• HHT enabled precise classification of AE signals from different CMP mechanisms in fixed abrasive lapping. • Fixed agglomerated diamond abrasive pad produced stable, continuous AE signals (0.2–0.7 MHz), linked to effective material removal. • Correlated 0.05–0.7 MHz bands with matrix friction, abrasive sliding, particle motion, and removal mechanisms. Chemical mechanical planarization (CMP) of silicon carbide (SiC) wafers is a critical process determining surface quality, and achieving efficient, high-quality machining urgently requires reliable online monitoring technology. Acoustic emission (AE) technology, with its advantages of high sensitivity and fast response, holds unique application potential in the field of machining state monitoring. However, the complex abrasive-workpiece interaction and strong noise interference during CMP pose significant challenges to feature extraction and recognition of acoustic signals. This paper focuses on the fixed abrasive lapping process of 4H-SiC wafers, conducting research on the classification of machining behaviors based on acoustic emission signals. By comparing AE signals from pure resin pads, fixed single crystal diamond abrasive pads (FSCDAPs), and fixed agglomerated diamond abrasive pads (FADAPs) under different machining conditions, and combining Hilbert-Huang Transform (HHT) for time-frequency analysis, the results show: FADAP, benefiting from the self-sharpening capability of the abrasives, achieves a stable material removal rate (MRR). Its corresponding AE signals are continuous and stable in the time domain, with frequency domain energy concentrated in a continuous band of 0.2∼0.7 MHz. In contrast, the machining process of FSCDAP is nearly ineffective, and its acoustic signals exhibit burst-pulse characteristics with discrete and disordered spectral distribution. Through designed control experiments, specific frequency band signals were further associated with different physical mechanisms: the 0.05∼0.1 MHz band suggests a correlation with to the sliding/rubbing action of the resin matrix, the 0.1∼0.2 MHz band suggests a correlation with to the sliding/rubbing behavior of blunted diamond abrasives, the 0.1∼0.5 MHz band is related to particle motion, and the 0.2∼0.7 MHz band can be interpreted as effective material removal behaviors such as yielding, plastic deformation, and fracture. This study provides effective feature identification methods and a frequency-band assignment framework for real-time acoustic monitoring of 4H-SiC lapping.
Chen et al. (Fri,) studied this question.