Synergy of a Genetic Algorithm and TRIZ in Conceptual Design

Abstract

A design task usually begins with a conceptual design phase where requirements and objectives are defined and synthesized into design alternatives. Although there have been successful accomplishments in developing computational methods and tools in the conceptual design of spatial structures, the initial step of problem structuring needs more consideration. A designer may spend several hours to build a parametric model and explore the suitable solutions through an iterative process of generation, evaluation, and modification of design alternatives. But the parametric model may not initially include some determining variables. The process of exploring design alternatives and choosing the suitable one/s may be more of a challenge when contradictory design objectives exist in a project. In such a case, the designer may give privilege to only one design criterion over the others, or compromise (trade-off) and choose a solution among a group of suitable ones. This study addresses the extent to which TRIZ can contribute to computational early design exploration of spatial structures. The focus of the paper is on presenting a map of the design process which allows a better problem structuring and, also, elimination of the contradiction among design objectives. The proposed model is founded on the Lawson's map of design process. The matrix of contradictions and 40 Inventive Principles of TRIZ, and a genetic algorithm in the framework of ParaGen are incorporated in this model.

Full text

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
July 16-20, 2018, MIT, Boston, USA
Caitlin Mueller, Sigrid Adriaenssens (eds.)
Copyright © 2018 by Anahita KHODADADI
Published by the International Association for Shell and Spatial Structures (IASS) with permission.
Synergy of a Genetic Algorithm and TRIZ in Conceptual Design
Anahita KHODADADI*
*University of Michigan
Ann Arbor, MI, USA
anahitak@umich.edu
Abstract
A design task usually begins with a conceptual design phase where requirements and objectives are
defined and synthesized into design alternatives. Although there have been successful accomplishments
in developing computational methods and tools in the conceptual design of spatial structures, the initial
step of problem structuring needs more consideration. A designer may spend several hours to build a
parametric model and explore the suitable solutions through an iterative process of generation,
evaluation, and modification of design alternatives. But the parametric model may not initially include
some determining variables. The process of exploring design alternatives and choosing the suitable one/s
may be more of a challenge when contradictory design objectives exist in a project. In such a case, the
designer may give privilege to only one design criterion over the others, or compromise (trade-off) and
choose a solution among a group of suitable ones. This study addresses the extent to which TRIZ can
contribute to computational early design exploration of spatial structures. The focus of the paper is on
presenting a map of the design process which allows a better problem structuring and, also, elimination
of the contradiction among design objectives. The proposed model is founded on the Lawson’s map of
design process. The matrix of contradictions and 40 Inventive Principles of TRIZ, and a genetic
algorithm in the framework of ParaGen are incorporated in this model.
Keywords: conceptual design, architecture design, TRIZ, genetic algorithm, parametric modeling.
1. Introduction
Within the conceptual design phase, designers proceed through a course of reasoning, argumentation,
and decision-making [1]. Since designing is too complex to be described by a single plan of work, several
attempts have been made to provide designers different maps of the design process to facilitate
proceeding through it. Lawson describes a general one to be employed for a wide variety of design
problems. It includes problem structuring, generating solutions, evaluating, and making decisions. His
model is similar to that of many others, but his emphasis is on the cyclic relation between every two
steps. He believes that a design problem, unlike a mathematical problem, is not apparent and must be
uncovered and defined. He states that a design problem can become clearer while designers attempt to
solve it. Thus, there is no linear relation between problem structuring and generating solutions [2].
Furthermore, Guilford addresses a divergent approach to generate solutions and a convergent process to
make decisions. Guilford’s concept of divergent and convergent process allows exploring the widest
possible range of alternatives and, also, being able to manage the search space meaningfully. Through
the divergent phase, the designer explores multiple solutions and their unexpected combinations. The
convergent phase may include sorting, filtering, or ranking the several generated solutions and
concentrating on a few suitable ones [3] [4] [5] [6].
In the field of conceptual design of buildings, the concept of divergent and convergent process has been
implemented in several computer-aided design methods. Computational parametric modeling tools,
based on the consistent structure of dependencies within a problem, allow the generation of a population
of solutions conveniently and relatively fast. Furthermore, evolutionary form exploration methods allow
searching the solution space for unexpected alternatives while using a “fitness function” to manage the

