generated design
alternatives and the performance results to be comparable with
the results generated through EEPFD, the Excel objective function
calculator is provided using the same SPC and NPV formulae.
For the manual exploration process, see Fig. 3, participants are
first given a lecture on how to generate design alternatives, cal-
culate the three objective functions, Pareto rank their results, and
record their explorations using the given Excel calculator. Next par-
ticipants are given 3 h of exploration time to explore their design.
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441 431

Table 1
Educational background summary of the Experiment I participants.

ID Educational Prior Revit® Other prior BIM or Prior experience Other tool Prior Prior Other sim. Other sim.
background experience parametric tool length (months) exposure energy sim. energy sim. experience tools used
(months) experience (Y/N) experience tool used

ST01 MArch, BS in 12 N N/A N/A N N/A Y SAP, PKPM,

Civil 3D3S
ST02 BS in LA 0 Y 6 Grasshopper N N/A N N/A
ST03 BS in UP 0 N N/A N/A Y Ecotect N N/A
ST04 MArch, BArch 0 Y 6 Grasshopper N N/A N N/A
ST05 MBA, BS in 0 N N/A N/A N N/A N N/A
Economics
ST06 BArch 0 N N/A N/A Y HEED Y HEED
ST07 MArch, BArch 0 Y 12 Grasshopper Y Ecotect Y Autodesk
Algor.

Note: BS in LA: Bachelor of Science in Landscape Architecture; BS in UP: Bachelor of Science in Urban Planning; BArch: Bachelor of Architecture; BSArch: Bachelor of Science
in Architecture; MS in LA: Master of Science in Landscape Architecture; MArch: Master of Architecture.

At the end of the manual exploration process, a questionnaire is
given to collect feedback regarding their explorations. After the
manual EEPFD exploration process, an EEPFD pre-computed data
set is given to participants in order to measure how the feedback
from EEPFD is able to assist participants in their design decision-
making. A lecture and detailed handout regarding how to use the
data set are also given to the participants prior to implementing
their design decision process. The given dataset, as summarized in
Fig. 4, includes an Excel file containing 730 pre-computed design
alternatives with their objective scores and overall rank informa-
tion, four tradeoff study images, and the 3D images of all Rank 1
design alternatives.
This decision-making process consists of three steps the partic-
ipants are asked to follow. First, participants are asked to record

Fig. 2. Simulation process comparison between EEPFD’s experimental process and manual benchmark process. The observation focus of Experiment I is highlighted.
432 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441

Fig. 3. Illustration of Experiment I’s procedure, design problem and the explored parametric model.

their design decision-making strategies based on the given data
set. Second, participants are given a chance to explore the given
dataset and select a design alternative. Next, participants are
given the opportunity to improve their selected design based
on the given dataset. During the process of utilizing the EEPFD
pre-generated dataset, a questionnaire is given to the partici-
pants asking them to record their final selection and specific
feedback regarding their decision-making process enabled by
EEPFD. The collected data from participants is summarized in
Table 2.
5.2. Experiment I: results and analysis
Participants provided responses by completing the provided
questionnaire. The first two sections of the questionnaire focus
on whether the participants were able to select a better-fit final
design at the end of their exploration process through the EEPFD
pre-generated data. Table 3 summarizes the comparison of design
selections from each participant’s manual exploration process and
EEPFD pre-generated data. When ranking their selected design
based on the manual process compared to that of using the EEPFD
pre-generated data, the majority of the participants were able to
identify a Pareto solution in both approaches. Exceptions include
student (ST) ST03’s manual exploration and ST06’s EEPFD explo-
ration process. ST03 was able to identify a higher performing design
through EEPFD, which improved ST03’s original design from Rank
3 to Rank 1. Overall, six out of the seven participants were able
to use the EEPFD pre-generated data to identify Pareto solutions
through the EEPFD process, with the exception of ST06’s selection,
which was designated as Rank 2. The subsequent analysis of their
design decision-making strategy revealed that their documented
selection criteria were not based on the use of the three objective
functions as the primary goal, but rather on the rotation factors, as
the participant desired to emphasize their esthetic objective. This
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441 433

Fig. 4. Summary of provided EEPFD pre-generated data set for Experiment I.

