Cognitive Digital Twin for industrial maintenance: operational framework for fault detection and diagnosis

• Review on Cognitive Digital Twins (CDTs)/DTs for industrial maintenance • Definition of the CDT cognitive capabilities and key enabling technologies • Proposal of a CDT framework to support Fault Detection and Diagnosis (FDD) • Development of a novel ontology formalizing FDD expert knowledge • Application of the CDT framework on a laboratory platform Digital Twin is a cutting-edge technology designed to address disruptions in manufacturing operations by supporting humans in complex maintenance decisions via advanced data analytics and real-time synchronization. However, as the complexity of decisions increases, enhanced capabilities are required, such as reasoning and context awareness, leading to the Cognitive Digital Twin (CDT) concept. In this context, this work offers two contributions. First, it presents a state-of-the-art review on CDT for maintenance in manufacturing, identifying Fault Detection and Diagnosis (FDD) as a relevant investigation area. Second, it proposes a novel CDT framework specifically tailored to support FDD in industrial maintenance. The contributions are twofold: (i) an ontology that formalises maintenance expert knowledge and supports diagnostic reasoning; and (ii) data-driven algorithms that elaborate data from the physical system, and instantiate or update the proposed ontology. The structured integration of ontology and data analytics into an operational CDT framework enables and properly places all six cognitive capabilities - perception, attention, memory, reasoning, problem-solving, and learning - within a domain-specific framework tailored to maintenance, and especially to support FDD decisions. The CDT output is the augmented information flowing to the maintenance decision-making process, which is held by the maintenance staff, who, after the completion of the FDD activity, can act back on the physical asset with the required maintenance interventions. The CDT framework is finally tested in a laboratory setting, demonstrating its functional effectiveness in supporting maintainers in the FDD decision-making process by formalizing knowledge and guiding reasoning.