Conference PaperPDF Available

Multi-criteria evolutionary optimisation of building enveloped during conceptual stages of design

Authors:

Abstract and Figures

This paper focuses on using evolutionary algorithms during conceptual stages of design process for multi-criteria optimisation of building envelopes. An experiment is carried out in optimising a panelled building envelope. The design scenario for the experiment is based on the scenario described in Shea et al. (2006) for the building envelope of the Media Centre Building in Paris. However, in their research, the optimisation process only allowed panel configuration to be optimised. In this paper, the task is to approach the optimisation of the envelope of the same building, assuming it to be in the early phases of the design process. The space of possible solutions is therefore assumed to be much wider, and as a result both external building form and internal layout of functional activities are allowed to vary. The performance intent of the experiment remains the same as that of Shea et al. (2006), which was to maximise daylight and minimise afternoon direct sun hours in the building at certain specific locations.
Content may be subject to copyright.
T. Fischer, K. De Biswas, J.J. Ham, R. Naka, W.X. Huang, Beyond Codes and Pixels: Proceedings of
the 17th International Conference on Computer-Aided Architectural Design Research in Asia, 497–506.
©2012, Association for Computer-Aided Architectural Design Research in Asia (CAADRIA), Hong Kong
MULTI-CRITERIA EVOLUTIONARY OPTIMISATION
OF BUILDING ENVELOPES DURING CONCEPTUAL
STAGES OF DESIGN
VIGNESH SRINIVAS KAUSHIK AND PATRICK JANSSEN
National University of Singapore (NUS), Singapore
vigneshkaushik@gmail.com
Abstract. This paper focuses on using evolutionary algorithms during
conceptual stages of design process for multi-criteria optimisation of
building envelopes. An experiment is carried out in optimising a pan-
elled building envelope. The design scenario for the experiment is based
on the scenario described in Shea et al. (2006) for the building envelope
of the Media Centre Building in Paris. However, in their research, the
optimisation process only allowed panel conguration to be optimised.
In this paper, the task is to approach the optimisation of the envelope
of the same building, assuming it to be in the early phases of the design
process. The space of possible solutions is therefore assumed to be much
wider, and as a result both external building form and internal layout of
functional activities are allowed to vary. The performance intent of the
experiment remains the same as that of Shea et al. (2006), which was
to maximise daylight and minimise afternoon direct sun hours in the
building at certain specic locations.
Keywords. Multi-criteria optimisation; building envelopes; conceptual
stages of design evolutionary algorithms; parametric design.
1. Introduction
Building envelopes can be optimised for performance criteria that are often
in conict with each other and are also difcult, if not impossible, to optimise
manually. Hence, there is an opportunity to explore the use of computational
design and optimisation tools that facilitate the design of optimised building
envelopes for multiple criteria. Previous approaches to computational building
envelope optimisation include both evolutionary and numerical optimisation
approaches (Bouchlaghem 2000, Gagne and Anderson 2010). The problem
498 V. S. KAUSHIK AND P. JANSSEN
with most of these approaches is that they target the detail design stage when
most of the design is already xed, thereby losing on the opportunity for a
more holistic exploration of design alternatives.
This research explores the use of evolutionary optimisation techniques
(Holland 1975, Goldberg 1989) to optimise a building envelope during the
conceptual stages of the design process for multiple conicting performance
criteria. To deal with the evaluation of multi-objective optimisation problems,
Pareto ranking techniques are used (Goldberg 1989, Fonseca and Fleming
1993).
The design scenario for the experiment is based on the scenario described
in Shea et al. (2006) for the building envelope of the Media Centre Building
in Paris. The building consists of a 12 m × 20 m × 8 m parallelepiped shaped
space, split into ve internal spaces namely, gallery wall 1, gallery wall 2,
meeting area, reception and ofce (RP1, RP2, RP3, RP4 and RP5 respectively)
all with different lighting performance requirements as shown in Figure 1.
The building envelope to be optimised can be imagined as a pixel grid
of panels that wrap the walls and roof of a building, where each panel can
accommodate several types of panels with different lighting and cost proper-
ties (e.g. opaque, clear glass etc.).
