Automation in Construction 32 (2013) 125–135

Contents lists available at SciVerse ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Performance-driven architectural design and optimization technique

from a perspective of architects
Xing Shi a,⁎, Wenjie Yang b
a
Key Laboratory of Urban and Architectural Heritage Conservation, Ministry of Education, School of Architecture, Southeast University, #2 Si Pai Lou, Nanjing, 210096, China
b
Tianhua Architecture Design Company Limited, 27 Floor, Zhaofeng Global Mansion, 1800 Zhongshan West Road, Shanghai, 200235, China

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

Article history: Performance-driven architectural design emphasizes on integrated and comprehensive optimization of
Accepted 25 January 2013 various quantifiable performances of buildings. As the leading profession of a project team, architects play
Available online 28 February 2013 a vital role in guiding and conducting the performance-driven design. Methodology and techniques start
emerging both in literature and practice. However, architects often find them difficult to use for various rea-
Keywords:
sons. Therefore, developing an effective technique to conduct performance-driven design and optimization
Performance
Optimization
from the perspective of architects is necessary. This paper starts from discussing the concept of
Simulation performance-driven architectural design. Existing methodology and techniques are reviewed. The focus is
Rhinoceros on selecting a basic platform suitable for architects, upon which the technique can be developed. Rhinoceros,
Ecotect an architectural modeling program, is used, along with its graphical algorithm editor Grasshopper, to estab-
Radiance lish such technique by incorporating three performance simulation programs, namely Ecotect, Radiance, and
EnergyPlus EnergyPlus. Design cases are presented to demonstrate the technique and its effectiveness.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction building project team, urgently need to study and grasp the new de-
sign philosophy and the supporting technique to ensure the design
Performance-driven architectural design emphasizes on integrated quality while keeping the good essence of the conventional design.
and comprehensive optimization of various quantifiable performances
of buildings. It is an important research subject and a design philosophy
being practiced by many architects and design firms [1]. Compared with 1.1. Green buildings and architects
the conventional architectural design methodology, which focuses
on space and form, performance-driven design takes a holistic view Reviewing the history of green buildings shows a clear pattern
towards ecological and environmental performances of buildings that it is largely the research, development, and utilization of new
while ensuring that the functions and esthetics of the design are not materials and/or mechanical equipments that lead the progress of
overlooked. It is particularly important in countries undergoing fast the field. As a consequence, the green building is gradually becoming
urbanization such as China [2] since the performance of many newly a high-tech architectural machine and architects, the supposedly
built buildings affects the overall quality of urbanization. leading professional, are somewhat lost. The conventional architec-
A major boost for performance-driven architectural design is the tural design methodology is often powerless facing the scientifically
implementation of green building standards internationally such as rigorous and quantifiable performance criteria. The implementation
LEED in the US [3], BREEAM in the UK [4], and the newly launched of green building standards exacerbates this problem. Two kinds of
green building evaluation standard in China [5]. These green building so-called green architects can be identified.
standards establish many quantifiable performance requirements
to guide and control the design. Thus, performance-driven design is
encouraged and more rational thinking and scientific analysis are • The first kind is an architect who follows the conventional design
brought into the field of architectural design. As more and more approach to complete the conceptual design. He then turns the
green buildings emerge, architects, the leading professional of a work to other professionals such as consulting engineers or me-
chanical engineers to apply various green technologies. In this
way, the green building is designed by rigidly adding technologies
without adequate integration and optimization.
• The second kind is an architect who aims at designing a green
⁎ Corresponding author. Tel.: +86 25 83790037; fax: +86 25 83790682. building in the conceptual stage. Due to the lack of specialized
E-mail address: [email protected] (X. Shi). knowledge and technique, he uses conceptual, non-quantifiable,

0926-5805/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.autcon.2013.01.015
126 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135

and sometimes even vague methods to design. The end result is a can be made. The performative outcome of different designs should
quite subjective design which may or may not be truly green. be quantifiable and visible to the client and the architect.

