- The paper introduces automated SPARQL-based mechanisms that detect unit mismatches and incomplete data bindings in ontology-based process models.
- It extends knowledge graphs with explicit unit expectation relations, ensuring rigorous error detection following industry-standard ontologies.
- The approach automates validation of parameter interdependencies, enhancing model reliability and reusability in scalable engineering systems.
Consistency Verification in Ontology-Based Process Models with Parameter Interdependencies
This paper systematically addresses the verification of semantic consistency in ontology-based process models that formalize parameter interdependencies, with a focus on their application to manufacturing systems, particularly Resin Transfer Molding (RTM). Leveraging Semantic Web technologies, the approach integrates standardized process semantics, formal mathematical constructs, and rigorous verification mechanisms to support machine-interpretable engineering models.
Core Contributions
Three primary verification mechanisms are introduced, each targeting a specific area of potential inconsistency in ontology-based process models:
- Contextual Data Filtering with SPARQL: A SPARQL-based filtering mechanism ensures that only data elements contextually relevant to a specific process and its associated interdependencies are retrieved. This is critical for reusable interdependency definitions, such as generic physical laws applied in various contexts, to avoid erroneous parameter binding arising from data belonging to unrelated process instances.
- Unit Consistency Verification: The approach extends the knowledge graph (ParX alignment ontology) by introducing an explicit
ParX:expectsUnit relation, linking mathematical variables to their semantically expected units based on the UNECE ontology. Unit consistency is verified using SPARQL queries that identify and flag mismatches between the expected unit and the actual unit declared for each data element. This formalizes a robust, ontological method for detecting modeling errors introduced through heterogeneous unit conventions, inconsistent data sources, or manual data mapping.
- Data Completeness Checking for Interdependency Evaluation: The availability of all necessary input data for computing process outputs is ensured through recursive SPARQL queries traversing the mathematical expressions modeled in OpenMath-RDF. The process traverses the dependency chain for each variable, confirming that every input variable in any mathematical relation is grounded in an actual data element of the appropriate type and context.
Theoretical and Practical Implications
The methodology demonstrates an alignment with, and significant extension of, existing industrial standards such as VDI/VDE 3682 for process semantics and DIN EN 61360 for data element definitions. By embedding verification mechanisms at the ontology layer, the approach enables:
- Early Detection of Modeling Errors: Errors related to context, data binding, or unit consistency are identified at the design stage, rather than at runtime, increasing model robustness.
- Reusability of Knowledge and Equations: The ability to define formulas generically and apply them reliably across multiple contexts addresses a key bottleneck in model-based engineering, supporting scalable and maintainable knowledge systems.
- Automated and Machine-Interpretable Reasoning: The integration of OpenMath-RDF allows for the precise modeling of mathematical interdependencies, laying the groundwork for automated calculation, simulation, and validation tasks directly from structured process knowledge.
The approach was validated using an RTM use case, modeling each process step according to VDI/VDE 3682 and mapping data and unit semantics through standard-based ontologies and OpenMath-RDF. Empirical results demonstrate that the proposed mechanisms reliably filter context-relevant data, detect unit mismatches (e.g., differences between cm³ and litres in cavity volume specifications), and flag incomplete parameter bindings that would impede model evaluation.
Strong Results & Claims
- Automated SPARQL-based verification is shown to effectively identify not only unit mismatches but also missing or incorrectly bound data elements, even in scenarios with highly reusable and generic interdependency formulations.
- The proposed filtering queries demonstrate necessary selectivity, ensuring only context-relevant data elements—supported by numerical results comparing filtered versus unfiltered query outputs.
- Unit mismatch scenarios generated intentionally (as negative examples in the RTM case paper) were correctly identified, supporting the system's efficacy for real-world error detection in multi-source engineering environments.
Limitations
A notable limitation is the reliance on the explicit, formal definition of parameter interdependencies. In scenarios where mathematical relations are unknown or only empirically specified, manual modeling becomes infeasible. This restricts the immediate applicability of the approach to domains where expert knowledge has already been codified in formal expressions.
Future Directions
The authors identify several key avenues for future work:
- Assistance Systems for Non-Experts: Semantic technology and ontology engineering remain challenging for typical domain engineers. Integration of assistance systems powered by LLMs has shown promise for lowering the barrier to knowledge graph interaction and semantic modeling.
- Closing the Knowledge Gap with Machine Learning: There is significant opportunity to extend the current framework by integrating ML techniques to discover or approximate parameter interdependencies from data and subsequently formalize them for ontological integration.
- Scalability and Robustness: Applying these mechanisms to large-scale, high-dimensional knowledge graphs—characteristic of complex manufacturing organizations—will help surface potential performance and usability constraints, particularly around SPARQL query performance and the maintenance of large ontological structures.
Broader Implications
By tightly integrating standard-compliant ontological modeling with formalized consistency verification, the approach provides a solid foundation for actionable, reliable, and reusable engineering knowledge. In the context of Industry 4.0 and increasing process flexibility, such formalization and validation mechanisms are increasingly necessary to unlock the potential of automation, simulation, and optimal control within knowledge-based engineering environments.
The presented methodology offers a clear path to higher assurance levels in automated model-based reasoning—an essential prerequisite for the practical deployment of digital twins, advanced simulation environments, and semi-autonomous engineering design systems. The extension of the knowledge graph to more deeply integrate LLM-powered assistants and machine-discovered interdependencies signals a maturing convergence between semantic technologies and modern AI approaches.