ConceptualDesignOfHighRises Gane Haymaker
ConceptualDesignOfHighRises Gane Haymaker
ConceptualDesignOfHighRises Gane Haymaker
1. Introduction
Conceptual design is a challenging part of the design process as it requires reconciling
the design with a complex set of goals and constraints such as program requirements,
construction cost, environmental performance, and more abstract notions such as
aesthetics and usability. When faced with such complexity, designers are taught to
generate and test a large number of options. However, faced with time and budget
constraints, Architecture Engineering and Construction (AEC) project teams today
often generate and test relatively few options for a design problem. Analyses of these
options are limited, and tend to favour space programming and aesthetics over other
criteria. Two possible causes of these problems lie in the tools the AEC industry is
using. First, these tools were designed to support generating single, static solutions,
they do not efficiently support exploring and managing many potential design
alternatives. Second, these tools do not yet integrate well with many analysis tools,
therefore discouraging multidisciplinary analysis. Potentially better building solutions
are neglected by the current inability of design teams to rapidly generate and analyse a
wider range of design options. This paper deals principally with the first issue:
generating design options.
Our research investigates whether parametric tools can help formalize the complex
rules that describe a high-rise and how might this improve the current design process.
Variables are the primary drivers of geometric variations. We distinguish between two
types of variables: independent and dependent. An independent variable is a user-
defined numeric input whose value can actively be controlled and changed (i.e.
triangle height, etc.) while the dependent variable is the output whose value changes
as a result (i.e. triangle area). Variables can also be global or local depending on how
these have been attributed to geometric elements. For example, by attributing a
variable to the radii of all columns in a building one would establish a global variable,
since modifying its value will propagate globally to all the columns. In contrast, a
local variable will always affect only a single geometric element to which it is
attributed.
Constraints help delineate the range of variations that a parametric model can sustain.
The extent of the range and the exact outcome of each geometric variation will be
determined by the type of constraints used in the geometry definition process. We
3
differentiate between two types of constraints: dimensional and geometric.
Dimensional constraints are equivalent to the above discussed variables, in which an
attributed numeric value acts as a constraint until it is modified. Dimensional
constraints are essential in defining the geometry of a design concept. For example
one might define an arc by constraining its radius, and length. Such constraints
establish a dependency of the geometric elements on the variable(s) that defines them.
Geometric constraints help determine how the geometric components relate to each
other, establishing another level of dependency. For example, one might use
geometric constraints to tangentially inter-constrain two arcs such that the final
definition of one arc depends on the other and vice versa. Geometric and dimensional
constraints work in tandem to enable the dynamic character of parametric models. A
summary of constraints and the types of geometric elements to which these apply is
given in Fig. 2. The symbolic representation of each constraint is derived from CATIA
– one of our research tools.
Figure 2 - Dimensional and Geometric constraints and types of geometric elements to which these apply
A PowerCopy (PoC) is a set of building components that are grouped and intended to
be used in a context. A PoC definition process may include specifying input variables
that allow this PoC to be instantiated according to the context into which it is pasted.
For example, one might define a column extrusion along a centerline with a height and
inclination angle; the latter two become input variables and when the PoC is pasted
into a floor with a different height and inclination angle, it uses the specification of
new variables representing that specific condition leading to an automated update of
the PoC geometry.
A Rule is used to construct a component from inputs. Input types range from numeric
and geometric variables to constraints and other components.
4
A design Theme is a collection of rules determining the geometric topology and the
global behavior of the parametric model.
Should the choice of geometric topology be opposite (i.e. curvilinear vs. rectilinear),
the nature of this rule would change and a different set of constraints would be used.
Should the choice of the theme be different (i.e. flaring vs. twisting tower), the core
rule would change, leading to a different set of constraints and variables being used
(i.e. rectangular vs. circular core may become optimal).
Together, these rules establish the building at the ground plane. Similar rules were
then used to construct descriptions of the building’s envelope, columns, beams, slabs,
walls, windows and shading fins. Following is a brief description of these rules:
• Construct Footprint PoC – assembles footprint and column profiles into a PoC
that is instantiated twice to create the mid and top tower footprints. The ground,
mid and top footprints as well as the column profiles are controlled by global
variables used to initially construct the ground footprint and the column profile
PoC. Tower twist variable is the only exception. It requires formulaic adjustment
for the mid footprint PoC instantiation to always be half the value of the global
variable (tower twist = tower twist/2) and for the ground footprint to always
maintain a zero value by subtraction (tower twist = tower twist - tower twist).
• Construct Tower envelope – creates the tower envelope as a surface by lofting
the ground, mid and top footprints.
• Construct Column extrusions – creates 3D extrusions from column profiles at
ground, mid and top footprints with tangency dependency on the tower envelope.
• Construct Single Floor Slabs – creates two slabs from intersecting the envelope
with two planes controlled by the floor height variable.
• Construct Single Floor Columns – slices column extrusions with the floor slab
definition planes and extracts an output variable calculating surface area.
• Construct Spandrel Beams – creates a spandrel beam outer footprint, offset from
the floor slab perimeter and controlled by the offset from slab variable. The inner
footprint is then offset from the outer footprint and its position controlled by the
width variable. The rule manages the beam extrusion height with the height
variable. An output variable calculating the beam’s volume is then extracted.
