Leeza SOHO — walking the completed building
Six minutes in the finished tower, and how the coordination model was tested in mock-up, fabrication and construction.
From computation to construction
I worked on Leeza SOHO from the competition in 2013, initially as design lead, through to completion in 2019, becoming project architect as the project developed and relocating to Beijing to carry it through design development, documentation and construction. The project became an unusually direct encounter with the consequences of computational design: the same geometric systems that controlled the digital model eventually had to be translated into façade components, structural assemblies, fabrication processes and construction sequences.
As Project Architect I worked with consultants, the client and contractors through design coordination, façade mock-up reviews, value engineering and site supervision. Those encounters were continuous feedback between the digital development of the building and its physical realisation.
A computational model became the project's coordination instrument.
The project was developed through a large Grasshopper definition linking the primary geometry to a range of downstream systems. The sequence below shows one aspect of this process: the basic geometry is generated and manipulated while the façade system responds to changes in the underlying geometry.
The main geometry was developed between two operators. I built the overall design, the Grasshopper system, and the detailed models of façades FS1 and FS2. Kaloyan Erevinov modelled FS3 and FS6 in Digital Project. Pavilion façades, lobby interior and landscape were only partly automated and mostly modelled by hand. Those workflows were then combined in a master model for coordination with the client, local architect and consultants.
The FS3 façade illustrates why different approaches were necessary. Its geometry was generated from the intersection of the atrium surface with the core volumes, producing a less regular geometry that was considerably more difficult to automate and detail systematically.
In a project of this scale, automation did not eliminate specialist modelling; it changed how the available expertise could be organised. Maintaining computational systems while requirements evolved was itself a substantial task, with new requirements regularly forcing established relationships to be adapted or rebuilt. The practical limit was therefore not simply what could be automated, but how much complexity a small number of people could develop, maintain and verify.
From digital model to physical system
The digital surface was not itself a construction method. Each façade system imposed different constraints on panel size, curvature, tolerances, assembly and installation.
Façade System 2 (FS2), for example, used flat glass panels that were cold-bent on site. The subdivision of the surface therefore became a question of manufacturing and installation rather than purely geometric resolution. At the same time, achieving a lightweight and transparent structure required the façade, its support structure and the installation sequence to be considered together. The resulting system suspended the façade from bridge trusses while transferring lateral loads to the adjacent main columns.
The project also involved physical testing at scales where the digital model ceased to be sufficient on its own. FS1, FS2 and FS3 were tested in large-scale mock-ups before tendering. These tests informed details ranging from the final colour and appearance of the systems to the selection of connection hardware for FS2. They also exposed practical issues in assembly and detailing that were difficult to anticipate from the digital model and drawing details alone.
For the structural system, the building was also tested as an 8-metre-tall physical model on a shaking table as part of the structural verification process.
Source: China Academy of Building Research (CABR)
From model to building
The value of the computational model was ultimately tested outside the computer. Construction introduced material behaviour, tolerances, sequencing, labour, cost and regulatory requirements that could not be reduced to geometry alone.
The project taught me that maintaining a relationship between digital modelling, physical verification and construction can be more consequential than increasing the sophistication of any individual model.
The building in 2022
The following footage was recorded in 2022, when I was able to revisit the building after its completion in late 2019.
Construction timelapse, 2017–2019
I set up a camera overlooking the construction site from an apartment in the adjacent building and recorded the construction process over 2 years. The footage was subsequently assembled into a timelapse.
