Temporal feature analysis of audio signal for instability identification in high-speed micro milling of thin-walled Ti6Al4V

Abstract

Ti6Al4V is widely used in aerospace and biomedical components due to its high strength, low weight, and biocompatibility. However, machining this material presents challenges, particularly due to its low heat dissipation capacity. In thin-walled micro milling, regenerative chatter is inevitable. The present study proposes a novel method for chatter detection during high-speed micro milling of thin-walled Ti6Al4V, utilizing Euclidean Distance (ECD) and Principal Component Analysis (PCA). Audio signals recorded during machining have been segmented into equal parts, and ECD has been used to detect variations between segments. Root mean square (RMS) is extracted for each segment and the same has been used in PCA to classify the cutting conditions. It has been observed that a minimum of three principal components (PCs) are essential for accurate and reliable chatter detection. Radial micro milling experiments have been conducted at spindle speeds ranging from 20,000 to 80,000 rpm, with radial depths of cut varying between 30 µm and 130 µm. A threshold of 3% has been established, where ECD variations below the threshold indicate stable cutting and variations above the threshold show chatter onset. These predictions have been validated through surface analysis for the presence or absence of imprinted chatter marks for different cutting conditions. The proposed approach can be used as real-time feedback on machining stability without requiring complex signal transformations or solving governing equations, making it a computationally efficient solution for industrial applications.