The first kind of green architects already loses control of the de- 1.3. Performance-driven design and digital technology
sign. The final quality and the performance achieved are determined
primarily by the effectiveness of technology summation. The cost is Conducting performance analysis and optimizing the design effec-
significantly higher and more importantly, no integration or optimi- tively and efficiently used to be challenging. Lately, the rapid progress
zation of technologies is conducted. The second kind of green archi- of digital technology and its application in architecture have changed
tects is still the leading professional in the design team. However, the field dramatically. The emergence and development of perfor-
since the design methods he applies are conceptual and non- mance simulation tools make rapid performance analysis possible.
quantifiable, he has no real control over building performances. His More and more architects and engineers become familiar with these
interpretation of the design can only circle around ideas and/or tools and proactively use them in the design. More powerful personal
concepts. Performance criteria cannot be rigorously checked, not to computers shorten the time needed for analysis. In recent years,
mention satisfy green building standards. building performance simulation has become a very active research
Therefore, what architects urgently need is a new methodology field. On one hand, researchers and specialists are studying and pro-
and the supporting technique of performance-driven design which ducing more powerful simulation tools. On the other hand, practi-
can be applied in the early conceptual design stage and are reason- tioners start realizing the value of incorporating them into the
ably precise, efficient, and dependable. design process. This combined force pushes performance-driven ar-
chitectural design to the forefront.
1.2. Conventional methodology and performance-driven architectural This paper aims to analyze the concept of performance-driven archi-
design tectural design and its current status. We argue that the performance-
driven architectural design is a design philosophy that must be
The conventional architectural design methodology is, in essence, supported by effective and efficient design technique. Equally impor-
an approach involving some basic design principles, mainly based on tant, the technique should be viewed familiar and practical from the
functions and forms. The driving force is the combination of the perspective of architects. First and foremost, the technique must involve
architect's rationality and sensibility. When performance criteria a modeling platform that architects feel comfortable with. Secondly, the
must be met, this design methodology is facing unprecedented chal- simulation tools that can be integrated into the design flow should
lenges. Architects have to deal with the following three problems. cover the most important performances that architects need to consid-
er. Last but not least, the optimization algorithms should be readily
• The prerequisite for performance analysis is a building model that available. Very few architects can and want to write their own codes
can be analyzed. However, the complexity and variance of buildings for optimization. Such performance-driven design flow and technique
make an analyzable model quite difficult to obtain. The current can assist the architect to explore many design possibilities and
practice usually involves setting up a model using design software their corresponding performances in a convenient way. The end result
and then importing the model into performance simulation pro- is a design that is esthetically pleasing, spatially efficient, and per-
grams. This process is time-consuming and labor-intensive. formatively sound.
• The model created in most modeling programs only contains geomet-
ric information (the latest development and application of building 2. Literature review
information modeling might change it). Many non-geometric param-
eters have to be input in the simulation program. This, combined with Many architects, engineers, and researchers have aimed at achiev-
the previous point, discourages the engineer to use the architect's ing performance-driven design. Research works conducted by dif-
model for performance simulation purposes. Rather, he prefers to di- ferent professionals show distinguishable features. Analyzing these
rectly set up the model in the simulation program for he can input features helps us understand what kind of approach and technique ar-
both geometric and non-geometric information at once. chitects need. The following section presents a short literature review
• However, the modeling capability of most simulation programs is not on performance-driven architectural design. The literature is grouped
on the level of commonly used architectural modeling programs, es- into three categories based on who the leading professional is.
pecially when dealing with complex shapes and forms. Here a dilem- The first category of the research work is led by computer scien-
ma arises, i.e., the engineer does not want to use the architect's model tists or software engineers. Performance-driven design is achieved
because he has to import it and add many parameters before a simu- by developing source codes from the very bottom. Very few architects
lation can be run; on the other hand, the architect is not satisfied are involved since they don't possess the special knowledge and skills
with the engineer's simplified model and believes that it lacks details to write a significant amount of computer codes. For instance, Ellis et
and is not esthetically pleasing. al. [7] developed an automated multivariate optimization tool to per-
form energy efficient building design. The tool employs multiple
These three problems are difficult to overcome using conventional modules, including a graphical user interface, a database, a preproces-
design methodology. New approaches and techniques must be devel- sor, a simulation engine, an optimization engine, and a simulation run
oped to assist the architect to carry out performance-driven design. manager. All modules are customer written.
An architectural design process can be divided into three steps, The second category of the research work is led by consulting pro-
namely, conceptual design, detailed design, and construction docu- fessionals. The framework is to use commercially available optimiza-
ment design. It is widely agreed that design decisions made in the tion program and integrate building performance simulation tools to
conceptual stage have the largest impact on the final overall perfor- conduct performance-driven design. Some computer code writing is
mance of the building. Guillemin and Morel conducted a survey on usually needed to set up the design flow. Shi [8] used modeFRONTIER,
67 buildings and found that 57% of technological decisions were a commercially available optimization program, and integrated ener-
made in the conceptual design stage, compared with only 13% in gy simulation program EnergyPlus to study the optimal insulation
the detailed design stage [6]. Therefore, the methodology adopted strategy for an L-shape, one story building. Also using modeFRONTIER,
by the aforementioned two types of green architects clearly has lim- Manzan and Pinto [9] integrated ESP-r, an energy simulation program,
itations. The right paradigm is to incorporate performance analysis and Radiance, a lighting simulation program, to design an external
into the early conceptual design stage so that right technical decisions shading device in an office with a window and different glazing
 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135 127

