Early chatter detection in end milling based on multi-feature fusion and 3σ criterion

Chatter detection and suppression/control is an effective way to ensure the quality of workpiece in machining process. In this paper, a new method is proposed to detect early chatter before the workpiece surface is deteriorated. As a preparation work, various samples in stable cutting conditions are collected to train the self-organizing map (SOM) neural network. During online chatter monitoring, firstly, vibration signals are measured in real time and multi-features are extracted with signal processing methods to form a feature vector. Then, the feature vector is input to the trained SOM neural network. The minimum quantization error (MQE), namely, the Euclidean distance between the best matching unit (BMU) of the SOM neural network and the feature vector, is calculated as a new chatter detection indicator. Lastly, the MQE is compared with a 3σ criterion-based threshold, which is independent of cutting conditions, to determine the chatter occurrence. The proposed method was verified with end milling tests, and the results showed that the chatter can be detected before serious marks were left on the workpiece surface.