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
2
generated solutions. However, the initial step of problem structuring has been less developed in the field
of computational building design. “A wrong problem can be proposed and solved correctly” [7]. “A
problem clearly defined is already half solved” [8]. These are examples of statements indicating on the
significant effect of problem structuring on the final design outcome. Designers may spend several hours
to proceed through the conceptual design phase. Later, they may realize that the solution could have
been found around a design parameter that is not included in the parametric model and form exploration
procedure. Then, they have to disregard what has accomplished previously and make a new parametric
model and redo everything again from the beginning. One may make the argument that the more design
parameters are included through a search process, the better the solutions will be. However, defining a
gigantic parametric model and running more diverse simulations require more time and will be
prohibitive.
In the phase of desision making, choosing suitable solution/s is more of a challenge when contradictory
design objectives exist in a project. In such a case, the designer may give privilege to only one design
criterion over the others, or compromise (trade-off) and choose a solution among a group of suitable
ones [9]. Another approach may be the attempt to eliminate the contradiction innovatively. The Theory
of Inventive Problem Solving (TRIZ), originally introduced by Genrich Altshuller (1926-1998), is
known to support this approach. The main concept of TRIZ can be described in four steps. First, defining
our specific problem. Second, synthesizing a specific problem to a more generic one. Third, learning
from past and finding generic solutions for the generic problem. Finally, interpreting the generic solution
to prescribe some specific solutions for our specific problem. TRIZ includes several tools and concepts
[10] [11]. The design model introduced in this paper incorporates the matrix of contradictions and 40
Inventive Principles of TRIZ. The proposed model is founded on Lawson’s design model and embodied
a genetic algorithm in the framework of ParaGen.
In the following, it is explained how the GA+TRIZ design model is developed. To demonstrate the
contribution of applying this method to the field, designing a folded plate dome has been carried out.
First, the design is accomplished using only the ParaGen method. Second, the GA+TRIZ method is
employed to explore suitable solutions. Finally, the results are discussed and the benefits gained using
the combination of GA and TRIZ are concluded.
2. ParaGen method
ParaGen, originally developed by Peter von Buelow, combines a Non-Destructive Dynamic Population
Genetic Algorithm (NDDP-GA) with a SQL database to provide a directed exploration of a solution
space [12]. It starts by generating an initial population of solutions and follows by an iterative cycle. The
cycle includes steps of selection of a population of parents, a half-uniform crossover (HUX) [13] to
breed new solutions, geometry generation, and evaluation via the objective functions. In each cycle, the
outputs are uploaded and ranked in a database. ParaGen generates a searchable database of solutions
which can be explored interactively for different and multiple combinations of objectives. Further
exploration toward more desired solutions takes place by setting some criteria for breeding new
solutions. The criteria can be set through a SQL query or by Pareto sets [14] [15]. Moreover, the
association of each solution with at least one visual representation, allows the designer to include
subjective judgments in the exploration [16]. Figure 1 demonstrates the parallelism of the ParaGen
method, Lawson’s map of the design process, and Guildford’s concept of convergent and divergent
process.
3. Designing a folded plate dome using ParaGen
As the first case study of this research project, the design alternatives of a folded plate dome have been
explored using the ParaGen method. The dome is considered to be built by cross-laminated timber plates
(CLT). Furthermore, the effects of a variety of geometrical properties on structural efficiency, interior
lighting, and acoustics have been studied. The topology of the dome was inspired by some of Yoshimura
origami patterns. Then, the configuration of a diamatic dome was processed using the concepts of
Formex algebra and its associated programming software, Formian 2.0 [17]. The geometrical,
environmental, and structural properties of this design problem are described in figure 2. The structural
analysis has been carried out based on FEA method using STAAD.Pro. Daylight evaluation is done in

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
3
DIVA, a plugin for Rhino and Grasshopper. Acoustics analysis is based on Sabine’s equation [18]. In
this paper, the process of form exploration and decision making of this design problem is compared with
that of the second case study described in section 5. Further discussions on potentials of CLT-based
structures, and also the effects of geometrical properties of the folded-plate dome on structural and
environmental performances can be reviewed in a paper presented in IASS 2015, Amsterdam [19].
Figure 2 demonstrates the distribution of solutions regarding the reverberation time (RT60) versus the
total weight of the dome. The dome is designed for an exhibition pavilion. Therefore, it requires having
a reverberation time of 2.5s or less. Lower total weight is also preferred. The graph shows that the two
objectives are conflicting. Thus, the designer can choose the preferred solutions exploring the Pareto
sets that are commonly expected to be the best trade-off solutions.
!
Figure 1: Parallelism of Lawson’s map of design process, Guildford’s concept of convergent and divergent process,
and the ParaGen method.
4. The matrix of contradictions and 40 Inventive Principles of TRIZ
A glance at historical or contemporary architecture shows that well-known architects did not merely rely
on their talents. They have been exposed to a large number of problems and solutions. Their experiences
enable them to recognize the underlying principles and influential parameters within a design problem.
Also, they have become able to make preliminary evaluations of their tentative decisions [20]. TRIZ
tools, such as the 40 inventive principles, are derived from the observation and analysis of over 2.5
million inventions. Thus, TRIZ practitioners can benefit from the treasures of the past engineering
experiences. Here, one may argue how TRIZ can contribute to the field of architecture, while its
principles have been primarily developed to assist inventors in proposing patentable technical systems.