Table 1
Comparison of the three categories of research work on performance-driven architectural design.

Category 1 Category 2 Category 3

Leading Computer scientists Building engineers and consultants Architects

professional and software engineers
Advantages Open source code Integrated platform, complete performance optimization Powerful modeling capability, visible feedback of performance
and fast processing simulation results, familiar to architects
Shortcomings Difficult coding User interface not friendly, geometric modeling capability limited Multiple programs required, slow speed, coding platform limited
Tools used Coding languages Commercial optimization programs such as modeFRONTIER and Geometric modeling programs such as Rhinoceros, performance
ModelCenter, performance simulation programs such as Ecotect simulation programs such as Ecotect

characteristics. The same approach can also be found in urban studies. terms of thermal and lighting performance in a building. The building
Bruno et al. [10] used modeFRONTIER and integrated CATIA, a performance simulation program used is DOE.
modeling program originated from aerospace engineering, to explore The three categories of research work on performance-driven
a non-linear design space whereby multiple objectives may be architectural design share some commonalities and also differ in
optimized concurrently. The authors defined and applied quantitative some ways. Table 1 presents a comparison.
metrics in order to examine the potential for a new workflow It should be noted that there are currently some ambitious and
in urban design. Flager and Haymaker [11] compared so-called large-scale research projects in progress that are intended to establish
Multidisciplinary Design Analysis (MDA) and Optimization (MDO) a comprehensive and powerful integrated performance-driven archi-
processes in the building and aerospace industries based upon case tectural design platform. A noteworthy example is the Virtual Design
data gathered on recent projects in each industry. Further, Flager et Studio project funded by the Department of Energy of the US and
al. [12] introduced multidisciplinary design and optimization technol- conducted at Syracuse University [19,20].
ogy and applied it on a classroom building design using another com- Based on the literature review, it is clear that performance-driven
mercially available optimization program called ModelCenter. Chronis architectural design has attracted much attention from architects,
et al. [13] integrated climatic and site data into a dynamic model of a engineers, and researchers. As the technology progresses, different
large student housing complex project using parametric and optimiz- approaches have been developed. A new trend is to provide more
ing technique. powerful architectural modeling capability and more user friendly in-
The above two categories of research work share one commonality, terfaces in the design flow. We believe it is valuable to develop a tech-
i.e., the leading professionals are not architects. The computer tools nique that architects feel comfortable with since they are the leading
involved in the design flow, especially the space and form modeling professional of a design team. In this paper, Rhinoceros, an architectural
programs, are either not familiar to architects or capable of generating modeling software, and its script coding component Grasshopper are
complex geometric shapes. Therefore, some architects and researchers selected as the platform, based on which a design workflow linking
with an architectural background start exploring performance-driven three performance simulation programs, namely Ecotect, Radiance,
design and its technology from the perspective of architects. Sargent and EnergyPlus is established.
et al. [14] from the Harvard Design School presented a new approach
called SHADERADE. The authors actually developed an eponymous
tool to assess the performance of different shading designs. Stravoravdis 3. Methodology
and Marsh [15] proposed a method to use LUA-scripting to control
and manipulate the model and data within a building performance 3.1. Selection of the design platform
simulation program. A similar approach was taken by Kawakita to opti-
mize the window design [16]. Mark [17] used Bentley's Generative Sketchup, 3dsMax, Maya, etc. are the commonly used modeling
Component and linked it with Ecotect to evaluate whether larger trans- programs among architects. Lately, Rhinoceros has gained popularity
formations to the structure as a whole or smaller movement in the fab- because of its powerful modeling capability, especially for complex
ric would help to optimize the solar insolation benefits. The author geometric shapes, flexible expandability, and relatively low require-
acknowledged that the technique was initially developed by DeBiswas ments on computer hardware. In addition, Rhinoceros provides an
at MIT. Caldas and Norfork [18] applied genetic algorithm as a genera- effective user development platform called Grasshopper, thus enabling
tive and search procedure to look for optimized design solutions in architects to customize for some complex projects with special needs.

