articulo 10 adaptabilidad
DESCRIPTION
asdfsfTRANSCRIPT
-
This article was downloaded by: [Universidad Del Rosario]On: 25 July 2014, At: 07:39Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Construction Management and EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rcme20
Evaluating the adaptability of an industrializedbuilding using dependency structure matricesRobert Schmidt IIIa, Kasper Sanchez Vibaekb & Simon Austinaa School of Civil and Building Engineering, Loughborough University, Loughborough,Leicestershire LE11 3BQ, UKb Centre for Industrialised Architecture, The Royal Danish Academy of Fine Arts,Copenhagen, DenmarkPublished online: 20 Mar 2014.
To cite this article: Robert Schmidt III, Kasper Sanchez Vibaek & Simon Austin (2014) Evaluating the adaptability of anindustrialized building using dependency structure matrices, Construction Management and Economics, 32:1-2, 160-182, DOI:10.1080/01446193.2013.847274
To link to this article: http://dx.doi.org/10.1080/01446193.2013.847274
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
-
Evaluating the adaptability of an industrialized building usingdependency structure matrices
ROBERT SCHMIDT III1*, KASPER SANCHEZ VIBAEK2 and SIMON AUSTIN1
1School of Civil and Building Engineering, Loughborough University, Loughborough, Leicestershire LE11 3BQ, UK2Centre for Industrialised Architecture, The Royal Danish Academy of Fine Arts, Copenhagen, Denmark
Received 9 December 2012; accepted 17 September 2013
A relatively neglected aspect of sustainable development is the creation of an enduring built environment that can
be adapted to suit changing circumstances. This presents a significant challenge: how to evaluate a buildings
adaptability. The premise is introduced that adaptability is enhanced through the use of analytical tools which
can provide better control of the buildings system architecture that enables easier, and less costly, user-driven
adaptations. More specifically it investigates what a dependency structure matrix (DSM), a square (N N)
matrix-based model that visualizes the relationships between elements within a system, can reveal about the
capacity for an industrialized building to accommodate change, through clustering and impact analyses.
Clustering analysis attempts to assess the system architecture on generic principles of change by organizing it into
discrete modules, while the impact analysis examines propagation in 30 change scenarios by tracing dependencies
within the DSM. Feasibility assessments of the scenarios are compared with the knowledge of a system expert.
The results indicate the DSM analysis provided insights beyond the intuition of the system expert regarding
change propagation, while the system experts knowledge of component characteristics and overall composition
of the building proved beyond the capacity of the DSM. Additional conclusions are drawn from the case study
regarding DSM construction and the analytical process.
Keywords: Adaptability, building design, change management, industrialized building, sustainable construction.
Introduction
The transition of parts of the construction industry
from craft-based professions to producers of industrial-
ized products has been ongoing since the mid-twentieth
century. While providing clear benefits, this movement
has been met with significant obstacles and remains an
elusive goal for many. As the industry focuses its atten-
tion on creating a more sustainable built environment,
for which adaptability along with durability play a key
role in prolonging the longevity of our built assets
(Graham, 2005), industrialized methods of construc-
tion provide a rational alternative to traditional meth-
ods. The Cellophane House is an exemplar of anindustrialized building designed to be built primarily
off site, assembled quickly on site, disassembled and
moved to another location. It was designed with a
system architecture intended to fulfil a variety of client
and site demands by easily generating an array of
configurations (mass customization). On the other
hand, how the building could adapt to changing needs
while in operation wasnt overtly considered, e.g. can the
user move an internal partition or change an external
wall panel?
Buildings are complex objects constructed from
parts and components with varying service lives that
demand design and construction strategies to militate
against the cost and time of accommodating change
(cf. Brand, 1994). As a starting point, an industrialized
building would seem to reduce the adaptability of a
building (i.e. you cant implement the ad hoc kind of
changes that a bespoke solution may allow for). We
suggest, however, that systematic deployment of
*Author for correspondence. E-mail: [email protected]
2014 Taylor & Francis
Construction Management and Economics, 2014Vol. 32, Nos. 12, 160182, http://dx.doi.org/10.1080/01446193.2013.847274
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
well-defined and documented integrated product
deliveries (an industrialized building) can reduce the
complexity of the design task while at the same time
enhancing design control when changes occur. Follow-
ing Mikkelsen et al. (2005, p. 3) an integrated product
can be defined as a multi-technological complex part
of a building that can be configured and customized
to a specific construction project, produced as a sepa-
rate product process and when delivered to a customer
becomes an integrated product delivery (IPD). Thus,
the hypothesis set forth is that adaptability, namely
the capacity to accommodate change (cf. Schmidt III
et al., 2010), can be enhanced through such increased
control over the system architecture, and where the
application of analytical tools can further enhance the
capacity to disassemble (partially or completely)
the building as discrete integrated products. Two
research questions are explored:
(1) What can a product DSM model reveal about
the capacity for an industrialized building to
accommodate change? and subsequently,
(2) What can a worked example tell us about
constructing a product DSM model and the
analytical techniques used?
In response to these questions, a succinct review of
the industrialization of buildings is presented and sets
forth concepts and approaches from product architec-
ture and construction literature which coalesce as part
of the analysis. The research methodology is explained
and two models for visualizing system architectures are
introduced: system structures (SSs) and dependency
structure matrices (DSMs). Both models visualize the
Cellophane House with the former supporting theelaboration of the latter. Two types of analysis were
carried out on the product DSM produced: clustering
and impact analysis. The first analysis assesses the
system on generic principles of change by identifying
and isolating functional modules. Meanwhile the sec-
ond analysis examines change propagation for various
components with specific change scenarios by tracing
dependencies within the DSM. A critical evaluation
of the two analyses and the system architecture allows
conclusions to be drawn regarding both analyses them-
selves and the capacity for the specific building solution
to accommodate change.
Industrialization of buildings
Since the early descriptions and theories of first Adam
Smith and later Frederick Taylor (Smith, 1776; Taylor,
1911), a pronounced division of labour and ensuing
industrialization have spread to all areas of society.
The construction industry is no exception, and there
have been several attempts at industrialization in the
industry during the twentieth century; however, they
have never led to industrialization of the way buildings
(in their entirety) are produced in the proper sense of
the word. Standardized buildings are still erected in
fairly traditional ways following divisions along those
of the old crafts and true industrialized production is
only found on the scale of building materials and to
some extent at the building component and subsystem
levels. As pointed out by Bergdoll and Christensen
(2008), on the one hand prefabrication (the prepara-
tion of buildings off site) can be traced back to antiquity
(e.g. a Mediterranean shipwreck containing structural
members for an entire classical temple), while on the
other hand, a culture of prefabrication arises with
modernity in the early twentieth century and is born
from the union of architecture and industry (ibid.,
p. 12). While this might be the ideological birth of what
can be called an industrialized building the maturation
of its physical manifestation, a truly industrially pro-
duced building, is still underway.
It was not until the massive industrialization
attempts of building processes and products in the
1960s that the division between the crafts and profes-
sions on the one hand and the modularization of con-
struction on the other evolved to partially break the
traditional craft segregations. Previously, the building
crafts could be seen as independent systems of coherent
expert knowledge with defined interfaces. Today, the
crafts and construction skills have almost disappeared
from the construction industry in their traditional form
due to increased technical and economic demands.
Large standardized quantities, extreme technical preci-
sion and a need for increased productivity with less
manpower dissolve the essentials of a traditional
Figure 1 Cellophane House, designed by KieranTimber-lake (left photo by Peter Aaron/OTTO; right photo by Albert
Vecerka/Esto)
Evaluating adaptability 161
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
manually based production and onsite adaptation. At
the same time, the explosion in the number of choices
within the building materials industry has made it
impossible for anyone to cope with all the possible
combinations in a traditional non-explicit (tacit) man-
ner (cf. Alexander, 1964; Utida, 2002). Although the
fundamental challenge remains relatively unchanged,
the premise for solving this task has evolved consider-
ably as buildings have become much more complex
both as objects (material) and design tasks (process)
(Alexander, 1964). If we assume that industrialization
is (a) a condition (not just an option) that architects
and other stakeholders in the industry have to respond
to; and (b) a means not a goal in itself, then we have
framed what an industrialized building is about. The
concept can be clarified further through the introduc-
tion of some terminology from the product industry.
