Online Incipient Chatter Detection Based on Once-Per-Revolution Sampling and Dynamic Threshold Variant
Abstract
Online incipient chatter detection is a crucial task to improve surface quality, reduce dimensional errors, and protect the cutting tool and machine tool in the high-speed milling process. Since variable background noise in measured signal impacts the accuracy of the incipient chatter detection, it is necessary to develop an appropriate and efficient threshold to detect incipient chatter reliably. This paper proposes an online incipient chatter detection method for high-speed milling process based on once-per-revolution sampling and dynamic threshold variant. First, the once-per-revolution sampling technique is used to reduce the data size and acquire independent data of machining parameters. Next, an incipient chatter indicator MaxEnt, i.e., maximum entropy, is extracted from the once-per-revolution sampled data by using the MaxEnt principle. Finally, a dynamic threshold variant and sequential probability ratio test (SPRT) are used for incipient chatter detection. To validate the effectiveness of the proposed method, a simulated chatter signal, including seven components, is employed. The validated results show that the proposed method could detect the incipient chatter effectively in the variable noise environment.