Fig. 1. The workflow of performance driven conceptual design.

128 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135

Fig. 2. Diagram of the workflow for integrating Ecotect into Rhinoceros/Grasshopper.

Studying the technique of performance-driven architectural de- simulation results can be automatically fed back to the modeling pro-
sign based on Rhinoceros/Grasshopper is valuable for the following gram to guide the design optimization controlled by certain algorithms.
reasons. Thus, the key to the workflow is a data exchange and communication
system to control the entire design and analysis process. Its procedure
• Performance-driven architectural design, while emphasizing on can be decomposed as follows.
performance optimization, must simultaneously consider space
and shape, two of the major design considerations that architects (1) Use the modeling program to generate one or multiple initial
will never neglect. Therefore, the design workflow should incorporate designs and parameterize them.
modeling programs familiar to architects. Rhinoceros/Grasshopper is (2) Select the needed geometric, material, physical, and environ-
such a program. Design workflow and technique based on it would mental parameters and transport them to the performance
be friendly to architects. simulation program.
• The powerful modeling capability of Rhinoceros/Grasshopper makes (3) Analyze the performance in the simulation program based on
it an adaptable platform for performance-driven design since it can the received data.
handle various conceptual designs from linear to non-linear and (4) Compare the simulation results to the predefined optimization
from simple to complex. objectives. If the objectives were met, end the design work-
flow. Otherwise, enter the next step.
For these reasons, we select Rhinoceros/Grasshopper as the (5) Automatically, usually driven by an optimization algorithm, or
modeling program to study the performance-driven design technique manually adjust the design and repeat the process.
and use Galapagos, a third-party program built in Rhinoceros, to
achieve the control of optimization. Fig. 1 shows a diagram of the design workflow. It is worth noting
that,

3.2. Design workflow • This design workflow is iterative. The number of iterations needed
to reach an optimal or desired design depends on various factors,
The essence of the Rhinoceros/Grasshopper-based performance including the initial conceptual design, the optimization objective,
driven design technique is to establish a workflow so that performance the optimization algorithm, and the manual adjustment, etc.

Fig. 3. The names and functions of the five user-defined components to integrate Ecotect into Rhinoceros/Grasshopper.
 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135 129

Fig. 4. Total yearly insolation and four parameters defining the roof surface shape.

• The effectiveness of manual adjustment is highly dependent on the

operator's knowledge and skills in the field of building performance.
• Automatic adjustment is usually driven by a predefined optimiza- Table 2
tion algorithm. Design parameters of the 108th conceptual design.
• The time and resources needed to complete a successful workflow ID OffsetBig OffsetSmall BiArc Angle Insolation
are highly dependent on the complexity of the performance simula-
108 2 4 0.56 −45 20,445,000
tion. If the optimized performance is insolation or such that does
130 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135

Fig. 5. The final shape of the roof surface with the optimized solar radiation.

