Content uploaded by Robert Aish
Author content
All content in this area was uploaded by Robert Aish on Sep 16, 2022
Content may be subject to copyright.
© 2020 XXX-X-XXXX-XXXX-X/XX
Representations in Transition
Subtitles (are not captured in Xplore)
ROBERT AISH, Bartlett School of Architecture, UCL, London, robert.aish@ucl.ac.uk
Fig. 1. TopoFacade – Topology driven digital fabrication [Fagerström 2014]. The whole process of decomposition, component design and connector
design can be expressed as design rules and encoded in a design computation program. This provides a user-defined ‘change propagation’ mechanism
which frees the designer to explore changes in form and construction strategy. The critical enabler of this ‘change propagation’ mechanism is topology.
By using lightweight topology as the primary representation [in this case a mesh topology], appropriate topological queries can report which components
are adjacent and therefore which connectors are shared between components and how each can become the context which the other has to respond to.
In this example the cross-sectional geometry of the skeletal edge-based frame structure is driven by the bisectors of the adjacent face normals, and the
mitred end-treatment of the frame structure is driven by the pair-wise edge bisectors at each vertex. This is topologically enabled digital craft.
Architecture is predominantly project driven, combining the
objective and the subjective and spanning problem solving, social
concerns and cultural impact. This suggests that tangible architectural
progression, whether radical or incremental, occurs one project at a time.
While architects as digital tool users can express one form of
creativity, there is also creativity to be found within the digital tool
builders. These are the Building Physicists, Computer Scientists,
Software Engineers and Application Developers who explore new and
potentially more expressive architectural representations which are
intended to be applicable across architectural projects in general.
The argument is that new representations offer architects new
abstractions, encourage new ways of thinking, support new types of
expression and thereby create the condition for the emergence of new
tangible forms of architecture.
Experience shows that the development and the successful adoption
of new representations and related tools is not just a question of technical
innovation. It is a very slow process.
The gap between an initial research paper to a working prototype, to
a deployable industry strength implementation and finally to the wide
spread adoption of such representations can be measured in decades.
The process of adoption often involves quite complex issues and
trade-offs within the architectural user community during the transition
from a previously established representation. What lessons can be learnt?
REFERENCES
© 2020 XXX-X-XXXX-XXXX-X/XX
1. THE TRANSITION FROM 2D DRAFTING TO BIM
One of the representations we want to discuss is BIM and one of the
first papers to describe what we now know as BIM and its advantages to
the construction industry was published over three decades ago [Aish
1986]. Yet it is only recently that the rationale presented in this paper is
beginning to be accepted and BIM is being more widely adopted.
Essentially, BIM is the application of the principles of data
normalisation to construction industry information [Codd 1990]. Prior to
BIM, construction information was distributed across a number of
independently editable source documents (2D drawings). There is a
general awareness that there are two fundamental weakness here:
incompleteness and inconsistency. It is often difficult to completely
represent a 3D object such as a building with 2D drawings. It is difficult
to maintain consistency across multiple independently editable source
documents: both rely on human intervention. We are all fully aware that
the consequences of incomplete and inconsistent information are
construction errors and increased construction costs.
In contrast, BIM makes a clear distinction between the defining data
(the 3D building model) and derived data (drawings and reports).
The consistency of this derived data can be contrasted with the inevitable
inconsistencies associated with the prior 2D drafting. However, the
successful adoption of BIM assumes that users are willing to accept a
certain level of ‘cognitive retooling’ [Flemming et al 2002] and process
reengineering across the different actors within a construction project.
The availability of BIM coincided with initiatives within the
construction industry to move from an adversarial to a more collaborative
project structure. BIM can be seen as both an agent of these organisational
changes and a beneficiary of these changes. So from an informational,
organisational and construction process efficiency perspective there can
be few arguments against the adoption of BIM.
It could be argued that the availability of low cost and easily used 2D
CAD systems which completed with the early BIM applications delayed
the transition to BIM and consequently the overall industrial advantages.
2D CAD systems offered immediate but limited benefits while BIM
required changes in practices and offered deferred but more extensive
benefits. In the intense ‘project-focused’ world of architectural practice it
is this trade-off between immediate but limited benefits and deferred but
more substantial benefits which becomes a recurring theme.
