The effectiveness of cognitive manufacturing systems in agile production environments heavily depends on the automatic assessment of various levels of interoperability between manufacturing resources. For taking informed decisions, a semantically rich representation of all resources in a workcell or production line is required. OPC UA provides means for communication and information exchange in such distributed settings. This paper proposes a semantic representation of a resource's properties, in which we use OWL ontologies to encode the information models that can be found in OPC UA NodeSet specifications. We further combine these models with an OWLbased description of the resource's geometry and-if applicable-its kinematic model. This leads to a comprehensive semantic representation of hardware and software features of a manufacturing resource, which we call semantic digital twin. Among other things, it reduces costs through virtual prototyping and enables the automatic deployment of manufacturing tasks in production lines. As a result, small-batch assemblies become financially viable. In order to minimize the effort of creating OWL-based UA NodeSet descriptions, we provide a software tool for the automatic transformation of XML-based NodeSet specifications that adhere to the OPC Foundation's NodeSet2 XML schema.