Research on tool wear prediction for milling high strength steel based on DenseNet-ResNet-GRU

Tool condition monitoring is an important basis to ensure workpiece quality and machining efficiency. It is also a key factor in improving machining efficiency, ensuring machining accuracy. Therefore, a new method for predicting tool wear based on DenseNet-ResNet-GRU is proposed. Firstly, statistical theory and an improved wavelet threshold denoising method are used to improve the signal quality. In addition, the asymptotic semi-soft threshold function is applied to reduce the noise of the cutting force signal. Secondly, DenseNet, ResNet, and GRU (gate recurrent unit) deep learning networks are integrated to create a new tool wear prediction model to realize the nonlinear mapping relationship between the tool wear amount and the cutting force characteristic. Finally, the tool wear prediction model is verified by high-strength steel experiment. The experimental results verify the accuracy and reliability of the method, which has a better training effect and higher prediction accuracy compared with the CNN-GRU model.