Documente Academic
Documente Profesional
Documente Cultură
H O S T E D BY
Alexandria University
ORIGINAL ARTICLE
a
Alexandria University, Alexandria, Egypt
b
Mansoura University, Mansoura, Egypt
KEYWORDS Abstract This research aims at investigating the potentials of parametric design optimization pro-
Building optimization; cess over the residential building to achieve more sustainable design as well as the computational
Parametric design; design methodology in the design process to activate the role of the computer in the design process
Genetic algorithms; not just as a drafting or visualization tool. It also investigates the essentiality of computational
Building performance simu- design and the state of art computer tools like Grasshopper, Octopus, Energy plus, Open studio,
lation; Radiance and Daysim in achieving an optimized parametric design. A residential building consists
Daylighting; of five stories is studied to find the best parameters which lead to the optimum performance in day-
Energy modelling lighting and energy performance. Building parameters such as WWR, construction material, glass
material, shading device configurations are optimized through closed-loop framework which con-
sists of three main portions; Parametric modelling which depends on Grasshopper software, build-
ing performance simulation is performed via Ladybug and Honeybee plugins which depend on
Radiance and Daysim in daylighting simulation while uses EnergyPlus in energy simulation, and
genetic algorithm which is performed via Octopus plugin. The generations of solutions formed in
Octopus are studied one by one to clarify by how much there is development in optimization pro-
cess and when the optimization is ended. Also, the optimum solution is illustrated in addition to the
best daylighting and energy performance ones.
Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
residential compounds and new communities in addition to the ‘simulation-based optimization’ [4]. Revelation and compar-
widespread use of domestic appliances especially the air condi- ison are the two main merits of the exploration of design trade-
tioners in the summer [1]. offs. Firstly, Tradeoffs clarifies issues you don’t take into your
In Egypt, the extreme sensitivity to changes in temperature consideration and in turn, future ways of exploration could be
(Gleick, 1991) caused an increase in electricity consumption to proposed by making new parts of the design solution space
put up with necessary cooling loads. This is evident in the fact attainable to further examination. Secondly, the computer
that the number of air conditioners used in Egypt has risen plays an essential role in realizing that the design satisfies speci-
from 700,000 in 2006 to 3 million in 2010 (MOEE, 2010). fic criteria which are better than others [6]. The possibility to
Air conditioners consume around 12% of the maximum pro- perform transformations that result in different configurations
ductive capacity of power stations. Resulting in a total con- of the same geometric components is also one of the principal
sumption of 22% of Egypt’s overall energy production in the advantages of a parametric model. This amplifies the effort of
building sector (see Fig. 1-1). These energy needs could be min- building a representation by providing an interpretation of a
imized by an intelligent choice of envelope materials [2]. The model as a typically infinite set of instances, each determined
proposed approach is developed by using the daylight and by a particular selection of values for the model’s independent
thermal simulation in Ladybug and Honeybee as the main dri- variables [6]. Parametric analysis is not a novel concept but
ven tool for the design development, and it will show the ben- automation and software availability has facilitated it to pur-
efits of using the parametric tools in the architecture design sue [7]. Parametric modeling is gradually adjusted to design
process in order to achieve a specific daylight and thermal practice by means of tools like Grasshopper, Dynamo, and
quality, then it explains the logic behind combining a validated Generative Components. The capability of adding many
analysis tool together with a parametric modelling tool in plug-ins to parametric modeling could be a practical approach
order to develop a closed-loop process between design and to ease holistic simulation support. Examples of plug-ins for
simulation which optimizes the architecture design based on the possibly most broadly used tool, Grasshopper, include:
the daylight and thermal requirements. (a) Honey- bee which links to Radiance, Daysim, Energy Plus,
and Open studio; (b) Mr. Comfy which facilitate interactive
visualizations of thermal simulations [8].