not require a complex physical model, the workflow can be • Conceptual design generation and data input module. The conceptual
straightforward. However, if the architect wants to optimize for design generation and data input module provide capability of generat-
some performances that require complex analysis or involves ad- ing designs and inputting data. Design generation starts from setting
vanced tools such as CFD (Computational Fluid Dynamics) program, up geometric models, such as mesh, surface, solid, etc. Since perfor-
the design workflow can be time-consuming and resource inten- mance simulation programs have specific requirements on the format
sive. In addition, the pre-defined optimization objectives also affect of geometric models, one should select the type of geometric models
the time and resources needed to complete a successful workflow. appropriately. Input data include geometric data, physical data, and al-
For example, a weak threshold can be quickly reached while a strin- gorithm data. Geometric data describe the geometric model and need
gent one may take much longer. to be readily transferred into the performance simulation program.
Physical data include material properties, environmental parameters,
The aforementioned design process is achieved by connecting and others that are necessary to conduct performance simulation in ad-
various modules into one coherent workflow. These modules are as dition to geometric data. The data format can be integer, double, long,
follows: or string. Algorithm data are important to the optimization process
control and the workflow efficiency.
• Graphic User Interface (GUI) module. The GUI module is the beginning • Communication module. The communication module links the GUI
of the workflow and the one that architects are most familiar with. It module to the performance simulation module. For close source pro-
performs functions like modeling geometry, setting up the workflow, grams, one needs to establish a client program to enable the communi-
inputting commands, graphically showing the performance simulation cation between processes. In this paper, we call for Ndde.dll to achieve
results, etc. The commonly used GUI modules in architectural design the data exchange between Rhinoceros and Ecotect and Radiance. For
are Rhinoceros, Maya, 3dsMax, Processing, Houdini, etc. open source programs such as EnergyPlus, we write DOS batch file to
• Optimization module. The optimization module generates new sets of achieve that.
design parameters, usually driven by optimization algorithms, to define • Performance simulation module. The performance simulation module
a new design and attempt to reach the predefined design objectives. calculates and analyzes various performances that will be used to

Fig. 6. Workflow of integrating Radiance into Rhinoceros/Grasshopper.

 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135 131

Fig. 7. The names and functions of the six user-defined components to integrate Radiance into Rhinoceros/Grasshopper.

optimize the conceptual design. Most of performance simulations re- • Integrate Radiance into Rhinoceros/Grasshopper and study a win-
quire specialized and complex programs. Commonly used performance dow design strategy to achieve optimal natural lighting,
simulation programs include energy simulation programs such as • Integrate EnergyPlus into Rhinoceros/Grasshopper and study
EnergyPlus, lighting simulation programs such as Radiance, integrated the conceptual design of a simple building to minimize its energy
performance simulation programs such as Ecotect, etc. Most of the per- consumption.
formance simulation programs are based on Windows operating sys-
tem and are closed source. Few are based on Linux or DOS and are The focus is on establishing the workflow and the technique
open source. For the latter, users can write commands to run applica- and demonstrating their applications. In addition, the examples
tions and obtain more control and flexibility. presented, although simplified, all have practical design backgrounds.
• Performance simulation results feedback and analysis module. This
module is to store the performance simulation results and feed them 4.1. Rhinoceros/Grasshopper and Ecotect integration
back to the optimization algorithm module. For open source programs,
the user needs to access the output file and select data needed. When a The design problem considered originates from a project located
large number of data are produced, special database may be needed for in Nanjing, China. It is a building with a complex roof. The geometric
storage. For instance, the technique presented in this paper involves shape of the roof is defined using a NURBS surface. For more informa-
using Slingshot, an add-in in Grasshopper, to store all relevant data in tion on NURBS surfaces and curves, one can refer to [21]. The client
the workflow. wants to integrate a photovoltaic system on the roof and produce
electricity as much as possible. Thus, the shape of the roof surface be-
4. Technique and application of the performance-driven comes a critical design issue and is obviously performance-driven.
design workflow The problem would be much easier if the roof had a regular shape.
With a complex shape like NURBS surface, it is quite challenging. A
The previous section presents a workflow and its modules for technique of integrating Ecotect into Rhinoceros/Grasshopper was
performance-drive design. This section discusses in detail the tech- developed and a performance-driven workflow was established to as-
nique of applying such workflow in different situations, namely, sist the architect to determine the optimal roof surface shape.
To integrate Ecotect into Rhinoceros/Grasshopper and create the
• Integrate Ecotect into Rhinoceros/Grasshopper and study the con- workflow, we used Inter-process Communication in Windows system
ceptual design of a roof with a complex geometric shape to achieve to enable communication between processes and simultaneously
maximum insolation, sent commands from Rhinoceros/Grasshopper to Ecotect. In this way,

Table 3
Material reflectance and optimization parameters.

