Active suppression of chatter in machining

摘要

This paper presents a novel approach for active, on-line suppression of chatter in machining operations. While traditionally passive methods for chatter suppression have been studied extensively, for higher speeds, lighter weights, smaller tolerances, and greater process flexibility and efficiency, these methods have proven less than adequate. The active chatter suppression system presented in this paper is comprised of a vibration source with continually adjustable parameters. The goal is to select, on-line, the optimum values for amplitude, phase angle, and frequency of the vibration source to achieve maximum chatter suppression. The on-line selection of the optimum parameters for the active vibration system is based on a differentiable model of the cutting process. This differentiable model is constructed using a neural network. The neural network provides a differentiable model of the cutting process and the dynamic effects of the tool-workpiece interface. The neural network used for this purpose is a multi-layered feedforward network with tapped delay lines in frequency domain. The input to the model are estimates of the frequency response at the start of a sampling period. The outputs are the predicted frequency response at the end of a sampling period. Other inputs to the neural network model include cutting conditions, and parameters of the active vibration source. Once the presence of chatter is detected the suppression system will be activated. The neural network model can be used to calculate current gradient values with respect to the parameters of the active vibration source. These gradient information will be used by an optimization module to find the optimal set of parameters for the active vibration source. The overall result is an on-line, adaptable chatter suppression system. The methodology described was evaluated through simulation studies for both boring and milling operations. The results show that the system is capable of suppressing chatter in the presence of noise and uncertainty in the signals.