434 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
Table 2
Summary of the recorded data for the Experiment I.
Recorded data
Background 1. Education background
questionnaire
2. Experience using Revit
3. Experience using other parametric
software
4. Experience using energy simulation
programs
5. Experience using other analytic tools
Manual 1. Explored parametric values for each
exploration iteration
questionnaire
2. The three objective function scores
of explored design alternatives
3. Time required to obtain and
calculate each objective function
4. Time required to obtain the final
design
5. Design exploration process
6. Final thoughts and suggestions
regarding the manual exploration
process.
Decision-making 1. Decision-making strategies prior to
through EEPFD the availability of the provided dataset
questionnaires
2. Decision-making strategies after
reviewing the pre-generated dataset
3. Strategies change factors
4. Final selection’s offspring ID and
objective scores
5. Parametric values and objective
scores of the improved design after the
selection
6. Design improvement strategy
7. The considered relevance of the
provided data during the improving
design process
8. New exploration ranges for each
parameter
Table 3
Performance comparison of the selected designs from the manual exploration process and subsequent design selected from the EEPFD pre-generated data. Improvements in
the objective performances through EEPFD pre-generated data are highlighted.
shift in focus resulted in their design selection not being ranked as
a Pareto solution.
Data type If the design alternatives for both approaches are compared, no
Enumeration clear dominance of the design alternatives provided by EEPFD is
revealed. However, when put in the context of each participant’s
Number design decision-making strategy, it is evident that EEPFD was able
Y/N, number, to provide improved performance in alignment with each partic-
enumeration
ipant’s outlined priorities, with the exception of ST07’s results.
Y/N, number,
enumeration ST07’s final selection based on the EEPFD pre-generated data does
Y/N, time, not provide a clear performance improvement consistent with their
enumeration designated design goals when compared with their manual explo-
Number
ration process. However, this exception resulted from violating the
provided site constraints in pursuit of maximizing the SPC score
Number during the manual exploration process. Therefore the results from
ST07’s manual exploration process are considered as an outlier by
Number this research and thereby incomparable to the EEPFD pre-generated
data set. These findings demonstrate an initial ability of novice par-
Number
ticipants to find better-fit solutions based on each participant’s
Image individual goals while maintaining design alternative within all
Description designated constraints through EEPFD. In comparison this direct
relationship between the outlined design decision strategy and the
Description
resulting design alternative’s performance is not readily observable
in the participants’ manual exploration process (Fig. 5).
The third section of the questionnaire asked participants to
Description improve their design by utilizing the EEPFD pre-generated dataset.
The responses provided in this section are summarized in Fig. 6.
Description
Number The ranking information provided is based on a solution pool con-
sisting of all the designs generated by all three approaches. In this
Number case, no new Pareto solutions were generated during the improve-
ment process, but all the solutions received Rank 1. The observed
Description
exception is ST06, since their first declared objective was to main-
Y/N, tain the rotation parameters for reasons of esthetic preference.
description Although the new generated solutions do not clearly dominate their
prior design selection, the data further supports the conclusion that
Number
when the participants utilized the EEPFD pre-generated dataset,
they were able to improve their design performance over their pre-
vious design. Conversely, this is not observed during their manual
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441 435

Fig. 5. Participants’ selected design via the three exploration approaches: (1) selection from their manual exploration process; (2) selection from EEPFD pre-generated
solution space; and (3) improved.

exploration process based on the fact that their final selections goal. For example, for participants who designated a reduction in
are not usually the last explored. Another observation from the EUI as a design goal, were not always successful through their
manual exploration process is that the participants are not always manual exploration process. However, the data illustrates when
able to improve their designs toward their overall indicated design given an EEPFD pre-generated data set, participants were able to

Fig. 6. Illustrated diversity among participants’ parameterization/problem formulation for the same design problem. Summary of participants’ parametric designs with the
comparison of design parameter problem scale, coupling, and geometric complexity.
436 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
improve the EUI in accordance with their stated design goal. In addi-
tion, using the EEPFD pre-generated data enabled participants to
improve their design toward their intended goal while considering
the EUI, SPC, and NPV scores simultaneously. This is in contrast to
their observed manual exploration process where typically a single
objective was concentrated on at one time. In addition, participants
were able to further improve their stated design intent based on
using the provided data set, which enabled them to identify the
relative impact of each parameter on each objective performance
score.
5.3. Experiment II: a best practice for early stage design
application of MDO
Aside from addressing the applicability issues of current MDO
applications for designer use, the previously identified critical issue
preventing MDO from being adopted for early stage design is
the long analysis time necessary for identifying solutions through
mathematical convergence of competing objectives. This is a com-
mon goal of prior MDO applications to design but the unique
demands of early stage architectural design imply a divergence
