Neural Network Feed Rate Modulation for Ball-end Milling of Inclined Surfaces

Abstract

This paper proposes a feed rate modulation strategy to achieve desired surface roughness for ball-end milling an inclined surface. The system includes two models. First, the surface roughness prediction model is built by adopting a backpropagation neural network (NN). The model inputs include 3 cutting parameters, the inclined surface angle, and the characteristics of the acceleration for the spindle, while the output is the surface roughness. 0° and 30° machining inclination angles are considered. The backpropagation NN is also utilized for the feed rate modulation model. The model inputs include 3 cutting parameters, the error between the targeted and real surface roughness (i.e., ΔRa), and the inclined surface angle, and the output is the modulating value of the feed rate. Finally, some case study is conducted to show that the proposed NN feed rate modulation strategy can provide a more suitable feed rate to realize demanded surface roughness in inclined surface machining.