IoT-Driven Digital Twins for Manufacturing Optimization: Hybrid Modelling, Reinforcement Learning and Sustainable Operations

Abstract

Digital twins are virtual representations of physical objects and processes that are emerging as a new foundation technology for manufacturing, enabling real-time monitoring, predictive maintenance and industrial optimization. U.S. manufacturing faces growing pressure to improve productivity and reduce its environmental footprint. To overcome these problems, companies integrate Internet of Things data with digital twins. Trends are identified in IoT-enabled digital twins and hybrid modeling, which integrates physics-driven modeling and data-driven modeling (machine/AI learning) to optimize performance in applications utilizing digital twins. Multi-fidelity simulations, reinforcement learning controllers and real-time data fusion can enable adaptive scheduling and energy-efficient operation of production processes. Twin construction comprises the sensor architecture, data integration framework design using industrial communication protocols, model calibration as well as model validation. The performance of the twin is compared to that of customary control, and twin use cases such as predictive maintenance, virtual commissioning and sustainability optimization in discrete manufacturing and process industries are described. This paper synthesizes a structured implementation framework that integrates multi-fidelity modeling hierarchies, adaptive reinforcement learning control strategies, and sustainability-aware optimization objectives into a cohesive architecture for IoT-enabled manufacturing twins. The framework provides systematic decision guidance for practitioners on sensor network design, hybrid model selection patterns, and performance-sustainability tradeoffs across varying manufacturing contexts. Elsewhere, there is impact on re-shoring of manufacturing to the US, reskilling of the workforce, data governance, the future of cybersecurity, model interoperability, and artificial intelligence-based automation of twin generation.