from this approach may be pertinent. Based on the prior obser-
vations of the authors [13], it is revealed that during the GA
optimization run the optimal performance boundaries of each
objective score can typically be obtained after only a few gen-
erations. Most significantly, this implies that the performance
potential for each design scenario is identifiable prior to reach-
ing an optimal solution defined by mathematical convergence. If
this supposition is true, it could result in providing the desired
context synchronously in which to gauge any individual solutions’
performance by the designer. Moreover, it may potentially be more
relevant to supporting early design decision making than provid-
ing the often over populated Pareto solution pool. As a result,
this experiment focuses on observing the obtaining of the optimal
objective score performance boundaries for each design problem
to validate a conceptualized best practice for implementing MDO
in the early stage architectural design process.
This experiment is formatted as an assignment for a course
where participants are in the process of learning Revit. The assign-
ment is given after a 1 h parametric conceptual mass lecture which
provides step by step instructions to familiarize participants with
creating a parametric mass identical to the one used in the previous
experiment. For this experiment, participants are asked to provide
their individual parametric design by formulating their design
problem based on the provided design criteria, site constraints,
and consistent parameter count during their design exploration.
The intent is to observe if there is a measurable variance in the
design performance potential between the participants’ concep-
tual designs and to understand the required generation and time
necessary to obtain performance boundaries, i.e. the maximum
value of NPV and SPC and the minimum value of EUI. The two sig-
nificant measurements of the experiment are: (1) if a significant
quantifiable difference in the performance ranges of these varying
conceptual designs for an identical design problem is observed, this
can be considered as evidence of the value of exploring competing
design concepts during the early stages of design; and (2) if there is
a significant reduction in the generation number and time required
to obtain the performance boundaries in comparison to a mathe-
matical convergence, this reduction can be viewed as indicating
support for the use of performance boundaries to provide the con-
text synchronously in which to gauge competing design concepts
by designers. For these purposes, this research collected each par-
ticipant’s parametric design and engaged H.D.S. Beagle for up to 50
generations for each. The data of interest recorded by this research
are listed below:
5.3.1. Design problem measurements
Project complexity: Project complexity refers to the project size
as measured in square feet and the number of types of pro-
gram spaces, such as parking, commercial or residential, that are
included within the design problem. In this case, all the partici-
pants’ designs have the same design problem as documented in
Fig. 3.
Design complexity: Design complexity refers to the amount of sur-
faces (i.e. tessellation) required to be included in the energy model
along with the number of available parameters as provided by the
design problem.
5.3.2. Design process measurements
Speed: Speed measures the average run time required to obtain an
energy analysis for each parametric design. It also records the over-
all time required for a GA run up to 50 generations in a normalized
condition.
GA settings: In this experiment all the scenarios were run with the
same GA settings and therefore normalized:
Initial population: 10 Population size: 20
Crossover ratio: 0.6 Selection size: 20
Mutation ratio: 0 Maximum iteration: 50
5.3.3. Design product measurements
Feedback quantity measures: This records the total time required for
each design to run up to 50 generations. Based on the GA settings
of the experiment, this means the time required to generate and
analyze 1010 design alternatives. This value is also translated into
the number of design alternatives generated during an 8 h work
period.
Feedback quality: The feedback quality is defined by the boundary
condition of the EUI, NPV, and SPC performance after running up
to 50 generations and the Pareto solution numbers based on these
three objective rankings.
Performance boundaries: This records the generation number and
time for each scenario’s boundary condition to be found, i.e. max-
imum NPV and SPC and minimum EUI.
5.4. Experiment II: results and analysis
Fig. 6 provides a summary of the scenarios provided by the 16
participants for this experiment. Similar nomenclature and ST#
acronym is used though these participants are different than Exper-
iment I. It should be noted they are of similarly normalized and
comparable in terms of experience and backgrounds as those in
experiment I. Fig. 7 summarizes the resulting ranges in perfor-
mance boundaries for each scenario’s solution space along with
the GA optimization generation number where these boundaries
were identified. These results show a significant difference in the
potential performance of varying conceptual designs generated for
the same design problem. While an improvement from the initial
design’s performance was achieved in each case by using EEPFD,
a limit to the potential for further improvement was noted. For
example, the highest performing design for the participant ST01 is
only capable of reaching an NPV score of 275 million USD while
the highest performing NPV for ST03 is 461 million USD. Therefore,
it can be extrapolated that a medium performing design solution
for ST03 could be on par with a high performing design solution
for ST01. These findings further support the potential relevance of
using EEPFD during the early stages of design, since EEPFD provides
the ability to rapidly explore a wide variety of geometrically diverse
conceptual designs and generates the context in which the low or
high performance potential of design concepts can be identified and
pursued in accordance with the design problem objectives. Table 4
shows that for the majority of the tested scenarios the performance
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441 437

