1. From Building Information Modelling to Digital Twins: Digital Representation for a Circular Economy

BIM is a frequently misrepresented and therefore misunderstood technology. Many poor definitions describe not the phenomenon itself but its applications and effects (Sacks et al. 2018), often from the perspective of existing analogue practices. The production of drawings and other conventional documents to incrementally improve efficiency or reduce errors takes up a disproportionate amount of the BIM literature but does not explain how BIM is structured and how its structure helps to achieve certain objectives. Instead, it makes BIM appear as a mere step in AECO computerisation. The truth is more revolutionary: BIM marks the transition to symbolic representation (Koutamanis 2022). While earlier technologies like computer-aided design (CAD) focused on the graphic implementation mechanisms of building representations, BIM makes explicit the symbols described by these mechanisms.

Symbolic representation is already the norm in many computer applications. In a digital text, the capital ‘A’ is not a group of three strokes, as in handwriting, but the Unicode symbol U+0041, explicitly entered through a keyboard and stored as such, regardless of how it appears on the screen. Any change to the symbol does not come from changing the three strokes but from changing the properties of the symbol (e.g. a different font or size) or switching to a different symbol (e.g. U+1D434 for the mathematical capital ‘A’). Symbolic representation underlies a lot of machine intelligence. In digital texts, knowing each letter allows computers to recognise words and sentences and subsequently understand grammar and syntax.

Similarly, in BIM, a window is not the group of line segments one sees in a graphic view like a floor plan but a symbol explicitly entered in a specific location of a wall. One can reposition the window in the wall, but changing its type or even its size may require switching to a different symbol. The interfaces of BIM software tend to depart from facsimiles of analogue drawing, which confuse users into thinking that they are drawing and obscure the symbolic structure of the model. We should think of BIM models not as 2D or 3D drawings with additional data but as graphs of interconnected symbols. In fact, connections are between specific symbol properties (Fig. 1.1): the co-termination of two walls links the endpoints of their axes, while the orientation of a wall is inherited by the windows it hosts.

The structure of the model displays symbol at the two ends. The symbols are connected with various properties. The properties of the two symbols are linked through a constraint.

BIM is often called ‘object-oriented’. This is misleading because the term has a different meaning in computer science but also because we should not equate symbols with real things. In English, the letter ‘a’ corresponds to five different sounds (phonemes). Knowing how to pronounce the letter depends on the context (the word). When considering representations in building, the correspondence between symbols and things can be even fuzzier. A window may be considered a discrete component, but a wall is an assemblage with variable composition and indeterminate form. Its material layers often continue into other walls, forming construction networks that are not captured by wall symbols in BIM. A main reason for this is geometric bias: continuous walls are segmented into separate symbols by the geometry of their axes.

Despite such fuzziness and resulting ambiguities, the symbolic representation underlying BIM remains the obvious choice for AECO computerisation, with a potential similar to that of the Latin alphabet or the Hindu-Arabic numerals. The graph of symbols and their relations is a transparent, consistent and efficient foundation for any application. The capacity for integration and continuity means that information efforts can be consolidated into a single representation that caters for all aspects, goals and disciplines.

While the use of BIM has yet to reach a satisfactory level or achieve significant efficiencies, AECO has already adopted a new buzzword: digital twinning. In contrast to BIM, digital twinning has yet to consolidate into a recognisable technology. Quite frequently, any virtual model seems to qualify as a digital twin, purely on the basis of intent. However, a digital twin is more than a model: it is a digital replica of something physical. It describes the form, behaviour and performance of the thing, including uses, users and direct context – all that is required for precise and accurate analyses and forecasts of future states of the physical twin.

Information in a digital twin is dynamic and reciprocal: sensors in the physical twin that monitor temperature, light, sound, occupancy, vibration, etc., send their data to the digital twin, where they become attached to relevant properties of the appropriate symbols. The products of the digital twin travel in the reverse direction, guiding actuators in operational adaptations, e.g. the functioning of heating systems, and informing users through displays (Fig. 1.2). In other words, the twins are connected in both directions in near real time and are capable of communication and synchronisation (Chen 2017; Liu et al. 2018). Consequently, we can distinguish between representations (static models, as in BIM), shadows (representations which are updated by data from the physical things) and twins (full two-way synchronisation) (Fuller et al. 2020; Sepasgozar 2021).