As the 1986 paper suggested, BIM depends on a particular
representation of architecture. BIM is based on the assumption originated
a decade earlier, that “a building is considered as the spatial composition
of a set of parts” [Eastman 1974]. Even in the pioneering days in the mid-
1980s during the development of the proto-BIM systems such as
RUCAPS, there were software developers who thought that BIM was
architecture and others who understood that architecture embraced a
whole range of ideas, intermediate representations and processes which
went far beyond the ‘assembly of components’ paradigm.
Central to this is to understand the difference between architecture
and building. Both are concerned with the realisation of a physical
artefact but with architecture, ideas and intensions of form and space
necessarily precede consideration of materials and making. The latter is
normally in the service of the former. So wWhile the BIM ‘component
assembly’ paradigm can usefully represent the material and physical
building, the consequence is that form and space become a secondary by-
product of the material model. BIM forces the creative user to select how
the building concept is to be decomposed into components (micro level)
before the concept can be expressed (macro level). The representation is
forcing the architectural user to reverse the natural thought process. It is
also forcing the use of precise dimensions too early in an uncertain design
process, thereby confusing precision with certainty.
This raises important questions: What is the effect of the
representation on creative thinking? Specifically, does the technology
expand or limit architectural originality? So while there can be few
arguments against the adoption of BIM as an efficient construction
delivery mechanism, we might want to recognise its limitations:
It is a technology. It is a methodology. It is not a philosophy of
design.
2. THE TRANSITION FROM BIM TO TOPOLOGY MODELLING
‘Beyond BIM’, there are important opportunities to re-establish the
creative freedom of architecture based on the direct conceptual modelling
of form and space.
Geometry plays a critical role in architecture. Geometry has such a
direct impact that it is not immediately apparent that there is something
hidden within geometry which if understood can be even more important
and fundamental to architecture, and that is topology. Indeed topology is
one of the principle ways in which geometric relationships can be
articulated in architecture.
The challenge is this: How can topology be harnessed as an
expressive and creative architectural representation in a way which is
completely accessible to architectural users? As with BIM, the
development and adoption of architectural topology will be a multi-
decade project with many false starts and blind alleys.
In the late 1990s (at Bentley Systems) we started to experiment with
the concept of an ‘idealised’ design model which was distinct from the
geometric representation of the material or physical design. We explored
a radical use of solid modelling techniques, not to directly model solid
objects, but more abstractly to harness the underlying topology to control
the configuration of other building components. In this approach the
components of the material model are ‘parented’ to the faces, edges and
vertices of the idealised topological model. Changes to the idealised
model could then be automatically propagated to enforce corresponding
changes to the material model. The underlying topology need not be seen:
what was seen was its effect. One of the first explorations of this direct
use of topology in architectural modelling was presented at a NASA
conference on Next Generation CAD/CAM/CAE systems [Aish 1997].
Over a decade later (at Autodesk) we had the opportunity to revisit
and extend the concept of the idealised topology model to include a more
complete non-manifold topology functionality and to create a fully user
accessible application class library. This enabled the spatial enclosure and
partitioning of an architectural concept model to be used as a direct and
lightweight input into Energy Plus. The advantage is that this lightweight
model allowed the user to bypass the need to build a BIM model prior to
energy analysis. In this context, a BIM model would be unnecessarily
detailed and arduous to change, potentially inhibiting the exploration of
alternative design options [Aish et al 2012].
In addition, the topology of the idealised model can be used to drive
BIM, to manage the connectivity of the components of the material model
and to programmatically define the detail fabrication geometry at the
interface between adjacent material components [Fagerström 2014].
There is also an interesting user generated video demonstrating the
DesignScript IDE driving non-manifold topology [anonymous 2012].
There are important differences between manifold and non-manifold
topology. A solid object can be represented using manifold topology
because it is defined by a single continuous boundary [a manifold] with
material on the inside and a void on the outside. It would also be possible
to represent the external envelope of a building using manifold topology
if there is a single continuous boundary separating the inside and outside.
However, there are other aspects of architecture which cannot be
represented by manifold topology, for example the partitioning of the
enclosed interior region into separate spaces. This is the role of non-
manifold topology. In addition to the single external boundary between
the enclosed inside region and the external void found in manifold
topology, non-manifold topology supports internal boundaries between
sub-regions which can represent spatial partitioning.
Essentially, non-manifold topology provides a number of new and
convenient ways of working and related benefits:
First, the ease with which the user can directly model the complete
representation of the building without having to consider
components or the component decomposition associated with BIM.
