Vibration-Based Machine Learning for Tool Fault Diagnosis and Real-Time Monitoring in Machining Operations

Defects in tool inserts have major effect on productivity and product quality in contemporary machining operations, which are frequently missed by conventional condition monitoring techniques. To overcome this, the study investigates a machine learning-driven method in line with Industry 4.0’s self-monitoring paradigm for vibration-based fault diagnosis of tool inserts during machining. To distinguish between defective and fault-free tool conditions, vibration data gathered during face machining are analyzed. The Decision Tree (DT) algorithm is used to reduce dimensionality after the statistical feature extraction technique. Two supervised Bayesian family classifiers, Naïve Bayes and Bayes Net, are then used to classify the processed data. Comparatively speaking, the Naïve Bayes classifier outperforms Bayes Net (93.75%) in terms of accuracy (96.25%) and training time (0.2 s versus 0.25 s). These results open the door for intelligent and self-governing machining systems by proving that Naïve Bayes is more effective and dependable algorithm for real-time tool condition monitoring during turning operations, wherein the lathe’s vibration-sensor-based control system reduces defects, decreases inspection turns, and increases productivity by enabling early tool-wear effect detection, automatic shutdown under excessive vibration, and real-time operator guidance for decision-making during roughing and finishing. Applications for real-time monitoring are further supported by thresholds established during manual turning.