LAUSR

Thin-walled workpieces frequently appear to chatter during milling, causing the creation of vibration patterns on the surface of the workpiece, which affects the machining quality and workpiece performance. Identification of online chatter is important for thin-walled parts. The vibration signal associated with chatter changes dynamically throughout the milling process due to changes in tool position, removal of workpiece material, and transmission-related vibration signal attenuation characteristics, making reliable chatter identification difficult. This research proposes a unique online chatter identification approach for the machining of thin-walled parts based on improved multisensor signal fusion and multiscale entropy. In order to acquire the fused vibration signals, the vibration signals produced during the machining of thin-walled parts are first acquired by sensors at various test sites and adaptively given weight coefficients by the improved Hausdorff distance and distance factor. Second, using the multiscale sample entropy (MSSE) and multiscale weighted permutation entropy (MSWPE), the eigenvalues of the fused vibration signals are obtained. Finally, the trained logistic regression (LR) classification model is used to determine the vibration status of thin-walled parts. The analysis’s results demonstrated that the technique can properly identify the vibration state of thin-walled parts and that the fusion signal can reflect the vibration state at the tool contact site more accurately.