Linking Behaviour Data to Knowledge: Contextualization and De‐Contextualization

Sensor informatics provides extensive data about system behaviour, which we would ideally like to link to system models and the underlying knowledge. The nature of systems knowledge formation is that we observe the behaviour of particular systems in their context, and incrementally de‐contextualize it to arrive at formal knowledge. Conversely, the nature of engineering synthesis is that we develop and reason about the behaviour of system configurations by identifying and combining all the knowledge applicable to that particular configuration and context i.e. by successive levels of contextualization. This systems science paper proposes a conceptual framework for linking knowledge, system models and behaviour data. It inquires into the levels of system description involved, and the relationships between the levels that drive contextualization and de‐contextualization. We propose organizing knowledge about individual entities and interactions into type knowledge frames , complemented by patterns and other knowledge related to configurations. We use assume‐guarantee concepts to formulate self‐contained type knowledge frames with internal consistency relationships between structures, interactions and behaviours. We suggest how observations data can be abstracted into observable behaviour models . Together, these insights point towards the possibility of tooling support for populating knowledge frames, and creating bi‐directional relationships between behaviour data, system models and domain knowledge, based on contextualization and de‐contextualization.