Proceedings of Building Simulation 2011:

12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

PROCESS-DRIVEN BIM-BASED OPTIMAL DESIGN USING INTEGRATION OF

ENERGYPLUS, GENETIC ALGORITHM, AND PARETO OPTIMALITY

Se-Min Oh1, Young-Jin Kim 2, Cheol-Soo Park2 and In-Han Kim3,

1
Department of u-City Design and Engineering, SungKyunKwan University, South Korea
2
Department of Architectural Engineering, SungKyunKwan University, South Korea
3
Department of Architecture, College of Engineering, Kyung Hee University, South Korea
E-mail: cheolspark@skku.ac.kr

interoperability of BIM in the schematic design stage.

ABSTRACT The use of BIM-based energy performance
This paper addresses optimal architectural design assessment during the design stages is believed to be
using BIM at the design stage. We present a essential to achieve energy-conscious green buildings.
comparative study of the differences between data- BIM allows architects and engineers to receive
centric and process-driven interoperability building information as a detailed and digitalized
approaches during the design phases. The authors virtual model. However, the use of BIM-based
will elaborate on two approaches, summarize their energy performance simulation in the design phase of
consequences, and present a real application of the a building project is still controversial. In other words,
approach using a DAI (Design Analysis Integration) for an improved BIM-based simulation, one of the
prototype (Augenbroe et al, 2003) based on process- important issues is how architects and HVAC
driven interoperability in a schematic design stage. engineers cope with interoperability between design
To solve a multi-criteria optimal design problem of a and energy performance tools.
BIM-based energy performance simulation model, In order to analyze interoperability results between
we use Genetic Algorithm (GA) and Pareto BIM-based CAD applications and building energy
optimality. The term ““optimal design”” refers to the analysis tools, Revit Architecture and EnergyPlus 6.0
selection of various exterior double glazing were selected. For this study, one library building (5
systemsfor minimizing energy use and stories with a 7,235m2total floor area) was selected.
satisfyingthermal comfort. The software EnergyPlus It should be noted that in the process of the building
6.0 developed by DOE (U.S Department of Energy) energy performance assessment at each design stage,
was used for energy and comfort simulation. The the problem of selecting optimal design alternatives
integration of GA, Pareto optimality and EnergyPlus is a multi-criteria optimization problem. In the study,
6.0 simulation runs was automatically performed in the GA (Genetic Algorithm) was chosen to find
the MATLAB platform. The results of this study optimal design alternatives and Pareto optimality was
indicate that meaningful information can be delivered selected to deal with multi-criteria. The GA is known
to DM based on process-driven interoperability using for solving constrained discontinuous non-linear
BIM and GA with Pareto optimality. optimization problems. The selected architectural
design variables include glazing material, thickness,
INTRODUCTION cavity gas, etc. The chosen performance criteria for
With the recently increased expectation of low- decision-making include energy use and
energy building design, there is a growing need for PMV(Predicted Mean Vote). In particular, the
building elements analysis and searching for optimal authors developed an automatic data transformation
design using energy performance simulation for the (from gbXML to IDF) as well as input file generation
design analysis integration process in the design in the MATLAB GUI (Graphical User Interface)
stages. However, some important issues are involved platform since genetic algorithm requires numerous
in the application of design analysis integration in simulation samplings.
actual building design practice as follows: (1) the This paper describes the utilization of BIM-based
difficulty of effective exchange of information simulation and the searching of optimal design
amongthe stakeholders (2) the amount of design alternatives on multi-criteria problems for building
information/knowledge (3) the significant amount of energy performance assessment in a schematic
time and effort required for finding the optimal design stage.
design alternatives, and (4) the different assessment
method required at each building design stage. BIM-BASED ENERGY PERFORMANCE
Building energy consumption is affected by SIMULATION
architectural elements (geometry, type of envelope,
floor plan and site plan) which are determined from Interoperability
the design stages. Therefore, the goal of this study is It is known that interoperability requires data
to search for an optimal design based on the exchange and decision-making between the domain

