A Tool Remaining Useful Life Prediction Model Based on Improved TCN Combined with BiGRU

To address the issue of low accuracy in traditional tool remaining useful life prediction, this paper proposes a prediction model based on an improved Temporal Convolutional Network (TCN) combined with a Bidirectional Gated Recurrent Unit (BiGRU). The proposed method enhances the conventional TCN residual module by introducing a parallel pooling structure and incorporating a Convolutional Block Attention Module (CBAM) to improve the extraction capability of tool wear features. Additionally, BiGRU leverages its bidirectional characteristics and gating mechanisms to capture complex temporal dependencies within the sequence, thereby enhancing the prediction capability and accuracy of the model. The proposed method is evaluated on the PHM2010 tool wear dataset, where the coefficient of determination for tools C1, C3, and C4 on the validation set reaches 0.9978,0.9843, and 0.9886, respectively. Compared to TCN, TCN-LSTM, and other baseline methods, the proposed approach achieves significantly higher prediction accuracy. Experimental results demonstrate that the proposed method outperforms existing approaches in terms of prediction precision.