Material reflectance (%) Optimization parameters

Ground Ceiling Interior wall Window (transparency in %) Max. stagnant Population Initial boost Maintain Inbreeding

20 90 50 70 20 10 1 5% 75%
132 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135

Ecotect was kept running in the background and the modeling environ- The workflow uses four parameters, namely OffsetBig, OffsetSmall,
ment, i.e. the user interface of Rhinoceros, that the architect is familiar BiArc, and Angle to control the roof surface shape. The driving perfor-
with was kept in the foreground. This is an important advantage that mance criterion is to obtain the largest total yearly insolation on the
should not be overlooked. The workflow is shown in Fig. 2. Note that roof which is simulated by Ecotect. After initiating the workflow, the
LUA in Fig. 2 is a scripting language used with Ecotect. user can observe the converging curve and decide when to terminate
The workflow was run in Grasshopper via user-defined com- the design process or let the program stop at the preset number of iter-
ponents written in VB.net. There are five user-defined compo- ations. Fig. 4a shows a plot of the total yearly isolation of each generated
nents, namely RUN ECO, WEATHERDATA, MODEL & MATERIAL, conceptual design. Fig. 4b, c, d, and e shows the plots of the four geo-
CALCULATION (OBJ or GRID), and RESULT. Their names and functions metric parameters defining the roof surface shape, respectively.
are summarized in Fig. 3. Note that IPC in Fig. 3 means Inter-Process Based on the observation of the data, the conceptual design with
Communication. the largest yearly insolation is the 108th one. Its corresponding de-
sign parameters are shown in Table 2. The final shape of the roof sur-
face is shown in Fig. 5.

4.2. Rhinoceros/Grasshopper and Radiance integration

Using natural lighting creatively has long been one of the favorite
design approaches for architects since it provides physical lighting
and offers spiritual values as well. Natural lighting is also an effective
way to minimize energy consumption of buildings. The design prob-
lem considered is a simple box shape residential building with a
6 m × 6 m floor plan. The design objective is to find the window de-
sign, i.e. height and width, to satisfy the requirement that the average
daylight illuminance at 900 mm above the floor is 500 lx, a standard
requirement in Chinese code for residential buildings. Note that the
Chinese standard requires that the illuminance due to natural lighting
shall be analyzed using a fully overcast sky model. Therefore, the ori-
entation of the window, assuming only one window for the building,
is not a factor in this case.
The development of natural lighting simulation tools can be dated
back to the 1970s. Nowadays, commonly used programs include
Lumen Micro 2000, AGI32, Lightscape, Ecotect, Radiance, etc. Radi-
ance was selected to demonstrate the workflow partly because it is
open source.
As an advanced daylighting simulation program, Radiance is not a
common tool for architects. It does not have a graphic user interface
and requires an accurate model with many parameters for simulation
purposes. The technique presented below uses Rhinoceros/Grasshopper
to set up the model, materials, and other simulation parameters.
Grasshopper makes it possible to add more parameters to the workflow
as the design and simulation evolve. In other words, one can start from a
simplified model and gradually make it more complex and closer to the
real model. Fig. 6 shows a diagram of the workflow.
We used VB.net to define modules that can automatically generate
parameters in Rhinoceros/Grasshopper. Files generated include *.rad
file, *.rif file, and *.bat file. The aim is to enable Radiance to be running
in the background while Rhinoceros/Grasshopper is running in the
foreground. To generate the required parameters for simulation, six
components were created in Grasshopper. Their names and functions
are shown in Fig. 7.
The input parameters are height and width of the window. The
output parameter is the average illuminance on the plane at
900 mm above the floor. The objective is to find the window design
in which the output parameter is no less than 500 lx. Other important
parameters are shown in Table 3.
The workflow was initiated and terminated when fitness showed
a clear trend of converging. Fig. 8a, b, and c shows the plots of fitness,

Table 4
Design parameters of the 547th and 636th design.

Design ID Window width Window height Average illuminance Fitness

(mm) (mm) (lux)

547/636 4992 1161 500 0.006051

Fig. 8. Fitness, illuminance, and width and height of the window.
 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135 133

China. The building has one window on each wall with a constant
total window area. Fig. 9 illustrates the building and its four windows.
The objective is to find the window area on each wall to minimize the
energy consumption. Let AS, AE, AW, AN denote the window area on the
south, east, west, and north walls respectively. The total window area
on four walls is set to be 7.29 m 2. The following equations are used in
the workflow to determine each window area:

AS ¼ s·7:29
AE ¼ e·ð7:29−AS Þ
ð1Þ
AW ¼ w·ð7:29−AS −AE Þ;
AN ¼ 7:29−AS −AE −AW

where s, e, w are numbers between 0.01 and 0.99.