The structure of the model displays connections between symbols in a digital twin and things in a physical twin.

Digital twins of buildings are invariably based on BIM (Boje et al. 2020; Sacks et al. 2020; Begić and Galić 2021; Mêda et al. 2021; Shahat et al. 2021; Tagliabue et al. 2021; Alibrandi 2022; Shaharuddin et al. 2022). At the same time, it is stressed that digital twinning is more than BIM, as it includes sociotechnical and process aspects, especially in use (Boje et al. 2020; Davila Delgado and Oyedele 2021; Sepasgozar 2021). This makes it significantly more demanding than as-built BIM in terms of reliability, precision and completeness. Furthermore, it is questionable whether BIM can accommodate and process the big data produced by sensors in the built environment. Rather than a foundation, BIM is a predecessor to digital twinning, based on the same symbolic approach to representation (Boje et al. 2020; Koutamanis et al. 2021).

More than on BIM, digital twinning relies on the Internet of Things (IoT): the networks that connect sensors, actuators and displays in a building, making it ‘smart’, i.e. automating certain operations, such as opening doors and regulating ventilation systems. In addition to such local automation, the IoT also collects data from all sources to capture the history and the overall conditions in a building. This improves local operations by connecting them to global goals and constraints. The IoT is not just an enabler but a necessity because digital twinning presupposes a building heavily populated by IoT for bidirectional communication and synchronisation, including feedback to users and operators (Farsi et al. 2020; Fuller et al. 2020; Lu et al. 2020; Sepasgozar 2021). The collection of data for digital twinning could be much more extensive than in most smart buildings, resulting in a lack of suitable physical twins and possibly rendering digital twinning a pipe dream. Alternatively, one could tolerate low-fidelity solutions as early deployment stages and encourage incremental development (Mêda et al. 2021). However, experience with BIM maturity levels suggests that such tolerance is self-defeating because it provides alibis for not taking the trouble to use the technology properly while continuing processes that actually undermine it. The degree of validation and verification required in digital twinning makes any attempt to pass off static models as twins as misguided as calling 2D drawings BIM.

BIM and digital twinning are general-purpose technologies. There are also stand-alone information technologies specifically developed for circularity in AECO. These are referred to by terms such as building or material passport or logbook. Chapter 5 by Honic et al. in this book describes the potential of such technologies and relevant life cycle and standardisation challenges in detail. Therefore, from the perspective of this chapter, it suffices to emphasise that BIM, as an integrated information environment, is more than a useful source of data (Durmišević 2018; Bertin et al. 2020). There is a significant overlap between BIM and material or building passports (Charef and Emmitt 2021), even when the latter are based on other sources for product composition breakdown.

The advantage of BIM is that it makes materials situated and connected to life cycle processes (Honic et al. 2019). This supports design for deconstruction and disassembly (Minunno et al. 2018; Xing et al. 2020; Marzouk and Elmaraghy 2021; O’Grady et al. 2021) and other circularity goals. Translating manufacturers’ disassembly instructions into simulations in BIM improves legibility and completeness, especially concerning resources that may be available or required. It also verifies the disassembly procedures and validates designs with respect to them. Including the location of a component among its metaproperties in a passport does not offer the same advantages.

In conclusion, passports and logbooks are amenable to the integrating power of BIM and digital twins, which can accommodate product information (Kebede et al. 2022), life cycle energy data (Shah et al. 2023) and other key information in their properties and relations. In BIM, information collections such as material passports can become views of the model, similarly to bills of quantities. Linking their goals and constraints to all activities in design, construction and operation through BIM returns connections to information sources that help make material flow registration and analysis realistic and reliable (Miatto et al. 2022).

The first step in a coherent approach to circularity with BIM or digital twins is to learn to rely on symbolic representation. Any full model or twin can easily cover circularity information needs without additional investment, but in practice representations can be selective or opportunistic and hence incomplete or inconsistent. Deferring the information burden to any particular goal and its stakeholders (as with passports) is not a viable option. Instead, all AECO stakeholders should insist on joint, permanent working environments, not disconnected repositories or documentation for different phases. There can be no half-hearted BIM or digital twin deployment: economising on investment means severely limited potential and low returns.