Product architecture
According to Ulrich and Eppinger (2008, p. 165)
product architecture is the assignment of the
functional elements of a product to the physical build-
ing blocks of the product. The result of this assign-
ment, the product architecture, makes up the
structural organization of constituent elements of the
product. For Ulrich and Eppinger any product archi-
tecture is function-wise a trade-off between modular-
ization (isolated chunks) and integration which are
considered opposites in the literature. In building pro-
jects the direct transfer of the concept is problematic
for several reasons. First, the use of architecture gets
an inappropriate double meaning referring to both the
structural organization of elements (as above) as well
as a holistic notion of a built structure with artistic
intention being more than just the sum of its parts (or
products). While system structure (SS) is suggested later
as an alternative denomination system architecture is
generally used throughout when referring to the (holis-
tic) building level rather than the (sub) product level.
Secondly, industrial products usually have a limited
number of well-defined (technical) functions while, in
buildings, the issue of modularization vs. integration
becomes blurred due to the fact that buildings are pri-
marily frames around liveable space and only secondar-
ily a coupling of discrete functional elements. Thus,
functions can only to a limited extent be related directly
to physical elements. Finally, in terms of industrializa-
tion, the relative decoupling of physical and functional
elements combined with the size of buildings produces
a paradox between the need to reduce chunk sizes into
manageable dimensions and the advantage of combin-
ing chunks with functional purposes, thus reducing
inter-modular complexity and interfaces.
According to Baldwin and Clark (2000), who deal
specifically with the impacts of modularization in the
product industry, integration requires a high degree of
overall design coordination for each specific develop-
ment of a product, whereas modularization, in the
sense of isolating discrete functions or systems within
chunks, makes it possible to change pieces of a system
without redoing the whole. Design becomes flexible
and capable of evolving at the module level (ibid.,
p. 6). For Baldwin and Clark modularity cannot be
limited to the physical structure of the product; it
integrates process and organization.
The appearance of modular designs, a process that
according to Baldwin and Clark (2000, p. 16) began
around 1970, has led to the forming of modular clusters
which are group[s] of firms and markets that play
host to the evolution of a set of modular designs. If
certain modules and their particular interface defini-
tions become sufficiently established, industry will
adapt to and emerge around them. In the construction
industry, this effect is currently mainly manifested at a
relatively simple component level (e.g. bricks, win-
dows). There seems to be little incentive to specialize
in more complex building products in the form of
clearly delimited, discrete modules (Beim et al.,
2010). Thus adaptability is consequently achieved
through bespoke solutions on the project level rather
than through industrialized modular design and inte-
gration on a product level.
Many products based on modular principles are
organized hierarchically involving a combination of a
basic structure of common components (a core tech-
nology) plus modules that are added, attached or
inserted into this structure. This basic structure is often
called a product platform, while the range of product
configurations including the added modules constitutes
a product family. Meyer and Lehnerd (1997, p. 16)
define a product family as a set of products that share
Figure 2 Chunks assembled on site (left photo by Albert
Vecerka/Esto; right photo by KieranTimberlake)
162 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
a common technology and address a related set of mar-
ket applications. In construction a structural frame
could be used as a product platform. In the case study,
the Cellophane House, the structural aluminium framesystem was originally developed for production line
environments but has been modified to serve as a
multi-storey system. It is rare, however, to see such
highly industrialized platforms in construction. Most
offsite produced elements are either unstandardized
and labour intensive or standardized with subsequent
labour intensive adaptation on site. This suggests that
industrialized product platforms used in construction
represent a potential for enhanced quality control of
new products. Furthermore, the concept of mass
customization theoretically bridges the value of unique
(customized) one-off solutions for specific customers
with the efficiency of industrialized (standardized) mass
production (Pine, 1993). Mass customization in con-
struction may seem to provide the missing link for a
truly industrialized building: unique context-sensitive
solutions based on an efficient industrialized production
apparatus. The concepts defined within the product
architecture literature of modularity, integration,
product platform and mass customization provide a
theoretical foundation from which an approach for
analysing a buildings capacity to adapt can be built.
Measuring adaptability
Approaches in product architecture, as defined above,
are often concerned with modularity and product
platforms stressing the connections between elements
(cf. Engel and Browning, 2008; Li et al., 2008). Fricke
and Shultz (2005) discuss changeability across phases
of the product life cycle (design, manufacture and
operation), products and product families. Such
inter-phase preoccupations are typical of the literature
and its concern with design and production change,
where products evolve within the same organization
(product versions). Buildings, however, are generally
one-offs, changed in use with no second product
release. Hence the focus here is on attempting to
prolong the use/operational life cycle phase rather than
between iterations of design and production.
Within the product architecture literature, adapt-
able design is considered a relatively new discipline
(Fletcher et al., 2009) in that the principles are superfi-
cially understood, and product architecture lacks the
theory and tools for applying the principles and evalu-
ating success (Li et al., 2008; Saleh et al., 2009). As
mentioned in the introduction, difficulty in measuring
adaptability arises because, unlike other design charac-
teristics, adaptability is not observable under normal
operating conditions (Saleh et al., 2009, p. 313). Both
Olewnik and Lewis (2006) and Ross et al. (2008) sug-
gest metrics for flexibility that rely on the specification
of possible changes from the outset and are analogous
to Saleh et al.s (2009, p. 309) definition of flexibility
as the number of remaining alternatives after a first
commitment is made. They suggest that its possible
to define all possible changes at the project outset,
which has obvious limitations for long-life, complex
and contextually sensitive artefacts as buildings. While
their method has merit in short-term considerations,
its unsuitable for the extended periods and unpredict-
able situations considered here. Hashemian (2005)
expands this metric by proposing two typologies: spe-
cific adaptability which covers the foreseeable changes
and general adaptability which facilitates unforeseen
change. Hashemians (2005) assessment of general
adaptability is based on segmentation which is character-
ized by decomposing the system into discrete (autono-
mous) and functional modules, consistent with the
product architecture literature. The work, while pro-
posing a novel system (built upon by Li et al., 2008)
for measuring adaptability, focuses on the cost savings
of developing an adaptable product vs. multiple prod-
ucts which limits its applicability to the design phase
context. In summary, these evaluation techniques for
adaptable design in product architecture dont provide
objective, quantifiable techniques applicable to the
scale, complexity and lifespan of buildings, nor do they
reveal much about how to improve adaptability.
In contrast, much of the construction literature
concentrates on defining critical physical and spatial
parameters (e.g. storey height, plan depth, grid spacing)
in the form of design guides (cf. Graham, 2005;
Canadian Standards Association, 2006). Modularity is
often mentioned for specific elements, e.g. use
Figure 3 Cellophane House structural frame made withBosch Rextroth extruded aluminium members (photo by
KieranTimberlake)
Evaluating adaptability 163
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
modular components (Iselin and Lamer, 1993) or design
an additive structure using modules or lattices (Lynch,
1958), but rarely expanded upon in terms of achieving
it. While providing a helpful checklist this strand of
the literature doesnt provide suitable metrics. A second
branch of the literature applies a more systematic
approach to evaluating a buildings capacity to adapt
(cf. Kincaid, 2002; Geraedts and de Vrij, 2004). One
example is Larssen and Bjorbergs (2004) software,
which systematically evaluates the technical capacity
of adaptability with regard to flexibility (functionally),
generality (spatially) and elasticity (overall size). The
software associates physical parameters (e.g. structural
span, heating capacity, building size) with the three
above-mentioned strategies and rates them 0 to 3 with
0 for highly dynamic (good adaptability) and 3 for very
static (poor adaptability). While the tool attempts an
objective evaluation by providing quantifiable parame-
ter descriptions, it provides no insights for a path of
action. In summary, these approaches score building
characteristics to produce a ranking; none, however,
targets the system holistically and in particular the rela-
tionships between components to understand how
change would propagate as a result of general change
principles and more specifically change scenarios.
