Concept Learning in AutomationML with Formal Semantics and Inductive Logic Programming

Technologies for Industry-4.0 are evolving rapidly, and the term semantics is widely used. Various standardization groups claim that they provide mechanisms to express semantics or allow the integration of semantic information. The emerging data format AutomationML (as IEC 62714) proposes a role-based approach to encode semantics in engineering models and has already standardized fundamental engineering concepts as role classes. However, for concrete data processing tasks the currently standardized role classes are not sufficient to unambiguously express the meaning of various vendor or application specific concepts, while user-defined role classes rely on hand wired semantics encoded in dedicated software, e.g. importer/exporters. Yet AutomationML system unit classes represent reusable engineering objects as relational models of AutomationML roles, attributes, interfaces, internal elements and links, which can be used to describe complex user-specific concepts. To enable an automatic machine interpretation of these unstandardized relational models, we transform AutomationML data to a formal and declarative semantic representation using the Web Ontology Language (OWL), and propose a rule mining approach to learn the intended meaning of user selected system unit classes, i.e. to identify the common relational structure shared by the selected engineering objects.