부산대학교 도서관

Deep learning methods have developed rapidly in the field of tool condition monitoring, but due to the complexity and diversity of working conditions, it is difficult to ensure the high accuracy, strong generalization performance, and wide applicability of monitoring models. Therefore, this article proposes a deep transfer learning method with center loss for tool condition monitoring under different processing conditions. Firstly, Deep Extreme Learning Machine (DELM) is used to extract sample features and tool condition monitoring, and deep coral is integrated into the last feature extraction layer of DELM for domain adaptation. Secondly, the center loss is introduced into the transfer learning model to improve the intra-class compactness by minimizing the center loss, thereby obtaining a broader decision boundary. A tool wear experiment was conducted on a milling machine. Research shows that the proposed method can achieve the tool condition monitoring under different processing conditions. The introduction of central loss is beneficial for promoting the separation of samples from different categories in the target domain and effectively improves the model's applicability. Compared with other domain adaptation methods, this method has better accuracy and generalization ability. [ABSTRACT FROM AUTHOR]