Abstract: Effective monitoring of tool wear is crucial for ensuring machining quality and improving production efficiency. To address the issue of tool wear monitoring accuracy and the limitations of single deep learning models in extracting wear features and capturing temporal dependencies, a tool wear monitoring method based on Transformer-BiGRU is proposed. Firstly, the collected sensor signals are preprocessed. Secondly, the time-domain, frequency-domain and time-frequency-domain features of the signals are extracted. Finally, the multi-head self-attention mechanism of Transformer is utilized to efficiently and in parallel learn global dependencies between signal features, combined with the BiGRU network to bidirectionally capture the temporal dynamic characteristics of wear evolution. This approach fully explores the deep correlation between signal features and wear states, thereby improving the prediction accuracy of the model. Experimental results show that the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination R² of the proposed model are 4.31, 5.33 and 0.93, respectively, which can effectively monitor tool wear.