Introduction
In industrial sectors such as petroleum, mining, and renewable energy, geological interpretation relies heavily on expert field observations. These observations are traditionally captured through hand-drawn sketches that distill complex underground structures—like faults and stratigraphic layers—into qualitative insights. However, these sketches often remain "trapped" as static images in PDF reports, creating a significant digitization gap. While 3D modeling tools exist for quantitative data, the conceptual, expert-driven understanding captured in sketches is not machine-readable, hindering automated search, comparison, and large-scale analysis.
This paper presents a system designed to bridge this gap by transforming qualitative geological sketches into machine-readable Knowledge Graphs (KGs) using a specialized "GeoStructure" ontology. By capturing information formally at the point of creation, the system ensures that industrial data adheres to FAIR (Findable, Accessible, Interoperable, and Reusable) principles from its inception.
Methodology
The proposed system utilizes a three-layered architecture to facilitate the transition from a visual sketch to a formal data structure.
- User Sketch Interface Layer: This front-end allows geologists to input their interpretations directly. Unlike raw sensor data, sketching acts as a cognitive filter where experts highlight critical entities and relationships while ignoring environmental noise.
- Domain Knowledge Ontology Layer (GeoStructure Ontology): This layer serves as the system's conceptual backbone. It provides a formal hierarchy of geological concepts, including classes for Rocks, Geological Objects, Processes, and Structures. The ontology was developed by aligning professional vocabulary with existing standards like GeoCore and the Fracture Ontology.
- Intermediate Mapping Layer: This layer acts as the bridge, translating graphical elements from the sketch into formal ontological concepts to automatically generate the final Knowledge Graph.
Results and Validation
The primary output of this research is the GeoStructure Ontology, which provides a robust framework for describing structural geological features. The system's efficacy was validated through a real-world use case involving a complex fault system.
As shown in the validation scenario, the system successfully converts field observations into a semantically rich Knowledge Graph. This graph captures both qualitative relationships (e.g., identifying which rock unit constitutes the "footwall" of a specific fault) and quantitative properties (e.g., strike and dip values). This dual capture allows for complex semantic queries that are impossible with traditional flat databases or image-based reports.
Discussion and Future Work
The shift toward ontology-based capture represents a move toward "FAIR at the source." By focusing on qualitative reasoning—such as the spatial relationships between geological units—the system provides a high-level interpretive layer that complements traditional 3D geometric models.
However, several challenges remain. The current user interface is in a preliminary stage, and making formal data entry as intuitive as pen-and-paper sketching is a significant hurdle. Furthermore, while the current ontology focuses heavily on structural faults, future work will expand its scope to other geological disciplines. The authors also plan to conduct extensive field studies with industrial partners to refine the system's usability and mapping complexity. Ultimately, this approach promises to unlock decades of "trapped" geological expertise, enabling automated reasoning and more informed decision-making in critical industrial applications.
Summary updated