Abstract
Modern automated manufacturing networks are increasingly dependent on the operational reliability of distributed physical assets, ranging from industrial transformers and robotic subsystems to sensor-driven production modules. As manufacturing systems evolve toward fully digitalized and interconnected architectures, the need for data-centric asset health management (DCAHM) becomes critical for ensuring uninterrupted production efficiency, minimizing downtime, and optimizing lifecycle costs. This paper proposes a comprehensive analytical synthesis of data-driven prognostics and health management strategies adapted from power system reliability frameworks and extended into automated manufacturing environments.
The study integrates machine learning-based predictive maintenance approaches, statistical reliability modeling, and fuzzy logic-driven decision systems to establish a unified conceptual foundation for asset health evaluation. Techniques such as sequence learning-based health index prediction (Dong et al., 2021), modified Weibull reliability modeling (Dong & Nassif, 2018), and neural network-based diagnostic frameworks (Murad et al., 2023) are examined and contextualized within manufacturing networks. Additionally, fuzzy logic-based remnant life estimation models (Bakar & Abu-Siada, 2016) and multiparameter health assessment techniques (Mharakurwa & Goboza, 2019) are incorporated to address uncertainty in heterogeneous industrial environments.
The paper further explores how data-centric predictive architectures can be generalized beyond electrical asset management into broader manufacturing ecosystems characterized by distributed intelligence and real-time operational feedback loops. A comparative synthesis of existing methodologies highlights the transition from rule-based maintenance systems to adaptive, AI-driven prognostic frameworks. The findings emphasize that hybrid modeling approaches combining statistical reliability theory and machine learning significantly improve prediction accuracy and operational decision-making.
Finally, the study identifies key limitations, including data heterogeneity, scalability constraints, and interpretability challenges in deep learning models. The research concludes that integrating structured health indices with real-time data streams can substantially enhance output efficiency in automated manufacturing networks, while also supporting sustainable asset lifecycle management strategies.