The technique of integrating EnergyPlus into Rhinoceros/Grasshopper
is similar to that presented previously in the Radiance section. VB.net
was used to define components and automatically generate parame-
ters. The EnergyPlus input file, i.e. *.idf file, contains a lot of informa-
tion. To demonstrate the capability of the technique, we only created
four components, namely ZONE, GEOMETRY, RUN PERIOD, and PROJECT
INFORMATION. Fig. 10 summarizes the four components and their
functions.
Fig. 9. The rectangular building and its four windows. The optimization results are shown in Fig. 11. Fig. 11a is the ener-
gy consumption by assuming an ideal load system with a COP of 1.
Fig. 11b, c, d, e shows the window areas of the south, east, west,
average illuminance, window height and width, respectively. Table 4 and north walls, respectively. It is clear, based on Fig. 11a, that the
shows the two optimal designs and their parameters. The window overall energy consumption converges to a minimum value during
height and width are rounded to be 5000 mm and 1200 mm, the optimization process. One can select the best window design
respectively. strategy based on the optimization results and other practical design
considerations. In this case, the solution at Step 750, shown in
Fig. 11a, has the minimum overall energy consumption. Its corre-
4.3. Rhinoceros/Grasshopper and EnergyPlus integration sponding window design is 0.233 m2 for the east window,
0.046 m2 for the south window, 6.16 m2 for the west window, and
Energy consumption is one of the primary concerns for green 1.4 m2 for the north window. At his point, the architect can make a
building design. EnergyPlus is a standard energy simulation engine practical decision of not designing any window on the south elevation
developed by the US Department of Energy. EnergyPlus is open since the area is approximately zero and designing the other three
source and provides detailed descriptions, engineering references, ex- windows as shown above. The final design may not be strictly optimal
amples, etc. in a freely downloaded package. This section presents a but should be satisfactorily close and practical.
technique of integrating EnergyPlus into Rhinoceros/Grasshopper
and setting up a performance-driven workflow with similar features 5. Conclusions and future work
as discussed previously.
The considered design problem is a rectangular building with This paper studies the concept, methodology, and technique of
three dimensions of 4200 × 4200 × 3000 mm, located in Nanjing, performance-driven architectural design from the perspective of

Fig. 10. The names and functions of the four user-defined components to integrate EnergyPlus into Rhinoceros/Grasshopper.
134 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135

Fig. 11. Overall energy consumption and the window areas on four walls.
 X. Shi, W. Yang / Automation in Construction 32 (2013) 125–135 135