The perennial question in AECO is not so much who makes a BIM model but who maintains it, especially in the life cycle of a building. If this does not happen collaboratively by conjoining the core processes of all actors, and preferably automatically, there is little hope for success. Collaborative solutions also lower the participation threshold for smaller enterprises and offer enticing benefits in terms of digital support and room for fruitful specialisation. In return, the enterprises contribute to the completeness and up-to-dateness of information simply by using it.

The second reason for a lack of digital solutions for circularity is selectivity: any information solution motivated primarily or exclusively by circularity inevitably remains restricted to circularity factors and aspects. It may even suffer from inattentional blindness, which causes omissions of important data simply because we concentrate on other matters (Chabris and Simons 2010). One can naturally work with conscious concentration towards a full, inclusive solution, but then the results would amount to something akin to BIM or digital twins, i.e. a comprehensive solution that could only justify costs and improve returns by being open to other goals and priorities, too.

The second step towards circularity with BIM or digital twins is to utilise their capacities for exploring deconstruction and disassembly (Akanbi et al. 2019; van den Berg et al. 2021). In the same way that we simulate construction processes, we can also simulate the expected maintenance, refurbishment, renovation and deconstruction processes with the accuracy and precision required for feasibility, effectiveness and efficiency. This provides direct support for construction-related circularity goals (narrow and regenerate through efficiency improvement) and a useful background for others (slow and close through reliable life cycle projections). It also stresses the necessity of detail and realism. For deconstruction in particular, we should acknowledge that it is not a mere reversal of construction. As Van den Berg explains in Chap. 11 in the relevant chapter in this volume, information is a key issue in organising reverse logistics. As-is representations are essential for the identification and harvesting of reusable resources from existing buildings because as-built models (i.e. construction documentation) are neither sufficient nor reliable enough. Closing loops requires certainty about the state of components and materials, as well as about their physical context, which has changed from an accommodating construction site to a finished, functioning building. This calls for solutions that are full and realistic, including all details of deconstruction in space and time, e.g. how cranes and scaffolds would function in the existing building. Van den Berg (Chap. 11) describes a number of focused explorations and demonstrations that must graduate from opportunistic demonstrations of potential to standard facilities in BIM and digital twinning.

Based on the second step, we should explicitly describe circularity dependencies and constraints in properties and relations of symbols (e.g. constraints on interfacing between components for effective deconstruction). Relations are of particular importance in this respect because they link interfacing between components to symbol behaviours. They can ensure that the building design and construction allow for deconstruction (Sanchez et al. 2021), e.g. avoid additions that spoil interfaces designed for disassembly, such as equalising layers of in situ concrete over demountable floor slabs. If symbols refuse to accept such additions to their properties or relations, similarly to a door not accepting positioning outside a wall, the scope for human error becomes much smaller. This is particularly important in the use phase, where changes are only too frequently improvised, in both refurbishment and maintenance. The representation can also anticipate circularity operations, such as the replacement of some components when they fall below a certain performance level, by including among the symbol triggers that adjust the timing of loops.

The final step is an extension of the previous three: use 4D symbolic representations to monitor the detailed history of a building, preferably in near-real time. As symbol properties and relations can register the activities and effects of maintenance, refurbishment, etc., material flows are measured and managed not by questionable proxies but with primary, precise and accurate data (Minunno et al. 2018; Chen and Huang 2020; Marzouk and Elmaraghy 2021). Up-to-date information is essential for the planning of circularity operations: narrowing, slowing, closing and regenerating can be based not just on initial assumptions and projections but on constantly refinable and dynamic decision frameworks that include permanent validation and verification facilities for making sense of the existing building conditions for deconstruction (Van den Berg, Chap. 11). The bidirectional relation between digital twins and buildings is clearly advantageous in this respect, as it covers not only monitoring but also adaptations in the behaviour of the physical twins, e.g. adjusting the heating and ventilation of a building in order to reduce the extent of material ageing in specific components.