by David Andreen, 8 March 2012
Design is the pursuit of performance – the performance of objects, structures and processes. Whether that performance is structural, economic, aesthetic, artistic or social matters little; the job of the designer is to balance the outcome for the task.
In the realms where architects and computational designers mostly operate, performance is derived from the organisation of material. Over the centuries, we have learned what we can expect from a multitude of materials and how to manipulate these according to our needs. When one material fails to deliver all the desired properties, we create assemblies composed of multiple materials in predictable, homogeneous and discrete arrangements. Better performance has come to mean more complex assemblies and more exotic materials.
This method has proven extremely successful as it is well adapted to standardisation – on the level of production as well as in situ over product life span. In order to continually improve the performance, increasingly complex factories push miniaturisation and material science. For architects this is a problem. Primarily because building economics quickly place a ceiling on how advanced these assemblies can be, and the inherent need for customisation is at odds with the required standardisation. But also because we deal with flows of matter, e.g. air, water, people, traffic, and these cannot be miniaturised.
In the context of vastly increased demands on architectural performance our profession is in dire need of a new approach. Can we accomplish architecture which allows for more innovation (by orders of magnitude!) and is affordable given constrained resources? Considering the limitations that come with discrete assemblies, I would be suggest that an alternative lies in continuous material differentiation which discards the component method in favour of a deeply integrated approach.
Continuous material differentiation implies a continuous treatment of materials which are functionally graded without any abrupt discontinuities in shape or material. As in biological organisms, such as a human body, it allows one structure to serve multiple functions, often at several scales simultaneously. Such structures can be self-similar, or fractal, and tend to behave in very favourable ways.
Unikabeton by Per Dombernowsky and Asbjorn Sondergaard using topology optimisation to shape a continuously differentiated pavilion.
The advantages of such a process are numerous. A graded structure can be locally optimised to focus the resources where they are needed as opposed to component based assemblies which are necessarily compromises due to their discrete nature. This can be illustrated most easily by techniques such as topology optimisation, as in the case of the Unikabeton project shown above. The shape of the pavilion is determined by an evolutionary algorithm that optimises the geometry to have the greatest structural integrity with a minimal amount of material. This leads to uniform stress profile without local points of weakness (as found in standardised components) at the cost of a more complex manufacturing process. This cost is however decreased, and potentially completely eliminated, with some digital fabrication techniques. The same principle applies equally to other properties such as acoustics, heat transfer, etc.
Beyond basic optimisation, a continuous structure has the potential to facilitate one of nature's greatest advantages over engineered objects: the combination of multiple functionalities in the same physical space or structure. Traditionally architecture uses separate component placed adjacent to one another in order to deliver multiple functions, e.g. steel or timber studs for structure, mineral wool for insulation, and plastic pipes for wiring. But in using a continuously differentiated and largely uniform material, many of these functions can be achieved by manipulating the geometry itself. A void inside a wall may be there to optimise the weight to strength ratio, but by limiting its size to prevent convective flow it can reduce heat transfer though the wall.
Equally the same void could be used to channel a flow of air or house electric wiring. Changes intended for one purpose can have implications elsewhere. In controlling a void's shape the designer may affect the acoustic properties of the adjacent spaces. But this is not always the case and at what scale the designer operates is key: Manipulating the texture of a surface may alter its properties with regards to moisture absorption or appearance, but leave structural capacity unaffected. Modulating the overall height of a surface may affect it acoustic properties and the programming of the space without losing its micro-properties.
The benefit of component assemblies for a building is that separated teams of experts can design different systems such as cladding, structure or Mechanical and Electrical in parallel. Apart from agreeing key interfaces they can act largely independently. Similarly, as each component performs independently from other components, and to a great extent from its context, this performance can me measured and guaranteed in advance. However, with the increasing complexity of contemporary buildings this linear approach is becoming a bottleneck for efficiency in both design and performance; take for example the issues of clash detection between different systems for a large, freeform building or complex.
