Modeling Dynamic Context for Automation Systems Using Semantically Enriched Property Graphs

In light of the ongoing digitalization, automation systems are converging to cyber-physical systems to increase their intelligence and autonomy during runtime. In order to improve the operational efficiency and adapt to unforeseeable operating conditions, considering contextual relations between data as well as the physical operational context is pivotal. Addressing the need to elicit the operational context of a system and relate it to available operational data and system models to enhance its operation, this thesis proposes a process and method composed of five steps to identify, acquire, model, learn and apply context via a middleware-based architectural framework. To evaluate the proposed process and method for context modeling, three use cases are demonstrated, namely a medication assistance system, a wash-dryer as well as a flexible production system. Each automation system has a varying application scope and system modeling approach. Applying the proposed process and considering context shows an added value in contextualizing anomalies, generating user-centric recommendations, attributing failures in system diagnosis and enhancing production planning.