In the design experiment by Shea et al. (2006), only the panel conguration
was optimised. The external envelope geometry and the internal plan congu-
ration were predened and were not allowed to vary. However, during the
conceptual stages of the design process, only optimising the panel congura-
tion may not be sufcient. It may be possible to nd better solutions by opti-
mising other aspects of the design in tandem with the panel conguration. In
this paper, the task is to approach the optimisation of the envelope of the same
building, assuming it to be in the early phases of the design process. The space
of possible solutions is therefore assumed to be much wider, and as a result
both external building geometry and internal plan layout are allowed to vary.
Figure 1. Paris media centre space
and response point (RP1–RP5) specication.
499MULTI-CRITERIA EVOLUTIONARY OPTIMISATION
The demonstration is divided into two experiments. In the rst experiment
the panel conguration and external geometry of the building are varied. In
the second experiment, the panel conguration and the internal oor plan
of the building are varied. The performance intent of both the experiments
remains the same as that of Shea et al. (2006), which was to maximise daylight
and minimise afternoon direct sun hours in the building at certain specic
locations.
2. Design method
An Evolutionary system called Dexen is used to execute the optimisation
process. This system is coupled to Sidefx Houdini, an advanced procedural
CAD application that is used for both design development and design evalua-
tion procedures (Janssen et al. 2011). This application includes a visual data-
ow modeling (VDM) interface that allows users to create complex parame-
trised procedures. With the evolutionary design method, the set of parameters
are referred to as the genes (which together from the genotype), the model is
referred to as the phenotype, and the results from the evaluations are referred
to as a set of tness scores. The evolutionary design method requires three key
procedures to be dened; namely the development procedure that generates a
phenotype from a set of genes, one or more evaluation procedures that calcu-
lates tness scores by analysing and simulation the phenotype, and a feedback
procedure that performs genetic reproduction based on the tness scores. The
developmental and evaluation procedures are dened using Houdini using the
VDM approach. The feedback procedure is generated automatically by Dexen
based on various standardised settings. This procedure will rank groups of
phenotypes using a standard Pareto ranking method, and will then create new
genotypes using crossover and mutation operators.
3. Lighting analysis
For the evaluation procedures, the Radiance simulation program is used for
calculating two different lighting performance criteria: lighting level and
number of sun hours. Calculations are performed for certain predened points
in the model. For lighting levels, the Daylight Factor is calculated for each
point. For sun hours, the total number of hours that a point receives direct
irradiance from the sun of at least 120 W/sq-m is measured. (The World Mete-
orological Organisation refers to this as ‘sunshine duration’.)
Within Houdini VDM nodes have been created for linking Houdini to Radi-
ance. These nodes generate the required input les from the Houdini model
and then execute the radiance program.
500 V. S. KAUSHIK AND P. JANSSEN
4. Experiment A
In the rst experiment, parameters related to the panel conguration and exter-
nal geometry are varied.
4.1. DEVELOPMENTAL PROCEDURE
The developmental procedure for Experiment-1 involves creating the para-
metric model in Houdini. Though the oor plan and the internal arrangement
of spaces is maintained as in Shea et al. (2006), the parameters governing the
shape of the parallelepiped (the envelope) are varied along with the panel
types. The three planes are divided into a total of 192 panels, each of which
can be assigned a different material. For the purpose of this experiment, the
materials are assumed to be the same as those described in Shea et al. (2006)
i.e. opaque panels, clear glass panels, diffusing glass panels and shaded glass
panels, as indicated in Table 1. In this experiment, the parameters that dene
the phenotype are the 192 panels and the movement of points P1, P2, P3 and
P4 in the X, Y and Z direction as shown in Figure 2. The movement of points
in X and Y directions controls the slope of envelope in that direction and
is given a range to move towards and away from the building by 3 m. The
movement of points in Z direction controls the height of the internal spaces
in the building and is given a range of 4 m to 8 m. As the shape and form of
the envelope changes, the size and shape of the panel change accordingly, still
maintaining the same number of panels throughout the evolutionary process.
Figure 2. Points that dene the geometry of the envelopes.
TABLE 1. Panel materials.
501MULTI-CRITERIA EVOLUTIONARY OPTIMISATION
4.2. EVALUATION PROCEDURE
There are in total ve performance criteria for which the strength of the pheno-
types will be assessed. The evaluation procedures are dened using Houdini.
