A Hybrid Digital Twin Optimization Framework for Predictive Maintenance Scheduling in Smart Manufacturing Systems

Modern manufacturing systems increasingly depend on interconnected industrial assets whose operational continuity is essential for productivity, supply chain resilience, and infrastructure reliability. Conventional preventive maintenance approaches often rely on fixed maintenance intervals that fail to capture dynamic equipment degradation, while reactive maintenance results in costly production interruptions and reduced asset utilization. This study proposes a hybrid digital twin optimization framework that integrates stochastic degradation modeling, virtual asset representation, Bayesian state estimation, and multi-objective maintenance optimization to improve maintenance scheduling in smart manufacturing environments. The proposed framework constructs a digital representation of industrial equipment capable of continuously estimating asset health through probabilistic degradation models and virtual operational states. Equipment reliability is represented using continuous-time degradation processes, while maintenance decisions are formulated as a constrained optimization problem minimizing lifecycle maintenance cost, production downtime, and reliability risk. Bayesian state estimation continuously updates component health estimates, and a multi-objective optimization algorithm determines optimal maintenance actions under operational constraints. The proposed framework was analytically evaluated using the NASA CMAPSS turbofan degradation dataset and the PHM Society challenge datasets. Performance evaluation includes maintenance cost reduction, remaining useful life prediction accuracy, equipment availability, maintenance frequency optimization, and production continuity metrics. The proposed framework provides a scalable analytical foundation for predictive maintenance in smart manufacturing systems without requiring proprietary operational data.