- 894 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

experts involved in the building project by using a x BIM data at each design stage
shared data model (Eastman, 2008). Especially, BIM x Building energy performance simulation tools
can be used for decision-making and for an effective
exchange between building information stakeholders x Prepared data (preconditions, analysis functions,
using various simulation tools in the building design etc.) and subjective judgment with knowledge of
stage. Additional work is necessary (for instance, 3D domain experts for BIM-bases simulation
geometry modeling, reentering energy input variables, Efforts of design analysis integration process
etc.) when analysis experts or simulationists conducta
Application of various performance simulation tools
building energy performance simulation using 2D-
(EnergyPlus, Esp-r, Ecotect, etc.) during design
based CAD applications. In contrast, BIM can be
stages is the most importantat aspect of sustainable
used for successful information exchange because it
growth and green buildings. For the abovementioned
already contains information for building energy
state, it is necessary to conduct a building energy
performance assessment.
performance assessment onan integratedsimulation in
A variety of interoperability methods are being used the design process.
between BIM based CAD applications and
Over the last 20-30 years, various efforts have been
performance simulation tools (for instance, using
made to integrate design and energy analysis,such as
BIM standard file format, Plug-in, data-centric
the BDA (Building Design Advisor) project
exchange). However, these have many problems
(Papamichael et al, 1997). The BDA project was a
involving limitation and supposition such as the
seamless integration simulation of design tools for
difficulty in the automatic data mapping of different
non-specialists (e.g., architects). However, this leads
domain software, the uncertainty of influential inputs,
to a decline in the quality of energy performance
assumptions and simplifications involved in the
analysis due to assumptions and simplifications of
processes of modeling and simulation, deformation
reality, and the reduction of analysis domain
and missing data (Augenbroe et al, 2003; Bazjanac,
expertise. Lawrence Berkeley National Laboratory
2008; Treldal, 2008). For this reason, it is necessary
(LBNL) suggested a middleware program (IDF
to involve the analysis expert’’s subjective judgment
Generator and IFC HVAC interface) for converting
based on experience and expertise in order to
IFC to the EnergyPlus input file (Bazjanac, 2003;
complement the insoluble problem of BIM data or
Maile et al, 2007). The middleware program is an
deciding on input variables related to energy
alternative for saving the time and cost of building
performance simulation when the simulationist
energy performance analysis. However, the approach
conductsa BIM-based simulation(Fig 1).
misses a framework for DM to find optimal design
based on the whole Performance Indicators (PI)
(energy use, thermal comfort, indoor air quality, etc.).
The COMBINE project is based on a process-driven
data exchange model using sub schema (Augenbroe,
1995). However, the COMBINE project has a
constraint in providingthe process for integration
design analysis. Augenbroe et al (2003) suggested
the Design Analysis Integration (DAI) prototype to
exchange information between the design and
analysis domains from the process-driven
interoperability of COMBINE. The DAI prototype
has a number of characteristics that will guarantee
objectivity and reliability of simulation results
G
according to the preconditions and analysis scenario
when domain experts attempt to exchange
Figure 1 BIM-based simulation model based on information using a workbench comprised of four
judgment and knowledge of actors layers (design information, structured simulation
In other words, automatically digitalized virtual models, analysis scenario, and software application
models of the BIM-based simulation require a and tools).
combination of the knowledge and efforts of all the Process-driven interoperability
domain experts (architect, mechanical engineer,
simulationist, etc.), based on the correlation of Interoperability of BIM can be divided into data-
input/output variables between BIM and simulation centric and process-driven interoperability
tools and interoperability through (Augenbroe, 2002). Data-centric interoperability can
modifying/complementing the building information be used for ‘‘structured idealizations’’ of a shared
with proper control. Therefore, the elementary information model built with simplification and
requirements for BIM-based simulation methods by assumption of reality with various dynamic
dynamic building energy simulation are as follows: simulations (Fig 2(a)). This method is problematic
because of the difficulty with the information of

