A novel tool wear monitoring approach based on attention mechanism and gated recurrent unit
Lei Zhang, Zhengcai Zhao, Shichen Zheng, Ning Qian, Yao Li, Jiuhua Xu, Haixiang Huan
Abstract
Obtaining tool wear value quickly and accurately during machining is crucial for maintaining the surface quality of machined parts, enhancing machining efficiency and reducing operational costs. To improve the accuracy of the tool wear monitoring, this article proposes an Attention - GRU tool wear monitoring approach, which considers the importance of features extracted from multi-sensor signals and the previous machining processes in achieving accurate monitoring. Firstly, signals from the machining process are acquired, segmented and sliced. Second, an Attention - GRU tool wear monitoring model consisting of a 6-layer gated recurrent unit neural network optimized by the attention mechanism is constructed. Third, feature extraction is performed for the preprocessed signals, and the features which have a strong correlation with tool wear condition are selected as the inputs of the constructed approach using the extreme gradient boosting algorithm. Finally, the tool wear monitoring accuracy of the proposed approach is experimentally verified, and the monitoring performance is compared with back propagation neural networks and support vector regression. The results demonstrate that the proposed approach can achieve a tool wear monitoring accuracy of 97%.