The handover problem: why two standards describing the same plant do not connect

The challenge of integrating industrial data across different standards is often misunderstood. Many assume that failures in data handover stem from differences in file formats or syntax. However, the deeper issue lies in semantic incompatibility, even when both systems conform to established standards. This lesson explores why two seemingly compatible industrial ontologies fail to interoperate at the semantic level.

The Semantic Disconnect in Practice

The scenario involves a process plant designed under the ISO 15926 standard and handed over to an operator using an IFC or CFIHOS-based asset system. Despite both systems being standards-compliant, integration fails. The common assumption is that different file formats are to blame. However, in many cases, both sides can generate RDF. The real issue is that the same term denotes different concepts, and different terms denote the same one.

Measuring Ontological Incompatibility

When comparing the lexical overlap between CFIHOS V2.0 and IFC4 ADD2, the result is extremely limited. Across 1,397 and 1,286 classes respectively, only one class name, PhysicalQuantity, is shared exactly. This demonstrates a significant disconnect in how concepts are modelled, even when both systems are valid and conformant. The mismatch is not one of syntax but of conceptualisation.

Comparing ISO 15926 and IFC4 Ontologies

When comparing the IDO (ISO 15926-14) line with IFC4, the results are similarly unconvincing. Out of 1,286 classes in IFC4, only 7 token matches are found with IDO. These matches are largely false friends, meaning they superficially align but refer to different semantic constructs. This highlights the limitations of relying on surface-level alignment for semantic interoperability.

Implications for Crosswalks and Alignment

This level of incompatibility makes lexical or string-based alignment ineffective. The use of similarity metrics or string matching is not just suboptimal, it is close to useless in practice. Real-world crosswalks require a deeper understanding of the models and are fundamentally exercises in semantic engineering. The process must move beyond matching to one of modelling alignment.

Building a Reproducible Ontology Pipeline

This course introduces a reproducible pipeline for handling such challenges. It includes fetching standards, measuring their structures, proposing semantic correspondences, validating these with SHACL, promoting them to OWL, and finally reasoning over the integrated model. This structured approach ensures that semantic interoperability is not left to chance or approximation.

What to take away

The failure of industrial data handover is not due to file format incompatibility but to deep semantic misalignment. Even when both systems are standards-conformant, as few as 1 out of 1,286 classes may match between standards. This course provides a methodological framework to address such issues through structured reasoning and validation rather than surface-level matching.

Reference

Apply it to your work