Complex things in an IIoT system require more advanced information models with well-defined underlying ontologies, extensible semantic description capabilities and data structures.
Sharing and reuse of data between applications is a big challenge in building IIoT systems. Knowledge models for a specific application tend to be tightly coupled with the application-specific concepts, which makes moving data between applications a difficult task to undertake.
Knowledge models created using fixed data structures or relational database schemas are limited to a pre-defined set of concepts and relationships.
This is because this approach addresses the structural representation of data, not the semantics of data. Traditional methods such as database schemas and XML schemas provide limited expressivity in modeling knowledge, which forces business logic that should really be part of the knowledge model to be embedded into applications. This makes it difficult and expensive for users to switch to new software, often leaving them stuck with obsolete software.
Decoupling the knowledge model from the application is critical to enabling data sharing and reuse between systems or across knowledge domains. All of the business logic in the knowledge model should be captured in the model itself and not in the applications.
With two systems each using a different schema for its database or information model, data sharing is achieved by translating from one schema into the other. Data sharing would occur through a process of translation. A dump of data would be mapped and translated from one schema to the other. The original information is replaced with new information in a form conforming to the target schema. This means any data and metadata that cannot be directly translated will be lost.
On the other hand, information semantically described in a knowledge model can be shared across domain boundaries by defining the concepts of the foreign domain using the concepts of the local domain. This is a descriptive approach instead of the translational approach seen for data sharing based on databases and other data structures. With the mapping being descriptive, the data has meaning in both domains and the full fidelity of the original data is maintained.
For example, Semantic Web technologies such as Resource Description Framework (RDF), RDF Schema (RDFS), Web Ontology Language (OWL) and SPARQL support this descriptive, semantics-based approach of data sharing. Resource Description Framework (RDF) provides a way to model information in a graph. However, it does not provide a way of specifying the meaning of that information by itself.
Other means of defining a vocabulary of terms with semantics for describing the information are therefore necessary:
- A simple vocabulary is a collection of defined terms used in communication.
- A taxonomy is a vocabulary in which terms are organized in a hierarchical manner. In a taxonomy, for example, you can define that both Speedometer and GPS are sensors, or both mirrors and doors belong to a body in a vehicle.
- An ontology is vocabulary of terms to define concepts and the relationships between them. Taxonomies are more expressive than vocabularies, and ontologies are more expressive than taxonomies. Ontologies allow you to express the semantics behind vocabulary terms, their interactions, and context of use.
In Semantic Web technologies, RDFS supports taxonomies, and OWL extends RDFS to provide a language for defining ontologies that capture semantics of domain knowledge. Ontologies are the core element of Semantic Web.
Not all existing information models are based on Semantic Web technologies. For example, the OPC UA information model uses object-oriented model paradigms with features such as classes and entities (Object Types and Objects), properties, attributes and methods. It is more descriptive than a taxonomy. There are successful examples of attempts to map those non-Semantic Web information models to a formal ontology such as OWL. Mature Semantic Web Frameworks such as Jena are available, and mapping to OWL ontology allows complex queries on the information model and automatically apply reasoners to generate new knowledge using those frameworks.
Our new IIC white paper, “Characteristics of IIoT Information Models,” surveys a subset of information models that are relevant to IIoT and characterizes those information models using a meta-model developed for this purpose. We also look at the layers of abstraction in an information model, ranging top down from the Conceptual Layer at the highest level, to the Semantic Layer, the Base Layer and finally the Serialization Layer for the actual data transfer of a medium. With this work, we capture commonalities and can begin to address the challenge of integrating subsystems that use different information models.
This blog is excerpted from the white paper, which is posted on the IIC website here.
Kym Watson is principal scientist at Fraunhofer IOSB.