In-process tool wear monitoring through time series modelling and pattern recognition

Abstract

This paper describes an indirect in-process technique for monitoring the cutting tool condition in a turning process. Here an attempt has been made to extract maximum information from force/vibration signal acquired during machining. Statistical methods like time series modelling technique are used to extract parameters called features which represent the state of the cutting process. Autoregressive (AR) model parameters and AR residual signals are investigated and found to show effective and consistent trend towards tool wear. Other parameters such as static cutting force and power of the dynamic signal (force/vibration) are also studied here as features. Once the features are extracted through preliminary processing of the signal, tool state is decided through a pattern recognition technique.