Fig. 7. Illustration of resulting ranges in performance boundaries for each scenario’s solution space along with the generation and required runtime where these boundaries
were identified.

boundary could be obtained before the 17th generation. The excep-
tions to this include the EUI for ST08 and ST09 and the SPC for
ST07 and ST13. However for these exceptions it was observed that,
while continuing improvements were made in later generations,
these gains could be viewed as potentially unsubstantial in terms
of overall scores. For example, in the case of ST08 the EUI’s per-
formance boundary is technically reached at the 49th generation
with EUI = 41.62 kBtu/ft2 /year. However, the solution space with
EUI = 41.64 kBtu/ft2 /year was found at the 14th generation. While
there was continued improvement, the 0.02 kBtu/ft2 /year reduc-
tion over 35 generations resulted in less than a 0.05% impact on the
overall EUI performance boundary and therefore can be viewed as
insignificant. Similar circumstances were observed through anal-
ysis of the other exception cases. As a result, for this experiment
any changes in the performance boundaries that had less than
a 0.1% impact on the overall performance range are defined as
insignificant. Using the 0.1% performance tolerance, all boundary
conditions were obtained within 17 generations with an average
requirement of 7 generations. Further research and analysis are
needed to understand any potential correlation between the quan-
tity of parameters explored and the required generations to identify
performance boundaries.
Based on these research results, the limits on the performance
potential of a solution space can be identified with an average of
438 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
Table 4
Original and adjusted generation numbers required to define the objective performance boundaries per participant. Highlighted in pink are required generation numbers
exceeding 17. Highlighted in yellow are generation numbers altered per the introduced 0.1% impact tolerance.
7 generations, as opposed to reaching the mathematically conver-
gent criteria, which typically requires run times of more than 50
generations. Consequently, the performance boundaries of a solu-
tion space are much more rapidly available and useful to provide
the context synchronously in which to gauge any individual design
concept than otherwise possible. In addition, these boundaries
can be considered more relevant than the potential of any indi-
vidual design solution to the early stage design process, when
various conceptual designs are being explored in parallel. Even
under the strict conditions of this experiment, which limited con-
ceptual designs to a single volume with a fixed number of allowable
design parameters, a wide range in the conceptual design perfor-
mance boundaries was observed. Therefore, there is the potential
that, with a greater diversity between explored conceptual designs,
there would be an even more dramatic range in the performance
potential between competing conceptual designs. This potential in
performance would be a new variable that EEPFD would be able to
add to the design decision-making process, which has previously
not been offered through the precedent tools and methods.
6. Conclusion and discussion
With the overall aim to enable energy performance feedback to
support and influence designers’ design decision-making during
the early stage of the design process, an early stage designer-
centered MDO framework, EEPFD, is presented. This paper first
highlights the theoretical foundation of EEPFD through the lit-
erature review and gap analysis between current design and
energy simulation domains and the applications of optimiza-
tion. By categorizing and mapping the obstacles and solutions
the three major components of EEPFD are isolated: problem for-
mulation, automation loop and design decision support. Through
the three components, EEPFD is able to support designers with
a designer-centric user interface, bypassing the interoperability
issues, providing rapid design solution spaces with performance
ranking, evaluation, sensitivity analysis and tradeoff study in par-
allel with geometric and data visualization. This research then
focuses on addressing two specific gaps found in precedents
through two experiments using EEPFD, and the prototyped tool,
H.D.S. Beagle. The first experiment observes whether the provided
feedback from EEPFD supports and influences designers’ decision
making; while the second experiment explores the proposed best
practice of utilizing the performance boundaries of an EEPFD-
generated solution pool to gauge the potential performance of
different design concepts during the early stage of the architectural
design process.
To address the first identified gap, the first experiment gauges
whether or not the energy feedback generated by EEPFD supports
or influence designers’ decision-making process through an analy-
sis of responses collected from the participants regarding their use
of the provided EEPFD pre-generated dataset in their design explo-
ration. Also included in the analysis is a comparison of their design
decisions based on their manual design exploration with the use
of the EEPFD pre-generated dataset. All participants provided pos-
itive responses regarding the ability of the EEPFD pre-generated
data set to support their design decision making. In addition, the
influence of the EEPFD provided energy performance feedback on
their design decisions is also demonstrated in the recorded data,
which indicates that participants were able to improve the energy
performance of their final designs based on their design goals. Fur-
ther improvement was also observed when participants used the
given dataset to improve the energy performance of their design.
This is in contrast to the manual exploration process, which demon-
strated that participants were able to improve their SPC and NPV
consistently but not their EUI. These findings confirm that the
use of EEPFD is able to provide better energy performance feed-
back and improve guidance when improving energy performance
is an intended design goal. While the limited sample size of this
experiment is insufficient for identifying statistical significance, the
results demonstrate descriptively that, when participants utilize
