Time-Domain Vibration-Based Fault Diagnosis of Round Insert Face Milling Tools Using SVM and XGBoost: A Machine Learning Approach

Abstract

Tool Condition Monitoring (TCM) plays a significant role in maintaining machining quality, reducing equipment idle time, and improving TCM often affords near-real-time detection of wear-out phenomena. A new method for vibration-based fault diagnosis of round insert face milling tools based on analysis in the time-domain and machine learning, the vibration signals were obtained using a spindle-integrated piezoelectric accelerometer during milling with controlled conditions. Healthy, flank wear, edge chipping, built-up edge and mixed fault conditions were studied from the statistical features extracted. Feature selection was implemented using Recursive Feature Elimination and Mutual Information ranking. Distributed computing was used to pilot SVM and XGBoost classifiers. The output indicated that the test accuracy of XGBoost was better (92.4% accuracy) than SVM (89.2%), while both had lower errors and shorter estimates as well. The presented approach is a cost-effective and real-time applicable method for intelligent monitoring of the condition of the monitoring tool in CNC machining. This study contributes to the development of data-driven predictive maintenance systems for smart manufacturing.