Transfer learning for autonomous chatter detection in machining

Large-amplitude chatter vibrations are one of the most important phenomena in machining processes. It is often detrimental cutting operations causing a poor surface finish and decreased tool life. Therefore, detection using machine learning has been an active research area over last decade. Three challenges can be identified applying for at large industry: insufficient understanding universality features across different processes, need automating feature extraction, existence limited data each specific workpiece-machine combination. These three grouped under umbrella transfer learning. This paper studies by evaluating prominent as well novel methods. We investigate classification accuracy variety extracted from turning milling experiments with configurations. The studied methods include Fast Fourier Transform (FFT), Power Spectral Density (PSD), Auto-correlation Function (ACF), Wavelet Packet (WPT), Ensemble Empirical Mode Decomposition (EEMD). also examine more recent approaches based on Topological Data Analysis (TDA) similarity measures time series Discrete Time Warping (DTW). evaluate potential approach training testing both within sets. Our results show that carefully chosen time-frequency lead to high accuracies albeit cost requiring manual pre-processing tagging expert user. On other hand, we found TDA DTW provide F1 scores par without preprocessing.