- 895 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

automatic mapping among various domain tools and based on the principles of Charles Darwin’’s‘‘survival
the discontinuity with updating of BIM data due to of the fittest’’. Although GA, based on probabilistic
piled information with time. In contrast, process- search, is more computationally intensive than the
driven interoperability is pursued in close association gradient-based method, it can be effectively used in
with defining the standard procedures on finding the global minima and solving discontinuous
performance indicators and module-based and non-linear objective functions (Schaffer, 1985).
modification for the energy analysis and missing data
Pareto optimality
for each standard procedure (Fig 2(b)). This method
can be used as a leading means of objective exchange Optimal design generally consists of decision and
data through the correlation of input/output variables search, involving finding a solution based on trade-
and analysis process for determining a set of analysis off among multi-objective functions. A correlation
scenarios (or processes) over the collaborative work between decision and search can be divided as
of domain experts (architect, engineer, simulationist, follows (Van Veldhuized and Lamont, 2000;
etc.) prior to the exchange of data among domain Miettnen, 2001):
expert tools. This method can overcome data-centric
interoperability through the meshing of information x Approachช: a prior preference articulation
and adequate events using process definition in (decide ĺ search)
energy analysis and precondition because analysis x Approachซ: progressive preference articulation
domain experts can easily become subjective in
(decide ļsearch)
exchange data. With this in mind, this study will
show an application of the BIM-based simulation x Approachฌ: a posteriori preference articulation
model of process-driven interoperability with multi- (search ĺ decide)
criteria optimization.
In this study, the Pareto optimization using
““approach ฌ”” was selected since it allows the DM to
select their preferred solution after finding optimal
design alternatives.
Fig. 3 shows the concept of Pareto optimality. A, B,
and C always dominates A1, B2, and C2 (for
example, CO2 concentration and initial cost of A
show better results than those of A1). However, A, B,
and C are not dominated by any other solution within
G the feasible regions. The set of results of A, B, and C
(a) ‘data-centric’ (b) ‘process-driven’ is called the ““Pareto optimal set”” or ““efficient frontier””
Figure 2 Two way view of interoperability and the set of A1, B1, and C1 is called the
(Malkawi & Augenbroe, 2004) ““dominated Pareto”” (Deb et al, 2000). This approach
will indicate that meaningful information (Pareto
MULTIOBJECTIVE OPTIMAL DESIGN optimal set) can be delivered to the DM and the
weights between utilities are not necessary. This
Optimal design on architectural design study focuses on finding the Pareto optimal set using
Architectural design needs to be a goal-directed the multi-criteria optimization.
activity in order to achieve the goal of making
decisions about components and various physical
forms of building (Antony and John, 1998).
Standards for building performance assessments have
multi-objective functions such as energy
consumption (cooling, heating), PMV, CO2
concentration, initial cost, etc. The aforementioned
assessment standards are affected by many design
variables (for instance, window-wall ratio, form,
materials, etc). Consequently, architectural optimal
design involves searching for a solution set of
optimal designs that have multi-objective functions G
rather than one objective function (Wang et al, 2005;
Hopfe, 2009). Figure 3 Pareto Optimality (Kim and Park, 2009)

Genetic algorithm (GA) APPLICATION

Genetic Algorithm (GA) is one method of searching BIM and optimal algorithm for decision-making
for the optimal solution for repeated generations of The building design process is developed through
initial population, selection, crossover, and mutation sequential decision making (Mourshed et al, 2003).

- 896 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

Architects carry out decision-making when they interior zones. Perimeter and interior zones are
repeatedly combine assessment and analysis on divided by the virtual wall without heat transfer
various design alternatives. Necessary information (Stein et al, 2006).
for design phases continues to change as data In this study, the HVAC system modelling,
necessary for building energy analysis changes ““IdealLoadAirSystem”” (assumption of ideal
through the cooperation of the decision-maker. In operation such as mechanical, plant system), was
particular, it is necessary to provide a design used as provided by EnergyPlus 6.0 since it was
alternative and to exchange information with difficult to obtain specific information at the
objectivity and credibility through analysis of the schematic design phase. The window to wall ratio is
expert’’s knowledge and judgment. This is 48%.
particularly important for non-specialists (e.g.,
architects) since it is not possible to acquire all
information for the analysis in the schematic design
stage. For this reason, the goal of this study is to find
meaningful information using a BIM-based
simulation model base on an application of a DAI
prototype with process-driven interoperability and an
optimal algorithm for the decision-maker (Fig 4).

