Method for edge chipping monitoring based on vibration polar coordinate image feature analysis

Cutting-edge chipping is a randomly happened but critical tool failure, which could seriously affect machining quality without effective and timely treatment especially in the automated machining process. Accurate and timely monitoring chipping is the first key step for properly handling edge chipping issues, while there are still gaps between academic research and industrial application as chipping initiation and propagation are always stochastic especially for milling complex structural parts with mass of ever-changing cutting conditions. This paper proposes a chipping monitoring method for automated milling process based on cutting vibration polar coordinate image features and residual neural network (ResNet) model, which lies on the change of cutting vibration signals that reveal just after an event of edge chipping has occurred. With the creatively proposed polar coordinate transform method, cutting vibration signals are first preprocessed for real-time chipping detection. The preprocessed signals with easily distinguished shape features of polar coordinate images are then employed for developing a ResNet50-based classification model. And consequently, the cutting-edge chipping could be monitored in almost real time. Experimental validation for real-time chipping detection has been conducted with the proposed method. The accuracy of No. 5~8 experiments are 92%, 88%, 83% and 84%, respectively. More importantly, the method is suited for edge chipping monitoring for milling complex structural parts, where the effects of constantly changing cutting conditions have been eliminated.