The AJ takes a look at how Computational Design and Optimisation can underpin design and efficiency across a range of architectural projects Thanks to computational optimisation, architectural design can now rely more on hard facts and less on intuition. Though that may sound distasteful, it can be incredibly liberating. So say architect Gianni Botsford and Arup's Andrew Sedgwick. Parameters as varied as views, programmatic adjacencies, daylight levels, thermal efficiency and costs can be incorporated into complex computer models which, like an ant colony emitting pheromone to find the optimal route through unknown terrain, identify solutions which best meet the objectives established by the design team.
Once a computational solution set has been built, alternate designs can be explored by varying the parameters. While this approach has been widely used in the automotive, aerospace and boating industries, it has only recently taken hold in building engineering.
A handful of examples illustrate how this can revolutionise architectural design, and particularly the way engineers input into the design process.
After pioneering work on Gianni Botsford Architects' Light House in Notting Hill, west London (AJ 15.12.05), where three-dimensional computational studies of daylight levels across the site were used to define parts of the brief - such as the best place for the garden on a tight urban site, Arup is now applying Computational Design and Optimisation (CDO) on a variety of schemes in different stages of design. Initially this approach was used for detail design such as determining structurally effcient bracing for Kohn Pedersen Fox Architects' DIFA Tower or optimising panel layouts on a curvy building envelope without sacrificing design intent. Arup also developed CDO tools to optimise building-envelope design using often conicting lighting and energy-performance criteria. When detail design variables are combined with cost information, the power of this approach to achieve time and costs savings is indisputable.
What is more interesting and what lies at the heart of Gianni Botsford's design approach of 'local adaptation', which he presented recently as part of the Architecture Foundation's Winter Nights series, is the use of CDO to inform design at the outset of a project. Sedgwick explains that CDO 'enables an understanding of a project's constraints with no presupposition about its form and increases confidence about finding the fibestfl solution'. What was at the outset primarily a two-dimensional tool for optimising building envelope has become three-dimensional. A building or a site can be divided into voxels, three-dimensional pixels with bits of data attached. The design team can develop an optimisation model which encompasses client, architectural, engineering, fabrication, construction and cost parameters to generate and evaluate thousands of design alternatives. The results are graphed three-dimensionally as a parato-optimal set or 'point cloud' of solutions that provide a starting point for further design work.
Arup worked with LAB FAC architects to find the 'bestfit' solution for their Paris media centre design. Arup developed an algorithm incorporating 14 design variables which searched over hundreds of millions of designs to find 14,000 pareto optimal solutions - each of which, for some performance parameter, gives an optimal result. Sliders for each variable could be used to prioritise different parameters in order to examine alternate designs. Similarly, for a masterplanning project for Parc Rendel near Fréjus in the south of France, Botsford overlaid variables such as solar orientation, areas of protected ora, and wind exposure to identify buildable sites.
In a more ambitious application early in the design process, Botsford teamed up with Arup to use CDO for a competition entry for the Estonian National Museum in Tartu.
The basic programme requirements, along with their appropriate proximities, environmental control, daylighting requirements and views, were used in an optimisation analysis to generate the geometrical form of the building.
Sedgwick explains that CDO is turning the way engineers design on its head. He refers to the old way of working as 'brute force'. Architects would draw a scheme, then the engineer would test it and the scheme would be revised. Now, rather than testing a solution and redesigning, design parameters are used to write an objective function which defines a range of optimal solutions.
CDO is replacing traditional hand calculation methods with computer models. On the Light House, the computation model was revised manually during each iteration of the design as Botsford prioritised the variables. Each round could take a couple of weeks. Now, with the development of more sophisticated modeling techniques, what took several weeks to do a few years ago can be done in an hour.
Arup Lighting has used CDO on several projects recently.
CDO has been used in detail design with Allies and Morrison to determine the bafe design for cove lighting in the redesign of King's Cross underground station. For Renzo Piano's Los Angeles County Museum of Contemporary Art, Arup used CDO to establish the optimal profile for north-facing light scoops in the roof. Arup Lighting engineer Giulio Antonutto-Foi says: 'The beauty of [CDO] is that you don't know the outcome - it's a very Zen approach.' Arup is also using CDO in masterplanning studies to determine building form based on a variety of parameters.
Interestingly, to date, wind has defied CDO models because, according to Sedgwick, wind simulation 'takes too long'.
The appeal of CDO for architects is that it provides an objective basis for design, but it is in no way a replacement for design itself. The architect must control the subjective process of selecting and weighting the parameters. With the input of the design team, the architect can overlay the brief with data, in order to - in Botsford's words - 'interrogate the space'. It's like starting with a primal soup of data points from which the solution emerges.
This is a 21st-century soup whose recipe includes developments in algorithmic design, 3D parametric and associative geometry, performance-based criteria, and whose chefs are an increasingly computer-literate generation of architects and engineers.