Enhancing Electric Machinery Fault Prognostics with Human-in-the-Loop AI for Industry 5.0

Human industry has evolved from isolated, empirical, manually operated systems to fully interconnected, data-driven, automated ones. The key advancements in this undertaking categorize this progress into industrial revolutions: mechanization (Industry 1.0) to mass production (Industry 2.0) to computer automation (Industry 3.0) to interconnectivity (Industry 4.0). As tools responsible for the latter are growing ever more complex, their inner workings are increasingly obscured to users, reducing trust and dependability compared to known, understandable methods, which remain the standard in industry. Real-world, actual deployments showcase limited adoption, guiding researchers towards systems that offer transparency regarding their decision variables, processing, and output. Over-reliance on the latest hallmark of Artificial Intelligence (AI) poses issues with ethical concerns, reduced human oversight, and untraceable mistakes. Industry 5.0 is poised as the next revolution which responds to this need by shifting focus toward practical integration, human-machine collaboration, and interpretability. A new literature paradigm, the Human-in-the-Loop (HITL), is introduced to put human expertise in the center of the decision-making pipeline. When combined with the other advancements that characterize Industry 5.0, seamless human participation enhances system reliability and maintains transparency throughout the diagnostic workflow. This chapter documents an integration of the Industrial Internet of Things (IIoT) with AI concerning the evolving priorities of Industry 5.0, in the subfield of advanced electrical machine fault diagnostics. A case study is presented as a practical example to illustrate the nuances encountered in industrial settings as discussed. It includes an analysis of data availability and system dynamics using the two prevalent tasks of: state classification and index regression. Challenges and limitations are thoroughly analyzed. Model output uncertainty quantification and result interpretability are explained. The chapter demonstrates how these innovations contribute to a more resilient, efficient, and sustainable industrial sector. It offers valuable insight for industry professionals to achieve better practices in line with the future of industrial development.