EEPFD, they are able to include the energy performance feedback
in their considerations and improve the energy performance dur-
ing their design exploration process. Perhaps most importantly,
this experiment demonstrates that participants who do not possess
expert energy domain knowledge are able to observe the impact
on energy performance during their exploration process through
the ranking, sensitivity analysis and visualizations of the expanded
solution provided by EEPFD. These observations and responses
provide the basis of initial confirmation for the value and usability
of EEPFD, when employed by designers during the early stages of
the design process.
Irrespective of the limitations of the sample size of the current
experiment, it is qualitatively and descriptively evident during the
experiment that participants were not motivated to participate in
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
the tedious manual design exploration process as a part of their
overall design approach. This observation reinforces the need for
an automated and integrated design exploration approach to facili-
tate inclusion of energy performance as feedback during designers’
early stage design explorations. While the participants indicated
that the EEPFD pre-generated data was useful in assisting them in
deciding on a better performing design, difficulties were observed
regarding the use of the provided interface. Currently, the EEPFD
generated data is manually organized by the authors, where the
users use built-in functionalities of Excel to browse, filter, and orga-
nize the data as desired. It is noted that the participants needed
more instructions regarding use of this platform and the manner in
which they could use the data to assist their decision making. These
are known issues of the experimental framework that are expected
to be addressed through future improvements to the user interface
and interactive guidance. However, even with these known issues
the participants were able to use the data to improve the energy
performance of their designs.
The second experiment described in this paper explores the pro-
posed best practice of using performance boundaries to provide
the context synchronously for comparing competing design con-
cepts during the early stages of design rapidly and visually pairing
geometry with analytical data. Through this second experiment
two significant observations were made confirming the potential
of the proposed best practice. First, when provided identical design
requirements and energy parameter settings, but with significantly
varying conceptual designs, a wide range of resulting performance
boundaries are observed. This implies a direct relationship between
the initial conceptual design and parametric variations and the
resulting performance boundaries of generated design iterations.
Therefore, the ability of EEPFD to rapidly aid in determining the per-
formance potential of multiple competing conceptual designs for
the same design requirements can be considered more applicable
to conceptual architectural design, which demands diversity. This
is an important finding when compared to the typical MDO applica-
tion in other fields more commonly demanding a single optimized
solution space or more formulized trade off.
The second significant observation made during this experiment
was that the optimal performance boundaries for a design could
be obtained after an average of only seven generations of GA runs
while the more commonly sought after mathematically convergent
solution set required more than 50 generations. Considering that
time constraints are often a dominant factor during the early stage
design process, this experiment confirms the potential of the pro-
posed best practice, that these performance boundaries can provide
the context in which to gauge competing conceptual designs and
that these boundaries can be reached significantly more quickly
than the mathematically convergent solution set. It is only after
the conceptual design phase has been completed, and the design
form roughly finalized, that a mathematically convergent solution
might be applicable to later design phases. While the current exper-
iment initially supports this proposition, further experiments and
observations of this proposed best practice concept are planned.
Another issue directly dependent on the unique composition
of every design problem is the optimization settings utilized dur-
ing a GA run. While a best practice would imply that a particular
collection of settings should yield the most efficient means of pro-
ducing results, these have yet to be identified. However, despite the
absence of a universally applicable GA setting list, throughout the
experimental process the GA was able to consistently deliver solu-
tion pools with improving performances. Whether a universally
applicable optimized collection of GA settings exists or each design
problem requires individualized settings remains the subject of
future research. In the case of optimal GA settings needing to be
individually identified for each design problem it should be noted
that due to the infinite diversity and infinite combinations available
439
at the conceptual stage of design the pursuit of optimizing the
GA settings becomes both impractical and irrelevant, highlighting
the importance of the boundary condition for gauging conceptual
architectural designs. Finally, it is acknowledged that MDO applica-
tions are not restricted to including energy performance feedback.
However, among the design objectives, only energy performance
required the use of EEPFD to influence design decision-making,
as demonstrated in Experiment I. Therefore, although EEPFD is
potentially capable of including other performance criteria energy
savings of the first and second costs for example, the major contri-
bution of the framework is in its ability to demonstrate that MDO
can successfully enable energy performance feedback to influence
design decision-making during the early stages of the design pro-
cess. Our future work acknowledges that the sample size to date
is insufficient to statistically validate the results. However, this
research is qualitatively and descriptively a contribution in that it
provides a research tool and method to measure the design process