• Construct Glazing Single Bay – creates within a single window bay (space
between columns) a glazing position footprint dependent on the slab perimeter
through the offset from slab variable. The rule then divides the glazing vertically
into 3 equal panels by first creating a dependent variable (window length =
column spacing – column width) that feeds into another dependent variable
(vertical division = window length / 3). It then divides the window panels
horizontally into vision glass and shadow box. The division is generated with a
plane controlled by the spandrel beam height variable, which establishes an
explicit dependency to the spandrel beam. Two output variables are then extracted
calculating the vision glass and shadow box areas.
7
• Construct Fins – creates 3 fin panels within a single bay by using the vertical
division variable and then extracts an output variable calculating surface area.
• Construct Window Bay PoC – creates an assembly out of the single bay glazing
and fins, controlled formulaically by the selected PoC input variables. The
window bay PoC is then instantiated multiple times to fill in all the bays of a
single floor. Because of the twisting geometry, each column has a different
inclination angle, making the use of a window bay power copy highly effective.
The model supported efficient exploration of design options within the range of the
chosen theme. Since the tower had four unique floor types, rationalizing the exterior
wall as a single floor allowed rapid generation of new floors by adjusting the value of
the floor height variable (Fig. 5). In conjunction with structural engineers the model
was used to iteratively refine the design in response to the evolving structural solution
(i.e. modified columns spacing, column size, spandrel beam height and width). The
area calculations, as well as the glass and fin solutions were automated and each floor
type exported for assembly in traditional CAD tools.
Figure 5 - Low level parametric manipulations: a) typ. 3.7m floor; b) 6.4m mechanical floor – floor
height adjusted; c) spandrel beam and slab heights adjusted
The completed model took one person a week to construct. It became an essential part
of the design process with the client’s visit to the architect’s office. In two days the
design team was able to assemble and visually present three options based on the
client’s comments (45deg, 60deg, 90deg rotation and glass repositioned to exterior
face for maximizing sellable area) (Fig. 6).
Figure 6 - High-level parametric manipulations: a) 60deg rotation; b) 90deg rotation; c) 90deg rotation
with glazing pushed 200mm from exterior face of the columns from original 1200mm.
In retrospect, as constructed the model can respond to relatively simple changes in the
design theme (i.e. from twisting tower to vertically extruded by modifying the tower
twist value to zero or from twisting to leaning by modifying the distance to origin
variable in the mid and top footprints or from uniform twist to tapered twist by
modifying tower side length variables at the mid and top footprints, etc). However, the
model will not support investigation of alternate, more complex design themes (i.e.
blob tower) without more pre-rationalization in constructing the parametric skeleton to
include a wider range of the solution space – a challenging task in the usually fast
8
paced conceptual design process. Many of the underlying rules for the current twisting
tower theme will not apply for a “blob tower” and much of the parametric skeleton
and dependent geometry will need to be reconstructed with a modified set of rules.
Furthermore, with simple parametric techniques, the current model cannot easily deal
with modifying the number of bays or vertical division of glazing panels. To achieve
this level of flexibility, a parametric model must be complemented by custom
reasoning (i.e. scripting to allow the use of loops).
5. Conclusions
This paper presented a documented case study of a parametric model constructed for
one high rise building. More case studies and theory are needed to understand the
ways in which parametric thinking affects the design process. The case study
illustrates that parametric tools can enable architects to shift from creators of single
designs to designers of systems of inputs and outputs that generate design spaces.
Such tools provide a foundation to formalize design spaces in terms of explicit theme-
dependent rules and manage information in a unified environment. However,
designers need to understand and reformulate the design process in these terms, which
implies new ways of thinking and communicating about design. Parametric models
can provide important geometric flexibility and support an iterative design refinement
process. However, the choice of rules used to construct the model will limit its
usability to a specific range, often making extreme design variations unfeasible.
With the ability to rapidly generate design options comes the need to analyze them all.
To enable multidisciplinary integration, parametric tools can be used to construct and
manage layers of geometric information optimized for discipline specific analysis
tools (i.e. stick models for structural analysis, meshed geometry for CFD or Incident
Solar Radiation analysis, etc). As long as the dependency of each layer is established
to the primary parametric skeleton (i.e. in Infinity Tower case to the ground, mid and
top footprints and the envelope), the geometry update process can be automated.
Additional rules establishing the data schema and the information flow between
parametric and multidisciplinary tools, and methods to manage and optimize these
processes also need to be developed.
References
Alexander, Christopher, Ishikawa, Sara, Silverstein, Murray (1977): A Pattern Language. Oxford
University Press
Haymaker, J., (2006). “Communicating, Integrating and Improving Multidisciplinary Design and
Analysis Narratives”, JS Gero (ed), Design Computing and Cognition'06, Eindhoven, Netherlands,
Springer. pp 635 – 654.
Marques, D., Woodbury, R., (2006). “Using Rule Based Selection to Support Change in Parametric CAD
Models” Smart Graphics, Volume 4073/2006, pp 230-235.
Shah, J., Mantyla, M., (1995). “Parametric and Feature-based CAD/CAM
Kilian, A., (2006). “Design Exploration through Bidirectional Modeling of Constraints”, PhD Thesis,
MIT.
Shelden, D., (2002) “Digital Surface Representation and the Constructability of Gehry’s Architecture.
Ph.D. thesis, MIT
Stiny G, (1980), "Introduction to shape and shape grammars" Environment and Planning B: Planning and
Design 7 343-351
Whitehead, H., Burry, M., (2003). “Laws of Form”, “Between Intuition and Process: Parametric Design
and Rapid Prototyping” B Kolarevic (ed), Architecture in the Digital Age: Design and Manufacturing. pp
61 – 100, pp 147-162.
9