G
Figure 5 Revit Architecture modelling

Figure 6 Simulation model &fifth floor zoning

Table 1
The preconditions and analysis scenario

GENERAL INFORMATION
Project Library building energy analysis
Assessment 1. Energy use
Figure 4 Process of decision-making using BIM and criteria 2. PMV
optimal algorithm SIMUALTION CONDITIONS
Weather file Incheon (*epw)
Simulation model Simulation Heating design day : January 7th
th
A sample library building was selected for this study period Cooling design day : July 1
(5 stories, with a total floor area of 7,235 m2) in the Internal heat people Lighting Equipment
schematic design stage. The specific floor plan and gain 140W/person 12.4W/ m2 16.1W/ m2
types of windows has yet to be decided. The Revit Occupancy
0.1 person/m2
Architecture program was selected in this study since density
it supports information on the architecture of the Schedule people Lighting Equipment
campus, rooms (space division for thermal zoning), 00:00-09:00 : 0%
and BIM data (gbXML). gbXML, one of the BIM 09:00-18:00 : 100%
standard formats, can be used because it contains 18:00-24:00 : 0%
information about energy analysis, space zoning, Thermostat Heating Setpoint Cooling Setpoint
thermal information, HVAC system, and schedule control type 20ଇ 26ଇ
(Dong et al, 2007; gbXML, 2007). A Revit HVAC
Ideal Load Air System
Architecture model is shown in Fig. 5. system
As a building energy performance tool, EnergyPlus
6.0, developed by DOE, is utilized. In the simulation Representative constructions were classified into
model, the thermal (perimeter & interior) zones of three groups by ASHRAE (2009): Light, Medium,
each floor are separated (Fig.6). The outdoor air and Heavy. The authors selected Medium
affects the perimeter zones, but does not affect the

- 897 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

construction, the physical properties of which are necessary computerization must be handled by
shown in Table 2 & Table3. analysis expertsin order to automatically generate the
BIM-based simulation model. This is analogous to
Table 2 the fact that the high-tech robotic surgery could not
Class of Medium constructions be carried out without a doctor’’s adjustment
(expertise and judgment by human beings).
CLASS LAYER ID TYPE
M01 100mm brick
Exterior I02 50mm insulation board
Wall F04 Wall air space resistance
G01a 19mm gypsum board
G01a 19mm gypsum board
Partitions F04 Wall air space resistance
G01a 19mm gypsum board
M14a 100mm heavyweight concrete
Roof &
F05 Ceiling air space resistance
Ceiling
F16 Acoustic tile
F16 Acoustic tile
Floor F05 Ceiling air space resistance
M14a 100mm heavyweight concrete G
(a) User interface
Table 3
Properties of Medium constructions

RESIST
CONDUCT DEN SPECIFIC
ANCE,
ID IVITY, SITY, HEAT,
(m2ȝ
W/(mȝK) kg/m3 KJ/(kgȝ K)
K)/W
M01 0.89 1,920 0.79 ȟ
M14a 1.95 2,240 0.9 ȟ
F16 0.06 367 0.59 ȟ
I02 0.03 43 1.21 ȟ
G01a 0.16 800 1.09 ȟ
F04 ȟ ȟ ȟ 0.15
F05 ȟ ȟ ȟ 0.18
Exchange data (gbXML to IDF)
Using MATLAB GUI, the authors developed a
gbXML-IDF Converter that converts a gbXML file
to the EnergyPlus 6.0 input file (*.idf) (Fig 7).
Exchanged data can be classified into four groups: (1)
automatically possible mapped information, (2) G
missing data between BIM, (3) required data for (b) MATLAB coding
simulation environment/setting, and (4) unstructured Figure 7 gbXML-IDF Converter program
information.gbXML-IDF Converter completely
support automatic converting 3D geometry of Design variables and objective function
gbXML. It is necessary to input properties of Selection of building component base on simulation
constructions from the gbXML-IDF Converter base is import to design alternatives assessment in
as a precondition due to the limitation of Revit building design process. One example, the selected
Architecture, where by input properties of design variables included, such as the type of (1) 62
constructions cannot be enteredfor simulation and exterior window glazing (X1), (2) 16 cavity gas (X2),
recognized complexity ina single wall. The gbXML- and (3) 62 interior window glazing (X3) are provided
IDF Converter can also input requirements for the by EnergyPlus 6.0 for optimal double-glazing system
EnergyPlus 6.0 simulation (version, Time Step, alternatives for this study (Fig.8, total combination of
simulation period, convection algorithm, ground alternatives=62×16×62=61,504). The energy use
temperature, output variable type). At present, the (cooling energy, heating energy) and PMV during the
BIM standard format cannot fully support HVAC design day were chosen as elements of the objective
systems (Bajzanac, 2008). This study assumes that function. The boundary conditions of the design
analysis experts modified unstructured data such as variables for encoding as chromosomes in GA are
HVAC system information, internal heat gain, and shown Table 4. In this study, the selected object
the schedule in IDF Editor. It should be noted that the functions (Eq. 1) were solved by GA and Pareto