for the wide variance between individuals within an MDO tuned
to improve energy outcomes. These experiments demonstrate an
ability to quantify and evaluate highly individualized designer pro-
cesses and to explore further the value of performance boundaries
as a likely impactful design decision making innovation in the con-
text of cloud based computing and large solution space design
exploration. The results thus far, serve as a critical benchmark to
compare the results of the effect of these performance boundaries
from larger sampling of novice and professional designers with
varying levels of technical expertise as future work.
Acknowledgement
The authors thank Autodesk Inc. for the generous financial sup-
port of the research within the IDEA Studio program.
References
[1] Architecture 2030, Energy – buildings consume more energy than any other
sector, 2011.
[2] The European Union, Directive 2012/27/EU of the European Parliament and
of the Council of 25 October 2012, Official Journal of the European Union 55
(2012).
[3] I. Obanye, Integrating building energy simulation into the architectural design
process, in: BEECON 2006, The Centre for Education in the Built Environment
(CEBE), London, 2006.
[4] A. Mahdavi, S. Feurer, A. Redlein, G. Suter, An inquiry into the building perfor-
mance simulation tools usage by architects in Austria, in: Building Simulation
2003, Eindhoven, Netherlands, 2003, pp. 777–784.
[5] C.A. Morbitzer, Towards the Integration of Simulation into the Building Design
Process, Department of Mechanical Engineering, University of Strathclyde,
2003.
[6] D. Robinson, Energy model usage in building design: a qualitative assessment,
Building Services Engineering Research & Technology 17 (2) (1996) 89–95.
[7] J. Hensen, Energy related design decisions deserve simulation approach, in:
1994 DDSS, Vaals, The Netherlands, 1994.
[8] S. Bambardekar, U. Poerschke, The architect as performer of energy simulation
the early design stage, in: Building Simulation 2009, IBPSA, Glasgow, Scotland,
2009, pp. 1306–1313.
[9] U. Bogenstätter, Prediction and optimization of life-cycle costs in early design,
Building Research & Information 28 (5–6) (2000) 376–386.
[10] D.B. Crawley, J.W. Hand, M. Kummert, B.T. Griffith, Contrasting the capa-
bilities of building energy performance simulation programs, Building and
Environment 43 (4) (2008) 661–673.
[11] F. Flager, B. Welle, P. Bansal, G. Soremekun, J. Haymaker, Multidisciplinary
process integration and design optimization of a classroom building, ITcon
14 (38) (2009) 595–612.
[12] A. Schlueter, F. Thesseling, Building information model based energy/exergy
performance assessment in early design stages, Automation in Construction
18 (2) (2009) 153–163.
[13] S.-H.E. Lin, D.J. Gerber, Designing-in performance: a framework for evolu-
tionary energy performance feedback in early stage design, Automation in
Construction 38 (2014) 59–73.
[14] D.J. Gerber, Parametric Practices: Models for Design Exploration in Architec-
ture (dissertation), Architecture, Harvard Graduate School of Design, 2007.
[15] P. de Wilde, G. Augenbroe, M. van der Voorden, A strategy to provide compu-
tational support for the selection of energy saving building components, in:
Building Simulation 2001, Rio de Janeiro, Brazil, 2001, pp. 653–660.
440 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
[16] K.P. Lam, Y.C. Huang, C. Zhai, Energy Modeling Tools Assessment for Early
Design Phase, CMU Center for Building Performance and Diagnostics, Pitts-
burgh, PA, 2004.
[17] C. Morbitzer, P. Strachan, J. Webster, B. Spires, D. Cafferty, Integration of build-
ing simulation into the design process of an architectural practice, in: Building
Simulation 2001, IBPSA, Rio de Janeiro, Brazil, 2001.
[18] A. Pedrini, S. Szokolay, The architects approach to the project of energy effi-
cient office buildings in warm climate and the importance of design methods,
in: Building Simulation 2005, IBPSA, Montréal, Canada, 2005, pp. 937–944.
[19] M. Trebilcock, B. Ford, R. Wilson, Integration of sustainability in the design
process of contemporary architectural practice, in: PLEA 2006, Geneva,
Switzerland, 2006.
[20] B. Welle, J. Haymaker, Z. Rogers, ThermalOpt: A methodology for automated
BIM-based multidisciplinary thermal simulation for use in optimization envi-
ronments, Building Simulation 4 (4) (2011) 293–313.
[21] S. Attia, E. Gratia, A. De Herde, J.L.M. Hensen, Simulation-based decision sup-
port tool for early stages of zero-energy building design, Energy and Buildings
49 (2012) 2–15.
[22] T. Hong, S.K. Chou, T.Y. Bong, Building simulation: an overview of devel-
opments and information sources, Building and Environment 35 (4) (2000)
347–361.
[23] G. Augenbroe, Trends in building simulation, Building and Environment 37
(8–9) (2002) 891–902.
[24] M. Pilgrim, N. Bouchlaghem, D. Loveday, M. Holmes, Towards the efficient
use of simulation in building performance analysis: a user survey, Building
Services Engineering Research & Technology 24 (3) (2003) 149–162.
[25] B.J. Urban, The MIT Design Advisor: Simple and Rapid Energy Simulation of
Early-Stage Building Designs, Department of Mechanical Engineering, Mas-
sachusetts Institute of Technology, 2007.
[26] B. Toth, F. Salim, J. Burry, J.H. Frazer, R. Drogemuller, M. Burry, Energy-oriented
design tools for collaboration in the cloud, International Journal of Architec-
tural Computing 9 (4) (2011) 339–360.
[27] S. Attia, L. Beltrán, A. De Herde, J. Hensen, “Architect friendly”: a comparison of
ten different building performance simulation tools, in: Building Simulation
2009, Glasgow, Scotland, 2009, pp. 204–211.
[28] L. Weytjens, S. Attia, G. Verbeeck, A. De Herde, The ‘architect-friendliness’ of
six building performance simulation tools: a comparative study, International
Journal of Sustainable Building Technology and Urban Development 2 (3)
(2011) 237–244.
[29] G. Riether, T. Butler, Simulation space: a new design environment for
architects, in: 26th eCAADe Conference: Computation: The New Realm of
Architectural Design, Antwerpen, Belgium, 2008, pp. 133–142.
[30] S. Attia, A Tool for Design Decision Making: Zero Energy Residential Build-
ings in Hot Humid Climates, Architecture et climat, Université catholique de
Louvain, 2012.
