See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334425507

Modularity in architectural design: Lessons from a housing case

Chapter · July 2019

DOI: 10.1201/9781315229126-78

CITATIONS READS

0 223

3 authors:

Antonio Lopes Correia Luís Simões da Silva

University of Coimbra University of Coimbra
17 PUBLICATIONS 33 CITATIONS 562 PUBLICATIONS 4,482 CITATIONS

SEE PROFILE SEE PROFILE

Vitor Murtinho
University of Coimbra
28 PUBLICATIONS 7 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

international conferences View project

SBRI - Sustaible steel- concrete composite bridges View project

All content following this page was uploaded by Antonio Lopes Correia on 26 July 2019.

The user has requested enhancement of the downloaded file.

Modularity in Architectural Design: Lessons from a Housing Case

A. L. Correia
ISISE (Institute for Sustainability and Innovation in Structural Engineering), UC (University of Coimbra), Portugal
V. Murtinho
CES (Centro de Estudos Sociais), UC, Portugal
L. Simões da Silva
ISISE, UC, Portugal

ABSTRACT: This paper presents a housing case devised with modular principles set towards an
algorithmic description. Aside illustrating the case, the broader aim is to show both the potential
and limitations of a modular approach from an architectural perspective, on the way to an im-
provement of decision-making processes in the complex, multi-dimensional problems that arise
during architectural creation.

1 INTRODUCTION
Within a modular system, the array of hypotheses underlays a reference frame with production
control and scale purposes, that nonetheless enables variability in the outputs. Current technolo-
gies allow a more feasible production of non-standard material dimensions than it used to in the
past (e.g. 3D printing). Nonetheless, normalized sizes are still cheaper in most cases. Moreover,
modularity of physical components is conceivable under a discrete aegis, which envisions a po-
tential complexity achievable in the fashion of the discrete math of algorithms.
Research in modularity in the context of the built environment has been following three main
trends: (1) deepening on to the nature and/or features of module(s), from which a network of
relations can evolve to create a range of shapes (e.g. Shahabian and Osman, 2017); (2) focusing
on modular applications, i.e., in the engineering and assembly processes implied for modular
structures to be brought about in real-life applications (e.g. Lawson et al., 2014); (3) onto broader
aspects of modular systematization (e.g. Kamali and Hewage, 2016). Following previous inves-
tigation (Correia et al., 2017), the research hereby presented reflects a synthesis of these concerns,
as it deepens to the modular nature of the study object while addressing its applicability in housing
production (Fig. 1).

Figure 1. Illustration of a presumable expanded set of modular housing units based on the case-study design.
2 CASE-STUDY
2.1 Background
The development of a modular design for single-family housing purposes started from a previ-
ously developed design (Murtinho et al., 2010) based in units (u) of 60x60cm, forming rectangular
modules of order u:(7,8), disposable in a modular matrix of order m:(2,3) (Fig. 2a). From this
scheme, dozens of combinations are conceivable between two consecutive floors of an m order
scope (Fig. 2b), or thousands, if considering different internal variations within the u order scope.
In a superior m order, say e.g. m:(4,2), any possible derivation can be made by assembled deriva-
tions of a previous m order, such as the m:(2,3) and, in the very limit, any shape can be made from a
single module (Fig. 3). This is an elementary conclusion, that nonetheless puts in evidence two rele-
vant aspects: it reinforces the need to impose subsequent constrains, according to an intended scope;
and the need to focus the research in the modular singularity.

2.2 Addressing dependencies

A modular optimization implies the tackling of spatial/constructive dependencies. In the case, that
was primarily materialized by a restriction of the internal space fit-out to the scope of each module
and concomitantly to use symmetry for improved constructive efficiency (Fig. 4).

2.3 Clarifying interfaces

It also matters to clarify interfaces, again both at a spatial and constructive level. For instance, in Figure
4, given the unnecessary constructive duplication in the contact area (Fig. 5a), it must be opted (Fig.
5b) by privileging either one side (Fig. 5c) or the other (Fig. 5d). As a by-product of these small dif-
ferences, subtle asymmetries are introduced. From the subjective standpoint of the design intentions
this may either be regarded positively (as was the case) or as an undesirable feature.

2.4 Optimizing similarity

The modules of u:(8,7) intrinsically share dimensional difficulties that inhibit certain operations, as
the possibility to rotate modules without spatial repercussions. By enlarging to u:(8,8), the dimensional
issue is mitigated. Besides, given that it is obtained a square base, symmetries are doubled (Fig. 6).

