Vibration-based turning tool wear classification using time- and frequency-domain features and SVM–RFE

Tool wear during turning operations can compromise surface integrity, dimensional precision, and production efficiency. This study introduces a machine learning-based framework for identifying tool wear conditions using the Support Vector Machine (SVM) algorithm. Vibration signals were acquired through impact excitation and analyzed in both time and frequency domains to extract statistical features representative of wear stages. To improve classification accuracy, Recursive Feature Elimination (RFE) was applied for optimal feature selection. The model was evaluated using a 5-fold cross-validation technique. Results indicate that the highest accuracy was achieved using frequency-domain features with RFE, yielding 100% accuracy in binary classification and 97.2% in multiclass classification. These outcomes emphasize the potential of the proposed approach as a reliable and non-invasive solution for tool wear monitoring in turning applications.