Second, the automatic propagation of topology changes to the
dependent material model. It is much easier to model the simplified
REFERENCES
© 2020 XXX-X-XXXX-XXXX-X/XX
idealised model and allow propagation to change the material model
than to laboriously edit the material model one component at a time.
Third, by building a topologically coherent model, the user now has
the facilities to make topological queries and discover relationships
within the model which can drive analysis and fabrication.
Nearly another decade later, and with the enlightened support of the
Leverhulme Trust, researchers at the Bartlett and the Welsh School of
Architecture, have completely re-engineered and extended the
architectural application of non-manifold topology to include mixed
dimensional topology and finally published an open source, cross
platform plug-in called Topologic [Aish et al 2018].
In this re-implementation we see not just the distinction between an
idealised and a material model, but the idealised model can now be the
foundation for a number of important analytical models, as follows:
Structural analytical model, using mixed dimensional topology:
A structural analytical model does not need to represent the full geometry
of the structural members based on the thickness or cross-section
dimensions, but only the connectivity of the structure, for example
between 1D linear elements such as columns and beams, 2D elements
such as slabs and 3D elements such as cores. Each type of element can
be idealised respectively as a 1D edge, 2D face or 3D cell. In order to
represent the structural analysis model, a collection of such elements
needs to be combined into a single topological system. However, this is
not possible with conventional manifold topology, where a model can
only represent a collection of lower dimension entities as the constituents
(or building blocks) of a single higher dimensional entity.
In a mixed dimensional model, instances of different types of
topological entities can be the constituents of higher dimensional
topological entities (as before) and other instances can exist as
independent entities all within the same mixed dimensional topological
model. In addition, all entities, even of different dimensions, are able to
participate in any Boolean operation and be the subject of topological
queries. This represents an important increase in the generality of
modelling over manifold topology and specifically enables an analytical
model of a typical structure to be represented and queried as a mixed
dimensional collection of entities called a "Cluster". The topology is a
lightweight representation of the connectivity of the structure. The
material dimensions (cross-sections and thickness) and other properties
are not explicitly modelled but abstracted as attributes attached to the
topology.
Energy analytical model, using non-manifold topology:
An energy analytical model does not need to represent the cross-section
dimensions of the building fabric (which can be abstracted as attributes),
but should represent: (1) the connectivity of the spaces, (2) how these
spaces are bounded by the building fabric and (3) whether the fabric is
between an internal space and the outside or between two internal spaces.
In manifold topology a single space can be represented as a cell and
the bounding fabric as faces, but buildings are usually a collection of
spaces and therefore cannot be represented by manifold topology.
However in non-manifold topology, a collection of cells can be modelled
as a CellComplex. Cells within the Complex can be queried to find
adjacent cells or shared faces between cells, thus supporting the exact
lightweight representation of architecture required in energy analysis.
Circulation analytical model using a 'dual graph’ of a CellComplex:
The spatial enclosure and partitioning of a building, represented as a
CellComplex, can be further abstracted as a dual graph. Cells can be
represented as nodes and the relationship between adjacent cells (through
shared internal faces) can be represented as the arcs of the graph. In the
energy analytical model (above), the shared face can be interpreted as
‘connecting’ the adjacent cells. In a circulation analytical model the same
shared face may represent a barrier to circulation.
In addition, an internal boundary within a shared face between two
adjacent cells can represent an aperture (such as a door). Apertures allow
separated adjacent spaces to be conditionally connected (for circulation)
without changing the underlying spatial topology, i.e. without unifying
the two adjacent spaces into a single space.
Prior architectural representations (such as drawings and physical
models) only had to be human readable and writable. However in design
computation, an architectural representation has to function as a
bidirectional channel of man-machine communication and therefore has
to be readable and writable by both humans and machines. A ‘machine
only’ representation can be succinct, but a representation for man-
machine communication may need to include additional features which
assist comprehension by the intended users.
We propose that non-manifold topology fulfils these requirements by
providing the most succinct description of architectural space and
materials which is also cognitively valid. It is a formalism which most
closely corresponds to an intuitive view of architecture including spatial
composition and the connectivity of construction. Non-manifold
modelling unifies exploratory -free form modelling with the data integrity
required to build consistent analytical models. It supports contrasting
ways of design thinking: expression, exploration, informed decision
making and reasoning about architectural space and construction.