2.5 Expanding the modular application

The expansion to multi-family and multi-story purposes, led to the consideration of a minimum
unit optimized for applications developed over with two or more floors. In the case, the devised
solution allows four entrance directions and 4 development directions on the subsequent floor
(Fig. 7a). As shown, the minimum for this is a 1–2 shape type (i.e., 1 module required for entrance
zone and 2 for minimum development of a housing typology). Figure 7b depicts some of the
multiple formal possibilities from a nucleus of the 1–2 type.

3 ENCODING THE MODULAR DESIGN

The housing program constrains were developed to define the number of admissible house spaces
(Table 1) corresponding to a house typology. The first step of the algorithm (Fig. 8) clarifies the
modular scope, setting the modular constraints input. Then, from a user inputted brief, a two-fold
primary functional allocation stage is informed, with a part for allocation and the other to establish
topological relations. The output of this step informs shape assembly at a modular scope. Subse-
quently, this can be further detailed, thus proceeding to the secondary functional allocation.
Following a modular principle, it is defined that at a primary level is allocated a main functional
space to each module m of an m:(2, 3) scope. In a subsequent step, secondary functional spaces
are to be allocated within each of these main functional spaces. The secondary spaces S and T,
are admissible within any primary space; I only admissible in K; and V only within O and only if
a second level is required. The circulation spaces are not assignable, as they can be functionally
Figure 2. (a) Modular matrix scopes, and (b) example of combinations in a m scope.

Figure 3. Instances of a m:(4,2) assembled through combined derivations of the initial m:(2,3).

Figure 4. Addressing dependencies: (a) previous solution; (b) optimized solution.

Figure 5. Clarifying interfaces.

Figure 6. Optimizing similarity: where (a) is the initial solution and (b) the optimized design.

Figure 7. (a) Minimum unit using a 1–2 type; (b) formal possibilities from a 1–2 type.
defined by exclusion of the remaining primary and secondary functional allocation. Nevertheless,
these have a key role in the definition of dimensional constraints that integrate both the primary and
secondary functional spheres. The number of instances of m required by a housing brief is defined
with function to bedrooms in the house (tb, i.e. typology and number of bedrooms within).
The total number of m required by a housing brief is defined in function to number bedrooms in the
house (tb, standing for typology and number of bedrooms within). Following the methodology, rules for
primary functional assignment can thus be defined (Table 2). With this set of rules, and given an inputted
housing brief, it becomes possible to compute a primary housing program for functional assignment.
Without loss of generalization, based on the previous set of rules, Table 3 shows the manually computed
results for the simpler one level case house in a m:(2, 3). The subsequent step of program assignment is
to use this information (e.g. Table 2) as input to placement within modules, which calls for further rules
defining the relative positioning of each functionally allocated module, as in the pseudocode described
in Figure 9i. Using a similar formalism, it is possible to transform the outputted matrix of this algorithm
into a set of topological relations between main functions, which define the general circulation flows in
the typology (Fig 9ii). In Figure 10 applications of these rules are illustrated.
Since shapes and sub-shapes are previously clarified in the modular design, it is not necessary
to develop a full algorithm for shape creation, as it is, e.g., in a classical shape grammar fashion
(cf. Stiny, 2006). Instead, shapes are a consequence of a clarified functional mapping acting on
the previously defined modules’ physical and dimensional features. That is, a primary functional
allocation, defines the broad volumetric features, over which acts a secondary functional alloca-
tion, with its subset of rules, and eventually introducing subsequent levels.
Secondary spaces are describable as subsets of the primary spaces. These must follow their own
dedicated brief and obey a hierarchy of allocation. Each house requires an I, a T and an S as minimum
secondary spaces. Given its specificity, the kitchen is considered both as a primary (labelled K) and a
secondary (labelled I) space. The number of I (i.e. the iI) thus comes in first and is always iI=1. It
follows the number of T, that can be defined it in relation to the number of B, where iTmax=iB, iTmin=(iB-
1) for 2≤iB≤3 or iTmin=(iB-2) for iB≥3, and finally in a t0 typology, iT=1. As to the number of S, every
typology has an iSmin=1. For further cases, we can define it in relation to T, where 1≤iS≤(1+iTmax–iTmax).
The output conjunction of these elements is the brief of the secondary functional spaces.
Finally, the spatial integration of the primary and secondary elements follows the dimensional con-
straints imposed by the circulation spaces (H and C). Both can be defined implicitly or explicitly.
Implicitly means that they occur integrated in a wider space than its own minimum dimensional re-
quirements, e.g. occurring via an L space. In this sense, H or C will be implicit within any primary
functional space. Explicitly means they go from a module to another through a third module, thus
having to physically separate the functional space from C. This is also what occurs between a primary
and a secondary space. Thus, explicit circulation is defined from a secondary space, whereas the im-
plicit circulation is non-definable. Finally, in terms of housing brief, H has the specificity of occurring
only once, and thus must be associated with a module or a pair of modules which are closer to a certain
requested entrance direction, definable through a precedent input.

