Review Comment:
The paper presents the AutomationML Ontology (AMLO) that covers the top level of the AutomationML standard (i.e. the Computer Aided Engineering eXchange format - CAEX) focusing on modeling topology information (i.e. object hierarchies, properties, and libraries) of Cyber-Physical Production Systems (CPPS).
The AutomationML markup language is an IEC standard that has been developed to describe complex and heterogeneous CPPS, and for addressing data exchange and integration issues between engineering tools during manufacturing systems design.
As stated by the authors, the main benefits of having an ontology based representation of the AutomationML standard are: the ability to 1. interlink and integrate heterogeneous data efficiently; 2. enhance the engineering process of CPPS system by leveraging semantic technology; 3. facilitate the connection to other industrial standards for which a semantic format already exists (e.g. the GoodRelations ontology or the eCl@ss catalog).
The paper describes the design process and motivates the main decisions taken during the AMLO development, provides an overview of the major implementation details, and discusses two two application exaples for the AMLO. The paper is well focused and reads well. Moreover, as stated in Section 6, many recent works paid attention on providing a semantic representation of the automation domain through ontologies (testifying the interest of researchers and the industry on the topic), but most of them are not mature, incomplete and not publicly available.
However, the paper has some issues that I’d like to be considered and resolved:
1) the authors claim that the AMLO "has been developed following best practices for ontology engineering", such as the application of Ontology Design Patterns (ODPs - Section 4.2) and the re-use of existing ontologies (Section 4.3). However only two basic patterns have been used (i.e. PartOf and Constituency), while other useful patterns have been ignored (e.g. for modeling connections and relations, to represent functionalities and behaviors, etc.). Moreover the authors should clarify how patterns have been applied (i.e. have they specialized/extended the pattern? have they directly/inderectly reused the pattern?). Also the reuse of existing ontologies is quite elementary (as far as I understand, a few SKOS and PROV-O properties have been used for basic metadata, such as definitions, labels, and to declare the tool that generated the AutomationML document, and the Ontology of Units of Measure for encoding unit of measurement).
2) the code excerpts are not clear and should be reviewed. For instance, Listing 2: what syntax has been used? If it is Turtle, I think that the code is not correct. Also Listing 4 should be reviewed: what does "RO_MechatronicAssembly" mean? Perhaps something is missing... There is also an issue in Listing 3, line 3. In addition I suggest to indent Listing 2 and 3 to improve readability, and to check prefix declarations (e.g. what does "aml:" mean in Listing 3?). These inaccuracies are not unacceptable for a paper submitted to the Semantic Web journal.
Another critical (and more general) remark is that it is not clear what is the contribution in terms of design and modeling solutions applied to develop the ontology. Is it a mere conversion of XML elements to OWL classes and properties? What are the design and modeling problems solved? What is the added value?
Minor comments and typos:
- Section 3.1.: "we manually implemented several AutomationML data samples, provided on the official AutomationML web-site by means of AMLO". I think that it would be useful to publish these examples (for instance online in the AMLO documentation), so that the reader can compare the XML version with the developed ones;
- Section 4.1., InternalLink: it may be useful to provide an example of InternalLink use. As far as I understand, ODPs may provide better solutions for modeling direct connections between AMLO constructs;
- Section 5: invilved -> involved
|