Adaptive Network Based Inference System for Estimation of Surface Roughness in End-milling

This paper presents a new approach for surface roughness (Ra) prediction during milling by using dynamometer to measure cutting forces signals and cutting conditions. End milling machining process of hardened die steel with carbide end mill, was modeled in this paper using the adaptive neuro fuzzy inference system (ANFIS) to predict the effect of machining variables (spindle speed, feed rate and axial/radial depth of cut) on surface roughness. In this contribution we also discussed the construction of a ANFIS system that seeks to provide a linguistic model for the estimation of surface roughness from the knowledge embedded in the neural network. The predicted surface roughness values determined by ANFIS were compared with experimental measurements. The comparison indicates that the performance of this method turned out to be satisfactory for evaluating Ra, within a 6% mean percentage error and 96% accuracy rate.