Part I
Architectural Design and Theoretical Tenets
This part incorporates multiple facets, ethos, and relevant meanings of the multidisciplinary framework of todayâs architectural design. It also expands on the meaning of cognitive design thinking, and the augmenting role of research trajectories, speculating on potential paths for design methodology.
The focus of this part is on the current duplicitous realms of design and theoretical tenets of architecture; one to be disciplinary and cultural, and the other, interdisciplinary, meaning technological and scientific, determining the vast boundaries of the design domains of this time.
1
Design Parameters to Parametric Design
Patrik Schumacher
Parametric design is a computer-based design approach that treats the geometric properties of the design as variables. The dimensions, angles and geometric properties (like curvature) remain malleable as the design progresses. Although at any time the âparametric modelâ displays a determinate shape according to the set of currently chosen values, the essential identity of the parametric design resides in the malleable objectâs topology rather than its momentary determinate shape. This means that the design consists in the relationships that are maintained between the various elements of the composition. In fact the parametric design model is conceived as a network of relations or dependencies. This way of building up a design has the important advantage that the build-up of complexity and the detail resolution of the design can progress while simultaneously maintaining the malleability to adapt to changing requirements as new information is fed into the design process. The generation of alternative options remains viable and economical deep into the detail design without requiring abortive modeling and drafting work. This parametric malleability is advantageous both for the sake of continuous design adjustments as the design progresses, and for the sake of the generation of options and variations. The parametric model can be conceived as general building plan or geno-type for the generation of many different versions or pheno-types that might co-exist (rather than substitute each other as options). Optioniering thus leads to versioning. Mechanical repetition is being replaced by mass customization. Versioning might also be applied within a single building design via the versioning of components, via âgenerative componentsâ. The components adjust their individual shapes in relation to their placement within the encompassing model. These components are small parametric models, i.e. sets of interdependent parts with adjustable shapes. The component adapts to (and fits into) local constraints via the adjustment of its internal parameters. For instance an array of façade componentsâcomplete with glazed openings, frames and fixing detailsâmight be made to populate the surface of a volume with changing curvature. The components are to be set up in such a way that they auto-fit to the surface. Each component will assume an individually fitted âpheno-typicalâ shape, on the basis of the same underlying âgenotypeâ Thus parametric design is a powerful methodology to achieve a new architectural morphology, namely a morphology of continuous differentiation. However, the potential for such differentiation is not confined to the achievement of scaling and geometric fit with respect to complex forms with continuously changing surface curvature. This kind of differentiation might also be driven by performance parameters like structural parameters or environmental parameters like sun exposure or wind-loads. For instance the opening within a façade panel or the shape of a shading element might vary according to the differential sun exposure of a curved façade at each point of its surface. The parametric designer might set up the following dependency: the higher the sun exposure of a certain surface patch, the smaller should be the opening of the façade component at this location. A sun-exposure map imported from an environmental analysis tool might then deliver the data input for the component differentiation. The sun-exposure map is thus being âtranscodedâ into a differentiated field of façade panels that âoptimizesâ the sunlight penetration within brackets set out by the parametric design. The resultant façade articulation is thus a function, mapping or indeed a representation of the façadeâs differential exposure to the sun.
Figure 1.1 Transcoding of sun exposure map into a differentiated brise soleil pattern. Study for Soho China Wang Jing towers, Beijing by Zaha Hadid Architects.
Similarly, a designed architectural volume might be structurally articulated via the transcoding of structural analysis parameters into differentiated geometric components. For this purpose the results of a finite elements stress analysis might become the input for a framing pattern that differentiates either member density or member size or both. Again, the result achieves a relative structural optimization (if compared to an undifferentiated framing pattern) and the differentiated structure represents the underlying stress distribution.
Figure 1.2 Studies in the translation of primary and secondary stress lines from structural analysis software Karamba into a configuraion of ribs reinforcing a slab with two linear supports and two point loads. Philipp Ostermaier, Zaha Hadid Architects.
Figure 1.3 Studies in the translation of stress lines from structural analysis software Karamba into a configuration of a tubular skeleton for high-rise structures. Philipp Ostermaier, Zaha Hadid Architects.