The overall optimisation strategy can be dened as follows:
Maximise daylight factor at response points (maximum of 15%)
Achieve desired sun hours at response points
Minimise average U-value of the panels
Minimise overall cost of the panels
Achieve desired height at each of the ve internal spaces
The rst four criteria were part of the performance criteria dened by Shea
et al. (2006). The last criterion of height has been added for the purpose of this
experiment. Each internal space has different height requirements, taking into
consideration factors such as air-conditioning, privacy, lighting etc.
4.3. ANALYSIS OF RESULTS
The 3-D chart in Figure 3 represents 10,000 solutions, with daylight factor,
cost and U-value plotted along the X, Y and Z axis respectively. Sun hours are
indicated with colour coding. (On this chart, the fth performance criterion -
space height, also represented through colour coding - is not shown.) A score
of 5 indicates that the solution has satised all the rules set for that evaluation
criterion of sun hours or height. As shown in Figure 3, the solutions with the
maximum score of 5 for ‘sun hours’ gradually get denser towards the origin.
The Pareto front of these 10,000 solutions can be derived by ltering points
in the 3D graph.
Figure 3. 3D graphs representing the 10,000 solutions for Experiment A.
502 V. S. KAUSHIK AND P. JANSSEN
The Pareto optimal solutions shown in Figure 4 represent the nal outcome
of this experiment. The score for ‘sun hours’ and ‘height’ criteria has reached
the maximum for all the solutions in this graph. Hence it results in a 2D graph
with daylight factor and U-value plotted along the X and Y axis and the third
dimension of cost indicated with colour coding. For the purpose of compari-
son, six solutions are picked from the Pareto front and compared against each
other on their performance. Table 2 compares the values of the various per-
formance criteria of these 6 solutions. But as suggested by Caldas and Rocha
(2001), these solutions must not be interpreted as optimal answers, but as
diagnoses of potential problems and as suggestions for further architectural
explorations.
TABLE 2. Comparison of evaluation criteria for the 6 selected solutions of Experiment A.
Figure 4. Experiment A – Study of Pareto optimal solutions.
5. Experiment B
In the second experiment, parameters related to the panel conguration and
internal oor plan layout are varied along with the panel types.
503MULTI-CRITERIA EVOLUTIONARY OPTIMISATION
5.1. DEVELOPMENTAL PROCEDURE
The geometry of the building is maintained as a 12 m × 20 m × 8 m (constant
height) parallelepiped shaped space as in Shea et al. (2006). In this experi-
ment, the parameters that dene the phenotype are the panels and the move-
ment and rotation of gallery walls in the oor plan as shown in Figure 5.
In this experiment, the panelled roof and the two panelled walls are divided
into 1 m × 1 m panels, thereby yielding a total of 496 panels, each of which
can be assigned a different material as indicated in Table 1.
The positioning of the two internal partitions denes two spaces: the space
between these partitions and the solid walls is the gallery space, and the space
between these partitions and the panelled walls is the ofce space. (The ofce
space combines the ofce area, meeting area and reception area.) Paintings/
artefacts may be hung either on the solid walls or the side of the partition
facing the solid wall. Both partitions can move in X and Y directions within
the available oor plate of 20 m × 12 m, and can be rotated to a maximum of
90 degrees. In order to avoid generating overlapping partitions, the second
partition is actually positioned relative to the rst partition. The 90 degree
limitation to the rotation ensures that the two partitions always face the solid
walls. However for a given position of the gallery walls, there are 2 methods
of partitioning the internal space as shown in Figure 5. The method that results
in the desired ratio of partition of ofce and gallery areas is chosen.
Figure 5. Movement of gallery walls and the 2 methods of area partitioning.
5.2. EVALUATION PROCEDURE
The overall optimisation strategy is the same as Experiment A except that
the fth criterion in this experiment is to achieve desired ratio of oor area
between ofce space and gallery space. The desired ratio of partition of ofce
and gallery areas is specied in the evaluation procedure.
