Toward a modular ontology of process supervision

RODA, F.; BRACCIA, L.; BOTTARI, A.

Buenos Aires

Congreso; WCCE11 - 11th WORLD CONGRESS OF CHEMICAL ENGINEERING; 2023

Asociación Argentina de Ingenieros Químicos

The management and exploitation ofexpert knowledge are critical in thesupervision and control of industrial processes. According to theAbnormal Situation Consortium (ASM) most incidents in industries areattributable tohuman errorsrelated to inadequate information and misunderstandings of theprocess and its underlying operational condition. In that sense,semantic technologies and mainly ontology-based frameworks haveproven to be valuable tools as they may provide a consistent andunambiguous description of the process and it context. Beside theremarkable contributions made by large-scale ontologies such asONTOCAPE and the ISO 15926 ontology, to the best of our knowledgea sound and formal modeling of supervisionconcepts (e.g. fault, failure,symptom ormalfunction) isstill missing.In this work, anovel modular ontology of process supervision is presented. Itinvolves technical terms of fault detection, diagnosis and protectiontasks, whose communication and correct interpretation is critical forthe plant safety. In the pursuit of a high reusability knowledgemodel, an extensive study of the semantic of these concepts wasperformed looking for a common understanding regarding theirmeanings. The conceptualization required a review of more than ahundred of process supervision concepts. Most of the terminology wastaken from the works realized by the IFAC-TechnicalCommittee SAFEPROCESS(http://www.asmconsortium.net) andcontributions done bytheauthors such as RolfIsermann and Venkat Venkatasubramanian. Sincesupervision concepts definitions are not unique, an effort has beenmade to come to consistency and standardization. In order to improvethe ontologyadaptability andreusability, it has been organized in five modules: i)Supervision Tasks: a taxonomy of faultdiagnosis and detection methods; ii) SystemStates: defines the plant conditions and states; iii) SafetyProperties: process features with regards to the safety conditions;iv) Fault Conditions: defines faultsand failures clasess using a multiple inheritance taxonomy; v)Feature abstractions: featureextraction-related concepts (e.g. featureand symptom).The ontology was implementedusing OWL2(Web Ontology Language) underthe SROIQDescription Logic(DL) expresive power. Beyonadressing interoperability issues, the proposed model aim attransfering some key reasoningresponsibilities that usually reside in specialized distributedagents to DL reasoners whose outputs are self-explaineddeducctions based on logic axioms. Among others, the axioms supportthe following use cases: a) the automatic classification of plantevents and conditions; b) thecharacterization of process safety; c) the work flow trace of processsupervision tasks and d) the searchfor fault detection and diagnosis methods for a given situation. Thiswork ilustrates how the proposed ontologycanbe used to address the aforementioned use cases allowing expert andnot expert agents to get a better understanding of the plantcondition and state.p { line-height: 115%; margin-bottom: 0.25cm; background: transparent }