architects. The main arguments and conclusions are summarized as References

follows:
[1] X. Shi, Performance-based and performance-driven architectural design and opti-
mization, Frontiers of Architecture and Civil Engineering in China 4 (4) (2010)
• Performance-driven architectural design is attracting more and 512–518.
more attention from both practicing architects and researchers. It [2] C. Gu, L. Wu, I. Cook, Progress in research on Chinese urbanization, Frontiers of
Architectural Research 1 (2) (2012) 101–149.
is a response to the movement of green and sustainable design. [3] United States Green Building Council, LEED — New Construction and Major
• The conventional architectural design methodology cannot be fully Renovation V2.2 Reference Guide. Washington D.C. , 2007.
performance-driven because it primarily focuses on space, form, [4] D. Crawley, I. Aho, Building environmental assessment methods: applications and
development trend, Building Research & Information 27 (4–5) (1999) 300–308.
functions, and esthetics while performance-driven design needs to [5] Ministry of Construction, China, Evaluation Standard for Green Building, GB/T
develop and refine the design based on scientifically sound perfor- 50378. Beijing; 2006. In Chinese, 2006.
mance analysis. [6] A. Guillemin, N. Morel, An innovative lighting controller integrated in a self-adaptive
building control system, Energy and Buildings 33 (5) (2001) 477–487.
• The methodology and technique for performance-driven design
[7] P. Ellis, T. Griffith, N. Long, P. Torcellini, D. Crawley, Automated multivariate opti-
must address the needs of architects since they are the leading pro- mization tool for energy analysis, IBPSA SimBuild 2006 Conference, Cambridge,
fessional in a design team. Therefore, Rhinoceros/Grasshopper is Massachusetts, 2006.
considered an appropriate platform upon which such technique [8] X. Shi, Design optimization of insulation usage and space conditioning load using
energy simulation and genetic algorithm, Energy 36 (3) (2011) 1659–1667.
and workflow can be established. [9] M. Manzan, F. Pinto, Genetic optimization of external shading devices, The 11th
• Three important and commonly used performance simulation pro- International IBPSA Conference, Glasgow, Scotland, 2009.
grams, namely Ecotect, Radiance, and EnergyPlus, were integrated [10] M. Bruno, K. Henderson, H. Kim, Multi-objective optimization in urban design,
Proceedings of the 2011 Symposium on Simulation for Architecture and Urban
into Rhinoceros/Grasshopper to establish the workflows for Design, San Diego, USA, 2011.
performance-driven architectural design and optimization. Some [11] F. Flager, J. Haymaker, A comparison of multidisciplinary design, analysis and
program development and code writing capability were needed. optimization processes in the building construction and aerospace industries,
Conference on Information Technology in Construction, 2007, pp. 625–630.
• Each one of the three techniques and workflows was demonstrated [12] F. Flager, G. Soremekun, B. Welle, J. Haymaker, P. Bansal, Multidisciplinary process
using one practical design case. It is clear that they can facilitate the integration & design optimization of a classroom building, CIFE Technical Report
performance-driven design and assist the architect to make sound #TR175, Stanford University, 2008.
[13] A. Chronis, K. Liapi, I. Sibetheros, A parametric approach to the bioclimatic design
design decisions.
of large scale projects: the case of a student housing complex, Automation in Con-
struction 22 (2012) 24–35.
Although the methodology and techniques developed in the paper [14] J. Sargent, J. Niemasz, C. Reinhart, Shaderade: combining Rhinoceros and
EnergyPlus for the design of static exterior shading devices, Building Simulation
are valuable, they have some limitations that warrant future research
2011, Sydney, Australia, 2011.
work. First, the workflows were set up using customer written com- [15] S. Stravoravdis, A. Marsh, A proposed method for generating, storing and manag-
puter codes which are not familiar to architects. It should be worth- ing large amounts of modeling data using scripts and on-line databases, The 9th
while to standardize the codes and develop architect-friendly International IBPSA Conference, Montreal, Canada, 2005.
[16] G. Kawakita, Environmental Optimisation Methods in Sustainable Design Process:
interface so that no coding capability is required. Secondly, the opti- In Combination with Evolution-Based Digital Technology. Master thesis, Oxford
mizations demonstrated are all single objective. However, most, if Brooks University, 2008.
not all, architectural design problems are multi-objective. Therefore, [17] E. Mark, Optimizing solar insolation in transformable fabric architecture: a para-
metric search design process, Automation in Construction 22 (2012) 2–11.
expanding the technique from single objective to multi-objective is [18] L. Caldas, L. Norford, A design optimization tool based on a genetic algorithm,
a natural next step. Automation in Construction 11 (2) (2002) 173–184.
[19] P. Pelken, J. Zhang, W. Feng, Y. Chen, D. Rice, L. Gu, H. Henderson, Virtual design
studio — part 1: design process, Proceedings of CHAMPS 2011: The 8th Interna-
tional Forum and Workshop on Combined Heat, Air, Moisture and Pollutant
Acknowledgment Simulations, Nanjing, China, 2011.
[20] J. Zhang, P. Pelken, W. Feng, Y. Chen, D. Rice, L. Gu, H. Henderson, Virtual design
studio — part 2: platform design, Proceedings of CHAMPS 2011: The 8th Interna-
The financial support of this paper is provided by the National
tional Forum and Workshop on Combined Heat, Air, Moisture and Pollutant
Natural Science Foundation of China (51008058) and Ph.D. Programs Simulations, Nanjing, China, 2011.
Foundation of Ministry of Education of China (20100092120001). [21] L. Piegl, W. Tiller, The NURBS Book, Springer, Germany, 1996.