A novel approach for accurate in-situ chatter detection by iterative Vold-Kalman and LMS adaptive filtering of milling signals

The generation of milling chatter can result in a deterioration in the surface finish quality of the workpiece and a reduction in productivity. Therefore, timely and accurate detection of chatter is essential. Most existing chatter detection methods filter out the noise component of the acquired signal during the preprocessing stage. However, achieving complete separation of useful components from noise is challenging, which adversely affects the accuracy of chatter detection, especially in the early chatter stage. To address this issue, this paper employs the iterative VoldKalman and least mean square algorithms to effectively filter out uninterested components in the signal, thereby facilitating the accurate extraction and in-situ identification of chatter components. The periodic component is tracked via the iterative Vold-Kalman method, wherein the bandwidth is optimized through the Hooke-Jeeves direct search method. The results indicate that the tracking filtering error before and after optimization is reduced from 7.13 % to 0.21 %, enhancing the extraction accuracy. Following this, a least mean square adaptive filter is applied to the remaining signal to further diminish the noise. The findings reveal that the chatter component in the signal before and after filtering remains essentially unchanged, while the noise component is reduced by approximately 71.43 %, improving the signal-to-noise ratio. Finally, the cumulative energy ratio is utilized as the chatter indicator for detection. Through milling tests, it is demonstrated that the proposed method can accurately and promptly detect chatter at an early stage, making it adaptable to the complex working conditions encountered in actual machining.