文献类型：期刊文献

英文题名：Mill Vibration Signal-Based Milling Condition Recognition Using Transformer with Time-Frequency Decoupled Attention

作者：Li, Rongjun[1]; Zou, Huaiwen[1]; Zhou, Jiayi[1]; Li, Zhongmei[2]; Wang, Xiaoli[1]

机构：[1] School of Automation, Central South University, Changsha, 410083, China; [2] School of Information Science and Engineering, East China University of Science and Technology, Shanghai, 200237, China

年份：2025

卷号：59

期号：32

起止页码：36

外文期刊名：IFAC-PapersOnLine

收录：EI(收录号：20260520005989)

语种：英文

外文关键词：Condition monitoring - Signal analysis - Vibration analysis

摘要：In mill operation, vibrations are generated by interactions between internal materials and the mill shell, which can directly reflect the milling condition and have been used for recognition of milling condition. However, existing methods exhibit limitations in time-frequency feature modeling and complex condition recognition, which limit the accuracy and practicality of condition monitoring. In this paper, a novel method is proposed for milling condition recognition based on vibration signals using a Transformer architecture. In the method, a time-frequency decoupled attention mechanism is proposed to extract intrinsic time and frequency features from vibration signals, and a feature aggregation module is introduced for effective integration of different features. Experimental results on a laboratory-collected mill vibration dataset demonstrate that the proposed method outperforms existing CNN-based and Transformer-based approaches in recognition performance and show the effectiveness of the method. ? 2025 The Authors.