DrAFT: an Algorithmic Framework for Facade Design

Inês Caetano, António Leitão

INESC-ID/Instituto Superior Técnico

Figure 1 The selected case studies. First row: Library of Birmingham, in Birmingham, and the Yardmasters Building, in Melbourne; Second Row: Sheung Wan Hotel, in Hong Kong, and Hello House, in Melbourne: Last row: Formstelle Building, in Töging am Inn, and Precinct Energy Project, in Melbourne.

Architecture has always followed the times and their innovations and, currently, an architecture based on digital technologies has been emerging and has increasingly explored architectural facades.

In this paper we use DrAFT, a computational framework for the generation and exploration of facade designs, to explore a set of different examples of building skins.

Figure 2 Library of Birmingham. The set of algorithms used to produce the model: A. Units shape; B. Units Distribution; C. Two layers of different sizes and colours; D. Layered facade articulation.

DrAFT includes a classification of facades that helps in the identification of algorithms that best suits each design intent.

After combining the algorithms provided by this framework, the designer can more easily explore the solution space of the intended design.

Figure 4 Hello House model produced using DrAFT framework. A. Pattern unit; B. Unit distribution; C. Picture to control the bricks positioning; D. Final pictorial effect.

Nowadays, the architectural facade is characterized by complex shapes and patterns, mainly, due to the use of new design tools (Pell 2010) which promote further design exploration.

Figure 5 Precinct Energy Project model. A. Definition of the unit; B. Unit distribution in an alternated-grid. C. Pictorial effect with an image; D. Final effect, which was then subtracted from the facades’ surface.

The development of Generative Design tools, particularly, the use of algorithmic approaches, have had an important role in the generation of these contemporary skins because they simplified the design of complex and intricate architectural surfaces, which would not be viable to produce manually.

Figure 6 Formstelle building. A. Pattern unit; B. Unit distribution in an alternated-grid; C. Units size transformation; D. Facade with a perforated articulation.

In addition, they also increase the design efficiency and their evolution has been changing, not only the design process, but also the architectural thinking (Kolarevic 2003).

Figure 7 The Yardmasters Building model. A. Units geometry; B. Units distribution in an alternated-grid; C. Creation of the windows; D. Applied articulation, with the Islamic pattern on the facade’s surface; Below: the Yardmasters’ final model.

Unfortunately, algorithmic approaches do not make facade design trivial. On the contrary, they require the rigorous specification of all algorithmic steps, a task that requires specialized knowledge and that, in many cases, can be quite complex.

Figure 8 An instance of the pattern produced using the set of algorithms above. Figure 9 An instance of the pattern produce using the algorithms above, which now include units with a different shape (pyramidal) and a distribution in an regular-grid.

In this paper, we propose a computational framework designed to simplify the algorithmic specification of facade designs.

In practical terms, DrAFT – Draft Algorithmic Facade Tool – promotes the exploration of facade designs and simplifies the adaptation of the generated models to the everchanging design process conditions.

Here, we present a collection of examples developed using this framework, thereby demonstrating its usability and flexibility in facade design, and also showing other possible applications.

Figure 10 On the left: a graphic with the areas maximized for each smoothness value; On the right: the selected final model

Generative design (GD) is a computer-based approach to design that creates shapes through algorithms (Terzidis 2003).

Algorithmic design is a process that explores complex forms from simple and iterative methods/rules while preserving specified qualities (Meredith 2008).

For this, architects produce an intermediate algorithmic-based description of a design rather than its shape (Leitão 2013).

Figure 11 Re-execution of the same model in different backends (AutoCAD, Rhino5, SketchUP and Revit)

Parametric Design is a specific GD approach that generates different instances of a design, where each instance represents a particular set of values for the design parameters (Barrios 2005), allowing the designer to freely explore a large solution space of the design briefing/program.

Therefore, this allows architects to continuously evaluate several solutions, which would be difficult to do with traditional design methods.