Automatic feature extraction for online chatter monitoring under variable milling conditions

Chatter is an unpredictable self-excited vibration phenomenon in the milling process, which can seriously affect machining efficiency and quality. In the aerospace industry, the cutting process lasts for an extended period, and the cutting parameters continuously change. This paper presents an automated method for monitoring chatter in this field. Recurrence plot (RP) can accurately reflect dynamic changes in the cutting system, but its hyperparameters must be set in advance. This paper initially proposes a novel adaptive particle swarm algorithm (APSO) for calculating hyperparameters so that RP can be obtained automatically. Then, as the global and local features of RP show clear changes in different cutting states, a deep neural network architecture that can extract features from multiple scales is developed. Three categories of experiments are conducted to test the proposed method. Experimental results show that the proposed method can achieve accurate online chatter monitoring under different cutting conditions.