Resource-Aware Predictive Maintenance System for Injection Blow-Moulding Machines: Reproducibility Artefacts and Edge Deployment Evidence

This record contains reproducibility artefacts for a final-year engineering research project on a resource-aware predictive maintenance system for injection blow-moulding machines. The artefacts support a non-intrusive predictive maintenance pipeline combining reduced-order digital-twin residuals, machine-learning models, TensorFlow Lite compression, Raspberry Pi-class edge deployment, dashboard monitoring, and edge-to-server retraining. The package includes compressed model artefacts, benchmark summaries, Raspberry Pi deployment evidence, dashboard/API evidence, retraining workflow outputs, configuration files and documentation. The artefacts support methodological feasibility under synthetic, proxy-validated and Raspberry Pi feature-replay conditions. They do not constitute full industrial certification on a labelled physical blow-moulding production line.