2. Literature review The combinations of independent parameters that generate
well-performing instances within the solution space of the
Optimization is the procedure of finding the minimum or max- parametric model could be searched by the evolutionary prin-
imum value of a function by choosing a number of variables ciples of genetic algorithms [6]. The complex natures of build-
subject to a number of constraints [3]. Many sciences, i.e. engi- ing simulation outputs, the scale of the problems, the expensive
neering, mathematics, economics, environmental science, com- computational cost, multi-objective design problems, and the
puter science, etc relate to building design optimization [4]. uncertainty of many factors during the optimization, including
The parametric simulation method is the way of developing design variables, environmental variables, model and con-
building performance. In this method, all variables inputs are straint uncertainty, etc are the essential problems in solving
kept unchanged except one variable to see its influence on BOPs by simulation-based methods [4] The increasing compu-
design objectives [4]. The using of parametric modeling cou- tational power available to process problems that were infeasi-
pled with optimization techniques has been noticed as a poten- ble previously increased the popularity of optimisation for
tial solution to give a smart searching approach for effective sustainable building design, and of multi-objective optimisa-
feedback [5]. To get an optimal solution to a problem (or a tion in particular [9]. Evolutionary algorithms (EAs) are a
solution near the optimum) with less time and labor, the com- common meta-heuristic optimization algorithm [9] and
puter building model is often ‘‘solved” by iterative approaches, stochastic search methods in which the natural biological evo-
which perform infinite sequences, of progressively better lution and/or the social behavior of species is simulated. These
approximations to a ‘‘solution”, i.e. a point in the search- algorithms have been developed to find the near-optimum
space that satisfies an optimality condition. These approaches solutions to large-scale optimization problems, which cannot
are commonly automated by computer programming. Such be solved with traditional mathematical techniques [10]. The
approaches are usually known as ‘numerical optimization’ or applications of evolutionary multi-objective algorithms have
The parametric based optimization framework daylighting and energy performance 3597
3.3.1. Massing
C. Shading angle: set to be (75° to 75°) with 15° B. Wall construction was defined to be the 4th parameter.
increment. Six constructions were defined (single red brick wall,
double red brick wall, Bulk cement wall, hollow cement
Distance to glass was a constraint which set to be 0.05 m. wall, single brick with insulation, double brick with insu-
Output shading Breps were sent to (Honeybee_ EP Context lation). These materials were defined as the most com-
Surface) component which was in turn processed in Honeybee mon construction materials in residential buildings in
context in the final simulation stage. Egypt by Shady Atteia.
C. Roof construction was set as a constraint for all
3.3.4. Surrounding context conditions.
To form surround context seen in Fig. 3-9 in simulation pro-
cess the remaining geometry of block building was created.
In addition, the buildings surround case study building was 3.4. Energy simulation
created. All of those Breps were exported to Grasshopper. In
turn, These Breps were added to (Honeybee_ EP Context Sur- 1. After that, Zones were sent to (Honeybee _ Set Energy Plus
face) component. Zone Thresholds) component in which cooling and heating
setpoint was assigned as 24 and 21 respectively. The mid-
3.3.5. Materials rise apartment does not have cooling or heating set back.
All zones in both categories were sent to (Honeybee _ Set EP Cooling set point is effectively the indoor temperature
Zone Construction). Within this component in Fig. 3-10, con- above which the cooling system is turned on, while the heat-
struction materials were set to walls, floors, and windows: ing setpoint is effectively the indoor temperature below
which the heating system is turned on.
A. For window materials, five materials were defined (single 2. The next step is (Honeybee _ Set Energy Plus Zone Load)
clear, double clear with argon gab, double clear with an component. In which equipment load per area, lighting
air gap, double low-E with an air gap and double low-E density per area and number of people per area were set
with vacuum gap). All of these materials were created in to (14 w/m2, 8 w/m2, and 0.04 ppl/m2) respectively, other
‘‘Window 7.5 ” software which is which created and inputs are the default.
developed by Barkley Lab. Then IDF files were exported 3. All zones were assigned to (Honeybee _ Assign HVAC Sys-
to GH to define materials through (Honeybee _ Import tem) component in which HVAC system was assigned to
Window IDF Report). PTAC ‘‘Packaged Terminal Air Conditioner” as seen in
Fig. 3-11. This System is familiar to be used for residential 5. Hence, Energy simulation process has been ended and the
purpose. As well as, Heating, cooling and air details were next stage was the optimization process. Before thinking
set to default. about optimization EUI was calculated by summing all
4. After all, the final station in energy simulation was reached. output loads generated from (Honeybee _Read EP Result).