5.3. ANALYSIS OF RESULTS
The 10,000 solutions generated for this experiment are analysed in 3D graphs
and a set of solutions are picked from the Pareto Front and are compared
504 V. S. KAUSHIK AND P. JANSSEN
against each other on their performance. Figure 6 indicates both the internal
functional space arrangement and the external panel conguration for each of
the six selected solutions. Table 3 compares the values of the various perform-
ance criteria of these 6 solutions.
TABLE 3. Comparison of evaluation criteria for the 6 selected solutions of Experiment B.
Figure 6. Experiment B – Study of Pareto optimal solutions.
6. Conclusions
The experiments considered both geometry and oor plan based variations
to a panelled building envelope scenario presented by Shea et al. (2006).
Experiment-1 evolved solutions with envelopes that opened up to daylight
in certain parts and were self-shading in certain other parts of the building,
thereby minimising the use of costlier panel types. (It was noted that the
505MULTI-CRITERIA EVOLUTIONARY OPTIMISATION
actual cost of constructing such a façade may be higher due to the complexity
of the geometry.) Experiment-2 evolved solutions with better internal func-
tional space efciency for the same performance goals. The two experiments
demonstrate how designers can apply evolutionary techniques to explore dif-
fering combinations of design parameters and performance criteria early in
the design process during the conceptual stages. The VDM approach within
Houdini allowed complex developmental and evaluation procedures to be
developed without any advanced programming skills. The Dexen evolution-
ary system then allowed designers to setup and run advanced evolutionary
design explorations.
However, searching through the design variants within the archived evolu-
tionary data in order to make informed decisions was challenging due to both
the quantity and complexity of the data. When comparing solutions with more
than two performance criteria, it became difcult to visualise the strength of
each solution over the other. Alternative approaches that could be explored
include parallel projections, spider diagrams, and 3D and 4D graphs.
The experiments have demonstrated how designers could apply evolution-
ary algorithms to explore a wide range of design variants at early conceptual
stages of design process. Future research will focus on developing an improved
decision support system for analysing the archived evolutionary data.
References
Bouchlaghem, N.: 2000, Optimising the design of building envelopes for thermal performance,
Automation in Construction, 10(1), 101–112.
Caldas, L. and Rocha, R.: 2001, A generative design system applied to Siza’s School of Archi-
tecture at Oporto, in Gero, J. et al. (eds.), Proc. CAADRIA 2001, Sydney, 253–264.
Deb, K. et al.: 1998, Genetic programming 1998: proceedings of the third annual conference,
IEEE Transactions on Evolutionary Computation, 3(2), 159–161.
Fonseca, C. M. and Fleming, P. J.: 1993, Genetic algorithms for multiobjective optimization:
formulation, discussion and generalization, Genetic Algorithms: Proceedings of the Fifth
International Conference, San Mateo, 416–423.
Gagne, J. and Anderson, M.: 2010, Multi-objective facade optimisation for daylighting design
using a genetic algorithm, Proc. 4th National Conference, IBPSA-USA SimBuild 2010,
New York.
Goldberg, D.E.: 1989, Genetic Algorithms in Search, Optimization, and Machine Learning,
Addison Wesley, Reading.
Holland, J. H.: 1975, Adaptation in Natural and Articial Systems, University of Michigan
Press, Ann Arbor.
Janssen, P. H. T., Basol, C. and Chen, K. W.: 2011, Evolutionary developmental design for non-
programmers, Proc. 29th eCAADe Conference, University of Ljubljana, 886–894.
Shea, K., Sedgwick, A. and Antonuntto, G.: 2006, Multicriteria optimization of paneled build-
ing envelopes using ant colony optimization. Proceedings of EG-ICE, Ascona, 627–636.
“Distributed EXecution ENvironment”: 2011. Available from: <http://www. dexen.org/>
(accessed 05 December 2011).
... Shan Jie proposed a green building analysis method based on Grasshopper and completed the optimization design of the wind environment of building clusters. He also proposed plane energy-saving for individual buildings and exterior shading components [17]. Han Yunsong also proposed the method of building form generation under the influence of the environment. ...