- 898 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

optimality because of the non-linear and multi- The optimization process consists of the following
criteria problem. The input variables for PMV four steps:
calculation were Air Velocity (0.137m/s), 0.5Clo x Step 1: generate the intial population
(summer), 1Clo (winter), and the MRT (Mean
Radiant Temperature) are used as the Zone Average GA demands an initial population for virtual
option (assumption that a person is in the center of a evolution. In this study, X1-X3 are randomly selected
within the range of the boundary conditions and then
space) (refer to ANSI/ASHRAE standard 55-2004).
converted into a 500×1vector (subpopulation: 20,
individuals in the subpopulation: 25).
x Step 2: calculation of objective functions
EnergyPlus 6.0 is used to calculate the objective
functions. Step2 involves the integration of
EnergyPlus 6.0 with GA on the MATLAB platform.
The process is as follows: (1) The EnergyPlus 6.0
simulation input files are read into MATLAB, (2) a
new individual generated by the GA is overwritten
into the EnergyPlus 6.0 simulation input file by
Figure 8 The design of optimal double glazing system MATLAB m-codes, (3) EnergyPlus 6.0 is executed
from MATLAB, and (4) the EnergyPlus 6.0 result is
read in MATLAB. This procedure is repeatedly
Table 4
performed to find the Pareto optimal set.
Boundary conditions of design variables
x Step 3: fitness sharing
DESIGN LOWER UPPER INCRE A non-dominated sorting genetic algorithm ซ
VARIABL BOUND BOUND MENT (NSGA-ซ) method was used to increase the
E TYPE ARY ARY diversity of the initial generation from the Pareto
X1 Discrete 0 0.62 0.01 optimal solution. NSGA-ซ is performed as follows:
X2 Discrete 0 0.16 0.01 (1) fast-nondominated-sort, (2) crowding-distance-
X3 Discrete 0 0.62 0.01 assignment, and (3) crowded-comparison-operator
(Deb et al, 2000).
‫ܨܰܫܯ‬ሺܺሻ ൌ ‫ܨ‬ሺ݂ଵ ǡ ݂ଶ ሻሺͳሻ
x Step 4: selection, crossover, mutation
where,
݂ଵ ൌ ‫݁ݏܷݕ݃ݎ݁݊ܧ‬ሺܹ݄݇ሻ Step 4 involved finding the Pareto optimal solution
݂ଶ ൌ ܲ‫ܸܯ‬ሺ݀݅݉݁݊‫ݏݏ݈݁݊݋݅ݏ‬ሻ set. The number of generations was set at 50. Step 4
is performed for selection (stochastic universal
Optimization process sampling selection), crossover (ratio: 0.9), and
The optimization process of integration GA, Pareto mutation (ratio: 0.003-0.1) after which the
optimality, and simulation are shown in Fig 9. optimization process was repeated according to the
number of generations.
The entire process automatically proceeds in the
MATLAB platform.
RESULTS
Analysis of the optimal solutions
Table 5 shows the results of the integration GA,
Pareto optimality and EnergyPlus 6.0 simulation
based on process-driven BIM-based simulation. The
Pareto optimal set consisted of 24 solutions and
always dominated the other options in the feasible
regions. For validation’’s purpose, three approaches
were employed.
x The Pareto method is better understood for the
dominated-nondominated solutions through a
visual two-dimensional graph.
TheDecision-Maker can determine 24 solutions that
are superior to the other solutions among the Pareto
optimal set using a two dimensional graph (Fig 10).
Figure 9 Integration of the GA, Pareto optimality, x Comparisons between the optimal solutions
and EnergyPlus simulation (parts of 24 solutions).

