An adaptive, artificial intelligence-based chatter detection method for milling operations

Chatter is an uncontrollable and unattenuated vibration that results in large oscillations between the workpiece and the cutting tool and has a detrimental effect on the surface quality, the tool life and the health of the machine tool components. As a result, it is one of the key limitations that hinders the productivity and quality of the milling process and is a key barrier for the autonomous operation of milling machine tools. Therefore, systems that can detect chatter, based on process-generated signals, are of utmost importance for the formation of a closed-loop control system that can suppress chatter during the process. Most existing approaches lack adaptability to different machining scenarios, since they use manually defined thresholds for the decision-making between chatter and stable machining, while being validated in a limited set of machining operations, running the risk of overfitting. This work proposes a method for chatter detection based on vibration signals in milling. An optimized version of variational mode decomposition (VMD) is used, where its hyperparameters can be selected automatically online, making it fully adaptable to different machining scenarios. Through VMD, the vibration signals are decomposed, and the modes with chatter rich information are selected for further analysis. Features are extracted from these modes in the time and frequency domains and are used to train a support vector machine classifier to predict the stability status of the process. The proposed approach presents a high classification performance (93% accuracy) and rapid detection speed (26.1 ms), which makes it a promising solution for real-time implementation.

Automated summary

Chatter is a major limitation in milling processes, affecting surface quality and machine tool health. Existing detection methods lack adaptability and may be overfit. This study proposes a vibration-based detection method with high classification performance and fast detection speed.