Article
Full-text available
A complete evaluation of conflicting and nonlinear multiperformance is referred to as an evaluation system. Therefore, multiobjective optimization research for building performance is of great importance. Considering the above problems, this paper proposes a multiobjective performance optimization platform at the architectural design stage. The optimization research is combined with the basic algorithms. It is based on Grasshopper software. It aims to provide a fast and effective method for the optimization design of sustainable buildings. It is important to note that this workflow has no restrictions on what to optimize for or what to target, and can be determined on a case-by-case basis. To evaluate the use of the work platform and make the research results more extensive, the standard model of the inner corridor space building in office buildings in cold areas was extracted. The extracted standard model was taken as the final research object. At the same time, the convergence, distribution, and comprehensive evaluation indexes of the simulation results were analyzed. The quantitative influence parameter system of window parameters on the total building energy consumption, lighting uniformity, and glare probability was established. A series of parametric design models based on the comprehensive optimal solution of building energy consumption and lighting quality was obtained. The main purpose of this paper is to design a green building design method based on the principle of goal-oriented design. This paper, combined with the group search function of nondominated genetic algorithm (NSGA-II), the energy consumption of the building, and the evaluation function of indoor thermal and humid environment, is obtained and linked with MATLAB. The energy consumption of residential buildings and indoor thermal comfort were taken as the optimization objectives of the design scheme. The automatic optimization model of the architectural design scheme was established. Finally, the Pareto solution set of the architectural design scheme is obtained, and the designer can select the appropriate design scheme of the Pareto solution set according to the actual situation of the project, to meet the requirements of green building design. The design scheme optimized by this method can effectively reduce building energy consumption and improve indoor thermal comfort.
Article
A platform for experimenting with population-based design exploration algorithms is presented, called Dexen . The platform has been developed in order to address the needs of two distinct groups of users loosely labeled as researchers and designers . Whereas the researchers group focuses on creating and testing customized toolkits, the designers group focuses on applying these toolkits in the design process. A platform is required that is scalable and extensible: scalable to allow computationally demanding population-based exploration algorithms to be executed on distributed hardware within reasonable time frames, and extensible to allow researchers to easily implement their own customized toolkits consisting of specialized algorithms and user interfaces. In order to address these requirements, a three-tier client–server system architecture has been used that separates data storage, domain logic, and presentation. This separation allows customized toolkits to be created for Dexen without requiring any changes to the data or logic tiers. In the logic tier, Dexen uses a programming model in which tasks only communicate through data objects stored in a key-value database. The paper ends with a case study experiment that uses a multicriteria evolutionary algorithm toolkit to explore alternative configurations for the massing and façade design of a large residential development. The parametric models for developing and evaluating design variants are described in detail. A population of design variants are evolved, a number of which are selected for further analysis. The case study demonstrates how evolutionary exploration methods can be applied to a complex design scenario without requiring any scripting.
Conference Paper
Full-text available
Visual programming languages enable users to create computer programs by manipulating graphical elements rather than by entering text. The difference between textual languages and visual languages is that most textual languages use a procedural programming model, while most visual languages use a dataflow programming model. When visual programming is applied to design, it results in a new modelling approach that we refer to 'visual dataflow modelling' (VDM). Recently, VDM has becoming increasingly popular within the design community, as it can accelerate the iterative design process, thereby allowing larger numbers of design possibilities to be explored. Furthermore, it is now also becoming an important tool in performance-based design approaches, since it may potentially enable the closing of the loop between design development and design evaluation. A number of CAD systems now provide VDM interfaces, allowing designers to define form generating procedures without having to resort to scripting or programming. However, these environments have certain weaknesses that limit their usability. This paper will analyse these weaknesses by comparing and contrasting three VDM environments: McNeel Grasshopper, Bentley Generative Components, and Sidefx Houdini. The paper will focus on five key areas: * Conditional logic allow rules to be applied to geometric entities that control how they behave. Such rules will typically be defined as if-then-else conditions, where an action will be executed if a particular condition is true. A more advanced version of this is the while loop, where the action within the loop will be repeatedly executed while a certain condition remains true. * Local coordinate systems allow geometric entities to be manipulated relative to some convenient local point of reference. These systems may be either two-dimensional or three-dimensional, using either Cartesian, cylindrical, or spherical systems. Techniques for mapping geometric entities from one