Fault Classification of Face Milling Tool Using Vibration Signals and Histogram Features – A Machine Learning Approach

In the metal-cutting process, the condition of the cutting tool is critical. The tool condition is one of the factors that impact the surface finish. Monitoring the tool’s condition is necessary to ensure the quality of the end result and productivity. Because vibration signals have a strong relationship with tool state, vibration signals were captured in this investigation while milling mild steel specimens with carbide inserts in a vertical milling machine. Four tool conditions were considered in this study, namely, a good tool (G), a tool with nominal flank wear (FW), tool flaking on the rake face (FL), and tool breakage (B). Histogram features were extracted from the captured vibration signal. J48 algorithm is used to select relevant features, which are then fed into Support Vector Machine (SVM) and K-Nearest neighbourhood (KNN) algorithms. SVM and KNN classification abilities are compared. SVM classifies the tool condition with 88.75% accuracy, whereas KNN achieved the classification with 90% accuracy, suggesting that when it comes to monitoring tool conditions, KNN surpasses SVM.