Optimization of CNC ball end milling: a neural network-based model

Abstract

In this paper, an integrated product development system for optimized CNC ball end milling is presented. First, the developed model is extended from flat end milling to ball end milling. Second, the optimization is extended from 2D (speed and feed) to 3 (1/2) D (speed, feed, radial and axial depths of cut). Third, the modeling and simulation of the flat end milling is extended to include more input variables. Finally, a new, more efficient and practical, neural network technique is introduced to replace the back-propagation neural network (BPNN), and is successfully implemented for the case of ball end milling. The work is verified and validated using typical machining scenarios. A very good agreement between predicted and experimentally measured process parameters is found.