Shaping up nicely
Unless you are Foster or Grimshaw or one of a few like-minded practices, the whole business of parametric design probably seems a bit esoteric, something to do with designing blobs, a kind of mathematical post- or perhaps prehoc justification for doing architecture with funny shapes. Well, that is partly correct. And that is also what is so interesting about architectural parametrics.
It is probably legitimate to think of the current MicroStation/AutoCAD phase of CAD as its horseless carriage stage - think of the first cars which looked like broughams and landaus and were steered by a tiller, and then of how current CAD mimics the orthogonal, Euclidian rigidity of the T-square and set square. If you think of CAD in those terms, parametric design looks set to be part of future architectural computer modelling.
You have to make the distinction about architectural modelling because parametric design is well advanced in such areas as engineering and industrial design, especially aerospace and automotive design in which complex three-dimensional shapes often have to be analysed. Architectural CAD software houses have begun to catch on to the possibilities, and parametric design is sometimes used to describe the linking of data - such as mass, colour and temperature range - to specific elements in an architectural drawing.
But interest in the topic has yet to reach a critical mass and so what is available is expensive and tends to be adaptations of software already available to other disciplines. There is nothing particularly wrong or unusual with that: old software designed for solving such diverse problems as the topology of groundwater and the skinning of jet fighters (the origins of Dassault's CATIA) has recently been adapted for use by architectural designers.
So what is parametric design and modelling? Firstly, a parameter is not just a defined limit, as in the common but strictly incorrect 'What are the parameters of the problem?' A parameter is more complex than that. It is a variable to which other variables are linked, the variables of which can be defined by mathematical formulae. A parametric expression of design relationships, as British CAD guru Peter Szalapaj puts it, 'can represent a set of geometrical forms instead of just one'.
Action stations The example most commonly cited is Grimshaw's Waterloo terminal. where the array of three-pin arches - made up of bowstring trusses with tension rods on the outside of the larger trusses and outside on the smaller trusses - had not only to curve on plan but also to change profile to provide clearance for trains. By establishing the basic, but very detailed, rules for a single arch, it was possible to develop the others parametrically based on the key span parameter.
Designing all the individual arches by hand and then probably doing it over again several times would have been very laborious. Had the basic arch model been changed, all the other arches would have had to change too.
Because dependent shapes and forms can be modified by changing the original model, and because this can be done mathematically, designers can readily explore the implications of a variety of shapes. It is easy enough to do this in sketch form for both orthogonal and non-orthogonal shapes but it is very laborious to do it with any precision for non-orthogonal three-dimensional forms.
A recent trend has been to use generative techniques which deploy parametric modelling, scripting and genetic algorithms. Bentley has developed Generative Components, which has been released to selected designers and is due to be beta-tested at the end of this year. Greg Lynn, father of Blob, has been playing around with it. He has been reported by KPF London's Lars Hesselgren to have used it to deal with very large datasets involving as many as 30,000 components. These components are parametrically controlled by a set of variable trusses that are driven by adding escalators to the facade in question.
'Think Centre Pompidou, ' Hesselgren says, 'married to a snake swallowing an escalator and you get the idea.' According to Bentley, designers and engineers can work in parallel - the architect generating the components and their configuration while the engineer can read the design geometry and, ultimately, the building performance as well as the basics of digital fabrication.
Autodesk is still hovering on the sidelines, arguing that Revit's family editor has always allowed the creation of parametric components without programming, although the use of formulae seems necessary for more speculative design.
Parametric design is in its most interesting phase: a recent conference saw demonstrations of the development of a parametric dome based on the Fibonacci series, space-planning tools and interactions between design moves and structural limits. A body of theory and comment is emerging with a group including, apart from those mentioned above, Mark Goulthorpe at MIT, Mark Bury at the Royal Melbourne Institute of Technology, Neil Leach at Bath, Huw Whitehead at Fosters and Kristina O'Shea at Cambridge.
O'Shea has developed her own program, eifForm, which has already been used to design a structure. 'This is the first architectural structure built where both the form and related structure were generated by a computer via design parameters and conditions, rather than explicitly describing geometry, 'she says. 'The programme generates new structural forms that intelligently respond to given design conditions using a combination of structural analysis and optimisation.' With this in mind, maybe architects need to get to know about parametric design sooner rather than later.