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
4
In response, two matters should be addressed. First, the technological aspect of architecture and, second,
the benefits of interdisciplinary exchange of knowledge. A building is assumed to be a technical system
since it performs a function and like other technical systems it is composed of some sub-systems, such
as mechanical and structural systems, and relates to some super-systems such as a neighborhood, region,
and city. Therefore, application of scientific and other organized knowledge is acceptable as it is in other
practical and technological fields.
Figure 2: The geometrical, environmental, structural properties inputs; The distribution of solutions regarding the
reverberation time (RT60) and the total weight.
There are several problem-solving methods such as TRIZ tools, brainstorming, mind mapping, lateral
thinking or morphological analysis. Also, there are several TRIZ tools and concepts such as
contradiction matrix, ideality, and patterns of evolution [9]. In this study, the matrix of contradiction
and 40 inventive principles are chosen to be integrated with the GA-based method of ParaGen because
of two reasons. First, TRIZ is a systematic way of problem-solving while some of the other problem-
solving methods merely provide a list of discovered patterns that one may explore. Alexander’s 253
patterns of problem-solving or 15 geometrical properties are the examples of such lists [21] [22]. Also,
some of the other methods, are less abstract and more applicable in some specific disciplines or contexts.
Second, TRIZ principles are analytical, formulaic and teachable. In particular, the contradiction matrix,
have the potential to be represented in the form of software or a code. There, it is possible to combine it
with another computational design tool and benefit from their synergic effect.
The contradiction matrix addresses the incompatibility of desired features within a system. A designer
should study the pertinent functions, identify the useful functions and harmful ones, notice and uncover
the contradictions among the design parameters. Then, the designer can refer to the matrix of
contradiction where the functions to be improved and those are getting worse are listed along rows and
columns respectively. Within the corresponding cells of the matrix, some of the inventive principles
which may help to resolve the contradictions are proposed [23] [24] [25].
5. Developing the GA+TRIZ design model
Through the process of exploring design alternatives using the ParaGen method, TRIZ can contribute to
defining the design problem or making decisions. Before creating the parametric model, when the design
objectives are determined, TRIZ can come into play. There, it assists to recognize the relation among
design objectives, to see if they are in contradiction or accordance, and to find solutions. First, the
designer should define the demanded useful functions of the design and the harmful functions that are
to be avoided. As the second case study of this research project, the timber folded plate dome is re-
designed using the GA+TRIZ method. Lower weight, reverberation time of less than 2.5s, which can be
an appropriate value for an exhibition pavilion, are the design objectives discussed in this section.

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
5
Second, the designer should synthesize the useful and harmful functions into their basic factors and see
if it is possible to vary those factors. For instance, the designer may look for series of solutions with
lower reverberation time (RT60). Regarding Sabine’s equation, this harmful function is dependent on
the volume of the dome, the sound absorption coefficients of timber and glass plates, and the total area
of the dome surface. Furthermore, each of them may be dependent on their basic factors. For example,
the diameter of the base of the dome is determined to remain constant. Thus, the volume of the dome
depends only on the dome height. Figure 3 shows a graph of synthesizing the two conflicting design
objectives, reverberation time and total weight. The analysis reveals another conflict. Because of the
shape of the folds, when the depth of the folds increases to augment the total area of the surfaces,
drainage of the rainwater becomes complicated on the top part of the dome.
!
Figure 3: The graph of synthesizing the harmful and useful functions and the matrix of contradictions of the dome
project.
In the third step, the designer can refer to the matrix of contradiction to find some generic ideas for
resolving the conflicts. The fourth step is to interpret the generic solutions innovatively and restate
a/some specific solution(s) to resolve the conflicts. For example, the matrix of contradiction suggests
“other way round” to resolve the conflicts between increasing the depth of folds and eliminating the
drainage problem (see figure 3). In other words, the designer may think of creating the folds in a convex
shape and not a concave form. Also, there are four generic solutions to deal with the contradiction
between the desire to increase of the dome surface and lessen its total weight. Among the suggested
solutions, using “flexible thin films or membrane” may be applied by adding acoustic surfaces to control
the quality of sound in the pavilion. Moreover, one may utilize a kind of coating, as thin a film, on the
CLT or glass plates to increase their sound absorption coefficients. In addition, the designer can take the
idea of “separation/ extraction” into account. CLT plates are more massive than glass plates. The
designer may achieve lighter domes with a certain amount of surface area by considering the weight of
the two types of plates separately and reducing the ratio of CLT to glass plates.
6. Discussion
When the designers proceed through the four steps within the phase of problem structuring (see figure
5), they should see if all the pertinent design parameters, which are possible to be considered as variables,
are included. In the second case study, the height of the dome, depth of folds, areas of CLT and glass
plates, weights of CLT and glass plates, the ratio of the CLT plates to glass plates are included in the
database. Also, the designer may re-consider keeping the diameter of the base of the dome constant or
not. Moreover, the designers are recommended to see if their initial design sketches conform with the
design objectives. Hence, the folds of the dome change to be convex (compare solutions shown in figure
2 and figure 4). During the phases of decision making at each cycle, the designers may study the
distribution of solutions in XY graphs and compare certain performative parameters. If they take the
described four steps, they can set the fitness functions and select the breeding criteria with a better
understanding of parameters dependencies. For example, the designer recognizes increasing the depth
of the folds may reduce the reverberation time but it may provide a heavier dome. The ParaGen+TRIZ

