Chatter detection based on adaptive estimation of the fourth cumulant of the modeling error

This paper proposes a design method for chatter detection based on the fourth order statistics estimation applied to the modeling error. Usually, a test of hypothesis is built on an a priori knowledge of the statistical distribution. The frequency function is often assumed Gaussian. In the case of machining process without chatter, the Gaussian assumption is not valid. Consequently, the proposal of an adaptive estimation of the fourth cumulant on a sliding window is introduced for machining process monitoring with possible chatters. The modeling error of the cutting force is the difference between the output of nonlinear model of cutting force and a real force estimation, taking into account the stiffness of the sensor. A milling process illustrates the effectiveness of the proposed chatter detection strategy.