Thus in a tall building a parametrically designed skeleton responds to and displays the differentiation of structural forces. Both compressive stresses due to the accumulating vertical loads as well as the moments due to horizontal wind-loads accumulate at the bottom of the tower which will thus be rather different from the middle and top of the tower respectively. The respective variation of performance parameters of the various subsystems of the building like envelope and skeleton thus translates into the morphological differentiation of these subsystems.
The way performance parameters might be transcoded into morphologies is an open question that calls forth the creative designer. Further: these subsystemsâeach adaptively differentiated according to its own performance logicâalso might adapt to each otherâs differentiation. We might talk about subsystem âcorrelation.â To the extent that the envelopeâs differentiation is responsive to the skeletonâs differentiation according to a rule it becomes its âmappingâ or ârepresentation.â The particular rule or mode of correlation is again open to design invention. The same principles of adaptive system differentiation and multi-subsystem correlation might be applied to urbanism which thus becomes âparametric urbanism.â The initially considered subsystems here might be the circulation system (road network), the building fabric (massing) and the programmatic distribution (land use). The existing topography (topo-map) as well as the pre-existing roads might serve as underlying input data sets to be transcoded into a differentiated road network. The differentiation of the urban massing might initially follow its own logic of block differentiation, initially conceived as internal product variation without as yet responding to external data inputs. This internal differentiation could in a second step be âover-codedâ or correlated with the differentiation of the circulation network according to a certain rule. The fabric differentiation might be further adapted with respect to an agenda of morphological affiliation with the adjacent urban context. Each step requires the invention of a rule of differentiation or adaptive correlation. At the basis of these differentiations and correlations are the chosen geometric âprimitivesâ (or components built up from those primitives) with their respective variables and respectively chosen degrees of freedom.
Figure 1.4 Zaha Hadid Architects: Soho Galaxy, Beijing 2012. The project demonstrates the advantages of using parametric curves. Both volumes and urban spaces are easily identifiable, facilitating orientation and navigation despite the density and complexity of the urban scene.
Parametric design thus delivers a new powerful adaptive capacity to architectural design. This new capacity opens up a new domain of creative design invention, namely the invention of transcoding rules and rules of subsystem correlation. Design thus becomes ârule-basedâ design. Critics unfamiliar with this new world of parametric design sometimes presume that the new algorithmic design operations, somehow replaces or dis-empowers the designerâs creative freedom. The opposite is the case: a new realm of creative exploration with its new design challenges is opened up and calling for the designerâs creative ingenuity. The more computational design tools free the designer from the drudgery of drafting and modeling, the more the creative essence of the design process as a process of invention and decision making comes to the fore.
To design is to generate and to choose. All design is decision making, i.e. the making of choices. Choice presupposes a set of alternatives to choose from. The design process thus comprises two fundamental sub-processes: the generation of alternative solution candidates and the selection of an alternative according to test results on the basis of posited evaluation/selection criteria. Thus, the overall rationality and effectiveness of a design process depends on two principally independent factors: its power to generate and its power to test/select. The design principle of âgenerate and testâ conducted in a design medium or model in advance of physical construction stands in as economic (rational) substitute for the physical âtrial and errorâ process that is the principle of the biological evolution as well as of all pre-architectural building, i.e. building without architectural design aided by drawing. Both powers of design rationality are being massively enhanced by the computational aids that constitute parametric and algorithmic design in comparison with traditional design based on drawing according to precedent or intuition. The more the processes of generation and selection are themselves automated via algorithms, the more powerful does the design process become, as the designerâs creative choices shift to the meta-plane of choosing generative algorithms and evaluating selection criteria. These in turn might be looped into evolutionary algorithmic set ups. Parametric design and design via scripted rules is replacing design via the direct manipulation of individual forms. Computational processes can uniquely enhance both the design processâs generative power and its analytical power. The techniques of variation and versioning as well as the differentiation on the basis of transcoding and correlation advance the parametric designerâs efficiency as well as the rationality of his design.
