Description
The AI-based Demand Prediction and Asset Estimation Engine enables predictive and autonomous operation of the UniMaaS platform through the realization of closed-loop AI-driven mechanisms for manufacturing demand forecasting, asset availability estimation, anomaly detection, and fault diagnosis. The component leverages historical and real-time monitoring data originating from manufacturing resources, supply chains, devices, and industrial processes to identify dynamic behavioural patterns and operational deviations. The implemented mechanisms incorporate supervised learning, time-series forecasting, deep neural networks, transformer-based D10.2 UniMaaS Interim Dissemination, Communication and Exploitation Report 57 of 120 architectures, and anomaly detection techniques to support proactive maintenance, self-optimization, self-healing, and self-protection capabilities within MaaS environments. The produced predictions and diagnostics are continuously propagated to the planning, scheduling, Digital Twin, and decision-making mechanisms of UniMaaS to improve resource utilization, operational resilience, and quality assurance towards Zero-Defect Manufacturing (ZDM).It supports all four UniMaaS pilots: AEGEAN, ADIENT, ANV, and CATONE with pilot-specific Pydantic validation schemas and is designed for fully local, self-contained deployment within the consortium’s Kubernetes-based testbed.
Main Innovations
- Integration of AI-based demand prediction and asset estimation into a closed-loop autonomous MaaS platform;
- combination of deep learning, transformers, and anomaly detection for manufacturing optimization;
- self-* capabilities (self-optimization, self-healing, self-protection);
- real-time predictive analytics for dynamic manufacturing environments;
- integration with Digital Twins and scheduling mechanisms.
Licence
Apache-2.0
TRL
4 → 6
