CONICET | Buscador de Institutos y Recursos Humanos

In this work, a ontology-based supervision system [1] has been extended to support risk management in Hydrogen production processes. In knowledge engineering, ontology is a formal representation scheme to build Knowledge Bases (KBs) as a set of concepts and their relationships. An advantage of the proposed scheme, is the use of a description logic reasoner that checks the consistence of knowledge models and provide logical-formal justifications of all its results. It brings transparency to non expert users thanks to the use of the semantic of the plant units, control systems, fault and hazard analysis, captured in the ontology. The proposed system detects the fault conditions through a multi-variant data analysis. Then, it isolates faults using an inference diagnosis method based on fault signatures. Hazard classes are identified and risk measures are deduced based on the severity and likelihood of the situation. Finally, the system suggests some corrective actions to maintain the process in a safe and reliable operation zone. Alarm management models are also captured into the KB, allowing the reasoner to set and report the actual alarm states. The H2 production from bio-ethanol reforming considered here consists of five reactors and a PEM fuel cell for generating electric power. This plant is designed to work onboard of a vehicle. Fifteen failures were considered such as the catalyst deterioration in the reforming reactor, faults at the fuel cell and auxiliary equipments, sensors drifts and valves clamping. They have been simulated in order to identify deviation patterns and their associated hazards. The optimal measurements selection for building a proper monitoring system is done by using a fault detectability index. The results show how the system is capable of isolating faults involving catalyst problems at the reforming reactor.