A third and last strand within the construction liter-
ature considers the effects of change by decomposing
the building into discrete chunks, e.g. levels, layers or
subsystems. The categorization enables a hierarchical
structure between elements to be defined emphasizing
their (a) composition and (b) relationships (cf. Fried-
man, 2002). First, concerning composition, Schmidt
III et al. (2011) present several of the approaches to
illustrate various interpretations of how a building
could be decomposed. For example, the layers
approach separates the building as a set of shearing
layers that change at different rates (Figure 4); the more
layers are connected, the greater difficulty and cost of
adaptation (Brand, 1994). On the other hand, the
subsystems approach focuses explicitly on identifying
distinct functions as a method for stratification,
whereas the layers approach blends functionality with
a specific concern for differing component life cycles
(cf. Brand, 1994). The level concept attempts to
balance the physical with the social understanding that
one cannot achieve adaptability without both
(Habraken, 2008). Here decomposition is based
primarily on separating levels of ownership to enable
individual control.
The second aspect of this conceptualization is
concerned with the types of relationships that can occur
between discrete chunks. As an example, Rush (1986)
defines five physical/spatial connection types: remote,
touching, connected, meshed and unified. Schmidt III
et al. (2011) compare typologies and suggest three dis-
tinct types applicable to buildings: (1) structural (e.g.
gravitational, lateral); (2) spatial (e.g. adjacency, circu-
lation); and (3) service (e.g. energy, water). This third
strand within the construction literature concentrates
on the composition and relationships between compo-
nents which is beneficial for our analysis as it provides
a way of conceptualizing the building relative to change
and its system architecture. The layers analogy is help-
ful when analysing change because it sets rules regard-
ing relationships, similar to software design, whereby
relationships between predefined modules of code are
eliminated by software which can iteratively parse the
code for rule-breakers (Sangal and Waldman, 2005).
While layers (construction) and modules (product
architecture) should not be confused as being the same,
the two concepts are used harmoniously here as an
organizing (layers) and analytical method (modules).
Research method
Research design
The aim was to understand what a product DSM
model can reveal about the ability to control the adapt-
ability of an industrialized building, establishing an
inductive enquiry in which we sought analytical tech-
niques to examine a systems capacity to accommodate
change. A single case study was chosen given the
unique nature of the project and the importance of
providing a richer and more nuanced understanding
of it (Yin, 2003). While selecting a single case study
has limitations, e.g. it may not turn out as foreseen,
the increased chance of misrepresentation (Yin,
2003), the chosen buildings extreme level of industri-
alization at the building scale, combined with excellent
access to product personnel and documents provided
the opportunity for novel insights. The Cellophane
House (CH) is an exemplar industrialized buildingwhere it was explicitly sought to reorganize the produc-
tion and product conditions compared to conventional
construction and therefore is well suited to a discussion
of industrialization.
STUFF
SERVICES
SKIN
SITE Eternal
30 - 300 years
7-15 years
3-30 years
20 years
1 day - 1 month
STRUCTURE
SPACE PLAN
Figure 4 Building as a set of shearing layers (Brand, 1994)
164 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Cellophane HouseTM: the case
The chosen case was designed by Kieran Timberlake
Architects and the result of a competition held in 2007
by MoMA (the Museum of Modern Art in New York)
as a part of the exhibition Home Delivery: Fabricating
the Modern Dwelling in 2008 (Bergdoll and Christensen,
2008). The unique context of the competition allowed
for a more self-contained product reducing many of
the complexities found in a conventional building. Key
to the design concept was transparency, lightness, and
mass customization of a standardized (structural) prod-
uct platform through the use of standardized infill
products. The majority of the building was produced
off site as volumetric elements in New Jersey and came
to New York on trucks. The building was designed for
disassembly (DfD) through discrete industrialized
products joined primarily by dry connections.
One of the authors collected considerable
documentation (drawings, photos, interviews) while
undertaking earlier research co-located with the archi-
tects, adding contextual knowledge not present in the
actual data. While the design life of the Cellophane
House may be much shorter than that of a conven-tional building, the high degree of industrialization at
the building scale is a unique feature offering the
opportunity to uncover original findings regarding the
relationship between industrialization and adaptability
(Proverbs and Gameson, 2008). Furthermore, being
just a full-scale prototype makes the system architecture
relatively simple comparatively, while still being
sophisticated from the point of view of an industrialized
building, providing a clearer analysis and enabling a
more specific focus and discussion.
System structure: initial model
Related to the concept of product architecture, the
notion of system structure in architecture proffered by
Vibaek (2011) conceptualizes systemic levels that lie
between general construction techniques and specific
building results. The system structure becomes opera-
tional through the elaboration of a system structure model
that seeks on the one hand to analytically grasp and on
the other hand to make it possible to actively work with
system structure scenarios as part of building design.
System structures should be understood as abstract
(system) representations of buildings focusing on the
way buildings are put together as combinations of
thought (ideas), process and matter (materials/prod-
ucts). They are particularly, but not exclusively, suited
for industrially produced buildings with varying degrees
of offsite processes or prefabrication. The basic system
entity in a system structure is the delivery which closely
relates to, while simultaneously seeking to merge, the
two concepts from the product industry of product
architecture (physical) and supply chain (procedural).
More specifically, the model visualizes a system
structure as chains of deliveries with different degrees
of integrated complexity, which can be understood
through delivery tiers: spanning raw materials (tier 5),
building materials and standard components (tier 4), sub-
assemblies and system components (tier 3), assemblies (tier
2), building chunks (tier 1) and ending in the building (tier
0) (see Figure 5). A lower tier numbermeans higher inte-
gration in complex deliveries, while a higher tier number
means lower integration and more simple deliveries.
Simpler deliveries can be nested into more integrated
(and complex) deliveries (IPDs) such as sub-assemblies,
assemblies or even entire building chunks before reach-
ing the final building. A building can thus be decom-
posed into its (more or less integrated) systems as they
are actually produced and delivered, just as it can be
decomposed into its spaces or building elements. The
system structure of Cellophane House was the basisfor constructing the subsequent product DSM model.
Matrix-based product modelling
As Alexander (1964, p. 5) stated in his seminal book,
the intuitive resolution of contemporary design prob-
lems simply lies beyond a single individuals integrative
grasp. System architecture decomposes the problem as
a whole into a hierarchical structure of functional sub-
systems which can be looked at individually at a lower
level and cohesively at a higher level (Eppinger and
Browning, 2012). Managing the complexity of engi-
neered systems can be considered the role of a system
architect; according to Rechtin (1991, p. 1) the
essence of architecting is structuring. There are several
ways to visualize complex systems beyond conventional
2D/3D drawings, one architecting tool being a matrix.
Matrix-based modelling has been deployed successfully
in a range of applications including product modular-
ization (Sharman and Yassine, 2004) and change
impact analysis (Clarkson et al., 2004) as it provides a
concise way of visualizing a system (Malmqvist, 2002)
and lends itself to computational analysis through
sequencing or clustering, sometimes with dedicated
algorithms (Eppinger and Browning, 2012).
DSMs explained
Several product matrix modelling methods exist, which
model a variety of relationships between elements of
different domains (e.g. axiomatic design, quality
function deployment). A dependency structure matrix
(DSM) is a square N N matrix that highlights
relationships between elements within a single domain.