4 CONCLUSIONS

4.1 Modularity, variability and discrete description

Dimensional coordination for standardized components must be addressed towards the benefit of pro-
duction scale but will likely make variability in components harder to attain. When the latter is in-
tended, the use of CAD-CAM in non-linear production methods can be a valuable contributor, with
techniques such as 3D printing with high potential in this domain. However, the notion of modularity
is vaster than the notion of component, and with potentially more varied, complex and precise out-
comes. Modularity entails the possibility that things are ought to be encapsulated within an entity defin-
able by a self-contained functionality and an interface, easing its description in the fashion of a discrete
mathematical entity. That enhances the ability for an algorithmic applicability (e.g., for rule-based de-
signs). However, modular design is difficult, harder than “regular”, as it implies more constraints, which
must be tested and validated in advance—as was in this case with the previous construction of a real-
scale prototype (Figure 11 and 12)—before their discrete “dematerialization”. Anyhow, on its own, the
knowledge of the implications of modularity can bring an improved insight on intuitive design processes.
Figure 8. Algorithm development flowchart.

(a) (b) (c)

primary allocation label secondary allocation label circulation label
base module O supplement (storage, pantry, etc.) S horizontal circulation C
kitchen K toilet T entrance hall space H
bedroom B vertical circulation (stairs) V
living room L kitchen I
dining room D
Table 1. Functional labelling.

1: For each B in tB (except t0), m must be at least

per B: m=1
per K: m=1
per L: m=1 except if L and D are not differentiated, then m(L+D)=1
per D: m=1 except if L and D are not differentiated, then m(L+D)=1
2: In terms of tb definition, maximum number of m for each B in tB comprehends a maximum of (except in t0)
per B: m=2 meaning that for a certain of tb there is only one master-bedroom considered (i.e. a B with m=2). If there is more
than one master-bedroom with m=2, then tb increases level to tb+1, meaning there is only one master-bedroom for
the B value of a tb considered
per K: m=1
per L: m=1
per L: m=1
3: Minimum number of levels per house is 1
4: Maximum number of levels per house is 2
5: Per level, minimum m=2
6: Per level, maximum m=6
Table 2. Primary design brief assignment rules.

typology number of modules base program

t0 t0 min 2 (O, K)
t0 - -
t0 max 3 (O, K, B)
t1 t1 min 3 (O, K, B)
t1 4 (O, K, B, L)
t1 max 5 (O, K, B, L, D)
t2 t2 min 4 (O, K, B, B)
t2 5 (O, K, B, B, L)
t2 max 6 (O, K, B, B, L, D)
t3 t3 min 5 (O, K, B, B, B)
t3 - -
t3 max 6 (O, K, B, B, B, L)
t4 t4 min 6 (O, K, B, B, B, B)
t4 - -
t4 max - -
Table 3. Primary housing program for a simple one level house, given Table 1 rules and a m:(2, 3) grid.

4.2 Modular knowledgeability for improved design decisions

The knowledge of different aspects of modularity can be a valuable contribution to improve de-
cision-making processes in the complex, multi-dimensional decisions that are required during
architectural creation, namely having in mind aspects of dependencies, similarities and interfaces:
(a) Dependency reduction. Modularity is all about reducing dependencies, starting from a
clearly established functional map. Dependencies can be more obviously observed in physical or
constructive terms, but in fact these are essentially emanated from to the ability or intention to
discretize spatially attributed functionalities.
(b) Boosting similarity. Similarity aspects, such as the introduction of symmetries, must be a design
concern if aiming to attain constructive cost-efficiency while keeping up with quality standards. The
Figure 9. Pseudocode to (i) generate a random primary functional allocation from a given brief, for a generic
case of a single level typology of an m:(2, 3) scope and (ii) for conversion of functional allocation into
topological relations for a simple case of a single level typology.
Figure 10. Example of derivation, for given inputs, of the set of rules for (a) allocation and (b) topology,
respectively following rules described in Figure 7i and Figure 7ii.