Through (Honeybee _ Export to Open Studio) component This cooling, heating, electric lighting, and fan electric
in Fig. 3-11, HB zones were exported into Open Studio and loads were calculated monthly then summed to generate
run through Energy Plus. Otherwise, there were other EUI.
inputs needed to be assigned to this component:
A. HB Context which outputs of step 6 and 11 were sent
to it. 3.5. Daylighting simulation
B. Weather data file which contains all weather data
about specific location. Helwan EPW file was set as it Energy simulation model was exploited to setup daylighting
is the nearest location from 6th of October city which simulation process.
had weather data file.
C. Analysis Period, which was assigned through (Lady 1. Output zones from glazing component in step 5 in energy
bug _ Analysis Period) component from January to simulation setup were sent to (Honeybee_ Set Radiance
December. Material) component in Fig. 3-12 in which the materials
D. Simulation outputs, which was assigned through were set as follow:
(Honeybee _ Generate EP Output) component. This
component has controlled the type of calculation which A. For walls, floors, and ceilings, off-white plaster wall,
was performed and if these calculations were hourly, grey floor tiles and white paint room ceiling respectively
yearly or monthly as in that case. were placed as Constrain materials.
E. Filename, which is very helpful to find a specific simu- B. For window material, Radiance glass materials were
lation reports and results from hundreds of designed to match EP materials which were placed in
simulations. Energy simulation process.
F. Working Directories, is the path of Energy Simulation 2. Analysis Recipe was prepared to be processed in daylight-
files. ing simulation via (Honeybee_ Annual Daylight Simula-
G. Write OSM and run Simulation which is responsible tion) component in which points, vectors, and meshes of
for running energy simulation. test surfaces were added from (Honeybee_ Generate Test
Fig. 3–11 Set energy plus settings and run simulation (By Researcher).
Fig. 3–12 Assign radiance materials and set daylighting simulation settings (By Researcher).
The parametric based optimization framework daylighting and energy performance 3603
Points) component. The test surfaces were 0.076 m above 7. Shading Angle slider: (0–10) with 1 increment which rep-
finishing ground and 0.75 m * 0.75 m grid size. As recom- resents (75° to 75°) with 15° increment.
mended in LEED V4. In addition, weather data file of Hel-
wan was processed in recipe component. Octopus plug-in was running for 72 h on a laptop with
3. (Honeybee_ Run Daylight Simulation) is the main compo- (Intel(R) Core(TM) i7-6500U CPU @ 2.50 GHz (4 CPUs),
nent to generate a daylighting analysis. Output zones from 2.5 GHz) processor and 16 GB Ram. During this period
step 1 and Context buildings and Shading Breps which were many daylighting and energy simulations were running. About
output from step 6 and 11 in Energy simulation setup were 300 daylighting simulations and the same range for energy sim-
added to HB Objects. As well, analysis recipe, working ulation were run to form 6 generations which represent Pareto
directories and the number of CPU were set to this front in Fig. 3-14. The two axes Pareto front represent EUI on
component. the X-axis and SDI on Y axis. The fittest solution should be
4. Analysis results were converted to SDA values via (Honey- located near the Pareto optimal front curve which seen in
bee_ Read Annual Analysis Results) component. Fig. 3-14 as the two axes, a cross point represents the theoret-
ical optimum solution of this front. Due to the contradictive
relationship between Daylighting and energy performance,
3.6. Optimization there are many solutions which have high performance in
one objective but don’t have any an effective result in the sec-
In the process of optimizing for daylighting and energy con- ond objective and so on.
sumption, there are many conflicting parameters which are
interactive with each other and Octopus is trying to find the 4. Results and discussion
logical balance in between. The whole optimization Process
was performed via Octopus plug-in. As shown in Fig. 3-13, 4.1. Objective function
there were 4 inputs to be processed in Galapagos component
to run optimization: The following fitness function was used before by [7] to accu-
rately find the optimum solution in Pareto front while EUI and
A. Phenotype: it represents the geometry of zones used in SUDI were the objectives of this study:
the optimization process.
B. Objectives: EUI and SDI values were processed in this y ¼ ðsUDIi sUDImin ÞC1 þ 1ðEUIi EUImin ÞC2 ð4-1Þ
input.