DSMs can be static (product, organization) or
Evaluating adaptability 165
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
1. Concrete
Building materials andstandard components
Sub-assemblies andsystem components
Assemblies(IPDs by system)
1. Aluminum Extrusions +steel connectors
1. Curtain wall panels +door frames
Chunks(IPDs by zone)
Building(tiers nested on site)
2. BM
2. BM3. OTS
3. OTS4. ICI Paints5. Steel connectors6. T1, Kullman
4. 3form5. Flooring6. T1, Kullman + T0 Team
6. T1, Kullman + T0 Team
6. T1, Kullman + T0 Team
1. Polycarbonate plates
2. BM
2. BM3. C2F4. Burgess Steel5. Walways, balconies, roof
1. Aluminium Grate
2. BM
3. OTS
3. OTS4. Burnett5. Various
4. 3M
6. T1, Kullman
6. T1, Kullman
6. T1, Kullman
1. Bolts and fasteners
5. Smart wrap facade panels
1. Interior shading2. BM
2. BM
3. OTS4. CPI Daylighting5. Roof and Canopy
2. BM
6. T1, Kullman
3. OTS4. Total Plastics5. Stairs and drain pans
2. BM1. Acrylic panels
6. T2, Kullman
6. T2, Kullman3. OTS
3. OTS4. Duravit5. Bathrooms 5. Bathrooms
2. KOP
2. KOP3. CM
3. CM2. KOP (17 units)1. Chunks
3. CM2. ASM (of BM)
4. Sciame5. MoMa-site6. N/A
6. N/A
4. Sciame5. MoMa-site
1. Chunk Assembly
1. Final fit- out (partition walls+ glazing + int. Sm-panels)
6. N/A
1. Foundation
3. CM
3. CM3. ASM (of KOP)
4. Capital Plastics5. Interior stairway
1. Staircase
3. M2O/CM?
3. M2O2. ASM (of KOP)1. Kitchen installation
4. Valcucine
4. Offsite Solutions/Kullman
1. Bathroom pods2. COM
2. COM3. CM4. Maspeth Welding5. Structural frame
1. Steel connectors
2. COM3. CM/M2O4. Berkowitz
6. T0, Sciame/sub5. Curtain wall and N-facade
1. Insulated glass units
2. COM
4. Greenheck/Del Ren
1. Fixtures
1. Ventilation fans + louvers
6. T1, Kullman 6. T1, Kullman
6. T1, Kullman
2. ASM
4. Universal Services Ass.
1. Smart WrapTM facade panels
3. CM
5. E+W Facades
4. Kullman5. Cellophane House
5. Cellophane House
6. T0, Sciame
6. T1, Kullman + T0 Team5. Walways, balconies, roof4. Team (Science, Kullman, KT)
2. Team (of ASM+COM)
6. T1, Kullman + T0 Team
3. OTS
5. Ventilation shaft
6. T1, Kullman5. Interior + exterior
3. OTS2. KOP
2. KOP3. CM/M2O4. Schuco5. curtain wall and N-facade
4. Philips
1. Electrical fixtures
6. T1, Kullman
6. T1, Kullman
4. 3form5. Interior partitions
2. BM1. Interior wall panels
6. T2, Universal Services Ass.
6. T2, Universal Services Ass.
6. T2, Universal Services Ass.
3. OTS4. Dupont Teijin
1. Photovoltaic film
5. Smart wrap facade panels
2. BM1. PET film
3. OTS
3. OTS
5. Kitchen6. T0, Valcucine
5. Kitchen
1. Kitchen cabinets2. KOP3. M2O4. Valcucine
6. T0, Valcucine
4. Miele/Valcucine
1. Appliances2. COM/KOP
3. OTS
5. Structural frame6. T2, USA + T1/T2, Kullman
2. COM/KOP
4. Bosch/Airline Hydraulics
4. Power flim5. Smart wrap facade panels
5. Smart wrap facade panels
2. BM
3. OTS4. Manufacturer?
2. BM1. Copper tape
1. Polypropylene plates
1. Exterior paint
3. OTS4. 3M5. Various
1. Double sided tape
3. M2O4. Sciame/sub5. Foundation6. T0, Sciame
T4
T3
T2
T1
T0
2. COM3. OTS
5. Bathrooms4. AFNY
1. Plumbing accesories
6. T1, Kullman
1. Partition walls2. ASM3. CM4. Kullman5. Interior6. T1, Kullman + T0 Team
Figure 5 System structure model (Vibaek, 2011)
166 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
time-based (activity) (Browning, 2001). Static DSMs
are optimized through clustering, while the latter repre-
sent a temporal flow or process architecture and are
optimized through sequencing (Eppinger and Brown-
ing, 2012). Clustering involves rearranging elements
into chunks or modules which have a high amount of
interactions internally and few interactions externally
(Browning, 2001). Another strategic manoeuvre is to
isolate elements that have high interactions across sev-
eral chunks as bus or integrating components (Sharman
and Yassine, 2004). Alternatively, sequencing or parti-
tioning orders activities into a logical sequence identify-
ing sequential, parallel, coupled and conditional
relationships between tasks (Austin et al., 2000). DSMs
represent a single domain; however, as Eppinger and
Browning (2012) point out complex projects are often
a collection of inter-related complex systems, each with
its own architecture and thus, multi-domain matrices
(MDMs) help explore cross-domain effects. An
MDM is made up of a series of DSMs (along the diag-
onal) and a series of domain mapping matrices
(DMMs) either side, the label given to matrices which
map the domain of one DSM to that of another. Danil-
ovic and Browning (2007) propose a periodic table of
DSMs using Browning et al.s (2006) five domains
(goals, product, process, organization and tools).
MDMs and DMMs extend DSMs beyond a single
domain and create a framework for project architecting.
Most DSMs are binary but several authors have
proposed numerical DSMs which capture additional
attributes of the system applying numerical values, col-
our, or additional symbols to indicate the importance,
strength or type of interaction (Eppinger and Brown-
ing, 2012). Pimmler and Eppinger (1994) found that
identification of relationship types, while not always
used for analysis, helps gather and verify relationships
between elements. Once constructed, matrices can be
manipulated manually by the modeller or automatically
using a clustering algorithm. The danger of manual
clustering, which Sharman and Yassine (2004) point
out, is the ease with which different cluster boundaries
can be identified making the decision somewhat arbi-
trary and in need of some form of automated process.
However, manual clustering does offer several benefits
to the modeller, bringing into play their tacit knowl-
edge, but requires a systematic process to ensure logical
rules are followed. Sharman and Yassine (2004) define
a vocabulary to describe characteristics displayed in a
product DSM. For example, pinning refers to a
component which overlaps or is pinned in place between
compound elements (two modules). In addition, com-
ponents can be held away from each other when they
are part of a series of dependencies in which a linking
component holds one component away from another.
For automated clustering Loomeo v2.5 uses a
spectral clustering algorithm based upon Laplacian
matrices (Luxburg, 2007). There are three types, two
of which are normalized and one unnormalized. Loo-
meo offers a choice of such algorithms and the number
of clusters sought. In model-based clustering, where
knowledge of the system is known, there are a number
of well-justified approaches to selecting the number of
clusters based on log-likelihood of the data, whereas,
if few assumptions can be made about the system a
large variety of indices are available from ad hoc
measures (ratio of within-cluster and between-cluster
similarities) to stability approaches (Luxburg, 2007).
In addition to clustering, there are a growing num-
ber of studies using change propagation or impact anal-
ysis with DSM product models. The concept of
following dependencies within a DSM to assess the
impact of change is rooted in DSMs process origins
(cf. Steward, 1981). More recently, Eckert et al.
(2004) applied the concept to a product DSM to classify
components based on their behaviour during change
events, quantifying the number of changes a component
absorbed (dependencies in) against the number of
changes propagated (dependencies out). The work later
goes on to include a probability or risk factor associated
with each change by including a degree of likelihood the
change would occur (Giffin et al., 2009). The classifica-
tion system has recently been applied to a retail project
in the construction industry and proven applicable
despite its limitations to address the cost magnitude of
change (Grinnell et al., 2012).
DSM was chosen as the matrix technique because it
provides a narrower focus than other techniques which
may or may not be optimal for inter-related systems,
but can be expanded based on need (DMMs, MDMs).
As Eppinger and Browning (2012, p. 235) suggest,
Maintaining the distinctions between systems enables
focused modelling and the generation of insights that
might not have been as apparent otherwise. Addition-
ally, its worth noting that while DSMmodels are quan-
tifiable and relatively objective, the techniques are
subject to the modellers personal experiences and
understanding of the process and system. The method
relies on thorough validation of dependencies, compli-
cated by different ways (and degrees to which) elements
can depend on another. Thus, measures were put in
place to verify the modellers understanding of the sys-
tem through checks with system and modelling experts.
