CNC Machine Tool Wear Monitoring Based on Densely Connected Convolutional Networks

As an important part of machining, the condition of cutting tools will have a huge impact on the processing cost, efficiency and product quality, and the intelligent monitoring for tool becomes increasingly important. This paper proposes a tool wear monitoring model based on 2σ-RMS Method and Densenet (Densely Connected Convolutional Networks). The time domain signal collected during tool processing is used to divide the tool condition utilizing 2σ-RMS method. The Densenet is trained in two stages to extract features information to build a mapping relationship between features and tool wear. The method of secondary sampling and multiple estimation for averaging reduces the interference of abnormal values on the estimated results and realized estimation of total wear of the tool after each stroke. Comparing with other monitoring methods, the monitoring model proposed in this paper can more effectively characterize the wear of milling cutters, and greatly improves the accuracy of estimation.