Accurately predicting tool wear is a challenging problem.How to combine the advantages of various signals and fuse the sensor signal features to improve the prediction accuracy is a key issue.To solve the above problems,a multi-domain feature fusion tool wear prediction method was proposed based on temporal convolutional network-long short term memory (TCN-LSTM).The signals from different sensors were collected and features were extracted from different sensor signals in the time domain and frequency domain,respectively.The raw signal was decomposed in the time-frequency domain using the variational mode decomposition algorithm,and the energy of each component was calculated to form a multi-domain feature vector.The multi-domain features were optimized using the Pearson correlation coefficient method.The optimized composition of the multi-domain feature matrix was used as an input to the model,and the complex relationship between the obtained multi-domain feature matrix and real-time tool wear was efficiently learned by the TCN-LSTM model.Finally,the proposed method was validated by three sets of experimental on tool wear under dry milling conditions.The experimental results show that the proposed method has higher prediction accuracy and better generalization capability than the comparison model.