The building as a platform

Abstract

This chapter discusses the need and merit for a possible alternative view on digital twinning. First, the potential that the digital twin will not be IFC-based; in fact, the possibility that the digital twin will not be based on any common model of data. Second, and consequently, the fact that the process of digital twinning becomes more important than the digital twin itself. The arguments for these are based on the very role and definition of digital twins. In this chapter, they are not just a repository of data. In addition to data, a digital twin includes two major elements: representation of workflows, which are needed for supporting an automated/algorithmic operation; and simulation models of, for example, energy management and/or user comfort scenarios. Such data will span structured and unstructured data; building, operator, and user-generated data; historical, real-time, and simulated-futures data. Consequently, the hard-to-achieve interoperability within the traditional structured BIM data will be impossible. Of course, BIM will be a major component in any digital twin, but not the core. The role of a digital twin, in this context, is to discover knowledge: learning what the data is telling us; supporting predictive analysis. This stands in sharp contrast to IFC mentality: compliance to a pre-defined model of knowledge; and achieving interoperability. To this end, the process of composing, analysing and learning from digital twin data becomes the key contribution. We first present our proposition for predictive digital “twinning”, the business case for their existence and value. We situate that against recent advances in IFC-based twinning and, equally important, the criticism for IFC (model-driven) mentality. To illustrate the arguments made, we showcase an ongoing digital twinning project at the University of Toronto. It is built on top of a non-IFC legacy system for building automation. A no-model architecture is proposed to achieve the following: adding BIM to the digital twin as a component, not as the core element; using machine learning to discover patterns and support predictive analysis and business intelligence applications; and engaging all stakeholders in the learning process.