Figure 11. Interior view of prototype under construction.

Figure 12. External view of built prototype.

 eventual idea that this may result in monotonous environments is wrong. Indeed, if devised ac-
cordingly, it can also enact a wide and varied range of solutions.
(c) Addressing interfaces. Modules are to be built in real space, and thus their boundaries have
materiality, with a body or thickness. Indeed, if in diagram the junction of modules can work seamlessly,
when further detailed, it will often arise contiguously duplicate elements. This can be less of an issue
when these walls separate, say, different dwellings, yet it adds unnecessary elements when within the
same dwelling. When removing them in a drafting stage, particularly in the case of more complex
connections between modules’, often unexpected blank areas come up, disrupting previous formal
alignments, and so forth. Broadly, there are two main options in this respect, which are to either repeat
structural elements or not doing it. The first, privileges discreteness, and thus has potential gains in
terms of simplicity of production processes, although it might incur in material losses given the un-
necessary repetition of construction elements. The second option, is released of these issues, but will
disrupt alignments (which may not be desirable design-wise) and will have more dependencies attached.

4.3 Notes towards algorithmic description of modular designs

As designs become increasingly clarified in a modular fashion, they also become closer to an
algorithmic description. Following a typical evolution path of a design, in the early stages of the
presented case, the housing program was distributed a shape or combination of shapes, over which
finer adjustments followed, in a roughly intuitive process. The starting point of that kind of ap-
proach is a formal desire. However, the underlying logic of an algorithmic description works
inversely, that is, form is raised from the specification of an elementary structure of relations.
These are ought to speak a logical language, which may be far from intuitive. That cannot occur
without a design precedence, whose roots are necessarily “intuitive”, but most importantly that
raises the need to use a different approach. Generically, the way to do it seems to be from an
architectural program starting point (Duarte, 2000). In this case, it started by a housing program
that is emanated from a design brief. The envisioned process begins by a definition of a set of
needs, transcribed in a system-constrained housing program, from where subsequently functional
relations are built. Final shape will be accomplished by successively building up these relations
and will be terminated when all the program elements are accommodated in their respective spa-
tially predetermined entities (i.e., the modules). The logic is analogous to establishing a functional
mapping to assess a product modularity (Correia et al., 2017). That is, the topological relations
between different functional zones of a design can be outlined but only insofar as they are in a
scope that it can be defined, i.e., in the measure of the detailing required or intended to achieve.

5 REFERENCES

CORREIA, A. L., MURTINHO, V. & SIMÕES DA SILVA, L. (2017) Steel and Modularity in
Architectural Creation. XI Congresso de Construção Metálica e Mista. Coimbra.
DUARTE, J. P. (2000) Customizing mass housing: a discursive grammar for Siza's Malagueira
houses. Dept. of Architecture. Cambridge, Mass., Massachusetts Institute of Technology.
KAMALI, M. & HEWAGE, K. (2016) Life cycle performance of modular buildings: A critical
review. Renewable and Sustainable Energy Reviews, 62, 1171-1183.
LAWSON, M., OGDEN, R. & GOODIER, C. (2014) Design in Modular Construction, CRC Press.
MURTINHO, V., CORREIA, A. L., FERREIRA, H., SIMÕES DA SILVA, L., GERVÁSIO, H.,
REBELO, C., SANTIAGO, A., SANTOS, P., MATEUS, D. & RIGUEIRO, C. (2010)
Affordable Houses: Architectural Concepts of a Modular Steel Residential House. IN
CRUZ, P. (Ed. ICSA 2010 - International Conference Structures and Architecture.
Guimarães, Taylor & Francis Group.
SHAHABIAN, A. & OSMAN, B. (2017) This Satisfying Algorithm Combines Japanese-Inspired
Joints to Generate Any Form. Archdaily. www.archdaily.com/874153/this-satisfying-
algorithm-combines-japanese-inspired-joints-to-generate-any-form . Retrieved: 11.2018.
STINY, G. (2006) Shape: Talking About Seeing and Doing, Cambridge, Mass., MIT Press.

View publication stats