AI-Enhanced Process Digital Twins for Circular Manufacturing: Design, Architecture, and Deployment

The design and deployment of Industry 4.0 (I4.0)compliant systems remain challenging due to the complexity of integrating digital models, semantic interoperability, and scalable architectures. The Reference Architecture Model for Industry 4.0 (RAMI 4.0) and the Asset Administration Shell (AAS) offer standardized frameworks to bridge physical assets with their digital counterparts. However, widespread adoption is limited, particularly in dynamic and heterogeneous manufacturing environments. This paper presents the architecture, implementation, and validation of AI-enhanced Process Digital Twins (PDTs) tailored for circular manufacturing. The proposed PDT integrates real-time monitoring via AAS, and embedded AI capabilities for adaptive decision-making. A multi-viewpoint architectural model ensures modularity, scalability, and alignment with $\mathbf{I 4. 0}$ principles. The solution has been validated across three industrial pilots - WEEE, BATT, and PETRO - demonstrating interoperability, orchestration of distributed DTs, and support for circular economy goals. Results show improved responsiveness, autonomy, and sustainability in process operations. By sharing key design patterns, technological enablers, and deployment insights, this paper offers a reusable framework for advancing the adoption of $\mathbf{I 4. 0}$-compliant digital twin ecosystems.