C. Parameters\Genomes: 7 sliders were set to this where:
component: i = result of iteration,
1. WWRsouth slider: (0.1–0.9) 0.1 increment. min = minimum value of optimization set and
2. WWRsouth slider: (0.1–0.9) 0.1 increment. max = maximum value of optimization set
3. Window material slider: (0–4) with 1 increment which
100 100
represents the chosen five window materials. C1 ¼ ; C2 ¼
sUDImax sUDImin EUImax EUImin
4. Wall construction slider: (0–5) with 1 increment which
represents the six chosen wall construction materials. The values of fitness function were calculated for some solu-
5. Shading Depth slider: (0–5) with 1 increment which rep- tions of the Pareto front which represent varied cases of the
resents (0.05–0.3 m) with 0.05 increment. performance of daylighting and energy. These values were
6. Shading width slider: (0–4) with 1 increment which rep- arranged by descending in Table 4-1 which explains also the
resents (2–10) with 2 increments. objectives and parameters of these solutions.
4.3. Pareto front tion after generation 5 and 6 however, there is still improve-
ment in the density of Pareto optimal front region and
While generations of genomes were produced, every generation Pareto optimal front curve. Due to time limitation and there
contained genomes which are fitter than included in the previous is no improvement appeared in the optimum solution during
ones (see Figs. 4-3 and 4-4). As seen in Fig. 4-1 fitness function generations, Octopus optimization process was stopped at gen-
for the base case was 42.78. In turn, a great improvement was eration 6.
reached during generation 1 and 2 while there were solutions As seen in Table 4-1, the top ten solutions have similar
whose fitness reached up to 83.11. After generation 3 and 4 were characteristics which qualified them to have higher fitness
produced, the density of genomes in Pareto optimal front region function. However the WWR of southern elevation varies
and on Pareto optimal front curve became higher. between (0.6, 0.7 and 0.8), WWR of eastern elevation varies
As well, there is an improvement in the optimum solution between 0.2 and 0.3. In turn, the wall construction material
by 2.8%. There is not any improvement in the optimum solu- is two layers of 0.1 thickness red brick separated by
Egyptian glass wool isolation material with 0.05 thickness. In 4.4. The absolute optimum solution
addition, double low-E glass separated by a vacuum is the
material for glass construction. For shading devices, how- The optimum solution in Pareto front characterized by the bal-
ever, most of the louvers count differ between 2 and 4, their ance in the performance of daylighting and energy consump-
depth doesn’t have a definite pattern as it varied from 0.05 m tion. So, it achieved the best value of fitness function by
to 0.2 m. Louvers rotation angle is one of these two values 85.45. This genome was produced during the fifth generation
45° and 15° except for one solution which has a zero and remains the absolute optimum genome in Pareto front
rotation angle. during the sixth generation also. As seen in Fig. 4-5 energy
[10] E. Elbeltagi, T. Hegazy, D. Grierson, Comparison among five Autom. Constr. 22 (2012) 36–50, https://doi.org/10.1016/j.
evolutionary-based optimization algorithms, Adv. Eng. Inform. autcon.2011.08.001.
19 (2005) 43–53, https://doi.org/10.1016/j.aei.2005.01.004. [14] M. Peel, B.C.L. Finlayson, T.A. McMahon, Updated world
[11] J.H. Jo, J.S. Gero, Space layout planning using an evolutionary map of the Köppen-Geiger climate classification, Hydrol. Earth
approach, Artif. Intell. Eng. 12 (1998) 149–162, https://doi.org/ Syst. Sci. (2007) 439–473.
10.1016/S0954-1810(97)00037-X. [15] S. Attia, A tool for design decision making: Zero energy
[12] A. Wagdy, F. Fathy, A parametric approach for achieving residential buildings in hot humid climates, Presses univ de
optimum daylighting performance through solar screens in Louvain (2012).
desert climates, J. Build. Eng. 3 (2015) 155–170, https://doi.org/ [16] Climate-Data.org, climate: Sixth of October City, (2012)
10.1016/j.jobe.2015.07.007. <https://ar.climate-data.org/location/3677> (accessed
[13] M. Turrin, P. Von Buelow, A. Kilian, R. Stouffs, Performative November 18, 2017).
skins for passive climatic comfort: a parametric design process,