[31] E.J. van Dijk, P.G. Luscuere, An architect friendly interface for a dynamic build-
ing simulation program, in: Sustainable Building 2002, Oslo, Norway, 2002.
[32] K.P. Lam, N.H. Wong, F. Henry, A study of the use of performance-based simu-
lation tools for building design and evaluation in Singapore, in: N. Nakahara,
J.L.M. Hensen (Eds.), Building Simulation 1999, Kyoto, Japan, 1999.
[33] E. Gratia, A. De Herde, A simple design tool for the thermal study of an office
building, Energy and Buildings 34 (3) (2002) 279–289.
[34] L. Weytjens, G. Verbeeck, Towards ‘architect-friendly’ energy evaluation tools,
in: A. Khan (Ed.), SimAUD 2010, Society for Computer Simulation Interna-
tional, Orlando, FL, USA, 2010, pp. 179:171–179:178.
[35] S. Attia, J.L.M. Hensen, L. Beltrán, A. De Herde, Selection criteria for building
performance simulation tools: contrasting architects’ and engineers’ needs,
Journal of Building Performance Simulation 5 (3) (2012) 155–169.
[36] M.M. Mourshed, D. Kelliher, M. Keane, ArDOT: A tool to optimise environmen-
tal design of buildings, in: Building Simulation 2003, Eindhoven, Netherlands,
2003, pp. 919–926.
[37] R. Aish, A. Marsh, An integrated approach to algorithmic design and environ-
mental analysis, in: R. Attar (Ed.), SimAUD 2011, Boston, MA, USA, 2011, pp.
61–67.
[38] J.L.M. Hensen, R. Lamberts, Introduction to building performance simulation,
in: J.L.M. Hensen, R. Lamberts (Eds.), Building Performance Simulation for
Design and Operation, Spon Press, New York, NY, USA, 2011, pp. 1–14.
[39] C. Eastman, J.-M. Lee, Y.-S. Jeong, J.-K. Lee, Automatic rule-based checking of
building designs, Automation in Construction 18 (8) (2009) 1011–1033.
[40] G. Augenbroe, P. de Wilde, H.J. Moon, A. Malkawi, An interoperability work-
bench for design analysis integration, Energy and Buildings 36 (8) (2004)
737–748.
[41] J. Plume, J. Mitchell, Collaborative design using a shared IFC building model-
learning from experience, Automation in Construction 16 (1) (2007) 28–36.
[42] J. Pohl, L. Myers, A.J. Chapman, ICADS: an intelligent computer-aided design
environment, ASHRAE Transactions 96 (2) (1990) 473–480.
[43] G. Augenbroe, Integrated building performance evaluation in the early design
stages, Building and Environment 27 (2) (1992) 149–161.
[44] K. Papamichael, J. LaPorta, H. Chauvet, Building design advisor: automated
integration of multiple simulation tools, Automation in Construction 6 (4)
(1997) 341–352.
[45] Y.E. Kalay, P3: computational environment to support design collaboration,
Automation in Construction 8 (1) (1998) 37–48.
[46] A. Mahdavi, A comprehensive computational environment for performance
based reasoning in building design and evaluation, Automation in Construc-
tion 8 (4) (1999) 427–435.
[47] P.J.C.J. de Wilde, Computational Support for the Selection of Energy Saving
Building Components (Ph.D. dissertation), Architecture, Delft University of
Technology, 2004.
[48] K.P. Lam, N.H. Wong, A. Mahdavi, K.K. Chan, Z. Kang, S. Gupta, SEMPER-II:
an internet-based multi-domain building performance simulation environ-
ment for early design support, Automation in Construction 13 (5) (2004)
651–663.
[49] D. Holzer, Optioneering in collaborative design practice, International Journal
of Architectural Computing 8 (2) (2010) 165–182.
[50] D.T.J. O’Sullivan, M.M. Keane, D. Kelliher, R.J. Hitchcock, Improving building
operation by tracking performance metrics throughout the building lifecycle
(BLC), Energy and Buildings 36 (11) (2004) 1075–1090.
[51] S. Attia, M. Hamdy, W. O’Brien, S. Carlucci, Assessing gaps and needs for inte-
grating building performance optimization tools in net zero energy buildings
design, Energy and Buildings 60 (2013) 110–124.
[52] C.M. Clevenger, J. Haymaker, Metrics to assess design guidance, Design Stud-
ies 32 (5) (2011) 431–456.
[53] A. Yezioro, D. Bing, F. Leite, An applied artificial intelligence approach towards
assessing building performance simulation tools, Energy and Buildings 40 (4)
(2008) 612–620.
[54] A.D. Radford, J.S. Gero, Tradeoff diagrams for the integrated design of the
physical environment in buildings, Building and Environment 15 (1) (1980)
3–15.
[55] D.E. Grierson, Pareto multi-criteria decision making, Advanced Engineering
Informatics 22 (3) (2008) 371–384.
[56] F. Flager, J. Haymaker, A comparison of multidisciplinary design, analysis and
optimization processes in the building construction and aerospace industries,
in: I. Smith (Ed.), 24th W78 Conference on Bringing ITC knowledge to work,
Maribor, Slovenia, 2007, pp. 625–630.
[57] M.S. Al-Homoud, Computer-aided building energy analysis techniques, Build-
ing and Environment 36 (4) (2001) 421–433.
[58] G. Augenbroe, J. Hensen, Simulation for better building design, Building and
Environment 39 (8) (2004) 875–877.
[59] D. Kolokotsa, G.S. Stavrakakis, K. Kalaitzakis, D. Agoris, Genetic algorithms
optimized fuzzy controller for the indoor environmental management in
buildings implemented using PLC and local operating networks, Engineering
Applications of Artificial Intelligence 15 (5) (2002) 417–428.
[60] D. Kolokotsa, C. Diakaki, E. Grigoroudis, G. Stavrakakis, K. Kalaitzakis, Decision
support methodologies on the energy efficiency and energy management in
buildings, Advances in Building Energy Research 3 (1) (2009) 121–146.
[61] C. Diakaki, E. Grigoroudis, N. Kabelis, D. Kolokotsa, K. Kalaitzakis, G.
Stavrakakis, A multi-objective decision model for the improvement of energy
efficiency in buildings, Energy 35 (12) (2010) 5483–5496.
[62] L.G. Caldas, L.K. Norford, A design optimization tool based on a genetic algo-
rithm, Automation in Construction 11 (2) (2002) 173–184.
[63] L.G. Caldas, Generation of energy-efficient architecture solutions applying
GENE ARCH: an evolution-based generative design system, Advanced Engi-
neering Informatics 22 (1) (2008) 59–70.
[64] P.H.T. Janssen, An evolutionary system for design exploration, in: T. Tidafi, T.
Dorta (Eds.), Joining Languages, Cultures and Visions: CAAD Futures 2009, Les
Presses de l’Université de Montréal, Montréal, 2009, pp. 259–272.
[65] Y.K. Yi, A.M. Malkawi, Optimizing building form for energy performance based
on hierarchical geometry relation, Automation in Construction 18 (6) (2009)
825–833.
[66] Y.K. Yi, A.M. Malkawi, Site-specific optimal energy form generation based on