Analytical method
A product DSM was created for the Cellophane
House. As an initial step, the SS model was used toadvise the decomposition of the system into a product
breakdown structure (PBS). The SS model proved
Evaluating adaptability 167
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
extremely useful to grasp the hierarchical composition
of components in relation to their production and
assembly processes beyond the typical 2D/3D drawing
and helped inform the logics behind the PBS and the
chosen granularity visualized. However the SS model
did not present a full compositional understanding
and was supplemented where needed with building
drawings and photographs. Components were
organized in the DSM as building layers (e.g. skin)
and relationships were identified as spatial, structural
or service. Descriptions for each dependency were cap-
tured in the cells as a record. A discussion was then
held to verify the DSM with the system expert regard-
ing the granularity of elements, layer categorization,
and the dependencies between elements. A short list
of follow-up questions was generated for the architect
and the DSM revised accordingly. Lastly, the DSM
was verified with a DSM expert that resulted in some
duplication of structural dependencies across connect-
ing components being removed. The sophisticated
dependency typology is subsequently flattened upon
importation to Loomeo. A combination of manual
and automated clustering was undertaken in Loomeo
v2.5 to obtain the final matrix, as explained in the
results section.
Regarding impact analysis, 30 scenarios were
selected from the Adaptable Futures change scenario
database (Adaptable Futures, 2011) representing a
mixture of change strategies and actions that were more
likely to occur. An impact analysis was carried out on
each of the scenarios as follows:
(1) Identify component(s) which would be
affected.
(2) Trace the row of the component that is
identified, highlighting the horizontal compo-
nents dependencies (see Figure 9).
(3) Assess each dependency regarding the affect
the change of the horizontal component would
have on the vertical component.
(4) If the vertical component is physically affected
its highlighted and its row is assessed in a
subsequent iteration (e.g. round 2).
(5) Steps 2 to 4 are repeated until propagation
ends.
A feasibility rating for each of the scenarios was then
assigned based on the number of components affected
via the propagation and the nature of those changes
(e.g. amount of work and cost). A simple three-level
scale was adopted identifying whether the scenario
was feasible, somewhat feasible or not feasible.
In parallel to the DSM impact analysis, assessment
of the scenarios was carried out by the system expert
(an architect with extensive knowledge of the system)
who rated the feasibility of the system to accommodate
each scenario using the same scale and offered a ratio-
nale for that decision. The results are then aggregated
and compared.
Results
Clustering analysis
Figure 6 illustrates the imported DSM file highlighting
five layers (modules) plus connection materials. With
exception of the structure layer, the remaining layers
are sparsely defined. Inside the remaining layers
are small groups of components (e.g. 1214, 3031,
3233). The addition of connection materials as bus
components shows a split in that some carry dependen-
cies across layers while others appear sparse and limited
to a layer.
Figure 7 illustrates an initial manually clustered
matrix restricted to within the layers, with the exception
of components identified as buses and the distribution
of connection materials which were sparsely populated.
Steps taken were:
(1) Moved two bus components to the bottom:
structural frame (4) and electrical cables (16).
(2) Distributed connection materials (3540)
throughout with the exception of bolts and
fasteners (39).
(3) Reordered space plan elements closer to the
diagonal.
From this 10 clusters were identified. If a dominant
functionality or spatial adjacency was formed a cluster
is given an identifying name in brackets under its associ-
ated layer. Six dependencies were highlighted that
existed outside the identified clusters (floating depen-
dencies). In addition two components, wall partition
(19) and flooring (23) pin clusters (6), (7) and (8)
together respectively within the space plan layer. A quick
visual check also reveals that some of the IPD elements
are separated from each other (e.g. bathroom pod).
The next step used the original DSM as a starting
point to run Loomeos three spectral clustering algo-
rithms. Unnormalized and normalized symmetrical
algorithms generated minimal and repetitive results
on every attempt, regardless of the cluster size, provid-
ing relatively zero insight. On the other hand, the nor-
malized unsymmetrical algorithm (recommended by
Luxburg, 2007) manipulated the matrix extensively
and varied with the number of clusters chosen and
attempts. It was therefore selected as providing the
most appropriate results to investigate. A target of 2
to 12 clusters was chosen and 10 iterations were carried
out for each cluster size. Cluster sizes below 4 and
above 10 proved ineffective (visually there were little
signs of clean clusters) and were discarded. Within
168 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Pro
duct
ele
men
t
Thr
eade
d ro
ds a
nd n
uts
Ste
el b
ase
plat
es
Cur
tain
wal
l fra
me
PV
pan
els
(sou
th)
Non
-ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Ope
rabl
e IG
Us
(sou
th)
Ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Slid
ing
Doo
rs (
nort
h) -
IG
Us
and
fram
e
Sec
urit
y gl
ass
(bal
cony
and
ter
race
)A
cryl
ic b
atte
n sk
ylig
ht s
yste
m
PV
pan
el (
cano
py)
Ext
erio
r S
W p
anel
Ele
ctri
cal c
able
s (t
hrou
gh t
he s
truc
ture
)
Inte
rior
SW
pan
el
Par
titi
on w
all (
fram
e &
pan
el)
Par
titi
on w
all w
/ lou
vres
Bat
hroo
m p
ods
(BP
)
Sta
irca
se (
ST
) (a
cryl
ic p
anel
s)do
ors
(int
erio
r)
Sta
irs
LE
D f
ixtu
res
(acc
essi
bilit
y)
LE
D L
ight
ing
stri
ps (
stru
ctur
e)L
ED
Lig
hts
in b
athp
odP
lum
bing
acc
esso
ries
(fa
ucet
, wat
er p
ipes
)F
ixtu
res
(bat
hroo
m)
App
lianc
es (
kitc
hen)
Kit
chen
Cab
inet
s -
isla
nd
Dou
ble
side
d ta
pe (
conn
ecti
on m
ater
ial)
Fla
shin
g ta
pe (
conn
ecti
on m
ater
ial)
Scr
ews
Adh
esiv
e (g
lue)
Cop
per
tape
(C
onne
ctio
n m
ater
ial)
Bot
h &
Fas
tene
rs (
conn
ecti
on m
ater
ial)
Alu
min
um S
helv
ing
(att
ache
d pa
rtit
ion
wal
ls)
LE
D li
ghts
(ab
ove
kitc
hen,
wal
l pan
els)
Pol
ycar
bona
te p
late
s (f
loor
ing)
Ven
tila
tion
Fan
s (r
oof
leve
l, ce
ntre
fra
me)
Inte
rior
SW
pan
el w
/ lou
vres
Alu
min
um G
rate
(ba
lcon
y, g
roun
d fl
oor,
roo
f)
Str
uctu
ral f
ram
e (a
lum
inum
+ s
teel
)
Rei
nfor
ced
foun
dati
on w
/ epo
xy a
ncho
rs
LA
YE
R
ST
RU
CT
RE
SK
IN
SE
RV
ICE
S
SPA
CE
PL
AN
ST
UF
F
CO
NN
EC
TIO
NS
1
2
3
4
5
1
12
45
67
89
O
O
O
1011
1213
1415
1617
1819
2021
2223
2425
2627
2829
3031
3233
3435
3637
3839
403
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Figure6
InitialDSM
highlightinglayer
decomposition
Notes:Dependency
relationship:
X=Symmetrical
O=Asymmetrical
Evaluating adaptability 169
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Pro