A continuous structure can be designed through a negotiation process, which is free from the constraints and compromises of pre-fabricated component assemblies. This process is algorithmic in nature, and must incorporate the ability to manage the emergent complexity and interdependence of such a system. Each part of a building responds to the unique combination of local requirements across multiple scales, and depending on the function which is addressed the requirements for resolution and accuracy varies greatly. The question is how could this level of negotiation ever be achieved? Surely this interdependence and packing of multiple functions into the same material space is too complex to manage let alone compute?
Cast of internal tunnel network of Macrotermes Michaelseni. Image credit Rupert Soar.
The answer may lie in nature.
A striking example of continuously differentiated structures is the mounds of macro termites in sub-Saharan Africa. These structures are not the actual nest of the termites, as this is situated in the ground below it. Rather, they perform several critical functions in what can be considered the body of the termite 'super-organism', consisting of termites, the mound, and a species of symbiotic fungus. The mounds provide a defensive mechanism protecting the colony, they regulate the humidity and perhaps most importantly they act as a 'lung', enabling the exchange of respiratory gasses between the nest and the outside air while maintaining the controlled internal climate of the nest required by the termites. All of this is accomplished through a structure which is composed of a single, simple material most easily described as mud!
This functionality is enabled by the continual rearrangement by the worker termites of the mound geometry, originating initially from the pheromone field emitted by the termite queen. The thousands of tunnels in the mound are tuned to harness the energy available in turbulent and unpredictable winds found at ground level. This is not a static system, but changes in scale and topology both over time and space. Near the surface narrow and branching egress tunnels are found. Moving closer to the nest and further away from the surface, these tunnels change in a continuous manner to accommodate different processes, growing larger and more reticulated.
Slice and scan of a termite mound in Namibia for later reconstruction into 3D model. Image credit Rupert Soar.
So how do the termites do it? They have evolved a series of behavioural and physiological rules which guide this process. But these rules are not dictated in a top down fashion, instead they are simple actions in response to local environmental stimuli. Communication and coordination happens primarily indirectly: as the termites manipulate their environment they also effect the stimuli perceived by other termites. This is what is referred to as stigmergy. The result is process of construction and adaption which is emergent. While we may not yet (or potentially ever) be able to generate emergent building processes through the same manner of evolution as found in nature, much of what is observed in the termites could, I would suggest, be applied in architecture and construction. But, I believe, a considerable leap would need to be made in how we model and design before we could contemplate designing comparable continuous material structures.
Instead of focusing on the design and modelling of geometric components and shapes, I believe that the emphasis should be placed on the relations between performance and shape and how this can be defined algorithmically. This is a radical departure from the way design is currently conceived and practised. It implies a focus on the forces, flows, gradients and energies and how material modulates these flows; process driven design where form and performance are just expressions of the processes which underlie them. This might hybridise the roles of designer and engineer and bring design closer to both building science and simulation. The hope is to facilitate an architecture which is truly responsive to a complex context without diminishing the role of the designer.
Giving an example of recent research experiments, the image above shows a network which is designed to organise within an encountered design space. The network formed by the particle springs is modelled on the connectivity properties found in the mounds of Macrotermes michaelseni. It has been defined algorithmically to self-organise around an existing structure, finding pathways around obstacles in the predefined design space which may consist of structurally critical components, utility channels, windows etc.
When 'materialised' in the form of a voxel model, the output can be printed using additive manufacturing techniques and may ultimately exhibit properties which allows it to 'breathe' in a manner similar to the termite mounds. The question is to what extent a static structure such as we can fabricate using today's technology is able to functionally perform in manners similar to the dynamic structure found in the mounds. Will a continually differentiated building ultimately incorporate embedded sensors and dynamic feedback which allows and demands for its continual material rearrangement?
To conclude, I would suggest that continuous material differentiation may come about as a way of dealing with the design challenges that have followed the invention and adoption of additive fabrication, but the potential of the system extends far beyond that and could have an enormous impact on the way we design, fabricate and interact with the built environment. While considerable challenges are to be overcome, the solutions hold the potential for innovation, collaboration and sustainability at new levels. Aspects of buildings which have previously been separated by process, scale and profession can in such a model be interrelated in ways which allow us to meaningfully visualise and describe the multiplicity of dynamic feedback loops that we all understand to exist in a building or built environment.