hierarchical geometry relation, Automation in Construction 26 (2012) 77–91.
[67] E. Asadi, M.G. da Silva, C.H. Antunes, L. Dias, Multi-objective optimization for
building retrofit strategies: a model and an application, Energy and Buildings
44 (2012) 81–87.
[68] D. Tuhus-Dubrow, M. Krarti, Genetic-algorithm based approach to optimize
building envelope design for residential buildings, Building and Environment
45 (7) (2010) 1574–1581.
[69] D.J. Gerber, S.-H.E. Lin, Designing in complexity: simulation, integration, and
multidisciplinary design optimization for architecture, Simulation (2013)
(published online before print April 9, 2013).
[70] C. Struck, J. Hensen, P. Kotek, On the application of uncertainty and sensitiv-
ity analysis with abstract building performance simulation tools, Journal of
Building Physics 33 (1) (2009) 5–27.
[71] C. Struck, P.J.C.J. de Wilde, C.J. Hopfe, J.L.M. Hensen, An investigation of the
option space in conceptual building design for advanced building simulation,
Advanced Engineering Informatics 23 (4) (2009) 386–395.
[72] C.J. Hopfe, Uncertainty sensitivity analysis in building performance simula-
tion for decision support and design optimization, Technische Universiteit
Eindhoven, Eindhoven, 2009.
[73] Y. Yildiz, K. Korkmaz, T.G. Özbalta, Z.D. Arsan, An approach for developing
sensitive design parameter guidelines to reduce the energy requirements of
low-rise apartment buildings, Applied Energy 93 (2012) 337–347.
[74] S. Burhenne, O. Tsvetkova, D. Jacob, G.P. Henze, A. Wagner, Uncertainty
quantification for combined building performance and cost–benefit analyses,
Building and Environment 62 (2013) 143–154.
[75] K. Mela, T. Tiainen, M. Heinisuo, Comparative study of multiple criteria deci-
sion making methods for building design, Advanced Engineering Informatics
26 (4) (2012) 716–726.
[76] R. Mikučionienė, V. Martinaitis, E. Keras, Evaluation of energy efficiency meas-
ures sustainability by decision tree method, Energy and Buildings 76 (2014)
64–71.
 S.-H. Lin, D.J. Gerber / Energy and Buildings 84 (2014) 426–441
[77] A. Sabapathy, S. Maithel, A multi-criteria decision analysis based assessment
of walling materials in India, Building and Environment 64 (2013) 107–117.
[78] C.J. Hopfe, G.A. Augenbroe, J.L.M. Hensen, Multi-criteria decision making
under uncertainty in building performance assessment, Building and Envi-
ronment 69 (2013) 81–90.
[79] J.-J. Wang, Y.-Y. Jing, C.-F. Zhang, J.-H. Zhao, Review on multi-criteria decision
analysis aid in sustainable energy decision-making, Renewable and Sustain-
able Energy Reviews 13 (9) (2009) 2263–2278.
[80] J.C. Damski, J.S. Gero, An evolutionary approach to generating constraint-
based space layout topologies, in: R. Junge (Ed.), CAAD Futures 97, Springer,
Netherlands, 1997, pp. 855–864.
[81] W. Marks, Multicriteria optimisation of shape of energy-saving buildings,
Building and Environment 32 (4) (1997) 331–339.
[82] K. Peippo, P.D. Lund, E. Vartiainen, Multivariate optimization of design trade-
offs for solar low energy buildings, Energy and Buildings 29 (2) (1999)
189–205.
[83] L.G. Caldas, An Evolution-based Generative Design System: Using Adaptation
to Shape Architectural Form (Ph.D. dissertation), Department of Architecture,
Massachusetts Institute of Technology, 2001.
[84] J.A. Wright, H.A. Loosemore, R. Farmani, Optimization of building thermal
design and control by multi-criterion genetic algorithm, Energy and Buildings
34 (9) (2002) 959–972.
[85] P. Geyer, Component-oriented decomposition for multidisciplinary design
optimization in building design, Advanced Engineering Informatics 23 (1)
(2009) 12–31.
[86] L. Magnier, F. Haghighat, Multiobjective optimization of building design using
TRNSYS simulations, genetic algorithm, and artificial neural network, Build-
ing and Environment 45 (3) (2010) 739–746.
[87] K. Suga, S. Kato, K. Hiyama, Structural analysis of Pareto-optimal solution
sets for multi-objective optimization: an application to outer window design
problems using multiple objective genetic algorithms, Building and Environ-
ment 45 (5) (2010) 1144–1152.
[88] I. Keough, D. Benjamin, Multi-objective optimization in architectural design,
in: A. Khan (Ed.), SimAUD 2010, Orlando, FL, USA, 2010, pp. 5–12.
441
[89] R. Evins, A review of computational optimisation methods applied to sustain-
able building design, Renewable and Sustainable Energy Reviews 22 (2013)
230–245.
[90] S.-H.E. Lin, Designing-in performance: energy simulation feedback for early
stage design decision making (dissertation), School of Architecture, Univer-
sity of Southern California, 2014.
[91] D.J. Gerber, S.-H.E. Lin, X. Ma, Designing in performance: a case study of apply-
ing evolutionary energy performance feedback for design, in: ACADIA 2013 –
Adaptive Architecture, Cambridge, Ontario, Canada, 2013, pp. 79–86.
[92] S.-H.E. Lin, D.J. Gerber, Designing-in performance: a case study of a net zero
energy school design, in: PLEA 2013: Sustainable Architecture for a Renew-
able Future, Munich, Germany, 2013.
[93] S.-H.E. Lin, D.J. Gerber, Evolutionary energy performance feedback for design
(EEPFD): interaction and automation for a design exploration process frame-
work, in: R. Stouffs, S. Sariyildiz (Eds.), 31st eCAADe Conference: Computation
and Performance, Delft University of Technology, The Netherlands, 2013, pp.
175–184.
[94] Autodesk, Autodesk Revit, 2013.
[95] Autodesk, Autodesk Green Building Studio: Web-Based Energy Analysis Soft-
ware, 2013.
[96] Microsoft, Excel – Analyze, 2013.
[97] D.G. Ullmana, T.G. Diettericha, L.A. Stauffera, A model of the mechanical
design process based on empirical data, Artificial Intelligence for Engineering,
Design, Analysis and Manufacturing 2 (1) (1998) 35–52.
[98] C.J. Atman, J.R. Chimka, K.M. Bursic, H.L. Nachtmann, A comparison of fresh-
man and senior engineering design processes, Design Studies 20 (2) (1999)
131–152.
[99] T.J. McCabe, C.W. Butler, Design complexity measurement and testing, in:
Communications of the ACM, ACM, New York, NY, USA, 1989.
[100] C.M. Eastman, Explorations of the Cognitive Processes in Design (report),
Department of Computer Science, Carnegie Mellon University, Pittsburgh, PA,
1968.
[101] Ö. Akin, C. Lin, Design protocol data and novel design decisions, Design Studies
16 (2) (1995) 211–236.