duct
ele
men
t
Thr
eade
d ro
ds a
nd n
uts
Ste
el b
ase
plat
es
Non
-ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Cur
tain
wal
l fra
me
PV
pan
els
(sou
th)
Ope
rabl
e IG
Us
(sou
th)
Ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Slid
ing
Doo
rs (
nort
h) -
IG
Us
and
fram
e
Sec
urit
y gl
ass
(bal
cony
and
ter
race
)
Acr
ylic
bat
ten
skyl
ight
sys
tem
PV
pan
el (
cano
py)
Ext
erio
r S
W p
anel
Inte
rior
SW
pan
elIn
teri
or S
W p
anel
w/ l
ouvr
es
Fla
shin
g ta
pe (
conn
ecti
on m
arer
ial)
Cop
per
tape
(co
nnec
tion
mar
eria
l)
(str
uctu
ral/
spat
ial)
foun
daio
n an
d gr
ate
Par
tial
SW
pan
el
light
s an
dap
plia
nces
(spa
tial
)
floo
r an
dca
bine
ts(s
truc
tura
l/sp
atia
l)
bath
pod
and
stuf
f(s
truc
tura
l/sp
stia
l)bu
s el
emen
ts
surf
aces
pin
clu
ster
spa
rtit
ion
(19)
/ flo
orin
g (2
3)
Ele
ctri
cal c
able
s (t
hrou
gh t
he s
truc
ture
)
Str
uctu
ral f
ram
e (a
lum
inum
+ s
teel
)
Par
titi
on w
all (
fram
e &
pan
el)
Par
titi
on w
all w
/ lou
vres
Bat
hroo
m p
ods
(BP
)
Sta
irca
se (
ST
) (a
cryl
ic p
anel
s)
door
s (i
nter
ior)
Sta
irs
LE
D f
ixtu
res
(acc
essi
bilit
y)
LE
D L
ight
ing
stri
ps (
stru
ctur
e)
LE
D L
ight
s in
bat
hpod
Plu
mbi
ng a
cces
sori
es (
fauc
et, w
ater
pip
es)
Fix
ture
s (b
athr
oom
)
App
lianc
es (
kitc
hen)
Kit
chen
Cab
inet
s -
isla
nd
Dou
ble
side
d ta
pe (
conn
ecti
on m
ater
ial)
Scr
ews
Adh
esiv
e (g
lue)
Bot
h &
Fas
tene
rs (
conn
ecti
on m
ater
ial)
LE
D li
ghts
(ab
ove
kitc
hen,
wal
l pan
els)
Pol
ycar
bona
te p
late
s (f
loor
ing)
Ven
tila
tion
Fan
s (r
oof
leve
l, ce
ntre
fra
me)
Alu
min
um G
rate
(ba
lcon
y, g
roun
d fl
oor,
roo
f)
Alu
min
um S
helv
ing
(att
ache
d pa
rtit
ion
wal
ls)
Rei
nfor
ced
foun
dati
on w
/ epo
xy a
ncho
rs
LA
YE
R
ST
RU
CT
RE
SK
IN
SK
IN
SK
IN
SE
RV
ICE
S
SPA
CE
PL
AN
SPA
CE
PL
AN
SPA
CE
PL
AN
ST
UF
F
ST
UF
F(k
itch
en)
(bat
hroo
m)
(sta
irs)
(ven
tila
tion
)
(roo
f &
can
opy)
(sou
th f
acad
e)
(fou
ndat
ion)
(clu
ster
labe
l)
1
2 3 4
5
6 7 8
9 10
1 2 3 5 6 7 8 9 10 11 12 36 13 4014 15 17 18 21 20 22 34 28 27 19 25 26 23 38 24 35 27 29 30 31 32 33 39 16 4
12
56
78
910
1112
3613
1440
1517
1821
2022
3428
3719
2526
2338
2435
2729
3031
3233
3916
43
1
2
63-
5
O
O
O
Figure7
ManuallymanipulatedDSM
Notes:Dependency
relationship:
X=Symmetrical
O=Asymmetrical
170 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Pro
duct
ele
men
t
(str
uctu
ral/
spat
ial)
foun
daio
n an
d gr
ate
(str
uctu
ral/
spat
ial)
(str
uctu
ral/
spat
ial)
(spa
tial
)
(spa
tial
)
bath
room
and
ligh
ts
bath
room
and
wal
l
shel
ves
and
wal
l
(str
uctu
ral)
scre
ws
and
wal
l
shou
ld t
he c
ompo
nent
sbe
left
loos
e o
r sh
ould
they
m a
clu
ster
?
BU
S C
OM
PO
NE
NT
S
LE
D L
ight
ing
stri
ps (
stru
ctur
e)
Str
uctu
ral f
ram
e (a
lum
inum
+ s
teel
)
Ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Slid
ing
Doo
r (n
orth
) -
IGS
s an
d fr
ame
Non
- op
erab
le I
GU
s (n
orth
) in
clud
es f
ram
e
LA
YE
R
ST
RU
CT
RE
SK
IN
SK
IN(r
oof)
SPA
CE
PL
AN
SPA
CE
PL
AN
SPA
CE
PL
AN
ST
UF
F
(bat
hroo
m)
SE
RV
ICE
S(v
enti
lati
on)
(fou
ndat
ion)
(clu
ster
labe
l)
1
2
3
SK
IN(s
outh
fac
ade)
5
6
7
8 9
10
41
a
shou
ld t
he p
arti
tion
wal
l bec
ome
anau
xilia
ry b
us?
b
22
3
4
(coo
rdin
atio
n)fl
oor
and
stai
rs
floo
r an
d ca
bine
ts76
5
ST
UF
F(k
itch
en)
(sta
irs)
1
1
36
36
2
2
56
78
8
9
9
10
10
11
11
12
12
13
13
14
14
4015
17
17
18
18
21
21
20
20
2234
34
28
28
37
37
1925
2623
3824
3527
29
29
3031
3233
5 6 7 401522 19 25 26233824 35273031 32 33
39
39
16
16
4
4
3
3
o
Thr
eade
d ro
ds a
nd n
uts
Ste
el b
ase
plat
es
Cur
tain
wal
l fra
me
PV
pan
els
(sou
th)
Ope
rabl
e IG
Us
(sou
th)
Acr
ylic
bat
ten
skyl
ight
sys
tem
PV
pan
el (
cano
py)
Ele
ctri
cal c
able
s (t
hrou
gh t
he s
truc
ture
)
Ext
erio
r S
W p
anel
Par
titi
on w
all (
fram
e &
pan
el)
Par
titi
on w
all w
/ lou
vres
Bat
hroo
m p
ods
(BP
)
Sta
irca
se (
ST
) (a
cryl
ic p
anel
s)S
tair
s L
ED
fix
ture
s (a
cces
sibi
lity)
door
s (i
nter
ior)
LE
D L
ight
s in
bat
hpod
Plu
mbi
ng a
cces
sori
es (
fauc
et, w
ater
pip
es)
Fix
ture
s (b
athr
oom
)
App
lianc
es (
kitc
hen)
Kit
chen
Cab
inet
s -
isla
nd
Dou
ble
side
d ta
pe (
conn
ecti
on m
ater
ial)
Fla
shin
g ta
pe (
conn
ecti
on m
ater
ial)
Scr
ews
Adh
esiv
e (g
lue)
Cop
per
tape
(C
onne
ctio
n m
ater
ial)
Bol
ts &
Fas
tene
rs (
conn
ecti
on m
ater
ial)
Alu
min
um S
helv
ing
(att
ache
d pa
rtit
ion
wal
ls)
LE
D li
ghts
(ab
ove
kitc
hen,
wal
l pan
els)
Pol
ycar
bona
te p
late
s (f
loor
ing)
Ven
tila
tion
Fan
s (r
oof
leve
l, ce
ntre
fra
me)
Inte
rior
SW
pan
el w
/ lou
vres
Inte
rior
SW
pan
elS
ecur
ity
glas
s (b
alco
uny
and
terr
ace)
Alu
min
um G
rate
(ba
lcon
y, g
roun
d fl
oor,
roo
f)
Rei
nfor
ced
foun
dati
on w
/ epo
xy a
ncho
rsO
O
O
Figure8
DSM
ofLoomeosautomatedclustering
Notes:Dependency
relationship:
X=Symmetrical
O=Asymmetrical
Evaluating adaptability 171
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Pro
duct
ele
men
t
cont
rolli
ng
floo
r an
d st
airs
(slo
w t
o fa
st)
(coo
rdin
atio
n)
Str
uctu
ral f
ram
e (a
lum
inum
+ s
teel
)
LA
YE
R
SK
IN(r
oof)
(sm
art
wra
p)
(nor
th f
acad
e)
(sou
th f
acad
e)
SPA
CE
PL
AN
SPA
CE
PL
AN
(bat
hroo
m)
SE
RV
ICE
S(v
enti
lati
on)
(clu
ster
labe
l)
1
1
2 3S
KIN
4S
KIN
5
6
SK
IN
8
SPA
CE
PL
AN
(par
titi
on w
all)
9
10
7
ST
UF
F(k
itch
en)
(sta
irs)
1
1
36
36
2
2
5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
12
13
13
14
14
40
40
15
15
17
17
18
18
21
21
2837
3419
2523
2633
3216
3824
3527
28 37 34 19 25 23 26 33 32 163824 3527
22
22
31
31
30
30
29
29
20
20
39
39
4
4
3
3
Thr
eade
d ro
ds a
nd n
uts
Ste
el b
ase
plat
es
Acr
ylic
bat
ten
skyl
ight
sys
tem
PV
pan
el (
cano
py)
Fla
shin
g ta
pe (
conn
ecti
on m
ater
ial)
Bol
ts &
Fas
tene
rs (
conn
ecti
on m
ater
ial)
LE
D li
ghts
(ab
ove
kitc
hen,
wal
l pan
els)
Alu
min
um G
rate
(ba
lcon
y, g
roun
d fl
oor,
roo
f)
Rei
nfor
ced
foun
dati
on w
/ epo
xy a
ncho
rsS
TR
UC
TR
E(f
ound
atio
n)
Non
-ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Cur
tain
wal
l fra
me
PV
pan
els
(sou
th)
Ope
rabl
e IG
Us
(sou
th)
Ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Slid
ing
Doo
rs (
nort
h) -
IG
Us
and
fram
e
Sec
urit
y gl
ass
(bal
cony
and
ter
race
)
Ext
erio
r S
W p
anel
Inte
rior
SW
pan
elIn
teri
or S
W p
anel
w/ l
ouvr
es
Cop
per
tape
(co
nnec
tion
mar
eria
l)
Ele
ctri
cal c
able
s (t
hrou
gh t
he s
truc
ture
)
Par
titi
on w
all (
fram
e &
pan
el)
Par
titi
on w
all w
/ lou
vres
Bat
hroo
m p
ods
(BP
)
Sta
irca
se (
ST
) (a
cryl
ic p
anel
s)
door
s (i
nter
ior)
Sta
irs
LE
D f
ixtu
res
(acc
essi
bilit
y)
LE
D L
ight
ing
stri
ps (
stru
ctur
e)
Plu
mbi
ng a
cces
sori
es (
fauc
et, w
ater
pip
es)
Fix
ture
s (b
athr
oom
)
App
lianc
es (
kitc
hen)
Kit
chen
Cab
inet
s -
isla
nd
Dou
ble
side
d ta
pe (
conn
ecti
on m
ater
ial)
Scr
ews
Adh
esiv
e (g
lue)
LE
D li
ghts
in b
athp
od
Pol
ycar
bona
te p
late
s (f
loor
ing)
Ven
tila
tion
Fan
s (r
oof
leve
l, ce
ntre
fra
me)
_
Alu
min
um S
helv
ing
(att
ache
d pa
rtit
ion
wal
ls)
OO
O
Figure9
Finalclustered
DSM
Notes:Dependency
relationship:
X=Symmetrical
O=Asymmetrical
172 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
the 4 to 10 range, a handful of permutations was
selected to investigate further based on their cleanli-
ness, visual compactness to the diagonal and variety
of cluster sizes.
