Online Monitoring Method for Milling Chatter Based on Multi-Channel Parellel Convolutional Neural Networks and Attention Mechanisms

Milling chatter is a common issue in machining processes that significantly impacts both the quality and efficiency of machining. The traditional approach to monitor milling chatter depends on manually set feature parameters and thresholds to identify chatter incidents, resulting in issues such as feature redundancy and the subjectivity of threshold determination. To address these issues, a novel online monitoring method for milling chatter is proposed, based on Multi-Channel Parallel Convolutional Neural Networks (MC-CNN) and self-attention mechanisms. This method leverages MC-CNN to automatically extract and integrate signal features from multiple sensors, thereby reducing human intervention. The introduction of the Channel-wise Attention Mechanism (ECA) further enhances the model’s ability to identify key features by dynamically adjusting feature weights. Experimental results conducted on a horizontal machining center demonstrate that the proposed MC-CNN+ECA model significantly improves classification accuracy and reduces training time compared to traditional models such as LSTM and Transformer. This model exhibits superior classification performance and computational efficiency.

Recommended citation: Jin Yihan, Wei Chuang, Tang Yuzhe, et al. Online Monitoring Method for Milling Chatter Based on Multi-Channel Parellel Convolutional Neural Networks and Attention Mechanisms[J]. Machine Design & Research.(Accepted)
Download Paper