coordinate system to another also need to be considered. * Duplication includes three types: simple duplication, endogenous duplication, and exogenous duplication. Simple duplication consists of copying some geometric entity a certain number of times, producing identical copies of the original. Endogenous duplication consist of copying some geometric entity by applying a set of transformations that are defined as part of the duplication process. Lastly, exogenous duplication consists of copying some geometric entity by applying a set of transformations that are defined by some other external geometry. * Part-whole relationships allow geometric entities to be grouped in various ways, based on the fundamental set-theoretic concept that entities can be members of sets, and sets can be members of other sets. Ways of aggregating data into both hierarchical and non-hierarchical structures, and ways of filtering data based on these structures need to be considered. * Spatial queries include relationships between geometric entities such as touching, crossing, overlapping, or containing. More advanced spatial queries include various distance based queries and various sorting queries (e.g. sorting all entities based on position) and filtering queries (e.g. finding all entities with a certain distance from a point). For each of these five areas, a simple benchmarking test case has been developed. For example, for conditional logic, the test case consists of a simple room with a single window with a condition: the window should always be in the longest north-facing wall. If the room is rotated or its dimensions changed, then the window must re-evaluate itself and possibly change position to a different wall. For each benchmarking test-case, visual programs are implemented in each of the three VDM environments. The visual programs are then compared and contrasted, focusing on two areas. First, the type of constructs used in each of these environments are compared and contrasted. Second, the cognitive complexity of the visual programming task in each of these environments are compared and contrasted.
Article
Full-text available
A new generative design system based on a genetic algorithm is tested within the framework of Alvaro Siza?s School of Architecture at Oporto, Portugal. The system works over a detailed three-dimensional description of the building and uses natural lighting and overall environmental performance as objective functions to guide the generation of solutions. This paper researches the encoding of architectural design intentions into the system, using constraints derived from Siza?s original design. Experiments using this generative system were performed on three different geographical locations to test the algorithm?s capability to adapt solutions to different climatic characteristics within the same language constraints.
Article
Computer models for the simulation of the thermal performance of buildings have been in existence for many decades. Although these programs save a great deal of time and effort in computational terms they still rely on designers intuition and experience to achieve optimum solutions for a design problem. This paper presents a computer model which, not only simulates the thermal performance of the building taking into account design variables related to the building's envelope and fabric, but also applies numerical optimization techniques to automatically determine the optimum design variables, which achieve the best thermal comfort conditions. The main optimization program is supported by a graphical model for the design of window shading devices, which uses the numerical coefficients that define the window shading to model shading devices taking into account seasonal variations in solar angles and solar gains. The rationale and methodology used to develop the models is outlined and the resulting programs are described with examples of outputs. Finally, it is concluded that the models offer a valuable decision support system for designers at an early design state for the optimization of the passive thermal performance achieving optimum thermal comfort and savings in heating and cooling energy.
Conference Paper
Definition of building envelopes is guided by a large number of influences including structural, aesthetic, lighting, energy and acoustic considerations. There is a need to increase design understanding of the tradeoffs involved to create optimized building envelope designs considering multiple viewpoints. This paper presents a proof-of-concept computational design and optimization tool aimed at facilitating the design of optimized panelized building envelopes for lighting performance and cost criteria. A multicriteria ant colony optimization (MACO) method using Pareto filtering is applied. The software Radiance is used to calculate lighting performance. Initial results are presented for a benchmark and project-motivated scenario, a media center in Paris, and show that the method is capable of generating Pareto optimal design archives for up to 11 independent performance criteria. A preliminary GUI for visualizing the Pareto design archives and selecting designs is shown. The results illustrate that for desired values of lighting performance in different internal spaces, there is often a range of possible panel configurations and costs.
Conference Paper
The paper describes a rank-based fitness assignment method for Multiple Objective Genetic Algorithms (MOGAs). Conventional niche formation methods are extended to this class of multimodal problems and theory for setting the niche size is presented. The fitness assignment method is then modified to allow direct intervention of an external decision maker (DM). Finally, the MOGA is generalised further: the genetic algorithm is seen as the optimizing element of a multiobjective optimization loop,...
Multi-objective facade optimisation for daylighting design using a genetic algorithm
  • J Gagne
  • M Anderson
gagne, J. and Anderson, M.: 2010, Multi-objective facade optimisation for daylighting design using a genetic algorithm, Proc. 4th National Conference, IBPsA-usA simBuild 2010, New york.