Proposta de Tese de Doutorado de Nicolau Oyhenard dos Santos

Proposta de Tese de Doutorado

Aluno(a): Nicolau Oyhenard dos Santos
Orientador(a): Mara Abel

Título: COPI: A Core Ontology for Petroleum Installations—Development, Validation, and Application
Linha de Pesquisa: Aprendizado de Máquina, Representação de Conhecimento e Raciocínio

Data: 21/08/2026
Hora: 08:00
Local: Esta banca ocorrerá de forma remota. Acesso público disponibilizado pelo link https://conferenciaweb.rnp.br/sala/mara_abel.

Banca Examinadora:
-Joel Luis Carbonera (UFRGS)
-Melinda Hodkiewicz (UWA)
-Ismael Santos (Petrobras)

Presidente da Banca: Mara Abel

Resumo: Offshore petroleum production plants are complex industrial facilities where thousands of equipment items—separators, compressors, pumps, sensors—generate heterogeneous data across maintenance records, sensor time series, equipment hierarchies, and event histories. Achieving semantic interoperability across these data sources is essential for tasks such as production loss analysis, where engineers must trace failure causes by connecting information from systems that were designed under incompatible conceptual models. The core obstacle to interoperability is threefold: (i) industrial standards (ISO 14224, ISO 15926-4, CFIHOS, DEXPI) define equipment types and positions using natural-language descriptions without formal axioms, preventing automated reasoning; (ii) when physical equipment is replaced at a given position, historical data loses its semantic anchor because existing data models conflate the identity of the position with the identity of its occupant—the persistence problem; and (iii) industrial reference data libraries make incompatible ontological commitments about what equipment types are, causing cross-standard interoperability failures that vocabulary-level mappings cannot resolve. This thesis proposal introduces COPI (Core Ontology for Petroleum Installations), a BFO/IOF-Core aligned core ontology to be constructed through three complementary methodological instruments: (1) an ontology design pattern for functional locations that separates four concerns conflated in industrial practice—spatial site, organizational code, equipment tag, and functional specification—grounding persistence in BFO’s site semantics through an ontological elimination argument; (2) a five-step axiomatization method for enriching industrial reference data library classes with BFO/IOF-Core-grounded axioms (identity-giving function, genus definition, differentia specifica, necessary parts, and process participation signature) paired with a typology of five cross-standard conceptual divergence types that the method simultaneously diagnoses; and (3) the COPIeditor, an LLM-assisted ontology engineering platform that operationalizes the five-step method at industrial scale while preserving logical consistency, annotation completeness, and cross-standard coherence through integrated verification. While developed for petroleum production, these methodological instruments address general problems in applied ontology—identity persistence, cross-standard conceptual divergence, and tractable rigorous construction—applicable to any asset-intensive industrial domain. COPI is validated through an integrated deployment architecture over real operational data from a tower gas dehydration unit of an offshore platform, combining Virtual Knowledge Graph mediation via Ontop over a Trino federation layer (unifying access to relational data on PostgreSQL—equipment hierarchy, functional location codes and tags, maintenance work orders, and SmartPID topology—with sensor time series on InfluxDB) and a Model Context Protocol server generated automatically from the ontology, enabling natural language access to industrial data by AI agents. Validation is strengthened by executing a set of case studies derived from real production-loss investigation workflows, in which analysts resolve informal equipment references, navigate functional-location-to-sensor topology, and retrieve historical observations by posing questions in natural language. The methodology follows LOT (Linked Open Terms) combined with BFO design guidelines and ontology design patterns.

Palavras-Chave: Ontology Engineering Petroleum Installations Knowledge Graphs Semantic Interoperability Functional Locations Basic Formal Ontology