Non-manifold topology can become the new defining architectural
representation with higher levels of detail being algorithmically
generated from this topology. Its most important influence on
architectural practice is to enable the lightest possible model to be built
with the least effort to provide the most feedback earliest in design.
3. CONCLUSIONS
Looking at the full range of transitions over the previous half century,
from manual 2D drafting, to BIM and, to topologytopology modelling,
we see important changes in design thinking and workflows; some
facilitated by technology and others which may appear to be imposed by
technology.
Although BIM brought advantages, there is a change of emphasis:
the manual effort of creating and editing drawings ‘line by line’ is no
longer necessary. In addition, changes to the model are automatically and
consistently propagated to all drawings.
Topology modelling brings many new advantages and a further
change of emphasis: the manual effort of creating and editing BIM
models ‘component by component’ is no longer required. In addition,
changes to the topology are automatically propagated to the complete
BIM model (and hence to all the drawings and other derived reports).
This suggests that a representation is not just a static externalisation
of design. In the development of new representations (above) we see a
progressive decrease in manual effort and a progressive increase in the
role of ‘change propagation’ [Fig.1]. The value of ‘change propagation’
is that it frees the architect from being over-committed to initial decisions.
By automating and lowering the cost of change, it has the effect of
completely de-stressing the design process, allowing the architect to
return to the primary roles as an exploratory designer.
Anticipating and managing ‘change propagation’ provides an
additional opportunity to explicitly think about design as a computational
process: to think more abstractly and more consequentially. It is this
‘cognitive retooling’ which distinguished pre- and post-computational
design and understandably may present a dilemma when retrofitting
computational design to established architectural practice. Whether this
is an opportunity or an obligation is precisely captured in a conversation
with an architect who had only recently been introduced to computational
design: “The advantage of your system is that it enables me to think about
what I am doing” …then a slight pause… “The disadvantage of your
system is that it forces me to think about what I am doing”.
There are a number of important similarities in all these transitions:
First, a new representation is often slow to be adopted in preference
to the continued uses of an existing familiar representation. However,
once established, a representation may become the new orthodoxy and
often inhibits and delays the adoption of future representations, so that
architecture always seems to be 180° out of phase with innovation.
Second, while a new representation may offer potential benefits and
significantly less ‘manual’ effort, adoption often requires some form of
‘cognitive retooling’. New concepts, terminologies and processes may
REFERENCES
© 2020 XXX-X-XXXX-XXXX-X/XX
have to be acquired by users. More importantly, previous ways of
thinking and instinctive ways working may have to be unlearnt. Together
these may be quite challenging in the hectic project-oriented world of
architectural practice.
Third, we see in these transitions the trade-off between immediate
but limited benefits and deferred but substantial benefits. Is this a
recognised pattern? Apparently. It seems amazing that the trade-offs in
the way representational technologies are adopted by architecture bears a
striking resemblance to the Stanford Marshmallow experiment [Mischel,
1970]. Your choice: One marshmallow now or two marshmallows later?
REFERNCES
AISH, R. 1986. Building Modelling: The Key to Integrated Construction
CAD. Use of Computers for Environmental Engineering related
to Building, CIB, 55–67.
AISH, R. AND PRATAP, A. 2012. Spatial Information Modeling of
Buildings using Non-Manifold Topology with ASM and
DesignScript”, Advances in Architectural Geometry 2012.
AISH, R. et al. 2019. Topologic: Tools to explore architectural topology.
Advances in Architectural Geometry 2018, 315–341.
ANONYMOUS. 2012. DesignScript Non-Manifold Topology video
https://www.youtube.com/watch?v=kjvFkjUMvpU&list=PLgU
HXWXukRCsm_e9z1QNEZ26m2pG8xDcD&index=20&t=0s.
CODD, E. 1990. The relational model for database management: Version
2. Addison-Wesley. ISBN 9780201141924.
EASTMAN, C. 1974. An Outline of the Building Description System.
Research Report No. 50. Inst. of Physical Planning, CMU.
FAGERSTRÖM, G., VERBOON, B. AND AISH, R. 2014. TopoFacade:
Envelope Design and Fabrication Planning using Topological
Mesh Representations. Fabricate 2014, 37–43.
FLEMMING, U., ERHAN, H. AND, OZKAYA, I, 2002. Object-Oriented
Application Development in CAD, ACADIA 2002.
