An IoT-enabled hybrid machine learning approach for tool wear classification and power monitoring in legacy lathe machines

The transition toward Industry 4.0 presents significant challenges for small and medium-sized enterprises (SMEs), particularly in retrofitting legacy machinery for intelligent monitoring and decision-making. This study introduces an IoT-enabled framework for real-time power monitoring and tool wear classification in a legacy lathe machine. High-resolution current and voltage data were acquired using both industrial-grade and cost-effective sensors. Time-domain features were extracted from the acquired signals, and Shapley Additive Explanations (SHAP) were employed for feature selection, ensuring interpretability and model robustness. A hybrid machine learning model was developed by combining Support Vector Machines (SVM), XGBoost, and Multi-Layer Perceptron (MLP), with logistic regression serving as the meta-classifier. The proposed model achieved a classification accuracy of 96% on the C1 dataset, outperforming individual models across all evaluated metrics. Furthermore, a dynamic, web-based dashboard was implemented to visualize real-time and historical power consumption trends, facilitating data-driven decisions related to energy efficiency and predictive maintenance. The proposed framework provides an affordable and flexible way to improve operations in older manufacturing systems and helps small and medium-sized enterprises join smart manufacturing setups.