Tool Wear Estimation with a Self-learning Adaptive Neuro- Fuzzy System in Copy Milling

Abstract. Tool wear sensing plays an important role in the optimisation of tool exchange and tip geometry compensation during automated machining in flexible manufacturing system. The focus of this work is to develop a reliable method to estimate flank wear during end milling process. A neural-fuzzy scheme is applied to perform one-step-ahead prediction of flank wear from cutting force signals obtained from a dynamometer. Because cutting force signals have more informations than acoustic emission signals, the relationship between the cutting force components and flank wear was examined. In our research we also discussed the construction of a neuro-fuzzy system that seeks to provide a linguistic model for the estimation of tool wear from the knowledge embedded in the neural network. Neuro-fuzzy modeling proved to be effective in modeling such complex systems. With the developed approach it is also possible to estimate the tool condition pretty accurately if the feeed and thrust cutting forces are measured at identical cutting conditions.