<p>Effective tool wear monitoring in cutting machining is crucial for ensuring machining quality and enhancing productivity. Current tool wear monitoring typically relies on a data-driven approach, which involves integrating multi-sensor data with end-to-end deep learning models to predict tool wear values. However, different sensor channels exhibit varying sensitivities to wear. Existing methods often lack mechanisms to assess channel importance and emphasize the most relevant ones. This limitation reduces the overall predictive accuracy of the model. We present a novel tool wear monitoring model based on parallel residual channel attention and a bidirectional gated recurrent unit (PRes-GE-BiGRU). First, the collected multi-sensor signal data are input into a parallel residual network (PResNet) for the extraction of multi-scale local features. Subsequently, these features are processed by a channel attention mechanism (GENet) to assign varying weights to different channels. Finally, global time series features associated with tool wear are derived using a bidirectional gated recurrent unit (BiGRU). The fully connected layers map the input features to predict tool wear values. The model proposed in this study has been experimentally validated, showing high accuracy and significant advantages in predicting tool wear.</p>