Abstract
Predictive maintenance (PdM) is a cornerstone of smart manufacturing, aiming to reduce unplanned downtime and optimize maintenance scheduling. Digital twin (DT) technology offers a promising enabler by providing a virtual replica of physical assets for real-time monitoring and simulation. This paper proposes a hybrid digital twin framework that integrates physics-based models with machine learning algorithms to enhance predictive maintenance in manufacturing systems. The methodology involves developing a digital twin of a CNC machine tool using sensor data (vibration, temperature, and acoustic emissions) and historical maintenance logs. A convolutional neural network (CNN) is employed for fault detection, while a physics-based degradation model estimates remaining useful life (RUL). The framework is validated using a publicly available milling machine dataset. Results show that the hybrid approach achieves a fault detection accuracy of 96.2% and a mean absolute error (MAE) of 3.1 hours for RUL prediction, outperforming pure data-driven or physics-based methods. Comparative analysis reveals a 22% reduction in false positives and a 15% improvement in maintenance scheduling efficiency. The study demonstrates that integrating digital twins with predictive analytics can significantly enhance manufacturing reliability and operational efficiency. The proposed framework is scalable and adaptable to various industrial assets, offering a pathway toward autonomous maintenance in Industry 4.0 environments.