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
6
method hints to find the suitable solutions around some other design parameters. Thus, by minimizing
the ratio of CLT to glass plates, it is possible to gain suitable solutions. Also, it becomes obvious that
the lower the height of the dome is, the less reverberation time it may have. Therefore, the designer may
shift the conflicts among the two desirable parameters by bringing some other initial parameters into
play. Then, a diverse series of solutions may emerge.
!!!!!!!!! !
Figure 4: The distribution of solutions regarding the reverberation time (RT60) and the total weight. The values of
some pertinent parameters are displayed in the table.
Figure 5: The plan of work in the GA+TRIZ design method

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
7
7. Conclusion
The GA+TRIZ model has two wings: Matrix of contradictions along with 40 Inventive Principles of
TRIZ, and a genetic algorithm in the framework of ParaGen. Application of a genetic algorithm in the
framework of ParaGen assists designers in generating and studying a wide range of solutions and
expanding their search space. The framework of ParaGen provides a strong interaction with designers
to include their preferences and subjective concerns through the search process. TRIZ is a systematic
way of problem-solving and its principles are analytical. In particular, the contradiction matrix has the
potential to be represented in the form of software or code. Thus, it is possible to combine it with another
computational design tool such as ParaGen and benefit from the advantages of an interactive
evolutionary exploration technique and TRIZ at the same time.
The GA+TRIZ form exploration method includes the four general phases of a design procedure with
additional considerations and actions to take within the phases of problem structuring and decision
making (see figure 5). Taking the four steps of TRIZ allows one to:
- include all the possible pertinent design parameters in the parametric model;
- re-consider maintaining some parameters as constant or variable;
- check the conformity of the initial sketches with the design objectives;
- set the fitness function with a better understanding of parameters’ dependencies;
- overcome the conflicts among the design objectives using some innovative generic solutions,
and bring some other parameters into play to make a suitable final decision.
Acknowledgment
My thanks to Professor Peter von Buelow at the University of Michigan for providing the opportunity
to utilize ParaGen tool as a platform to create the GA+TRIZ design method. Also, I would like to thank
Negin Zarnegar for her contribution in preparing the PHP codes of the TRIZ addition to the ParaGen
website.
References
[1]
A. Rolvink, C. Mueller and J. Coenders, "State on the Art of Computational Tools for Conceptual
Structural Design," in Proceedings of the IASS-SLTE 2014 Symposium "Shells, Membranes and
Spatial Structures: Footprints", Brasilia, 2014.
[2]
B. Lawson, How Designers Think: Demystifying the Design Process, 4th ed., Taylor and Francis,
2005.
[3]
J. Guilford, The nature of human intelligence, New York: McGraw-Hill, 1967.
[4]
N. Cross, Engineering Design Methods: Strategies for Product Design, Chichester: John Wiley &
Sons, 1994.
[5]
S. Pugh, Total Design: Integrated Methods for Successful Product Engineering, Wokingham:
Addison Wesley, 1991.
[6]
Liu Y. C., T. Bligh and A. Chakrabarti, "Towards an ‘ideal’ approach for concept generation,"
Design Studies , vol. 24, no. 4, pp. 341-355, 2003.
[7]
M. Kiatake and J. R. D. Petreche, "A Case Study on the Application of the Theory of Inventive
Problem Solving in Architecture," Architectural Engineering and Design Management, vol. 8, pp.
90-102, 2012.
[8]
D. Hannan, "Value Methodology, Creative Problem Solving Strategies and TRIZ," in Save
International Conference Proceedings, 2000.
[9]
S. Savranksy, Engineering of Creativity—Introduction to TRIZ Methodology of Inventive
Problem Solving, BocaRaton, Florida: CRCPress, 2000.
[10]
V. Souchkov, "A Brief History of TRIZ," 2008.
[11]
M. A. Orloff, Inventive Thinking through TRIZ, A Practical Guide, 2nd ed., Springer, 2006.

