6. Parametric and generative design
Parametric design enables the designer to define relationships between elements or groups of elements, and to assign values or expressions to organize and control those definitions. It is usually applied within a three-dimensional CAD program. However, unlike the limitations raised earlier as a result of the independent nature of elements in the design, the underlying principle here is of connectivity and relationship. Parametric design addresses the constraints of traditional CAD operations by supplanting the designer’s direct interaction with the design elements—adding, subtracting, copying, etc.— with the development of a series of relationships by which elements connect and build up the design. The designer may, at any time, alter the values or equations that form the relationships between elements and the effects of these changes will be incorporated into the system, which reflects them visually. On this latter point, it should be stressed that not all developments will result in a perceptible, i.e. visual, change to the design. The relationships are subsequently edited as the designer observes the effects of the revisions, as the connected system of elements evolves and the desired results are chosen based on relevant “performative” and aesthetic criteria. In addition, complex assemblies of elements may be grouped or “collapsed” together to form a new, customized element defined by the designer. This process may undergo further iterations, allowing the designer to develop bespoke variations that may be used in future projects. Although often perceived as a relatively recent development, parametric design was actually one of the early concepts in CAD. Ivan Sutherland’s 1963 PhD thesis, Sketchpad: A Man-machine Graphical Communications System, proposed the first graphical user interface to enable a designer to draw on the computer and effect changes to the design parametrically. Despite the technology of the period limiting the full potential of this system, Sutherland foresaw the advantages of this type of interface: “A display connected to a digital computer gives us a chance to gain familiarity with concepts not realizable in the physical world. It is a looking glass into a mathematical wonderland.”10
As with any design tool, there are positive and negative aspects to parametric design. The primary advantage is twofold: once the relationships have been established the system may run autonomously within its parameters and explore novel solutions that may not be apparent to the designer; and, as the design is always kept consistent with these parameters, the designer has greater opportunity to explore them without time-consuming reworkings. However, it is in the very task of establishing all the relationships within the parametric system that the main disadvantage is to be found. The parametric-design process is initially very time-consuming, particularly for the inexperienced, but perhaps an even greater challenge is the shift in mindset it requires, as Robert Woodbury explains in Elements of Parametric Design: “Parametric design depends on defining relationships and the willingness (and ability) of the designer to consider the relationship-definition phase as an integral part of the broader design process ... This process of relationship creation requires a formal notation and introduces additional concepts that have not previously been considered as part of ‘design thinking’.”11 This is, of course, potentially exciting and rich territory, as parametric design and this attendant change in the user’s behavior may further the design possibilities by enabling ideas to be developed in an explicit, effective manner—thereby contributing to an understanding of the process and improving methods of its communication.
Future Cities Lab’s Vivisys installation is an experimental double-curved acrylic-lattice vault that responds to interactions within its environment. To achieve the complex geometry, the initial surfaces were parametrically modeled using GenerativeComponents software while the later files for fabrication were generated using Rhino.
Let us therefore consider what a parameter is. It is a quantity that is constant in a specific scenario but may vary in other situations. While sometimes seen as a limit or constraint, a parameter is neither of these but does possess a value. In basic design projects there would be negligible benefit in using a parametric system, since it would not be very time-efficient. However, where a situation has complexity, and thus many different parameters, this type of design generation is extremely useful as the practitioner is able to integrate all the different aspects into one large database and manipulate it accordingly. The latter issue, concerning the generative properties of this approach, is important since one of the key features of parametric design is the ability to describe the design as a series of relationships that may be used to iterate further versions. Here we see a principal distinction between physical modelmaking, or even traditional CAD software, and using a parametric design program. In the first two modes of inquiry, every different permutation requires a new model, whether physical or digital, to be made—or at least the original one to be extensively modified through disassembly, editing, and reconfiguration. By contrast, parametric design allows multiple options to be generated within values specified by the designer. That is not to imply that one approach is more successful than another, but clearly the time and effort in the first two methods is invested in the making of the designs themselves whereas, by contrast, in the latter, generative approach the designer directs time and effort into making the system that will subsequently reiterate the designs. This phenomenon is what Mark Burry refers to as “designing the design.”12 To understand this way of thinking and designing it is helpful to consider algorithms for architectural design in broader sense.
