Tool wear state recognition method based on improved ResNet and feature fusion

The tool is the executing part of the machine tool processing, and its wear state directly affects the production efficiency and product quality. This paper proposes a tool wear state recognition method based on improved ResNet and feature fusion. This method first extracts features from various fields from triaxial vibration signals, temperature signals and power signals, and uses PCA technology to fuse the features. In order to better process the feature fusion data, the convolutional layers and pooling layers in the ResNet architecture are replaced by the fully connected layers, and an improved ResNet model is constructed. Finally, the tool’s wear condition is identified with the improved ResNet model. Experiments have shown that the model has a high ability to identify and classify the three different states of tool initial wear, normal wear and severe wear, and can achieve an identification and classification accuracy of $\mathbf{9 7. 5 \%}$.