- 899 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

Table 5
Optimal designs (Non-dominated Pareto Solution Set)
PARETO OTIMAL SET
OBJECTIVE
OPTI DESIGN VARIABLE
FUNCTIONS
MAL
Energy
DESI X1 X2 X3
use PMV
GN (Exterior window glazing) (Cavity gas) (Interior window glazing)
(kWh)
1 ECABS-2 BLEACHED 6MM XENON 8MM LOW IRON 5MM 3,786 0.380
2 COATED POLY-66 XENON 6MM LoE SPEC SEL CLEAR 6MM Rev 3,249 0.396
3 COATED POLY-66 XENON 8MM LoE SPEC SEL CLEAR 6MM Rev 3,252 0.394
4 COATED POLY-88 XENON 8MM LoE SPEC SEL CLEAR 6MM Rev 3,373 0.383
5 COATED POLY-33 XENON 8MM COATED POLY-44 3,141 0.423
6 COATED POLY-33 XENON 6MM LoE SPEC SEL CLEAR 6MM Rev 3,154 0.419
7 COATED POLY-77 XENON 6MM LoE SPEC SEL CLEAR 6MM Rev 3,263 0.392
8 COATED POLY-55 XENON 8MM COATED POLY-33 3,147 0.420
9 COATED POLY-66 XENON 8MM COATED POLY-44 3,187 0.409
10 COATED POLY-55 XENON 8MM COATED POLY-44 3,154 0.416
11 COATED POLY-44 XENON 8MM COATED POLY-33 3,134 0.423
12 COATED POLY-44 XENON 8MM COATED POLY-44 3,144 0.421
13 COATED POLY-77 XENON 8MM COATED POLY-44 3,190 0.407
14 LoE CLEAR 6MM Rev XENON 8MM COATED POLY-44 3,208 0.404
15 LoE SPEC SEL CLEAR 6MM XENON 6MM ECREF-2 BLEACHED 6MM 3,330 0.387
16 LoE CLEAR 3MM XENON 6MM ECREF-2 BLEACHED 6MM 3,513 0.380
17 COATED POLY-77 XENON 8MM LoE SPEC SEL CLEAR 6MM Rev 3,266 0.391
18 LoE CLEAR 3MM Rev XENON 6MM LoE SPEC SEL CLEAR 6MM Rev 3,344 0.385
19 LoE CLEAR 3MM Rev XENON 8MM LoE SPEC SEL CLEAR 6MM Rev 3,347 0.384
20 LoE SPEC SEL CLEAR 6MM Rev XENON 8MM COATED POLY-55 3,182 0.409
21 LoE SPEC SEL CLEAR 6MM Rev XENON 8MM COATED POLY-33 3,156 0.415
22 LoE SPEC SEL CLEAR 6MM Rev XENON 8MM COATED POLY-44 3,167 0.410
23 LoE CLEAR 3MM Rev XENON 8MM COATED POLY-44 3,216 0.403
24 LoE SPEC SEL CLEAR 6MM Rev XENON 8MM LoE SPEC SEL CLEAR 6MM Rev 3,219 0.397

(modifications #1, #2, and #3 changes X3 (LoE

SPEC SEL CLEAR 6MM Rev), X2 (ARGON 6MM),
and X1 (LoE CLEAR 6MM)). As a result, optimal
design #2 dominated all the modifications (Table 6).
Table 6
Boundary condition of design variable

Objective
Modifications functions
Design
of the values Energy
PMV
use
Optimal design 2 - 3,249 0.396
1 X3 3,437 0.420
Modifications 2 X2 3,303 0.449
3 X1 3,297 0.404

Figure 10 24 Pareto optimal solution CONCLUSION

Optimal Design #1 vs. #10 (Table 5): Design #10 had This study presents an example of BIM-based multi-
criteria optimal design process. The gbXML-IDF
lower energy use than #1, whereas PMV of design #1
Converter is developed to exchange BIM data
was superior to #10.
(gbXML) with the EnergyPlus input file (*.idf).
x Comparisons of various values of design gbXML-IDF Converter is developed for exchange
variables between optimal solutions and BIM data to the EnergyPlus input file and additional
modifications (part of 24 solutions). work (handling missing data, modifications) was then
The authors compared Optimal Design #2 and the performed by analysis experts based on the
modifications of the values of design variables preconditions of a scenario. The GA and the Pareto
optimality were used to solve multi-criteria

- 900 -
 Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.

