Advanced Building Simulation

162 Mahdavi
7.3 Self-organizing models
7.3.1 Requirements
To serve effectively as the representational core of a sentient building, a self-organizing
model must fulfill at least two requirements. First, such a model must incorporate and
integrate both a rather static building product view and a rather dynamic behavioral
view of the building and its environmental systems. Second, to provide real-time
building operation support, the model must be easily adaptable, that is, it must
respond to changes in occupancy, systems, and context of the building. Ideally, the
model should detect and reflect such changes automatically, that is, it must update
(organize) itself autonomously (without intervention by human agents).
7.3.2 Building as product
Numerous representational schemes (product models) have been proposed to
describe building elements, components, systems, and structures in a general and
standardized manner. Thereby, one of the main motivations has been to facilitate
hi-fidelity information exchange between agents involved in the building delivery
process (architects, engineers, construction people, manufacturers, facility managers,
users). A universal all-purpose product model for buildings has not emerged
and issues such as model integration across multiple disciplines and multiple levels
of informational resolution remain unresolved (Mahdavi 2003). Nonetheless,
past research has demonstrated that integrated building representations may be
developed, which could support preliminary simulation-based building performance
evaluation. An instance of such a representation or a shared building model
(see Figure 7.2) was developed in the course of the SEMPER project, a research effort
toward the development of an integrated building performance simulation environment
(Mahdavi 1999; Mahdavi et al. 1999b, 2002). We submit here, without proof,
that such a shared building model can be adapted as part of a self-organizing building
model and provide, thus, a sentient building with the requisite descriptions of
building elements, components, and systems.
7.3.3 Performance as behavior
Building product models typically approach the building from a “timeless” point of
view. Their representational stance may be said to be decompositional and static. In
contrast, simulation allows for the prediction of buildings’ behavior over time and
may be thus said to provide a kind of dynamic representation. A comprehensive
building product model can provide simulation applications with necessary input
data concerning the building’s geometry, configuration, and materials. This information,
together with assumptions pertaining to the context (e.g. weather conditions,
available solar radiation) and basic processes (e.g. occupancy schedules, lighting, and
ventilation regimes) is generally sufficient to conduct preliminary simulation studies
resulting in predictions for the values of various performance indicators such as
energy use and indoor temperatures. However, more sophisticated simulations
involving detailed behavior of a building’s environmental systems and their control