In-process Evaluation of the Overall Machining Performance in Finish-turning via a Single Data Source

As a different approach to the condition monitoring of machining processes by sensor fusion, which has been of great interest over the recent time, this paper presents a novel approach of utilising a single data source to evaluate in-process the overall machining performance in finish turning. The overall machining performance includes the variations of machining performance (chip breakability, surface roughness, dimensional accuracy and cutting forces) with progressive tool wear of various types of tool wear (major flank, crater and minor flank wear). The 3-D cutting force measured by a tool dynamometer, perhaps the most reliable data source available for machining processes, is fully utilised through the combination of multivariate time series models and neural network techniques. Dispersion analysis based on the established multivariate time series model of 3-D cutting force is introduced to single out signal features corresponding to particular types of tool wear, and four dispersion patterns along with others are used to train the neural network to quantify the complex interrelationship -between machining performance (chip breakability, dimensional accuracy and surface roughness) and progressive tool wear for cutting tools with different chip control geometry. The approach has demonstrated to be a simple yet effective means in on-line monitoring of the overall machining performance for fmish-tuming operations.