Abstract
During cylindrical grinding processes, two types of regenerative chatter— workpiece and grinding wheel—may degrade the accuracy of the surface finish. To maintain productivity and quality, a closed-loop vibration control system should be provided for the grinding system. An algorithm for automated classification of chatter by type is essential in developing such a system. In this paper, an entropy technique is introduced to classify regenerative chatter by type based on the vibration spectrum, the performance of this scheme is compared with that of a neural network approach. Experimental results show that the entropy system yields the same performance as the neural network. Because the entropy technique has several practical advantages, it may be judged preferable for monitoring of cylindrical grinding chatter. This paper also describes the theoretical background of the entropy technique in classifying regenerative chatter.