The variety of components and the complexity of technical solutions in factory automation push information management based on relational databases to it's limits in terms of maintenance complexity and usage flexibility. Semantic Technologies account for maintainable, comprehensible and rich schema descriptions as well as state-of-the-art reasoning and SPARQL engines claim to deliver compelling performance. In the following we briefly report on applying ontologies and reasoning for managing complex product data in the automation domain.
Motivation
The ontology schema also contains properties to repre- portal drive train sent part-of relationships (! in Fig. 1) as well as aggregations ( I) to describe drive trains used gear in portals. Furthermore a 2D or 3D portal is concatenating sev- axis mounting motor controller eral drive trains ( BB). SWRL is used to describe constraints imposed by product management flange to restrict the technically possible to the economically reason- Fig. 1. Excerpt of the electric drives ontology able portal systems. The SWRL rules were iteratively worked out by ontology experts from informal statements verbalized on PowerPoint and Excel sheets. In addition, the ontology contains axioms to express implicit dependencies between components, e.g. to propagate information from drive elements to drive trains or portals. The schema part of the ontology was built using the Protégé ontology editor, contains about 500 classes and 60 object properties and is of horn-SRIF (D) expressivity fully complying with OWL 2 RL plus SWRL. 3
Fast and Flexible Knowledge Access with Reasoning and SPARQL The ontology is populated via an import pipeline with instance data from di erent sources including master data from SAP and product development data from an Oracle DB. The very first benefit of the semantic approach was an increase in data quality. Many data flaws and inconsistencies were regularly reported with the help of automated tests that incorporate integrity checks based on SPARQL queries and entailment tests.
All inferences are materialized with the help of the RDFox rule engine prior to querying. After materialization querying for all valid drive trains or for 3D portals comprising axes of a particular family or transmission type is a matter of a few seconds. Furthermore, since we have now a declarative, comprehensible product model SPARQL queries are much more easier to write than the SQL queries before. In addition, extensions or changes in the information model due to technical or business demands are much faster accomplished and validated. That is a significant increase in flexibility as well as saves time and computing resources. For daily business a set of Web services and a SPARQL endpoint allow to deliver on-demand “views” on the complete product line taking account of all technical and business related constraints in contrast to the previous solution that had to run overnight to materialize predefined export queries.
Since the whole architecture is based on standards such as OWL, SWRL, and SPARQL as well as REST, JSON, etc. on top of a reasoning aware SPARQL backend there are no dependencies on particular vendors or software products. Furthermore, due to the compactness of the schema and data part of the ontology as well as the availability of multi-platform, easy deployable OWL RL/SPARQL engines the semantic solution is also ready to serve any on- and o ine product configuration application.