optimization problems. This process automatically environments, Proceedings: Building Simulation,

proceeded in the MATLAB platform. As a result, the pp. 1530-1537.
DM can make rational decisions based on meaningful Eastman. C, Teicholz. P, Sacks. R, Liston. K. 2008
information using BIM, GA, and Pareto optimality. BIM Handbook, Wiley & Sons.
ACKNOWLEDGEMENT gbXML. 2007.http://www.gbXML.org/. last
"This work was supported by the National Research reviewed, 02/17/2007
Foundation of Korea(NRF) grant funded by the Hopfe, C.J. 2009. Uncertainty and sensitivity
Korea government(MEST) (No. 2010-0029196)." analysis in building performance simulation for
"This work is financially supported by the Korea design for decision support and design
Minister of Ministry of Land, Transport and Martime optimization, Ph.D. thesis, Technische
Affairs(MLTM) as ೻U-City Master and Doctor Universiteit Eindhoven, Netherlands.
Kim, Y.J., Park, C.S. 2009. Multi-Criteria Optimal
Course Grant Program೼"
Design of Residential Ventilation Systems,
REFERENCES Proceedings of the 11th IBPSA Conference, July
27-30, Glasgow, Scotland, pp. 1806-1813.
Antony D. Radford, John S. Gero.1998. Design by
Optimization in Architecture, Buliding, and Maile, T., Fischer, M., and Bazjanac, V. 2007.
Construction, Van Nostrand Reinhold Company. Building energy performance simulation tools: a
life-cycle and interoperable perspective,
ANSI/ASHRAE Standard 55-2004. 2004. Thermal
Stanford, Califonia: Center for Integrated
Enevironmental Conditions for Human
Facility Engineering.
Occupancy, Atlanta, USA.
Malkawi, A. and Augenbroe, G. 2004. Advance
ASHRAE. 2009 ––ASHRAE Handbook Fundamentals.
Building Simulation, Spon Press.
Atlanta: American Society of Heating,
Refrigerating and Air-Conditioning Engineers, Miettinen K. 2004. Some Methods for Non-linear
Inc. Multi-objective Optimization, Lecture Notes in
Computer Science: Evolutionary Multi-Criterion
Augenbroe, G. 1995. COMBINE 2, Final Report.
Optimizatioin, v1993, 1-20.
Commission of the European Communities,
Brussels, Belgium. Available from Mourshed, M.M., Kelliher, D., Keane, M. 2003.
http://dcom.arch.gatech.edu/bt/Combine/my_ww ArDOT: A tool to Optimise Environmental
w/document.htm Design of Buildings, Proceedings of the 8th
IBPSA Conference, August 11-14, Eindhoven,
Augenbroe, G. 2002. Trends in building simulation.
Netherlands, pp. 919-926.
Building and Environment, Vol.37, pp.891-902.
Papamichael, K., Laporta, J., Chauvet, H. 1997.
Augenbroe, G., de Wilde, P., Moon, H.J., Malkawi,
Building Design Advisor: Automated integration
A. 2003. The design analysis integration (DAI)
of multiple simulation tools, Automation in
initiative, Proceedings of the 8th IBPSA
Construction, Vol. 6, No. 4, pp.341-352.
Conference, August 11-14, Eindhoven,
Netherlands, pp. 79-86. Schaffer, J. D. 1985. Multi objective optimization
with vector evaluated genetic algorithm,
Bazjanac. V. 2003. Improving building energy
Proceedings of the First International
performance simulation with software
Conference on Genetic Algorithms, Hillsdale,
interoperability, Proceedings of the 8th IBPSA
New Jersey, pp.93-100.
Conference, August 11-14, Eindhoven,
Netherlands, pp. 87-92. Stein, B., Reynolds, J., Grondzik, W. and Kwok, A.
2006. Mechanical and electrical equipment for
Bazjanac, V. 2008. Response to the OGC Request for
buildings, John Wiley & Sons, Inc.
Technology in Support of the AECOO Testbed,
Technical document. Building Technologies Treldal, N. 2008. Intergrated Data and Process
Department LBNL, Berkeley, CA. Control During BIM Design, Master's Thesis,
Technical University of Denmark.
Deb, K., Pratap, A., Agrawal, S. and Meyarivan, T.
2000. A fast multi-objective optimization: Van Veldhuizen D A., Lamont G B. 2000. Multi-
NSGA-ซ, Technical Report N0. 200001, objective Evolutionary Algorithms: Analyzing
Kanpur: Indian Institute of Technology Kanpur, the State-of-the-Art, Evolutionary Computation,
India. 8(2), 125-147.
Dong, B., Lam, K.P., Huang, Y.C., and Dobbs, G.M. Wang, W., Zmeureanu, R., Rivard, H. 2005.
2007. A comparative study of the IFC and Applying multi-objective genetic in green
gbXML informational infrastructures for data building design optimization, Building and
exchange in computational design support Environment, Vol. 40, No. 11, pp.1512-1525.

- 901 -