This work presents a hybrid physical–data-driven modelling framework to relate melt spinning process parameters to the mechanical performance of polymeric yarns. The physical component is based on coarse-grained molecular dynamics approach that links polymer crystallinity and chain orientation to the resulting stress–strain behavior. Using experimental data provided by AITEX, a neural network model was trained to predict key mechanical properties of the yarns. By combining physics-based modelling with machine learning, the proposed approach reduces the computational burden of full physical simulations while preserving consistency between microstructural descriptors and macroscopic mechanical response. Process optimisation is performed using the hybrid model as a predictive core. First, an interactive tool was developed to visualize the influence of process variables on final yarn properties. Subsequently, a genetic algorithm was applied to identify optimal parameter configurations under multiple objectives and constraints. Finally, an AI-based optimisation assistant powered by large language models was implemented to automate the formulation of objective functions and constraints, enabling intuitive industrial deployment.

Industrial assets are increasingly expected to provide persistent, interoperable, and operationally meaningful digital representations across their lifecycle. This paper presents an implemented platform for industrial observability and operational machine passports oriented to heterogeneous assets. The platform supports machine registration, multi-protocol connectivity via OPC UA and MQTT, signal discovery, semantic normalisation, metric export for monitoring, early anomaly detection, and dynamic reconstruction of a machine passport per asset. The architecture combines an industrial exporter, a Prometheus-based observability layer, an analytics service, a persistent asset repository, and a React interface for machine management, monitoring, and diagnosis. Validation is performed in controlled scenarios using reproducible CNC simulators over OPC UA and MQTT. The results show the functional viability of integrating observability, semantic interoperability, and machine passport management within a single implemented platform.

The transition towards Manufacturing-as-a-Service (MaaS) requires not only advanced technological architectures but also a clear alignment with business value and industrial adoption requirements. This paper presents a Business Viewpoint for the MaaSAI project, aimed at integrating stakeholder needs, business processes, and value drivers into the design of interoperable manufacturing services. The proposed approach establishes a structured link between business processes, MaaSAI functional capabilities, and expected business impact, while also considering implementation barriers. The methodology is validated through a stakeholder-driven evaluation involving industrial companies, covering four key business processes across the MaaS lifecycle. The results show that Production and Delivery Monitoring is perceived as the most critical process and the one with the highest improvement potential. Additionally, operational efficiency emerges as the primary business driver, followed by customer satisfaction and profitability. The analysis also identifies key adoption challenges, particularly related to data management, system integration, and budget constraints, with a stronger impact on smaller companies. These findings demonstrate that the proposed Business Viewpoint provides a practical and structured foundation for aligning manufacturing services with real business needs, supporting the transition from architectural design to industrial adoption.