Rawlam from The North
The second phase of the RawLam 3 demonstrator assembled and exhibited at the Aarhus Architecture School.
CITA and the Eco-Metabolistic Model
Looking to apply a suite of digital technologies, Copenhagen architecture school’s digital architecture lab, CITA oversaw Rawlam, an ambitious research programme through 2021 and 2022 to radically re-think and optimise the supply chain process from forest through to finished glu-laminated construction material. The results, as CITA researcher Tom Svilanswrites, was as ground-breaking as it was provocative.
CITA – the Centre for IT and Architecture at the Royal Danish Academy in Copenhagen, Denmark – explores the intersection of digital culture and architecture. By deploying digital tools for modelling, simulation, and fabrication, it investigates how new design practices and techniques can emerge from a digitally-augmented understanding of materials and their processing. A key stream of research has been timber construction and the “extended timber value chain” – from how timber products are sourced to how they are processed and assembled into complex structures. As part of a new ERC-funded research project, CITA investigates the notion of an “eco-metabolistic model” for building with bio-materials. This project sets out to discover the necessary conceptual and technical frameworks required for engaging with the design and production of living and bio-based materials – the grown, the living, and the harvested. This requires a rethinking of the scope of design – towards a broader understanding and engagement with material value chains and lifecycles – as well as production – developing frameworks for adaptive fabrication that is able to respond to unpredictable material behaviour and heterogeneity.
The RawLam 3 demonstrator at the Architectures of Transition exhibition in Umeå, Sweden.
Details from the RawLam 1 prototype, showing the strategic allocation of wood qualities to different areas.
Current research into the timber value chain focuses on the ongoing digitization of the forestry and sawmill industries. Developments in material tracking and computed tomography (CT) scanning of the forest resource create a large quantity of material data that is seldom carried over to further processing, design, and manufacturing stages in a building’s production chain. While this has been geared primarily towards increasing the efficiency of value extraction from the forest and lowering waste from sawmilling, the wealth of deep material information gained in these processes is reduced to overall per-element grading of individual boards that proceed to the market.
This is where we see tantalizing opportunities: how can we leverage these under-utilized high-resolution datasets within processes of digitally-augmented architectural design? The ability to map internal material structures at sub-millimetre resolutions suggests new ways for steering the processing and usage of heterogeneous biomaterials and tailoring their inherent individuality towards specific functional goals – or vice versa. In this way, can we exploit our harvested resources more efficiently by finding appropriate places in our structures for a wider range of material qualities? Can we become more precise in our material specification by knowing more precisely what we are allocating and where it might be best suited? Timber is notorious for having a wide scatter of results across samples in mechanical strength testing, leading to the typical oversizing of cross-section dimensions and increasing of safety factors. If we are able to control the processing of the log such that we can precisely organize the material distribution in our designed architectural elements – by allocating knots to areas with low mechanical demand, for example – could we lower the variance and risk associated with specifying timber beams and achieve a leaner, more materially-economic approach to timber construction?
The Rawlam demonstrator at the Architectures of Transition exhibition in Umeå, Sweden.
Technologies such as LiDAR enable a detailed analysis of the forest, allowing more informed and nuanced decisions to be made about its use and stewardship.
These questions initiated a series of research probes and prototypes – dubbed RawLam – which sought to develop a clearer picture of what this new approach might look like, what constraints and issues it might raise, and what other questions it may lead to. At CITA, we have developed a “probe-prototype-demonstrator” research methodology, where probes are initial open-ended forays into a question or topic, prototypes are consolidations, clarifications, and proofs-of-concepts – a working through of the processes with a clearer picture in mind – and demonstrators are typically 1:1 built mock-ups that prove the feasibility and merit of the inquiry – objects that demonstrate the research findings.
In the RawLam series of work, RawLam 1 constitutes an initial probe of the previous questions. Its role was to identify the key steps in the log-to-product workflow, to determine what computational infrastructure and tooling is necessary, and to attempt a coarse but complete “walk-through” of the process. Glue-laminated (glulam) beams were chosen as a particular focus both because of prior research into the the design and specification of glulam structures and because they are an aggregated, designed material, built up of multiple individual pieces of timber. This means that that the material composition of such glulam beams could be controlled to a greater degree, at a higher resolution.
X-ray computed tomography (CT) scanning reveals the interior heterogeneity and features such as knots in harvested logs.
RawLam 1 therefore sketched out the workflow using approximate and stand-in datasets: a log was sourced close to Copenhagen, sawn into boards, and these boards were individually photographed to acquire a mapping of knots and fibre directions. The knots and wane of the boards were associated with a lower quality of material, and clear wood was associated with a higher quality. We designed a simple glulam assembly to have a variety of performance demands: areas of higher curvature that would require knot-free wood, areas around joints where knots and wane should be avoided, and a simulated stress distribution based on virtual loads placed on the design model – generated by a finite-element analysis (FEA). A heuristic algorithm was then developed which used this “material quality map” to position the lamellas of a glulam assembly onto the boards so that the performance demands matched the underlying material “quality” as best as possible.
Transferring the location and orientation of the lamellas onto the physical boards was then done by using measurements from reference edges and knots locations. After cutting out the lamellas from the boards, the glulam blanks were glued together and robotically machined.
