[Back]

S. Feldmann:
"Diagnosis and Handling of Inconsistencies in Heterogeneous Models of Automated Production Systems";
Supervisor, Reviewer: B. Vogel-Heuser, G. Kappel, C. Diedrich; Institute of Information Systems Engineering, Business Informatics Group, 2019; oral examination: 2019-09-20.

Companies in the automated production systems domain are more and more forced to manufacture
individualized goods in order to remain globally competitive. Thereby, the rapidly changing product
and system requirements impose an ever-increasing complexity on the engineering of automated
production systems. One opportunity to reduce this complexity and to support engineers in developing
automated production systems is the use of models. However, due to the manifold disciplines
involved in the engineering process - mechanical, electrical/ electronic and software engineering to
name only a few - and the disparate stakeholders participating in this process, manifold heterogeneous
models are created during engineering. To make things worse, these models heavily differ -
for instance, in their degree of formality, with regard to their level of abstraction or regarding their
viewpoint on the system. In addition, the models overlap - they incorporate elements, which refer
to common aspects of the system under investigation. As a consequence, inconsistencies - namely
states of conflicts within the models - are likely to occur. Whereas in the current state of practice,
a number of commercially available software tools already provide the means for checking the compliance
to syntactical or well-formedness constraints, none of these tools support the overarching
management of inconsistencies within engineering models of automated production systems. To
address these issues, manifold approaches have been developed in the related research, which focus
on the diagnosis of inconsistencies but not on the handling of inconsistencies in engineering models
of automated production systems. It is therefore inevitable to provide an approach, which allows
to diagnose inconsistencies within the multitude of models created during engineering and, in case
an inconsistency is diagnosed, to support engineers in choosing appropriate handling actions.
In this dissertation, based on the different heterogeneous types of models and inconsistencies in
the automated production systems domain, requirements to be fulfilled by an inconsistency management
approach are derived. Furthermore, a knowledge-based approach to diagnose and handle
inconsistencies by means of Semantic Web technologies - in particular the Resource Description
Framework (RDF) and the SPARQL Protocol and RDF Query Language (SPARQL) - is presented.
Inconsistency diagnosis rules are therein described by means of graph patterns (SPARQL Query
Language), which are matched against a graph-based representational formalism (RDF triples) that
contains the involved models. Analogously, inconsistency handling rules are described by means of
respective graph manipulation rules (SPARQL Update Language) and executed to handle diagnosed
inconsistencies by either ignoring, tolerating or resolving them. To keep the inconsistency
management approach as flexible as possible, semantic abstraction mechanisms are provided by
means of mediations to vocabularies, which allow for incorporating not only the different disciplines
involved during engineering, but also discipline-spanning concepts and background knowledge that
are common to multiple disciplines. Hence, instead of creating a "world model" of automated production
systems, models are mediated to a common model that represents only the elements that
are required for inconsistency management purposes.
For the purpose of validating the approach, a prototypical implementation is realized by means of
standard implementations within the fields of Model-based Engineering (Eclipse Modeling Framework
(EMF)) and SemanticWeb Technologies (Apache Jena). The overall feasibility of the approach
is evaluated by means of a laboratory automated production system. In addition, the approach is
compared to another inconsistency management approach, which relies on Model-based Engineering
technologies (EMF). Finally, in order to assess benefits and limitations of the approach, an evaluation
at the hand of application examples that are inspired from industrial application is provided.

Resource Description Framework, SPARQL Protocol