Out of the subsets explored, normalized 10.05 (10
indicates the cluster size and 5 the iteration number)
is presented as a permutation which led to interesting
results. Its worth noting that the software does not
explicitly identify clusters, which is at the discretion
of the modeller. Ten clusters were identified tightly
bound along the diagonal (Figure 8) placing anomalies
at the top left corner, including three integrating
components along with a handful of components with
no dependencies outside of the integrating
components. The exception to this is the first cluster
Services (ventilation) that appears to have been pinned
between bus components bolts and fasteners (39) and
structural frame (4). The cluster sizes are quite
compact, varying between two and four components.
Seven dependencies (two asymmetrical) were high-
lighted outside the clusters. LED lights kitchen (26)
appears to pin clusters (9) and (10) (binding the space
plan and stuff layers). This pinning (in relationship with
kitchen appliances (32)) appears to hold away kitchen
cabinets (33) from flooring (23). Upon completion of
the analysis, two questions were prompted:
(1) Should the components at the top of the matrix
be left independent or should they form a
cluster(s)?
(2) With approximately half the floating dependen-
cies, should wall partition (19) form an auxil-
iary bus which multiple clusters are pinned to?
Observations from the manual manipulation
(Figure 7) and automated clustering (Figure 8) led to
the final three steps of manual manipulations on top of
the automated clustered DSM. Step one attempted to
remove the floating dependencies as documented here:
Floater (1): Flashing tape (36) and aluminiumgrate (14) were swapped to allow for aluminium
grate (14) to become pinned between clusters
(3) roof and (4) foundation.
Floater (2): LED lights bathpod (29) was shiftedto cluster (8) space plan (bathroom).
Floaters (3 and 4): Cluster (2) stuff was shiftedadjacent with partition walls (19).
Floater (7): Flooring (23) and doors (25) wereswapped to enlarge cluster (10) stuff to include
LED lights (26) and flooring (23).
After step one floating dependencies (5) and (6)
remained:
Floater (5): Partition wall (19) is pinned betweentwo clusters within the space plan layer holding
bathroom pod (22) away.
Floater (6): Flooring (23) is pinned between aspace plan cluster and stuff cluster holding it away
from staircase (24).
The second step investigated the remaining clusters
and component classifications in comparison to the
IPDs.
Cluster 1 Services (ventilation): While notstructurally tied together functionally rely on
each other and represent an alternative grouping
compared to the IPDs.
Cluster 2 Stuff: Integrated with cluster (9). Cluster 3 Skin (roof): Could be separated into two
smaller clusters, roof and canopy mimicking their
separation as IPDs; linked by acrylic batten
system (12).
Cluster 4 Structure (foundation): Tightly boundand constructed together on site.
Cluster 5 Skin (south): Tightly bound andassembled on site.
Cluster 6 Space plan (bathroom): Tightly boundand assembled in the factory as an IPD, but has
a spatial coordination with partition walls (19)
which pin it to another cluster.
Cluster 8 Space plan (staircase): Tightly boundcluster and IPD.
Cluster 9 Space plan: With the addition of cluster(2), cluster (9) is largely connected through par-
tition wall (19) as an integrating component. The
order of components is adjusted to allow parti-
tion wall (19) to be pinned next to cluster (6)
(bathroom) which has swapped positions with
cluster (8) (stairs) removing floater (5). While
improving the clustered solution, cluster (9) con-
tains multiple IPDs.
Cluster 10 Stuff: Consists of multiple IPDs and ispinned to cluster (9) by LED lights kitchen (26)
along with flooring (23) which has dependencies
in both clusters as well.
Non-clustered elements (811, 17 and 18): Thesecomponents consist of skin layer components
Table 1 Strategy and action breakdown of change scenarios
Strategy Action 01 Action 02
Adjustable (3) Alter (2) Add (1)
Versatile (6) Alter (5) Add (1)
Refitable (13) Replace (8) Add (5)
Scalable (6) Add (5)
Convertible (1) Replace (1) Alter (1)
Moveable (1) Alter (1)
Evaluating adaptability 173
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Table 2 Results of scenario assessments
Scenario Strategy Action A B System expert (A) DSM modeller (B) R
1 Move a
partition
Versatile Alter 3 2 () Component scale, ()
relationship w/ installations (22)
() Flooring (23) D
2 Add a new
partition
Versatile Add 2 2 () Component scale, () lack of
space
() Flooring (23) D
3 Replace a
partition
Refitable Replace 2 1 () Component scale, () lack of
space
(+) Bolts (39) D
4 Move
bathroom pod
Versatile Alter 3 3 () Fixed position w/ partition
walls (19) and vertical shaft
() Partition walls (19), ()
assembly
S
5 Replace
bathroom pod
Refitable Replace 2 3 () Volumetric assembly (22),
(+) Structurally independent
() Partition walls (19), ()
volumetric assembly (22)
D
6 Change room
temperature
Adjustable Alter 2 2 (+) Natural ventilation, () poor
air tightness
(+) Natural ventilation, ()
system limitations
S
7 Add
mechanical
heating
Refitable Add 2 4
(3)
() Poor air tightness () Partition walls (19), ()
electrical cables (16)
D
8 Change
flooring
Refitable Replace 1 1 (+) Simply lifted out (+) Kitchen cabinets (33),
appliances (32)
S
9 Add ceiling Refitable Add 2 4
(1)
(+) Simple to install, (+)
enhance acoustics
(+) Bolts (39) S
10 Remove 3rd
floor
Scalable Remove 2 4
(x)
(+) Partial disassembly Wrong granularity X
11 Add space
horizontally
Scalable Add 2 4
(x)
() Creates a deep plan Insufficient information X
12 Replace facade
panel
Refitable Replace 1 1 (+) Easy to disassemble & insert
different solution
(+) Bolts (39), (+) copper tape
(40)
S
13 Change facade
panel
Refitable Replace 1 1 (+) Easy to disassemble & insert
different solution
(+) Bolts (39), (+) copper tape
(40)
S
14 Change
material of
panel
Refitable Replace 2 4
(x)
Take panel off & change in
factory
Wrong granularity X
15 Add shading
device
Refitable Add 1 4
(1)
(+) Could be installed on
exterior or integrated into panel
(+) Bolts (39) S
16 Change to
commercial use
Convertible Replace
& alter
1 2 () Acoustics, () thermal
control, () lighting control
(+) Panels, () lighting, ()
bathroom pod, () kitchen
cabinets
D
17 Change
locations
Moveable Alter 1 1 (+) Good reusability of
materials, () climate conditions
(+) Bolts (39) S
18 Replace
ventilation
fans
Refitable Replace 1 2 (+) Unbolt () Partition wall (20) D
19 Move staircase Versatile Alter 3 3 () Structural composition () Flooring (23), () assembly
(24)
D
20 Enclose terrace
for winter
Refitable Add 2 4
(x)
(+) Lightweight solution Insufficient information X
21 Enclose terrace
complete
Scalable Add 3 4
(x)
() Insufficient load capacity Insufficient information X
22 Make lower
floor an
interior
Scalable Add 3 4
(x)
() Insufficient foundation; ()
floor height
Insufficient information X
23 Adjust lighting Adjustable Alter 1 3 (+) Capacity within fixtures
(2629)
() Electrical cables (16), ()
fixtures (2629)
D
24 Add new
lighting
Refitable Add 1 3 (+) Redundancy in electrical
cables (16)
() Electrical cables (16) D
(Continued)