The design development of Dragonfly by EMERGENT/Tom Wiscombe in collaboration with Buro Happold necessitated the use of generative design software to explore the honeycomb patterns analogous to those found in dragonflies’ wings. These patterns demonstrate rule-based interaction in relation to cell shape, depth, and density, and were investigated and optimized in response to these parameters (top and middle) and pattern configuration as illustrated in the Voronoi tiling distribution (above).
Case study Parametric urban design
Zaha Hadid Architects—One North Masterplan, Singapore, 2001–21.
This masterplan echoes a consistent theme within the designs of Zaha Hadid Architects for more than two decades: developing an urban architecture that uses the spatial repertoire and morphology of natural landscape formations to offer rich territory for public programming, identity, and flexibility. The proposed morphology is envisaged as a system that accommodates variation while retaining a strong formal coherence throughout. The gently undulating overall form accommodates a wide range of built volumes and public realm, in tandem with infrastructure and connective tissue with neighboring urban districts. Rather than using strictly Platonic geometry, the design was developed parametrically as a free, curvilinear, and malleable series of deformations, pliant yet resilient to the variety of forces and flows that occur across a city district.
A The original concept painting illustrates the masterplan’s dynamic, free-form geometry.
B A design-development model, laser-cut from acrylic glass to investigate the urban system as a series of landscape outlines.
C, D, E, and F The emergence of the design as urban formation through a series of physical study models (left) illustrates the project’s gradual development until such a point that the design is consolidated as mirrored in the final diagram for the urban fabric (right).
G The conceptual mass/form, as developed for implementation following an intensive process of design optimization and generative development using parametric software.
Case study Parametric design of 1:1 structure
EMERGENT/Tom Wiscombe—Dragonfly, SCI-Arc, Los Angeles, 2007.
A The geometrical evolution of the project from a trabeated structure, via a two-way plate and then honeycomb plate to the dragonfly composite. This development achieves an emerging structural hierarchy with localized inplane stiffness owing to the quad cells, localized flexible infill via the honeycomb cells, and the adaptive response to indeterminate force flow.
B The dragonfly geometry scaled to the plan of the exhibition space. The cell size and density of the design are controlled by the boundary condition, the “veins” taking on complex shapes in relation to force flow.
C Screenshot from Digital Project software, illustrating the design in a parametric scenario wherein the model may be manipulated and the resultant effects observed across the entire design.
The concept of the dragonfly wing, unique in its structural performance and exquisite formal variation, informed this installation. Nature offered a rich precedent for biomimetic design—in terms of formal and behavioral features rather than merely aesthetic considerations. In contrast to the wings found in nature—which respond to aerodynamics, lightness, and mechanical functions—this installation was determined in relation to parameters including gravity, specific support points, and flat material properties. The designer was keen to view the project within a broader context of research into cellularity in architecture as a departure from its pure form toward a tectonic based on emerging structural hierarchies within cellular aggregations.
In collaboration with Buro Happold, “populations” of random structural mutations were generated and fitness-tested based on the given support and loading data. These conditions were then run through a feedback loop comprising multiple generations, and the geometry evolved toward performance criteria and novel variation. Formal coherence was balanced with structural legibility in the choice of mutations.
The fabrication techniques for the installation reflect the adaptive model produced during the design process. Firstly, a CATIA model was generated to parametrically connect hundreds of two-dimensional unfolded bands to “live” three-dimensional geometry. Through the evolutionary nature of the design plus the addition of engineering data such as scoring, bending, and drilling, the information related to these bands was automatically updated via the “live” model. A nesting algorithm enabled the bands to be distributed onto standard 4 x 8-inch (100 x 200mm) aluminum sheets so as to reduce waste material. The sheets were inscribed and cut using CNC milling machines. The embedding into the bands of assembly data for the installation structure, including relative cell position and bending angles, ensures that the construction of Dragonfly is a bottom-up process without the need for conventional documentation.
D Von-Mises analysis, illustrating equivalent stress across the arrangement of cells.
E Development of cell connections as bespoke elements within the structure.
F Detailed view of the parametric model, showing cell positions, connection points, and bending angles.
G This information is used to assemble the physical installation. Reference data is clearly inscribed on the cut components for those constructing it.
H The final installation, seemingly defying gravity through its innovative structure and evolutionary geometry.