Proceedings of the IASS Symposium 2018
Creativity in Structural Design
8
[12]
P. von Buelow, "Genetically enhanced parametric design in the exploration of architectural
solutions," in Structures and Architecture, London, 2016.
[13]
G. Syswerda, "Uniform Crossover in Genetic Algorithms," in Third International Conference on
Genetic Algorithms, 1989.
[14]
P. von Buelow, "Choosing Parents to Produce Better Performing Children: A Comparison of
Selection Methods Used for Evolutionary Search," in Proceedings of the IASS Annual Symposium
2017 “Interfaces: architecture . engineering . science”, Hamburg, Germany, 2017.
[15]
N. Emami, H. Giles and P. von Buelow, "Continuous to Discrete: Computational Performative
Design and Search of Shell Structures," in Proceedings of the IASS Annual Symposium 2017
“Interfaces: architecture . engineering . science”, Hamburg, Germany, 2017.
[16]
P. von Buleow, "ParaGen: Performative Exploration of generative systems," Journal of the
International Association for Shell and Spatial Structures, vol. 53, no. 4, pp. 271-284, 2012.
[17]
H. Nooshin and P. Disney, "Formex Configuration Processing II," in International Journal of
Space Structures, vol. 16, no. 1, 2001.
[18]
W. Sabine, Collected Papers on Acoustics, Cambridge : Harvard University Press, 1922.
[19]
A. Falk, P. von Buelow and A. Khodadadi, "Form Exploration of Timber-based Folded Plate
Domes," in International Association for Shell and Spatial Structures (IASS) Symposium,
Amsterdam, 2015.
[20]
N. Cross, "Expertise in design: an overview," Design Studies, vol. 25, no. 5, p. 427–441, 2004.
[21]
C. Alexander, The Nature of Order: The phenomenon of life, Center for Environmental Structure,
2002.
[22]
C. Alexander, The Timeless Way of Building, New York: Oxford University Press, 1979.
[23]
M. A. Orloff, Inventive Thinking through TRIZ, A Practical Guide, 2nd ed., Springer, 2006.
[24]
G. Altshuller, The Innovation Algorithm: TRIZ, Systematic Innovation and Technical Creativity,
Technical Innovation Center, Inc., 1999.
[25]
G. Altshuller, L. Shulyak and S. Rodman, 40 Principles: TRIZ Keys to Innovation, Technical
Innovation Center, Inc., 1997.
[26]
S. Adriaenssens, P. Block, D. Veenendaal and C. Williams, Shell Structures for Architecture:
Form Finding and Optimization, Routledge, 2014.
[27]
J. Chilton, "Heinz Isler's Infinite Spectrum: Form-Finding in Design," Architectural Design, vol.
80, no. 4, pp. 64-71, 2010.
[28]
A. G. Tibert and S. Pellegrino, "Review of Form-Finding Methods for Tensegrity Structures,"
International Journal of Space Structures, vol. 26, no. 3, pp. 241-255, 2011.
[29]
M. Turrin, P. von Buelow and R. Stouffs, "Design Explorations of Performance Driven Geometry
in Architectural Design Using Parametric Modeling and Genetic Algorithms," Advanced
Engineering Informatics, vol. 25, p. 656–675, 2011.
[30]
J. de Chiara and M. Crosbie, Time-Saver Standards for Building Types, 4th ed., McGraw-Hill
Professional Publishing, 2012.
[31]
G. Syswerda, "Uniform Crossover in Genetic Algorithms," in Third International Conference on
Genetic Algorithms, 1989.
[32]
A. Khodadadi and P. von Buelow, "Performance Based Exploration Of Generative Design
Solutions Using Formex Algebra," International Journal of Architectural Computing, vol. 12, no.
3, pp. 321-338, 2014.
[33]
O. Oliyan Torghabehi and P. von Buelow, "Performance oriented generative design of structural
double skin facades inspired by cell morphologies," in IASS-SLTE 2014 Symp. Shells, Membranes
and Spatial Structures, Brazilia, 2014.