174 R. Schmidt et al.
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
from different facade orientations. As non-depen-
dent components to each other, it makes sense
for them to be nestled against bus elements from
a pure clustering perspective. However, given
knowledge of their production dependencies it
makes sense to cluster them with their associated
components as IPDs. Elements (17) and (18)
were moved together with the other SW panel
components to form cluster (3) skin (Smart-
Wrap panels, an IPD). The remaining fourcomponents are then identified as cluster (4) skin
(north facade) representing the elements of the
north facade, despite not having any dependen-
cies between each other (811).
The last manoeuvre sequenced the modules as an
indication of their rate of change; structure, the longest
lasting layer, at the top left to the quickest stuff layer at
the bottom right. The result of the three steps is pre-
sented as Figure 9. The relationship between flooring
(23) and stairs (24) remains as the only floating depen-
dency (floater 1). This is due to the flooring (23) being
pinned between two clusters, cluster (10) stuff and clus-
ter (9) space plan. An alternative solution would visual-
ize the flooring (23) as an auxiliary bus changing how
the dependency is viewed. On the other hand, partition
wall (19) also has dependencies within three defined
clusters, but is resolved to a certain extent by pinning
flooring (23) and LED lights kitchen (26) to cluster
(10), creating an overlap between the three clusters.
It is also worth noting that two overlaps in clusters
are not due to a component being pinned between mul-
tiple clusters, but because the same component has two
different applications in the system yet only represented
once in the DSM (aluminium grate (14) and acrylic
batten skylight (12)).
A last observation relates to bus components. The
system was designed for elements to tie back to the
structural frame (4), so it makes sense that this
component is acting as an integrating element. Bolts
and fasteners (39) acting as a second bus component
is a result of modelling connection materials and the
dependency between structural frame (4) and attached
components splitting into two types (structural and
spatial). The third bus component electrical cables
(16) are interwoven into the structural frame (4) and
can become a serious propagation issue.
Impact analysis
A breakdown of the 30 scenarios is presented in Table 1
providing an overview of the strategies investigated and
their associated action(s). Many demonstrate the refit-
able strategy (13) and the add action (12) which
reflects the relatively bare nature of the system as a
starting point.
Table 2 presents the results of the analysis for the 30
scenarios. The feasibility level assigned to each scenario
Table 2 (Continued)
Scenario Strategy Action A B System expert (A) DSM modeller (B) R
25 Add additional
furniture
Adjustable Add 1 4
(1)
(+) No problems Flooring (23) D
26 Split dwelling
into two
Scalable Add 3 4
(x)
() Circulation Insufficient information X
27 Alter structural
frame
Versatile Alter 3 3 () Specific capacity of each
member
() Number of dependencies D
28 Replace
structural
member
Refitable Replace 2 3 () Structural integrity () Number of dependencies D
29 Add structural
member
Scalable Add 1 4
(2)
() Depends on structural
capacity
() Number of dependencies D
30 Alter kitchen
layout
Versatile Alter 2 2 () Component specificity () Cables (16) D
Table 3 Feasibility ratings
System expert DSM modeller
Feasible (11) Feasible (8)
Somewhat feasible (12) Somewhat feasible (7)
Not feasible (7) Not feasible (8)
Unclear (7)
Table 4 Comparison of feasibility ratings and rationale
Rationale Agreed on feasibility Disagreed on feasibility
Same 7 0
Different 5 11
Evaluating adaptability 175
Dow
nloa
ded
by [U
nivers
idad D
el Ro
sario
] at 0
7:39 2
5 July
2014
-
Pro
duct
ele
men
t
RO
UN
D 0
2
RO
UN
D 0
4
RO
UN
D 0
4
RO
UN
D 0
3
RO
UN
D 0
1
RO
UN
D 0
2
Thr
eade
d ro
ds a
nd n
uts
Ste
el b
ase
plat
es
Str
uctu
ral f
ram
e (a
lum
inum
+ s
teel
)
Acr
ylic
bat
ten
skyl
ight
sys
tem
PV
pan
el (
cano
py)
Ext
erio
r S
W p
anel
Fla
shin
g ta
pe (
conn
ecti
on m
ater
ial)
Bol
ts &
Fas
tene
rs (
conn
ecti
on m
ater
ial)
LE
D li
ghts
(ab
ove
kitc
hen,
wal
l pan
els)
Alu
min
um G
rate
(ba
lcon
y, g
roun
d fl
oor,
roo
f)
Rei
nfor
ced
foun
dati
on w
/ epo
xy a
ncho
rs
Non
-ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Cur
tain
wal
l fra
me
PV
pan
els
(sou
th)
Ope
rabl
e IG
Us
(sou
th)
Ope
rabl
e IG
Us
(nor
th)
incl
udes
fra
me
Slid
ing
Doo
rs (
nort
h) -
IG
Us
and
fram
e
Sec
urit
y gl
ass
(bal
cony
and
ter
race
)
Inte
rior
SW
pan
el
Inte
rior
SW
pan
el w
/ lou
vres
Cop
per
tape
(co
nnec
tion
mar
eria
l)
Ele
ctri
cal c
able
s (t
hrou
gh t
he s
truc
ture
)
Par
titi
on w
all (
fram
e &
pan
el)
Par
titi
on w
all w
/ lou
vres
Bat
hroo
m p
ods
(BP
)
Sta
irca
se (
ST
) (a
cryl
ic p
anel
s)
door
s (i
nter
ior)
Sta
irs
LE
D f
ixtu
res
(acc
essi
bilit
y)
LE
D L
ight
ing
stri
ps (
stru
ctur
e)
Plu
mbi
ng a
cces
sori
es (
fauc
et, w
ater
pip
es)
Fix
ture
s (b
athr
oom
)
App
lianc
es (
kitc
hen)
Kit
chen
Cab
inet
s -
isla
nd
Dou
ble
side
d ta
pe (
conn
ecti
on m
ater
ial)
Scr
ews
Adh
esiv
e (g
lue)
LE
D li
ghts
in b
athp
od
Pol
ycar
bona
te p
late
s (f
loor
ing)
Ven
tila
tion
Fan
s (r
oof
leve
l, ce
ntre
fra
me)
Alu
min
um S
helv
ing
(att
ache
d pa
rtit
ion
wal
ls)
1
1
2
2
4
4
5
5
6
6
7
7
8
8
9
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
1011
1213
1415
1617
1819
2021
2223
2425
2627
2829
3031
3233
3435
3637
3839
403
3
O
O
O