The generation of design options can be opened up much further than the mere versioning proliferation of pheno-types on the basis of a pre-established geno-type. A much more open ended, generative technique of producing solution candidates is via an agent-based system whereby the elemental primitives (atoms) of a composition or multi-primitive components (molecules) are set free to roam within the modeling space where they aggregate and configure larger global structures according to local rules of attraction, repulsion, alignment, attachment, etc., set by the designer. Many of the properties of the resultant configuration are emergent and un-anticipated outcomes of the complex interaction of the rules. Prediction can only mean pattern prediction here in terms of general qualitative properties or in terms of quantitative brackets but hardly precise anticipation. Genuine surprise is possible. Some undesired properties might be prevented by giving the generation process respective constraints. Certain desired properties might be attainable in ways and to a degree that would have been difficult or impossible to attain via intuitive methods. Agent-based processes open up a huge field of exploration and arena for the designerâs creative ingenuity. They can also be used in the agenda of multi-subsystem correlation described above. A structural skeleton or an urban path network might be configured via agent-based aggregation processes. Urban fabric particles (agents) might interact and configure over the substrate of a topographic map that biases the migration and self-organization process of the agent population in ways that produce a transcoding of the underlying topography not unlike the more direct transcoding via a simple rule of correlation. The result of the agent-based model might display many unexpected variants and properties that might or might not be advantageous upon further analysis. The general advantage of these less predictable processes is that they might deliver in-built criteria in new, unexpected ways and offer up unusual properties that might stimulate the designerâs formulation of altogether new desires and criteria. However, the legibility of the transcoding as representation might be compromised relative to the technique with direct rules of correlation.
Parametric Software
By far the most widely used parametric design software in architecture is âGrasshopperâ developed by David Rutten for Robert McNeel Associates (founded in 1980, McNeel is a privately-held, employee-owned company based in Seattle) and first released in 2008. Grasshopper is a freely available graphical associative logic modeler and algorithm editor closely integrated with McNeelâs 3-D modeling tool âRhinoceros.â Grasshopper is a pertinent tool for the set-up parametric models as described here as networks of interdependent elements and manipulations. The network of relations is set up and visualized graphically so that the designer can keep track of and intervene in the relational network he is designing. The parametric designer usually opens two programs/windows: the 3D modeling space of Rhinoceros and Grasshopperâs graphical algorithm editor. The designer can now move between the modeling and the scripting environment to build up the parametric design, e.g. creating objects in Rhino, make them interdependent and manipulate the parameters in Grasshopper and then view the behavior of the dynamic interdependent configuration in Rhino, etc. Grasshopper might become the primary medium and site of the design work while the 3D geometric model visible in Rhinoceros (passive visual control) is driven or executed by the active definition/script visualized and manipulated in the grasshopper window. That the design is all about the set-up of topological parametric geno-types defined via networks of relations (both internal to the building/artefact and external in relation to context parameters) is thus evident in the constitution of the primary design medium. That indeed most parameters/values are treated as variables is evident in the ubiquitous use of sliders (with designated ranges of values).
Rhino/Grasshopper has also become the preferred platform for scripted plug-ins and for a new powerful set of integrated tools that push architectureâs design intelligence beyond the mere handling of geometry to include engineering logics and real time access to physics simulations that allow for sophisticated form finding and optimization processes to be seamlessly folded into the design process. Kangeroo is a physics engine created by Daniel Piker as a tool for interactive real time structural form-finding simulations like surface-relaxations. These simulations are implemented via particle-spring systems. With this particular tool Frei Ottoâs seminal physical form-finding experiments with tensile structures and shells via inverted catenary systems can be recreated in the much more versatile digital domain. Frei Ottoâs models represent but a small corner of the new space of possibilities that is put at the fingertips of parametric designers in a very intuitive, playful way that equals the intuitive play with real physical materials, however now unleashed from the narrow parametric bounds given by any chosen physical material. Karamba and Millipede are structural analysis and optimization tools for Grasshopper. They are interactive, parametric finite element analysis programs that display stress distributions and deformations of any geometric form under any imaginable load. Karamba was (and continues to be) developed by Clemens Preisinger and Robert Vierlinger a.o. within the structural engineering office Bollinger- Grohmann-Schneider. Millipede was (and ...