While a rough approximation of the process, this initial probe showed us a promising and feasible path forwards. Along the way, while working with the cutting, lamination, and machining of the glulam beams, another aspect quickly became apparent: exposing the wane and “rawness” of the original log in the finished elements revealed an interesting aesthetic and textural dimension. Whereas there can be a strong propensity in engineered timber architecture for smooth, knot-free surfaces and straight grain, we found the presence of bark and wane an alluring potential for a controlled textural and aesthetic expression. As a direct result of allocating as much of the log as we could into the finished elements, it also signified for us another symbolic link back to the forest and the biological rawness of the tree – allowing it to find a voice in the final architectural expression, even after a whole chain of industrial processes and transformations along the way.
The distribution of lamella on the sawn boards is informed by the varying density and features captured in CT scans of logs.
RawLam 2 was largely a shoring up and tightening up of the process, with some key improvements and developments – a prototype. While in RawLam 1 the material map was approximated by photography, our new collaboration with Microtec – the manufacturer of CT-scanners for the timber industry – allowed us to source scanned logs along with their CT-scan datasets from their network of partners. The logs arrived precut into boards – having been used in another CT-scanning experiment at another institution. The material map was therefore replaced with the CT-scan density mapping, and augmented by Microtec’s own dataset of digitally-extracted features such as knot vectors, the pith line, and sapwood / heartwood delineations. The heuristic algorithm was therefore also improved to respond to this difference in source data. The same glulam assembly design was used to test the improved workflow, in order to keep changes to the process manageable and controlled. The notable change during production was the usage of a laser projector to trace the lamella outlines onto the physical boards – greatly increasing the accuracy and expediency of transferring the digital cutting patterns to the physical material.
RawLam 3 is a three-phase installation commissioned by the Bildmuseet in Umeå, Sweden for the Architectures of Transition exhibition. It demonstrates the implementation of this integrative method and refines the necessary material logistics, modelling and simulation frameworks, and data management workflows for actuating the link between forest and architectural element. As the escalation of the RawLam studies, it demonstrates the implementation of this integrative method at a larger scale and refines the necessary material logistics, modelling and simulation frameworks, and data management workflows for actuating the link between forest and architectural element. The act of scaling up from a small material prototype to a larger structure consisting of multiple interconnected glulam beams raised a number of challenges: handling more logs posed problems both for the physical logistics of moving, scanning, and cutting them; as well as stressing the digital process of how this larger dataset and interrelations between models and scan datasets can be maintained intact and operable.
What the demonstrator also allowed us to do – importantly – was to develop new relationships with actors across the timber value chain. Logs were sourced from Norra Timber’s Kåge sawmill -– a short distance outside of Skellefteå in northern Sweden – and were CT-scanned at the Wood Science and Engineering division at Luleå University of Technology. Visiting both the sawmill and research centre afforded us further insights into the development and digitization of the timber value chain. Exhibiting at the Bildmuseet gave us the opportunity to present the work and thinking of CITA to a broader audience in design and forestry, initiating interesting dialogues with the regional forest research cluster and producers of timber products and forestry equipment. Learning from the diverse voices surrounding the forests of northern Sweden has opened up new questions for us: how does the regionality and specificity of the forest resource factor into its use as a commodity and product? How can we become more mindful of the impacts – ecologic, economic, social – of the various types of forest usage and the tensions that inevitably arise between them? While we have followed a very narrow and technical thread in the story of digitization in the forest and sawmill, there is a much broader discussion to be had there as well: about what digitization in the forest means for land owners, harvesters, sawmillers, inhabitants, municipalities, and so on and so forth.
Futures and Outlook
Our immediate take-away from this foray into the digital forest is that, through these tools and incredibly rich datasets, we can return – at a somewhat different scale of industrial operation – to a semblance of the way in which the forest resource was – and still is, in many cases – chosen and applied because of its specific forms and characteristics. Crooked trees, crotches, and forks have found specific places in the construction of boats and houses. Perhaps this integration of the digitally-acquired internal landscape of trees can help us adapt and steer their exploitation towards more ecologically sound and sustainable paths.
We further realize that it is impossible to disconnect the product from the place. Tracing the relations and lineages between the forest and the final built element only shows how interlinked and intertwined they really are. Our on-going efforts are to make these relations and disparate data points somehow visible and accessible to a wider audience, seeking ways of integrating the diversity and wealth of knowledge across the timber value chain into our design practice. At this point, initial steps are being taken to move towards some type of more formalized trialling. The immediate task here is to bring our glue-lamination process up to established standards to make comparisons more legitimate and meaningful with more standard products and processes. New partnerships are looking to leverage new research in computational mechanics of wood and how this could be linked to the wealth of acquired data in the CT-scanned logs, as well as to participate in the discussions around the semantic web and the automation of design and construction information. Further, from a design perspective, we believe that many exciting opportunities in this work remain to be captured. As a computational design system, the work by itself does not suggest radically new forms or building systems. However, as a way to tune, tailor, and expose the textural dimension of wood, it has the potential to introduce new spatial expressions and challenge existing design frameworks for building with timber.
Tom Svilans is an Assistant Professor at CITA, Royal Danish Academy, and was the lead researcher on the RawLam project.