Chatter diagnosis using Mel frequency cepstral coefficient of vibrational signal for various operating conditions

Abstract

The chatter reduces quality of the machined parts and induces vibration and sound that is different from those under normal cutting process. A variety of cutting conditions are significantly correlated with the occurrence of chatter. After changing of the cutting conditions such as rotational speeds and depths of machine tools blade, the vibration signals was measured using vibrational accelerometers. The chatter diagnosis algorithm using Mel Frequency Cepstral Coefficient (MFCC) and Deep belief network (DBN), one of the Deep learning algorithm is proposed. Deep belief network is effective on classifying complicated signals as it represents the hierarchical cognitive process of a human brain. To acquire features from the distinctive noise and vibration under normal and chatter status, the MFCCs was used. Chatter diagnosis algorithm, DBN using MFCCs as input features, is suggested and verified.