Smart Manufacturing Framework for Real-Time Process Monitoring, Predictive Maintenance, and Quality Control in Advanced Mechanical Production Systems

Abstract

This study develops an integrated smart-manufacturing framework that unifies real-time process monitoring, predictive maintenance, and predictive quality control within a closed-loop decision architecture for advanced mechanical production systems. The proposed framework combines multi-sensor acquisition, edge analytics, digital-twin state estimation, anomaly detection, remaining useful life (RUL) prediction, and quality-risk forecasting to support adaptive control of machining and assembly operations. Unlike fragmented monitoring strategies that treat maintenance and quality as separate functions, the present approach fuses equipment-health indicators and part-quality indicators into a common risk score used for supervisory decision-making. A simulation-based industrial case study was constructed to emulate a high-mix mechanical production cell with CNC machining, in-process sensing, and end-of-line inspection. The case study was configured using literature-informed process logic and representative parameter bounds for spindle speed, feed rate, thermal load, vibration, current, and dimensional deviation. The results show that the integrated framework can detect degradation earlier than threshold-only monitoring, improve RUL tracking stability, and reduce quality escape by linking machine-state evolution to downstream defect probability. In the simulated evaluation, anomaly-detection F1-score increased from 0.79 to 0.94, RUL mean absolute error decreased from 21.8 to 12.6 cycles, and quality-prediction AUROC increased from 0.84 to 0.96. At the operational level, the proposed strategy reduced monthly unplanned downtime from 18.6 h to 10.8 h, lowered scrap rate from 4.8% to 2.2%, and increased overall equipment effectiveness from 71.2% to 81.6%. These findings indicate that a unified monitoring-maintenance-quality architecture can provide stronger production resilience and more economically efficient decision support than isolated digital initiatives. The manuscript is intentionally written as an original-research draft built around a simulation-based validation study; plant-scale experimental verification is the next required step before journal submission.