SFE-O: An Ontology for Knowledge-Grounded AI in Supercritical CO$_2$ Biomass Extraction and Fractionation

AI summary

Overview

Research area: Semantic knowledge representation and ontology engineering for chemical process engineering, specifically supercritical CO₂ biomass extraction and fractionation, with an eye toward knowledge-grounded AI.

Technical level: Advanced. The paper assumes familiarity with OWL 2, ontology design, semantic web vocabularies, SPARQL querying, and supercritical fluid extraction concepts.

Scope: The paper presents SFE-O, a modular OWL 2 ontology that organizes heterogeneous knowledge about biomass, process conditions, mass transfer, and products into a machine-actionable model for supercritical CO₂ extraction.

What This Paper Is About

Supercritical CO₂ extraction is a sustainable way to recover valuable compounds from biomass, but designing and interpreting these processes requires knowledge scattered across reports, engineering models, databases, and expert experience. The paper introduces SFE-O to unify that knowledge in a formal ontology, so that process data, parameters, mechanisms, and provenance can be represented and queried consistently. The goal is a shared knowledge layer that supports explanation and consistency checking now, and ontology-grounded AI reasoning in the future.

Key Contributions

1. A modular OWL 2 ontology for CO₂-SFE: SFE-O is organized around three interconnected dimensions—Biomass, Process, and Product—covering the core entities of supercritical CO₂ extraction and fractionation.

2. Explicit representation of process mechanics and provenance: The ontology models process topology, the origin of parameters, external and internal mass-transfer phenomena, and the Constant Extraction Rate (CER) / Falling Extraction Rate (FER) mechanisms.

3. Declarative computational dependencies with traceability: Relationships among process quantities are represented explicitly without embedding numerical solving in the ontology, allowing input, measured, computed, externally retrieved, and fitted parameters to remain distinguishable and traceable.

4. Reuse of established semantic resources and an initial evaluation: SFE-O reuses ChEBI, QUDT, OWL-Time, and PROV-O, and is evaluated through a worked example instantiating a real pilot-scale extraction batch, a reasoner-based consistency check, and a competency-question SPARQL query.

Main Findings

• Three-dimension knowledge structure: The ontology connects biomass properties, process conditions and mechanisms, and product outcomes in a single modular framework rather than treating them as separate data silos.

• Parameter provenance is first-class: By distinguishing input, measured, computed, externally retrieved, and fitted parameters, SFE-O keeps the origin and status of each process quantity explicit.

• Mechanisms are represented, not simulated: CER/FER behavior and internal/external mass-transfer phenomena are captured conceptually, while numerical solving is deliberately left outside the ontology.

• Standards-based interoperability: Reuse of ChEBI, QUDT, OWL-Time, and PROV-O connects the ontology to existing chemical, unit, temporal, and provenance vocabularies.

• Demonstrative evaluation only: The abstract reports a worked example on a real pilot-scale batch, a reasoner consistency check, and a competency-question SPARQL query, but gives no quantitative results, performance figures, or comparison baselines.

• AI support is motivated, not evaluated: The authors position SFE-O as a knowledge layer for future ontology-grounded AI reasoning, while explicitly stating that this direction is not evaluated in the paper.

Methodology in Plain English

The authors built a formal vocabulary for a specific chemical process. They identified the key concepts involved in supercritical CO₂ extraction—what the biomass is, how the process is configured and operated, what happens during mass transfer, and what product results—and organized them into interconnected modules. Instead of writing equations into the ontology, they described which quantities depend on which others, so a computer can track where a value came from and how it relates to the rest of the process. They reused existing semantic standards for chemistry, units, time, and provenance rather than inventing new ones. To test the model, they encoded a real pilot-scale extraction batch, checked the ontology for logical consistency with a reasoner, and ran a SPARQL query corresponding to a competency question.

Why This Matters

Impact on research: SFE-O offers a shared, machine-readable representation for a process area where knowledge is fragmented, creating a foundation for comparing experiments, integrating models and databases, and eventually supporting AI systems that reason over extraction knowledge rather than only fitting data.

Real-world applications:

• Designing and documenting supercritical CO₂ extraction runs for high-value natural compounds.
• Supporting scale-up from laboratory or pilot batches by keeping operating conditions, biomass properties, and product outcomes linked.
• Integrating data from technical reports, engineering models, and external chemical databases into a consistent knowledge base.
• Providing mechanistic explanations and provenance-aware decision support for process engineers.

Industry relevance: Sectors using CO₂ extraction—such as food, nutraceuticals, cosmetics, and pharmaceuticals—depend on reproducible process knowledge and reduced reliance on organic solvents. A formal ontology can help companies organize proprietary process data, trace parameter origins, and build interoperable knowledge systems.

Future Directions

• Ontology-grounded AI reasoning: Develop and evaluate AI systems that use SFE-O as a knowledge layer, since the paper motivates this direction but does not test it.

• Broader validation: Apply the ontology to more extraction batches, biomass types, and operating conditions beyond the single worked example described in the abstract.

• Linkage to numerical models and simulation: Connect the declarative dependency structure to solvers or simulation tools while preserving parameter provenance.

• Extension and alignment: Extend SFE-O to related separation and fractionation processes and align it with other process-engineering and semantic-web ontologies.

Target Audience

This paper is most useful to ontology engineers and semantic-web researchers working on scientific and industrial knowledge representation; process engineers and chemists interested in formalizing supercritical CO₂ extraction knowledge; and AI researchers exploring knowledge-grounded reasoning for process systems. Readers seeking experimental extraction results or quantitative performance evaluations will not find them in this abstract.

Authors’ abstract

Supercritical CO$_2$ extraction (CO$_2$-SFE) offers a sustainable route for recovering high-value compounds from biomass while limiting the use of conventional organic solvents, but its design and interpretation require heterogeneous knowledge spanning biomass properties, operating conditions, mass transfer, and product fractionation, typically fragmented across technical reports, engineering models, external databases, and expert practice. This paper introduces SFE-O, a modular Web Ontology Language (OWL 2) ontology for supercritical CO$_2$ biomass extraction, structured around interconnected Biomass, Process, and Product dimensions and extended with explicit representations of process topology, parameter provenance, external and internal mass-transfer phenomena, and the Constant Extraction Rate (CER) / Falling Extraction Rate (FER) mechanisms. Computational dependencies among process quantities are represented explicitly, without embedding numerical solving in the ontology, so that input, measured, computed, externally retrieved, and fitted parameters remain traceable. SFE-O reuses established semantic resources (ChEBI, QUDT, OWL-Time, PROV-O) and is evaluated through a worked example that instantiates a real pilot-scale extraction batch, a reasoner-based consistency check, and a competency-question SPARQL query. The resulting knowledge model provides a machine-actionable foundation for querying, consistency checking, and mechanistic explanation of SFE processes, and establishes a knowledge layer that could support future ontology-grounded AI reasoning, a direction this paper motivates but does not evaluate.